SFI-Net: a scattering feature integration Network for aircraft target detection in SAR images

Synthetic Aperture Radar (SAR) can provide all-weather, all-time high-resolution detection of aircraft targets, which is crucial for aviation safety. However, detecting aircraft targets in SAR images often faces challenges such as missed detections, false positives, and incomplete detections. To address these issues, we propose a scattering feature integration network (SFI-Net) to enhance the extraction of scattering information in complex backgrounds and improve the ability to distinguish the aircrafts from the surrounding environments. Specifically, we introduce a Peak Enhancement Module (PEM) to strengthen the salient features of aircraft targets, effectively suppressing background noise interference. Additionally, a Spatial Context Enhancement Module (SCEM) is designed, combining spatial attention mechanisms to improve the model’s ability to capture contextual information, significantly reducing false and missed detection rates. Lastly, we introduce an innovative Topological Feature Extraction Module (TFEM) that leverages the stability of scattering topological features to achieve precise aircraft target localization. Experiment on SAR-Aircraft1.0 dataset demonstrates that the proposed algorithm significantly improves the high-precision detection performance of aircraft targets in SAR images.