Application Of Transmission Line Insulator Defect Detection Algorithm Based On Improved YOLOv5

With the increasing demand for electricity in society,the length of transmission lines is increasing year by year.Regular inspection of transmission lines to ensure their normal operation is of great significance for the smooth operation of the power system.Insulators,as components of overhead transmission lines,play a crucial role in electrical insulation and mechanical support.As an emerging electric power inspection method,UAV inspection is gradually replacing traditional inefficient manual inspection.Aiming at a large number of images taken during UAV inspection,deep learning object detection algorithm is used to detect insulator defects,which can improve inspection efficiency.This paper takes insulator images taken by UAV as the research object to carry out research on insulator defect detection methods.The main research contents are as follows:Firstly,this paper constructs insulator images taken by UAV and obtained from open source website as the dataset used in the experiment,and performs rotation,flip,adding noise and other pre-processing operations on it to enhance the generalization ability of the algorithm.Then,the experimental results of YOLOv5 algorithm and other commonly used object detection algorithms are analyzed.The YOLOv5 algorithm has better detection effect on insulator defects.On this basis,YOLOv5 s with smaller model and faster speed is selected for improvement.In response to the problem of low accuracy in detecting insulator damage defects,the C3 module of the YOLOv5 s backbone network is improved into the Rep Insulator module,which improves the accuracy of insulator defect detection while also improving the detection speed.To address the issue of high similarity between complex backgrounds and insulator objects,which can lead to missed and false detections,an improved bilinear spatial and channel attention mechanism BSC is introduced and added before the detection head.In response to the poor detection performance of multi-scale objects with both insulator strings and insulator defects in the same image,the feature fusion layer of YOLOv5 s is improved to a Bi-FPN structure to improve the algorithm’s detection accuracy for global multi-scale objects.Finally,the three improved methods are integrated together,and the effectiveness of the improved method in this paper is verified by analyzing the experimental results and the detection effect of insulator defects.The experimental results show that the improved YOLOv5s-Rep Insulator+BSC+Bi FPN algorithm has an average precision of 96.7% for detecting insulator self-explosion defects,which is 0.7% higher than before;the average precision of detecting insulator damage defects is 93.9%,which is 7.7% higher than before;the average precision of detecting complete insulators is 97.8%,which is 0.2% higher than before.The mean average precision of the comprehensive detection of the three objects is 96.1%,which is 2.8% higher than the algorithm before improvement.The detection speed on this experimental platform is 147.1frames/second,which is 13.3% higher than before improvement,meeting the task requirements of insulator defect detection.