Overall Accuracy Research Articles

A deep learning approach for deriving wheat phenology from near-surface RGB image series using spatiotemporal fusion

Accurate monitoring of wheat phenological stages is essential for effective crop management and informed agricultural decision-making. Traditional methods often rely on labour-intensive field surveys, which are prone to subjective bias and limited temporal resolution. To address these challenges, this study explores the potential of near-surface cameras combined with an advanced deep-learning approach to derive wheat phenological stages from high-quality, real-time RGB image series. Three deep learning models based on three different spatiotemporal feature fusion methods, namely sequential fusion, synchronous fusion, and parallel fusion, were constructed and evaluated for deriving wheat phenological stages with these near-surface RGB image series. Moreover, the impact of different image resolutions, capture perspectives, and model training strategies on the performance of deep learning models was also investigated. The results indicate that the model using the sequential fusion method is optimal, with an overall accuracy (OA) of 0.935, a mean absolute error (MAE) of 0.069, F1-score (F1) of 0.936, and kappa coefficients (Kappa) of 0.924 in wheat phenological stages. Besides, the enhanced image resolution of 512 × 512 pixels and a suitable image capture perspective, specifically a sensor viewing angle of 40° to 60° vertically, introduce more effective features for phenological stage detection, thereby enhancing the model’s accuracy. Furthermore, concerning the model training, applying a two-step fine-tuning strategy will also enhance the model’s robustness to random variations in perspective. This research introduces an innovative approach for real-time phenological stage detection and provides a solid foundation for precision agriculture. By accurately deriving critical phenological stages, the methodology developed in this study supports the optimization of crop management practices, which may result in improved resource efficiency and sustainability across diverse agricultural settings. The implications of this work extend beyond wheat, offering a scalable solution that can be adapted to monitor other crops, thereby contributing to more efficient and sustainable agricultural systems.

Moderate Red-Edge vegetation index for High-Resolution multispectral remote sensing images in urban areas

Urban green space (UGS) monitoring is significant in optimizing urban planning, protecting the ecological environment, and improving residents’ quality of life. However, in urban environments, shadow interference and the emergence of new construction materials pose challenges to monitoring green vegetation in high-resolution imagery. This study found that existing vegetation indices (VIs), such as normalized difference vegetation index (NDVI) and enhanced vegetation index (EVI), perform inadequately in extracting vegetation in urban areas, resulting in significant omissions and errors. By conducting in-depth analysis and quantitative experiments on the reflectance of typical urban ground objects, this study developed a new vegetation extraction method, the moderate red-edge vegetation index (MREVI), to enhance the extraction accuracy of UGS vegetation from unmanned aerial vehicle (UAV) high-resolution multispectral remote sensing (RS) images. Experimental results demonstrate that MREVI performs exceptionally well in complex urban environments, significantly suppressing non-vegetation areas, achieving an overall accuracy (OA) of 98.6% and a Kappa coefficient of 0.97. This study supports for urban planning, UGS monitoring, and the evaluation of urban plant carbon sequestration capacity.

Classification of hyperspectral and LiDAR data by transformer-based enhancement

ABSTRACT The integration of multi-modal data allows for a more accurate representation of the ground characteristics. For a comprehensive interpretation of remote sensing data, existing multi-modal data fusion research mainly focuses on the joint utilization of 3D Light Detection and Ranging (LiDAR) and 2D Hyperspectral Image (HSI) data. However, existing algorithms do not pay much attention to the interaction of high-level semantic information between different modal data before fusion. This paper proposes a novel multi-modal data fusion deep learning network with the Cross-Modal Self-Attentive Feature Fusion Transformer (SAFFT). The framework employs a multi-head self-attention layer to fuse various attention information from multiple heads, effectively enhancing advanced feature information from different modalities for comprehensive integration. Experimental results on the Houston 2013 dataset demonstrate the effectiveness of the proposed method, which achieves an overall accuracy (OA) of 94.3757% in classifying 15 semantic classes.

