基于边缘特征的绝缘子故障检测

Insulator fault detection based on edge features

针对绝缘子图片的一致性较差且绝缘子故障特征的不可预知性问题,提出了基于绝缘子边缘特征的故障检测算法。首先基于纵坐标方向连续单像素边缘提取算法,实现绝缘子骨架提取与伪边缘过滤;然后采用改进的Canny算法,通过基于多峰梯度直方图的阈值分割和边缘跟踪,解决边缘的连通性与伪边缘之间相斥的问题,实现边缘特征的精确提取;最后基于均匀采样的边缘特征点计算角度的算法,通过绝缘子边缘特征的按片匹配,实现绝缘子的缺陷检测。对高速铁路接触网巡检车采集的406张运营线路绝缘子图片进行测试验证,结果表明本算法检测的精确度达到99%,召回率达到98.7%。本方法适用于拍摄正常及过度曝光情况下的绝缘子检测,满足运营维修保养的要求,可为绝缘子故障检测提供参考。

In response to the problems of poor consistency of insulator images and unpredictability of insulator fault features,a fault detection algorithm was proposed on the basis of insulator edge features.Firstly,extraction of insulator skeleton and filtering of false edges were actualized by virtue of a continuous single pixel edge extraction algorithm in the vertical ordinate direction;secondly,the modified Canny algorithm was adopted by using the threshold segmentation based on multi-peak gradient histogram and edge tracking,to solve the problem of repulsion between the connectivity of the edge and the false edge,and to realize the precise extraction of edge features;finally,based on the angle calculation algorithm of edge feature points that are uniformly sampled,the insulator defect detection was performed by matching the insulator edge features according to insulator pieces.The results indicate that by testing and verifying 406 pictures of operating line insulators collected by the high-speed railway catenary,the detection accuracy of the algorithm reaches 99%with the recall rate up to 98.7%.This method is applicable to insulator detection under normal shooting and overexposure conditions,capable of meeting the requirements of operation and maintenance,and providing a reference for insulator fault detection.