花粉识别技术：利用图像识别技术实现花粉检测的技术与理论研究

Pollen recognition has been shown to be important for a number of areas ranging from criminal investigations to paleoclimate studies. However, these palynology studies rely on highly qualified professionals to analyze pollen grains, which have become scarce and costly. Therefore, the automation of this task using computational methods is promising. Deep learning has proven to be the ultimate technique in computer vision tasks, but is very difficult to build a pollen data set with size enough to train such networks from scratch. This study investigated the use of transfer learning from pre-trained deep neural networks for pollen classification and compared their results with training from scratch and with promising predesigned features. Additionally, we introduced the biggest data set of pollen to the date. Experimental results achieved up to 96.24% of classification accuracy, suggesting that the fine-tuned deep learning architectures can be successfully applied to pollen classification.

The risk of pollen-induced allergies can be determined and predicted based on data derived from pollen monitoring. Hirst-type samplers are sensors that allow airborne pollen grains to be detected and their number to be determined. Airborne pollen grains are deposited on adhesive-coated tape, and slides are then prepared, which require further analysis by specialized personnel. Deep learning can be used to recognize pollen taxa based on microscopic images. This paper presents a method for recognizing a taxon based on microscopic images of pollen grains, allowing the pollen monitoring process to be automated. In this research, a deep CNN (convolutional neural network) model was built from scratch. Publicly available deep neural network models, pre-trained on image data (not including microscopic pictures), were also used. The results show that even a simple deep learning model produces quite good results when the classification of pollen grain taxa is performed directly from the images. The best deep learning model achieved 97.88% accuracy in the difficult task of recognizing three types of pollen grains (birch, alder, and hazel) with similar structures. The derived models can be used to build a system to support pollen monitoring experts in their work.

… Deep learning has been shown to successfully solve challenging classification tasks [15]. In … present a pollen-classification method that uses deep learning to classify 30 pollen types of …

Pollen grains, the male gametophytes for reproduction in higher plants, are vulnerable to various stresses that lead to loss of viability and eventually crop yield. A conventional method for assessing pollen viability is manual counting after staining, which is laborious and hinders high-throughput screening. We developed an automatic detection tool (PollenDetect) to distinguish viable and nonviable pollen based on the YOLOv5 neural network, which is adjusted to adapt to the small target detection task. Compared with manual work, PollenDetect significantly reduced detection time (from approximately 3 min to 1 s for each image). Meanwhile, PollenDetect can maintain high detection accuracy. When PollenDetect was tested on cotton pollen viability, 99% accuracy was achieved. Furthermore, the results obtained using PollenDetect show that high temperature weakened cotton pollen viability, which is highly similar to the pollen viability results obtained using 2,3,5-triphenyltetrazolium formazan quantification. PollenDetect is an open-source software that can be further trained to count different types of pollen for research purposes. Thus, PollenDetect is a rapid and accurate system for recognizing pollen viability status, and is important for screening stress-resistant crop varieties for the identification of pollen viability and stress resistance genes during genetic breeding research.

Pollen grains play a critical role in environmental, agricultural, and allergy research despite their tiny dimensions. The accurate classification of pollen grains remains a significant challenge, mainly attributable to their intricate structures and the extensive diversity of species. Traditional methods often lack accuracy and effectiveness, prompting the need for advanced solutions. This study introduces a novel deep learning framework, PollenNet, designed to tackle the intricate challenge of pollen grain image classification. The efficiency of PollenNet is thoroughly evaluated through stratified 5-fold cross-validation, comparing it with cutting-edge methods to demonstrate its superior performance. A comprehensive data preparation phase is conducted, including removing duplicates and low-quality images, applying Non-local Means Denoising for noise reduction, and Gamma correction to adjust image brightness. Furthermore, Explainable AI (XAI) is utilized to enhance the interpretability of the model, while Receiver Operating Characteristic (ROC) curve analysis serves as a quantitative method for evaluating the model's capabilities. PollenNet demonstrates superior performance when compared to existing models, with an accuracy of 98.45 %, precision of 98.20 %, specificity of 98.40 %, recall of 98.30 %, and f1-score of 98.25 %. The model also maintains low Mean Squared Error (0.03) and Mean Absolute Error (0.02) rates. The ROC curve analysis, the low False Positive Rate (0.016), and the False Negative Rate (0.017) highlight the reliability and dependability of the model. This study significantly improves the efficacy of classifying pollen grains, indicating an important advancement in the application of deep learning for ecological research.

Pollen-induced allergies affect a significant part of the population in developed countries. Current palynological analysis in Europe is a slow and laborious process which provides pollen information in a weekly-cycle basis. In this paper, we describe a system that allows to locate and classify, in a single step, the pollen grains present in standard glass microscope slides. Besides, processing the samples in the z-axis allows us to increase the probability of detecting grains compared to solutions based on one image per sample. Our system has been trained to recognise 11 pollen types, achieving 97.6 % success rate locating grains, of which 96.3 % are also correctly identified (0.956 macro–F1 score), and with a 2.4 % grains lost. Our results indicate that deep learning provides a robust framework to address automated identification of various pollen types, facilitating their daily measurement.

In palynology, the visual classification of pollen grains from different species is a hard task which is usually tackled by human operators using microscopes. Its complete automatization would save a high quantity of resources and provide valuable improvements especially for allergy-related information systems, but also for other application fields as paleoclimate reconstruction, quality control of honey based products, collection of evidences in criminal investigations or fabric dating and tracking. This paper presents three state-of-the-art deep learning classification methods applied to the recently published POLEN23E image dataset. The three methods make use of convolutional neural networks: the first one is strictly based on the idea of transfer learning, the second one is based on feature extraction and the third one represents a hybrid approach, combining transfer learning and feature extraction. The results from the three methods are indeed very good, reaching over 97% correct classification rates in images not previously seen by the models, where other authors reported around 70.

Pollen analysis is an important tool in many fields, including pollination ecology, paleoclimatology, paleoecology, honey quality control, and even medicine and forensics. However, labour‐intensive manual pollen analysis often constrains the number of samples processed or the number of pollen analysed per sample. Thus, there is a desire to develop reliable, high‐throughput, automated systems. We present an automated method for pollen analysis, based on deep learning convolutional neural networks (CNN). We scanned microscope slides with fuchsine stained, fresh pollen and automatically extracted images of all individual pollen grains. CNN models were trained on reference samples (122,000 pollen grains, from 347 flowers of 83 species of 17 families). The models were used to classify images of different pollen grains in a series of experiments. We also propose an adjustment to reduce overestimation of sample diversity in cases where samples are likely to contain few species. Accuracy of a model for 83 species was 0.98 when all samples of each species were first pooled, and then split into a training and a validation set (splitting experiment). However, accuracy was much lower (0.41) when individual reference samples from different flowers were kept separate, and one such sample was used for validation of models trained on remaining samples of the species (leave‐one‐out experiment). We therefore combined species into 28 pollen types where a new leave‐one‐out experiment revealed an overall accuracy of 0.68, and recall rates >0.90 in most pollen types. When validating against 63,650 manually identified pollen grains from 370 bumblebee samples, we obtained an accuracy of 0.79, but our adjustment procedure increased this to 0.85. Validation through splitting experiments may overestimate robustness of CNN pollen analysis in new contexts (samples). Nevertheless, our method has the potential to allow large quantities of real pollen data to be analysed with reasonable accuracy. Although compiling pollen reference libraries is time‐consuming, this is simplified by our method, and can lead to widely accessible and shareable resources for pollen analysis.

Monitoring of airborne pollen concentrations provides an important source of information for the globally increasing number of hay fever patients. Airborne pollen is traditionally counted under the microscope, but with the latest developments in image recognition methods, automating this process has become feasible. A challenge that persists, however, is that many pollen grains cannot be distinguished beyond the genus or family level using a microscope. Here, we assess the use of Convolutional Neural Networks (CNNs) to increase taxonomic accuracy for airborne pollen. As a case study we use the nettle family (Urticaceae), which contains two main genera (Urtica and Parietaria) common in European landscapes which pollen cannot be separated by trained specialists. While pollen from Urtica species has very low allergenic relevance, pollen from several species of Parietaria is severely allergenic. We collect pollen from both fresh as well as from herbarium specimens and use these without the often used acetolysis step to train the CNN model. The models show that unacetolyzed Urticaceae pollen grains can be distinguished with > 98% accuracy. We then apply our model on before unseen Urticaceae pollen collected from aerobiological samples and show that the genera can be confidently distinguished, despite the more challenging input images that are often overlain by debris. Our method can also be applied to other pollen families in the future and will thus help to make allergenic pollen monitoring more specific.