Intelligent classification of water bodies with different turbidity levels based on Gaofen-1 multispectral imagery

The turbidity of water is crucial for the health of river and lake ecosystems, necessitating efficient monitoring for effective water management. Existing methods for studying water turbidity's spatial and temporal distribution rely mostly on measured data. There is limited research on the classification of water bodies with different turbidity levels. The main challenge lies in determining the boundaries of liquid water bodies at various turbidity levels, making it challenging to classify them accurately using traditional remote sensing image classification methods. This paper proposes and validates an intelligent turbidity classification method based on deep learning using GaoFen-1 multispectral remote sensing imagery. An adaptive threshold water extraction method based on the Normalized Difference Water Index is proposed to capture water boundaries more accurately to improve the accuracy of extracting nearshore water bodies. A semi-automatic semantic annotation method for water turbidity is introduced to reduce manual labeling costs. The paper applies mode filtering to address edge noise issues and establishes a high-quality training sample dataset. After comparing the accuracy of various neural network models, DeepLab V3+ is selected for intelligent turbidity classification. The results show high accuracy, with mean intersection over union (MIoU), mean F1 score (MF1), and overall accuracy (OA) reaching 94.73%, 97.29%, and 97.54%, respectively. The proposed method and experiments demonstrate the feasibility of intelligent classification of water bodies with different turbidity levels using deep learning networks. This provides a new approach for large-scale and efficient remote sensing water turbidity monitoring.

Advancing Skarn Iron Ore Detection through Multispectral Image Fusion and 3D Convolutional Neural Networks (3D-CNNs)

This study explores novel techniques to improve the detection accuracy of skarn iron deposits using advanced image-processing methodologies. Leveraging the capabilities of ASTER image, band ratio (BR) images, and principal component analysis (PCA) alongside the power of 3D convolutional neural networks (3D-CNNs), the research aims to enhance the precision and efficiency of ore detection in complex geological environments. The proposed method employs a specific 3D-CNN architecture accepting input as a 7 × 7 × C image patch, where C represents the combined number of selected ASTER image bands, principal component (PC) bands, and computed BR images. To evaluate the accuracy of the proposed method, five distinct image band combinations, including the proposed band combination, were tested and evaluated based on the overall accuracy (OA), average accuracy (AA), and kappa coefficient. The results demonstrated that while the incorporation of BR images alongside ASTER bands initially seemed promising, it introduced significant confusion in certain classifications, leading to unexpected misclassification rates. Surprisingly, utilizing solely ASTER bands as input parameters yielded higher accuracy rates (OA = 93.13%, AA = 91.96%, kappa = 90.91%) compared with scenarios involving the integration with band ratios (OA = 87.02%, AA = 79.15, kappa = 82.60%) or the integration of BR images to PC bands (OA = 87.78%, AA = 82.39%, kappa = 83.81%). However, the amalgamation of ASTER bands with selected PC bands showed slight improvements in accuracy (OA = 94.65%, AA = 92.93%, kappa = 93.45%), although challenges in accurately classifying certain features persisted. Ultimately, the proposed combination of ASTER bands, PC bands, and BR images (proposed band combination) presented the most visually appealing and statistically accurate results (OA = 96.95%, AA = 94.87%, kappa = 95.93%), effectively addressing misclassifications observed in the other combinations. These findings underscore the synergistic contributions of each of the ASTER bands, PC bands, and BR images, with the ASTER bands proving pivotal for optimal skarn classification, the PC bands enhancing intrusions classification accuracy, and the BR images strengthening wall rock classification accuracy. In conclusion, the proposed combination of input image bands emerges as a robust and comprehensive methodology, demonstrating unparalleled accuracy in the remote sensing detection of skarn iron minerals.

