基于红外图像识别的输电线路故障诊断方法.PDF

第39 卷 第4 期 红 外 技 术 Vol.39 No.4 2017 年4 月 Infrared Technology April 2017 基于红外图像识别的输电线路故障诊断方法 1 1 1 2 王 淼 ，杜 伟 ，孙鸿博 ，张 静 （1. 国网通用航空有限公司，北京 100005 ； 2. 天津航天中为数据系统科技有限公司，天津市智能遥感信息处理技术企业重点实验室，天津 300301 ） 摘要：基于红外图像研究了输电线路的故障诊断方法，首先采用LSD 线段检测法提取红外图像中的 导线，采用深度卷积神经网络提取红外图像中绝缘子，从而得到线路元件区域。在线路元件区域内 进行温度和灰度分析提取发热点，并利用漫水填充算法对过热区域进行分割，提取骨架扫描点数、 有效凸缺陷、引流线对缺陷类型进行识别。实验证明该算法有较高发热点定位准确率和缺陷类型识 别准确率。 关键词：红外图像；输电线巡检；导线提取；绝缘子检测；缺陷点定位；特征提取；缺陷识别 中图分类号：TN219 文献标识码：A 文章编号：1001-8891(2017)04-0383-04 Transmission Line Fault Diagnosis Method Based on Infrared Image Recognition 1 1 1 2 WANG Miao ，DU Wei ，SUN Hongbo ，ZHANG Jing (1. State Grid General Aviation Co. Ltd , Beijing 100005, China; 2. Tianjin Zhongwei Aerospace Data System Technology Co. Ltd, Tianjin Key Laboratory of Intelligent Information Processing in Remote Sensing, Tianjin 300301, China) Abstract ：A fault diagnosis method of transmission line based on infrared image recognition is proposed in this paper. Firstly, LSD(Line Segment Detector) is used to extract the conductor and CNN (convolutional neural network) is used to extract insulator in the infrared image, both conductor areas and insulator areas are thought as line component areas. Heating pixels in component areas are obtained according to their temperature and gray value. Heating regions are segmented applying Flood Fill algorithm. Target accounting, skeleton scanning points, effective convex defects and lead wire are extracted to ide