Jinze Yang;Youming Wu;Zining Zhu;Wenchao Zhao;Wenhui Diao;Kun Fu
{"title":"GFA: Generalized Feature Affinity for Semantic Segmentation in Resolution-Degraded Images","authors":"Jinze Yang;Youming Wu;Zining Zhu;Wenchao Zhao;Wenhui Diao;Kun Fu","doi":"10.1109/JSTARS.2025.3552117","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3552117","url":null,"abstract":"In the remote sensing field, resolution-degraded images are often captured under poor imaging conditions, leading to missing texture information for segmentation. To tackle these challenges, a parallel super-resolution branch can be incorporated into the segmentation pipeline to restore detailed features through a feature supplementation operation. However, existing methods based on the structure are mainly limited by the direct supplementation operation between the two tasks with different feature distributions, and the original semantic distribution may be damaged. Therefore, a novel learning method called generalized feature affinity (GFA) is proposed. It is expected to realize more comprehensive feature supplementation in a consistent feature distribution space between the two tasks. Specifically, a global distribution affinity module is developed to encourage uniformity in the intraclass distribution between restored and semantic features. Then, a local texture affinity module is designed to better transfer detailed information by enhancing the texture prominence in feature maps and exploring supplementation in spatial and channel aspects. To guide the supplementation direction more effectively, a results-oriented hard sampling strategy, which integrates auxiliary spatial attention is further proposed to enhance the performance of the aforementioned modules. Extensive experiments are conducted on the widely recognized benchmarks ISPRS Potsdam and Vaihingen, and LoveDA datasets. The proposed method results in an improvement of about 2% mIoU across many commonly used segmentation models, demonstrating the effectiveness and generalization of GFA.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8797-8812"},"PeriodicalIF":4.7,"publicationDate":"2025-03-17","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10930542","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143786313","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Local-Descriptors-Based Rectification Network for Few-Shot Remote Sensing Scene Classification","authors":"Anyong Qin;Bin Luo;Qiang Li;Cuiming Zou;Yu Zhao;Tiecheng Song;Chenqiang Gao","doi":"10.1109/JSTARS.2025.3551599","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3551599","url":null,"abstract":"Few-shot remote sensing scene classification has become a study that has attracted widespread attention and aims to identify new scene classes through one or a few labeled scene images. Nevertheless, due to the existence of unrelated complex background in scene images, local descriptors (LDs) that offer a more efficient representation than image-level features, will carry semantic information unrelated to the real semantics of the scene images. Concurrently, these irrelevant background LDs are also causing a large distribution bias in support and query sets, which leads to the problem of inaccurate feature representation of scene images. To address the aforementioned problems, in this article, we introduce an LD-based rectification network called LDRNet. Within this network, we first design an LD semantic rectification module. It performs semantic rectification on LDs that are unrelated to scene image semantics by obtaining a descriptor-level global-aware semantic representation. Second, we introduce a cross-set bias rectification module. It rectifies the query set by obtaining the offset between two sets (query and support) from a more detailed LD perspective. This operation can shorten the distance among the two sets (query and support), thereby obtaining a more accurate representation of scene image features. Furthermore, we employ an LD-based contrastive loss function to guarantee that the rectified LD semantics are consistent with the corresponding scene image. The comparative experimental result indicates that our LDRNet achieves state-of-the-art performance on three commonly used public datasets.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"9566-9581"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925636","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143835368","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jingming Wang;Chang-Qing Ke;Yu Cai;Jianwan Ji;Zhenqing Wang
{"title":"A Novel Convolutional Neural Network for the Extraction of Algal Bloom and Aquatic Vegetation in Typical Eutrophic Shallow Lakes","authors":"Jingming Wang;Chang-Qing Ke;Yu Cai;Jianwan Ji;Zhenqing Wang","doi":"10.1109/JSTARS.2025.3548589","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3548589","url":null,"abstract":"Under the hybrid impact of regional climate change and extensive human activities, lake eutrophication has become an increasingly serious problem, which causes a dramatic reduction in the area of aquatic vegetation (AV) and poses huge challenges to the balance of regional lake ecosystems. As an important freshwater resource, shallow lakes play an important role in balancing water resources, adjusting regional climate, and retaining clean water supply. However, in view of the complexity and variability of shallow lake environment, especially the similarity of spectral characteristics between algal bloom (AB) and AV in shallow lakes, the extraction results of AB and AV using most algorithms are not satisfactory. In response to these problems, this study utilized Landsat images as the dataset to accurately differentiate AB and AV by developing a new extraction network (AAENet) aiming at eutrophic shallow lakes. Next, the AAENet model was compared with three classic semantic segmentation models (UNet, Deeplab v3, and PSPNet) and the vegetation and bloom indices algorithm. Finally, the spatiotemporal distribution and area change in typical shallow lakes were analyzed based on the extraction results of the AAENet model. The results showed that: 1) the AAENet model achieved the highest accuracy in distinguishing AB and AV, with an overall accuracy of 87.85%, an F1 score of 0.85, and a Frequency Weighted Intersection-over-Union of 0.76 in the testing lakes. 