遥感建筑损毁评估

Abstract After the occurrence of various types of disasters, including natural disasters and man-made damage, aid workers need accurate and timely data, such as the damage status of buildings, in order to take effective measures for rescue. So as to solve this problem, this paper researches and designs a building damage classification system based on machine learning. The damage assessment system consists of two network models (building extraction network and damage classification network). This article analyzes and designs the structure of each network model, and discusses the principles related to computer vision in machine learning. Buildings in satellite images are segmented through Siamese Convolutional Neural Network, the BottleNeck Module and Feature Pyramid Network are used in the damage classification assessment network to detect damage to buildings in sub-temporal remote sensing images. Subsequently, the model was trained and tested on different disaster events on the xBD dataset. The results show that the building damage detection system based on Siamese-CNN achieves good detection accuracy, and the system has the advantages of simple operation, good timeliness and low resource consumption, and can well meet the needs of disaster assessment.

ABSTRACT Most existing remote sensing disaster assessment methods rely on convolutional neural networks (CNNs). Although CNNs can extract effective semantic features, determining global spatial relationships remains limited due to the locality of convolutional operations. The recently developed transformer-based method can extract global information from images effectively by encoding image tokens. However, its consumption of computational resources varies markedly with increasing image resolution. In this paper, we propose a novel transformer-based neural network for disaster assessment problems. The network uses a multihead neighbourhood attention (MNA) transformer as the base layer of the encoder to achieve more efficient self-attention computation. In addition, the bitemporal feature fusion module (BFFM) performs differential enhancement and injects the change information to the decoder via skip connections. The multiscale tokenizer generates multiscale image tokens to mitigate the loss of detail during encoding. Experimental results on three datasets show that the proposed method outperforms existing methods.

Accurate classification of building damage caused by disasters such as earthquakes, hurricanes, and floods is crucial for effective disaster response and recovery. Remote sensing technology provides timely and extensive spatial data, making it a valuable tool for large-scale damage assessment. However, conventional training sample selection methods, which often rely on random or evenly distributed sampling, can be inefficient and fail to capture the full spectrum of damage severities. These traditional approaches may lead to suboptimal classification performance and require extensive manual effort, limiting their practical applicability, especially in time-sensitive disaster scenarios.To address these challenges, we propose an adaptive training sample selection method that integrates disaster intensity as a key criterion. By prioritizing heavily affected areas, our approach ensures a more representative and diverse set of training samples, improving the accuracy and robustness of remote sensing-based building damage classification. This method not only reduces manual intervention but also enhances classification efficiency, making the process more scalable for rapid deployment in disaster-stricken regions. Additionally, our approach leverages automated data-driven techniques to dynamically refine sample selection, further optimizing performance. By improving training sample representativeness, this method enhances the effectiveness of remote sensing technology in disaster response, enabling quicker, more accurate damage assessments and facilitating faster, more targeted assistance to affected communities

No abstract available

Remote sensing and computer vision technologies are increasingly leveraged for rapid post-disaster building damage assessment, becoming a crucial and practical approach. In this context, the accuracy of employing various AI models in pixel-level change detection methods is significantly dependent on the consistency between pre- and post-disaster building images, particularly regarding variations in resolution, viewing angle, and lighting conditions; in object-level feature recognition methods, the low richness of semantic details of damaged buildings in images leads to a poor detection accuracy. This paper proposes a novel method, OCD-BDA (Object-Level Change Detection for Post-Disaster Building Damage Assessment), as an alternative to pixel-level change detection and object-level feature recognition methods. Inspired by human cognitive processes, this method incorporates three key steps: an efficient sample acquisition for object localization, labeling via HGC (Hierarchical and Gaussian Clustering), and model training and prediction for classification. Furthermore, this study establishes a change detection dataset based on Google Earth imagery of regions in Hatay Province before and after the Turkish earthquake. This dataset is characterized by pixel inconsistency and significant differences in photographic angles and lighting conditions between pre- and post-disaster images, making it a valuable test dataset for other studies. As a result, in the experiments of comparative generalization capabilities, OCD-BDA demonstrated a significant improvement, achieving an accuracy of 71%, which is twice that of the second-ranking method. Moreover, OCD-BDA exhibits superior performance in tasks with small sample amounts and rapid training time. With only 1% of the training samples, it achieves a prediction accuracy exceeding that of traditional transfer learning methods with 60% of samples. Additionally, it completes assessments across a large disaster area (450 km²) with 93% accuracy in under 23 min.

Sudden-onset natural disasters, such as destructive earthquakes, pose significant threats to human life and property. The use of high-resolution remote sensing (HRRS) images for automated assessment of building damage can rapidly and accurately provide spatial distribution information and statistical data on building damage, assisting in disaster response and relief efforts. However, the task is exceedingly challenging due to the diverse and intricate appearance of damaged buildings in HRRS images, coupled with interference from surrounding areas that exhibit certain damage characteristics as a result of the disaster. To overcome these issues, this article proposes a weakly supervised building damage assessment method based on scene change detection in pre- and post-disaster bitemporal images. This method fully leverages visual information of building boundaries and deeper semantic information of building scenes from pre-disaster images to guide the identification of building damage in post-disaster images. Specifically, the method first generates fine-grained subbuilding objects with detailed boundaries from pre-disaster images by combining semantic segmentation of buildings with superpixel segmentation. Then, bitemporal image scene blocks obtained using sub-building objects as clues are input into our proposed Siamese local-global visual transformer (SLgViT) network, enabling scene change detection guided by deep semantic information from pre-disaster images. Finally, the change detection results serve as the basis to depict pixel-level building damage in post-disaster images. The proposed SLgViT network is primarily composed of a specially designed local-global visual transformer (LgViT) module and a cross-Siamese interaction fusion (CSIF) module, both of which play a crucial role in the deep mining and integrated interaction of local and global semantic features from pre- and post-disaster images. It is noteworthy that our method operates in a weakly supervised manner. The training of the SLgViT network requires only scene patches centered around building objects from pre- and post-disaster bitemporal images, along with image-level annotations. Experiments conducted with satellite images from the 2010 Port-au-Prince, Haiti earthquake and unmanned aerial vehicle (UAV) images from the 2019 Changning, China earthquake have demonstrated the effectiveness and superior performance of the proposed method.

Most post-disaster damage classifiers succeed only when destructive forces leave clear spectral or structural signatures -- conditions rarely present after inundation. Consequently, existing models perform poorly at identifying flood-related building damages. The model presented in this study, Flood-DamageSense, addresses this gap as the first deep-learning framework purpose-built for building-level flood-damage assessment. The architecture fuses pre- and post-event SAR/InSAR scenes with very-high-resolution optical basemaps and an inherent flood-risk layer that encodes long-term exposure probabilities, guiding the network toward plausibly affected structures even when compositional change is minimal. A multimodal Mamba backbone with a semi-Siamese encoder and task-specific decoders jointly predicts (1) graded building-damage states, (2) floodwater extent, and (3) building footprints. Training and evaluation on Hurricane Harvey (2017) imagery from Harris County, Texas -- supported by insurance-derived property-damage extents -- show a mean F1 improvement of up to 19 percentage points over state-of-the-art baselines, with the largest gains in the frequently misclassified"minor"and"moderate"damage categories. Ablation studies identify the inherent-risk feature as the single most significant contributor to this performance boost. An end-to-end post-processing pipeline converts pixel-level outputs to actionable, building-scale damage maps within minutes of image acquisition. By combining risk-aware modeling with SAR's all-weather capability, Flood-DamageSense delivers faster, finer-grained, and more reliable flood-damage intelligence to support post-disaster decision-making and resource allocation.

Most post‐disaster damage classifiers perform best when destructive forces leave clear spectral or structural signatures. However, these signatures are often subtle or absent after inundation, where damage may be nonstructural and difficult to detect. Consequently, existing models perform poorly at identifying flood‐related building damage. The model presented in this study, Flood‐DamageSense, addresses this gap as the first deep learning framework purpose‐built for building‐level flood‐damage assessment. The architecture fuses pre‐ and post‐event synthetic aperture radar/interferometric synthetic aperture radar (SAR/InSAR) scenes with very high‐resolution optical basemaps and an inherent flood‐risk layer that encodes long‐term exposure probabilities, guiding the network toward plausibly affected structures even when compositional change is minimal. A multimodal Mamba backbone with a semi‐Siamese encoder and task‐specific decoders jointly predicts (1) graded building‐damage states, (2) floodwater extent, and (3) building footprints. Training and evaluation on Hurricane Harvey (2017) imagery from Harris County, Texas—supported by insurance‐derived property‐damage extents—show a mean F1 improvement of up to 19 percentage points over state‐of‐the‐art baselines, with the largest gains in the frequently misclassified “minor” and “moderate” damage categories. Ablation studies identify the inherent‐risk feature as the single most significant contributor to this performance boost. An end‐to‐end post‐processing pipeline converts pixel‐level outputs to actionable, building‐scale damage maps within minutes of image acquisition. By combining risk‐aware modeling with SAR's all‐weather capability, Flood‐DamageSense delivers faster, finer‐grained, and more reliable flood‐damage intelligence to support post‐disaster decision‐making and resource allocation.

Rapid building damage assessment following an earthquake is important for humanitarian relief and disaster emergency responses. In February 2023, two magnitude-7.8 earthquakes struck Turkey in quick succession, impacting over 30 major cities across nearly 300 km. A quick and comprehensive understanding of the distribution of building damage is essential for efficiently deploying rescue forces during critical rescue periods. This article presents the training of a two-stage convolutional neural network called BDANet that integrated image features captured before and after the disaster to evaluate the extent of building damage in Islahiye. Based on high-resolution remote sensing data from WorldView2, BDANet used pre-disaster imagery to extract building outlines; the image features before and after the disaster were then combined to conduct building damage assessment. We optimized these results to improve the accuracy of building edges and analyzed the damage to each building, and used population distribution information to estimate the population count and urgency of rescue at different disaster levels. The results indicate that the building area in the Islahiye region was 156.92 ha, with an affected area of 26.60 ha. Severely damaged buildings accounted for 15.67% of the total building area in the affected areas. WorldPop population distribution data indicated approximately 253, 297, and 1,246 people in the collapsed, severely damaged, and lightly damaged areas, respectively. Accuracy verification showed that the BDANet model exhibited good performance in handling high-resolution images and can be used to directly assess building damage and provide rapid information for rescue operations in future disasters using model weights.

No abstract available

A catastrophic Mw7.8 earthquake hit southeast Turkey and northwest Syria on February 6th, 2023, leading to more than 44 k deaths and 160 k building collapses. The interpretation of earthquake-triggered building damage is usually subjective, labor intensive, and limited by accessibility to the sites and the availability of instant, high-resolution images. Here we propose a multi-class damage detection (MCDD) model enlightened by artificial intelligence to synergize four variables, i.e., amplitude dispersion index (ADI) and damage proxy (DP) map derived from Synthetic Aperture Radar (SAR) images, the change of the normalized difference built-up index (NDBI) derived from optical remote sensing images, as well as peak ground acceleration (PGA). This approach allows us to characterize damage on a large, tectonic scale and a small, individual-building scale. The integration of multiple variables in classifying damage levels into no damage, slight damage, and serious damage (including partial or complete collapses) excels the traditional practice of solely use of DP by 11.25% in performance. Our proposed approach can quantitatively and automatically sort out different building damage levels from publicly available satellite observations, which helps prioritize the rescue mission in response to emergent disasters.

Accurate and efficient post-earthquake building damage assessment methods enable key building damage information to be obtained more quickly after an earthquake, providing strong support for rescue and reconstruction efforts. Although many methods have been proposed, most have limited effect on accurately extracting severely damaged and collapsed buildings, and they cannot meet the needs of emergency response and rescue operations. Therefore, in this paper, we develop a novel building damage heterogeneity enhancement network for pixel-level building damage classification of post-earthquake unmanned aerial vehicle (UAV) and remote sensing data. The proposed BDHE-Net includes the following three modules: a data augmentation module (DAM), a building damage attention module (BDAM), and a multilevel feature adaptive fusion module (MFAF), which are used to alleviate the weight deviation of intact and slightly damaged categories during model training, pay attention to the heterogeneous characteristics of damaged buildings, and enhance the extraction of house integrity contour information at different resolutions of the image. In addition, a combined loss function is used to focus more attention on the small number of severely damaged and collapsed classes. The proposed model was tested on remote sensing and UAV images acquired from the Afghanistan and Baoxing earthquakes, and the combined loss function and the role of the three modules were studied. The results show that compared with the state-of-the-art methods, the proposed BDHE-Net achieves the best results, with an F1 score improvement of 6.19–8.22%. By integrating the DBA, BDAM, and MFAF modules and combining the loss functions, the model’s classification accuracy for severely damaged and collapsed categories can be improved.

ABSTRACT Efficient building damage assessment after disasters is vital for emergency response and loss evaluation, but the task is complicated by diverse building structures and complex environments. Traditional methods using Convolutional Neural Networks (CNNs) struggle to capture global contextual features, limiting damage categorization accuracy. To address this, we introduce the High-Resolution Transformer Architecture for Building Damage Assessment (HRTBDA), which enhances multi-scale feature extraction. A Cross-Attention-Based Spatial Fusion (CSF) module is proposed to utilize the attention mechanism, improving the model’s ability to identify detailed associations in damaged buildings. Additionally, we propose a deep convolution network matching optimization strategy that integrates a multilayer perceptron and expands the receptive field, enhancing global feature perception. HRTBDA’s performance was evaluated on two public datasets and compared with five recent frameworks. The model achieved an F1-score of 86.0% in building localization and 78.4% in damage assessment, with a 4.8% improvement in detecting minor damages. These results demonstrate HRTBDA’s potential for improving building damage assessment and highlight its significant advancements over existing methods.

In the aftermath of a natural hazard, rapid and accurate building damage assessment from remote sensing imagery is crucial for disaster response and rescue operations. Although recent deep learning-based studies have made considerable improvements in assessing building damage, most state-of-the-art works focus on pixel-based, multi-stage approaches, which are more complicated and suffer from partial damage recognition issues at the building-instance level. In the meantime, it is usually time-consuming to acquire sufficient labeled samples for deep learning applications, making a conventional supervised learning pipeline with vast annotation data unsuitable in time-critical disaster cases. In this study, we present an end-to-end building damage assessment framework integrating multitask semantic segmentation with semi-supervised learning to tackle these issues. Specifically, a multitask-based Siamese network followed by object-based post-processing is first constructed to solve the semantic inconsistency problem by refining damage classification results with building extraction results. Moreover, to alleviate labeled data scarcity, a consistency regularization-based semi-supervised semantic segmentation scheme with iteratively perturbed dual mean teachers is specially designed, which can significantly reinforce the network perturbations to improve model performance while maintaining high training efficiency. Furthermore, a confidence weighting strategy is embedded into the semi-supervised pipeline to focus on convincing samples and reduce the influence of noisy pseudo-labels. The comprehensive experiments on three benchmark datasets suggest that the proposed method is competitive and effective in building damage assessment under the circumstance of insufficient labels, which offers a potential artificial intelligence-based solution to respond to the urgent need for timeliness and accuracy in disaster events.

No abstract available

Comparison of other satellite data, there are fewer scientific papers about building damage assessment using Sentinel-1 data. Because many scientists ignore it due to middle-spatial resolution, the general trend is using high-resolution data (TerraSAR-X, COSMO-SkyMed, etc.) for that purpose. It is related to the problem that middle-resolution data has lower overall accuracy than high resolution. Sentinel-1 data is more freely available than others. Pre-event data is always available. The application of texture-based change detection techniques can be used to improve overall accuracy. Better separation of completely destroyed and intact buildings was achieved using homogeneity and dissimilarity textural parameters computed from the grey-level co-occurrence matrices (GLCM). The backscattering coefficients with dual polarization (VV, VH) and the coherence coefficient (pre-earthquake and coseismic data) were exploited for this study. The best relevant GLCM textural parameter variables were determined to extract open areas (without buildings), and damaged and untouched buildings in urban areas using supervised classification methods. In this research work, the overall accuracy was achieved at 0.77. The producer's accuracy for open areas is 0.84, for the case of a damaged building 0.85, and for untouched building 0.64. Beijing-2 high-resolution optical data and Copernicus Emergency Management Service data were exploited for that classification. Amatrice town as a study area chose for investigation as an example that was significantly affected by the earthquake in Central Italy in 2016.

Disasters impose significant losses on human society, among which building damage constitutes a crucial factor. Leveraging the strengths of deep learning to promptly detect buildings after disasters with varying damage levels enables efficient disaster assessment. Current methodologies predominantly rely on pixel-level change detection approaches for post-disaster building damage assessment. However, high-resolution pixel-based analytical paradigms exhibit inherent limitations, particularly their susceptibility to pseudo-changes caused by variations in imaging angles, seasonal changes, and environmental interference. Furthermore, pixel-level building damage assessment faces the challenge of intra-instance damage heterogeneity, where pixels within a single building structure may belong to different damage categories. To address the technical constraints of pixel-level building damage identification paradigms in disaster loss assessment, this study proposes an object-level building damage detection framework. This work shifts the analytical perspective from pixel-wise comparison to instance detection, effectively reduces false alarms induced by environmental variations. The proposed network architecture operates through two stages: a building localization stage and a damage detection stage. The building locating stage performs structural identification from pre-disaster images, establishing location references for subsequent analysis. The damage detection stage then conducts comprehensive analysis by processing both pre-disaster and post-disaster image pairs, enabling classification of structural damage levels through temporal comparison. The proposed architecture integrates a feature fusion network enhanced with a frequency-spatial attention mechanism. This mechanism synergistically combines discrete cosine transform-based frequency analysis with spatial relationship modeling, significantly improving the extraction of critical building arrangement features while suppressing irrelevant background noise. To overcome the persistent challenge of small building target recognition, we implement an improved composite loss function that strategically balances localization accuracy and shape similarity. Experimental validation conducted on our dataset demonstrates that the proposed method achieves 65% mean Average Precision (mAP). The work provides an efficient technical solution for post-disaster building damage assessment.

Natural disasters demand swift and accurate impact assessment, yet traditional field-based methods remain prohibitively slow. While semi-automatic techniques leveraging remote sensing and drone imagery have accelerated evaluations, existing datasets predominantly emphasize Western infrastructure, offering limited representation of African contexts. The EDDA dataset (a Mozambique post-disaster building damage dataset developed under the Efficient Humanitarian Aid Through Intelligent Image Analysis project), addresses this critical gap by capturing rural and urban damage patterns in Mozambique following Cyclone Idai. Despite encouraging early results, significant challenges persist due to task complexity, severe class imbalance, and substantial architectural diversity across regions. Building upon EDDA, this study introduces a two-stage building damage assessment pipeline that decouples localization from classification. We employ lightweight You Only Look Once (YOLO)-based detectors—RTMDet, YOLOv7, and YOLOv8—for building localization, followed by dedicated damage severity classification using state-of-the-art architectures including Compact Convolutional Transformers, EfficientNet, and ResNet. This approach tests whether separating feature extraction tasks—assigning detectors solely to localization and specialized classifiers to damage assessment—yields superior performance compared to multi-class detection models that jointly learn both objectives. Comprehensive evaluation across 640+ model combinations demonstrates that our two-stage pipeline achieves competitive performance (mAP 0.478) with enhanced modularity compared to multi-class detection baselines (mAP 0.455), offering improved robustness across diverse building types and imbalanced damage classes.

Efficient and accurate building damage assessment is crucial for effective emergency response and resource allocation following natural hazards. However, traditional methods are often time consuming and labor intensive. Recent advancements in remote sensing and artificial intelligence (AI) have made it possible to automate the damage assessment process, and previous studies have made notable progress in machine learning classification. However, the application in postdisaster emergency response requires an end‐to‐end model that starts with satellite imagery as input and automates the generation of large‐scale damage maps as output, which was rarely the focus of previous studies. Addressing this gap, this study integrates satellite imagery, Geographic Information Systems (GIS), and deep learning. This enables the creation of comprehensive, large‐scale building damage assessment maps, providing valuable insights into the extent and spatial variation of damage. The effectiveness of this methodology is demonstrated in Galveston County following Hurricane Ike, where the classification of a large ensemble of buildings was automated using deep learning models trained on the xBD data set. The results showed that utilizing GIS can automate the extraction of subimages with high accuracy, while fine‐tuning can enhance the robustness of the damage classification to generate highly accurate large‐scale damage maps. Those damage maps were validated against historical reports.

No abstract available

Building damage assessment in the face of natural disasters is crucial for economic development, disaster relief, and post‐disaster reconstruction. However, existing algorithms often overlook the impact of the disaster class when extracting difference features from high‐resolution pre‐ and post‐disaster image pairs obtained through satellite remote sensing, without considering the influence of the disaster type, that is, the different ways in which different disasters affect buildings. To address this limitation, we propose U2DDS‐Net, a two‐stage model based on U2Net and Swin Transformer. In stage 1, U2Net locates and segments buildings in pre‐disaster images. In stage 2, we enhance the model with the disaster‐type token and the diff swin stage module, which consider the disaster type and extract difference information at multiple scales. Experimental results on the xBD dataset demonstrate the significant improvement achieved by our approach, surpassing state‐of‐the‐art performance. Our research fills the gap by considering specific disaster types, and our two‐stage model provides accurate building damage assessment across various disaster scenarios.