Pollen classification is considered an important task in palynology. In the Netherlands, two genera of the Urticaceae family, named Parietaria and Urtica, have high morphological similarities but induce allergy at a very different level. Therefore, distinction between these two genera is very important. Within this group, the pollen of Urtica membranacea is the only species that can be recognized easily under the microscope. For the research presented in this study, we built a dataset from 6472 pollen images and our aim was to find the best possible classifier on this dataset by analysing different classification methods, both machine learning and deep learning-based methods. For machine learning-based methods, we measured both texture and moment features based on images from the pollen grains. Varied feature selection techniques, classifiers as well as a hierarchical strategy were implemented for pollen classification. For deep learning-based methods, we compared the performance of six popular Convolutional Neural Networks: AlexNet, VGG16, VGG19, MobileNet V1, MobileNet V2 and ResNet50. Results show that compared with flat classification models, a hierarchical strategy yielded the highest accuracy with 94.5% among machine learning-based methods. Among deep learning-based methods, ResNet50 achieved an accuracy of 99.4%, slightly outperforming the other neural networks investigated. In addition, we investigated the influence on performance by changing the size of image datasets to 1000 and 500 images, respectively. Results demonstrated that on smaller datasets, ResNet50 still achieved the best classification performance. An ablation study was implemented to help understanding why the deep learning-based methods outperformed the other models investigated. Using Urticaceae pollen as an example, our research provides a strategy of selecting a classification model for pollen datasets with highly similar pollen grains to support palynologists and could potentially be applied to other image classification tasks.

The determination of daily concentrations of atmospheric pollen is important in the medical and biological fields. Obtaining pollen concentrations is a complex and time-consuming task for specialized personnel. The automatic location of pollen grains is a handicap due to the high complexity of the images to be processed, with polymorphic and clumped pollen grains, dust, or debris. The purpose of this study is to analyze the feasibility of implementing a reliable pollen grain detection system based on a convolutional neural network architecture, which will be used later as a critical part of an automated pollen concentration estimation system. We used a training set of 251 videos to train our system. As the videos record the process of focusing the samples, this system makes use of the 3D information presented by several focal planes. Besides, a separate set of 135 videos (containing 1234 pollen grains of 11 pollen types) was used to evaluate detection performance. The results are promising in detection (98.54% of recall and 99.75% of precision) and location accuracy (0.89 IoU as the average value). These results suggest that this technique can provide a reliable basis for the development of an automated pollen counting system.

… of different types of pollen grains/particles, it … pollen recognition in honey. This is the first step to finally achieving an objective tool that allows the correct labelling of any types of pollen in …

Automated pollen identification has become an increasingly important tool for palaeoecological research; however, its application to fossil pollen assemblages remains challenging due to complex backgrounds, morphological variability, and taxonomic similarity among pollen types. In this study, we propose Pollen-YOLO, a deep learning-based object detection framework designed for automated pollen identification from microscopic images, and evaluate its performance using the TPPOL23 dataset. The model integrates a tailored backbone architecture with attention-based feature enhancement and class-specific data augmentation strategies to address the characteristics of fossil pollen images. Experimental results indicate that Pollen-YOLO achieves stable and competitive detection performance for most pollen taxa under the tested conditions, particularly for dominant taxa with distinctive morphological features. Model behavior is further examined through ablation experiments and Grad-CAM-based interpretability analysis, which provide insights into feature learning and classification mechanisms. The applicability of the framework is explored using a fossil pollen sequence from the Shaqu profile on the Tibetan Plateau. Automated results show a high level of agreement with manual identification in capturing major stratigraphic trends and vegetation succession patterns, while discrepancies persist for morphologically similar or low-abundance taxa. Overall, this study suggests that object detection-based deep learning approaches have the potential to support fossil pollen analysis and palaeoecological reconstruction. Rather than replacing expert identification, Pollen-YOLO is intended as a complementary, high-throughput tool that may assist large-scale pollen analysis under appropriate quality control when combined with expert verification.

Traditional approaches to automatic classification of pollen grains consisted of classifiers working with feature extractors designed by experts, which modeled pollen grains aspects of special importance for biologists. Recently, a Deep Learning (DL) algorithm called Convolutional Neural Network (CNN) has shown a great improvement in performance in many computer vision tasks such as classification, due to this great performance the computational requirements have increased considerably; therefore, it is advisable to use new platforms such as the Graphics Processing Unit (GPU), which offer large computational resources for the development of new systems with CNN. This paper presents the GPU-Trained implementation of a DL system with the CNN algorithm, proposing a CNN model capable of running on a GPU in real-time for the automatic classification of 19 different pollen grains belonging to 14 different families, which are found in high concentrations in Mexico, and which are large interest in areas such as beekeeping, paleoecology, botany, allergology, agriculture among others. These areas seek to improve the collection of palynological data in terms of time and accuracy. In order to evaluate our model, evaluation tests were performed in the NVIDIA Jetson TX2 Developer Kit GPU. Experimental results achieves around 90% in CCR and Sensitivity in the proposed model. Additionally, the proposed model works at a processing speed of 6,826 Frames Per Second (FPS) and has approximately 50% fewer parameters than reported in related works.

Analysis of pollen material obtained from the Hirst-type apparatus, which is a tedious and labor-intensive process, is usually performed by hand under a microscope by specialists in palynology. This research evaluated the automatic analysis of pollen material performed based on digital microscopic photos. A deep neural network called YOLO was used to analyze microscopic images containing the reference grains of three taxa typical of Central and Eastern Europe. YOLO networks perform recognition and detection; hence, there is no need to segment the image before classification. The obtained results were compared to other deep learning object detection methods, i.e., Faster R-CNN and RetinaNet. YOLO outperformed the other methods, as it gave the mean average precision (mAP@.5:.95) between 86.8% and 92.4% for the test sets included in the study. Among the difficulties related to the correct classification of the research material, the following should be noted: significant similarities of the grains of the analyzed taxa, the possibility of their simultaneous occurrence in one image, and mutual overlapping of objects.

Accurate identification of pollen grains from Abies (fir), Picea (spruce), and Pinus (pine) is an important method for reconstructing historical environments, past landscapes and understanding human-environment interactions. However, distinguishing between pollen grains of conifer genera poses challenges in palynology due to their morphological similarities. To address this identification challenge, this study leverages advanced deep learning techniques, specifically transfer learning models, which are effective in identifying similarities among detailed features. We evaluated nine different transfer learning architectures: DenseNet201, EfficientNetV2S, InceptionV3, MobileNetV2, ResNet101, ResNet50, VGG16, VGG19, and Xception. Each model was trained and validated on a dataset of images of pollen grains collected from museum specimens, mounted and imaged for training purposes. The models were assessed on various performance metrics, including accuracy, precision, recall, and F1-score across training, validation, and testing phases. Our results indicate that ResNet101 relatively outperformed other models, achieving a test accuracy of 99%, with equally high precision, recall, and F1-score. This study underscores the efficacy of transfer learning to produce models that can aid in identifications of difficult species. These models may aid conifer species classification and enhance pollen grain analysis, critical for ecological research and monitoring environmental changes.

… due to airborne pollen, highlighting the urgent need for advanced pollen detection systems. This study presents a novel automated approach for pollen recognition using microscopy …

This study investigates the use of pollen elastically scattered light images for species identification. The aim was to identify the best recognition algorithms for pollen classification based on the scattering images. A series of laboratory experiments with a Rapid-E device of Plair S.A. was conducted collecting scattering images and fluorescence spectra from pollen of 15 plant genera. The collected scattering data were supplied to 32 different setups of 8 computer vision models based on deep neural networks. The models were trained to classify the pollen types, and their performance was compared for the test sub-samples withheld from the training. Evaluation showed that most of the tested computer vision models convincingly outperform the basic convolutional neural network used in our previous studies: the accuracy gain was approaching 10% for best setups. The models of the Weakly Supervised Object Detection approach turned out to be the most accurate, but also slow. However, even the best setups still did not provide sufficient recognition accuracy barely reaching 65%–70% in the repeated tests. They also showed many false positives when applied to real-life time series collected by Rapid-E. Similar to the previous studies, fusion of the new scattering models with the fluorescence-based identification demonstrated almost 15% higher skills than either of the approaches alone reaching 77–83% of the overall classification accuracy.

In this paper, we propose a system for authenticating local bee pollen against fraudulent samples using image processing and classification techniques. Our system is based on the colour properties of bee pollen loads and the use of one-class classifiers to reject unknown pollen samples. The latter classification techniques allow us to tackle the major difficulty of the problem, the existence of many possible fraudulent pollen types. Also presented is a multi-classifier model with an ambiguity discovery process to fuse the output of the one-class classifiers. The method is validated by authenticating Spanish bee pollen types, the overall accuracy of the final system of being 94%. Therefore, the system is able to rapidly reject the non-local pollen samples with inexpensive hardware and without the need to send the product to the laboratory.