A novel red-edge vegetable index for paddy rice mapping based on Sentinel-1/2 and GF-6 images

ABSTRACT Accurate paddy rice mapping is crucial for ensuring food security and guiding agricultural production. Vegetation indices are extensively employed to map paddy rice. However, most traditional normalized indices tend to be oversaturated during periods of lush vegetation due to normalization errors, resulting in uncertainties in paddy rice mapping. To address this issue, we introduce a novel red-edge rice index (RERI) in this study; this index comprises information from red, near-infrared, and red-edge bands without normalization. To extract single- and double-cropping rice features from potential rice areas, we employ GF-6 and Sentinel-2 images based on the proposed RERI and the random forest algorithm. The proposed method is validated in the Dingcheng District of Changde city, China, and the results are compared with those based on three normalized vegetation indices. The results show that the RERI yielded the highest levels of the accuracy for all the metrics, achieving an overall accuracy (OA) of 92.50% and a kappa coefficient of 0.8875. The RERI exhibited F1 scores of 92.26% for single-cropping rice, 93.00% for double-cropping rice, and 92.28% for non-rice areas. Our results highlight the potential of using the RERI for rice identification, and the effectiveness of our method for rice extraction is demonstrated.

An Optimized Semi-Supervised Generative Adversarial Network Rice Extraction Method Based on Time-Series Sentinel Images

In response to the limitations of meteorological conditions in global rice growing areas and the high cost of annotating samples, this paper combines the Vertical-Vertical (VV) polarization and Vertical-Horizontal (VH) polarization backscatter features extracted from Sentinel-1 synthetic aperture radar (SAR) images and the NDVI, NDWI, and NDSI spectral index features extracted from Sentinel-2 multispectral images. By leveraging the advantages of an optimized Semi-Supervised Generative Adversarial Network (optimized SSGAN) in combining supervised learning and semi-supervised learning, rice extraction can be achieved with fewer annotated image samples. Within the optimized SSGAN framework, we introduce a focal-adversarial loss function to enhance the learning process for challenging samples; the generator module employs the Deeplabv3+ architecture, utilizing a Wide-ResNet network as its backbone while incorporating dropout layers and dilated convolutions to improve the receptive field and operational efficiency. Experimental results indicate that the optimized SSGAN, particularly when utilizing a 3/4 labeled sample ratio, significantly improves rice extraction accuracy, leading to a 5.39% increase in Mean Intersection over Union (MIoU) and a 2.05% increase in Overall Accuracy (OA) compared to the highest accuracy achieved before optimization. Moreover, the integration of SAR and multispectral data results in an OA of 93.29% and an MIoU of 82.10%, surpassing the performance of single-source data. These findings provide valuable insights for the extraction of rice information in global rice-growing regions.

Spatio-temporal assessment of land use and land cover dynamics in Urmia lake basin of Iran: A bi-directional approach using optical and radar data on the Google Earth Engine platform

Sustainable development, a cornerstone in responsible land resource management, occupies a critical role in global efforts to balance economic progress with ecological preservation. This research explores the intricate dynamics of sustainable development and land resource management in the Lake Urmia watershed, northwestern Iran, with a specific focus on Land Use and Land Cover (LULC) alterations. Leveraging cutting-edge methodologies within the Google Earth Engine (GEE) framework, the study employs the Random Forest (RF) model, and multi-sensor satellite data to achieve meticulous LULC classification from the years 2016 to 2020.This strategy incorporates satellite images from Sentinel-1 (S1), Sentinel-2 (S2), coupled with ancillary data, and the implementation of the RF algorithm. It highlighted the intricate synergy between optical and Synthetic Aperture Radar (SAR) data for LULC classification in the Lake Urmia watershed. The Overall Accuracy (OA) values ranging from 88.28 % to 92.00 % affirmed the robustness of the RF model, demonstrating its potential for research objectives. In addition, Dynamic World, a near real-time global dataset, was applied to the study area, provides a convenient view to detect changes from the years 2016 to 2023, while enabling comparison with the widely used RF algorithm in past studies. Notably, this study goes beyond methodological developments to examine the concrete consequences of the observed changes. The decline in water bodies within the Urmia Lake basin and the parallel expansion of cultivated lands underscore the profound impact of human activities, particularly agricultural practices, on the ecological equilibrium. As we navigate the intricate terrain of sustainable development, this research serves as a clarion call for remote sensing approaches to monitor and manage land resources, emphasizing the delicate balance between economic growth and ecological preservation.