Nowadays, unmanned aerial vehicle (UAV) remote sensing (RS) data are key operational sources used to produce a reliable building damage map (BDM), which is of great importance in instant response and rescue operations after earthquakes. This study proposes a novel weighted ensemble transferred U-Net-based model (WETUM) consisting of two major steps to create a reliable binary BDM using UAV data. In the first step of the proposed approach, three individual initial BDMs are predicted by three pretrained U-Net-based composite networks. In the second step, these three individual predictions are linearly integrated through a proposed grid search technique so that an optimized hybrid BDM (OHBDM) incorporating complementary damage information is made. The proposed WETUM was then compared with several conventional deep learning (DL) and machine learning (ML) models. The models were compared across two pivotal scenarios, addressing the impact of diverse feature sets on model performance and generalizability. Specifically, the first scenario focused solely on spectral features (SFs), while the second incorporated both spectral and geometrical features (SGFs). To make the comparisons, this study conducted empirical analyses using UAV spectral and geometrical data acquired over Sarpol-e Zahab, Iran. The experimental findings showed that the synergic use of spectral and geometrical data boosted both DL- and ML-based approaches in damage detection. Moreover, the proposed WETUM with damage detection rate (DDR) values of 65.22% and 78.26%, respectively, for the first and second scenarios, outperformed all the compared methods. Notably, WETUM with only spectral data outperformed the random forest (RF) classifier equipped with many hand-crafted SGFs, indicating the highest potential and generalizability of the proposed WETUM for building damage evaluation in a new unseen earthquake-affected area.

Building change detection (BCD) holds significant value in the context of monitoring land use, whereas building damage assessment (BDA) plays a crucial role in expediting humanitarian rescue efforts post-disasters. To address these needs, we propose the bitemporal attention module (BAM) as an innovative cross-attention mechanism aimed at effectively capturing spatio-temporal semantic relations between a pair of bitemporal remote sensing images. Within BAM, a shifted windowing scheme has been implemented to confine the scope of the cross-attention mechanism to a specific range, not only excluding remote and irrelevant information but also contributing to computational efficiency. Moreover, existing methods for BDA often overlook the inherent order of ordinal labels, treating the BDA task simplistically as a multiclass semantic segmentation problem. Recognizing the vital significance of ordinal relationships, we approach the BDA task as an ordinal regression problem. To address this, we introduce a rank-consistent ordinal regression loss function to train our proposed change detection network, bitemporal attention transformer. Our method achieves state-of-the-art accuracy on two BCD datasets (LEVIR-CD+ and S2Looking), as well as the largest BDA dataset (xBD).

Driven by dangerous Santa Ana winds and fueled by dry vegetation, the 2025 Eaton and Palisades wildfires in California caused historic levels of devastation, ultimately becoming the second and third most destructive fires in California history. Burning at the same time and drawing from the same resources, these fires burned a combined total of 16,251 structures. The first several hours of an emerging wildfire are a crucial period for fire officials to assess potential damage and develop a timely and appropriate response. A method to quickly generate accurate estimates of structural damage is essential to providing this crucial rapid response to wildfires. In this paper, we present a machine learning approach for automated assessment of structural damage caused by wildfires. By leveraging multiple data sources in model development (satellite-based building footprints, expert-labeled post-fire damage points, fire perimeters, and aerial thermal imagery) and innovative data processing techniques, the approach can be used to identify various levels of structural damage from just aerial thermal imagery during operational use. The resulting system offers an effective approach for rapid and reliable assessment of burned structures, suitable for operational wildfire damage assessment. Results on the Eaton and Palisades Fires demonstrate the effectiveness of this method and its applicability to real-world scenarios.

Subsidence prediction is essential for preventing and controlling geohazards in coal mining areas. However, the Interferometric Synthetic Aperture Radar (InSAR) technique is limited in deriving the goaf displacements with a large gradient and fast deformation rates, hindering its application for potential risk evaluation over the mining areas. In this study, we proposed a three-dimensional and full parameter inversion (TDFPI) model to derive the large-gradient subsidence and then investigate its application for building damage assessment over coal mining areas. By taking the Guotun coal mine as the case study, the TDFPI model was demonstrated to have effectively predicted the large-gradient deformation of the mining areas and successfully evaluated the house damage in Chelou village, which agrees well with our field investigations. Specifically, the predicted subsidence results were validated with high fitting accuracy against field measurements, with RMSE of 0.083 m and 0.102 m, respectively, on observation line A and line F. In addition, the classified damage levels are highly consistent with in situ field surveys for the house cracks in Chelou village, presenting its practicality and effectiveness for building damage evaluation, and thus can provide a useful tool for potential risk assessment and prevention over the mining areas.

Rapid building damage assessment (BDA) is vital in guiding disaster response missions and estimating population distribution across impacted areas. While commercial satellite imagery providers have enabled near-daily monitoring of the Earth, near-realtime assessment of disaster scenarios frequently requires analysis of off-nadir imagery, as satellites are often far from impacted areas for at-nadir post-event imaging to occur Such scenarios are, however, underrepresented in existing BDA datasets and methodologies. With this motivation, we investigate generalization capabilities of current BDA practices across overhead view-angles and strategies for their improvement. Using a labeled dataset of images capturing conflict-related damages, we first train a baseline BDA architecture using imbalanced and balanced datasets with respect to view-angle. Then, we explore conditional convolutions parameterized on image features, image nadir, and their combination as a mechanism for conditioning on view-angles. Experiments demonstrate the limitations of current practice and the potential of conditional mechanisms to increase model robustness to view-angle variations.

The Turkey–Syria earthquakes that occurred on February 6, 2023, have caused significant human casualties and economic damage. Emergency services require quick and accurate assessments of widespread building damage in affected areas. This can be facilitated by using remote sensing methods, specifically all-day and all-weather Synthetic Aperture Radar (SAR). In this study, we aimed to improve the detection of building anomalies in earthquake-affected areas using SAR images. To achieve this, we employed Recurrent Neural Network (RNN) to train coherence time series and predict co-seismic coherence. This approach allowed us to generate a Damage Proxy Map (DPM) for building damage assessment. The results of our study indicated that the estimated proportion of building damage in Kahramanmaras was approximately 24.08%. These findings were consistent with the actual damage observed in the field. Moreover, when utilizing the mean and standard deviation of coherence time series, our method achieved higher accuracy (0.761) and a lower false alarm rate (0.136) compared to directly using coherence with only two views of SAR data. Overall, our study demonstrates that this method provides an accurate and reliable approach for post-earthquake building damage assessment.

We identify a bias in a commonly used dataset for building damage detection, evaluate its effects on existing deep learning models, and devise mitigation strategies to overcome it. We find that the data contains significantly more groups of damaged buildings than single ones leading to skewed machine learning evaluations. Consequently, deep learning models heavily rely on surrounding context rather than individual building damage when classifying supporting our claim. Specifically, the dataset includes extraneous damage surrounding buildings such as debris, fallen trees, and other damaged buildings which results in deep neural networks overfitting to these features. We analyze the top-5 solutions of the xView2 challenge, which focuses on building damage classification using satellite imagery as provided by the xBD dataset. Our experiments reveal that these models struggle to accurately identify isolated damaged buildings, potentially causing oversights in critical disaster scenarios and delaying humanitarian aid. Finally, we devise a new augmentation strategy to reduce this bias in disaster datasets and show it improves real-world outcomes.

Accurate damage estimation after earthquakes is crucial for effective post-disaster response and recovery. However, earthquakes often trigger various additional hazards, such as landslides and liquefaction, making accurate building damage estimation even more challenging. To date, despite significant research efforts, automated, accurate building-specific damage estimation has not been achieved. Our study tackles this challenge. We integrate multi-sourced global building footprints and InSAR coherence-based Change Detection Maps (CDMs) generated by the U.S. Geological Survey (USGS) within a variational causal Bayesian network, providing intricate maps of landslides, liquefaction, and building damage. Our key innovations include: 1) a novel masking strategy for the CDMs, derived from low pre-event mean coherence value and high pre-event coherence standard deviation to eliminate noisy signals in InSAR products induced by irrelevant noise sources (steep slopes, soil moisture and vegetation change, open water, etc.), and 2) variational inference to differentiate potential causes of the changes in InSAR coherence signals, specifically landslides, liquefaction, building damage, and non-hazard changes. Our strategy is critical for enhancing the accuracy of building damage and ground failure assessments, as noise from environmental or human-induced changes can obscure true damage signals. We provide reliable damage identification with attribution to specific causes by focusing on accurate building footprints and improving regional ground failure predictions using the 2023 M6.8 Morocco earthquake to validate our methodology. Our approach enables thorough damage analysis across numerous buildings, with the potential for significantly aiding disaster management and marking a substantial advancement of post-earthquake building damage assessment methods.

Sudden natural disasters and man-made disasters pose a threat to human life and property safety, and real-time semantic segmentation of high-resolution remote sensing images is crucial for disaster damage assessment applications. In recent years, with the wide application of high spatial resolution (HSR) remote sensing images and semantic change detection methods based on deep learning (DL), the acquisition of information on damaged areas has become more and more convenient and accurate. However, due to the black box characteristics of existing methods, the lack of interpretability and prior knowledge embedding (such as building positioning information), as well as the low utilization of damage conditions around the building, lead to automatically learned feature representations that still need to be improved. To solve these problems, we proposed the position-constrained differential attention Siamese network (PCDASNet). The main idea is to merge building extraction and disaster damage assessment into a cascaded framework to improve building damage recognition results under the constraints of building positioning information. In particular, the proposed differential attention module (DAM) adaptively extracts change information corresponding to buildings and surrounding environments from dual-temporal images, with interpretability and theoretical guarantee, which enables the integration of prior positioning knowledge into the design of network architecture. The objective metrics of the method on the building damage dataset show that the method achieves a test F1 score of more than 73% compared with other baseline methods, and also outperforms several state-of-the-art methods in terms of visual results.

No abstract available

The building damage status is vital to plan rescue and reconstruction after a disaster and is also hard to detect and judge its level. Most existing studies focus on binary classification, and the attention of the model is distracted. In this study, we proposed a Siamese neural network that can localize and classify damaged buildings at one time. The main parts of this network are a variety of attention U-Nets using different backbones. The attention mechanism enables the network to pay more attention to the effective features and channels, so as to reduce the impact of useless features. We train them using the xBD dataset, which is a large-scale dataset for the advancement of building damage assessment, and compare their result balanced F (F1) scores. The score demonstrates that the performance of SEresNeXt with an attention mechanism gives the best performance among single models, with the F1 score reaching 0.787. To improve the accuracy, we fused the results and got the best overall F1 score of 0.792. To verify the transferability and robustness of the model, we selected the dataset on the Maxar Open Data Program of two recent disasters to investigate the performance. By visual comparison, the results show that our model is robust and transferable.

No abstract available

No abstract available

Building damage assessment is a critical subtask within GeoAI-driven remote sensing semantic segmentation, where deep neural networks have been widely applied. Most existing works typically use pre- and post-disaster images as input of a siamese deep neural network under supervised learning, which requires a large amount of labeled data. However, in real-world scenarios, acquiring massive labeled datasets is often difficult, making fully supervised methods less practical. To overcome this, we propose a self-supervised pretraining framework based on pre-post remote sensing image pairs. In the first stage, a dual denoising autoencoder with Vision Transformer backbone is proposed for image representation learning. In the second stage, two downstream tasks-building localization and building damage severity-are performed. Additionally, we incorporate an edge guidance module and an edge detection loss to further enhance performance in downstream tasks. On the xBD dataset, the largest building damage assessment dataset, the proposed method achieves an F1 score of 0.895 for building localization, outperforming state-of-the-art image segmentation techniques. Additionally, it receives an F1 score of 0.704 in building damage severity compared to state-of-the-art in all self-supervised learning methods.

Building damage assessment is critical in regions facing geopolitical challenges. This paper explores the use of Synthetic Aperture Radar remote sensing data, specifically from the Sentinel-1 constellation, to improve the accuracy and operational efficiency of building damage assessment. The approach is based on the use of SAR backscatter time series for a pixel-wise change detection methodology. As a case study, we consider the state of Ukraine, which has experienced significant building damage due to the ongoing Russo-Ukrainian war. Using the presented approach, we demonstrate the feasibility of change detection from freely available SAR data with weekly temporal resolution and at a nationwide spatial scale.

Abstract. Access to very high-resolution (HR) satellite imagery is often limited, delayed, or cost-prohibitive, restricting accurate and timely post-disaster damage detection and recovery monitoring (PDDRM). Additionally, class imbalance in disaster classification datasets further complicates deep learning (DL)-based assessments. This study addresses these challenges by leveraging ESRGAN to enhance low-resolution (LR) satellite imagery, thereby improving damage classification accuracy and the ability to monitor post-disaster recovery over time with three state-of-the-art DL models: Vision Transformer (ViT), ConvNeXt, and MaxViT for PDDRM classification across four key recovery states: Not Damaged, Not Recovered, Recovered, and New Buildings. To generate super-resolution (SR) images, LR images were first paired with HR images to train ESRGAN. Numerical evaluations using Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM) between SR and HR images confirm that ESRGAN effectively reconstructs high-resolution features, with Not Damaged (PSNR: 29.2, SSIM: 0.78) and New Buildings (PSNR: 30.3, SSIM: 0.81) exhibiting the highest reconstruction quality. ESRGAN-generated SR images were then compared against LR images in terms of classification accuracy and reliability. The results demonstrate that SR improves classification accuracy and precision, particularly for ViT and ConvNeXt, with ViT achieving an accuracy of 84% and ConvNeXt 82% on SR images, compared to 79% and 78% on LR images. We also employed Grad-CAM++ visualizations to interpret model predictions, which highlighted reliability improvements in certain classes. This study demonstrates that SR is a scalable and cost-effective alternative to very high-resolution satellite imagery, reducing dependency on expensive data sources while improving classification accuracy for PDDRM.

On February 6, 2023, a major earthquake of 7.8 magnitude and its aftershocks caused widespread destruction in Turkey and Syria, causing more than 55,000 deaths, displacing 3 million people in Turkey and 2.9 million in Syria, and destroying or damaging at least 230,000 buildings. Our research presents detailed city-scale maps of landslides, liquefaction, and building damage from this earthquake, utilizing a novel variational causal Bayesian network. This network integrates InSAR-derived change detection with new empirical ground failure models and building footprints, enabling us to (1) rapidly estimate large-scale building damage, landslides, and liquefaction from remote sensing data, (2) jointly attribute building damage to landslides, liquefaction, and shaking, (3) improve regional landslide and liquefaction predictions impacting infrastructure, and (4) simultaneously identify damage degrees in thousands of buildings. For city-scale, building-by-building damage assessments, we use building footprints and satellite imagery with a spatial resolution of approximately 30 meters. This allows us to achieve a high resolution in damage assessment, both in timeliness and scale, enabling damage classification at the individual building level within days of the earthquake. Our findings detail the extent of building damage, including collapses, in Hatay, Osmaniye, Adıyaman, Gaziantep, and Kahramanmaras. We classified building damages into five categories: no damage, slight, moderate, partial collapse, and collapse. We evaluated damage estimates against preliminary ground-truth data reported by the civil authorities. Our results demonstrate the accuracy of our classification system, as evidenced by the area under the curve (AUC) scores on the receiver operating characteristic (ROC) curve, which ranged from 0.9588 to 0.9931 across different damage categories and regions. Specifically, our model achieved an AUC of 0.9931 for collapsed buildings in the Hatay/Osmaniye area, indicating a 99.31% probability that the model will rank a randomly chosen collapsed building higher than a randomly chosen non-collapsed building. These accurate, building-specific damage estimates, with greater than 95% classification accuracy across all categories, are crucial for disaster response and can aid agencies in effectively allocating resources and coordinating efforts during disaster recovery.

During the last decades, archaeological site looting throughout Iraq has increased significantly up to a point where some of the most famous and relevant ancient Mesopotamian cities are currently threatened in their integrity. Several important archaeological monuments and artifacts have been destroyed, due to ISIL attacks and associated looting. Since 2016, the policies of the European Union have been increasingly harsh to condemn these atrocious acts of destruction. In such a scenario, the European Union Satellite Centre can be an invaluable instrument for the identification and assessment of the damage in areas occupied by ISIL. A detailed view of the damage suffered by the Nineveh and Nebi Yunus ancient sites, in Iraq, was assessed via visual inspection. The analysis was conducted considering the main events that occurred in the city of Mosul, between November 2013 and March 2018. More than 25 satellite images, new acquisitions and archived, supported by collateral data, allowed the detection and classification of the damage occurred over time. A description of the methodology and the classification of category and type of damage is presented. The results of the analysis confirm the dramatic levels of destruction that these two ancient sites have been suffering since 2013. The analysis reported in this paper is part of a wider study that the SatCen conducted in cooperation with the EU Counter-Terrorism Office and PRISM Office. The whole activity aimed at confirming to EU institutions the massive looting and trafficking operated in the area. The results have been provided to archaeologists in the field as well in support of local authorities who are trying to evaluate the current situation in the area.

Automated feature extraction from high-resolution satellite data has great potential to locate lucrative resources on Earth. The main elements that are being extracted include Land, Buildings, Roads, Water bodies, and all other necessary elements seen in high-resolution satellite photography. The input dataset includes high-resolution satellite images of New York together with matching map pictures of buildings, roads, landscapes, and water bodies, all in common file formats like.jpg. This paper introduces an innovative feature extraction approach utilizing Deep learning techniques such as Pix2Pix GAN (Generative Adversarial Network) for semantic segmentation and improving the object delineation, such as land, buildings, roads, and water features, in image-to-image translation tasks. U-Net enables precise semantic segmentation by retaining spatial information. Simple CNN (Convolutional Neural Netwok) processes image patches sequentially, extracting relevant features such as edges, textures, and patterns, enabling accurate classification and segmentation of objects in satellite imagery. Pix2Pix GAN is shown to perform better than other models based on Structural Similarity Index Measure (SSIM), a metric used to quantify the similarity between two pictures. It achieved the highest SSIM score of 0.904.

Satellite imagery plays a crucial role in land use mapping, land cover classification, and the detection of an- thropogenic interference in natural environments. However, the effectiveness of these applications depends significantly on image quality, which is often compromised by atmospheric conditions such as clouds and shadows, as well as radiometric incon- sistencies. This paper explores the application of Generative Adversarial Networks (GANs) as a state-of-the-art approach to satellite image super-resolution. By employing a two-component architecture—a generator that upscales low-resolution inputs and a discriminator that ensures photorealistic outputs—GANs can overcome the hardware limitations of satellite sensors while preserving critical spatial details. The proposed methodology incorporates specialized loss functions to maintain structural integrity and enhance feature extraction capabilities. Results demonstrate that GAN-based super-resolution techniques not only improve visual quality but also significantly enhance the accuracy of downstream analytical tasks in urban planning, environmental monitoring, and disaster response. This approach represents a promising direction for advancing geospatial intelli- gence and remote sensing applications where high-resolution data is essential but often unavailable through conventional means. Index Terms—GAN, Planet Imagery, Normalization

Automatic separation of buildings from built-up area has attracted considerable interest in computer vision and digital photogrammetry field. While many efforts have been made for building extraction, none of them address the problem completely. This even a greater challenge in low-contrast very-high resolution (VHR) panchromatic satellite images. To alleviate this issue, a framework for automatic building detection approach using dominant structural feature (DSF) is proposed in this study. Firstly, in order to suppress noise while enhancing structural feature, contourlet transform based image contrast enhancement is employed followed by directional morphological filtering operation. Considering the structural characteristics of buildings which are significantly different from the other non-manmade objects. We then exploit DSF by means of windowed structure tensor analysis. Candidate building edges are generated using multi-seed classification technique in DSF space, subsequently. Finally, a series rule- and knowledge-based criterions are elaborate designed for false alarm reduction procedures.

Natural disasters pose significant challenges to timely and accurate damage assessment due to their sudden onset and the extensive areas they affect. Traditional assessment methods are often labor-intensive, costly, and hazardous to personnel, making them impractical for rapid response, especially in resource-limited settings. This study proposes a novel, cost-effective framework that leverages aerial drone footage, an advanced AI-based video super-resolution model, Video Restoration Transformer (VRT), and Gemma3:27b, a 27 billion parameter Visual Language Model (VLM). This integrated system is designed to improve low-resolution disaster footage, identify structural damage, and classify buildings into four damage categories, ranging from no/slight damage to total destruction, along with associated risk levels. The methodology was validated using pre- and post-event drone imagery from the 2023 Turkey earthquakes (courtesy of The Guardian) and satellite data from the 2013 Moore Tornado (xBD dataset). The framework achieved a classification accuracy of 84.5%, demonstrating its ability to provide highly accurate results. Furthermore, the system's accessibility allows non-technical users to perform preliminary analyses, thereby improving the responsiveness and efficiency of disaster management efforts.

