The focusing of bistatic synthetic aperture radar (BiSAR) data is more challenging than the traditional monostatic counterparts because of the strong range-azimuth coupling of echo signal, and the range cell migration (RCM) and Doppler frequency modulation (FM) rate, which are caused by complex geometric configuration. Although several monostatic algorithms have been modified to handle general bistatic cases, these algorithms are derived from the assumption that the flying platforms are moving along a linear trajectory with uniform velocity. In practical situation, the flight path of the spaceborne SAR platform inevitably deviates from the ideal trajectory in the long integration time. In this case, besides the influence of the inherent geometric configuration of BiSAR, the curved trajectory of the platforms also causes an additional range-azimuth coupling and the spatial variance of RCM and Doppler FM rate, which cannot be processed by the traditional algorithms. In this article, considering the curved orbit of the SAR platforms and the motion of ground targets caused by the Earth's rotation, a high-order motion range model is proposed. Based on the range model, the spatial variance characteristic of the BiSAR with curved trajectory is analyzed. Then, an extended azimuth nonlinear chirp scaling (EANLCS) algorithm with an addition of highly varying residual Doppler centroid correction for BiSAR with curved trajectory is proposed. Simulation results show the effectiveness of the range model and the modified algorithm.

The advent of deep learning has significantly propelled the utilization of neural networks for Synthetic Aperture Radar (SAR) ship detection in recent years. However, there are two main obstacles in SAR detection. Challenge 1: The multiscale nature of SAR ships. Challenge 2: The influence of intricate near-shore environments and the interference of clutter noise in offshore areas, especially affecting small-ship detection. Existing neural network-based approaches attempt to tackle these challenges, yet they often fall short in effectively addressing small-ship detection across multiple scales and complex backgrounds simultaneously. To overcome these challenges, we propose a novel network called SwinT-FRM-ShipNet. Our method introduces an integrated feature extractor, Swin-T-YOLOv5l, which combines Swin Transformer and YOLOv5l. The extractor is designed to highlight the differences between the complex background and the target by encoding both local and global information. Additionally, a feature pyramid IEFR-FPN, consisting of the Information Enhancement Module (IEM) and the Feature Refinement Module (FRM), is proposed to enrich the flow of spatial contextual information, fuse multiresolution features, and refine representations of small and multiscale ships. Furthermore, we introduce recursive gated convolutional prediction heads (GCPH) to explore the potential of high-order spatial interactions and add a larger-sized prediction head to focus on small ships. Experimental results demonstrate the superior performance of our method compared to mainstream approaches on the SSDD and SAR-Ship-Dataset. Our method achieves an F1 score, mAP_0.5, and mAP_0.5:0.95 of 96.5% (+0.9), 98.2% (+1.0%), and 75.4% (+3.3%), respectively, surpassing the most competitive algorithms.

Target detection based on time reversal effectively uses multipath effects to achieve the space-time focusing effect and enhance radar detection performance by adaptively matching the propagation channel. In the context of ultra-wideband (UWB) radar applications, detection targets are often considered as extended targets comprised of multiple scattering points. The echoes from these scattering points carry rich information about the target’s characteristics, and thus the target’s response can not be ignored. In this paper, considering the target’s response characteristics, a single-station radar target detection method combining time-reversal (TR) and time-frequency (TF) analysis algorithms is proposed for complex environments characterized by clutter, multipath signals, and target responses. In the conventional time-reversal algorithm, we incorporate time-frequency analysis and optimize the energy distribution of the time-reversed transmission signal by comprehensively analyzing the time-frequency domain information of the target echo signals. Consequently, we propose a time-reversal and time-frequency likelihood ratio test (TFTR-LRT) detector that leverages target response characteristics to improve detection probability. In addition, we compare the proposed detector with several conventional detectors under the same conditions. Theoretical derivations and Monte Carlo simulation results demonstrate that the TFTR-LRT detector significantly outperforms other detectors in a mixed environment of clutter, multipath signals, and target responses.