Evaluation of machine learning algorithm capability for Bosten Lake Wetland classification based on multi-temporal Sentinel-2 data

As crucial carbon sinks within terrestrial ecosystems, wetlands have been extensively studied in terms of spatio-temporal distributions. However, existing methods for classifying wetlands are of limited accuracy, and it is difficult to acquire consistent samples over time. Therefore, precise classification methods are required to facilitate wetland conservation and ecological restoration. In this study, multiple machine learning (ML) algorithms in combination with feature sets based on Sentinel-2 data were used to accurately classify the land-use types (LUTs) of the Bosten Lake Wetland (BLW) in Xinjiang, China. The enhanced water index (EWI), modified normalised difference water index (MNDWI), and normalised difference water index (NDWI) were selected to extract water information and distinguish water bodies from land surfaces in the BLW. Three classification plans based on vegetation indices, water indices, and textural features were developed using artificial neural network (ANN), support vector machine (SVM), random forest (RF) algorithms. Plan 9 combined vegetation water and texture with the highest overall accuracy (OA) 91.02 % and kappa coefficient (KC) 0.89. This plan obtained a producer accuracy of over 90 % for lake wetlands, river wetlands, grassland wetlands, mud flats, and farmland and > 83 % for construction land and bareland. According to Plan 9, the wetland area during 2018–2023 showed noticeable seasonal fluctuations but stable interannual changes. Conversely, non-wetland areas demonstrated significant interannual fluctuations, particularly in bareland and farmland, which might have been influenced by urbanisation and ecological policies. This study provides insights into the data sources, feature selection, and methodological approaches for wetland information extraction in arid regions.

An Improved LandTrendr Algorithm for Forest Disturbance Detection Using Optimized Temporal Trajectories of the Spectrum: A Case Study in Yunnan Province, China

Forest disturbance mapping plays an important role in furthering our understanding of forest dynamics. The Landsat-based detection of Trends in Disturbance and Recovery (LandTrendr) algorithm is widely used in forest disturbance mapping. However, it neglects the quality of the temporal trajectory and its change trends for forest disturbance mapping. Therefore, the aim of this paper is to improve LandTrendr (iLandTrendr) for forest disturbance mapping by optimizing its temporal trajectories and the post-processing of detection results. Specifically, the temporal trajectory of complex forest disturbance types was optimized using the Savitzky–Golay (SG) filter with constraints. That is, the smooth value generated from the SG filter for the disturbance year was replaced by the satellite observations when the nonlinear abrupt signals were included in the multi-temporal data. The forest disturbance detected by LandTrendr was further modified using the consistency of spectral variation trends. A case study using iLandTrendr to detect forest disturbance in Yunnan Province was conducted. Compared to the LandTrendr method, which has an overall accuracy (OA) of 35.88%, iLandTrendr generated forest disturbance mapping with an OA of 89.32%, which was significantly higher. The total mapped area of disturbance was 1,985,820.9 km2, accounting for 49.69% of the total area. The disturbances were predominately caused by natural factors, such as wildfires, pests and diseases, and forest degradation, accounting for 85.31% of the total disturbed area. iLandTrendr can quickly and accurately detect the occurrence year of complex forest disturbance types and can be extended for the forest disturbance mapping of a large area.

Exploring long term Impervious Surface Areas (ISA) dynamics using Landsat imagery, Μachine Learning and GEE: The case of Attica, Greece