Detailed structural building information is used to estimate potential damage from hazard events like cyclones, floods, and landslides, making them critical for urban resilience planning and disaster risk reduction. However, such information is often unavailable in many small island developing states (SIDS) in climate-vulnerable regions like the Caribbean. To address this data gap, we present an AI-driven workflow to automatically infer rooftop attributes from high-resolution satellite imagery, with Saint Vincent and the Grenadines as our case study. Here, we compare the utility of geospatial foundation models combined with shallow classifiers against fine-tuned deep learning models for rooftop classification. Furthermore, we assess the impact of incorporating additional training data from neighboring SIDS to improve model performance. Our best models achieve F1 scores of 0.88 and 0.83 for roof pitch and roof material classification, respectively. Combined with local capacity building, our work aims to provide SIDS with novel capabilities to harness AI and Earth Observation (EO) data to enable more efficient, evidence-based urban governance.

Developments in Geographic Information Systems and Remote Sensing (RS) technologies and innovative approaches emerging in deep learning (DL) supported analysis methods have an important place in disaster research as in every field. Convolutional neural networks such as Mask RCNN, U-NET, one of the deep learning methods for disaster damage impact assessment and classification, have started to show successful results. However, high-resolution geospatial imagery and drones provide faster and more accurate detection of structural damage.  In this study, damaged building detection was performed using Göktürk-1 satellite images from 6 February 2023 using Mask-RCNN architecture. In this study, deep learning methods were used to detect the collapsed buildings in the city of Malatya during the 6 February 2023 earthquakes. The study aims to emphasize the significance of GIS and remote sensing for the timely and efficient evaluation of building damage after a disaster. Considering this, high quality images of Malatya city before and after the earthquake were analyzed and data sets were created by masking using Mask RCNN deep learning method through ArcGIS Pro 3.4.0 software. According to the results of the research, it quickly detected damaged buildings with an accuracy rate of 70% according to satellite images after the earthquake. As a result, GIS and deep learning models were used to detect and map the initial damage after the earthquake.

Timely and accurate urban change detection is vital for sustainable urban development, infrastructure management, and disaster response. Traditional two-dimensional approaches often overlook vertical and structural variations in dense urban areas. This study proposes a three-dimensional (3D) change detection framework that integrates high-resolution optical imagery and Digital Surface Models (DSMs) from two time points to capture both horizontal and vertical transformations. The method is based on a deep learning architecture combining a ResNet34 encoder with a UNet++ decoder, enabling the joint learning of spectral and elevation features. The research was carried out in two stages. First, a binary classification model was trained to detect areas of change and no-change, allowing direct comparison with conventional methods such as Principal Component Analysis (PCA), Change Vector Analysis (CVA) with thresholding, K-Means clustering, and Random Forest classification. In the second stage, a multi-class model was developed to categorize the types of structural changes, including new building construction, complete destruction, building height increase, and height decrease. Experiments conducted on a high-resolution urban dataset demonstrated that the proposed CNN-based framework significantly out-performed traditional methods, achieving an overall accuracy of 96.58%, an F1-score of 96.58%, and a recall of 96.7%. Incorporating DSM data notably improved sensitivity to elevation-related changes. Overall, the ResNet34–UNet++ architecture offers a robust and accurate solution for 3D urban change detection, supporting more effective urban monitoring and planning.

No abstract available

Accurate assessment of post-disaster damage is essential for prioritizing emergency response, yet current practices rely heavily on manual interpretation of satellite imagery.This approach is time-consuming, subjective, and difficult to scale during large-area disasters. Although recent deep-learning models for semantic segmentation and change detection have improved automation, many of them still struggle to capture subtle structural variations and often perform poorly when dealing with highly imbalanced datasets, where undamaged buildings dominate. This thesis introduces Satellite-to-Street:Disaster Impact Estimator, a deep-learning framework that produces detailed, pixel-level damage maps by analyzing pre and post-disaster satellite images together. The model is built on a modified dual-input U-Net architecture that strengthens feature fusion between both images, allowing it to detect not only small, localized changes but also broader contextual patterns across the scene. To address the imbalance between damage categories, a class-aware weighted loss function is used, which helps the model better recognize major and destroyed structures. A consistent preprocessing pipeline is employed to align image pairs, standardize resolutions, and prepare the dataset for training. Experiments conducted on publicly available disaster datasets show that the proposed framework achieves better classification of damaged regions compared to conventional segmentation networks.The generated damage maps provide faster and objective method for analyzing disaster impact, working alongside expert judgment rather than replacing it. In addition to identifying which areas are damaged, the system is capable of distinguishing different levels of severity, ranging from slight impact to complete destruction. This provides a more detailed and practical understanding of how the disaster has affected each region.

Spiking Neural Networks (SNNs) have shown potential for building damage detection after natural disasters. However, most existing methods focus solely on basic segmentation tasks, limiting their ability to assess detailed post-disaster damage. The conversion from Artificial Neural Networks (ANNs) to SNNs presents challenges, such as energy inefficiency and information loss. To address these limitations, we propose a novel Spiking-Siamese-UNet model that combines SNNs, Siamese Networks, and UNet architectures to enhance building damage classification from satellite imagery. By introducing multi-threshold spiking neurons and connection-wise normalization, and utilizing the xBD dataset, the model captures structural changes and damage severity with high precision. Experimental results demonstrate that the Spiking-Siamese-UNet outperforms traditional approaches in both damage localization and classification, offering an efficient solution for scalable disaster damage assessment, Our code is available in https://github.com/slzhang01/Spiking-Siamese-UNet.

No abstract available

Recent advancements in remote sensing technologies have enabled the development of innovative methods to improve search and rescue efforts, assess structural damage, and streamline disaster management following earthquakes. Various imaging modalities—such as satellite imagery, UAVs, airborne platforms, and terrestrial systems—are utilized for damage evaluation and debris detection. Distinguishing post-earthquake imagery from pre-event data captured by UAVs is essential for effective disaster response. In this context, deep learning models have been increasingly employed to rapidly classify images as "damaged" or "undamaged." This study offers a comprehensive comparison of multiple deep learning architectures for image classification, including six convolutional neural networks (CNNs) and three variants of Vision Transformers (ViTs). The results show that Vision Transformer models, especially the ViT Large variant, consistently outperformed CNN counterparts, achieving an accuracy of 96.12%, sensitivity of 99.18%, and MCC of 0.9187. The superior performance of ViTs highlights their enhanced ability to discriminate between damaged and undamaged classes, providing balanced and reliable results. Furthermore, the ViT Large model surpassed results from previous earthquake image classification studies, demonstrating strong generalization on domain-specific datasets. These outcomes underscore the promise of transformer-based architectures as leading-edge solutions for complex and imbalanced image classification challenges in disaster assessment.

Detecting damaged buildings quickly and accurately is of great significance for the disaster area to carry out emergency rescue and decision-making. However, the majority of traditional methods based on scene classification uses a fixed-size sliding window to traverse the very-high-resolution (VHR) satellite images, ignoring the scale difference among the geographical objects, which may lead to poor objects delineation and large time consumption. In this paper, we propose a combined multi-scale segmentation and scene change detection (MSSCD) strategy for detecting damaged buildings from pre- and post-disaster VHR satellite images. We compared the performance of four different scene classification structures in detecting damaged buildings in the 2010 Haiti earthquake, and the experimental results show that the proposed strategy is better than traditional methods.

Compared with optical sensors, Synthetic Aperture Radar (SAR) can provide important damage information due to its ability to map areas affected by earthquakes independently from weather conditions and solar illumination. In 2013, a new TerraSAR-X mode named staring spotlight (ST), whose azimuth resolution was improved to 0.24 m, was introduced for various applications. This data source made it possible to extract detailed information from individual buildings. In this paper, we present a new concept for individual building damage assessment using a post-event sub-meter very high resolution (VHR) SAR image and a building footprint map. With the building footprint map, the original footprints of buildings can be located in the SAR image. Based on the building imaging analysis of a building in the SAR image, the features in the building footprint can be extracted to identify standing and collapsed buildings. Three machine learning classifiers, including random forest (RF), support vector machine (SVM) and K-nearest neighbor (K-NN), are used in the experiments. The results show that the proposed method can obtain good overall accuracy, which is above 80% with the three classifiers. The efficiency of the proposed method is demonstrated based on samples of buildings using descending and ascending sub-meter VHR ST images, which were all acquired from the same area in old Beichuan County, China.

Building damage identification shortly after a disaster is crucial for guiding emergency response and recovery efforts. Although optical satellite imagery is commonly used for disaster mapping, its effectiveness is often hampered by cloud cover or the absence of pre-event acquisitions. To overcome these challenges, we introduce a novel multimodal deep learning (DL) framework for detecting building damage using single-date very high resolution (VHR) Synthetic Aperture Radar (SAR) imagery from the Italian Space Agency's (ASI) COSMO-SkyMed (CSK) constellation, complemented by auxiliary geospatial data. Our method integrates SAR image patches, OpenStreetMap (OSM) building footprints, digital surface model (DSM) data, and structural and exposure attributes from the Global Earthquake Model (GEM) to improve detection accuracy and contextual interpretation. Unlike existing approaches that depend on pre- and post-event imagery, our model utilizes only post-event data, facilitating rapid deployment in critical scenarios. The framework's effectiveness is demonstrated using a new dataset from the 2023 Kahramanmaraş earthquake in Türkiye, covering multiple cities with diverse urban settings. Results highlight that incorporating geospatial features significantly enhances detection performance and generalizability to previously unseen areas. By combining SAR imagery with detailed vulnerability and exposure information, our approach provides reliable and rapid building damage assessments without the dependency from available pre-event data. Moreover, the automated and scalable data generation process ensures the framework's applicability across diverse disaster-affected regions, underscoring its potential to support effective disaster management and recovery efforts.

No abstract available

No abstract available

No abstract available

Earthquakes have devastating effects on densely urbanised regions, requiring rapid and extensive damage assessment to guide resource allocation and recovery efforts. Traditional damage assessment is time-consuming, resource-intensive, and faces challenges in covering vast affected areas, often limiting timely decision-making. Space-borne synthetic aperture radars (SAR) have gained attention for their all-weather and day-night imaging capabilities. These advantages, coupled with wide coverage, short revisits and very high resolution (VHR), have created opportunities for using SAR data in disaster response. However, most SAR studies for post-earthquake damage assessment rely on change detection methods using pre-event SAR images, which are often unavailable in operational scenarios. Limited studies using solely post-event SAR data primarily concentrate on city-block-level damage assessment, thus not fully exploiting the VHR SAR potential. This paper presents a novel method integrating solely post-event VHR SAR imagery and machine learning (ML) for regional-scale post-earthquake damage assessment at the individual building-level. We first used supervised learning on case-specific datasets, and then introduced a combined learning approach, incorporating inventories from multiple case studies to assess generalisation. Finally, the ML model was tested on unseen study areas, to evaluate its flexibility in unfamiliar contexts. The method was implemented using datasets collected during the Earthquake Engineering Field Investigation Team (EEFIT) reconnaissance missions following the 2021 Nippes earthquake and the 2023 Kahramanmaraş earthquake sequence. The results demonstrate the method’s ability to classify standing and collapsed buildings, achieving up to 72% overall accuracy on unseen regions. The proposed method has potential for future disaster assessments, thereby contributing to more effective earthquake management strategies.

An accurate and quick detection and seismic classification of the building earthquake damage is significant for disaster emergency and rescue. This paper aimed at quickly, precisely and efficiently detecting building damage information. According to the analysis of the problems existing in present research, a technical process for the VHR remote sensing image of the building earthquake damage information object-oriented change detection was proposed for its extraction through different degrees of improvement and innovation for the key technologies. The final building earthquake damage information result was output by accuracy evaluation through using classification evaluation criteria and other indicators. The change detection accuracy of the Yushu small area is 88.45% with the Kappa coefficient was 0.8411. It was proved and verified that the proposed method can make up for the classification deficiency based on the single data source, and realized the complementary advantages among the classifiers, which can improve the classification accuracy.

No abstract available

No abstract available

No abstract available

No abstract available

No abstract available

ABSTRACT Natural and man-made disasters take place around the world and cause significant financial and human losses. An accurate and fast post-disaster building damage mapping could play a crucial role in rapid rescue planning and operations. Remote sensing satellite images are the main source of building damage map generation. Usually, both pre-disaster and post-disaster satellite images are used for the generation of building damage maps, which encounter some challenges such as registration errors, noise, and atmospheric conditions. This study proposed a new method for building damage detection based only on post images with the UNet architecture network and Global Context Vision Transformer blocks. The deep learning network proposed in this research is automatic without any further processing. The proposed method comprises four main steps: (1) pre-processing, (2) network training, (3) building damage map generation, and (4) accuracy assessment for the final damage map. This network is applied to three different natural and man-made disaster datasets. The first dataset is the post-satellite image of the 2023 Turkey earthquake, the second one is the post-satellite image of the 2021 Bata explosion and the last one is the post-satellite image of the 2011 haiti earthquake. Results of the final building damage map indicate that the proposed method is highly effective with OA above 96%, which is superior to the other deep learning methods.

The Mw 7.7 Myanmar earthquake on March 28, 2025, caused widespread building damage along the Sagaing Fault. Rapid, accurate assessment of building-level damage is critical for emergency response and resource allocation. This study integrates Sentinel-1 SAR-derived Damage Proxy Maps (DPM), Modified Mercalli Intensity (MMI) values, and field-verified ground truth observations for 325 buildings in Nay Pyi Taw (125 severe, 67 moderate, 133 undamaged) to evaluate machine learning-based damage detection. SVM, Random Forest, and XGBoost classifiers were trained on mean DPM and MMI features, with class imbalance addressed through random oversampling. XGBoost achieved the highest precision for severe and undamaged buildings, Random Forest excelled in moderate damage, while misclassifications occurred primarily for intermediate cases. DPM was confirmed as the most influential predictor, with MMI enhancing performance across classes. The framework demonstrates that combining SAR-derived structural changes and seismic intensity enables rapid, reliable building damage detection and can be adapted to other regions, earthquake scenarios, and multi-hazard contexts, supporting timely disaster response and mitigation planning.

Building damage detection is crucial for post-disaster assessment and relief planning. Damage detection tasks have seen considerable enhancement through the application of deep learning techniques, notably convolutional neural networks (CNNs). This study presents a unique dual-branch ConvNeXt model enhanced with an Attention-Based Fusion (ABF) mechanism and a Cross-Attention Module (CAM) for damage assessment using the xBD dataset. Our method utilizes a two-stream architecture, where pre-disaster and post-disaster images are processed independently through ConvNeXt encoders, and their feature representations are refined and merged using an attention-based fusion mechanism. The Cross-Attention Module (CAM) is introduced to model the interdependencies between pre- and post-disaster images, effectively capturing structural differences while preserving unchanged regions. By learning spatial and temporal patterns jointly, this method enhances classification precision for damaged structures. We evaluate our method on the xBD dataset and achieve superior performance compared to existing CNN-based architectures. Our model achieves F1b of 89.3%, IoU of 68.5%, and F1d of 80.5%, outperforming BDANet and SiamixFormer by approximately 2% and 1.4% respectively.

Wildfires are a growing global concern, causing significant damage to urban infrastructure each year. This study presents a novel approach for building damage assessment using generative artificial intelligence, focusing on the transferability of high-resolution satellite imagery models to lower-resolution datasets. Our diffusion-based model is trained on the xView2 Wildfire Building Damage Benchmark, a dataset specifically designed for wildfire-induced building damage detection. The model is further evaluated on real-world wildfire incidents in Lahaina, Hawaii, and Athens, Greece, demonstrating its effectiveness in damage localization across varying spatial resolutions. With competitive performance on benchmark datasets and practical utility in real-world scenarios, this work highlights the potential of generative AI for geospatial disaster assessment and urban resilience.

The increasing frequency of natural disasters neces-sitates efficient building damage detection for effective disaster response. This study addresses limitations in deep learning models, particularly their inability to classify minor as well as major damage classes due to inadequate detection of structural features like edges and corners in satellite images. To overcome these challenges, we propose the utilization of a fusion-based data augmentation technique that combines edge detection, contrast enhancement, and unsharp masking to enhance structural feature detection. We further evaluate the generalizability of this approach using domain adaptation techniques, including supervised fine-tuning and unsupervised Deep CORAL to address domain shifts between source (xBD) and target (Ida-BD) datasets. Experimental results demonstrate that the proposed augmentation improves damage classification accuracy by 5–7% in minor and major damage classes and enhances localization accuracy by 2.5%. Additionally, the integration of domain adaptation techniques validates the robustness in handling out-of-domain datasets. By improving structural feature detection and mitigating domain discrepancies, the proposed methodology enhances performance and adaptability of deep learning models for disaster response. This study demonstrates the potential of fusion-based augmentation and domain adaptation to enable reliable and efficient building damage detection in diverse disaster scenarios.

To address the limitations of existing models in multi-scale feature extraction and boundary localization, this study proposes DeepCrack, a deep learning model that integrates a spatial attention mechanism with dilated convolutions to enhance automation and accuracy in building damage detection. The model adopts a lightweight MobileNetV3 as its backbone, incorporating FPN-based multi-scale feature fusion, a DeepLabV3+ atrous convolution module, boundary-aware loss optimization, and CRF-based post-processing to form an efficient segmentation framework. Performance evaluations and ablation studies conducted on the Crack500 dataset demonstrate that DeepCrack outperforms UNet, PSPNet, and DeepLabV3+ in terms of precision, recall, and IoU. Each integrated module contributes significantly to performance gains. The proposed model offers a scalable and high-performance solution for intelligent structural damage recognition.

No abstract available

The detection and statistics of damaged buildings are crucial for rescue work, especially when detecting object-level and small buildings. Existing detection technologies mainly focus on pixel-level analysis, often ignoring the significance of object-level analysis and the identification of small buildings. To address these challenges, a cross-temporal progressive enhancement model (CTPEM) is proposed. In CTPEM, the visual center guided enhancement (VCGE) module is proposed, which uses multiple visual centers to learn the binary relationship between pretemporal and post-temporal features to capture global and local information. Meanwhile, the progressive small object enhancement (PSOE) module is proposed, which is used to capture tiny and small-sized features through multiple feature maps at varying depths, aiming to reduce the influence of gradually vanishing features. Compared with state-of-the-art (SOTA) methods, CTPEM achieves superior performance in detecting damaged buildings, particularly small and tiny targets. Extensive experimental results demonstrate the effectiveness and practical value of our approach for accurate casualty assessment, and further facilitate more efficient disaster response and optimal resource allocation. The CTPEM code is available at https://github.com/GZPLHJ181107/CTPEM.

Natural disasters continuously threaten populations worldwide, with hydrometeorological events standing out due to their unpredictability, rapid onset, and significant destructive capacity. However, developing countries often face severe budgetary constraints and rely heavily on international support, limiting their ability to implement optimal disaster response strategies. This study addresses these challenges by developing and implementing YOLOv8-based deep learning models trained on high-resolution satellite imagery from the Maxar GeoEye-1 satellite. Unlike prior studies, we introduce a manually labeled dataset, consisting of 1400 undamaged and 1200 damaged buildings, derived from pre- and post-Hurricane Maria imagery. This dataset has been publicly released, providing a benchmark for future disaster assessment research. Additionally, we conduct a systematic evaluation of optimization strategies, comparing SGD with momentum, RMSProp, Adam, AdaMax, NAdam, and AdamW. Our results demonstrate that SGD with momentum outperforms Adam-based optimizers in training stability, convergence speed, and reliability across higher confidence thresholds, leading to more robust and consistent disaster damage predictions. To enhance usability, we propose deploying the trained model via a REST API, enabling real-time damage assessment with minimal computational resources, making it a low-cost, scalable tool for government agencies and humanitarian organizations. These findings contribute to machine learning-based disaster response, offering an efficient, cost-effective framework for large-scale damage assessment and reinforcing the importance of model selection, hyperparameter tuning, and optimization functions in critical real-world applications.

This study addresses the approaches for satellite image analysis to assess infrastructure damage. The main aim is to conduct a comprehensive comparative analysis of the effectiveness of two key machine learning approaches: specialized semantic segmentation based on the U-Net architecture and generalized visual analysis using large vision-language models. The object of the research is the process of quantitatively benchmarking these two distinct approaches to determine their practical applicability for multi-class damage classification. The research material is the publicly available xView2 dataset. The methods involved two parallel experiments. For the semantic segmentation approach, a U-Net model with an EfficientNet-B4 encoder was implemented and trained on 6-channel input data ("before" and "after" images) using a combined Dice and Focal loss function. For the vision-language models approach, the open-source LLaVA-1.5-7B model was evaluated in a zero-shot mode using advanced prompt engineering for an aggregative counting task. To enable a direct comparison, the standard Jaccard index was calculated based on the aggregated object counts for each damage class. The results of the experiments revealed a significant performance disparity. The specialized U-Net model demonstrated high effectiveness, achieving an intersection over union score of 0.6141 on the test set. In contrast, the LLaVA model proved unsuitable for accurate quantitative analysis, yielding an extremely low Jaccard index of approximately 0.063, primarily due to its systemic failure to correctly identify and count objects (Recall ≈ 0.07). The scientific novelty lies in being the first study to quantitatively document this order-of-magnitude capability gap, confirming that for tasks requiring high-precision mapping, specialized segmentation models remain the indispensable tool.

