A lightweight intelligent detection method for SAR images named FCDIS-YOLOv11s

Aiming at the problem that the synthetic aperture radar (SAR) image detection model is difficult to balance the detection accuracy and model lightweight, this study proposes a lightweight SAR image target intelligent detection method based on YOLOv11 s. This method first replaces the backbone network with an efficient FasterNet structure, which significantly reduces the number of model parameters; secondly, the independently developed EMIBC module is innovatively integrated into the C3K2 module, which effectively improves the recognition ability of the model for small targets and multi-scale targets. Thirdly, the dynamic upsampling (DySample) is used to replace the traditional upsampling method to optimize the processing efficiency of the feature fusion stage. Finally, the Inner-SIoU loss function is introduced to replace the original CIoU bounding box loss, which further improves the training effect and feature extraction ability of the model. The experimental results on the HRSID dataset show that the improved model reduces the computational complexity index GFLOPs by 2.79 %, and the detection accuracy index mAP is increased by 7.35 %, which better realizes the balance optimization of model lightweight and detection accuracy.