Geosynchronous Satellite Synthetic Aperture Radar (GEO SAR) plays a crucial role in various fields such as crop growth monitoring, irrigation management, terrain and soil analysis, and natural disaster warning. However, GEO SAR images often suffer from interference caused by moving targets, which affects the accurate interpretation of agricultural scenes. This interference mainly manifests in two aspects: first, the moving targets deviate from their true positions and cause severe defocusing in GEO SAR images; second, the defocusing from the shifted moving targets can obscure the actual agricultural scenes. To address these issues, we analyze the shift and secondary phase errors caused by moving targets based on the non-straight slant imaging geometry of GEO SAR and propose a subaperture based Enhanced Back-Projection (E-BP) imaging algorithm. This algorithm effectively eliminates the interference from moving targets during imaging, restores the obscured background targets, and improves image quality. To solve the problem of separating moving targets, we introduce a semantic segmentation algorithm from deep learning into the BP imaging algorithm, achieving pixellevel segmentation and removal of moving targets in subaperture images. Furthermore, to restore the content of the obscured background images, we perform amplitude complementation using the background information in the subaperture images. By employing coherent fusion technology, we restore the complete full-aperture complex data, successfully generating high-quality background scene images without moving target interference. Finally, we validate the proposed E-BP imaging algorithm using simulated and real data. The results show that the proposed algorithm has significant advantages in improving image quality and accuracy, demonstrating its potential in GEO SAR practical applications.

Semantic segmentation of aerial images is crucial yet resource-intensive. Inspired by human ability to learn rapidly, few-shot semantic segmentation offers a promising solution by utilizing limited labeled data for efficient model training and generalization. However, the intrinsic complexities of aerial images, compounded by scarce samples, often result in inadequate feature representation and semantic ambiguity, detracting from the model's performance. In this paper, we propose to tackle these challenging problems via dual semantic metric learning and multi-semantic features fusion and introduce a novel few-shot segmentation Network (DSMF-Net). On the one hand, we consider the inherent semantic gap between the feature of graph and grid structures and metric learning of few-shot segmentation. To exploit multi-scale global semantic context, we construct scale-aware graph prototypes from different stages of the feature layers based on graph convolutional networks (GCNs), while also incorporating prior-guided metric learning to further enhance context at the high-level convolution features. On the other hand, we design a pyramid-based fusion and condensation mechanism to adaptively merge and couple the multi-semantic information from support and query images. The indication and fusion of different semantic features can effectively emphasize the representation and coupling abilities of the network. We have conducted extensive experiments over the challenging iSAID-5 $^{i}$ and DLRSD benchmarks. The experiments have demonstrated our network's effectiveness and efficiency, yielding on-par performance with the state-of-the-art methods.

Current signal processing for Synthetic Aperture Radar (SAR) cannot satisfy the requirements for high resolution, high squint angle, and wide swath due to the limitations of the traditional stop-and-go model. This model does not take into account the continuous motion between the satellite and the ground, leading to incomplete compensation for the corresponding phase. To address this issue, we first introduce an Accurate Range Equation (ARE) based on the continuous satellite and ground motion, which allows for more precise slant range calculations. Secondly, we derive the point target spectrum according to the accurate range equation, revealing additional phase components not considered in the traditional stop-and-go model. Third, we propose an advanced two-step processing algorithm to compensate for phase errors due to non-stop-and-go continuous motion. In the first-step processing, we eliminate range-independent components in the two-dimensional frequency domain. In the second-step processing, we compensate for residual phase errors related to range and azimuth frequency in the range-Doppler domain. We verified the proposed compensation method through simulation and compared it with the two-step compensation method used in high-resolution SAR data processing. Experimental results show that we provide an effective compensation method for high-resolution, high-squint-angle, wide-swath spaceborne SAR, which can correct azimuth shifts and solve the problem of defocusing at the edges of large scenes.