No abstract available

ABSTRACT Deep learning has been extensively utilized in the assessment of building damage after disasters. However, the field of building damage segmentation faces challenges, such as misjudged regions, high network complexity, and long running times. Hence, this paper proposes a two-stage building damage assessment network called the Efficient Channel Attention and Depthwise Separable Convolutional Neural Network (ECADS-CNN). It aims to quickly detect the types of disaster damage in buildings. Deep object segmentation and deep damage classification networks were integrated into a unified building damage detection network. In this study, the efficient channel attention (ECA) module was used to enhance the performance of building semantic segmentation, and a depthwise separable (DS) module was added to the dimension upscaling process. Finally, untrained disaster dataset images were used to test the robustness of the proposed model by comparing the evaluation results of each disaster. The experiments involve testing a total of five common deep learning models, and the results indicate that the ECADS-CNN model improves the speed by 7.4% and the overall F1 score by 5.2% compared with the baseline model. The comprehensive performance is better than that of mainstream deep learning models.

Natural disasters commonly occur in all regions around the world and cause huge financial and human losses. One of the main effects of earthquakes and floods is the destruction of buildings. Photogrammetric and remote sensing (RS) data track changes and detect damages in these events. Considering the evolution in deep learning (DL) techniques, the possibility of accurate information extraction from the RS-based data is increased. DL methods effectively show the damaged regions for decision making and immediate actions for crisis management. The present study is based only on postevent RS images, which apply an encoder–decoder network composed of pretrained EfficientViTB and Yolov8 network blocks as encoder path and the modified-Unet blocks as decoder path for building damage detection (BDD). Compared with methods that use only one network in their encoder path, the presented method achieves better results. To investigate the performance of the proposed method, three datasets affected by different natural disasters are considered. The first dataset is the satellite images of the 2023 Turkey earthquake, the second dataset is associated with the satellite images of the 2023 Morocco earthquake, and the third dataset contains the satellite images of the 2023 Libya flood. The proposed method ultimately reaches the overall accuracy of 97.62%, 98.63%, and 96.43% and the kappa coefficient of 0.86, 0.85, and 0.84 for the first, second, and third dataset, respectively, which shows the proper performance of the proposed method for BDD.

We construct a strong baseline method for building damage detection by starting with the highly-engineered winning solution of the xView2 competition, and gradually stripping away components. This way, we obtain a much simpler method, while retaining adequate performance. We expect the simplified solution to be more widely and easily applicable. This expectation is based on the reduced complexity, as well as the fact that we choose hyperparameters based on simple heuristics, that transfer to other datasets. We then re-arrange the xView2 dataset splits such that the test locations are not seen during training, contrary to the competition setup. In this setting, we find that both the complex and the simplified model fail to generalize to unseen locations. Analyzing the dataset indicates that this failure to generalize is not only a model-based problem, but that the difficulty might also be influenced by the unequal class distributions between events. Code, including the baseline model, is available under https://github.com/PaulBorneP/Xview2_Strong_Baseline

Abstract. Building collapse is a major cause of casualties after an earthquake, so accurately extracting building damage information is critical for post-earthquake assessment and rescue. Currently, most deep learning methods focus on the end-to-end detection of building collapse. However, in real-world earthquake scenarios, the end-to-end computational process often lacks flexibility and struggles to meet the requirements of rapid emergency response. To address this issue, this paper proposes a cascaded framework that combines pre-earthquake building extraction and post-earthquake building damage classification. The proposed framework includes two sections: (1) Progressive building semantic segmentation model in the joint frequency domain. This model is designed to accurately extract buildings prior to an earthquake, with the goal of minimizing error propagation throughout the cascading process. The model addresses the spatial similarity of buildings under complicated backgrounds, as well as the high internal heterogeneity of buildings, by utilizing frequency domain techniques. It compensates for the shortcomings of traditional models in terms of incomplete information extraction through the effective integration of global and local information. Finally, the model employs edge priors for edge regularization. (2) Rapid building damage classification process. Based on the accurate building extraction results, a fast and efficient classification process is developed. This process uses a simple and lightweight classification network to effectively extract building damage information caused by the earthquake. The superiority of the proposed framework is validated through comparison with traditional cascading architectures and end-to-end models. The results show that the cascading framework not only provides accurate pre-earthquake building extraction, but also enables efficient and accurate post-earthquake damage classification, which meets the requirements of rapid post-earthquake emergency response. This balance of accuracy and speed is essential for effective disaster management and recovery.

Human life is significantly impacted by ongoing natural disasters. The global impact of the building damage they inflict is deeply significant. Deep learning can effectively analyze the extent of damage to buildings. However, the challenge is the availability of the dataset. An innovative, challenging dataset collected from satellites has been presented, encompassing the extent of structural harm caused by the earthquake in Morocco. The dataset is from a complex, practical, arid environment where the building colour is almost like the background. The dataset has been tested using U-Net, FCN, and ResUnet models. The ResUnet demonstrates superior performance metrics with pixel accuracy, recall, precision, Fl-score, kappa score, and IoU values of 96.2%, 86.4%, 86%, 86.2%, 75.7%, and 85.5%, respectively, which promotes it for practical application.

One of the most destructive natural disasters, hurricanes destroy a great deal of property, harm people, and disrupt the ecosystem and world economy. Earlier, through ground surveys, the impact of the hurricane was measured, by locating the damaged and undamaged infrastructure in the affected area, which is a very time-consuming and labor-intensive method. Damage detection is carried out for rapid response and recovery. A fast and enhanced method is proposed in this study that uses a hybrid ensemble model on the post-hurricane satellite images of the 2017 Hurricane Harvey in Greater Houston. The structured approach that includes data handling, model development, and image preparation is introduced. Leveraging a hybrid ensemble model combining EfficientNetB0, VGG19, and InceptionResNetV2, the proposed model is built, trained, fine-tuned, and tested, achieving improved accuracy of 93%, 90% precision, 99% recall, 85% AUC, and 95% F1-score in automated damage detection. This framework not only accelerates disaster recovery but also offers valuable insights for future resilience planning.

No abstract available

Turkey is located in a region with a high density of fault lines, which makes it susceptible to a significant earthquake risk. The Kahramanmaraş earthquake on February 6, 2023, was one of the most devastating in recent years, causing extensive damage and loss. This study aims to support post-disaster rapid response and rescue operations by using deep learning techniques to detect and classify damaged and intact buildings from satellite images. Satellite images of the Kahramanmaraş and Antakya regions, with a resolution of 8192x4537, were obtained via Google Earth Pro. The images were labeled as damaged or undamaged using the Labelme editor, which generated JSON format files for the labeled images. Using Google Colab, the JSON files and unlabeled images were merged, and buildings were cropped and categorized into two classes: damaged and undamaged. As a preprocessing step, interpolation was applied, resulting in 2211 images with a size of 128x128. A Convolutional Neural Network [2] algorithm was created using TensorFlow, a Python library, via Google Colab. The performance metrics, including accuracy, loss, F1 score, ROC curve, precision, recall, and confusion matrix values, were compared based on the experiments.

Satellites offer a unique, comprehensive viewpoint for information retrieval from conflict zones worldwide. This study leverages medium-resolution (GSD 3 m) satellite imagery from the commercial provider Planet Labs to analyze the immediate infrastructural changes resulting from the recent conflict in the Gaza Strip, starting in October 2023. Utilizing an empirical model, we detect and quantify these changes, providing a detailed assessment of the conflict’s impact on the region’s infrastructure. The results of our investigation are evaluated, revealing its reliability in change detection in the affected area. This paper underscores the critical role of satellite imagery in conflict analysis and offers insights for future humanitarian and reconstruction efforts.

Current deep learning‐based building damage detection methods often suffer from limited accuracy and high computational complexity. To address these limitations, we propose a novel framework based on the YOLOv11 architecture, termed adaptive hybrid dynamic YOLO (AHD‐YOLO). AHD‐YOLO introduces three key innovations. Omni‐dimensional dynamic fusion (ODFusion) enhances the adaptability and precision of feature extraction. Adaptive in‐scale feature interaction (AIFI) captures fine‐grained damage features. Adaptive high‐level screening feature fusion pyramid network (AHSFPN) emphasizes critical damage regions while maintaining a lightweight design. Experiments conducted on the building damage dataset show that AHD‐YOLO achieves 70.5% mAP, 60.2% Recall, and 48.2% mAP@0.5:0.95, representing respective improvements of 2.1%, 1.3%, and 1.7% over YOLOv11s. Moreover, the model also reduces the number of parameters and GFLOPs by 11.0% and 13.0%, respectively. Comparative experiments indicate that AHD‐YOLO outperforms current state‐of‐the‐art detection methods. In generalization tests on a structural damage dataset, the model achieves 78.3% detection accuracy, exceeding YOLOv11s by 2.7%. These results confirm that AHD‐YOLO effectively balances detection precision and computational efficiency, enabling accurate and real‐time identification of multiple damage types in practical building inspection scenarios.

Urban infrastructures have become imperative to human life. Any damage to these infrastructures as a result of detrimental activities would accrue huge economical costs and severe casualties. War in particular is a major anthropogenic calamity with immense collateral effects on the social and economic fabric of human nations. Therefore, damaged buildings assessment plays a prominent role in post-war resettlement and reconstruction of urban infrastructures. The data-analysis process of this assessment is essential to any post-disaster program and can be carried out via different formats. Synthetic Aperture Radar (SAR) data and Interferometric SAR (InSAR) techniques help us to establish a reliable and fast monitoring system for detecting post-war damages in urban areas. Along this thread, the present study aims to investigate the feasibility and mode of implementation of Sentinel-1 SAR data and InSAR techniques to estimate post-war damage in war-affected areas as opposed to using commercial high-resolution optical images. The study is presented in the form of a survey to identify urban areas damaged or destroyed by war (Islamic State of Iraq and the Levant, ISIL, or ISIS occupation) in the city of Mosul, Iraq, using Sentinel-1 (S1) data over the 2014–2017 period. Small BAseline Subset (SBAS), Persistent Scatterer Interferometry (PSI) and coherent-intensity-based analysis were also used to identify war-damaged buildings. Accuracy assessments for the proposed SAR-based mapping approach were conducted by comparing the destruction map to the available post-war destruction map of United Nations Institute for Training and Research (UNITAR); previously developed using optical very high-resolution images, drone imagery, and field visits. As the findings suggest, 40% of the entire city, the western sectors, especially the Old City, were affected most by ISIS war. The findings are also indicative of the efficiency of incorporating Sentinel-1 SAR data and InSAR technique to map post-war urban damages in Mosul. The proposed method could be widely used as a tool in damage assessment procedures in any post-war reconstruction programs.

The goal of the two year Ground European Network for Earth Science Interoperations-Digital Repositories (GENESI-DR) project was to build an open and seamless access service to Earth science digital repositories for European and world-wide science users. In order to showcase GENESI-DR, one of the developed technology demonstrators focused on fast search, discovery, and access to remotely sensed imagery in the context of post-disaster building damage assessment. This paper describes the scenario and implementation details of the technology demonstrator, which was developed to support post-disaster damage assessment analyst activities. Once a disaster alert has been issued, response time is critical to providing relevant damage information to analysts and/or stakeholders. The presented technology demonstrator validates the GENESI-DR project data search, discovery and security infrastructure and integrates the rapid urban area mapping and the near real-time orthorectification web processing services to support a post-disaster damage needs assessment analysis scenario. It also demonstrates how the GENESI-DR SOA can be linked to web processing services that access grid computing resources for fast image processing and use secure communication to ensure confidentiality of information.

Timely and accurate structural damage assessment is essential for effective post‐earthquake response, especially in large‐scale disasters such as the February 2023 Türkiye earthquake. Manual inspections are slow and subjective, while current deep learning (DL) approaches remain limited by binary classification, weak contextual modeling, and high computational demands. The proposed model is evaluated on a high‐resolution UAV‐based earthquake damage dataset collected from post‐disaster urban regions in Türkiye. STCHMDA‐CVT achieves 99.27% precision, recall, and F1‐score, outperforming six traditional machine learning models, six deep CNNs, and five state‐of‐the‐art attention‐based architectures. These results position STCHMDA‐CVT as a robust, efficient, and interpretable solution for automated structural damage assessment in post‐earthquake scenarios. Gradient‐weighted class activation mapping (Grad‐CAM) visualizations further enhance interpretability by highlighting structural regions critical to the model's decision‐making. While the framework is validated on Türkiye data, it can be adapted to other seismic contexts through region‐specific calibration, such as fine‐tuning with local building typologies, construction materials, and seismic intensity distributions.

No abstract available

No abstract available

No abstract available

Post-disaster damage mapping is an essential task following tragic events such as hurricanes, earthquakes, and tsunamis. It is also a time-consuming and risky task that still often requires the sending of experts on the ground to meticulously map and assess the damages. Presently, the increasing number of remote-sensing satellites taking pictures of Earth on a regular basis with programs such as Sentinel, ASTER, or Landsat makes it easy to acquire almost in real time images from areas struck by a disaster before and after it hits. While the manual study of such images is also a tedious task, progress in artificial intelligence and in particular deep-learning techniques makes it possible to analyze such images to quickly detect areas that have been flooded or destroyed. From there, it is possible to evaluate both the extent and the severity of the damages. In this paper, we present a state-of-the-art deep-learning approach for change detection applied to satellite images taken before and after the Tohoku tsunami of 2011. We compare our approach with other machine-learning methods and show that our approach is superior to existing techniques due to its unsupervised nature, good performance, and relative speed of analysis.

Accurate mapping of hurricane-induced damage is essential for guiding rapid disaster response and long-term recovery planning. This study evaluates the Three-Dimensional Multi-Attributes, Multiscale, Multi-Cloud (3DMASC) framework for semantic classification of pre- and post-hurricane Light Detection and Ranging (LiDAR) data, using Mexico Beach, Florida, as a case study following Hurricane Michael. The goal was to assess the framework’s ability to classify stable landscape features and detect damage-specific classes in a highly complex post-disaster environment. Bitemporal topo-bathymetric LiDAR datasets from 2017 (pre-event) and 2018 (post-event) were processed to extract more than 80 geometric, radiometric, and echo-based features at multiple spatial scales. A Random Forest classifier was trained on a 2.37 km2 pre-hurricane area (Zone A) and evaluated on an independent 0.95 km2 post-hurricane area (Zone B). Pre-hurricane classification achieved an overall accuracy of 0.9711, with stable classes such as ground, water, and buildings achieving precision and recall exceeding 0.95. Post-hurricane classification maintained similar accuracy; however, damage-related classes exhibited lower performance, with debris reaching an F1-score of 0.77, damaged buildings 0.58, and vehicles recording a recall of only 0.13. These results indicate that the workflow is effective for rapid mapping of persistent structures, with additional refinements needed for detailed damage classification. Misclassifications were concentrated along class boundaries and in structurally ambiguous areas, consistent with known LiDAR limitations in disaster contexts. These results demonstrate the robustness and spatial transferability of the 3DMASC–Random Forest approach for disaster mapping. Integrating multispectral data, improving small-object representation, and incorporating automated debris volume estimation could further enhance classification reliability, enabling faster, more informed post-disaster decision-making. By enabling rapid, accurate damage mapping, this approach supports sustainable disaster recovery, resource-efficient debris management, and resilience planning in hurricane-prone regions.

Accurate building damage assessment using bi-temporal multi-modal remote sensing images is essential for effective disaster response and recovery planning. This study proposes a novel Building-Guided Pseudo-Label Learning Framework to address the challenges of mapping building damage from pre-disaster optical and post-disaster SAR images. First, we train a series of building extraction models using pre-disaster optical images and building labels. To enhance building segmentation, we employ multi-model fusion and test-time augmentation strategies to generate pseudo-probabilities, followed by a low-uncertainty pseudo-label training method for further refinement. Next, a change detection model is trained on bi-temporal cross-modal images and damaged building labels. To improve damage classification accuracy, we introduce a building-guided low-uncertainty pseudo-label refinement strategy, which leverages building priors from the previous step to guide pseudo-label generation for damaged buildings, reducing uncertainty and enhancing reliability. Experimental results on the 2025 IEEE GRSS Data Fusion Contest dataset demonstrate the effectiveness of our approach, which achieved the highest mIoU score (54.28%) and secured first place in the competition. The source code will be available at BGPLL.

Abstract. Accurately forecasting disaster impacts before they occur is crucial for effective emergency preparedness and response. This study presents a dual approach utilizing the Pix2Pix conditional Generative Adversarial Network (cGAN) to leverage pre-disaster satellite imagery for enhanced disaster risk management. Firstly, we employ Pix2Pix to predict post-disaster damage levels from pre-disaster satellite images. By training on the xBD dataset, the model learns to generate spatially distributed damage predictions, enabling proactive planning and resource allocation in high-risk areas. Secondly, Pix2Pix is used to generate synthetic post-disaster images from pre-disaster inputs, allowing for scenario visualization without reliance on actual post-disaster imagery. The model's performance is evaluated using accuracy, precision, recall, and F1-score for damage prediction, achieving an accuracy of 79% and an F1-score of 76%. For synthetic image generation, structural similarity index measure (SSIM) and peak signal-to-noise ratio (PSNR) are used, yielding average values of 0.57 and 23.5, respectively. These results indicate the potential of our framework in anticipating disaster damage and generating realistic post-disaster visualizations. The framework's performance depends on the quality and availability of pre-disaster satellite imagery, which may affect prediction reliability. Further evaluation across different disaster types, including earthquakes, and wildfires, is needed to assess robustness and generalizability. This study demonstrates the potential of generative AI-based approaches in enhancing disaster preparedness by providing both damage forecasting and post-disaster image generation. The proposed framework supports decision-makers in emergency response, urban resilience planning, and risk mitigation strategies, contributing to more effective disaster management.

Rapid assessment of damage and situational information is essential for effective restoration and recovery following a natural disaster. Addressing the current limitation with manual interpretation, significant efforts have been made to automate the process by leveraging satellite imagery with an unbiased overhead view of pre- and post-disaster scenarios. Existing approaches in segmenting and classifying damaged buildings often fall short in scenarios where damaged and undamaged buildings coexist within close proximity. In this work, we propose a two-stage disaster damage framework that localizes buildings through segmentation and then classifies the associated damage level at an object level. The framework employs a Siamese Vision Transformer (ViT-Small) backbone to create a rich feature representation that is informed by the entire scene. From this map, building-specific features are extracted, enabling robust identification of subtle and complex damage patterns often missed by pixelbased models. Experimental results demonstrate a localization F1 score of 0.8485 and a damage classification F1 score of 0.71, yielding an overall F1 score of 0.7507. These results surpass the baseline by a significant margin and outperform existing state-of-the-art solutions, while ensuring generalizability under high class imbalance and domain shift.

This research aims to map the level of potential and analyze and map the microzonation of building damage after an earthquake disaster in Nagari Kajai, West Pasaman Regency. The method used in mapping the potential level of earthquake hazard is scoring analysis, while in modeling the microzonation map of building damage is done by buffer method and matching the zone of danger level, then validating the model with field survey method which is integrated into the results of on-screen digitization of high-resolution satellite image interpretation of each distribution of damaged buildings at the research location. The results of this study are in the form of a map model of the potential level of earthquake disaster hazard which is divided into three zones, namely unstable areas with a score of 46-60, less stable areas have a score of 31-45, and stable areas against earthquakes have a score of 15-30. The results of the analysis of the microzoning map of post-earthquake building damage in Nagari Kajai, West Pasaman Regency obtained a microzoning area, in the unstable class covering 3583.63 ha, located in the jorong rimbo batu, kampung alang, pasa lamo area. In the less stable microzonation class covering an area of 6349.78 ha, located in the limpato jorong area, most of the lubuak sariak jorong area, and part of the tanjuang beruang jorong area, and in the stable class has an area of 1251.04 ha, located in the mudiak simpang jorong area, part of the timbo abu jorong, limpato, and some of the lubuak sariak jorong area.

Post-disaster change detection plays a crucial role in emergency response and urban damage assessment. This study introduces four CNN-based encoder-decoder models designed with distinct feature fusion strategies: early fusion, middle fusion, late fusion, and an improved middle fusion model integrated with a custom upsampling module. All models were trained and tested on the TUE-CD dataset, which includes pre- and post-disaster satellite images from the 2023 Türkiye earthquake. Among the models, the middle fusion approach achieved the best overall performance by combining intermediate-level features from bitemporal inputs. The enhanced model further improved segmentation accuracy by preserving spatial detail. Results indicate that the choice of fusion level significantly affects model performance and generalization. Middle fusion, in particular, offers a promising solution for reliable and accurate change detection in disaster scenarios. Future work may focus on integrating more efficient architectures and evaluating performance across different types of disasters.

