基于协同深度学习的二阶段绝缘子故障检测方法  

Two Stage Insulator Fault Detection Method Based on Collaborative Deep Learning

作　　者：王卓 王玉静[1] 王庆岩[1] 康守强[1] V.I.Mikulovich Wang Zhuo;Wang Yujing;Wang Qingyan;Kang Shouqiang;V.I.Mikulovich(School of Electrical and Electronic Engineering Harbin University of Science and Technology,Harbin,150080,China;Belarusian State University,Minsk,220030,Belorussia)

机构地区：[1]哈尔滨理工大学电气与电子工程学院,哈尔滨150080 [2]白俄罗斯国立大学,明斯克220030

出　　处：《电工技术学报》2021年第17期3594-3604,共11页Transactions of China Electrotechnical Society

基　　金：黑龙江省自然科学基金(LH2019E058);黑龙江省本科高校青年创新人才培养计划(UNPYSCT-2017091);黑龙江省普通高校基本科研业务专项资金(LGYC2018JC022)资助项目。

摘　　要：针对现有绝缘子故障检测模型受航拍图像中复杂背景干扰导致准确率低的问题,提出一种基于协同深度学习的二阶段绝缘子故障检测方法。该方法将全卷积网络(FCN)与YOLOv3目标检测算法相协同,第一阶段,利用FCN算法对航拍图像预处理,设计跳跃结构融合浅层图像特征与深层语义特征,构建8倍上采样的绝缘子分割模型,结合图像像素逻辑运算,实现绝缘子目标的初步分割,避免背景区域对绝缘子故障检测的干扰。在此基础上,第二阶段构建YOLOv3模型进行绝缘子故障检测,以深度神经网络Darknet-53作为特征提取器,借鉴特征金字塔思想,在三个尺度的输出张量上对绝缘子故障进行标记和类别预测,保证模型对不同尺寸的绝缘子故障准确检测。利用K-means++聚类算法优化YOLOv3的锚点框参数(Anchor Boxes),进一步提升检测精度。实验结果表明,基于协同深度学习的二阶段方法能够有效克服复杂背景的干扰,在绝缘子故障检测中平均准确率(MAP)高达96.88%,较原始YOLOv3算法MAP值提升了4.65%。For the problem of low accuracy caused by complex background interference in aerial images of existing insulator fault detection models,a two-stage insulator fault detection method based on collaborative deep learning is proposed.The method combines fully convolutional networks(FCN)with YOLOv3 target detection algorithm.The first stage,the FCN algorithm is used to preprocess the aerial image.The jump structure is designed to fuse the shallow image features and deep semantic features,and an 8-fold up-sampling insulator segmentation model is constructed.Combining with the image pixel logic operation,the initial segmentation of insulator target can be realized to avoid the interference of background area on insulator fault detection.On this basis,the second stage,YOLOv3 model is constructed to detect the insulator faults.The deep neural network Darknet-53 is used as the feature extractor,and referring to the idea of feature pyramid,the insulator faults can be marked and classified on three scales of output tensors to ensure the accurate detection of insulator faults of different sizes.K-means++clustering algorithm is used to optimize the anchor boxes parameters of YOLOv3 to further improve the detection accuracy.The experimental results show that the two-stage method based on collaborative deep learning can effectively overcome the interference of complex background,the mean average precision of insulator fault detection is as high as 96.88%,which is 4.65% higher than the original YOLOv3 algorithm.

关 键 词：绝缘子 故障检测 全卷积网络 YOLOv3 K-means++ 

分 类 号：TM507[电气工程—电器]