Pub Date : 2025-09-05DOI: 10.1109/LGRS.2025.3606521
Runbo Yang;Huiyan Han;Shanyuan Bai;Yaming Cao
Multiscale object detection in remote sensing imagery poses significant challenges, including substantial variations in object size, diverse orientations, and interference from complex backgrounds. To address these issues, we propose a scale-aware detection and feature fusion network (SADFF-Net), a novel detection framework that incorporates a Multiscale contextual attention fusion (MCAF) module to enhance information exchange between feature layers and suppress irrelevant feature interference. In addition, SADFF-Net employs an adaptive spatial feature fusion (ASFF) module to improve semantic consistency across feature layers by assigning spatial weights at multiple scales. To enhance adaptability to scale variations, the regression head integrates a deformable convolution. In contrast, the classification head utilizes depth-wise separable convolutions to significantly reduce computational complexity without compromising detection accuracy. Extensive experiments on the DOTAv1 and DIOR_R datasets demonstrate that SADFF-Net outperforms current state-of-the-art methods in Multiscale object detection.
{"title":"SADFF-Net: Scale-Aware Detection and Feature Fusion for Multiscale Remote Sensing Object Detection","authors":"Runbo Yang;Huiyan Han;Shanyuan Bai;Yaming Cao","doi":"10.1109/LGRS.2025.3606521","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3606521","url":null,"abstract":"Multiscale object detection in remote sensing imagery poses significant challenges, including substantial variations in object size, diverse orientations, and interference from complex backgrounds. To address these issues, we propose a scale-aware detection and feature fusion network (SADFF-Net), a novel detection framework that incorporates a Multiscale contextual attention fusion (MCAF) module to enhance information exchange between feature layers and suppress irrelevant feature interference. In addition, SADFF-Net employs an adaptive spatial feature fusion (ASFF) module to improve semantic consistency across feature layers by assigning spatial weights at multiple scales. To enhance adaptability to scale variations, the regression head integrates a deformable convolution. In contrast, the classification head utilizes depth-wise separable convolutions to significantly reduce computational complexity without compromising detection accuracy. Extensive experiments on the DOTAv1 and DIOR_R datasets demonstrate that SADFF-Net outperforms current state-of-the-art methods in Multiscale object detection.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-05","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145036868","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-04DOI: 10.1109/LGRS.2025.3605910
Renfang Wang;Kun Yang;Feng Wang;Hong Qiu;Yingying Huang;Xiufeng Liu
Deep learning is a powerful technique for semantic change detection (SCD) of bitemporal remote sensing images. In this work, we propose to improve SCD accuracy using deep learning with frequency feature enhancement (FFE). Specifically, we develop an FFE module that aims to enhance the performance of both binary change detection (BCD) and semantic segmentation, two main key components for obtaining high SCD accuracy, by integrating the Fourier transform and attention mechanisms. Experimental results on the SECOND and LandSat-SCD datasets demonstrate the effectiveness of the proposed method, and it achieves high resolution for change boundaries.