Identifying the botanical origin of honey is essential for ensuring its quality, preventing adulteration, and protecting consumers. Traditional techniques, such as melissopalynology, physicochemical analysis, and PCR, are often labor-intensive, time-consuming, or limited to the detection of only known species, while advanced DNA sequencing remains prohibitively costly. In this study, we aim to develop a deep learning-based approach for identifying pollen grains extracted from honey and captured through microscopic imaging. To achieve this, we first constructed a dataset named VNUA-Pollen52, which consists of microscopic images of pollen grains collected from flowers of plant species cultivated in the surveyed area in Hanoi, Vietnam. Second, we evaluated the classification performance of advanced deep learning models, including MobileNet, YOLOv11, and Vision Transformer, on pollen grain images. To improve performances of these model, we proposed data augmentation and hybrid fusion strategies to improve the identification accuracy of pollen grains extracted from honey. Third, we developed an online platform to support experts in identifying these pollen grains and to gather expert consensus, ensuring accurate determination of the plant species and providing a basis for evaluating the proposed identification strategy. Experimental results on 93 images of pollen grains extracted from honey samples demonstrated the effectiveness of the proposed hybrid fusion strategy, achieving 70.21% accuracy at rank 1 and 92.47% at rank 5. This study demonstrates the capability of recent advances in computer vision to identify pollen grains using their microscopic images, thereby opening up opportunities for the development of automated systems that support plant traceability and quality control of honey.

With the spread of technology in several fields, there is an increasing demand to automate specialized tasks that usually require human involvement in order to maximize efficiency and reduce processing time. Pollen identification and classification is a proper example to be treated in the Palynology field, which has been an expensive qualitative process, involving observation and discrimination of features by highly qualified experts. Although it is the most accurate and useful method, it is a time-consuming process that slowed down the research progress. In this paper, we present a dataset composed of more than 13.000 objects, identified by an appropriate segmentation pipeline applied on aerobiological samples. Besides, we present the results obtained from the classification of these objects by taking advantage of several Machine Learning techniques, discussing which approaches have produced the most satisfactory results, and outlining the challenges we had to face to accomplish the task.

… a computer vision-driven framework designed for the segmentation and classification of pollen … to segmentation and 5842 images for classification across 28 distinct classes, which have …

Abstract The classification of pollen species and types is an important task in many areas like forensic palynology, archaeological palynology, and melissopalynology. This paper presents a new public annotated image dataset for the Brazilian Savanna called POLLEN73S composed of 2523 images from 73 pollen types. Using the state-of-the-art Convolutional Neural Networks (CNNs), we provide a baseline for pollen grain classification. Our experiments showed evidence that DenseNet-201 and ResNet-50 have superior performance against the other CNNs tested, achieving precision results of 95.7% and 94.0%, respectively. Due to its category coverage and satisfactory diversity of examples, POLLEN73S offers a diversity of pollen grain to guide progress in computer vision to solve Palynology problems.

… PollenBB16 is an RGB pollen image dataset of Chilean flora with pixel-accurate instance … networks can be trained with more accurate classification. The operational quality of the …

In a large number of scientific areas, such as immunology, forensics, paleoecology, and archeology, the study of pollen, i.e., palynology, plays an important role: from tracking climate changes, studying allergies, to forensic investigations or honey origin analysis. Since the mid-nineties of the last century, the idea for an automated solution to the problem of pollen identification and classification was formulated and since then, several attempts and proposals have been made and presented, based on different technologies, in particular in the field of Computer Vision. However, as of 2021 microscopic analyses are performed mainly manually by highly trained specialists, although the capabilities of artificial intelligence, especially Deep Neural Networks, are steadily increasing. In this work, we analyzed various state-of-the-art research work concerning pollen detection and classification and compared their methods and results. The problems, such as data accessibility, different methods of Machine Learning, and the intended applicability of the proposed solutions are explored. We also identified crucial issues that require further work and research. Our work will provide a thorough view on the current state of the art, its issues, and possibilities for the future.

Pollen identification is necessary for several subfields of geology, ecology, and evolutionary biology. However, the existing methods for pollen identification are laborious, time-consuming, and require highly skilled scientists. Therefore, there is a pressing need for an automated and accurate system for pollen identification, which can be beneficial for both basic research and applied issues such as identifying airborne allergens. In this study, we propose a deep learning (DL) approach to classify pollen grains in the Great Basin Desert, Nevada, USA. Our dataset consisted of 10,000 images of 40 pollen species. To mitigate the limitations imposed by the small volume of our training dataset, we conducted an in-depth comparative analysis of numerous pre-trained Convolutional Neural Network (CNN) architectures utilizing transfer learning methodologies. Simultaneously, we developed and incorporated an innovative CNN model, serving to augment our exploration and optimization of data modeling strategies. We applied different architectures of well-known pre-trained deep CNN models, including AlexNet, VGG-16, MobileNet-V2, ResNet (18, 34, and 50, 101), ResNeSt (50, 101), SE-ResNeXt, and Vision Transformer (ViT), to uncover the most promising modeling approach for the classification of pollen grains in the Great Basin. To evaluate the performance of the pre-trained deep CNN models, we measured accuracy, precision, F1-Score, and recall. Our results showed that the ResNeSt-110 model achieved the best performance, with an accuracy of 97.24%, precision of 97.89%, F1-Score of 96.86%, and recall of 97.13%. Our results also revealed that transfer learning models can deliver better and faster image classification results compared to traditional CNN models built from scratch. The proposed method can potentially benefit various fields that rely on efficient pollen identification. This study demonstrates that DL approaches can improve the accuracy and efficiency of pollen identification, and it provides a foundation for further research in the field.

… cannot meet the requirements of pollen forecasting. Recently, … attention in vision tasks, such as image classification. However, … a new Vision Transformer pipeline for image classification. …

The classification of pollen species and types is an important task in many areas like forensic palynology, archaeological palynology and melissopalynology. This paper presents the first annotated image dataset for the Brazilian Savannah pollen types that can be used to train and test computer vision based automatic pollen classifiers. A first baseline human and computer performance for this dataset has been established using 805 pollen images of 23 pollen types. In order to access the computer performance, a combination of three feature extractors and four machine learning techniques has been implemented, fine tuned and tested. The results of these tests are also presented in this paper.

We present results from the development and validation campaign of an optical pollen monitoring method based on time-resolved scattering and fluorescence. Focus is first set on supervised learning algorithms for pollen-taxa identification and on the determination of aerosol properties (particle size and shape). The identification capability provides a basis for a pre-operational automatic pollen season monitoring performed in parallel to manual reference measurements (Hirst-type volumetric samplers). Airborne concentrations obtained from the automatic system are compatible with those from the manual method regarding total pollen and the automatic device provides real-time data reliably (one week interruption over five months). In addition, although the calibration dataset still needs to be completed, we are able to follow the grass pollen season. The high sampling from the automatic device allows to go beyond the commonly-presented daily values and we obtain statistically significant hourly concentrations. Finally, we discuss remaining challenges for obtaining an operational automatic monitoring system and how the generic validation environment developed for the present campaign could be used for further tests of automatic pollen monitoring devices.

ABSTRACT The aim of this paper is to describe a solution suitable for the automation of standard pollen information service (EN 16868:2019). We are describing the RealForAll integrated information system developed for automatic airborne pollen detection and real-time data delivery to end-users. This solution is based on the measurements from the Rapid-E airborne particle monitor. The system incorporates an AI-enabled subsystem based on a convolutional neural network that continuously retrieves raw data from Rapid-E and performs the classification of airborne pollen. The main advantages of this system reflect in real-time data delivery and independence of aerobiology experts during the pollen season.

… Afterward, we present an automatic algorithm to distinguish the species of the family Urticaceae … We have implemented the pollen detection with the method described in Section IV. An …

Pollen is the most common cause of seasonal allergies, affecting over 33 % of the European population, even when considering only grasses. Informing the population and clinicians in real-time about the actual presence of pollen in the atmosphere is essential to reduce its harmful health and economic impact. Thus, there is a growing network of automatic particle analysers, and the reproducibility and transferability of implemented models are recommended since a reference dataset for local pollen of interest needs to be collected for each device to classify pollen, which is complex and time-consuming. Therefore, it would be beneficial to incorporate the reference dataset collected from other devices in different locations. However, it must be considered that laser-induced data are prone to device-specific noise due to laser and detector sensibility. This study collected data from two Rapid-E bioaerosol identifiers in Serbia and Italy and implemented a multi-modal convolutional neural network for pollen classification. We showed that models lost their performance when trained on data from one and tested on another device, not only in terms of the recognition ability but also in comparison with the manual measurements from Hirst-type traps. To enable pollen classification with just one model in both study locations, we first included the missing pollen classes in the dataset from the other study location, but it showed poor results, implying that data of one pollen class from different devices are more different than data of different pollen classes from one device. Combining all available reference data in a single model enabled the classification of a higher number of pollen classes in both study locations. Finally, we implemented a domain adaptation method, which improved the recognition ability and the correlations of transferred models only for several pollen classes.