2) In terms of improving the performance of the AAENet model, the ConvNeXt encoder played the most significant role. 3) During 2013–2023, the area of AB in Chaohu Lake and Taihu Lake decreased annually by 0.73 km<sup>2</sup> and 3.29 km<sup>2</sup>, respectively. In particular, the area of AV in Chaohu Lake steadily increased at a rate of 0.27 km<sup>2</sup>/year, whereas the area of AV in Taihu Lake exhibited an initial decline followed by an increase. This study can provide important technical support for monitoring the dynamics of AB and AV in lakes.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8099-8111"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925549","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143706704","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Adaptive Gap-Filling of Multispectral Images at Coarse and Fine Spatial Resolution","authors":"Seyedkarim Afsharipour;Li Jia;Massimo Menenti;Hamid Reza Ghafarian Malamiri","doi":"10.1109/JSTARS.2025.3551360","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3551360","url":null,"abstract":"Optical fine and coarse spatial resolution multispectral images are essential for monitoring land surface processes but are often affected by gaps due to cloud contamination and other factors. Gap-filling methods are vital for overcoming these issues, yet existing approaches struggle to accurately reconstruct pixels impacted by undetected thin clouds and shadows, particularly in fine spatial resolution images. This study introduces a comprehensive gap-filling method that identifies and reconstructs invalid pixels in both fine and coarse spatial resolution images. The method combines different spatial and temporal gap-filling methods. The specific combination of methods is orchestrated to adapt to each image, mainly on the basis of the fractional abundance and spatial pattern of cloud cover. To evaluate the performance, experiments were conducted using MODIS (coarse-resolution) and Landsat/OLI (fine-resolution) images with artificial gaps (10% –90% ) introduced at varying positions in cloud-free reference images. For coarse-resolution images, the blue band showed the lowest root mean square error (RMSE) of 0.004 to 0.03, while the near-infrared (NIR) band had higher RMSE (0.01–0.05). The structural similarity index measure (SSIM) ranged from 0.96 to 0.73 as gap percentages increased. For fine-resolution images, random gaps were reconstructed most effectively, with RMSE values for the blue band between 0.005 and 0.01, and NIR errors ranging from 0.01 to 0.05. SSIM values ranged from 0.90 to 0.83 (blue) and 0.86 to 0.71 (NIR), confirming the method reliability for time-series analysis and data fusion applications.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8729-8746"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925634","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143786357","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Xiaoyang Zhang;Kaihui Dong;Dapeng Cheng;Zhen Hua;Jinjiang Li
{"title":"STWANet: Spatio-Temporal Wavelet Attention Aggregation Network for Remote Sensing Change Detection","authors":"Xiaoyang Zhang;Kaihui Dong;Dapeng Cheng;Zhen Hua;Jinjiang Li","doi":"10.1109/JSTARS.2025.3551093","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3551093","url":null,"abstract":"Existing change detection techniques exhibit significant deficiencies in the recognition of building edges and detailed textures, making it challenging to accurately distinguish building boundaries from the background. Consequently, these methods struggle to precisely capture complex building contours and subtle texture variations. To address this problem, a spatio-temporal wavelet attention aggregation network (STWANet) is proposed in this article. This network uses a pretrained Resnet18 to extract multiscale features to obtain features with sufficient spatial details and semantic information. We introduce the spatio-temporal differential self-attention module to extract the spatio-temporal difference information between two multiscale temporal features, and the introduction of the self-Attention mechanism is able to focus on the regions with the most significant changes in the multiscale feature maps. In order to extract the changes of detailed features such as building edges, we introduce the wavelet feature enhancement module (WFEM) to enhance the representation of the frequency domain feature information of the changing features, especially the enhancement of high-frequency detail information (e.g., building edges). In order to make up for the shortcomings of WFEM in capturing specific details and global spatial features, we also introduce the dual attention aggregation module to extract the feature information of the changing areas in parallel with WFEM, which can process the spatial context information in a more detailed way, and can better retain the detailed features, especially the complex spatial structure and shape information. spatial structure and shape information. We verify the effectiveness and advancement of STWANet on three classical datasets (LEVIR-CD, WHU-CD, GZ-CD), and the experimental results show that STWANet reaches the state-of-the-art performance level.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8813-8830"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10924762","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143786318","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Rangel Daroya;Luisa Vieira Lucchese;Travis Simmons;Punwath Prum;Tamlin Pavelsky;John Gardner;Colin J. Gleason;Subhransu Maji
