基于深度学习的航拍电网绝缘子缺陷检测  

Defect detection for grid insulator using aerial image based on deep learning

作　　者：杨肖辉 绳飞 薛鹏 谷峰颉 米新 YANG Xiao-hui;SHENG Fei;XUE Peng;GU Feng-jie;MI Xin(Urumqi Power Supply Company,State Grid Xinjiang Electric Power Co.,Ltd.,Urumqi 830000,China)

机构地区：[1]国网新疆电力有限公司乌鲁木齐供电公司,乌鲁木齐830000

出　　处：《信息技术》2020年第4期37-40,45,共5页Information Technology

基　　金：国家自然科学基金(41574072)。

摘　　要：电网绝缘检测缺陷的准确检测是电网运行状态有效监测及故障诊断前提,基于无人机航拍电网绝缘子图像,为解决深度学习缺陷检测存在的误检测和局部信息丢失问题,提出基于改进深度学习全卷积网络的缺陷自动检测算法,算法通过改进FCN的VGG结构、扩展滤波器尺寸、取消全连接层Dropout及模型深度,实现FCN模型在绝缘子缺陷检测方面的有效改进,实验结果表示,改进模型在较少运行时间增加基础上,有效提高了对绝缘子缺陷检测的性能和对背影的鲁棒性,取得了比已有算法更有优势的检测结果。Accurate detection of grid insulation detection defects is the prerequisite for effective monitoring and fault diagnosis of grid operating conditions.Based on the underwater vehicle insulator images of UAV,in order to solve the problem of false detection and local information loss in deep learning defect detection,an automatic defect detection algorithm based on improved deep learning full convolution network is proposed.In the proposed algorithm,the FCN model is improved in the detection of insulator defects by improving the VGG structure of the FCN,expanding the filter size,eliminating the full connection layer Dropout and the model depth,which realized the effective improvement of the FCN model in the detection of insulator defects.The experimental results show that,the proposed model improves the performance of the insulator defect detection and the robustness under the less increase of running time and obtains better detection results than the existing algorithms.

关 键 词：深度学习 电网绝缘子 缺陷检测 全连接网络模型 

分 类 号：TM726[电气工程—电力系统及自动化]