Abstract. Pollen-induced allergies are among the most prevalent non-contagious diseases, with about a quarter of the European population being sensitive to various atmospheric bioaerosols. In most European countries, pollen information is based on a weekly-cycle Hirst-type pollen trap method. This method is labour-intensive and requires narrow specialized abilities and substantial time, so that the pollen data are always delayed and subject to sampling- and counting-related uncertainties. Emerging new approaches to automatic pollen monitoring can, in principle, allow for real-time availability of the data with no human involvement. The goal of the current paper is to evaluate the capabilities of the new Plair Rapid-E pollen monitor and to construct a first-level pollen recognition algorithm. The evaluation was performed for three devices located in Lithuania, Serbia and Switzerland, with independent calibration data and classification algorithms. The Rapid-E output data include multi-angle scattering images and the fluorescence spectra recorded at several times for each particle reaching the device. Both modalities of the Rapid-E output were treated with artificial neural networks (ANNs) and the results were combined to obtain the pollen type. For the first classification experiment, the monitor was challenged with a large variety of pollen types and the quality of many-to-many classification was evaluated. It was shown that in this case, both scattering- and fluorescence-based recognition algorithms fall short of acceptable quality. The combinations of these algorithms performed better, exceeding 80 % accuracy for 5 out of 11 species. Fluorescence spectra showed similarities among different species, ending up with three well-resolved groups: (Alnus, Corylus, Betula and Quercus), (Salix and Populus) and (Festuca, Artemisia and Juniperus). Within these groups, pollen is practically indistinguishable for the first-level recognition procedure. Construction of multistep algorithms with sequential discrimination of pollen inside each group seems to be one of the possible ways forward. In order to connect the classification experiment to existing technology, a short comparison with the Hirst measurements is presented and the issue of false positive pollen detections by Rapid-E is discussed.

Automated pollen analysis is not yet efficient on environmental samples containing many pollen taxa and debris, which are typical in most pollen-based studies. Contrary to classification, detection remains overlooked although it is the first step from which errors can propagate. Here, we investigated a simple but efficient method to automate pollen detection for environmental samples, optimizing workload and performance. We applied the YOLOv5 algorithm on samples containing debris and c. 40 Mediterranean plant taxa, designed and tested several strategies for annotation, and analyzed variation in detection errors. About 5% of pollen grains were left undetected, while 5% of debris were falsely detected as pollen. Undetected pollen was mainly in poor-quality images, or of rare and irregular morphology. Pollen detection remained effective when applied to samples never seen by the algorithm, and was not improved by spending time to provide taxonomic details. Pollen detection of a single model taxon reduced annotation workload, but was only efficient for morphologically differentiated taxa. We offer guidelines to plant scientists to analyze automatically any pollen sample, providing sound criteria to apply for detection while using common and user-friendly tools. Our method contributes to enhance the efficiency and replicability of pollen-based studies.

… (PDMs) pattern detection modules which are responsible for … This layer builds up a set of templates, or pattern detection … system but are generated automatically during training. When …

There is high demand for online, real-time and high-quality pollen data. To the moment pollen monitoring has been done manually by highly specialized experts. Here we evaluate the electronic Pollen Information Network (ePIN) comprising 8 automatic BAA500 pollen monitors in Bavaria, Germany. Automatic BAA500 and manual Hirst-type pollen traps were run simultaneously at the same locations for one pollen season. Classifications by BAA500 were checked by experts in pollen identification, which is traditionally considered to be the "gold standard" for pollen monitoring. BAA500 had a multiclass accuracy of over 90%. Correct identification of any individual pollen taxa was always >85%, except for Populus (73%) and Alnus (64%). The BAA500 was more precise than the manual method, with less discrepancies between determinations by pairs of automatic pollen monitors than between pairs of humans. The BAA500 was online for 97% of the time. There was a significant correlation of 0.84 between airborne pollen concentrations from the BAA500 and Hirst-type pollen traps. Due to the lack of calibration samples it is unknown which instrument gives the true concentration. The automaticBAA500 network delivered pollen data rapidly (three hours delay with real-time), reliably and online. We consider the ability to retrospectively check the accuracy of the reported classification essential for any automatic system.

Airborne pollen monitoring has been conducted for more than a century now, as knowledge of the quantity and periodicity of airborne pollen has diverse use cases, like reconstructing historic climates and tracking current climate change, forensic applications, and up to warning those affected by pollen-induced respiratory allergies. Hence, related work on automation of pollen classification already exists. In contrast, detection of pollen is still conducted manually, and it is the gold standard for accuracy. So, here we used a new-generation, automated, near-real-time pollen monitoring sampler, the BAA500, and we used data consisting of both raw and synthesised microscope images. Apart from the automatically generated, commercially-labelled data of all pollen taxa, we additionally used a manually created partial classification test set of bounding boxes and pollen taxa, so as to more accurately evaluate the real-life performance. For the pollen detection, we employed two-stage deep neural network object detectors. We explored a semi-supervised training scheme to remedy the partial labelling. Using a teacher-student approach, the model can add pseudo-labels to complete the labelling during training. To evaluate the performance of our deep learning algorithms and to compare them to the commercial algorithm of the BAA500, we created a manual test set, in which an expert aerobiologist corrected automatically annotated labels. For the novel manual test set, both the supervised and semi-supervised approaches clearly outperform the commercial algorithm with an F1 score of up to 76.9 % compared to 61.3 %. On an automatically created and partially labelled test dataset, we obtain a maximum mAP of 92.7 %. Additional experiments on raw microscope images show comparable performance for the best models, which potentially justifies reducing the complexity of the image generation process. Our results bring automatic pollen monitoring a step forward, as they close the gap in pollen detection performance between manual and automated procedure.

Pollen identification is required in different scenarios such as prevention of allergic reactions, climate analysis or apiculture. However, it is a time-consuming task since experts are required to recognize each pollen grain through the microscope. In this study, we performed an exhaustive assessment on the utility of texture analysis for automated characterisation of pollen samples. A database composed of 1800 brightfield microscopy images of pollen grains from 15 different taxa was used for this purpose. A pattern recognition-based methodology was adopted to perform pollen classification. Four different methods were evaluated for texture feature extraction from the pollen image: Haralick's gray-level co-occurrence matrices (GLCM), log-Gabor filters (LGF), local binary patterns (LBP) and discrete Tchebichef moments (DTM). Fisher's discriminant analysis and k-nearest neighbour were subsequently applied to perform dimensionality reduction and multivariate classification, respectively. Our results reveal that LGF and DTM, which are based on the spectral properties of the image, outperformed GLCM and LBP in the proposed classification problem. Furthermore, we found that the combination of all the texture features resulted in the highest performance, yielding an accuracy of 95%. Therefore, thorough texture characterisation could be considered in further implementations of automatic pollen recognition systems based on image processing techniques.

… almost identical to the results of the image analysis. This result demonstrates that the B/R ratio obtained from the image analysis is a reliable parameter for classifying pollen grains. …

… uorescent pollen. Classi®cation by the image analysis system showed the pollen of images … These data provide further evidence that classi®cation is dif®cult when pollen uorescence …

… Then, taking into account that the analysis of pollen can offer another tool for … pollen shape and, in particular, exine pattern using the scanning electron microscopy and image analysis (…

… the use of digital image analysis of static images of germinated pollen grains as a quantitative method for rapidly assessing pollen germination characteristics such as pollen tube area. …

Melissopalynology is an important analytical method to identify botanical origin of honey. Pollen grain recognition is commonly performed by visual inspection by a trained person. An alternative method for visual inspection is automated pollen analysis based on the image analysis technique. Image analysis transfers visual information to mathematical descriptions. In this work, the suitability of three microscopic techniques for automatic analysis of pollen grains was studied. 2D and 3D morphological characteristics, textural and colour features, and extended depth of focus characteristics were used for the pollen discrimination. In this study, 7 botanical taxa and a total of 2482 pollen grains were evaluated. The highest correct classification rate of 93.05% was achieved using the phase contrast microscopy, followed by the dark field microscopy reaching 91.02%, and finally by the light field microscopy reaching 88.88%. The most significant discriminant characteristics were morphological (2D and 3D) and colour characteristics. Our results confirm the potential of using automatic pollen analysis to discriminate pollen taxa in honey. This work provides the basis for further research where the taxa dataset will be increased, and new descriptors will be studied.

… in the pollen analysis problem domain have been based on samples from the Australian National University’s pollen reference collection (2,890 grains, 15 species) and images bundled …

Pollen identification and quantification are crucial but challenging tasks in addressing a variety of evolutionary and ecological questions (pollination, paleobotany), but also for other fields of research (e.g. allergology, honey analysis or forensics). Researchers are exploring alternative methods to automate these tasks but, for several reasons, manual microscopy is still the gold standard. In this study, we present a new method for pollen analysis using multi-spectral imaging flow cytometry in combination with deep learning. We demonstrate that our method allows fast measurement while delivering high accuracy pollen identification. A dataset of 426,876 images depicting pollen from 35 plant species was used to train a convolutional neural network classifier. We found the best-performing classifier to yield a species-averaged accuracy of 96 %. Even species that are difficult to differentiate using microscopy could be clearly separated. Our approach also allows a detailed determination of morphological pollen traits, such as size, symmetry or structure. Our phylogenetic analyses suggest phylogenetic conservatism in some of these traits. Given a comprehensive pollen reference database, we provide a powerful tool to be used in any pollen study with a need for rapid and accurate species identification, pollen grain quantification, and trait extraction of recent pollen.