Timely assessment of structural damage is critical for disaster response and recovery. However, most prior work in natural disaster analysis relies on 2D imagery, which lacks depth, suffers from occlusions, and provides limited spatial context. 3D semantic segmentation offers a richer alternative, but existing 3D benchmarks focus mainly on urban or indoor scenes, with little attention to disaster-affected areas. To address this gap, we present 3DAeroRelief--the first 3D benchmark dataset specifically designed for post-disaster assessment. Collected using low-cost unmanned aerial vehicles (UAVs) over hurricane-damaged regions, the dataset features dense 3D point clouds reconstructed via Structure-from-Motion and Multi-View Stereo techniques. Semantic annotations were produced through manual 2D labeling and projected into 3D space. Unlike existing datasets, 3DAeroRelief captures 3D large-scale outdoor environments with fine-grained structural damage in real-world disaster contexts. UAVs enable affordable, flexible, and safe data collection in hazardous areas, making them particularly well-suited for emergency scenarios. To demonstrate the utility of 3DAeroRelief, we evaluate several state-of-the-art 3D segmentation models on the dataset to highlight both the challenges and opportunities of 3D scene understanding in disaster response. Our dataset serves as a valuable resource for advancing robust 3D vision systems in real-world applications for post-disaster scenarios.

Natural disasters demand rapid damage assessment to guide humanitarian response. Here, we investigate whether medium-resolution Earth observation images from the Copernicus program can support building damage assessment, complementing very-high resolution imagery with often limited availability. We introduce xBD-S12, a dataset of 10,315 pre- and post-disaster image pairs from both Sentinel-1 and Sentinel-2, spatially and temporally aligned with the established xBD benchmark. In a series of experiments, we demonstrate that building damage can be detected and mapped rather well in many disaster scenarios, despite the moderate 10$\,$m ground sampling distance. We also find that, for damage mapping at that resolution, architectural sophistication does not seem to bring much advantage: more complex model architectures tend to struggle with generalization to unseen disasters, and geospatial foundation models bring little practical benefit. Our results suggest that Copernicus images are a viable data source for rapid, wide-area damage assessment and could play an important role alongside VHR imagery. We release the xBD-S12 dataset, code, and trained models to support further research.

Natural disasters usually affect vast areas and devastate infrastructures. Performing a timely and efficient response is crucial to minimize the impact on affected communities, and data-driven approaches are the best choice. Visual question answering (VQA) models help management teams to achieve in-depth understanding of damages. However, recently published models do not possess the ability to answer open-ended questions and only select the best answer among a predefined list of answers. If we want to ask questions with new additional possible answers that do not exist in the predefined list, the model needs to be fin-tuned/retrained on a new collected and annotated dataset, which is a time-consuming procedure. In recent years, large-scale Vision-Language Models (VLMs) have earned significant attention. These models are trained on extensive datasets and demonstrate strong performance on both unimodal and multi-modal vision/language downstream tasks, often without the need for fine-tuning. In this paper, we propose a VLM-based zero-shot VQA (ZeShot-VQA) method, and investigate the performance of on post-disaster FloodNet dataset. Since the proposed method takes advantage of zero-shot learning, it can be applied on new datasets without fine-tuning. In addition, ZeShot-VQA is able to process and generate answers that has been not seen during the training procedure, which demonstrates its flexibility.

No abstract available

Abstract. Post-disaster rapid response relies on the timely acquisition of information. Specifically, assessing structural integrity of building requires fast and reliable analysis, which can be supported by quantitative damage assessment based on the complete 3D geometry knowledge of structures. Such information can be derived either from quickly acquired imagery (e.g., video frames or panoramas) or directly from 3D data. Today, data collection is relatively straightforward thanks to UAV surveys and mobile mapping systems; however, extracting actionable information remains time-consuming when performed manually. This highlights the need for automated methods that can localize damage, identify critical issues, and support interpretation and decision-making. In recent years, artificial intelligence (AI) has attracted substantial attention in this domain, driven by the emergence of models that deliver fast and effective performance across a range of perception tasks. Yet the success of these approaches remains strongly conditioned by data availability and quality. In many application settings—especially those involving rare or highly specific post-disaster scenarios—representative training samples are scarce, underscoring the need for dedicated datasets construction, together with methods capable of learning from limited data. In this research, a dataset dedicated to the detection of cracks was collected. The aim is to have quick and straightforward identification of decay related to the structural stability of the building. Trained YOLOv11 object detection and segmentation model were tested on two case studies collected in Lebanon. These two case studies feature structural damage under severe external force, representative for the post-disaster scenarios analysis. The research evaluated the proposed hybrid approach (involving deep learning and machine learning methods and integrated with photogrammetric workflow), retrieving architectural severe damage location from 2D and localizing them effectively in 3D to assess global analysis, by comparing data with the ground truth in both 2D and 3D data.

Nature disasters play a key role in shaping human-urban infrastructure interactions. Effective and efficient response to natural disasters is essential for building resilience and a sustainable urban environment. Two types of information are usually the most necessary and difficult to gather in disaster response. The first information is about disaster damage perception, which shows how badly people think that urban infrastructure has been damaged. The second information is geolocation awareness, which means how people whereabouts are made available. In this paper, we proposed a novel disaster mapping framework, namely CVDisaster, aiming at simultaneously addressing geolocalization and damage perception estimation using cross-view Street-View Imagery (SVI) and Very High-Resolution satellite imagery. CVDisaster consists of two cross-view models, where CVDisaster-Geoloc refers to a cross-view geolocalization model based on a contrastive learning objective with a Siamese ConvNeXt image encoder, and CVDisaster-Est is a cross-view classification model based on a Couple Global Context Vision Transformer (CGCViT). Taking Hurricane IAN as a case study, we evaluate the CVDisaster framework by creating a novel cross-view dataset (CVIAN) and conducting extensive experiments. As a result, we show that CVDisaster can achieve highly competitive performance (over 80% for geolocalization and 75% for damage perception estimation) with even limited fine-tuning efforts, which largely motivates future cross-view models and applications within a broader GeoAI research community. The data and code are publicly available at: https://github.com/tum-bgd/CVDisaster.

Reliable assessment of building damage is essential for effective disaster management. Synthetic Aperture Radar (SAR) has become a valuable tool for damage detection, as it operates independently of the daylight and weather conditions. However, the limited availability of high-resolution pre-disaster SAR data remains a major obstacle to accurate damage evaluation, constraining the applicability of traditional change-detection approaches. This study proposes a comprehensive framework that leverages generated SAR data alongside optical imagery for building damage detection and further examines the influence of elevation data quality on SAR synthesis and model performance. The method integrates SAR image synthesis from a Digital Surface Model (DSM) and land cover inputs with a multimodal deep learning architecture capable of jointly localizing buildings and classifying damage levels. Two data modality scenarios are evaluated: a change-detection setting using pre-disaster authentic SAR and another using GAN-generated SAR, both combined with post-disaster SAR imagery for building damage assessment. Experimental results demonstrate that GAN-generated SAR can effectively substitute for authentic SAR in multimodal damage mapping. Models using generated pre-disaster SAR achieved comparable or superior performance to those using authentic SAR, with F1 scores of 0.730, 0.442, and 0.790 for the survived, moderate, and destroyed classes, respectively. Ablation studies further reveal that the model relies more heavily on land cover segmentation than on fine elevation details, suggesting that coarse-resolution DSMs (30 m) are sufficient as auxiliary input. Incorporating additional training regions further improved generalization and inter-class balance, confirming that high-quality generated SAR can serve as a viable alternative especially in the absence of authentic SAR, for scalable, post-disaster building damage assessment.

Cyber-physical disaster systems (CPDS) are a new cyber-physical application that collects physical realm measurements from IoT devices and sends them to the edge for damage severity analysis of impacted sites in the aftermath of a large-scale disaster. However, the lack of effective machine learning paradigms and the data and device heterogeneity of edge devices pose significant challenges in disaster damage assessment (DDA). To address these issues, we propose a generative adversarial network (GAN) and a lightweight, deep transfer learning-enabled, fine-tuned machine learning pipeline to reduce overall sensing error and improve the model’s performance. In this paper, we applied several combinations of GANs (i.e., DCGAN, DiscoGAN, ProGAN, and Cycle-GAN) to generate fake images of the disaster. After that, three pre-trained models: VGG19, ResNet18, and DenseNet121, with deep transfer learning, are applied to classify the images of the disaster. We observed that the ResNet18 is the most pertinent model to achieve a test accuracy of 88.81%. With the experiments on real-world DDA applications, we have visualized the damage severity of disaster-impacted sites using different types of Class Activation Mapping (CAM) techniques, namely Grad-CAM++, Guided Grad-Cam, & Score-CAM. Finally, using k-means clustering, we have obtained the scatter plots to measure the damage severity into no damage, mild damage, and severe damage categories in the generated heat maps.

Effective post-disaster damage assessment is crucial for guiding emergency response and resource allocation. This study introduces DamageScope, an integrated deep learning framework designed to detect and classify building damage levels from post-disaster satellite imagery. The proposed system leverages a convolutional neural network trained exclusively on post-event data to segment building footprints and assign them to one of four standardized damage categories: no damage, minor damage, major damage, and destroyed. The model achieves an average F1 score of 0.598 across all damage classes on the test dataset. To support geospatial analysis, the framework extracts the coordinates of damaged structures using embedded metadata, enabling rapid and precise mapping. These results are subsequently visualized through an interactive, web-based platform that facilitates spatial exploration of damage severity. By integrating classification, geolocation, and visualization, DamageScope provides a scalable and operationally relevant tool for disaster management agencies seeking to enhance situational awareness and expedite post-disaster decision making.

Rapid building damage assessments are vital for an effective earthquake response. In Japan, traditional Earthquake Damage Certification (EDC) surveys—followed by the issuance of Disaster Victim Certificates (DVCs)—are often inefficient. With advancements in remote sensing technologies and deep learning algorithms, their combined application has been explored for large-scale automated damage assessment. However, the scarcity of remote sensing data on damaged buildings poses significant challenges to this task. In this study, we propose an Uncertainty-Guided Fusion Module (UGFM) integrated into a standard decoder architecture, with a Pyramid Vision Transformer v2 (PVTv2) employed as the encoder. This module leverages uncertainty outputs at each stage to guide the feature fusion process, enhancing the model’s sensitivity to collapsed buildings and increasing its effectiveness under diverse conditions. A training and in-domain testing dataset was constructed using post-earthquake aerial imagery of the severely affected areas in Noto Prefecture. The model approximately achieved a recall of 79% with a precision of 68% for collapsed building extraction on this dataset. We further evaluated the model on an out-of-domain dataset comprising aerial images of Mashiki Town in Kumamoto Prefecture, where it achieved an approximate recall of 66% and a precision of 77%. In a quantitative analysis combining field survey data from Mashiki, the model attained an accuracy exceeding 87% in identifying major damaged buildings, demonstrating that the proposed method offers a reliable solution for initial assessment of major damage and its potential to accelerate DVC issuance in real-world disaster response scenarios.

Accurate detection and identification of destructed urban areas are essentially important for rescue planning after a large-scale natural disaster (e.g., earthquake, tsunami). Spaceborne synthetic aperture radar (SAR) can provide quick response and huge area monitoring to mitigate the further loss. Due to the layover, speckle effect, multiple scattering, etc., accurate mapping of urban damage conditions with SAR is even challenging. Fully polarimetric SAR (PolSAR) has the better potential to understand urban damage conditions from the viewpoint of scattering mechanism investigation. Polarimetric coherence is an important source for PolSAR data investigation. This paper focuses on the hidden signature exploration of co-polarization coherence in the rotation domain along the radar line of sight for urban damage investigation. A damage index is proposed based on the interpretation tool of co-polarization coherence pattern. It can successfully discriminate urban patches with various damage levels. Then, a damage level inversion relationship and an urban damage level mapping approach are established. The study case is the great 3.11 East Japan earthquake and tsunami using multitemporal ALOS/PALSAR PolSAR data. Experimental studies validate the efficiency of the proposed co-polarization coherence pattern technique and the mapping approach.

On February 6, 2023 (local time), two earthquakes (Mw7.8 and Mw7.7) struck central and southern Turkey, causing extensive damage to several cities and claiming a toll of 40,000 lives. In this study, we propose a method for seismic building damage assessment and analysis by combining SAR amplitude and phase coherence change detection. We determined building damage in five severely impacted urban areas and calculated the damage ratio by measuring the urban area and the damaged area. The largest damage ratio of 18.93% is observed in Nurdagi, and the smallest ratio of 7.59% is found in Islahiye. We verified the results by comparing them with high-resolution optical images and AI recognition results from the Microsoft team. We also used pixel offset tracking (POT) technology and D-InSAR technology to obtain surface deformation using Sentinel-1A images and analyzed the relationship between surface deformation and post-earthquake urban building damage. The results show that Nurdagi has the largest urban average surface deformation of 0.48 m and Antakya has the smallest deformation of 0.09 m. We found that buildings in the areas with steeper slopes or closer to earthquake faults have higher risk of collapse. We also discussed the influence of SAR image parameters on building change recognition. Image resolution and observation geometry have a great influence on the change detection results, and the resolution can be improved by various means to raise the recognition accuracy. Our research findings can guide earthquake disaster assessment and analysis and identify influential factors of earthquake damage.

Change detection is instrumental to localize damage and understand destruction in disaster informatics. While convolutional neural networks are at the core of recent change detection solutions, we present in this work, BLDNet, a novel graph formulation for building damage change detection and enable learning relationships and representations from both local patterns and non-stationary neighborhoods. More specifically, we use graph convolutional networks to efficiently learn these features in a semi-supervised framework with few annotated data. Additionally, BLDNet formulation allows for the injection of additional contextual building meta-features. We train and benchmark on the xBD dataset to validate the effectiveness of our approach. We also demonstrate on urban data from the 2020 Beirut Port Explosion that performance is improved by incorporating domain knowledge building meta-features.

Building change detection and building damage assessment are two essential tasks in post-disaster analysis. Building change detection focuses on identifying changed building areas between bi-temporal images, while building damage assessment involves segmenting all buildings and classifying their damage severity. These tasks play a critical role in disaster response and urban development monitoring. Although supervised learning has significantly advanced building change detection and damage assessment, its reliance on large labeled datasets remains a major limitation. In contrast, self-supervised learning enables the extraction of meaningful data representations without explicit training labels. To address this challenge, we propose a self-supervised learning approach that unifies denoising autoencoders and contrastive learning, enabling effective data representation for building change detection and damage assessment. The proposed architecture integrates a dual denoising autoencoder with a Vision Transformer backbone and contrastive learning strategy, complemented by a Feature Pyramid Network-ResNet dual decoder and an Edge Guidance Module. This design enhances multi-scale feature extraction and enables edge-aware segmentation for accurate predictions. Extensive experiments were conducted on five public datasets, including xBD, LEVIR, LEVIR+, SYSU, and WHU, to evaluate the performance and generalization capabilities of the model. The results demonstrate that the proposed Denoising AutoEncoder-enhanced Dual-Fusion Network (DAEDFN) approach achieves competitive performance compared with fully supervised methods. On the xBD dataset, the largest dataset for building damage assessment, our proposed method achieves an F1 score of 0.892 for building segmentation, outperforming state-of-the-art methods. For building damage severity classification, the model achieves an F1 score of 0.632. On the building change detection datasets, the proposed method achieves F1 scores of 0.837 (LEVIR), 0.817 (LEVIR+), 0.768 (SYSU), and 0.876 (WHU), demonstrating model generalization across diverse scenarios. Despite these promising results, challenges remain in complex urban environments, small-scale changes, and fine-grained boundary detection. These findings highlight the potential of self-supervised learning in building change detection and damage assessment tasks.

Earthquake is a very serious geological disaster. At present, the pre-earthquake prediction is not very accurate and effective in the world, but the quick, accurate and comprehensive understanding of the earthquake situation after the earthquake, and the arrangement of the implementation of rescue work, as far as possible to reduce the loss of recovery, has become the focus of the work of earthquake prevention and mitigation. The application of remote sensing technology in earthquake disaster monitoring and emergency rescue can greatly avoid the time-consuming and laborious shortcomings of the traditional ground survey method. Optical remote sensing has always been the mainstream means in the field of remote sensing. Collapsed buildings pose a direct threat to human life, so it is an urgent task to quickly extract the collapse of buildings in the earthquake. The object-oriented detection method of building damage change is to classify the pre-earthquake and post-earthquake images by object-oriented method, and then compare and analyze them on the basis of object to determine the change area. It can extract high-resolution buildings effectively and accurately. In the post-earthquake emergency rescue, the extraction of building damage information can quickly understand the traffic situation of each disaster area, and more effectively carry out targeted rescue operations.

A near-real-time change detection network can consistently identify unauthorized construction activities over a wide area, empowering authorities to enforce regulations efficiently. Furthermore, it can promptly assess building damage, enabling expedited rescue efforts. The extensive adoption of deep learning in change detection has prompted a predominant emphasis on enhancing detection performance, primarily through the expansion of the depth and width of networks, overlooking considerations regarding inference time and computational cost. To accurately represent the spatio-temporal semantic correlations between pre-change and post-change images, we create an innovative transformer attention mechanism named Sliding-Window Dissimilarity Cross-Attention (SWDCA), which detects spatio-temporal semantic discrepancies by explicitly modeling the dissimilarity of bi-temporal tokens, departing from the mono-temporal similarity attention typically used in conventional transformers. In order to fulfill the near-real-time requirement, SWDCA employs a sliding-window scheme to limit the range of the cross-attention mechanism within a predetermined window/dilated window size. This approach not only excludes distant and irrelevant information but also reduces computational cost. Furthermore, we develop a lightweight Siamese backbone for extracting building and environmental features. Subsequently, we integrate an SWDCA module into this backbone, forming an efficient change detection network. Quantitative evaluations and visual analyses of thorough experiments verify that our method achieves top-tier accuracy on two building change detection datasets of remote sensing imagery, while also achieving a real-time inference speed of 33.2 FPS on a mobile GPU.

In war and explosion scenarios, buildings often suffer varying degrees of damage characterized by complex, irregular, and fragmented spatial patterns, posing significant challenges for remote sensing–based change detection. Additionally, the scarcity of high-quality datasets limits the development and generalization of deep learning approaches. To overcome these issues, we propose CMSNet, an end-to-end framework that integrates the structural priors of the Segment Anything Model (SAM) with the efficient temporal modeling and fine-grained representation capabilities of CNN–Mamba. Specifically, CMSNet adopts CNN–Mamba as the backbone to extract multi-scale semantic features from bi-temporal images, while SAM-derived visual priors guide the network to focus on building boundaries and structural variations. A Pre-trained Visual Prior-Guided Feature Fusion Module (PVPF-FM) is introduced to align and fuse these priors with change features, enhancing robustness against local damage, non-rigid deformations, and complex background interference. Furthermore, we construct a new RWSBD (Real-world War Scene Building Damage) dataset based on Gaza war scenes, comprising 42,732 annotated building damage instances across diverse scales, offering a strong benchmark for real-world scenarios. Extensive experiments on RWSBD and three public datasets (CWBD, WHU-CD, and LEVIR-CD+) demonstrate that CMSNet consistently outperforms eight state-of-the-art methods in both quantitative metrics (F1, IoU, Precision, Recall) and qualitative evaluations, especially in fine-grained boundary preservation, small-scale change detection, and complex scene adaptability. Overall, this work introduces a novel detection framework that combines foundation model priors with efficient change modeling, along with a new large-scale war damage dataset, contributing valuable advances to both research and practical applications in remote sensing change detection. Additionally, the strong generalization ability and efficient architecture of CMSNet highlight its potential for scalable deployment and practical use in large-area post-disaster assessment.

No abstract available

No abstract available

The increasing frequency and intensity of natural disasters call for rapid and accurate damage assessment. In response, disaster benchmark datasets from high-resolution satellite imagery have been constructed to develop methods for detecting damaged areas. However, these methods face significant challenges when applied to previously unseen regions due to the limited geographical and disaster-type diversity in the existing datasets. We introduce DAVI (Disaster Assessment with VIsion foundation model), a novel approach that addresses domain disparities and detects structural damage at the building level without requiring ground-truth labels for target regions. DAVI combines task-specific knowledge from a model trained on source regions with task-agnostic knowledge from an image segmentation model to generate pseudo labels indicating potential damage in target regions. It then utilizes a two-stage refinement process, which operate at both pixel and image levels, to accurately identify changes in disaster-affected areas. Our evaluation, including a case study on the 2023 Türkiye earthquake, demonstrates that our model achieves exceptional performance across diverse terrains (e.g., North America, Asia, and the Middle East) and disaster types (e.g., wildfires, hurricanes, and tsunamis). This confirms its robustness in disaster assessment without dependence on ground-truth labels and highlights its practical applicability.

