Bistatic SAR Target Recognition Using Transferred Monostatic Knowledge and Cross-Domain Feature Fusion

Bistatic synthetic aperture radar (SAR) systems offer advantages for target recognition but suffer from severe data scarcity, particularly for maritime targets, which hinders the development of data-driven deep learning methods. To address this, this paper constructs a comprehensive electromagnetic simulation dataset containing both monostatic and bistatic SAR images for three ship classes. Leveraging this dataset, we propose a novel target recognition framework that integrates transfer learning with a cross-domain dynamic feature fusion strategy. The core innovation is a hybrid attention-based feature fusion network, combined with domain-adversarial training, designed to effectively transfer and fuse knowledge from abundant monostatic SAR data to assist recognition under limited bistatic samples. Experimental results demonstrate the effectiveness of the proposed method, which achieves an F1-score of 68.08%, outperforming baseline approaches and confirming its capability to enhance bistatic SAR target recognition with scarce annotated data.