{"title":"Semantic Change Detection of Bitemporal Remote Sensing Images Using Frequency Feature Enhancement","authors":"Renfang Wang;Kun Yang;Feng Wang;Hong Qiu;Yingying Huang;Xiufeng Liu","doi":"10.1109/LGRS.2025.3605910","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605910","url":null,"abstract":"Deep learning is a powerful technique for semantic change detection (SCD) of bitemporal remote sensing images. In this work, we propose to improve SCD accuracy using deep learning with frequency feature enhancement (FFE). Specifically, we develop an FFE module that aims to enhance the performance of both binary change detection (BCD) and semantic segmentation, two main key components for obtaining high SCD accuracy, by integrating the Fourier transform and attention mechanisms. Experimental results on the SECOND and LandSat-SCD datasets demonstrate the effectiveness of the proposed method, and it achieves high resolution for change boundaries.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145036861","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-04DOI: 10.1109/LGRS.2025.3605993
Pengxiong Zhang;Yi Jiang;Xinguo Zhu
Due to its superior recognition accuracy, deep learning has been widely adopted in synthetic aperture radar (SAR) ship detection. Nevertheless, significant variations in ship target scales pose challenges for existing detection architectures, frequently leading to missed detections or false positives. Moreover, high-precision detection models are typically structurally complex and computationally intensive, resulting in substantial hardware resource consumption. In this letter, we introduce LSAR-Det, a novel SAR ship detection network designed to address these challenges. We propose a lightweight residual feature extraction (LRFE) module to construct the backbone network, enhancing feature extraction capabilities while reducing the number of parameters and floating-point operations per second (FLOPs). Furthermore, we design a lightweight cross-space convolution (LCSConv) module to replace the traditional convolution in the neck network. In addition, we incorporate a multiscale bidirectional feature pyramid network (M-BiFPN) to facilitate multiscale feature fusion with fewer parameters. Our proposed model contains merely 0.985M parameters and requires only 3.3G FLOPs. Experimental results on the SAR ship detection dataset (SSDD) and high-resolution SAR image dataset (HRSID) datasets demonstrate that LSAR-Det outperforms other models, achieving detection accuracies of 98.2% and 91.8%, respectively, thereby effectively balancing detection performance and model efficiency.
{"title":"LSAR-Det: A Lightweight YOLOv11-Based Model for Ship Detection in SAR Images","authors":"Pengxiong Zhang;Yi Jiang;Xinguo Zhu","doi":"10.1109/LGRS.2025.3605993","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605993","url":null,"abstract":"Due to its superior recognition accuracy, deep learning has been widely adopted in synthetic aperture radar (SAR) ship detection. Nevertheless, significant variations in ship target scales pose challenges for existing detection architectures, frequently leading to missed detections or false positives. Moreover, high-precision detection models are typically structurally complex and computationally intensive, resulting in substantial hardware resource consumption. In this letter, we introduce LSAR-Det, a novel SAR ship detection network designed to address these challenges. We propose a lightweight residual feature extraction (LRFE) module to construct the backbone network, enhancing feature extraction capabilities while reducing the number of parameters and floating-point operations per second (FLOPs). Furthermore, we design a lightweight cross-space convolution (LCSConv) module to replace the traditional convolution in the neck network. In addition, we incorporate a multiscale bidirectional feature pyramid network (M-BiFPN) to facilitate multiscale feature fusion with fewer parameters. Our proposed model contains merely 0.985M parameters and requires only 3.3G FLOPs. Experimental results on the SAR ship detection dataset (SSDD) and high-resolution SAR image dataset (HRSID) datasets demonstrate that LSAR-Det outperforms other models, achieving detection accuracies of 98.2% and 91.8%, respectively, thereby effectively balancing detection performance and model efficiency.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145061895","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The 11-month data gap between gravity recovery and climate experiment (GRACE) and GRACE Follow-On (GRACE-FO) hinders monitoring long-term ice mass change and its further analysis. While many attempts have been made to bridge water storage gaps, few unified frameworks exist to bridge the ice mass change gaps for both Greenland ice sheet (GrIS) and Antarctic ice sheet (AIS). This study combined partial least squares regression (PLSR) and the Sparrow Search Algorithm optimized back propagation (SSA-BP) to fill this gap in GrIS and AIS. During this process, seasonal autoregressive integrated moving average (MA) with exogenous variables (SARIMAX) and multiple linear regression (MLR) were introduced as comparison. PSLR is utilized to select key variables for constructing predictive models. We found SSA-BP outperformed SARIMAX and MLR, with correlation coefficients (CCs) and root mean square error (RMSE) at 0.99 and 39.22 Gt for GrIS, and 0.95 and 189.85 Gt for AIS within the testing period. SSA-BP demonstrated a reasonable mass change trend with less noise than other methods. SSA-BP reconstructed result shows superiority than other researches. Moreover, the reconstructed seasonal signals highlight the importance of filling the gap, showing decreased mass loss for GrIS and continuous mass loss acceleration for AIS post-2016.