ABSTRACT Rapidly estimating post-disaster building damage via high-resolution remote sensing (HRRS) imagery is essential for initial disaster relief. However, the complex appearance of building damage poses challenges for existing methods. Specifically, relying solely on post-disaster images lacks building boundary guidance, while change detection methods using dual-temporal imageries are prone to introducing false changes. To address these issues, this paper presents a novel weakly supervised approach that leverages pre- and post-disaster HRRS images for building damage detection. The contributions of this paper are twofold. Firstly, a unique framework is proposed to utilize dual-temporal images. Precisely, the proposed method initially extracts fine-grained sub-building-level individuals from pre-disaster images by combining a fully convolutional neural network (FCN)-based method with superpixel segmentation. Then, these details serve as cues to effectively guide the detection of damaged building areas on post-disaster images, thereby enhancing accuracy. Secondly, we propose a weakly supervised method that solely relies on labeling building damage based on image patches but can ultimately yield pixel-level building damage results. Experiments conducted using HRRS images captured during the 2010 Haiti earthquake demonstrate that the proposed method outperforms existing methodologies. This effort of this paper will contribute to the sustainable development of cities and human settlements.

Building change detection (CD) from remote sensing images (RSI) has great significance in exploring the utilization of land resources and determining the building damage after a disaster. This article proposed an attention-based multiscale input–output network, named AMIO-Net, for building CD in high-resolution RSI. It is able to overcome partial drawbacks of existing CD methods, such as insufficient utilization of information (details of building edges) of original images and poor detection effect of small targets (small-scale buildings or small-area changed buildings that are disturbed by other buildings). In AMIO-Net, the input image is scaled down to different sizes, and performed the convolution to extract features. Then, the feature maps are fed into the encoding stage so that the network can fully utilize the feature information (FI) of the original image. More importantly, we design two attention mechanism modules: the pyramid pooling attention module (PPAM) and the Siamese attention mechanism module (SAMM). PPAM combines a pyramid pooling module and an attention mechanism to fully consider the global information and focus on the FI of changed pixels in the image. The input of SAMM is the parallel multiscale output diagram of the decoding portion and deep feature maps of the network so that AMIO-Net can utilize the global contextual semantic FI and strengthen detection ability for small targets. Experiments on three datasets show that the proposed method achieves higher detection accuracy and F1 score compared with the state-of-the-art methods.

Convolutional neural networks (CNNs) and Transformers have made impressive progress in the field of remote sensing change detection (CD). However, both architectures have inherent shortcomings: CNN is constrained by a limited receptive field that may hinder their ability to capture broader spatial contexts, while Transformers are computationally intensive, making them costly to train and deploy on large datasets. Recently, the Mamba architecture, based on state space models (SSMs), has shown remarkable performance in a series of natural language processing tasks, which can effectively compensate for the shortcomings of the above two architectures. In this article, we explore for the first time the potential of the Mamba architecture for remote sensing CD tasks. We tailor the corresponding frameworks, called MambaBCD, MambaSCD, and MambaBDA, for binary CD (BCD), semantic CD (SCD), and building damage assessment (BDA), respectively. All three frameworks adopt the cutting-edge Visual Mamba architecture as the encoder, which allows full learning of global spatial contextual information from the input images. For the change decoder, which is available in all three architectures, we propose three spatiotemporal relationship modeling mechanisms, which can be naturally combined with the Mamba architecture and fully utilize its attribute to achieve spatiotemporal interaction of multitemporal features, thereby obtaining accurate change information. On five benchmark datasets, our proposed frameworks outperform current CNN- and Transformer-based approaches without using any complex training strategies or tricks, fully demonstrating the potential of the Mamba architecture in CD tasks. Further experiments show that our architecture is quite robust to degraded data. The source code is available at: https://github.com/ChenHongruixuan/MambaCD.

Summary Accurate and effective identification, determination of the location, and classification of damaged buildings are essential after destructive earthquakes. However, the accuracy of image change detection is limited because of the many texture features and changes in non-building information. In this context, a model for single-building damage detection based on multi-feature fusion is proposed. First, the normalized Digital Surface Model (nDSM) was extracted from the DSM through iterative filtering and point cloud thinning, followed by the extraction of building contour information. Next, single-building images were generated from different data sources through the region of interest (ROI), and the optimal texture feature parameters were extracted for fusion. Afterward, principal-component analysis (PCA) was conducted to suppress multi-feature correlation-induced information redundancy. Finally, the damage to buildings was quantitatively evaluated, and the model was compared with 13 models. The results confirmed the practicability of the model for the Yangbi MS6.4 and Honghe MS5.0 earthquakes.

ABSTRACT Building detection and change detection using remote sensing images can help urban and rescue planning. Moreover, they can be used for building damage assessment after natural disasters. Currently, most of the existing models for building detection use only one image (pre-disaster image) to detect buildings. This is based on the idea that post-disaster images reduce the model’s performance because of the presence of destroyed buildings. In this paper, we propose a siamese model, called SiamixFormer, which uses pre- and post-disaster images as input. Our model has two encoders and has a hierarchical transformer architecture. The output of each stage in both encoders is given to a temporal transformer for feature fusion in a way that query is generated from pre-disaster images and (key, value) is generated from post-disaster images. To this end, temporal features are also considered in feature fusion. Another advantage of using temporal transformers in feature fusion is that they can better maintain large receptive fields generated by transformer encoders compared with CNNs. Finally, the output of the temporal transformer is given to a simple MLP decoder at each stage. The SiamixFormer model is evaluated on xBD, and WHU datasets, for building detection and on LEVIR-CD and CDD datasets for change detection and could outperform the state-of-the-art.

Earthquakes cause large-scale structural damage and loss of life, making rapid and accurate damage assessment critical in disaster management processes. In this context, remote sensing (RS) and deep learning-based change detection (CD) methods have the potential to provide decision-makers with effective information during post-disaster response efforts. This study presents a comparative analysis of segmentation-based CD models for detecting building collapses following the earthquakes that struck Turkey on February 6, 2023. The high-resolution TUE-CD dataset, consisting of pre- and post-disaster satellite images, was utilized. Three different deep learning models were designed for change detection: an adaptation of the DeepLabv3 architecture for CD tasks, a U-Net-based model, and a custom architecture enhanced with attention mechanisms and multi-scale feature fusion. These models were evaluated to compare the impact of different architectural choices and components on performance. Additionally, a comprehensive damage report was generated to support post-disaster building damage analysis with quantitative data and facilitate interpretation. The report includes calculations based on segmentation masks derived from model predictions.

No abstract available

In the aftermath of disasters, building damage maps are ob-tained using change detection to plan rescue operations. Cur-rent convolutional neural network approaches do not consider the similarities between neighboring buildings for predicting the damage. We present a novel graph-based building dam-age detection solution to capture these relationships. Our pro-posed model architecture learns from both local and neigh-borhood features to predict building damage. Specifically, we adopt the sample and aggregate graph convolution strategy to learn aggregation functions that generalize to unseen graphs which is essential for alleviating the time needed to obtain predictions for new disasters. Our experiments on the xBD dataset and comparisons with a classical convolutional neu-ral network reveal that while our approach is handicapped by class imbalance, it presents a promising and distinct advan-tage when it comes to cross-disaster generalization.

Aerial bombardment of the Gaza Strip beginning October 7, 2023 is one of the most intense bombing campaigns of the twenty-first century, driving widespread urban damage. Characterizing damage over a geographically dynamic and protracted armed conflict requires active monitoring. Synthetic aperture radar (SAR) has precedence for mapping disaster-induced damage with bi-temporal methods but applications to active monitoring during sustained crises are limited. Using interferometric SAR data from Sentinel-1, we apply a long temporal-arc coherent change detection (LT-CCD) approach to track weekly damage trends over the first year of the 2023- Israel-Hamas War. We detect 92.5% of damage labels in reference data from the United Nations with a negligible (1.2%) false positive rate. The temporal fidelity of our approach reveals rapidly increasing damage during the first three months of the war focused in northern Gaza, a notable pause in damage during a temporary ceasefire, and surges of new damage as conflict hot-spots shift from north to south. Three-fifths (191,263) of all buildings are damaged or destroyed by the end of the study. With massive need for timely data on damage in armed conflict zones, our low-cost and low-latency approach enables rapid uptake of damage information at humanitarian and journalistic organizations.

A prompt and accurate assessment of buildings' damage is critical for disaster management and emergency response. With the development of high-resolution synthetic aperture radar (SAR) and deep-learning methods, more efficient damage assessment techniques based on building-units are possible. This paper proposes a new building damage assessment method using high-resolution SAR images based on semantic change detection. It utilizes a Siamese-based module for damage change detection together with an attention mechanism-based module for semantic segmentation of the damage maps. To evaluate the proposed model, a new damage assessment dataset is constructed from the SAR imagery originated from the battle of Aleppo, Syria, for model training and testing. The experiments performed on this dataset show an overall accuracy of 88.3%. The proposed method effectively identifies the damaged areas of the buildings and grade the damage condition.

No abstract available

Natural disasters pose significant harm to society. As an important place for social activities and economic development, the degree of damage to building areas is directly related to disaster loss assessment and emergency rescue. Remote sensing image data, characterized by its wide coverage and multi-temporal features, provides important data support for post-disaster loss assessment. However, imaging differences caused by factors such as shooting time, imaging angle, and different sensors can interfere with the extraction of damage features and loss assessment. This paper proposes a Dual-Exchange-Attention U-Net (DERU-Net) model, which transforms the identification of building damage levels into intra-class semantic change detection. The DFMA feature attention fusion module is introduced to enhance the ability of dual-temporal feature extraction and achieve end-to-end assessment of building damage. The proposed method is comprehensively evaluated and tested on the xBD dataset. Experimental results show that compared with other methods, the DERU-Net proposed in this paper exhibits better stability and evaluation accuracy in assessing the degree of building damage.

Post-flood building damage assessment is critical for rapid response and post-disaster reconstruction planning. Current research fails to consider the distinct requirements of disaster assessment (DA) from change detection (CD) in neural network design. This paper focuses on two key differences: 1) building change features in DA satellite images are more subtle than in CD; 2) DA datasets face more severe data scarcity and label imbalance. To address these issues, in terms of model architecture, the research explores the benchmark performance of attention mechanisms in post-flood DA tasks and introduces Simple Prior Attention UNet (SPAUNet) to enhance the model's ability to recognize subtle changes, in terms of semi-supervised learning (SSL) strategies, the paper constructs four different combinations of image-level label category reference distributions for consistent training. Experimental results on flood events of xBD dataset show that SPAUNet performs exceptionally well in supervised learning experiments, achieving a recall of 79.10% and an F1 score of 71.32% for damaged classification, outperforming CD methods. The results indicate the necessity of DA task-oriented model design. SSL experiments demonstrate the positive impact of image-level consistency regularization on the model. Using pseudo-labels to form the reference distribution for consistency training yields the best results, proving the potential of using the category distribution of a large amount of unlabeled data for SSL. This paper clarifies the differences between DA and CD tasks. It preliminarily explores model design strategies utilizing prior attention mechanisms and image-level consistency regularization, establishing new post-flood DA task benchmark methods.

Accurate change detection is critical for applications like disaster assessment, building damage analysis, and information retrieval. However, existing methods predominantly rely on homologous remote sensing data (e.g., optical/SAR image pairs), limiting their ability to achieve timely and precise change extraction. To address this limitation, this paper introduces the SAE_UNet++, a change detection network specifically designed for heterologous remote sensing images. The architecture integrates: (1) heterologous image translation module and (2) a Siamese attention enhanced detection module. The heterologous image translation module utilizes a conditional GAN to align optical and SAR images in a unified feature space, minimizing information loss during translation. The Siamese attention enhanced change detection module processes dual-temporal images through weight-shared encoders for local feature extraction. Multi-scale convolutional block attention mechanisms (CBAM) augment the Siamese architecture to model global context information and two branch feature fusion. The decoder subsequently generates multi-level predictions for optimization. To evaluate SAE_UNet++'s efficacy in heterogeneous change detection, we evaluated it on two flood event datasets. The model achieved F1 scores of 95.91% (Gloucester I) and 77.99% (California), with visualization results showing closer alignment to ground truth than comparative methods, demonstrating state-of-the-art performance.

This paper presents damage assessment using a hierarchical transformer architecture (DAHiTrA), a novel deep‐learning model with hierarchical transformers to classify building damages based on satellite images in the aftermath of natural disasters. Satellite imagery provides real‐time and high‐coverage information and offers opportunities to inform large‐scale postdisaster building damage assessment, which is critical for rapid emergency response. In this work, a novel transformer‐based network is proposed for assessing building damage. This network leverages hierarchical spatial features of multiple resolutions and captures the temporal differences in the feature domain after applying a transformer encoder to the spatial features. The proposed network achieves state‐of‐the‐art performance when tested on a large‐scale disaster damage data set (xBD) for building localization and damage classification, as well as on LEVIR‐CD data set for change detection tasks. In addition, this work introduces a new high‐resolution satellite imagery data set, Ida‐BD (related to 2021 Hurricane Ida in Louisiana) for domain adaptation. Further, it demonstrates an approach of using this data set by adapting the model with limited fine‐tuning and hence applying the model to newly damaged areas with scarce data.

Indonesia suffers frequent natural disasters, including tsunamis, which cause significant damage to infrastructure. One major event is the 2004 Aceh tsunami, which claimed over 170,000 lives and destroyed many buildings. In this context, improving emergency response and enabling post-disaster recovery plans are crucial for accurate and quick categorization of building damage levels. This study employs Convolutional Neural Networks (CNN), a deep learning approach for image data processing, thereby allowing the identification of complex patterns like damage levels and building footprints. The study uses a local dataset including preand post-disaster tsunami images from Indonesia, split into “damaged” and “not damaged” categories. Built on the EfficientNet architecture—known for its efficiency in handling visual data—the CNN model was trained on 702 images. Subsequently, the model was tested for accuracy, precision, recall, and F1-score metrics. The results display an accuracy of 80.14%; precision of 73.32%; recall of 77.57%; and an F1-score of 74.78%. Although these findings are promising, the model's damage detection sensitivity still has room for improvement. This study offers a great contribution by using a local dataset, thus producing a model more relevant for the social and geographical settings of Indonesia. The deployment of CNN substantially reduces damage assessment time compared to hand-operated techniques, which then enables faster and more effective decision-making in rehabilitation projects. Additionally, these findings lead to opportunities for further development in technology-based disaster mitigation.

Abstract Automated detection of building destruction in conflict zones is crucial for human rights monitoring, humanitarian response, and academic research. However, existing approaches (i) rely on proprietary satellite imagery, both expensive and of limited availability at wartime, (ii) require labeled training data, usually not available in war-affected regions, or (iii) use optical imagery, regularly obstructed by cloud cover. This study addresses these challenges by introducing an unsupervised method to detect destruction at the building level using freely and globally available Sentinel-1 synthetic aperture radar images from the European Space Agency. By statistically assessing interferometric coherence changes over time, unlike existing approaches, our method enables the detection of destruction from a single satellite image, allowing for near real-time destruction assessments every 12 days. We provide a continuous, statistically grounded probability measure for the likelihood of destruction at both the building and pixel level, thereby quantifying the level of uncertainty of the detection. Using ground truth data and reported sequences of events, we validate our approach both quantitatively and qualitatively, across three case studies in Beirut, Mariupol, and Gaza, demonstrating its ability to accurately identify the spatial patterns and timing of destruction events. Using open-access data, our method offers a scalable, global, and cost-effective solution for monitoring building destruction in conflict zones.

No abstract available

Remote sensing is an effective method of evaluating building damage after a large-scale natural disaster, such as an earthquake or a typhoon. In recent years, with the development of computer vision technology, deep learning algorithms have been used for damage assessment from aerial images. In April 2016, a series of earthquakes hit the Kyushu region, Japan, and caused severe damage in the Kumamoto and Oita Prefectures. Numerous buildings collapsed because of the strong and continuous shaking. In this study, a deep learning model called Mask R-CNN was modified to extract residential buildings and estimate their damage levels from post-event aerial images. Our Mask R-CNN model employs an improved feature pyramid network and online hard example mining. Furthermore, a non-maximum suppression algorithm across multiple classes was also applied to improve prediction. The aerial images captured on 29 April 2016 (two weeks after the main shock) in Mashiki Town, Kumamoto Prefecture, were used as the training and test sets. Compared with the field survey results, our model achieved approximately 95% accuracy for building extraction and over 92% accuracy for the detection of severely damaged buildings. The overall classification accuracy for the four damage classes was approximately 88%, demonstrating acceptable performance.

Building damage maps can be generated from either optical or Light Detection and Ranging (Lidar) datasets. In the wake of a disaster such as an earthquake, a timely and detailed map is a critical reference for disaster teams in order to plan and perform rescue and evacuation missions. Recent studies have shown that, instead of being used individually, optical and Lidar data can potentially be fused to obtain greater detail. In this study, we explore this fusion potential, which incorporates deep learning. The overall framework involves a novel End-to-End convolutional neural network (CNN) that performs building damage detection. Specifically, our building damage detection network (BDD-Net) utilizes three deep feature streams (through a multi-scale residual depth-wise convolution block) that are fused at different levels of the network. This is unlike other fusion networks that only perform fusion at the first and the last levels. The performance of BDD-Net is evaluated under three different phases, using optical and Lidar datasets for the 2010 Haiti Earthquake. The three main phases are: (1) data preprocessing and building footprint extraction based on building vector maps, (2) sample data preparation and data augmentation, and (3) model optimization and building damage map generation. The results of building damage detection in two scenarios show that fusing the optical and Lidar datasets significantly improves building damage map generation, with an overall accuracy (OA) greater than 88%.

No abstract available

ABSTRACT Convolution in convolutional neural network(CNN) essentially uses a filter (kernel) with shared parameters to achieve feature extraction by computing the weighted sum of the centre pixel and adjacent pixels. The transformer divides the input image into patches and adds position encodings, then learns global semantic information and performs remote modelling through a self-attentive mechanism. However, CNNs are good at extracting local features but have difficulty in capturing global cues; the Transformer uses the self-attention mechanism for remote modelling. However, relative to CNN, local feature details are ignored to a certain extent. We believe that CNN and Transformer are complementary and will show better results if they are fused. Therefore, in this work, we propose a Hybrid Transformer-CNN Networks based on the fusion of CNN and Transformer branches for remote sensing change detection. In the CNN branch, we use the classical U-Net architecture to learn local semantic features. In the Transformer branch, we use Transformer-based progressive sampling to focus the model’s attention on objects of interest and prevent corrupting object structure. Subsequently, we propose an adaptive feature merging module to fully fuse the features of CNN and Transformer to enhance feature representation. At the same time, we introduce a differentiable superpixel branch to take advantage of the superpixel segmentation algorithm to accurately identify object boundaries, preserve boundary information and reduce noise in pixel-level features. We supplement the fused enhanced features into the superpixel branch features using a feature refinement module. After our experiments, we demonstrate the superiority of our model over other State of the art methods.

Building footprint (BFP) extraction focuses on the precise pixel-wise segmentation of buildings from aerial photographs such as satellite images. BFP extraction is an essential task in remote sensing and represents the foundation for many higher-level analysis tasks, such as disaster management, monitoring of city development, etc. Building footprint extraction is challenging because buildings can have different sizes, shapes, and appearances both in the same region and in different regions of the world. In addition, effects, such as occlusions, shadows, and bad lighting, have to also be considered and compensated. A rich body of work for BFP extraction has been presented in the literature, and promising research results have been reported on benchmarking datasets. Despite the comprehensive work performed, it is still unclear how robust and generalizable state-of-the-art methods are to different regions, cities, settlement structures, and densities. The purpose of this study is to close this gap by investigating questions on the practical applicability of BFP extraction. In particular, we evaluate the robustness and generalizability of state-of-the-art methods as well as their transfer learning capabilities. Therefore, we investigate in detail two of the most popular deep learning architectures for BFP extraction (i.e., SegNet, an encoder–decoder-based architecture and Mask R-CNN, an object detection architecture) and evaluate them with respect to different aspects on a proprietary high-resolution satellite image dataset as well as on publicly available datasets. Results show that both networks generalize well to new data, new cities, and across cities from different continents. They both benefit from increased training data, especially when this data is from the same distribution (data source) or of comparable resolution. Transfer learning from a data source with different recording parameters is not always beneficial.

This study proposes an automatic building footprint extraction framework that consists of a convolutional neural network (CNN)-based segmentation and an empirical polygon regularization that transforms segmentation maps into structured individual building polygons. The framework attempts to replace part of the manual delineation of building footprints that are involved in surveying and mapping field with algorithms. First, we develop a scale robust fully convolutional network (FCN) by introducing multiple scale aggregation of feature pyramids from convolutional layers. Two postprocessing strategies are introduced to refine the segmentation maps from the FCN. The refined segmentation maps are vectorized and polygonized. Then, we propose a polygon regularization algorithm consisting of a coarse and fine adjustment, to translate the initial polygons into structured footprints. Experiments on a large open building data set including 181 000 buildings showed that our algorithm reached a high automation level where at least 50% of individual buildings in the test area could be delineated to replace manual work. Experiments on different data sets demonstrated that our FCN-based segmentation method outperformed several most recent segmentation methods, and our polygon regularization algorithm is robust in challenging situations with different building styles, image resolutions, and even low-quality segmentation.