Pollen monitoring is of great importance for the prevention of allergy. As this activity is still largely carried out by humans, there is an increasing interest in the automation of pollen monitoring. The goal is to reduce monitoring time in order to plan more efficient treatments. In this context, an original device based on computer vision is developed. The goal of such a system is to provide accurate measurement of pollen concentration. This information can be used as well by palynologists, clinicians or by a forecast system to predict pollen dispersion. The system is composed of two modules: pollen grain extraction and pollen grain recognition. In the first module, the pollen grains are observed in light microscopy and are extracted automatically from a microscopic slide dyed with fuchsin and digitised in 3D. The colour segmentation techniques implemented on a hardware architecture are presented. In the second module, the pollen grains are analysed for recognition. To accomplish recognition, it is necessary to work on 3D images and to use deep palynological knowledge. This knowledge describes the pollen types according to their main visible characteristerics and to those which are important for recognition. Some pollen structures are identified, like the pore with annulus in Poaceae, the reticulum in Olea and similar pollen types or the cytoplasm in Cupressaceae. Preliminary results show correct recognition of some pollen types, like Urticaceae or Poaceae, and some groups of pollen types, like reticulate group.

Pollen is used to investigate a diverse range of ecological problems, from identifying plant–pollinator relationships to tracking flowering phenology. Pollen types are identified according to a set of distinctive morphological characters which are understood to capture taxonomic differences and phylogenetic relationships among taxa. However, categorizing morphological variation among hyperdiverse pollen samples represents a challenge even for an expert analyst. We present an automated workflow for pollen analysis, from the automated scanning of pollen sample slides to the automated detection and identification of pollen taxa using convolutional neural networks (CNNs). We analysed aerial pollen samples from lowland Panama and used a microscope slide scanner to capture three‐dimensional representations of 150 sample slides. These pollen sample images were annotated by an expert using a virtual microscope. Metadata were digitally recorded for ~100 pollen grains per slide, including location, identification and the analyst's confidence of the given identification. We used these annotated images to train and test our detection and classification CNN models. Our approach is two‐part. We first compared three methods for training CNN models to detect pollen grains on a palynological slide. We next investigated approaches to training CNN models for pollen identification. Because the diversity of pollen taxa in environmental and palaeontological samples follows a long‐tailed distribution, we experimented with methods for addressing imbalanced representation using our most abundant 46 taxa. We found that properly weighting pollen taxa in our training objective functions yielded improved accuracy for individual taxa. Our average accuracy for the 46‐way classification problem was 82.3%. We achieved 89.5% accuracy for our 25 most abundant taxa. Pollen represents a challenging visual classification problem that can serve as a model for other areas of biology that rely on visual identification. Our results add to the body of research demonstrating the potential for a fully automated pollen classification system for environmental and palaeontological samples. Slide imaging, pollen detection and specimen identification can be automated to produce a streamlined workflow.

Significance We demonstrate that combining optical superresolution imaging with deep learning classification methods increases the speed and accuracy of assessing the biological affinities of fossil pollen taxa. We show that it is possible to taxonomically separate pollen grains that appear morphologically similar under standard light microscopy based on nanoscale variation in pollen shape, texture, and wall structure. Using a single pollen morphospecies, Striatopollis catatumbus, we show that nanoscale morphological variation within the fossil taxon coincides with paleobiogeographic distributions. This new approach improves the taxonomic resolution of fossil pollen identifications and greatly enhances the use of pollen data in ecological and evolutionary research. Taxonomic resolution is a major challenge in palynology, largely limiting the ecological and evolutionary interpretations possible with deep-time fossil pollen data. We present an approach for fossil pollen analysis that uses optical superresolution microscopy and machine learning to create a quantitative and higher throughput workflow for producing palynological identifications and hypotheses of biological affinity. We developed three convolutional neural network (CNN) classification models: maximum projection (MPM), multislice (MSM), and fused (FM). We trained the models on the pollen of 16 genera of the legume tribe Amherstieae, and then used these models to constrain the biological classifications of 48 fossil Striatopollis specimens from the Paleocene, Eocene, and Miocene of western Africa and northern South America. All models achieved average accuracies of 83 to 90% in the classification of the extant genera, and the majority of fossil identifications (86%) showed consensus among at least two of the three models. Our fossil identifications support the paleobiogeographic hypothesis that Amherstieae originated in Paleocene Africa and dispersed to South America during the Paleocene-Eocene Thermal Maximum (56 Ma). They also raise the possibility that at least three Amherstieae genera (Crudia, Berlinia, and Anthonotha) may have diverged earlier in the Cenozoic than predicted by molecular phylogenies.

Recognizing the types of pollen grains and estimating their proportion in pollen mixture samples collected in a specific geographical area is important for agricultural, medical, and ecosystem research. Our paper adopts a convolutional neural network for the automatic segmentation of pollen species in microscopy images, and proposes an original strategy to train such network at reasonable manual annotation cost. Our approach is founded on a large dataset composed of pure pollen images. It first (semi-)manually segments foreground, i.e. pollen grains, and background in a fraction of those images, and use the resulting annotated dataset to train a universal pollen segmentation CNN. In the second step, this model is used to automatically segment a large number of additional pure pollen images, so as to supervise the training of a pollen species segmentation model. Despite the fact that it has been trained from pure images only, the model is shown to provide accurate segmentation of species in pollen mixtures. Our experiments also demonstrate that dedicating a model to the segmentation of a subset of the available pure pollen species makes it possible to train a bin pollen class, corresponding to pollen species that are not in the subset of species recognized by the model. This strategy is useful to cope with unexpected species in a mixture.

The automation of pollen identification has seen vast improvements in the past years, with Convolutional Neural Networks coming out as the preferred tool to train models. Still, only a small portion of works published on the matter address the identification of fossil pollen. Fossil pollen is commonly extracted from organic sediment cores and are used by paleoecologists to reconstruct past environments, flora, vegetation, and their evolution through time. The automation of fossil pollen identification would allow paleoecologists to save both time and money while reducing bias and uncertainty. However, Convolutional Neural Networks require a large amount of data for training and databases of fossilized pollen are rare and often incomplete. Since machine learning models are usually trained using labelled fresh pollen associated with many different species, there exists a gap between the training data and target data. We propose a method for a large-scale fossil pollen identification workflow. Our proposed method employs an accelerated fossil pollen extraction protocol and Convolutional Neural Networks trained on the labelled fresh pollen of the species most commonly found in Northeastern American organic sediments. We first test our model on fresh pollen and then on a full fossil pollen sequence totalling 196,526 images. Our model achieved an average per class accuracy of 91.2% when tested against fresh pollen. However, we find that our model does not perform as well when tested on fossil data. While our model is overconfident in its predictions, the general abundance patterns remain consistent with the traditional palynologist IDs. Although not yet capable of accurately classifying a whole fossil pollen sequence, our model serves as a proof of concept towards creating a full large-scale identification workflow.

Abstract Pollen grains are valuable paleoclimate and paleovegetation proxies which require extensive knowledge of morphotypes and long acquisition time under the microscope. The abundance of damaged, folded, and broken pollen grains in the fossil register and sometimes also in modern soil and sediment samples, has so far prevented automation of pollen identification. Recent improvements in machine learning, however, have allowed reconsidering this approach. Here we present an automated approach which is capable of assisting palynologists with poorly preserved pollen samples. Called multi-CNNs, this approach is based on multiple convolutional neural networks (CNNs) integrated in a decision tree system. To test it, we built a system designed for three botanical families very common in the modern and fossil pollen assemblages of Eastern Africa, namely Amaranthaceae, Poaceae, and Cyperaceae. Our system was tested on stacked optical images of 8 pollen types (6 Amaranthaceae, 1 Poaceae, 1 Cyperaceae) using a training dataset of 1102 intact pollen grains and three validation datasets of intact (276 grains), damaged (223 grains), and fossil pollen (97 grains). We show that our system successfully recognizes intact, damaged, and fossil pollen grains with very low misclassification rates of 0%, 2.8%, and 3.7%, respectively. The use of augmentation on stacked optical images during the training increases classification accuracy. Following a palynologist's approach, our system allows grains without obvious characters to be classified into a class of high taxonomic level or as indeterminable pollen. This is the first software able to process grains with a wide range of taphonomical stages, which makes it the first truly applicable to automated pollen identification of fossil material.

ABSTRACT Allergenic pollen affects the quality of life for over 30% of the European population. Since the treatment efficacy is highly related to the actual exposure to pollen, information about the type and number of airborne pollen grains in real-time is essential for reducing their impact. Therefore, the automation of pollen monitoring has become an important research topic. Our study is focused on the Rapid-E real-time bioaerosol detector. So far, vanilla convolutional neural networks (CNNs) are the only deep architectures evaluated for pollen classification on multi-modal Rapid-E data obtained by exposing collected pollen samples of known classes to the device in a controlled environment. This study contributes to the further development of pollen classification models on Rapid-E data by experimenting with more advanced concepts of CNNs, residual, and inception networks. Our experiments included a comprehensive comparison of different CNN architectures, and obtained results provided valuable insights into which convolutional blocks improve pollen classification. We propose a new model which, coupled with a specific training strategy, improves the current state-of-the-art by reducing its relative error rate by 9%.