{"title":"A Unified Framework for Bridging the Data Gap Between GRACE/GRACE-FO for Both Greenland and Antarctica","authors":"Zhuoya Shi;Zemin Wang;Baojun Zhang;Nicholas E. Barrand;Manman Luo;Shuang Wu;Jiachun An;Hong Geng;Haojian Wu","doi":"10.1109/LGRS.2025.3605913","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605913","url":null,"abstract":"The 11-month data gap between gravity recovery and climate experiment (GRACE) and GRACE Follow-On (GRACE-FO) hinders monitoring long-term ice mass change and its further analysis. While many attempts have been made to bridge water storage gaps, few unified frameworks exist to bridge the ice mass change gaps for both Greenland ice sheet (GrIS) and Antarctic ice sheet (AIS). This study combined partial least squares regression (PLSR) and the Sparrow Search Algorithm optimized back propagation (SSA-BP) to fill this gap in GrIS and AIS. During this process, seasonal autoregressive integrated moving average (MA) with exogenous variables (SARIMAX) and multiple linear regression (MLR) were introduced as comparison. PSLR is utilized to select key variables for constructing predictive models. We found SSA-BP outperformed SARIMAX and MLR, with correlation coefficients (CCs) and root mean square error (RMSE) at 0.99 and 39.22 Gt for GrIS, and 0.95 and 189.85 Gt for AIS within the testing period. SSA-BP demonstrated a reasonable mass change trend with less noise than other methods. SSA-BP reconstructed result shows superiority than other researches. Moreover, the reconstructed seasonal signals highlight the importance of filling the gap, showing decreased mass loss for GrIS and continuous mass loss acceleration for AIS post-2016.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145036860","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-04DOI: 10.1109/LGRS.2025.3605916
Yuquan Gan;Siyu Wu;Xingyu Li;Zhijie Xu;Yushan Pan
Hyperspectral image (HSI) classification faces critical challenges in effectively modeling the intricate spectral–spatial structures and non-Euclidean relationships. Traditional methods often struggle to simultaneously capture local details, global contextual dependencies, and graph-structured correlations, leading to limited classification accuracy. To address the above issues, this letter proposes a graph-aware hybrid encoding (GAHE) framework. To fully exploit the spectral–spatial characteristics and graph structural dependencies inherent in HSI, the proposed method is structured into three key components: a multiscale selective graph-aware attention (MSGA) module, a hybrid projection encoding module, and a graph sensitive aggregation (GSA) module. The three modules work in a complementary manner to progressively refine and enhance feature representations across multiple scales and modalities. Compared with advanced classification methods, the experimental results demonstrate that the proposed GAHE method shows better classification performance.
