Attention-Augmented YOLO11 for High-Precision Aircraft Detection in Synthetic Aperture Radar Imagery

Synthetic Aperture Radar (SAR) imaging has emerged as a pivotal technology in remote sensing, offering unparalleled capabilities for high-resolution imaging under all-weather conditions, day or night. This versatility has rendered SAR indispensable for various critical applications, including ground-based aircraft monitoring. Despite its advantages, the complex nature of SAR imagery, characterized by speckle noise and intricate geometric distortions, presents significant challenges for object detection tasks. In response, advanced object detection algorithms have become essential for extracting actionable insights from SAR data. This study focuses on ground-based aircraft detection using the latest addition to the YOLO family, YOLO11, a state-of-the-art object detection algorithm. While YOLO11 exhibits cutting-edge performance in standard detection tasks, its built-in attention mechanisms were found insufficient for addressing the unique complexities of SAR imagery. To overcome this limitation, we augmented the neck architecture of YOLO11 with Enhanced Channel Attention (ECA) modules, achieving significant performance improvements. The ECA-enhanced YOLO11 demonstrated an average increase in precision of 1.7% and mAP50 of 2.30% across various input image resolutions, with gains of up to 3.4% in lower-resolution scenarios $({256}\times {256}\ \text{pixels})$. Furthermore, the F1 score consistently improved by up to 2.40%, showcasing the model's ability to balance precision and recall effectively. The results highlight the potential of integrating ECA modules into detection architectures to mitigate the challenges of SAR imagery, making it possible to achieve more reliable and accurate aircraft detection under complex imaging conditions.