… -guided deep feature learning (PK-DFL) for real-world optical microscope … Pollen classification uses a deep network (CNN) to classify pollen via imitating the pollen identification …

Simple Summary Pollen allergy is a highly prevalent disease affecting humans worldwide. Early pollen identification can help allergic individuals to prevent pollinosis. Recently, automatic pollen identification (API) has been shown to play a prominent role in pollen concentration monitoring. Developing an accurate and effective identification system may provide new insights for pollinosis prevention. This paper presents a novel automatic pollen identification method integrating localization tasks and classification tasks, thus perfectly mimicking the observation process from palynologists. The inter-task dependence and intra-task reliability are simultaneously considered in this method to effectively enhance the pollen identification performance. We believe that our study will contribute to enhancing symptom control of pollen allergy and maintaining the life quality of allergic patients. Abstract Existing API approaches usually independently leverage detection or classification models to distinguish allergic pollens from Whole Slide Images (WSIs). However, palynologists tend to identify pollen grains in a progressive learning manner instead of the above one-stage straightforward way. They generally focus on two pivotal problems during pollen identification. (1) Localization: where are the pollen grains located? (2) Classification: which categories do these pollen grains belong to? To perfectly mimic the manual observation process of the palynologists, we propose a progressive method integrating pollen localization and classification to achieve allergic pollen identification from WSIs. Specifically, data preprocessing is first used to cut WSIs into specific patches and filter out blank background patches. Subsequently, we present the multi-scale detection model to locate coarse-grained pollen regions (targeting at “pollen localization problem”) and the multi-classifiers combination to determine the fine-grained category of allergic pollens (targeting at “pollen classification problem”). Extensive experimental results have demonstrated the feasibility and effectiveness of our proposed method.

Pollen grains are microscopic structures produced by plants in order to reproduce. These grains are necessary for the pollination and fertilization processes, which are vital for the continued existence and diversification of plant species. Botanists and scholars have been interested in the identification and categorization of pollen grains for many years. A convolutional neural network (CNN) model is proposed in this study for the classification of pollen grains from a dataset of 805 photos with 23 annotated classes and an image resolution of 128 × 128 pixels with 3 color channels. The images of pollen grains from 23 plant species make up the publicly accessible Pollen Grain Dataset, which provided the dataset for this study. Based on the physical traits and properties of the pollen grains, the suggested CNN model is intended to properly categorize the pollen grains. The performance of the suggested CNN model is evaluated using various performance metrics. The CNN model’s categorization of pollen grains had an accuracy rate of 87.60%. The outcomes of this study’s investigation show how well CNN models can categorize pollen grains. Compared to manual categorization, which may be error-prone and time-consuming, this automated technique can save a lot of time and effort. According to the observations, the suggested CNN model is efficient at classifying pollen grains and may be applied to several tasks, including recognizing diverse pollen species for use in environmental study, agricultural practice, and scientific inquiry.

: Pollen is an important substance produced by seed plants. They contain the male gametes which are necessary for fertilization and the reproduction of flowering plants. The scientific study of pollen, palynology, plays a crucial role in a number of disciplines, such as allergology, ecology, forensics, as well as food-production. Current trends in climate research indicate an increasing importance of palynology, partly due to a projected rise in allergies. Pollen detection and classification in microscopic images via deep neural networks has been studied and researched, however, pollen data is often sparse or imbalanced, especially when compared to the number of plant species, which is estimated to be between 330,000 and 450,000, of which only a small percentage is investigated. In this work, we present a solution that does not require a large number of data samples by employing Few-Shot Learning. Our work shows, that by utilizing Prototypical Networks, an average classification accuracy of 90% can be achieved on state-of-the-art pollen data sets. The results can be further improved by fine-tuning the net, achieving up to 98% accuracy on novel classes. To our best knowledge, this is the first attempt at applying Few-Shot Learning in the field of pollen analysis.

Automatic pollen images recognition is crucial for pollinosis symptoms prevention and treatment. The problem of pollen recognition can be efficiently solved using deep learning, however neural networks require tens of thousands of images to generalize. At the same time, the existing open pollen images datasets are very small. In this paper, we present a novel open pollen dataset annotated for both detection and classification tasks. Based on our dataset we study learning from a small data using different state-of-the-art approaches. For the detection task we propose to use our new Bayesian RetinaNet network, which models aleatoric uncertainty. We compare it with the baseline RetinaNet and demonstrate that our model allows for higher detection precision. For the classification task we compare the impact of pre-training on the synthetic images from generative adversarial networks (GANs) and metric-based few-shot learning. Namely, we pre-trained our small convolutional neural network and Siamese neural network classifiers on synthetic pollen images generated by two GANs: StyleGAN and Self-attention GAN. The best classifier is the convolutional neural network pre-trained on StyleGAN images. Our best models achieved 96.3% of mean average precision for the detection task and 97.7% of F1 measure for the classification task on 13 pollen plant species. We have implemented the best models in our pollen recognition web service, which is available for palynologists on request.

Real-time bioaerosol monitoring is improving the quality of life for people affected by allergies, but it often relies on deep learning models which pose challenges for widespread adoption. These models are typically trained in a supervised fashion and require considerable effort to produce large amounts of annotated data, an effort that must be repeated for new particles, geographical regions, or measurement systems. In this work, we show that self-supervised learning and few-shot learning can be combined to classify holographic images of pollen grains using a large collection of unlabelled data and only a few identified particles per type. We first demonstrate that self-supervision on pictures of unidentified particles from ambient air measurements enhances identification even when labelled data are abundant. Most importantly, it greatly improves few-shot classification when only a handful of labelled images are available. Our findings suggest that real-time bioaerosol monitoring workflows can be substantially optimized, and the effort required to adapt models for different situations considerably reduced.

Pollen identification is critical in a number of fields. However, existing studies on automatic pollen identification mainly rely on image information without the consideration of the spatio and and temporal information of pollen, which can provide valuable knowledge to reduce the decision space of pollen classifier. This paper proposes a transformer-based framework to identify pollen by combining images and temporal and spatial features in the text. First, the pretrained BERT model is utilized to extract text features and the pretrained DenseNet model is used to extract image features. Then, a transformer aggregator is proposed to fuse multi modal features by superposing different channels for pollen classification. Experiments with many real-world datasets show that our transformer-based pollen recognition model achieves excellent performance compared to other state-of-the-art baselines, which validates the effectiveness of our new approach.

The phylogenetic interpretation of pollen morphology is limited by our inability to recognize the evolutionary history embedded in pollen features. Deep learning offers tools for connecting morphology to phylogeny. Using neural networks, we developed an explicitly phylogenetic toolkit for analyzing the overall shape, internal structure, and texture of a pollen grain. Our analysis pipeline determines whether testing specimens are from unknown species based on uncertainty estimates. Features of novel specimens are passed to a multi-layer perceptron network trained to transform these features into predicted phylogenetic distances from known taxa. We used these predicted distances to place specimens in a phylogeny using Bayesian inference. We trained and evaluated our models using optical superresolution micrographs of 30 Podocarpus species. We then used trained models to place nine fossil Podocarpidites specimens within the phylogeny. In doing so, we demonstrate that the phylogenetic history encoded in pollen morphology can be recognized by neural networks and that deep-learned features can be used in phylogenetic placement. Our approach makes extinction and speciation events that would otherwise be masked by the limited taxonomic resolution of the fossil pollen record visible to palynological analysis. Significance Statement Machine learned features from deep neural networks can do more than categorize and classify biological images. We demonstrate that these features can also be used to quantify morphological differences among pollen taxa, discover novel morphotypes, and place fossil specimens on a phylogeny using Bayesian inference. Deep learning can be used to characterize and identify and morphological features with evolutionary significance. These features can then be used to infer phylogenetic distance. This approach fundamentally changes how fossil pollen morphology can be interpreted, allowing greater evolutionary inference of fossil pollen specimens. The analysis framework, however, is not specific to pollen and can be generalized to other taxa and other biological images.

… Additionally, the rich feature vectors generated by the model can be used for few-shot … to classify pollen grains we select the dataset used in [14]. This dataset contains pollen grains from …

. Identifying all the pollen species present on earth, and more particularly in a territory, is a major concern for palynologists. This is an arduous task that can be automated using artificial intelligence. Many studies have tried to solve this problem by using machine learning and deep learning. In this paper, we present three pollen recognition approaches: Classification with no examples, recognition with a sufficient number of examples, recognition with un-sufficient number of examples. For each of them, we propose respectively to use Visual Bag of Word and expectation-maximization clustering algorithms, Classification using Local Binary patterns and the Gabor Filter Feature, Local Binary Patterns and Prototypical Networks. We find 77,38% recognition for 10 pollen species rate for the first one, 90.80 % for training with a sufficient number of examples and 80 species, and 20 different pollen species and finally 84,30% for the third approach with one example for training and 20 species.

