Research on Defect Detection of Insulators in Transmission Lines Based on Improved YOLOv5

Abstract

In order to solve the problems of low detection accuracy and high missed detection rate caused by complex background and scale changes in insulator images in aerial photography, this study proposes an improved YOLOv5-CMW algorithm to achieve accurate detection of target insulators. Firstly, utilizing cross local connectivity to achieve multi-scale feature fusion and obtain feature data of the target at different scale levels. In addition, for insulator defects with smaller targets, the backbone network combines a multi-scale adaptive feature attention module to dynamically adjust the importance of feature channels, enhancing the model's ability to represent features. Subsequently, the WIoU loss function was used to train the network, improving the stability of the model and accelerating the convergence process. Experimental data shows that compared with current mainstream single-level detection algorithms, our model can achieve detection accuracy of 94.2%, detection integrity of 93%, and average accuracy of 94.9%. Compared with YOLOv5s, the accuracy has increased by 3.0%, the recall rate has increased by 4.1%, and the average accuracy has increased by 4.5%. This improved technology has shown broad application potential in the field of insulator defect detection in power systems.