Accurate data on Impervious Surface Areas (ISA) are essential for various studies concerning urban environments, as the constant proliferation of these surfaces is a noticeable result of urbanization, especially in metropolitan cities. The present study proposes a methodology approach in performing a long-term mapping of ISA changes in Attica Prefecture, Greece, from 1984 to 2022, exploiting the Landsat archive and contemporary machine learning (ML) methods of geospatial data processing, namely Support Vector Machines (SVM) and Random Forests (RF). Using Google Earth Engine cloud platform, the SVM and RF classifiers are developed and implemented for four single dates (in years 1984, 1999, 2013 and 2022). Accuracy assessment of the classification maps was based on the computation of a series of statistical metrics based on the confusion matrix, ans the McNemar's chi-square test which was used to evaluate the statistical significance of the difference in the classification maps, derived from SVM and RF classifiers. Both SVM and RF provided very accurate results, with Overall Accuracy (OA) higher than 90% and kappa coefficient (Kappa) higher than 0.8 for all classification maps, with SVM performing better in 1984 and 2022 and RF outperforming SVM in 2013. In addition, the McNemar's test confirmed the statistical significance of the research findings reported herein. Change detection results, highlighted the wide sprawl of the urban fabric, especially in sub-urban areas, surrounding the metropolitan center of Athens. The employed methodology represents a significant advancement in the application of GEE, beyond their general use, by integrating cutting-edge ML techniques with available remote sensing data to create an automated analysis process. This innovative fusion not only enhances the precision and efficiency of ISA mapping but also establishes the basis for a pioneering standard in the field by harnessing the power of advanced technologies and accessible data sources.

Enhancing hydrological model efficiency through satellite image classification

ABSTRACT This paper aims to evaluate the performance of a hydrological model by using satellite image classification (SIM) to extract Land Use (LU) information. Four methods, namely Naive Bayes (NB), Classification and Regression Trees (CART), Support Vector Machine (SVM), and Random Forest (RF), are assessed for SIM using satellite images. CART demonstrated the highest Overall Accuracy (OA) of 0.996, followed by RF (OA=0.994), SVM (OA=0.797), and NB (OA=0.543). The Soil and Water Assessment Tool (SWAT) model was employed in this study to construct and refine the hydrological model for simulating the streamflow in the basin. For this purpose, the ground-based layer was replaced by LU generated from the classification algorithms. Integrating the four classification approaches yielded a significant improvement in Nash-Sutcliffe efficiency of the SWAT model, increasing it from 0.37 to approximately 0.76. These findings highlight the effectiveness of using satellite image classification in enhancing the efficiency of hydrological models.

Assessment of Machine Learning Algorithms for Land Cover Classification in a Complex Mountainous Landscape

Mapping land cover (LC) in mountainous regions, such as the Gilgit-Baltistan (GB) area of Pakistan, presents significant challenges due to complex terrain, limited data availability, and accessibility constraints. This study addresses these challenges by developing a robust, data-driven approach to classify LC using high-resolution Sentinel-2 (S-2) satellite imagery from 2019 within Google Earth Engine (GEE). The research evaluated the performance of various machine learning (ML) algorithms, including classification and regression tree (CART), maximum entropy (gmoMaxEnt), minimum distance (minDistance), support vector machine (SVM), and random forest (RF), without extensive hyperparameter tuning. Additionally, ten different scenarios based on various band combinations of S-2 data were used as input for running the ML models. The LC classification was performed using 2759 sample points, with 70% for training and 30% for validation. The results indicate that the RF algorithm outperformed all other classifiers under scenario S1 (using 10 bands), achieving an overall accuracy (OA) of 0.79 and a kappa coefficient of 0.76. The final RF-based LC mapping shows the following percentage distribution: barren land (46.7%), snow cover (22.9%), glacier (7.9%), grasses (7.2%), water (4.7%), wetland (2.9%), built-up (2.7%), agriculture (1.9%), and forest (1.2%). It is suggested that the best identified RF classifier within the GEE environment should be used for advanced multi-source data image classification with hyperparameter tuning to increase OA. Additionally, it is suggested to build the capacity of various stakeholders in GB for better monitoring of LC changes and resource management using geospatial big data.