{"title":"Graph-Aware Hybrid Encoding for Hyperspectral Image Classification","authors":"Yuquan Gan;Siyu Wu;Xingyu Li;Zhijie Xu;Yushan Pan","doi":"10.1109/LGRS.2025.3605916","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605916","url":null,"abstract":"Hyperspectral image (HSI) classification faces critical challenges in effectively modeling the intricate spectral–spatial structures and non-Euclidean relationships. Traditional methods often struggle to simultaneously capture local details, global contextual dependencies, and graph-structured correlations, leading to limited classification accuracy. To address the above issues, this letter proposes a graph-aware hybrid encoding (GAHE) framework. To fully exploit the spectral–spatial characteristics and graph structural dependencies inherent in HSI, the proposed method is structured into three key components: a multiscale selective graph-aware attention (MSGA) module, a hybrid projection encoding module, and a graph sensitive aggregation (GSA) module. The three modules work in a complementary manner to progressively refine and enhance feature representations across multiple scales and modalities. Compared with advanced classification methods, the experimental results demonstrate that the proposed GAHE method shows better classification performance.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145110260","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-04DOI: 10.1109/LGRS.2025.3605978
Allan A. Nielsen;Henning Skriver;Knut Conradsen
We report on a complex Wishart distribution-based test statistic $boldsymbol {Q}$ for block-diagonality in Hermitian matrices such as the ones analyzed in polarimetric synthetic aperture radar (polSAR) image data in the covariance matrix formulation. We also give an improved probability measure $boldsymbol {P}$ associated with the test statistic. This is used in a case with simulated data to demonstrate the superiority of the new expression for $boldsymbol {P}$ and to illustrate the dependence of results on the choice of covariance matrix, its dimensionality, the equivalent number of looks, and two parameters in the improved $boldsymbol {P}$ measure. We also give two cases with acquired data. One case is with airborne F-SAR polarimetric data, where we test for reflection symmetry, another case is with (spaceborne) dual-pol Sentinel-1 data, where we test if the data are diagonal-only. The absence of block-diagonal structure occurs mostly for man-made objects. In the example with Sentinel-1 data, some objects (e.g., buildings, cars, aircraft, and ships) are detected, others (e.g., some bridges) are not.
{"title":"A Test Statistic for Block-Diagonal Covariance Matrix Structure in polSAR Data","authors":"Allan A. Nielsen;Henning Skriver;Knut Conradsen","doi":"10.1109/LGRS.2025.3605978","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605978","url":null,"abstract":"We report on a complex Wishart distribution-based test statistic <inline-formula> <tex-math>$boldsymbol {Q}$ </tex-math></inline-formula> for block-diagonality in Hermitian matrices such as the ones analyzed in polarimetric synthetic aperture radar (polSAR) image data in the covariance matrix formulation. We also give an improved probability measure <inline-formula> <tex-math>$boldsymbol {P}$ </tex-math></inline-formula> associated with the test statistic. This is used in a case with simulated data to demonstrate the superiority of the new expression for <inline-formula> <tex-math>$boldsymbol {P}$ </tex-math></inline-formula> and to illustrate the dependence of results on the choice of covariance matrix, its dimensionality, the equivalent number of looks, and two parameters in the improved <inline-formula> <tex-math>$boldsymbol {P}$ </tex-math></inline-formula> measure. We also give two cases with acquired data. One case is with airborne F-SAR polarimetric data, where we test for reflection symmetry, another case is with (spaceborne) dual-pol Sentinel-1 data, where we test if the data are diagonal-only. The absence of block-diagonal structure occurs mostly for man-made objects. In the example with Sentinel-1 data, some objects (e.g., buildings, cars, aircraft, and ships) are detected, others (e.g., some bridges) are not.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-04","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145061854","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-03DOI: 10.1109/LGRS.2025.3605792
Siyuan Ding;Xun Wang;Deshan Feng;Cheng Chen;Dianbo Li
Ground penetrating radar (GPR) is a powerful tool for exploring the shallow subsurface due to its effective and noninvasive features. Recently, the accurate and high-resolution characterization of subsurface properties in 3-D GPR investigations calls for a quantitative and high-resolution imaging approach. However, the full-waveform inversion (FWI) method for GPR data was performed mostly in 2-D and rarely discussed the polarizations. To fully utilize 3-D GPR polarization data, this letter proposes a frequency-domain FWI algorithm for simultaneous inversion of both the co-polarized and cross-polarized data. Detail derivations and vital processes in our inversion workflow were described in detail, before applying it to the numerical experiments and analyzing the potential impacts of the polarizations on inversion results with a synthetic model. Results showed that the cross-polarized data are more sensitive than the co-polarized data in inversion, and the behaviors in the inversion of the multipolarized data with different values in the weighting matrix suggest that larger weights for co-polarized data are of benefit to a better inversion result.
