Improved YOLOv11 for transmission line foreign object detection based on multibranch feature fusion and hybrid attention

Abstract

Aiming at the problems of variable scales, strong background interference and high miss rate of small objects in transmission line foreign object detection, an improved YOLOv11 algorithm fusing the multi-branch feature extraction network MB-FEN and the CBAM attention mechanism is proposed. The algorithm replaces the original Neck structure with MB-FEN to enhance multi-scale feature extraction. The CBAM module is embedded to focus on key features of foreign objects and suppress background interference. Experimental results show that the detection accuracy of the proposed algorithm is significantly better than that of YOLOv5, YOLOv8 and the baseline YOLOv11. It can meet the actual requirements of UAV inspection and provide reliable technical support for transmission line foreign object detection.