查看更多>>摘要：In recent years, with the continuous development of remote sensing technology, maritime ship monitoring has become an important research area. Accurately counting the number of ships in remote sensing images is crucial for maritime traffic safety, fisheries management, and marine environmental protection. Existing methods typically use Gaussian kernel functions to generate density maps; however, due to the varied shapes of ships that do not conform to the Gaussian kernel, the resulting density maps fail to accurately reflect the true forms of ships, thereby affecting counting performance. To overcome these limitations, we introduce the cross-domain density map-generated ship counting network (CDDMNet). This network innovatively incorporates a cross-domain feature fusion module (CDFFM), which effectively adapts to ships of varying sizes and shapes. In addition, we have introduced the feature correlation regularization constraint (FCRC) and the integrated loss function, which effectively overcome the disturbances that may arise from variations in ship sizes and enhance the model’s adaptability to changes in ship types and environmental conditions. Experimental results show that the CDDMNet has achieved excellent performance across multiple remote sensing image datasets. Finally, on the RSOC dataset, the mean absolute error (MAE) reached 52.80 and the root mean squared error (RMSE) reached 69.77.
Barbara CaiffiLuca AlfonsoAndrea BersaniLuca Bottura...
1-5页
查看更多>>摘要：Deep-learning-based object detection has been increasingly attractive in various intelligent edge applications, including remote sensing and autonomous driving. However, achieving an optimal tradeoff between computing efficiency and detection accuracy is challenging. YOLO-series networks provide fast and lightweight detection but often sacrifice accuracy. In this work, we propose the multibranch cascading aggregation YOLO (MCA-YOLO) model, designed for multiscenario object detection tasks on edge devices. MCA-YOLO enhances detection accuracy by learning comprehensive feature information while maintaining computational efficiency through three key components: multibranch spatial pyramid pooling (MSPP), Ghost-convolution (GC)-based efficient layer aggregation network (G-ELAN), and hierarchical aggregation neck (HAN) that integrates MSPP and G-ELAN with GC. We train and validate MCA-YOLO using four benchmark datasets: VOC, COCO, SIMD, and VisDrone. The experimental results demonstrate significant improvements in detection accuracy and inference speed. Besides, we deploy the MCA-YOLO on a Jetson Xavier NX edge device embedded in an unmanned aerial vehicle (UAV), creating a remote sensing system capable of real-time, high-accuracy object detection. Our code is publicly available at https://github.com/lawlawCodes/MCA-YOLO.
Z. D. KauffmanC. D. CoatsG. D. MayP. M. McIntyre...
1-5页
查看更多>>摘要：The synthetic aperture radar (SAR) imagery plays a critical role in flood mapping due to its ability to capture data under all-weather and day-and-night conditions. However, the existing SAR segmentation methods, including the state-of-the-art HisSegNet, face challenges, such as limited generalization, insufficient utilization of SAR-specific features, and suboptimal performance on diverse datasets. To address these limitations, we propose enhanced-HisSegNet, a multimodal fusion strategy that builds upon HisSegNet by integrating learnable histogram layers (HLs) tailored for SAR data with active contour models (ACMs) for precise boundary refinement. These components are embedded into fine-tuned deep segmentation neural networks (DSNNs) to improve segmentation accuracy. Our model was evaluated on real SAR datasets, employing cross-dataset validation for robustness. Experimental results demonstrate significant performance gains, with up to 10% improvement in intersection over union (IoU)—a key metric that measures segmentation accuracy by computing the ratio of intersection to union between the predicted and ground truth regions—on internal datasets and 4% on external datasets, showcasing enhanced accuracy, robustness, and applicability. The code for this work is available at https://github.com/Mohsena1990/Enhanced-HistSegNet.
