Efficient Target Detection of Monostatic/Bistatic SAR Vehicle Small Targets in Ultracomplex Scenes via Lightweight Model

Military operations often demand considerable concealment and raid capabilities, particularly at night or in adverse weather conditions. However, the use of synthetic aperture radar (SAR) technology provides early warning and target localization capabilities. While spaceborne or airborne SAR systems can capture expansive SAR scenes, they frequently encounter challenges in delivering timely and high-resolution data, thereby limiting their effectiveness in detecting small ground vehicle targets. To address this issue, our research has developed a low-cost, high-resolution, and real-time monostatic MiniSAR system for the effective detection of small targets, such as vehicles. Furthermore, to enhance the stealthiness of the MiniSAR, a bistatic MiniSAR system has been developed to accomplish detection tasks. Nevertheless, despite the utilization of MiniSAR systems for ground armored target detection, two primary challenges persist: the presence of highly ultracomplex scene interference making accurate target detection difficult; and poor real-time performance resulting in slow detection and tracking. To overcome these challenges, this article proposes a ground vehicle target recognition method based on an improved lightweight anchor-free detection network using monostatic/bistatic SAR images. The method initially leverages the inherent features of SAR targets for localization, embedding these features into SAR images, and then outputs detection results through the improved lightweight anchor-free network. We validate the effectiveness of this method on our self-constructed monostatic/bistatic SAR datasets and verify the algorithm’s robustness on publicly available ship datasets. Experimental results demonstrate that this method outperforms other representative methods in detecting SAR vehicle small targets, exhibiting higher detection accuracy and timeliness.