Research on Transmission Line Insulator Defects Detection using YOLOv7
Authors:
- 17 December 2022
- conference paper
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 384-388
Abstract
Electrical component inspection has been a significant problem in the power distribution system. If not thoroughly investigated, poorly connected transmission lines could result in catastrophic power outages or blackouts. Much research on transmission line insulators, transformers, cables, twisted conductors, electric poles, and other power grid equipment and parts has been conducted to prevent such failures. In this work, a total of 1975 insulator images are used as the dataset, and an image augmentation technique is used to overcome the data insufficiency problem. This article’s main contribution is applying the recently proposed YOLOv7 single-stage object detector and its variants for accurately and efficiently detecting outdoor insulator defects. The experimental results show that the object detectors successfully identify insulator string as the primary class and three other subclasses, such as flashover damage, good, and broken insulator shells. The detection accuracy of the YOLOv7x object detector, among different YOLOv7 variants, can reach up to 97.5% with proper hyperparameter tuning.Keywords
This publication has 6 references indexed in Scilit:
- Research on edge intelligent recognition method oriented to transmission line insulator fault detectionInternational Journal of Electrical Power & Energy Systems, 2022
- InsuDet: A Fault Detection Method for Insulators of Overhead Transmission Lines Using Convolutional Neural NetworksIEEE Transactions on Instrumentation and Measurement, 2021
- A Review on State-of-the-Art Power Line Inspection TechniquesIEEE Transactions on Instrumentation and Measurement, 2020
- IN-YOLO: Real-Time Detection of Outdoor High Voltage Insulators Using UAV ImagingIEEE Transactions on Power Delivery, 2019
- Detection of Power Line Insulator Defects Using Aerial Images Analyzed With Convolutional Neural NetworksIEEE Transactions on Systems, Man, and Cybernetics: Systems, 2018
- Automatic autonomous vision-based power line inspection: A review of current status and the potential role of deep learningInternational Journal of Electrical Power & Energy Systems, 2018