AGGA-MAT: Angle-Geometric Graph Attention Enhanced Multi-agent Transformer for Collaborative SAR Target Recognition

Synthetic Aperture Radar (SAR) Automatic Target Recognition has proven valuable in military reconnaissance, disaster monitoring, and other applications. However, SAR imaging is sensitive to azimuth angles, causing significant variations in scattering characteristics for the same target under different views. This limits the effectiveness of single-view feature representations in complex scenarios. To address this, we propose a multi-agent collaborative SAR target recognition algorithm enhanced by Angle-Geometric Graph Attention (AGGA) and trained with a contrastive loss function. On the MSTAR dataset, our method achieves 99.23% accuracy in a 10-class recognition task. Experiments demonstrate a strong balance between efficiency and accuracy, supporting real-time decision-making and precision strikes in intelligent swarm systems.