{"title":"Potential Impacts of 3-D Polarized GPR Data on Full-Waveform Inversion","authors":"Siyuan Ding;Xun Wang;Deshan Feng;Cheng Chen;Dianbo Li","doi":"10.1109/LGRS.2025.3605792","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605792","url":null,"abstract":"Ground penetrating radar (GPR) is a powerful tool for exploring the shallow subsurface due to its effective and noninvasive features. Recently, the accurate and high-resolution characterization of subsurface properties in 3-D GPR investigations calls for a quantitative and high-resolution imaging approach. However, the full-waveform inversion (FWI) method for GPR data was performed mostly in 2-D and rarely discussed the polarizations. To fully utilize 3-D GPR polarization data, this letter proposes a frequency-domain FWI algorithm for simultaneous inversion of both the co-polarized and cross-polarized data. Detail derivations and vital processes in our inversion workflow were described in detail, before applying it to the numerical experiments and analyzing the potential impacts of the polarizations on inversion results with a synthetic model. Results showed that the cross-polarized data are more sensitive than the co-polarized data in inversion, and the behaviors in the inversion of the multipolarized data with different values in the weighting matrix suggest that larger weights for co-polarized data are of benefit to a better inversion result.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145090173","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
The Surface Water and Ocean Topography (SWOT) satellite provides high-resolution observations of the ocean surface topography and elevation of inland waters. Measurements from the two onboard Advanced Microwave Radiometers (AMRs) are used to compute the wet tropospheric correction (WTC), accounting for the radar signal delay due to water vapor and cloud liquid water content in the troposphere. This study presents the first implementation of the Global Navigation Satellite System (GNSS)-derived Path Delay Plus (GPD+) algorithm for SWOT to estimate the WTC when AMR observations are absent or invalid. Using the first 15 science-phase cycles between 50°N and 50°S, GPD+ retrieves the WTC for approximately 7% of points per cycle that would otherwise be excluded. Retrieval rates per cycle range from less than 5% of the points in passes mostly over open ocean, where the WTC derived from the radiometers is usually preserved, to up to 15% in passes including coastal zones. These results indicate that GPD+ can recover WTC values otherwise unavailable from SWOT’s radiometers, increasing the availability of valid WTC for SWOT measurements, in particular over coastal regions. Further refinements will focus on improving the accuracy of the WTC along the KaRIn swath and the Poseidon-3C nadir track.
{"title":"First Implementation of GPD+ Wet Tropospheric Correction on SWOT Side 1 and Side 2 Radiometer Tracks","authors":"Isabel Cardoso;Clara Lázaro;Telmo Vieira;M. Joana Fernandes","doi":"10.1109/LGRS.2025.3605854","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605854","url":null,"abstract":"The Surface Water and Ocean Topography (SWOT) satellite provides high-resolution observations of the ocean surface topography and elevation of inland waters. Measurements from the two onboard Advanced Microwave Radiometers (AMRs) are used to compute the wet tropospheric correction (WTC), accounting for the radar signal delay due to water vapor and cloud liquid water content in the troposphere. This study presents the first implementation of the Global Navigation Satellite System (GNSS)-derived Path Delay Plus (GPD+) algorithm for SWOT to estimate the WTC when AMR observations are absent or invalid. Using the first 15 science-phase cycles between 50°N and 50°S, GPD+ retrieves the WTC for approximately 7% of points per cycle that would otherwise be excluded. Retrieval rates per cycle range from less than 5% of the points in passes mostly over open ocean, where the WTC derived from the radiometers is usually preserved, to up to 15% in passes including coastal zones. These results indicate that GPD+ can recover WTC values otherwise unavailable from SWOT’s radiometers, increasing the availability of valid WTC for SWOT measurements, in particular over coastal regions. Further refinements will focus on improving the accuracy of the WTC along the KaRIn swath and the Poseidon-3C nadir track.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145036199","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Predictively synthesizing high-quality, close-range asteroid surface views from distant optical remote sensing imagery is critical for mission planning and landing-site selection in asteroid exploration missions. However, distant observations inherently lack sufficient resolution and surface detail, limiting the existing novel view synthesis (NVS) methods. To address this, we introduce, to the best of our knowledge, the first framework for distant-to-close NVS, tailored for asteroid surface imaging. Our method features two key innovations. First, a 3-D Gaussian splatting (3D-GS) super-resolution (SR) module applies 2-D SR to generate high-resolution virtual close-range views from distant images, enriching the 3-D scene model with finer details. Second, an entropy-driven residual refinement strategy adaptively emphasizes structurally complex regions by assigning higher loss weights based on residual image entropy. This strategy triggers targeted subdivisions of 3-D Gaussians in the areas of high structural complexity. Experiments conducted on datasets from Hayabusa (Itokawa), Dawn (Vesta), Rosetta (67P/Churyumov-Gerasimenko), Hayabusa2 (Ryugu), and OSIRIS-REx (Bennu) missions demonstrate substantial improvements over baseline methods in quantitative metrics, such as peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and learned perceptual image patch similarity (LPIPS).
