摘要
绝缘子是输电线路上重要的元件之一,因其工作环境长期暴露在外界,需要承受风吹、日晒、雨雪等因素,难免会出现破损以及严重污秽等缺陷。缺陷绝缘子会危害电力传输系统的正常运行,因此绝缘子缺陷检测工作对供电系统的安全性与稳定性有着重要作用。针对绝缘子表面缺陷问题,提出一种基于自校准卷积YOLOv7的绝缘子缺陷检测方法。该方法以YOLOv7网络模型为基础,通过自校准卷积替换部分常规卷积,增强模型的特征提取能力,扩大感受野,将原模型的CIOU损失函数替换成SIOU损失函数,用于修正检测框的框选能力不足问题。利用现有数据集进行训练、检测,最终比原YOLOv7模型精度提升5%,降低了输电线路绝缘子缺陷误检率,提高了线路巡检效率。
Insulators are one of the important components of power lines.Due to long-term exposure to the outside environment,such as wind,sun,rain,and snow,it is inevitable to have defects such as damage and serious pollution.The operation of defective insulators will probably harm the normal operation of the power transmission system,so the detection of defective insulators plays an important role in the safety and stability of the power supply system.Aiming at insulator surface defects,an insulator defect detection method based on self calibrating convolution YOLOv7 is proposed.Based on the YOLOv7 network model,the method replaces part of the conventional convolution by self-calibrating convolution,enhances the feature extraction capability of the model,expands the receptive field,and replaces the CIOU loss function of the original model with SIOU loss function to correct the insufficient box selection capability of the detection box.After training and detection on the existing data set,the final accuracy is improved by 5%compared with the original YOLOv7 model,which reduces the false detection rate of insulator defects of transmission lines and improves the efficiency of line inspection.
作者
霍一凡
衣丽葵
李佳玉
范乃心
HUO Yifan;YI Likui;LI Jiayu;FAN Naixin(School of Electric Power Engineering,Shenyang Institute of Engineering,Shenyang 110136;Anshan Power Supply Company,Anshan 114000;State Grid Liaoning Ultra-high Pressure Branch,Shenyang 110000,Liaoning Province)
出处
《沈阳工程学院学报（自然科学版）》
2024年第2期54-59,67,共7页
Journal of Shenyang Institute of Engineering:Natural Science
关键词
绝缘子
YOLOv7
输电线路
缺陷检测
自校准卷积
Insulator
YOLOv7
Transmission line
Defect detection
Self-calibrating convolution