C. Martins JardimJ. A. García-MatosF. ToralS. Izquierdo Bermudez...
1-5页
查看更多>>摘要：The global navigation satellite system (GNSS) reflectometry synthetic aperture radar (SAR) interferometry (GNSS-R InSAR) system enables elevation deformation retrieval using a single satellite. However, variations in bistatic configurations and the generally low accuracy of most satellites necessitate a refined satellite selection method. Thus, this letter proposes a satellite selection algorithm for GNSS-R InSAR, aiming to optimize satellite selection and data acquisition time to improve the precision of elevation deformation monitoring. First, the interferometric phase model based on the repeat-pass concept was established using GPS L5 signals. Second, a satellite selection algorithm was proposed that incorporates constraints on resolution cells, spatial baseline, and phase sensitivity for elevation deformation, derived from an analysis of the repeat-pass spatial baseline of GNSS satellites, interferometric phase sensitivity, and the maximum deformation range. Third, 24 sets of repeat-pass data were collected, and the experimental results validate the effectiveness of this single-satellite selection approach.
José V. C. VargasJuan C. OrdonezSastry V. PamidiChul H. Kim...
1-5页
查看更多>>摘要：As an effective clutter suppression method for airborne radar, the atomic norm minimization (ANM)-based space-time adaptive processing (STAP) method suffers from high computational complexity and parameter setting difficulty. To solve these problems, a deep unfolded (DU) ANM algorithm is proposed for STAP in this study. First, the clutter estimation problem based on ANM is established. Then, the problem is solved via the alternating direction method of multipliers (ADMMs) and a deep neural network (DNN), which is trained by designing an appropriate loss function and constructing a complete dataset. At last, the clutter-plus-noise covariance matrix (CNCM) and the STAP weighting vector are obtained by processing the training range cell data via the trained network. Simulation results show that the proposed DU-ANM-STAP method can achieve higher clutter and noise suppression performance with lower computational cost than the existing ANM-STAP methods.
查看更多>>摘要：The selection of statistically homogeneous pixels (SHPs) is essential for precise deformation monitoring in distributed scatterer synthetic aperture radar interferometry (DS-InSAR). Current SHP selection methods face challenges in test efficacy under small-sample and low-contrast conditions, resulting in imbalanced Type I and Type II errors and poor detection of weak heterogeneity. To address these issues, the wide-scope homogeneous testing (WSHT) algorithm is introduced, which enhances the hypothesis test of confidence interval (HTCI) by calculating the extremes of the confidence interval length and integrating Baumgartner-Weiss–Schindler (BWS) to minimize this length and improve the accuracy of the reference pixel mean. Simulations demonstrate that WSHT outperforms BWS and HTCI, achieving accuracy improvements of 71.00% and 34.71%, respectively. Analysis of Sentinel-1 images from Shenzhen City further highlights WSHT’s performance, achieving the highest SNR of 0.2755, a balanced speckle suppression index (SSI) of 5.5486, and the lowest mean squared error (MSE) of 0.2089, outperforming BWS and HTCI in noise suppression, resolution preservation, and robustness to sample size variations.
Martina CascielloNicolò RivaDaniele PlacidoZach Hartwig...
1-5页
查看更多>>摘要：With the increasing use of uncrewed aerial vehicles (UAVs), improving the accuracy of semantic segmentation is becoming critical. Depth information preserves geometric structure, serving as an invaluable supplement to color-rich UAV imagery. Inspired by this, we proposed a novel framework named INVITATION, which exclusively takes original UAV imagery as input, yet is capable of obtaining complemented depth information and fusing into RGB semantic segmentation models effectively, thereby enhancing UAV semantic segmentation accuracy. Concretely, this framework supports two distinct depth generation approaches: high-precision multiview stereo (MVS) depth reconstruction using multiple views or video sequences via structure from motion (SfM) and monocular depth estimation using individual images. Our empirical evaluations conducted on the UAVid dataset showed that mIoU metric of INVITATION used precise reconstructed depth maps via MVS improved from 66.02% to 70.57%, while used depth predictions from pretrained models reached 69.69%, which supports the effectiveness of extracting and fusing depth information from original imagery in enhancing UAV semantic segmentation. This study explores a novel approach to acquire UAV multimodal information at low data cost, highlights the advantages of incorporating depth information into UAV semantic analysis, and paves the way for further studies on the integration of multimodal UAV information. Our code is available at https://github.com/CVEO/INVITATION.