Automated pollen recognition is a foundational tool for diverse scientific domains, including paleoclimatology, biodiversity monitoring, and agricultural science. However, conventional methods create a critical data bottleneck, limiting the temporal and spatial resolution of ecological analysis. Existing deep learning models often fail to achieve the requisite localization accuracy for microscopic pollen grains, which are characterized by their minute size, indistinct edges, and complex backgrounds. To overcome this, we introduce HieraEdgeNet, a novel object detection framework. The core principle of our architecture is to explicitly extract and hierarchically fuse multi-scale edge information with deep semantic features. This synergistic approach, combined with a computationally efficient large-kernel operator for fine-grained feature refinement, significantly enhances the model’s ability to perceive and precisely delineate object boundaries. On a large-scale dataset comprising 44,471 annotated microscopic images containing 342,706 pollen grains from 120 classes, HieraEdgeNet achieves a mean Average Precision of 0.9501 (mAP@0.5) and 0.8444 (mAP@0.5:0.95), substantially outperforming state-of-the-art models such as YOLOv12n and the Transformer-based RT-DETR family in terms of the accuracy–efficiency trade-off. This work provides a powerful computational tool for generating the high-throughput, high-fidelity data essential for modern ecological research, including tracking phenological shifts, assessing plant biodiversity, and reconstructing paleoenvironments. At the same time, we acknowledge that the current two-dimensional design cannot directly exploit volumetric Z-stack microscopy and that strong domain shifts between training data and real-world deployments may still degrade performance, which we identify as key directions for future work. By also enabling applications in precision agriculture, HieraEdgeNet contributes broadly to advancing ecosystem monitoring and sustainable food security.

… requires not only learning the relationship among multi-scale features but also refining the … -based multi-scale feature fusion network (AMFF-Net) for automatic pollen detection on real-…

Pollen allergies are seasonal epidemic diseases that are accompanied by high incidence rates, especially in Beijing, China. With the development of deep learning, key progress has been made in the task of automatic pollen grain classification, which could replace the time-consuming and laborious manual identification process using a microscope. In China, few pioneering works have made significant progress in automatic pollen grain classification. Therefore, we first constructed a multi-class and large-scale pollen grain dataset for the Beijing area in preparation for the task of pollen classification. Then, a deblurring pipeline was designed to enhance the quality of the pollen grain images selectively. Moreover, as pollen grains vary greatly in size and shape, we proposed an easy-to-implement and efficient multi-scale deep learning architecture. Our experimental results showed that our architecture achieved a 97.7% accuracy, based on the Resnet-50 backbone network, which proved that the proposed method could be applied successfully to the automatic identification of pollen grains in Beijing.

Pollen allergy has emerged as a critical global health challenge. Proactive pollen monitoring is imperative for safeguarding susceptible populations through timely preventive interventions. Current manual detection methods suffer from inherent limitations: notably, suboptimal accuracy and delayed response times, which hinder effective allergy management. Therefore, we present an automated pollen concentration detection system integrated with a novel GGD-YOLOv8n model (Ghost-generalized-FPN-DualConv-YOLOv8), which was specifically designed for allergenic pollen species identification. The methodological advancements comprise three components: (1) combining the C2f convolution in Backbone with the G-Ghost module, this module generates features through half-convolution operations and half-symmetric linear operations, enhancing the extraction and expression capabilities of detailed feature information. (2) The conventional neck network is replaced with a GFPN architecture, facilitating cross-scale feature aggregation and refinement. (3) Standard convolutional layers are substituted with DualConv, thereby reducing model complexity by 22.6% (parameters) and 22% GFLOPs (computational load) while maintaining competitive detection accuracy. This systematic optimization enables efficient deployment on edge computing platforms with stringent resource constraints. The experimental validation substantiates that the proposed methodology outperforms the baseline YOLOv8n model, attaining a 5.4% increase in classification accuracy accompanied by a 4.7% enhancement in mAP@50 metrics. When implemented on Jetson Nano embedded platforms, the system demonstrates computational efficiency with an inference latency of 364.9 ms per image frame, equating to a 22.5% reduction in processing time compared to conventional implementations. The empirical results conclusively validate the dual superiority in detecting precision and operational efficacy when executing microscopic pollen image analysis on resource-constrained edge computing devices; they establish a feasible algorithm framework for automated pollen concentration monitoring systems.

Automated pollen detection is essential for ecological monitoring, allergy forecasting, and biodiversity research. However, existing methods rely heavily on manual or semi-automated annotations, limiting scalability and broader applicability. We introduce a highly automated training dataset generation pipeline that combines one-shot detection with systematic refinement, producing tens of thousands of high-quality annotations from bright-field microscopy while significantly reducing manual effort and annotation costs. Using multi-regional datasets from France, Hungary, and Sweden, we trained object detection models on seven pollen taxa and evaluated their performance on both external pure and mixed species slides and real-world airborne samples. We assessed the reusability of pretrained vision models for pollen detection, aiming to reduce the need for extensive retraining. Using linear probing, we identified foundational Vision Transformers (ViTs) as the most effective feature extractors and integrated them into Faster R-CNN detection models. We benchmarked these models against ResNet50, a widely adopted backbone in biological imaging. On held-out regions of the training datasets, our models achieved high performance in both classification and detection tasks. On independent reference slides from other datasets, ViTs continued to outperform ResNet50 in classification. However, in full object detection and under real deployment conditions, ResNet50-based models remained competitive and achieved the highest accuracy for detecting Ambrosia, a major allergen with public health significance. Cross-dataset generalization remains a challenge, underscoring the need for domain adaptation techniques such as stain normalization and data augmentation. This study establishes a scalable framework for AI-assisted pollen monitoring, supporting large-scale slide digitization and enabling applications in long-term ecological research, allergen surveillance, and automated biodiversity assessment.

Potato pollen segmentation is a pivotal task in agricultural image analysis, underpinning critical applications in plant breeding, disease monitoring, and genetic diversity assessment. Conventional methods for pollen viability detection via traditional image processing often encounter bottlenecks in efficiency and accuracy, limiting their utility in high-throughput scenarios. Addressing these challenges, this paper introduces a novel multi-scale attention U-Net architecture tailored for precise potato pollen segmentation and viability assessment. Our methodology leverages a rigorously annotated dataset of 45 highresolution microscopic images ($3600 \times 3600$ pixels), capturing diverse pollen morphologies under standardized conditions. At the core of this method lies a distinctive symmetric architecture complemented by skip connections, which enables the seamless integration of contextual and detailed information. This design effectively captures both the broad semantic context and finegrained details, laying a solid foundation for precise segmentation. Furthermore, multi-scale dilated convolutions are strategically employed to dynamically extract hierarchical features across diverse scales, adeptly adapting to the morphological variability of pollen grains. Simultaneously, a channel attention mechanism is incorporated to selectively emphasize crucial information channels, significantly enhancing the discriminative power of the model. These innovative components work in synergy to elevate the accuracy of segmentation results, thereby enabling highly efficient pixel-level image segmentation. Experimental results demonstrate the model's superior performance, achieving an accuracy of 99.15% in pixel-wise segmentation a significant improvement over existing methods.

… a Transformer-based pollen detection framework named ECF-… Extensive experiments conducted on a self-constructed pollen … -based methods for practical pollen detection applications. …

Airborne allergenic pollen can trigger various hay fevers such as seasonal allergic rhinitis and bronchial asthma. Accurate and timely pollen forecasting services play a crucial role in enabling individuals with hay fever to take preventive measures proactively. Currently, automatic pollen recognition research have provided new insights into timely pollen forecasting services. However, the forecasting results of existing automatic pollen recognition methods fail to convince owing to the pollen data characteristics in real scenes. Hence, we fully simulate the observation strategy of palynologists (namely, localization-before-classification) to address the challenges encountered in real-scenes. This strategy comprises two key steps: (1) to determine the location information of each pollen grain; (2) to distinguish the category information of pollen grain (using the key features of pollen grain, such as contour, color and texture). Motivated by this strategy, we propose a computer-aided system for eight airborne allergenic pollens recognition to a specific area in Beijing called PBJ-Sys. Pollen whole-slide imaging images are utilized as input, and four components (Image Prepocessing, Multi-scale Fusion Pollen Localization, Knowledge-guided Pollen Classification and Result Statistics) within the PBJ-Sys are integrated to output the total pollen concentration and single pollen category quantities results. The PBJ-Sys helps to reduce the burden of manual microscopy, enhance symptom control and maintain quality of life in pollen allergy.

Pollen identification is a critical task across various scientific disciplines, including geology, ecology, evolutionary biology and botany. However, existing methods for pollen identification are often labour-intensive, time-consuming and dependent on highly skilled experts, highlighting the need for an automated and precise system. This study introduces an innovative approach that combines Gabor Filters (GF) with Convolutional Neural Networks (CNN) to enhance the accuracy of pollen classification. The Gabor filters are applied to high-resolution images of diverse pollen species, accentuating texture-specific details essential for differentiation. These pre-processed images are subsequently analysed using a CNN architecture with multiple layers designed to discern hierarchical features critical for precise classification. The proposed GF-CNN model demonstrates exceptional proficiency, achieving remarkable accuracy rates of 99.85% for the Malaysian Pollen Dataset (MPD) and 99.43% for the New Zealand Pollen Dataset (NPD). These results underscore the model's ability to balance precision and recall effectively. Additionally, the model exhibits high sensitivity, indicating an increased true-positive rate, which is essential for detailed ecological studies. Furthermore, the model's improved specificity scores highlight its success in minimizing false positives, emphasizing its relevance for precision-focused research.

