An Anchor-free Method for Aircraft Detection in SAR Images Based on Density Map

The accurate detection of aircraft targets in Synthetic Aperture Radar (SAR) imagery is pivotal for aviation safety monitoring and military strategic planning. This paper addresses the challenges in detecting densely arranged aircraft targets within SAR images by introducing a novel detection network based on aircraft density map prediction. Initially, we propose a method for predicting the aircraft density map, which incorporates an innovative coding and decoding approach for aircraft targets, effectively mitigating the issue of overlapping target labels caused by dense distributions. Subsequently, we devise an aircraft density loss function, tailored specifically for the aircraft density map. This method abandons the reliance on preset anchor boxes and significantly enhances the model’s detection performance in scenarios with high-density targets. Comprehensive experimental validation on the SADD public dataset demonstrates that the proposed method exhibits high detection efficiency and accuracy when dealing with small-sized and densely distributed aircraft targets.