Transmission Line Hidden Danger Detection Based on Attention Model

In order to improve the efficiency of transmission line inspection and solve the problem of low accuracy of overhead hanging objects and bird nests detection, a transmission line hidden danger detection algorithm integrating SE attention model into YOLOv5 network is proposed to obtain channel-level global features, improve the sensitivity of the model to channel features, and increase the accuracy of overhead overhang and bird's nest detection. Extensive experiments were conducted on a set of transmission line hazard images, and the results show that the YOLOv5 network with the attention model has an average accuracy of 84.2% for overhead overhangs and 87.4% for bird nests, and the mAP value detected by the proposed method is 2% higher than that directly using the YOLOv5 algorithm.