Research on transmission line defect identification method based on computer vision

Abstract

This paper proposes a computer vision-based transmission line defect identification method, which establishes a sample data set of typical line defects based on the images and video data provided by the surveillance cameras deployed on the transmission lines. Based on slicing, equalization and broadening of the sample data set, this paper establishes a ResNet-based transmission line defect recognition algorithm, and verifies through testing that the algorithm can accurately and reliably identify line defects such as broken or dispersed wires and foreign object attachment. Based on the identification algorithm established in this paper, the automatic classification and identification of line defects can be realized, the line status can be monitored in real time for 24 hours, and the level of line operation and maintenance intelligence can be improved.