Recently, the combination of Transformers and convolutional neural networks (CNNs) has witnessed significant advancements in change detection (CD) tasks. However, it remains unexplored how to interactively integrate long-range dependency and local information to enhance the model’s global-local context awareness for effectively mitigating pseudo-changes. In addition, accurate identification and distinction of building changes from complex backgrounds still pose challenges due to the insufficient semantic context modeling across time between bi-temporal images. To address these issues, we propose a hybrid model ConvFormer-CD with parallel convolution and multihead self-attention (MSA). This combination enables better interaction of global and local information, thereby enhancing the adaptability to complex scenarios. Moreover, we introduce a novel module called Temporal Attention to establish cross-temporal semantic relationships between image pairs, effectively highlighting change regions by learning shared and nonshared semantics. This enables our model to accurately detect changed targets even in scenarios characterized by intricate geo-spatial arrangements and distributions. To further refine the differences in bi-temporal images, we propose a difference integration module (DIM) that connects the encoder and the decoder to fuse high-level semantic features across channels. We conduct extensive experiments on four benchmark datasets, including LEVIR-CD, LEVIR-CD+, WHU-CD, and S2Looking-CD, which demonstrates that the proposed ConvFormer-CD outperforms other state-of-the-art (SOTA) methods. Our codes will be available at https://github.com/taomi-lab/ConvFormer-CD.

3D building reconstruction from monocular remote sensing imagery is a promising and economical way to generate 3D city models at a large scale, yet the task is rarely touched. The paper tackles the problem via an end-to-end network. The goal is achieved by a modified network, named Mask-Height R-CNN, based on Mask R-CNN, with an additional height prediction head in the Region Proposal Network (RPN). Unlike most deep learning based methods, the height estimation is done on the instance level instead of pixel level, which does not require the assembly of the height maps and building masks. The proposed network gains good performances on ISPRS datasets, with 3D F1 scores of over 0.8.

In this paper, we consider building extraction from high spatial resolution remote sensing images. At present, most building extraction methods are based on artificial features. However, the diversity and complexity of buildings mean that building extraction methods still face great challenges, so methods based on deep learning have recently been proposed. In this paper, a building extraction framework based on a convolution neural network and edge detection algorithm is proposed. The method is called Mask R-CNN Fusion Sobel. Because of the outstanding achievement of Mask R-CNN in the field of image segmentation, this paper improves it and then applies it in remote sensing image building extraction. Our method consists of three parts. First, the convolutional neural network is used for rough location and pixel level classification, and the problem of false and missed extraction is solved by automatically discovering semantic features. Second, Sobel edge detection algorithm is used to segment building edges accurately so as to solve the problem of edge extraction and the integrity of the object of deep convolutional neural networks in semantic segmentation. Third, buildings are extracted by the fusion algorithm. We utilize the proposed framework to extract the building in high-resolution remote sensing images from Chinese satellite GF-2, and the experiments show that the average value of IOU (intersection over union) of the proposed method was 88.7% and the average value of Kappa was 87.8%, respectively. Therefore, our method can be applied to the recognition and segmentation of complex buildings and is superior to the classical method in accuracy.

Automated classification of building damage in remote sensing images enables the rapid and spatially extensive assessment of the impact of natural hazards, thus speeding up emergency response efforts. Convolutional neural networks (CNNs) can reach good performance on such a task in experimental settings. How CNNs perform when applied under operational emergency conditions, with unseen data and time constraints, is not well studied. This study focuses on the applicability of a CNN-based model in such scenarios. We performed experiments on 13 disasters that differ in natural hazard type, geographical location, and image parameters. The types of natural hazards were hurricanes, tornadoes, floods, tsunamis, and volcanic eruptions, which struck across North America, Central America, and Asia. We used 175,289 buildings from the xBD dataset, which contains human-annotated multiclass damage labels on high-resolution satellite imagery with red, green, and blue (RGB) bands. First, our experiments showed that the performance in terms of area under the curve does not correlate with the type of natural hazard, geographical region, and satellite parameters such as the off-nadir angle. Second, while performance differed highly between occurrences of disasters, our model still reached a high level of performance without using any labeled data of the test disaster during training. This provides the first evidence that such a model can be effectively applied under operational conditions, where labeled damage data of the disaster cannot be available timely and thus model (re-)training is not an option.

In this study, the Faster R-CNN model is trained to recognize the building based on the monitoring video image data captured by the high-resolution camera (high point) of the 40 meter communication tower. Aiming at the problem that the building target in the foreground is small and the detection effect is not good, the Faster R-CNN network structure is improved, and the low-level features and high-level features are used to detect the target in different scales. The experimental results show that the average accuracy of this method is 75.58%, which can effectively detect buildings. It can be applied to the law enforcement and inspection of illegal occupation of land, improve the work efficiency of government departments, and reduce the cost of manual inspection.

In this paper, they present a hybrid CNN-SVM model for the classification of building types based on image data. The model was trained and tested on a dataset of 10,340 images that were almost evenly divided into Apartment buildings (5,200 images) and Industrial buildings (5,140 images), with a support proportion of 0.50 for each in the case. The results were exceptional for both classes, with Apartment buildings reaching 99.03% Precision, 98.08% Recall, and 98.55% F1; while for Industrial buildings, it was Precision 98.07%, Recall 99.03%, and F1 98.55%. Cumulated accuracy was 0.99 in both classes, signifying very consistent and reliable performance. The models ended up with the same F1-Score and, thus, plainly exhibited that it has balanced models in terms of minimizing false and positive predictions. These two processes of deep learning feature extraction using Convolutional Neural Networks (CNN) and Support Vector Machines (SVM)—the latter being used for classification—imply very high-level representations and decision-making. Overall accuracy of 98.54% reinforces the model's promise and suitability for real-time implementation in automated building classification, urban planning, and infrastructure monitoring.

Numerous Heritage buildings are critical cultural assets susceptible to damage and deterioration. In this study, we explored machine learning techniques for the automated assessment of damage severity in heritage buildings. Specifically, this study has trained the Convolutional Neural Networks (CNN) and Support Vector Machine (SVM) on a dataset of 4,500 images of heritage buildings, each with a resolution of 224x224x3 pixels. The results demonstrate that both CNN and SVM can effectively classify damage severity levels in heritage buildings, with CNN showing slightly better overall performance. This study also found that both the models' performance improved as the damage severity remain increased. These findings suggest automated damage severity assessment using machine learning as a good heritage building management and preservation approach. Further research is needed to explore the feasibility of implementing this approach in real-time and to address potential ethical and technical challenges.

Urban areas are hotspots of complex and dynamic alterations of the Earth’s surface. Using deep learning (DL) techniques in remote sensing applications can significantly contribute to document these tremendous changes. Open source building data at a very high level of detail are still scarce or incomplete for many regions, therefore, hindering research and policy to properly provide knowledge on urban structures. In this study, we use a convolutional neural network to extract buildings for the city of Santiago de Chile. We deploy the recently released Mask R-CNN and use a pretrained model (PM) which already has been trained with remote sensing imagery. We fine-tune PM with very high-resolution (VHR) airborne RGB images from our study region and generate the fine-tuned model (FM). To extend the number of training data, we test several data augmentation methods for training purposes and evaluate their performance in context of urban environments. We achieve highest overall accuracy of 92 % by using augmentations and the generated FM. Our findings encourage to use DL methods in the urban context. The presented method can be adapted and applied to other global urban regions, and, help to overcome lacks in open source building data to assess urban environments.

No abstract available

In order to provide a timely response and recovery operations to a disaster event, executing precise building damage estimations is crucial. Traditional damage assessment techniques mainly depend on on-site inspections which are inefficient and subjective. This study explores the application of deep learning, and more specifically, Convolutional Neural Networks (CNNs), in the automatic classification of building damage from remote sensing images. The CNN model analyzes aerial or satellite images and assigns building damage levels as no damage, moderate, or severe. Unlike the customary methods, the approach proposed in the study has more rapid and precise evaluation of damages. This astonishing acceleration in the evaluation of damage can be very beneficial to management teams that handle the disaster. With accurate and timely damage evaluations performed by the system, decisions like when and how to allocate resources, when to plan for the response and recovery efforts become much easier and reliable. This study illustrates how disaster management operations can be improved with the implementation of models based on deep learning techniques and ensures timely recovery after an event. This method automates the assessment process which mitigates the need for manual evaluations, providing a comprehensive solution for assessing damages in areas affected by mass disasters.

No abstract available

Efficient and accurate post‐earthquake damage assessment of building structures is critical for ensuring the human safety and structural integrity of affected buildings. However, manual inspections and traditional visual damage identification methods are often constrained by the inaccessibility of hard‐to‐reach regions and the subjectivity of human inspectors. To overcome these issues, this paper proposes a novel approach for rapid and precise post‐earthquake damage identification and evaluation using unmanned aerial vehicle (UAV) and point cloud techniques, significantly reducing the time, labor, and errors associated with traditional methods. A comprehensive testing on full‐scale reinforced concrete shear walls was conducted to validate the precision and feasibility of this method. The point cloud models of the shear wall were generated leveraging UAV imagery and laser scanning technology with millimeter‐scale accuracy. The proposed algorithm effectively segmented each target plane of the shear wall, achieving a relatively satisfactory overall Intersection over Union of 99.25%. The relative errors of deformation between the algorithm's identification and gauges measurements were within 5%. This study successfully segmented and quantified structural surface damage, including cracks and spalling. Finally, the structural safety of the shear wall was evaluated according to ATC‐20 guidelines, using indicators such as inclination, story drift ratio, crack width, and damage area. Furthermore, proposed method was also verified in real cases.

Computer vision has shown potential for assisting post-earthquake inspection of buildings through automatic damage detection in images. However, assessing the safety of an earthquake-damaged building requires considering this damage in the context of its global impact on the structural system. Thus, an inspection must consider the expected damage progression of the associated component and the component’s contribution to structural system performance. To address this issue, a digital twin framework is proposed for post-earthquake building evaluation that integrates unmanned aerial vehicle (UAV) imagery, component identification, and damage evaluation using a Building Information Model (BIM) as a reference platform. The BIM guides selection of optimal sets of images for each building component. Then, if damage is identified, each image pixel is assigned to a specific BIM component, using a GrabCut-based segmentation method. In addition, 3D point cloud change detection is employed to identify nonstructural damage and associate that damage with specific BIM components. Two example applications are presented. The first develops a digital twin for an existing reinforced concrete moment frame building and demonstrates BIM-guided image selection and component identification. The second uses a synthetic graphics environment to demonstrate 3D point cloud change detection for identifying damaged nonstructural masonry walls. In both examples, observed damage is tied to BIM components, enabling damage to be considered in the context of each component’s known design and expected earthquake performance. The goal of this framework is to combine component-wise damage estimates with a pre-earthquake structural analysis of the building to predict a building’s post-earthquake safety based on an external UAV survey.

Global natural disasters are becoming more frequent and severe as a result of climate change. Recent advances in computer vision, particularly deep learning-based techniques and unmanned aerial vehicle (UAV) remote sensing, can aid disaster response teams in assessing the damage. Prior methods appear to be ineffective or were designed with inductive biases, making them difficult to conduct during the disaster damage assessment. In this paper, we investigate deep-learning-based methods capable of rapidly assessing building damage that follows natural disasters. Furthermore, we examine Deformable DETR, which is an improvement upon DETR, an object detection method based on the Transformer architecture, in terms of efficiency and convergence time, while inheriting DETR’s simple implementation and adaptable architecture, making it suitable for the task of damage detection. We also experimented and analyzed the performance of several optimizers to improve the performance of Deformable DETR.

Rapid damage assessment is essential for emergency action and recovery because earthquakes seriously harm urban infrastructure. A dataset for identifying collapsed structures was developed using Unmanned Aerial Vehicle (UAV) photos from atlas.gov.tr after the February 6th Kahramanmaras earthquake. With an emphasis on collapsed structures, YOLOv9, a sophisticated object detection model, was trained to categorize various building states. Across data subsets, cross-validation approaches were used to improve model performance and generalization. The model's efficacy for post-earthquake damage detection was demonstrated by its maximal overall precision of 0.892 and detection precision of 0.82 for collapsed structures. Rapid evaluation is made possible by combining deep learning with UAV imagery, which aids in disaster recovery, urban planning, and emergency response. By guaranteeing precise structural damage identification and supporting well-informed decision-making, this method improves catastrophe management. The findings demonstrate deep learning's potential for automating extensive damage assessment.

Quickly and accurately assessing the damage level of buildings is a challenging task for post-disaster emergency response. Most of the existing research mainly adopts semantic segmentation and object detection methods, which have yielded good results. However, for high-resolution Unmanned Aerial Vehicle (UAV) imagery, these methods may result in the problem of various damage categories within a building and fail to accurately extract building edges, thus hindering post-disaster rescue and fine-grained assessment. To address this issue, we proposed an improved instance segmentation model that enhances classification accuracy by incorporating a Mixed Local Channel Attention (MLCA) mechanism in the backbone and improving small object segmentation accuracy by refining the Neck part. The method was tested on the Yangbi earthquake UVA images. The experimental results indicated that the modified model outperformed the original model by 1.07% and 1.11% in the two mean Average Precision (mAP) evaluation metrics, mAPbbox50 and mAPseg50, respectively. Importantly, the classification accuracy of the intact category was improved by 2.73% and 2.73%, respectively, while the collapse category saw an improvement of 2.58% and 2.14%. In addition, the proposed method was also compared with state-of-the-art instance segmentation models, e.g., Mask-R-CNN and YOLO V9-Seg. The results demonstrated that the proposed model exhibits advantages in both accuracy and efficiency. Specifically, the efficiency of the proposed model is three times faster than other models with similar accuracy. The proposed method can provide a valuable solution for fine-grained building damage evaluation.

Year after year, floods become more and more a frequent and destructive force of nature, causing significant infrastructure losses and loss of life. An accurate and rapid assessment is required to determine the degree of contamination. The present study proposes a modern method for building damage assessment using deep learning during the flash flood of Derna, Libya. For this reason, we first exploited SAR satellite data, captured before and after the flood, to accurately determine the flood extent. Next, the footprint of affected buildings within this extent was extracted using a deep learning approach (U-Net model) based on high-resolution satellite imagery (30 cm) from MAXAR. Finally, an additional analysis was carried out using VIIRS VNP46A2 data (500 m spatial resolution) to analyse the night light assessment. The results demonstrate the effectiveness of this method, given that 5877 buildings were submerged by water and 2002 buildings were totally or partially destroyed. Also taking into account the estimated night light, Derna's power supply was reduced by over 90% after the floods. The suggested approach is an effective tool for comprehending the global effects of floods and aiding in relief efforts.

No abstract available

Integrated with remote sensing technology, deep learning has been increasingly used for rapid damage assessment. Despite reportedly having high accuracy, the approach requires numerous samples to maintain its performance. However, in the emergency response phase, training samples are often unavailable. Since no ground truth data is available, deep learning models cannot be trained for this specific situation and, thus, have to be applied to unseen data. Previous research has implemented transfer learning techniques to solve data unavailability. However, many studies do not accurately reflect the rapid damage mapping in real-world scenarios. This study illustrates the use of Earth observation and deep learning technologies in predicting damage in realistic emergency response settings. To this aim, we conducted extensive experiments using historical data to find the best model by examining multiple neural network models and loss functions. Then, we evaluated the performance of the best model for predicting building damage due to two different disasters, the 2011 Tohoku Tsunami and the 2023 Türkiye–Syria Earthquake, which were independent of the training samples. We found that a transformer-based model with a combined cross-entropy loss (CEL) and focal loss generates the highest scoring values. The testing on both unseen sites illustrates that the model can perform well in no-damage and destroyed classes. However, the scores dropped in the middle class. We also compared our transformer-based approach with other state-of-the-art models, specifically the xView-2 winning solution. The results show that the transformer-based models have stable generalization toward multiclass classification and multiresolution imagery.

The health condition of a building is related to the safety of people's lives and properties, so it is especially important to accurately detect its surface structure damage. Initial assessment of a building's health condition often involves identifying these cracks on its surface images. However, capturing these images traditionally relies on human labor, and the small pixel size of cracks in two-dimensional images makes segmentation challenging. In this study, a pioneering approach is introduced for automatic detection of building surface structure damage based on multi-UAV collaboration. The method comprises two primary components. Firstly, we introduce a reconstruction mathematical model based on MVS reconstruction rules to estimate the reconstruction quality. This model guides the extension of multi-UAV trajectories, considering trajectory energy consumption and security, enabling automated image capture and the creation of high-quality 3D models. Secondly, we design a spatial attention mechanism that incorporates Canny edge detection information and improve the deeplabV3 model to realize the automatic recognition of small cracks on building surfaces. Real-world experiments demonstrate that our method facilitates collaborative multi-UAV image capture, supporting high-quality 3D reconstruction and achieving precise crack segmentation.

No abstract available

In recent decades, millions of people are killed by natural disasters such as wildfire, landslide, tsunami, and volcanic eruption. The efficiency of post-disaster emergency responses and humanitarian assistance has become crucial in minimizing the expected casualties. This paper focuses on the task of building damage level evaluation, which is a key step for maximizing the deployment efficiency of post-event rescue activities. In this paper, we implement a Mask R-CNN based building damage evaluation model with a practical two-stage training strategy. The motivation of Stage-l is to train a ResNet 101 backbone in Mask R-CNN as a Building Feature Extractor. In Stage-2, we further build on top the model trained in Stage-l a deep learning architecture that performs more sophisticated tasks and is able to classify buildings with different damage levels from satellite images. In particular, in order to take advantage of pre-disaster satellite images, we extract the ResNet 101 backbone from the Mask R-CNN trained on pre-disaster images in Stage-l and utilize it to build a Siamese based semantic segmentation model for classifying the building damage level at the pixel level. The pre- and post-disaster satellite images are simultaneously fed into the proposed Siamese based model during the training and inference process. The output of these two models own the same size as input satellite images. Buildings with different damage levels, i.e., ‘no damage’, ‘minor damage’, ‘major damage’, and ‘destroyed’, are represented as segments of different damage classes in the output. Comparative experiments are conducted on the xBD satellite imagery dataset and compared with multiple state-of-the-art methods. The experimental results indicate that the proposed Siamese based method is capable to improve the damage evaluation accuracy by 16 times and 80%, compared with a baseline model implemented by xBD team and the Mask-RCNN framework, respectively.

No abstract available

Abstract. The Philippines is highly vulnerable to tropical cyclones (TCs), which frequently cause devastating impacts on infrastructure and communities. Rapid and accurate identification of damage extent and location is essential to trigger appropriate post-disaster response, expedite recovery, and facilitate better reconstruction. This study developed a methodology to classify building damage using unmanned aerial vehicle (UAV) images collected in February 2014 after Typhoon Haiyan hit the study area located in Tacloban City in November 2013. The methodology employed Structure-from-Motion (SfM) technique, texture analysis, and topographic modeling to analyze and extract building damage information. A damage rating system using percent area of damage as basis for damage severity was also developed. Correlation-based feature selection algorithm was used to refine and reduce the possible building damage predictor attributes. Random Forest was used to predict each building’s level of damage. Binary model R-A1, which classified completely damaged and undamaged buildings, had an accuracy of 93.5%, average precision of 0.938, average recall of 0.935 and average f-measure of 0.935, while model R-A2, which classified damaged and undamaged buildings, had an accuracy of 80.3%, average precision of 0.803, average recall of 0.803 and average f-measure of 0.803. Ternary model R-B1, which classified completely damaged, partially damaged and undamaged buildings, had an accuracy of 81.6%, average precision of 0.812, average recall of 0.816, and average f-measure of 0.813. The R-A2 model had an accuracy of 73.4% when tasked to classify 613 previously unseen damaged buildings. The R-B1 model had an accuracy of 70.3% when tasked to classify 575 previously unseen partially damaged and completely damaged buildings. This study highlighted the challenges in identifying and classifying building damage markers, some of which are unique to the Philippine setting, and demonstrated the value of UAV-based assessments for rapid and high-resolution damage evaluation.

The rapid and accurate assessment of building damage after earthquakes is critically essential for search-and-rescue and humanitarian-aid operations. This study proposes a comprehensive hybrid intelligent system to classify buildings into three categories—intact, damaged, and destroyed—using the UAV-TEBDE dataset, which comprises high-resolution Unmanned Aerial Vehicle (UAV) images collected after earthquakes in Türkiye. The proposed methodology is based on the fusion of deep features extracted from five different pretrained Convolutional Neural Network (CNN) models, including ResNet50, EfficientNetB4, VGG16, DenseNet121, and MobileNetV2, using a transfer learning approach. These enriched, high-dimensional combined feature vectors were systematically used to compare the performance of 12 machine learning classifiers, including ensemble learning methods, support vector machines, and discriminant analyses. The experimental results, validated through a robust 10-fold Stratified Group Cross-Validation, demonstrated that the proposed feature-level (early) fusion strategy achieved outstanding success. The Quadratic Discriminant Analysis (QDA) model exhibited the highest performance, attaining a mean Weighted F1 Score of 99.53% (±0.09%), surpassing more complex ensemble models and neural networks. The exceptionally low standard deviation observed across the validation folds confirmed that the superior performance of the QDA model was statistically robust and consistent. This study revealed that CNN-based feature fusion yields a highly distinctive feature space for post-disaster damage assessment, thereby enabling rapid near-perfect automatic damage mapping.