… For the sake of running efficiency, the author in [10] proposed a light and fast face detector (LFFD) inspired by the one-stage and multi-scale object detector SSD. Instead of pre-defined …

Podocarpus pollen morphology is shaped by both phylogenetic history and the environment. We analyzed the relationship between pollen traits quantified using deep learning and environmental factors within a comparative phylogenetic framework. We investigated the influence of mean annual temperature, annual precipitation, altitude, and solar radiation in driving morphological change. We used trait-environment regression models to infer the temperature tolerances of 31 Neotropical Podocarpidites fossils. Ancestral state reconstructions were applied to the Podocarpus phylogeny with and without the inclusion of fossils. Our results show that temperature and solar radiation influence pollen morphology, with thermal stress driving an increase in pollen size and higher UV-B radiation selecting for thicker corpus walls. Fossil temperature tolerances inferred from trait-environment models aligned with paleotemperature estimates from global paleoclimate models. Incorporating fossils into ancestral state reconstructions revealed that early ancestral Podocarpus lineages were likely adapted to warm climates, with cool-temperature tolerance evolving independently in high-latitude and high-altitude species. Our results highlight the importance of deep learning-derived features in advancing our understanding of plant environmental adaptations over evolutionary timescales. Deep learning allows us to quantify subtle interspecific differences in pollen morphology and link these traits to environmental preferences through statistical and phylogenetic analyses.

… science, but traditional microscopic morphological analysis methods are inefficient and … based on deep learning to improve the accuracy and efficiency of pollen identification. We …

This study presents an image analysis framework coupled with machine learning algorithms for the classification of microscopy pollen grain images. Pollen grain classification has received notable attention concerning a wide range of applications such as paleontology and honey certification, forecasting of allergies caused of airborne pollen and food technology. It requires an extensive qualitative process that is mostly performed manually by an expert. Although manual classification shows satisfactory performance, it may suffer from intra and inter-observer variability and it is time consuming. This study benefits from the advances of image processing and machine learning and proposes a fully-automated analysis pipeline aiming to: A) calculate morphological characteristics from the images using a cost-effective microscope, and b) classify images into 6 pollen classes. A private dataset from the Department of Agriculture of the Hellenic Mediterranean University in Crete containing 564 images was used in this study. A Random Forest (RF) classifier was utilized to classify images. A repeated nested cross-validation (nested-CV) schema was used to estimate the generalization performance and prevent overfitting. Image preprocessing, extraction of geometric and textural characteristics and feature selection were implemented prior to the assessment of the classification performance and a mean accuracy of 88.24% was reported.

… An extensive study on pollen grain identification is presented in this work. A combination … pollen grain discriminative features as well as the usage of the most popular feature extraction …

Pollen and tracheophyte spores are ubiquitous environmental indicators at local and global scales. Palynology is typically performed manually by microscopic analysis; a specialised and time-consuming task limited in taxonomical precision and sampling frequency, therefore restricting data quality used to inform climate change and pollen forecasting models. We build on the growing work using AI (artificial intelligence) for automated pollen classification to design a flexible network that can deal with the uncertainty of broad-scale environmental applications. We combined imaging flow cytometry with Guided Deep Learning to identify and accurately categorise pollen in environmental samples; here, pollen grains captured within c. 5500 Cal yr BP old lake sediments. Our network discriminates not only pollen included in training libraries to the species level but, depending on the sample, can classify previously unseen pollen to the likely phylogenetic order, family and even genus. Our approach offers valuable insights into the development of a widely transferable, rapid and accurate exploratory tool for pollen classification in 'real-world' environmental samples with improved accuracy over pure deep learning techniques. This work has the potential to revolutionise many aspects of palynology, allowing a more detailed spatial and temporal understanding of pollen in the environment with improved taxonomical resolution.

… image analysis and visual analysis (Table 1). For Fig. 1E, multiple image analysis did not count the number of pollen … 2568 images, multiple image analysis failed to detect pollen grains …

… The processed image was saved as an ASCII file in a PBM (monochrome bitmap) format. … Experimental apparatus for particle counting by image analysis. The pollen grain drop is …

Abstract Bee pollen is a natural matrix widely studied in its nutritional and bioactive compounds, including carotenoids. That composition is usually identified by Rapid Resolution Liquid Chromatography (RRLC) coupled to UV–Vis spectrophotometry, an expensive method that requires complex sample preparation and long analysis time. In this work, a correlation between colorimetric coordinates and carotenoid composition was evaluated. Through Digital Image Analysis (DIA) by DigiEye, the color characteristics were determined, and carotenoids profile was done by RRLC. The correlations were made by multiple linear regression (MLR). From 12 carotenoids found in the samples, six had a coefficient R2 > 0.75 between reference and predict value. Heterogeneous mixtures of bee pollen samples were analyzed, and the suitability of the mathematical models could be corroborated because the relative error for most of the compounds was less than 20%. It has been demonstrated that union of Tristimulus Colorimetry and Image Analysis represent an effective tool to estimate the chemical composition in food industry.

Anthesis prediction is crucial for breeding wheat. While current tools provide estimates of average anthesis at the field scale, they fail to address the needs of breeders who require accurate predictions for individual plants. Hybrid breeders have to finalize their plans for pollination at least 10 days before such flowering is due and biotechnology field trials in the United States and Australia must report to regulators 7–14 days before the first plant flowers. Currently, predicting anthesis of individual wheat plants is a labour-intensive, inefficient, and costly process. Individual wheat of the same cultivar within the same field may exhibit substantial variations in anthesis timing, due to significant variations in their immediate surroundings. In this study, we developed an efficient and cost-effective machine vision approach to predict anthesis of individual wheat plants. By integrating RGB imagery with in-situ meteorological data, our multimodal framework simplifies the anthesis prediction problem into binary or three-class classification tasks, aligning with breeders' requirements in individual wheat flowering prediction on the crucial days before anthesis. Furthermore, we incorporated a few-shot learning method to improve the model's adaptability across different growth environments and to address the challenge of limited training data. The model achieved an F1 score above 0.8 in all planting settings.

This paper examines the potential of using few-shot learning and computer vision techniques for detecting, identifying, and counting agricultural pests and diseases in images. A …

The temporal and spatial irregularity of plant diseases results in insufficient image data for certain diseases, challenging traditional deep learning methods that rely on large amounts of manually annotated data for training. Few-shot learning has emerged as an effective solution to this problem. This paper proposes a method based on the Feature Adaptation Score (FAS) metric, which calculates the FAS for each feature layer in the Swin-TransformerV2 structure. By leveraging the strict positive correlation between FAS scores and test accuracy, we can identify the Swin-Transformer V2-F6 network structure suitable for few-shot plant disease classification without training the network. Furthermore, based on this network structure, we designed the Plant Disease Feature Calibration (PDFC) algorithm, which uses extracted features from the PlantVillage dataset to calibrate features from other datasets. Experiments demonstrate that the combination of the Swin-Transformer V2F6 network structure and the PDFC algorithm significantly improves the accuracy of few-shot plant disease classification, surpassing existing state-of-the-art models. Our research provides an efficient and accurate solution for few-shot plant disease classification, offering significant practical value.

… This study is the first in archaeobotany to apply and evaluate deep learning techniques for the automatic identification and species-level classification of carbonized millet seeds. A …

In response to the need for precise blossom identification and optimization of key operational parameters in intelligent cherry spraying pollination, the SMD-YOLO (You Only Look Once with spatial and channel reconstruction convolution, multi-scale channel attention, and dual convolution modules) cherry blossom detection model is proposed, along with a pollination experiment platform for parameter optimization. The SMD-YOLO model, built upon YOLOv11, enhances feature extraction through the ScConvC3k2 (spatial and channel reconstruction convolution C3k2) module, incorporates the MSCA (multi-scale channel attention) attention mechanism, and employs the DualConv module for a lightweight design, ensuring both detection accuracy and operational efficiency. Tested on a self-constructed cherry blossom dataset, the model delivered a precision of 87.6%, a recall rate of 86.1%, and an mAP (mean average precision) reaching 93.1% with a compact size of 4765 KB, 2.28 × 106 parameters, a computational cost of 5.8 G, and a detection speed of 75.76 FPS, demonstrating strong practicality and potential for embedded real-time detection in edge devices, such as cherry pollination robots. To further enhance pollination effectiveness, a dedicated pollination experiment bench was designed, and a second-order orthogonal rotational combination experiment method was employed to systematically optimize three key parameters: spraying distance, spraying time, and liquid flow rate. Experimental results indicate that the optimal pollination effect occurs when the spraying distance is 3.4 cm, spraying time is 1.9 s, and liquid flow rate is 339 mL/min, with a deposition amount of 0.18 g and a coverage rate of 97.25%. This study provides a high-precision image detection algorithm and operational parameter optimization basis for intelligent and precise cherry blossom pollination.

… Total processing speed of multi-class flower detection and its distribution identification was 112.46 ms per image including 15.50 ms for image detection by YOLOv5l. Results showed …