从遥远的光学遥感影像中预测合成高质量的近距离小行星表面图像对于小行星探测任务的任务规划和着陆点选择至关重要。然而，远程观测本身缺乏足够的分辨率和表面细节，限制了现有的新视图合成（NVS）方法。为了解决这个问题，我们介绍了，据我们所知，为小行星表面成像量身定制的第一个远距离到近距离NVS框架。我们的方法有两个关键的创新。首先，3d高斯飞溅（3D-GS）超分辨率（SR）模块应用2d SR从远处图像生成高分辨率虚拟近景视图，以更精细的细节丰富3d场景模型。其次，熵驱动残差细化策略通过基于残差图像熵分配更高的损失权值，自适应地强调结构复杂的区域。该策略触发了高结构复杂性区域的三维高斯函数的目标细分。在Hayabusa (Itokawa), Dawn (Vesta), Rosetta (67P/Churyumov-Gerasimenko), Hayabusa2 （Ryugu）和OSIRIS-REx （Bennu）任务的数据集上进行的实验表明，在峰值信噪比（PSNR），结构相似指数测量（SSIM）和学习感知图像patch相似度（LPIPS）等定量指标上，比基线方法有了实质性的改进。
{"title":"Distant-to-Close Novel View Synthesis for Asteroid Surface Imaging","authors":"Xiaodong Wei;Linyan Cui;Xinyu Zhao;Gangzheng Ai;Jihao Yin","doi":"10.1109/LGRS.2025.3605777","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605777","url":null,"abstract":"Predictively synthesizing high-quality, close-range asteroid surface views from distant optical remote sensing imagery is critical for mission planning and landing-site selection in asteroid exploration missions. However, distant observations inherently lack sufficient resolution and surface detail, limiting the existing novel view synthesis (NVS) methods. To address this, we introduce, to the best of our knowledge, the first framework for distant-to-close NVS, tailored for asteroid surface imaging. Our method features two key innovations. First, a 3-D Gaussian splatting (3D-GS) super-resolution (SR) module applies 2-D SR to generate high-resolution virtual close-range views from distant images, enriching the 3-D scene model with finer details. Second, an entropy-driven residual refinement strategy adaptively emphasizes structurally complex regions by assigning higher loss weights based on residual image entropy. This strategy triggers targeted subdivisions of 3-D Gaussians in the areas of high structural complexity. Experiments conducted on datasets from Hayabusa (Itokawa), Dawn (Vesta), Rosetta (67P/Churyumov-Gerasimenko), Hayabusa2 (Ryugu), and OSIRIS-REx (Bennu) missions demonstrate substantial improvements over baseline methods in quantitative metrics, such as peak signal-to-noise ratio (PSNR), structural similarity index measure (SSIM), and learned perceptual image patch similarity (LPIPS).","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-03","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145090172","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
Pub Date : 2025-09-02DOI: 10.1109/LGRS.2025.3605331
Fan Ye;Xiaoning Zhang;Zhengjie Wang;Yifei Wang;Zhaoyang Peng;Tengying Fu;Ziti Jiao;Yanxuan Wu;Yue Wang
Considering the simplicity of flight route planning, orthorectified images obtained from nadir observations are widely used in remote sensing. However, they are always insufficient to represent the anisotropic reflectance and 3-D structural information of objects. Therefore, multiangle observation information can enhance target information and potentially improve the accuracy of target classification and recognition. In this study, we investigated the potential of anisotropic reflectance information in land cover classification. By employing the DJI P4M multispectral observation system, multiangle multispectral reflectance images for five land cover types were captured at bare soil, concrete roads, grassland, apricot tree, and red broom cypress areas. Subsequently, the anisotropic flat index (AFX)-based bidirectional reflectance distribution function (BRDF) archetypes model and the kernel-driven model were used