改进混合YOLO的配电线路瓷瓶缺陷识别算法  

Improved Hybrid YOLO Algorithm for Insulator Defect Identification in Transmission Lines

作　　者：杨增祥 陈琳 叶建慧 YANG Zengxiang;CHEN Lin;YE Jianhui

机构地区：[1]国网浙江省电力有限公司温岭市供电公司,浙江台州318000

出　　处：《电力系统装备》2023年第6期62-65,共4页Electric Power System Equipment

摘　　要：利用无人机在线路上空开展配电线路通道巡视可大幅提高巡检效率。文章将用于瓷瓶检测的YOLO模型和用于瓷瓶分类的卷积神经网络相结合,提出了一种优于标准模型的混合YOLO改进方法,对无人机拍摄的配电线路瓷瓶图像进行缺陷识别,相比VGG、ResNet、DenseNet、YOLOv5等架构有更优异的表现。Using UAV to carry out transmission line channel inspection over the line can greatly improve the inspection efficiency.This paper combines the YOLO model used for insulator detection and the convolution neural network used for insulator classification,and proposes a hybrid YOLO improved method superior to the standard model.Through the defect identification of transmission line insulator images taken by UAV,it has better performance than VGG,ResNet,DenseNet,YOLOv5 and other architectures.

关 键 词：深度学习 电网巡检 目标检测算法 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]