A lightweight CNN-Transformer network for pixel-based crop mapping using time-series Sentinel-2 imagery

Deep learning approaches have provided state-of-the-art performance in crop mapping. Recently, several studies have combined the strengths of two dominant deep learning architectures, Convolutional Neural Networks (CNNs) and Transformers, to classify crops using remote sensing images. Despite their success, many of these models utilize patch-based methods that require extensive data labeling, as each sample contains multiple pixels with corresponding labels. This leads to higher costs in data preparation and processing. Moreover, previous methods rarely considered the impact of missing values caused by clouds and no-observations in remote sensing data. Therefore, this study proposes a lightweight multi-stage CNN-Transformer network (MCTNet) for pixel-based crop mapping using time-series Sentinel-2 imagery. MCTNet consists of several successive modules, each containing a CNN sub-module and a Transformer sub-module to extract important features from the images, respectively. An attention-based learnable positional encoding (ALPE) module is designed in the Transformer sub-module to capture the complex temporal relations in the time-series data with different missing rates. Arkansas and California in the U.S. are selected to evaluate the model. Experimental results show that the MCTNet has a lightweight advantage with the fewest parameters and memory usage while achieving the superior performance compared to eight advanced models. Specifically, MCTNet obtained an overall accuracy (OA) of 0.968, a kappa coefficient (Kappa) of 0.951, and a macro-averaged F1 score (F1) of 0.933 in Arkansas, and an OA of 0.852, a Kappa of 0.806, and an F1 score of 0.829 in California. The results highlight the importance of each component of the model, particularly the ALPE module, which enhanced the Kappa of MCTNet by 4.2% in Arkansas and improved the model’s robustness to missing values in remote sensing data. Additionally, visualization results demonstrated that the features extracted from CNN and Transformer sub-modules are complementary, explaining the effectiveness of the MCTNet.

A new transferable deep learning approach for crop mapping

ABSTRACT Accurate crop mapping provides important information for government decision-making and agricultural management. Recent advancements in deep learning have significantly enhanced the capabilities of remote sensing-based crop mapping. In this study, we developed a new deep learning approach, DSH, which comprises three modules, DeepLabV3+ (D), channel self-attention (S), and histogram matching (H). The DeepLabV3+ module learns the spatial distribution of crops in the spatial and spectral dimensions. The channel self-attention module was used to enhance the weights of the important features. The histogram-matching module addresses the domain gap between the training and testing images and enhances the transferability of the DSH. In addition, the input data of the DSH are monthly synthesized Sentinel-2 data, thus relaxing the data collection requirements. The performance of the DSH was evaluated and benchmarked against HRNetV2, DeepLabV3+, and random forest (RF) at eight sites in the U.S. China, and France with different farming structures, climate conditions, and landscape complexities. Temporal transfer revealed that the classification performance of DSH improved as the growth season progressed, peaked in the peak growth season (August), and then declined at each study site. The DSH model trained during the peak growth season demonstrated superior temporal transferability. Spatial transfer experiments demonstrated that transferring DSH to sites with similar planting structures achieved higher precision than full spatial transfers. Spatiotemporal transfer experiments conducted at eight sites over three years (2020–2022) demonstrated an average overall accuracy (OA) of 88.1% for DSH, highlighting the importance of similar planting structures for effective spatiotemporal transfer. DSH outperformed HRNetV2, DeepLabV3+, and RF in OA, producer’s accuracy, and user’s accuracy at all eight study sites. Notably, DSH exhibited comparable performance during both the peak and full growth seasons, with an average OA of 93.9%. The classification accuracy of test images after histogram matching remained stable from 2020 to 2022, whereas the accuracy of raw test images fluctuated annually. This indicates that histogram matching is effective in improving the transferability of DSH. Embedding the channel self-attention module into the position of the high-level features had a more significant impact on improving DSH’s performance than other configurations. Visualizing features from the channel self-attention and high-level layers revealed that integrating the channel self-attention module with DeepLabV3+‘s high-level features significantly improved crop mapping by enhancing relevant semantic information. Overall, DSH exhibits robust spatiotemporal generalizability and requires minimal input data, making it suitable for large-scale crop mapping.