Foundation models have achieved remarkable breakthroughs across various domains, with the widely use of Masked Image Modeling (MIM) and Self-Supervised Learning (SSL). However, these models lack language comprehension, which limits their ability to generalize in downstream tasks and zero-shot applications. The CLIP (Contrastive Language-Image Pretraining) model overcomes these challenges by integrating textual information and performs exceptionally in remote sensing. However, no such multimodal model exists for the SAR (Synthetic Aperture Radar) domain. To bridge this gap, we propose SARCLIP, a vision-language model specially designed for SAR, which leverages textual supervision to enhance visual representation and achieves fine-grained vision-language alignment through modality alignment. Firstly, to overcome the lack of SAR image-text datasets, we propose the Detection-to-Caption (D2C) algorithm, which transforms heterogeneous format of object detection into diverse captions, ultimately creating the vision-language dataset, MMSAR. Additionally, to address the distortion and information loss caused by downsampling in high-resolution SAR images, we propose AnyReader, a dynamic module capable of processing images of any size through a low-resolution encoder. We validate SARCLIP’s generalization and performance on five downstream tasks: image-text retrieval, zero-shot classification, few-shot classification, linear probing and object counting. Especially, we introduce a new counting benchmark, SARCount. Our results demonstrate that SARCLIP consistently outperforms the CLIP baseline across various model sizes. Impressively, SARCLIP achieved a 7.07% higher average recall on the retrieval task compared to the largest CLIP model. In zero-shot classification, SARCLIP surpasses CLIP by 7.8% in average accuracy across three downstream datasets.

High-frequency surface wave radar (HFSWR) is a tool used for maritime surveillance and plays an important role in sea surface target detection. However, multiple-target detection in HFSWR is a challenging issue due to the presence of various types of clutter (such as sea clutter, ionospheric clutter, and ground clutter) in the actual range-Doppler map. To address this problem, a mixed model is established to describe the statistical characteristics of HFSWR echoes. The mixed model uses the Weibull distribution and the Swerling I fluctuation model to fit the amplitudes of clutter and sea surface target echoes, respectively. This paper proposes a robust constant false alarm rate (CFAR) detector for HFSWR, as the detection performance of many CFAR detectors is unsatisfactory in such detection backgrounds. The proposed CFAR detector constructs the objective function considering the fitting degree of HFSWR echoes while introducing a regularization term based on outlier sparsity to enhance the model’s generalization ability. To more accurately estimate the distribution parameters of detection background, this CFAR detector first employs the maximum likelihood method to estimate outliers and censors cells containing outliers, then uses the remaining units to estimate the parameters. The proposed CFAR detector outperforms conventional CFAR detectors in both Monte Carlo simulation tests and also performs well in measured data tests.

The High Frequency (HF) hybrid sky-surface wave radar has a wide surveillance range and relatively low distance resolution. In real environments, the detection background of the Range-Doppler (RD) spectrum is usually a mixed background of different types of clutter overlaid. Therefore, it is difficult to obtain stable and reliable background information as prior knowledge. In this case, traditional0 Constant False Alarm Rate (CFAR) detectors designed based on specific background models may not be fully adaptable. Our research group found through a large amount of measured data analysis that the actual RD spectrum detection unit has a good fitting effect on the Weibull distribution. Based on measured RD spectrum of HF hybrid sky-surface wave radar, this paper proposes a multi-target CFAR detection algorithm for HF hybrid sky-surface wave radar based on the autoencoder neural network under the Weibull clutter distribution model. This new method harnesses the pattern recognition ability of deep neural networks to estimate the background distribution parameters of detection units in the RD spectrum under different radar configurations and sea conditions, thereby obtaining the detection threshold under the specified false alarm probability and completing intelligent target detection. Finally, by verifying the RD spectrum of the measured data of the non-uniform clutter background of the HF hybrid sky-surface wave radar, the results show that the algorithm can well recognize the targets, and it can also recognize the simulated targets added under different signal-to-clutter ratios. And under the same simulation conditions, our proposed detector has better detection performance.