A Novel Defect Detection Method for Insulators of Power Transmission Line Based on YOLOv5

Insulators are widely used in power transmission lines. They need to go through defect detection before being used on site. The traditional method is through manual detection, which is inefficient and costly. This paper proposes a novel defect detection method based on YOLOv5 for insulators of power transmission line. First, in order to solve the problem of dataset richness, the insulator defect dataset is enhanced and annotated. Then the data in the defect dataset is converted into a unified format and input into YOLOv5 network for training. The optimal defect detection YOLOv5 network model is obtained by adjusting the learning rate and optimizing the network structure. Finally, the YOLOv5 network model has been compared with SSD, YOLOv4, and faster RCNN network model. The experimental results show that the defect detection method based on YOLOv5 has faster detection speed, higher detection accuracy in defect detection for insulators of power transmission line than SSD, YOLOv4, and faster RCNN network model.