Emergency responders require accurate and comprehensive data to make informed decisions. Moreover, the data should be acquired and analyzed swiftly to ensure an efficient response. One of the tasks at hand post-disaster is damage assessment within the impacted areas. In particular, building damage should be assessed to account for possible casualties, and displaced populations, to estimate long-term shelter capacities, and to assess the damage to services that depend on essential infrastructure (e.g. hospitals, schools, etc.). Remote sensing techniques, including satellite imagery, can be used to gathering such information so that the overall damage can be assessed. However, specific points of interest among the damaged buildings need higher resolution images and detailed information to assess the damage situation. These areas can be further assessed through unmanned aerial vehicles and 3D model reconstruction. This paper presents a multi-UAV coverage path planning method for the 3D reconstruction of postdisaster damaged buildings. The methodology has been implemented in NetLogo3D, a multi-agent model environment, and tested in a virtual built environment in Unity3D. The proposed method generates camera location points surrounding targeted damaged buildings. These camera location points are filtered to avoid collision and then sorted using the K-means or the Fuzzy C-means methods. After clustering camera location points and allocating these to each UAV unit, a route optimization process is conducted as a multiple traveling salesman problem. Final corrections are made to paths to avoid obstacles and give a resulting path for each UAV that balances the flight distance and time. The paper presents the details of the model and methodologies, and an examination of the texture resolution obtained from the proposed method and the conventional overhead flight with the nadir-looking method used in 3D mappings. The algorithm outperforms the conventional method in terms of the quality of the generated 3D model.

It is of vital importance to locate the extreme earthquake disaster area timely and immediately after the earthquake occurs, which is greatly helpful for first responders to rescue trapped victims. The development of deep learning, computer vision technology as well as the widespread availability of inexpensive unmanned aerial vehicles (UAVs) has brought a new opportunity for building damage detection. We aim to propose a SSD model integrated with Convolutional Block Attention Mechanism (SSD_CBAM) and use it to detect damaged buildings caused by an earthquake. On the basis of the building damage dataset derived from the UAV images in Wenchuan Ms8.0 Earthquake, we compared our model with the classical SSD model. The result shows that detection accuracy is increased by 3.72% after the attention mechanism is integrated. It is believed that the result will be even better in the case of a lager training dataset.

We present ARIES a multi-agent single and multi-Unmanned Aerial Vehicles (UAV) approach for full building inspection following disasters in search and rescue operations (SAR). Our system addresses compute challenges and time constraints of autonomous inspections. ARIES employs a two-pronged strategy. We deploy a robust single UAV for exterior inspection, identifying potential failure points from structural damage (cracks and spalls). Identified points are priority areas for intelligent interior swarm exploration. These interior UAVs use thermal imaging to identify people in low-light. ARIES combines deep neural networks for vision tasks and heuristic algorithms for path planning tasks. It leverages You Only Look Once model (YOLO) variants YOLOv8-nano and YOLO12-nano for inference on edge devices and efficient lightweight path planning using Travelling Salesman Problem, Rapidly-Exploring Random Tree*, A*, and D* Lite. This work compares trade-offs for efficient computational offloading in a real-world application. We show that onboard computation on a Raspberry Pi can detect exterior structural damage in ~2s at 0.64 and 0.67 recall for cracks and spalls respectively. Additionally, at Ground Control Station, we achieve 0.85 recall for human detection during interior inspection.

The timely and efficient generation of detailed damage maps is of fundamental importance following disaster events to speed up first responders’ (FR) rescue activities and help trapped victims. Several works dealing with the automated detection of building damages have been published in the last decade. The increasingly widespread availability of inexpensive UAV platforms has also driven their recent adoption for rescue operations (i.e., search and rescue). Their deployment, however, remains largely limited to visual image inspection by skilled operators, limiting their applicability in time-constrained real conditions. This paper proposes a new solution to autonomously map building damages with a commercial UAV in near real-time. The solution integrates different components that allow the live streaming of the images on a laptop and their processing on the fly. Advanced photogrammetric techniques and deep learning algorithms are combined to deliver a true-orthophoto showing the position of building damages, which are already processed by the time the UAV returns to base. These algorithms have been customized to deliver fast results, fulfilling the near real-time requirements. The complete solution has been tested in different conditions, and received positive feedback by the FR involved in the EU funded project INACHUS. Two realistic pilot tests are described in the paper. The achieved results show the great potential of the presented approach, how close the proposed solution is to FR’ expectations, and where more work is still needed.

Rapid, accurate, and descriptive building damage assessment is critical for directing post-disaster resources, yet current automated methods typically provide only binary (damaged/undamaged) or ordinal severity scales. This paper introduces DamageCAT, a framework that advances damage assessment through typology-based categorical classifications. We contribute: (1) the BD-TypoSAT dataset containing satellite image triplets from Hurricane Ida with four damage categories - partial roof damage, total roof damage, partial structural collapse, and total structural collapse - and (2) a hierarchical U-Net-based transformer architecture for processing pre- and post-disaster image pairs. Our model achieves 0.737 IoU and 0.846 F1-score overall, with cross-event evaluation demonstrating transferability across Hurricane Harvey, Florence, and Michael data. While performance varies across damage categories due to class imbalance, the framework shows that typology-based classifications can provide more actionable damage assessments than traditional severity-based approaches, enabling targeted emergency response and resource allocation.

In response to the urgent need for rapid and precise post-disaster damage evaluation, this study introduces the Visual Prompt Damage Evaluation (ViPDE) framework, a novel contrastive learning-based approach that leverages the embedded knowledge within the Segment Anything Model (SAM) and pairs of remote sensing images to enhance building damage assessment. In this framework, we propose a learnable cascaded Visual Prompt Generator (VPG) that provides semantic visual prompts, guiding SAM to effectively analyze pre- and post-disaster image pairs and construct a nuanced representation of the affected areas at different stages. By keeping the foundation model’s parameters frozen, ViPDE significantly enhances training efficiency compared with traditional full-model fine-tuning methods. This parameter-efficient approach reduces computational costs and accelerates deployment in emergency scenarios. Moreover, our model demonstrates robustness across diverse disaster types and geographic locations. Beyond mere binary assessments, our model distinguishes damage levels with a finer granularity, categorizing them on a scale from 1 (no damage) to 4 (destroyed). Extensive experiments validate the effectiveness of ViPDE, showcasing its superior performance over existing methods. Comparative evaluations demonstrate that ViPDE achieves an F1 score of 0.7014. This foundation model-based approach sets a new benchmark in disaster management. It also pioneers a new practical architectural paradigm for foundation model-based contrastive learning focused on specific objects of interest.

Automated damage evaluation is of great importance in the maintenance and preservation of heritage structures. Damage investigation of large cultural buildings is time-consuming and labor-intensive, meaning that many buildings are not repaired in a timely manner. Additionally, some buildings in harsh environments are impossible to reach, increasing the difficulty of damage investigation. Oblique images facilitate damage detection in large buildings, yet quantitative damage information, such as area or volume, is difficult to generate. In this paper, we propose a method for quantitative damage evaluation of large heritage buildings in wild areas with repetitive structures based on drone images. Unlike existing methods that focus on building surfaces, we study the damage of building components and extract hidden linear symmetry information, which is useful for localizing missing parts in architectural restoration. First, we reconstruct a 3D mesh model based on the photogrammetric method using high-resolution oblique images captured by drone. Second, we extract 3D objects by applying advanced deep learning methods to the images and projecting the 2D object segmentation results to 3D mesh models. For accurate 2D object extraction, we propose an edge-enhanced method to improve the segmentation accuracy of object edges. 3D object fragments from multiple views are integrated to build complete individual objects according to the geometric features. Third, the damage condition of objects is estimated in 3D space by calculating the volume reduction. To obtain the damage condition of an entire building, we define the damage degree in three levels: no or slight damage, moderate damage and severe damage, and then collect statistics on the number of damaged objects at each level. Finally, through an analysis of the building structure, we extract the linear symmetry surface from the remaining damaged objects and use the symmetry surface to localize the positions of missing objects. This procedure was tested and validated in a case study (the Jiankou Great Wall in China). The experimental results show that in terms of segmentation accuracy, our method obtains results of 93.23% mAP and 84.21% mIoU on oblique images and 72.45% mIoU on the 3D mesh model. Moreover, the proposed method shows effectiveness in performing damage assessment of objects and missing part localization.

Accurate assessment of building damage is very important for disaster response and rescue. Traditional damage detection techniques using 2D features at a single observing angle cannot objectively and accurately reflect the structural damage conditions. With the development of unmanned aerial vehicle photogrammetric techniques and 3D point processing, automatic and accurate damage detection for building roof and facade has become a research hotspot in recent work. In this paper, we propose a building damage detection framework based on the boundary refined supervoxel segmentation and random forest–latent Dirichlet allocation classification. First, the traditional supervoxel segmentation method is improved to segment the point clouds into good boundary refined supervoxels. Then, non-building points such as ground and vegetation are removed from the generated supervoxels. Next, latent Dirichlet allocation (LDA) model is used to construct the high-level feature representation for each building supervoxel based on the selected 2D image and 3D point features. Finally, LDA model and random forest algorithm are employed to identify the damaged building regions. This method is applied to oblique photogrammetric point clouds collected from the Beichuan Country Earthquake Site. The research achieves the 3D damage assessment for building facade and roof. The result demonstrates that the proposed framework is capable of achieving around 94% accuracy for building point extraction and around 90% accuracy for damage identification. Moreover, both of the precision and recall for building damage detection reached around 89%. Concluded from comparison analysis, the proposed method improved the damage detection accuracy and the highest improvement ratio is over 8%.

Interferometric Synthetic Aperture Radar (InSAR) technology uses satellite radar to detect surface deformation patterns and monitor earthquake impacts on buildings. While vital for emergency response planning, extracting multi-class building damage classifications from InSAR data faces challenges: overlapping damage signatures with environmental noise, computational complexity in multi-class scenarios, and the need for rapid regional-scale processing. Our novel multi-class variational causal Bayesian inference framework with quadratic variational bounds provides rigorous approximations while ensuring efficiency. By integrating InSAR observations with USGS ground failure models and building fragility functions, our approach separates building damage signals while maintaining computational efficiency through strategic pruning. Evaluation across five major earthquakes (Haiti 2021, Puerto Rico 2020, Zagreb 2020, Italy 2016, Ridgecrest 2019) shows improved damage classification accuracy (AUC: 0.94-0.96), achieving up to 35.7% improvement over existing methods. Our approach maintains high accuracy (AUC>0.93) across all damage categories while reducing computational overhead by over 40% without requiring extensive ground truth data.

The statistics of damaged buildings after natural disasters are crucial for rescue operations, especially for damaged buildings that are extremely challenging for object detection. There are unique spatial distribution problems in the existing damaged building datasets, and different categories of building targets show obvious imbalance, especially when the proportion of damaged buildings is less than 0.15%. To address the issues of extreme class distribution imbalance and spatial distribution uniqueness, this article proposes a new data augmentation method called the geospatial enhancement sampling (GES) algorithm. The GES algorithm performs precise data enhancement work by positioning the spatial information of the data. To enhance the robustness of the dataset for object-level detection tasks, the xFBD dataset is reconstructed into the xFBDTWC dataset in this article. The xFBDTWC dataset, featuring balanced samples and cloud occlusion, has demonstrated its excellence through experimental results. The experimental research on the proposed algorithm is conducted using the mainstream object detection models. The experimental results show that, at the object level, the detection accuracy of severely damaged buildings is 0.56, and the detection accuracy of damaged buildings is 0.65. Compared with the original detection accuracy, this method improves it by 39% and 22%, respectively. The outstanding experimental results demonstrate the effectiveness of the GES algorithm, which is crucial for the accuracy and reliability of postdisaster assessments, thereby promoting more efficient and effective disaster response and resource allocation.

Previous applications of machine learning in remote sensing for the identification of damaged buildings in the aftermath of a large-scale disaster have been successful. However, standard methods do not consider the complexity and costs of compiling a training data set after a large-scale disaster. In this article, we study disaster events in which the intensity can be modeled via numerical simulation and/or instrumentation. For such cases, two fully automatic procedures for the detection of severely damaged buildings are introduced. The fundamental assumption is that samples that are located in areas with low disaster intensity mainly represent nondamaged buildings. Furthermore, areas with moderate to strong disaster intensities likely contain damaged and nondamaged buildings. Under this assumption, a procedure that is based on the automatic selection of training samples for learning and calibrating the standard support vector machine classifier is utilized. The second procedure is based on the use of two regularization parameters to define the support vectors. These frameworks avoid the collection of labeled building samples via field surveys and/or visual inspection of optical images, which requires a significant amount of time. The performance of the proposed method is evaluated via application to three real cases: the 2011 Tohoku-Oki earthquake–tsunami, the 2016 Kumamoto earthquake, and the 2018 Okayama floods. The resulted accuracy ranges between 0.85 and 0.89, and thus, it shows that the result can be used for the rapid allocation of affected buildings.

The ability to evaluate the damage to buildings both accurately and precisely is essential for disaster recovery, planning, and rescue services. This paper proposes a new approach based on integrating machine learning algorithms in building damage classification. To achieve higher precision in classifying the level of building damage, this research proposes a new technique that employs machine learning strategies. The researchers were able to train the model to be able to differentiate the different levels of building damage and the feature extraction was performed through machine learning. The model effectively extracts and learns multiple complex signals which represent different degrees of damage from a well picked database which include several degrees of damage. In a single pass, the Siamese U-Net can perform feature extraction and similarity measurement between two different images. The efficiency and effectiveness of the Siamese U-net model can be increased by reducing inference time, thus increasing its ability to deliver faster predictions while also improving its accuracy. The suggested Enhanced U-Net (EU-Net) could greatly increase the accuracy of building-level classification. As it turned out, the results are very promising and reach beyond traditional approaches with bringing more sample opportunities of machine learning integration in the building damage assessment context. Additionally, this study believes that the accuracy of building damage classification can be further enhanced demonstrating the usefulness of machine learning in disaster management.

No abstract available

Large-scale optical sensing and precise, rapid assessment of seismic building damage in urban communities are increasingly demanded in disaster prevention and reduction. The common method is to train a convolutional neural network (CNN) in a pixel-level semantic segmentation approach and does not fully consider the characteristics of the assessment objectives. This study developed a machine-learning-derived two-stage method for post-earthquake building location and damage assessment considering the data characteristics of satellite remote sensing (SRS) optical images with dense distribution, small size, and imbalanced numbers. It included a modified You Only Look Once (YOLOv4) object detection module and a support vector machine (SVM) based classification module. In the primary step, the multiscale features were successfully extracted and fused from SRS images of densely distributed buildings by optimizing the YOLOv4 model toward the network structures, training hyperparameters, and anchor boxes. The fusion improved multi-channel features, optimization of network structure and hyperparameters have significantly enhanced the average location accuracy of post-earthquake buildings. Thereafter, three statistics (i.e., the angular second moment, dissimilarity, and inverse difference moment) were further discovered to effectively extract the characteristic value for earthquake damage from located buildings in SRS optical images based on the gray level co-occurrence matrix. They were used as the texture features to distinguish damage intensities of buildings, using the SVM model. The investigated dataset included 386 pre- and post-earthquake SRS optical images of the 2017 Mexico City earthquake, with a resolution of 1024 × 1024 pixels. Results show that the average location accuracy of post-earthquake buildings exceeds 95.7% and that the binary classification accuracy for damage assessment reaches 97.1%. The proposed two-stage method was validated by its extremely high precision in respect of densely distributed small buildings, indicating the promising potential of computer vision in large-scale disaster prevention and reduction using SRS datasets.

Flood damage data are needed for various applications. Structural damage of buildings can reflect not only the economic damage but also the life‐threatening condition of a building, which provide crucial information for disaster response and recovery. Since traditional on‐site data collection shortly after a disaster is challenging, remote sensing data can be of great help, cover a wider area and be deployed earlier in time than on‐site surveys. However, this has its challenges and limitations. We elucidate on that by presenting two case studies from flash floods in Germany. First, we assessed the reliability of an existing flood damage schema, which differentiates from minor (structural) damage to complete building collapse. We compared two on‐site raters of the 2016 Braunsbach flood, reaching an excellent level of reliability. Second, we mapped structural building damage after the flood in the Ahr valley in 2021 using a textured 3D mesh and orthophotos. Here, we evaluated the remote sense‐based damage mapping done by three raters. Although the heterogeneity of ratings using remote sensing data is larger than among on‐site ratings, we consider it fit‐for‐purpose when compared with on‐site mapping, especially for event documentation and as basis for financial damage estimation and less complex numerical modelling.

No abstract available

Building damage is the main cause of casualties and economic losses from earthquakes. Therefore, understanding building damage is critical for emergency handling. Current information acquisition methods for assessing earthquake damage using unmanned aerial vehicle (UAV) remote sensing systems offer great flexibility and high efficiency with the capability to obtain high-resolution images, which can reflect actual damage to affected areas intuitively. Consequently, UAV remote sensing has become a convenient and important means to acquire earthquake-induced building damage information. Although manual visual interpretation can achieve high recognition accuracy, it is extremely time-consuming. In contrast, although current automatic recognition methods require less time, they have relatively poor recognition accuracy. Neither approach can satisfactorily simultaneously meet efficiency and accuracy requirements for earthquake emergency handling. This article applies image classification algorithms based on the support vector machine (SVM) to earthquake-induced building damage recognition, and proposes a recognition method based on scale-invariant feature transform (SIFT) characteristics and SVM classification. We use the magnitude 6.4 earthquake at Yangbi (2021) as an example to validate the proposed method. Results verify that the proposed method can recognize earthquake damage quickly and accurately, providing effective support for decision-making regarding rescue actions.

In the aftermath of large-scale natural disasters, the accuracy of building damage detection (BDD) is of critical importance. Post-disaster high-resolution (post-HR) remote sensing imagery is fundamental for BDD; however, prompt acquisition of such imagery remains a significant challenge. To address this issue, we introduce a novel plug-and-play feature fusion (FF) module. This module, strategically situated between a pre-trained super-resolution (SR) model and a BDD model, ingeniously combines features from both pre-disaster high-resolution (pre-HR) and super-resolved post-disaster remote sensing imagery. The proposed approach is designed to maximize the utilization of pre-HR images, thereby enhancing BDD accuracy. Experimental validation confirms that this improvement in accuracy is attributable to the pragmatic extraction of features from pre-HR imagery, not just an increase in model complexity. Consequently, our approach holds substantial promise for real-world post-disaster scenarios and lays a solid foundation for future BDD research, demonstrating potential improvements in both efficacy and practicality.

Earthquakes and other disasters often cause substantial damage to health facilities, impacting short-term response capacity and long-term health system needs. Identifying health facility damage following disasters is therefore crucial for coordinating response, but ground-based evaluations require substantial time and labor. Artificial intelligence (AI) models trained on satellite imagery can estimate building damage and could be used to generate rapid health facility damage reports. There is little published about methods of generating these estimates, testing real-world accuracy, or exploring error. This study presents a novel method of overlaying model damage outputs with health facility location data to generate health facility damage estimates following the February 2023 earthquake in Turkey. Two models were compared for agreement, accuracy, and errors. Building-level damage estimates were obtained for Model A (Microsoft neural network model), and Model B (Google AI model), and overlaid with health facility location data to identify facilities with significant damage. Model agreement, sensitivity and specificity for damage detection were calculated. A descriptive review of common error sources based on selected satellite imagery was conducted. A spatially aggregated damage estimation, based on proportion of buildings damaged in a 0.125km2 area, was also generated and assessed for each model. Twenty-five hospitals, 13 dialysis facilities, and 454 pharmacies were evaluated across three cities. Estimated damage was higher for Model A (10.4%) than Model B (4.3%). Cohen’s kappa was 0.32, indicating fair agreement. Sensitivity was low for both models at 42.9%, while specificity was high (A:93.6%, B:96.8%). Agreement and sensitivity were best for hospitals. Common errors included building identification and underestimation of damage for destroyed buildings. Spatially aggregated damage estimates yielded higher sensitivity (A:71.4%, B:57.1%) and agreement (Cohen’s kappa 0.38). Leveraging remote-sensing models for health facility damage assessment is feasible but currently lacks the sensitivity to replace ground evaluations. Improving building identification, damage detection for destroyed buildings, and spatially aggregating results may improve the performance and utility of these models for use in disaster response settings.