YOLOv5 Transmission Line Fault Detection Based on Attention Mechanism and Cross-scale Feature Fusion

It is an important research direction and a challenging topic in the field of transmission line detection to use UAV for detecting high-voltage transmission lines and automatically and accurately detecting fault targets in the detection data based on computer vision technology. In order to solve the problem that the target to be detected has multi-scale characteristics and partial occlusion in complex inspection environment, a fault detection algorithm of YOLOv5 transmission line based on attention mechanism and cross-scale feature fusion is proposed. First, the YOLOv5 detection network is built. Based on the YOLOv5 detection network, the spatial and channel convolution attention model is introduced to suppress the complex background interference and enhance the significance of the target to be detected. Secondly, the FPN+PAN structure in the original YOLOv5 detection framework Neck is changed to BiFPN structure, so that the multi-scale features of the target can be fused effectively. Thirdly, to address the problems of missing and false detection caused by the insufficient feature expression ability of detection network, we design an adaptive weighted fusion module with multi-scale and same-scale features, which can enhance the detection accuracy of detection network to occluded fault targets. Finally, to verify the effectiveness of the proposed algorithm, the inspection data obtained by an inspection department using UAV in recent four years are used. The results show that the proposed method can accurately detect transmission line faults in complex environment, and the average accuracy of detection can reach 96.8%. ©2023 Chin.Soc.for Elec.Eng.