Multi-view Bistatic SAR Automatic Recognition Study based on Ground Small Target SAR Dataset

MiniSAR systems possess unique advantages in ground target detection compared to traditional spaceborne and airborne SAR systems, mainly in terms of lower cost, real-time performance, and portability. Although numerous studies have focused on target recognition based on SAR systems with unmanned aerial vehicle, most are limited to monostatic SAR systems. Bistatic SAR systems offer stronger stealth capabilities compared to monostatic systems, as they can receive electromagnetic reflections from any angle. However, due to the lower intensity of received electromagnetic signals in bistatic systems, targets in SAR images tend to experience deformation and attenuation. In this study, based on independently developed monostatic and bistatic SAR systems, we study the integration of multi-view and large-scale deep learning models to aiming to address the issue of hard to recognized small targets. Additionally, the paper validates our proposed method on the publicly available MSTAR dataset and the self-developed FAST-Vehicle dataset. Experimental results demonstrate that the proposed multi-view target recognition method effectively identifies SAR bistatic deformed targets.