to reconstruct the BRDF. Finally, land cover classification was performed using three types of machine learning algorithm considering different BRDF features and band combinations. The results indicate that, compared to nadir directional reflectance, multiangle feature sets can improve the overall classification accuracy up to 24%. Compared to using single-band information, band combinations can also improve that up to 54%. The overall accuracy using the feature set of kernel-driven model parameters and nadir reflectance was also enhanced significantly, which can reach 86% using green-red-near infrared band combinations. This work demonstrates the contribution of multiangle multispectral information to natural and artificial land cover classification.
{"title":"Application of Optical Multiangle Multispectral Reflectance in Land Cover Classification","authors":"Fan Ye;Xiaoning Zhang;Zhengjie Wang;Yifei Wang;Zhaoyang Peng;Tengying Fu;Ziti Jiao;Yanxuan Wu;Yue Wang","doi":"10.1109/LGRS.2025.3605331","DOIUrl":"https://doi.org/10.1109/LGRS.2025.3605331","url":null,"abstract":"Considering the simplicity of flight route planning, orthorectified images obtained from nadir observations are widely used in remote sensing. However, they are always insufficient to represent the anisotropic reflectance and 3-D structural information of objects. Therefore, multiangle observation information can enhance target information and potentially improve the accuracy of target classification and recognition. In this study, we investigated the potential of anisotropic reflectance information in land cover classification. By employing the DJI P4M multispectral observation system, multiangle multispectral reflectance images for five land cover types were captured at bare soil, concrete roads, grassland, apricot tree, and red broom cypress areas. Subsequently, the anisotropic flat index (AFX)-based bidirectional reflectance distribution function (BRDF) archetypes model and the kernel-driven model were used to reconstruct the BRDF. Finally, land cover classification was performed using three types of machine learning algorithm considering different BRDF features and band combinations. The results indicate that, compared to nadir directional reflectance, multiangle feature sets can improve the overall classification accuracy up to 24%. Compared to using single-band information, band combinations can also improve that up to 54%. The overall accuracy using the feature set of kernel-driven model parameters and nadir reflectance was also enhanced significantly, which can reach 86% using green-red-near infrared band combinations. This work demonstrates the contribution of multiangle multispectral information to natural and artificial land cover classification.","PeriodicalId":91017,"journal":{"name":"IEEE geoscience and remote sensing letters : a publication of the IEEE Geoscience and Remote Sensing Society","volume":"22 ","pages":"1-5"},"PeriodicalIF":4.4,"publicationDate":"2025-09-02","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":null,"resultStr":null,"platform":"Semanticscholar","paperid":"145021449","PeriodicalName":null,"FirstCategoryId":null,"ListUrlMain":null,"RegionNum":0,"RegionCategory":"","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":"","EPubDate":null,"PubModel":null,"JCR":null,"JCRName":null,"Score":null,"Total":0}
京公网安备 11010802042870号
京ICP备2023020795号-1
扫码关注我们
求助内容：
应助结果提醒方式：