{"title":"Improving Satellite Imagery Masking Using Multitask and Transfer Learning","authors":"Rangel Daroya;Luisa Vieira Lucchese;Travis Simmons;Punwath Prum;Tamlin Pavelsky;John Gardner;Colin J. Gleason;Subhransu Maji","doi":"10.1109/JSTARS.2025.3551620","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3551620","url":null,"abstract":"Many remote sensing applications require masking of pixels in satellite imagery for further analysis. For instance, estimating water quality variables such as suspended sediment concentration (SSC) requires isolating pixels depicting water bodies unaffected by clouds, their shadows, terrain shadows, and snow and ice formation. A significant bottleneck is the reliance on multiple data products (e.g., satellite imagery and elevation maps) and lack of precision in individual processing steps, which degrade estimation accuracy. We propose a unified masking system that predicts all necessary masks from harmonized landsat and sentinel (HLS) imagery. Our model leverages multitask learning to improve accuracy while sharing computation across tasks for added efficiency. In this article, we explore recent deep learning architectures, demonstrating that masking performance benefits from pretraining on large satellite imagery datasets. We present a range of models offering different speed/accuracy tradeoffs: MobileNet variants provide the fastest inference while maintaining competitive accuracy, whereas transformer-based architectures achieve the highest accuracy, particularly when pretrained on large-scale satellite datasets. Our models provide a 9% <inline-formula><tex-math>$F1$</tex-math></inline-formula> score improvement compared to previous work on water pixel identification. When integrated with an SSC estimation system, our models result in a 30× speedup while reducing estimation error by 2.64 mg/L, allowing for global-scale analysis. We also evaluate our model on a recently proposed cloud and cloud shadow estimation benchmark, where we outperform the current state-of-the-art model by at least 6% in <inline-formula><tex-math>$F1$</tex-math></inline-formula> score.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8777-8796"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925631","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143786325","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Jinyue Chen;Youming Wu;Wei Dai;Wenhui Diao;Yang Li;Xin Gao;Xian Sun
{"title":"Text-Enhanced Multimodal Method for SAR Ship Classification With Geometry and Polarization Information","authors":"Jinyue Chen;Youming Wu;Wei Dai;Wenhui Diao;Yang Li;Xin Gao;Xian Sun","doi":"10.1109/JSTARS.2025.3551239","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3551239","url":null,"abstract":"Synthetic aperture radar (SAR) ship classification is crucial for maritime surveillance. Most existing methods primarily focus on visual or polarimetric features, often constrained by a limited feature set and facing challenges in data diversity and multimodal information integration. This study introduces a text-enhanced multimodal framework for SAR ship classification (TeMSC), an extensible and unified approach that integrates multimodal information related to SAR ships. It consists of text-form geometry information embedding, polarization and visual information embedding, and a multimodal prediction module. By incorporating ship geometry information in text format, TeMSC leverages text representation to enhance feature expressiveness, compensating for the limited discriminative power of traditional visual and polarization features, especially in low-resolution scenarios. TeMSC effectively processes complementary multimodal information through a multimodal prediction module, while avoiding the complexity associated with traditional decision-level feature fusion strategies. In addition, a classification token mechanism is introduced to streamline the classification process. Through a two-stage training strategy, TeMSC captures information across multiple SAR datasets, enhancing its generalization and adaptability. Extensive experiments on the FUSAR-Ship and OpenSARShip datasets demonstrate the superior performance of TeMSC and highlight the benefits of multimodal integration for SAR ship classification. TeMSC provides a foundation for future research on SAR-focused multimodal learning applications.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8659-8671"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925632","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761431","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"Improved Deorientation Processing for Polarimetric SAR Data Using a Phenomenological Approach","authors":"Reza Bordbari;Andrew J. Hooper","doi":"10.1109/JSTARS.2025.3551246","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3551246","url":null,"abstract":"Polarimetric synthetic aperture radar data can be used to retrieve structural and textural information of the surface and is widely used in land cover classification. To extract relevant parameters, the data must first be compensated for the orientation of the polarimetric scattering targets. However, existing methods for deorientation processing are not able achieve this robustly and the data still suffer from orientation-induced scattering mechanism ambiguity. Here, we present a new approach to deorientation that greatly improves on existing methods. Our algorithm innovatively employs phenomenological target decomposition theory and the concept of polarization nulls to extract and deorient orientation-perturbed components of the target. The objective is to manipulate coherency matrix elements of a distributed target to obtain more descriptive target parameters, which are of critical importance in its deorientation. We applied our approach to C-, L-, and P-band datasets containing built-up areas with different orientations. In contrast to existing methods, our deorientation algorithm led to targets of all orientations being identified as having similar scattering characteristics. We also demonstrated that, an improved polarimetric target decomposition performance is achieved when the proposed deorientation processing is incorporated into model-based decompositions.