MLF-PointNet++: A Multifeature-Assisted and Multilayer Fused Neural Network for LiDAR-UAS Point Cloud Classification in Estuarine Areas

LiDAR-unmanned aerial system (LiDAR-UAS) technology can accurately and efficiently obtain detailed and accurate three-dimensional spatial information of objects. The classification of objects in estuarine areas is highly important for management, planning, and ecosystem protection. Owing to the presence of slopes in estuarine areas, distinguishing between dense vegetation (lawns and trees) on slopes and the ground at the tops of slopes is difficult. In addition, the imbalance in the number of point clouds also poses a challenge for accurate classification directly from point cloud data. A multifeature-assisted and multilayer fused neural network (MLF-PointNet++) is proposed for LiDAR-UAS point cloud classification in estuarine areas. First, the 3D shape features that characterize the geometric characteristics of targets and the visible-band difference vegetation index (VDVI) that can characterize vegetation distribution are used as auxiliary features to enhance the distinguishability of dense vegetation (lawns and trees) on slopes and the ground at the tops of slopes. Second, to enhance the extraction of target spatial information and contextual relationships, the feature vectors output by different layers of set abstraction in the PointNet++ model are fused to form a combined feature vector that integrates low and high-level information. Finally, the focal loss function is adopted as the loss function in the MLF-PointNet++ model to reduce the effect of imbalance in the number of point clouds in each category on the classification accuracy. A classification evaluation was conducted using LiDAR-UAS data from the Moshui River estuarine area in Qingdao, China. The experimental results revealed that MLF-PointNet++ had an overall accuracy (OA), mean intersection over union (mIOU), kappa coefficient, precision, recall, and F1-score of 0.976, 0.913, 0.960, 0.953, 0.953, and 0.953, respectively, for object classification in the three representative areas, which were better than the corresponding values for the classification methods of random forest, BP neural network, Naive Bayes, PointNet, PointNet++, and RandLA-Net. The study results provide effective methodological support for the classification of objects in estuarine areas and offer a scientific basis for the sustainable development of these areas.

A transfer learning-based network model integrating kernel convolution with graph attention mechanism for point cloud segmentation of livestock

Non-contact body size measurement has become a hot research topic in intelligent livestock farming. In regard to body size measurement of large livestock, such as cattle, collecting a substantial number of point clouds is frequently involved. The direct calculation of all point clouds for body size measurement can be impacted as point clouds of different body parts may interfere with each other, which poses huge challenges for the positioning of key points and induces inaccurate positioning, resulting in measurement errors. The accuracy of body size measurement can be improved by segmenting point clouds of different body parts from each other, key measurement points can be precisely located, thus enhancing the accuracy of body size measurement. In this paper, we propose a network model initially trained for pig point cloud segmentation based on the Kernel Convolution integrated with Graph Attention Mechanism (KCGATNet for short), which, through transfer learning techniques, can also be used to achieve successful segmentation of various cattle point clouds using only 7 training samples. The model utilizes two core modules, Kernel Convolution (KC) and Point-based Graph Attention Mechanism (P-GAT), to extract local neighborhood features of point clouds. When using pig body point clouds as training data, it achieved precise segmentation of the head, ears, limbs, torso, and tail of pigs through a downsampling-upsampling architecture. On the test set of pig point clouds, Overall Accuracy (OA) reached 98.1% and mean Intersection over Union (mIoU) was up to 90.5%. Furthermore, when this model served as a pre-trained model and underwent transfer learning using 7 sets of annotated data of Simmental cattle, it achieved a mIoU of 90.1% on a test set of 93 Simmental cattle, 89.6% on a test set of 439 dairy buffalo, 90.2% on a test set of 103 Hereford cattle, and 90.0% on a test set of 119 Black Angus cattle. The experimental outcomes fully demonstrate the robustness of the proposed livestock point cloud segmentation model, KCGATNet. With transfer learning of a small sample size, it can reliably perform point cloud segmentation on a wide range of different breeds of quadrupedal livestock, saving a significant amount of time spent on manual annotation and improving the efficiency of livestock point cloud segmentation models.