Jonathan LeeJeseok BangGriffin BradfordDmytro Abraimov...
1-5页
查看更多>>摘要：Deep learning-based high-resolution remote sensing for farmland extraction is a crucial method for obtaining large-scale farmland information. However, variations in crop types, growth conditions, and factors such as narrow edges in farmland lead to lower extraction accuracy and inaccurate boundaries in high-resolution remote sensing. Therefore, this letter proposes a multibranch convolutional neural network (FFENet) that employs a dense-feature overlay fusion module (FFM) and an information homogeneity enhancement module. This network facilitates rapid extraction and dense fusion of information at various scales through the implementation of the dense FM, thereby enhancing the model’s representation of global consistency and local features. The information homogeneity enhancement module further strengthens the information exchange between the bottom and top layers, improves the fusion of feature information across branches, and ensures consistent representation of internal farmland features while enhancing differentiation at the edges. The experimental results demonstrate that the proposed method effectively considers both internal global consistency and local variations in edge information, thereby ensuring the integrity of farmland plots and the continuity of the farmland edges. The quantitative evaluation of the dataset shows that the model performs well in farmland extraction, with overall accuracy (OA) and intersection over union (IoU) reaching 95.41% and 93.74% on the GF-2 dataset and 94.75% and 88.28% on the JL-1 dataset.
Swarn S. KalsiJames G. StoreyGrant A. LumsdenDuleepa Thrimawithana...
1-5页
查看更多>>摘要：Remote sensing rural building detection holds substantial practical value for the scientific management and unified planning of rural land. However, most existing methods struggle to achieve desirable feature representations due to the similarities and imbalances between underconstruction buildings (UBs) and completed buildings (CBs), as well as interference from background noise, which results in high rates of false positives and false negatives. To address these issues, we propose multiconstrained heterogeneous deep network (MHDN) for remote sensing rural building detection. Specifically, we propose a grid-based CNN-GNN hybrid (GCGH) model that incorporates the sparse connectivity graph into the CNN backbone to model global feature correlations for more robust feature representations. Furthermore, a cross-image multiscale contrastive constraint (CMCC) branch is introduced to supervise network training alongside the detection loss, which facilitates detector learning in the presence of category imbalance. Experimental results on our proposed dataset demonstrate that our MHDN outperforms state-of-the-art object detection methods. The code and dataset are available at https://github.com/Dongxu-Wang/MHDN.
查看更多>>摘要：Plenty of effort has been devoted to solving the nonlinear radiation distortions (NRDs) in planetary image matching. The mainstream solutions convert multimodal images into “single” modal images, which requires building the intermediate modalities of images. Phase congruency (PC) features have been widely used to construct intermediate modalities due to their excellent structure extraction capabilities and have proven their effectiveness on Earth remote sensing images. However, when dealing with large-scale planetary remote sensing images (PRSIs), traditional PC features constructed based on the log-Gabor filter take considerable time, counterproductive to global topographic mapping. To address the efficiency issue, this work proposes a fast planetary image-matching method based on efficient PC-based feature transform (EPCFT). Specifically, we introduce a method to calculate PC using Gaussian first- and second-order derivatives, called efficient PC (EPC). Different from the log-Gabor filter, which is sensitive to structures in a single direction, $\rm EPC$ uses circularly symmetric filters to equally process changes in all directions. The experiments with 100 image pairs show that compared with other methods, the efficiency of our method is nearly doubled without loss of accuracy.
NETL
NSTL
IEEE