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8685-8695"},"PeriodicalIF":4.7,"publicationDate":"2025-03-14","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925637","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143761432","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Seung-Bum Kim;Xiao-Lan Xu;Simon Kraatz;Andreas Colliander;Michael H. Cosh;Vicky Kelly;Paul Siqueira
{"title":"Soil Moisture Estimates Using -Band Airborne SAR Over Forests Replicating NISAR Observations","authors":"Seung-Bum Kim;Xiao-Lan Xu;Simon Kraatz;Andreas Colliander;Michael H. Cosh;Vicky Kelly;Paul Siqueira","doi":"10.1109/JSTARS.2025.3544095","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3544095","url":null,"abstract":"Airborne SAR observations of soil moisture conditions at 6-m resolution are analyzed over deciduous and evergreen forests in the U.S. Northeast during the 10-day spring and 14-day summer periods in 2022. During the summer, the dynamic range of HH is about 1 dB, associated mostly with soil moisture changes. Larger changes in backscattering are found between the two seasons, reflecting the vegetation effect. In spring, backscattering decreases in time, suggesting the impact of drying trunks and thickening foliage. In summer, σ° correlates highly with in situ soil moisture, consistently between ascending and descending viewing geometry on flat terrain and on slopes only when imaged at similar incidence angles. The consistency benefits NISAR's retrieval by allowing more frequent consistent retrievals of soil moisture. Soil moisture was retrieved using HH to replicate NISAR observations and its accuracy in the eight sites is 0.067 m<sup>3</sup>/m<sup>3</sup> in unbiased RMSE, assessed over a 140-m domain per in situ site. The results are very encouraging as an independent test of the retrieval algorithm under the challenging conditions of surface slope or forest vegetation. Deficiencies in the retrieval algorithm appear to originate from the modeling of vegetation effect and topography. As long as the two causes are temporally static, they introduce a bias error. However, the temporal range of the retrieval is the most useful property for applications and matches well with in situ observations.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"7364-7373"},"PeriodicalIF":4.7,"publicationDate":"2025-03-13","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10925521","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143698335","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
{"title":"3-D UXSE-Net for Seismic Channel Detection Based on Satellite Image Enhanced Synthetic Datasets","authors":"Xinke Zhang;Yihuai Lou;Naihao Liu;Daosheng Ling;Yunmin Chen","doi":"10.1109/JSTARS.2025.3550578","DOIUrl":"https://doi.org/10.1109/JSTARS.2025.3550578","url":null,"abstract":"Channels are essential indicators of sedimentary environments and play a vital role in geological applications, such as hydrocarbon exploration, sediment transport, and the study of ancient river geomorphology. Deep learning (DL) techniques have shown great potential in improving channel detection accuracy and efficiency. However, insufficient labeled training data remains a key challenge for refining DL models. To address this issue, we propose a workflow that automatically generates synthetic datasets by integrating channel features extracted from high-resolution satellite images. We first extract river channel features and grayscale values from satellite images. These extracted features are then used to construct reflectivity models, incorporating structural deformations based on seismic reflector dips. The reflectivity models are subsequently convolved with wavelets to generate synthetic datasets. These synthetic datasets are used to train the proposed 3-D UXSE-Net, which integrates the 3-D UX-Net architecture with the squeeze-and-excitation blocks. The model generates improved feature representations that enhance performance by combining convolutional neural networks for local feature extraction and Transformer-based modules for capturing global context. We validate our approach by applying the model to both synthetic and 3-D field seismic datasets. Our results show that 3-D UXSE-Net outperforms baseline methods, including the coherence-based approach and other DL models, and demonstrates strong generalization to field data even when trained solely on synthetic data. Comparisons of different methods highlight the effectiveness of the synthetic data generation approach for DL model training.","PeriodicalId":13116,"journal":{"name":"IEEE Journal of Selected Topics in Applied Earth Observations and Remote Sensing","volume":"18 ","pages":"8300-8311"},"PeriodicalIF":4.7,"publicationDate":"2025-03-12","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"https://ieeexplore.ieee.org/stamp/stamp.jsp?tp=&arnumber=10923698","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"143740374","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":2,"RegionCategory":"地球科学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"OA","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}