Extraction of River Water Bodies Based on ICESat-2 Photon Classification

The accurate extraction of river water bodies is crucial for the utilization of water resources and understanding climate patterns. Compared with traditional methods of extracting rivers using remote sensing imagery, the launch of satellite-based photon-counting LiDAR (ICESat-2) provides a novel approach for river water body extraction. The use of ICESat-2 ATL03 photon data for inland river water body extraction is relatively underexplored and thus warrants investigation. To extract inland river water bodies accurately, this study proposes a method based on the spatial distribution of ATL03 photon data and the elevation variation characteristics of inland river water bodies. The proposed method first applies low-pass filtering to denoised photon data to mitigate the impact of high-frequency signals on data processing. Then, the elevation’s standard deviation of the low-pass-filtered data is calculated via a sliding window, and the photon data are classified on the basis of the standard deviation threshold obtained through Gaussian kernel density estimation. The results revealed that the average overall accuracy (OA) and Kappa coefficient (KC) for the extraction of inland river water bodies across the four study areas were 99.12% and 97.81%, respectively. Compared with the improved RANSAC algorithm and the combined RANSAC and DBSCAN algorithms, the average OA of the proposed method improved by 17.98% and 7.12%, respectively, and the average KC improved by 58.38% and 17.69%, respectively. This study provides a new method for extracting inland river water bodies.

BCLNet: Boundary contrastive learning with gated attention feature fusion and multi-branch spatial-channel reconstruction for land use classification

The fusion of optical and synthetic aperture radar (SAR) images is not only a crucial method for enhancing land use classification tasks but also forms a fundamental basis for the interpretation of multimodal remote sensing imagery. However, the visual differences caused by the distinct nonlinear radiometric properties of optical and SAR images lead to challenges in effectively integrating multimodal features using current land use classification methods, resulting in lower classification accuracy. On one hand, these approaches may be affected by the heterogeneity among different source images, leading to an inability to fully utilize the semantic content of the fused features. On the other hand, the semantic embedding space does not consider the semantic relationships between different pixels from a global perspective (i.e., across the entire dataset). To address the aforementioned challenges, we propose the boundary contrastive learning classification network (BCLNet). In this framework, we introduce the gate attention fusion (GAF) module, which selectively considers distinctiveness and consistency features in optical and SAR imagery, facilitating effective modal representation fusion. Furthermore, we introduce the multi-branch spatial channel reconstruction (MSCR) module for enhanced feature augmentation, operating in both spatial and channel dimensions with a dual-dimensional feature reconstruction selection mechanism for the fused features. Furthermore, we adopt boundary contrast learning to address the global pixel semantic space embedding problem. Our method is evaluated on the WHU-OPT-SAR dataset and a multi-class scene classification dataset (MSD) that we constructed. Compared to state-of-the-art methods, our approach demonstrates improvements in pixel classification overall accuracy (OA) and mean intersection over union (mIOU).

PVF-10: A high-resolution unmanned aerial vehicle thermal infrared image dataset for fine-grained photovoltaic fault classification

Accurate identification of faulty photovoltaic (PV) modules is crucial for the effective operation and maintenance of PV systems. Deep learning (DL) algorithms exhibit promising potential for classifying PV fault (PVF) from thermal infrared (TIR) images captured by unmanned aerial vehicle (UAV), contingent upon the availability of extensive and high-quality labeled data. However, existing TIR PVF datasets are limited by low image resolution and incomplete coverage of fault types. This study proposes a high-resolution TIR PVF dataset with 10 classes, named PVF-10, comprising 5579 cropped images of PV panels collected from 8 PV power plants. These classes are further categorized into two groups according to the repairability of PVF, with 5 repairable and 5 irreparable classes each. Additionally, the circuit mechanisms underlying the TIR image features of typical PVF types are analyzed, supported by high-resolution images, thereby providing comprehensive information for PV operators. Finally, five state-of-the-art DL algorithms are trained and validated based on the PVF-10 dataset using three levels of resampling strategy. The results show that the overall accuracy (OA) of these algorithms exceeds 83%, with the highest OA reaching 93.32%. Moreover, the preprocessing procedure involving resampling and padding strategies are beneficial for improving PVF classification accuracy using PVF-10 datasets. The developed PVF-10 dataset is expected to stimulate further research and innovation in PVF classification.