Improved YOLOv5 Algorithm for Power Insulator Defect Detection

Abstract

Intelligent inspection of transmission lines by UAVs has become the mainstream of the industry, and insulator defect detection is a key part of the intelligent inspection operation. To address the problem of low accuracy of insulator defect detection in complex environments, this paper proposes an improved YOLOv5s-based insulator defect detection algorithm. First, the K-means algorithm is used to cluster the data set to obtain the best anchor frame size, which effectively improves the generalization ability and localization accuracy of the model; second, the Backbone part of YOLOv5s is embedded with the Coordinate Attention module to improve the feature extraction ability of the network to solve the influence of invalid features on the recognition accuracy; finally, the EIOU-Loss is used to improve the accuracy of insulator defect detection. Finally, the performance of the model is optimized using the EIOU-Loss function, and ablation experiments are set up to validate the proposed method. The experimental results show that the Precious and mAP of the improved YOLOv5s model are improved by2.S% and 1.6%, respectively, compared with the original YOLOv5s network.