Recently, deep learning has greatly promoted the development of detection methods for ship targets in synthetic aperture radar (SAR) images. However, existing detection networks are mostly based on large-scale models and high-cost computations, which require high-performance computing equipment to realize real-time processing and limit their hardware transplantation to onboard platforms. To address this problem, a lightweight ship detection network via YOLOX-s is proposed in this paper. Firstly, we remove the computationally heavy pyramidal structure and build a streamlined network based on a one-level feature for higher detection efficiency. Secondly, to expand the limited receptive field and enhance the semantic information of a single-feature map, a residual asymmetric dilated convolution (RADC) block is proposed. Through four branches with different dilation rates, the RADC block can help the detector to capture various ships in complex backgrounds. Finally, to tackle the imbalance problem between ships of different scales in the training stage, we put forward a balanced label assignment strategy called center-based uniform matching. To verify the effectiveness of the proposed method, we conduct extensive experiments on the SAR Ship Detection Dataset (SSDD) and High-Resolution SAR Images Dataset (HRSID). The results show that our method can achieve comparable performance to general detection networks with much less computational cost.

Speckle noise seriously affects synthetic aperture radar (SAR) image application. Speckle suppression aims to smooth the homogenous region while preserving edge and texture in the image. A novel speckle suppression method based on the combination of total variation and partial differential equation denoising models is proposed in this paper. Taking full account of the local statistics in the image, a quantization technique—which is different from the normal edge detection method—is supported by the variation coefficient of the image. Accordingly, a quantizer is designed to respond to both noise level and edge strength. This quantizer automatically determines the threshold of diffusion coefficient and controls the weight between total variation filter and anisotropic diffusion partial differential equation filter. A series of experiments are conducted to test the performance of the quantizer and proposed filter. Extensive experimental results have demonstrated the superiority of the proposed method with both synthetic images and natural SAR images.

Interrupted-sampling repeater jamming (ISRJ) is a new kind of coherent jamming for linear frequency modulation (LFM) signals. Based on digital radio frequency memory (DRFM), ISRJ can generate multiple false target groups by intercepting, storing, and retransmitting radar signal fragments, which significantly affects the postprocessing results of radar systems. Furthermore, due to the fragment interception of ISRJ, ISRJ false targets present a regular and discontinuous time-frequency (TF) distribution in contrast with real targets. Considering this intrinsic property and the coherent nature of ISRJ, this study proposes a new method based on TF analysis and target sparse reconstruction to address the ISRJ suppression issue. In this method, the echo signal is first sparsely represented to obtain both the real and false target positions. Then, according to the acquired target positions, information entropy features of targets are extracted in TF data for subsequent target identification. Finally, guided by the identification result, the real targets can be retained and reconstructed by adaptive filtering in the sparse domain to realize ISRJ suppression. Simulations have validated the effectiveness of the proposed method under various situations.

Radio frequency interference (RFI) poses major threats to synthetic aperture radar (SAR) systems. Due to the suppression of useful target signals via high-power RFI, the SAR imaging quality is severely degraded. Nevertheless, existing studies on RFI mitigation mainly focus on narrowband filtering, while wideband RFI mitigation methods are relatively lacking and perform non-robustly. In this paper, an RFI mitigation scheme is proposed based on instantaneous spectrum forward consecutive mean excision (FCME), which is suitable for both narrowband and wideband RFI mitigation. The SAR echo signal is first transformed into a time–frequency (TF) domain through a short-time Fourier transform (STFT). On this basis, the instantaneous spectra polluted via RFI are detected via a kurtosis-based statistical test and then filtered via FCME to achieve RFI mitigation. Finally, connected component analysis is applied as a safety measure so as to avoid the unnecessary loss of useful target signal. The combination of FCME and connected component analysis enables the proposed method to thoroughly filter out RFI while retaining more useful target signals compared with other competing methods. The experimental results on real SAR raw data validate the effectiveness of the proposed method.

Deep learning has achieved significant success in various synthetic aperture radar (SAR) imagery interpretation tasks. However, automatic aircraft detection is still challenging due to the high labeling cost and limited data quantity. To address this issue, we propose a multi-stage domain adaptation training framework to efficiently transfer the knowledge from optical imagery and boost SAR aircraft detection performance. To overcome the significant domain discrepancy between optical and SAR images, the training process can be divided into three stages: image translation, domain adaptive pretraining, and domain adaptive finetuning. First, CycleGAN is used to translate optical images into SAR-style images and reduce global-level image divergence. Next, we propose multilayer feature alignment to further reduce the local-level feature distribution distance. By applying domain adversarial learning in both the pretrain and finetune stages, the detector can learn to extract domain-invariant features that are beneficial to the learning of generic aircraft characteristics. To evaluate the proposed method, extensive experiments were conducted on a self-built SAR aircraft detection dataset. The results indicate that by using the proposed training framework, the average precision of Faster RCNN gained an increase of 2.4, and that of YOLOv3 was improved by 2.6, which outperformed other domain adaptation methods. By reducing the domain discrepancy between optical and SAR in three progressive stages, the proposed method can effectively mitigate the domain shift, thereby enhancing the efficiency of knowledge transfer. It greatly improves the detection performance of aircraft and offers an effective approach to address the limited training data problem of SAR aircraft detection.

Speckle noise exists inherently in the synthetic aperture radar (SAR) image. Its multiplicative property leads to lots of difficulties in SAR image processing. A novel guidance-aided triple-adaptive Frost filter is proposed in this paper, which has potential for real-time processing platforms. Firstly, a scale-adaptive sliding window sizing method is adopted to determine the neighborhood ranges for every point in the image. All the subsequent processing is based on it. Then, an adaptive calculation for the tuning factor in the Frost filter is embedded into the proposed method. Lastly, the feature information apertured from the original image is used to provide guidance for edge recovery automatically, which guarantees the satisfactory ability for feature preservation. Thus, a novel improved Frost filter is proposed with triple adaptabilities. Both the positioning accuracy and response sensitivity of the scale-adaptive sliding window sizing method are verified first. The superiority of the adaptive tuning factor combined with the scale-adaptive sliding window is confirmed by two comparison experiments. At last, the results of speckle suppression experiments on the synthetic images and two natural airborne SAR images present a better performance than other methods.