基于周期性结构特征的绝缘子掉片检测算法

An algorithm of detecting insulator piece missing by periodic structural featuresGong Hao,Wang Lexi,Feng Zhihui, Liang Wenyong and Yan Biwu

摘要:

由于受强电场、机械应力、污秽、温度以及湿度等影响,绝缘子经常出现内部裂缝、表面破损、绝缘阻抗降低及污闪等现象,进而造成电网停电事故,自动地检测绝缘子缺陷对保障电力网络的安全运行具有重要的意义。本文根据绝缘子相邻伞裙间距的不变性特点,提出一种绝缘子的掉片检测算法,首先,采用基于直方图的自适应分割方法,提取绝缘子区域图像,并进行水平倾斜校正,然后,利用灰度归一化相关匹配方法,估计绝缘子几何结构的周期性参数,最后,利用灰度归一化相关匹配方法,检测绝缘子掉片位置。针对无人机巡检的809张缺陷绝缘子数据集,绝缘子掉片检测的准确率和召回率分别达到95.8%和91.9%。与现有的方法相比,其优点是不需要事先利用大样本进行统计学习,对尺度、旋转、光照、背景以及绝缘子种类的变化均具有很强的适应性。

Abstract:

Due to the influence of strong electric field, mechanical stress, pollution, temperature and humidity, insulators often appear anomalies such as internal cracks, surface damage, insulation impedance reduction and pollution flashover, which may cause serious an accident of power cut, so automatic detection of insulator defects is of great significance to ensure the safe operation of power network. According to the distance invariance between two adjacent umbrellas in an insulator, this paper presents an algorithm of detecting the position of lost umbrellas. Firstly,an image of insulator regions is extracted using an adaptive histogram-based segmentation method(ACHS), and its horizontal tilt is corrected. Secondly, the periodic parameters of the geometric structure of insulators are estimated by gray normalized correlation matching method. Finally, gray normalized correlation matching method is also used to detect the position of missing pieces of insulators. This proposed method is tested on 809 data set for unmanned aerial vehicle inspection, the accuracy and recall of the insulator drop detection are 95.8% and 91.9%, respectively. Compared with the existing methods, the advantage of this method is that there is no need to use large samples for statistical learning in advance and it has strong adaptability to the changes of scale, rotation, illumination, background and the variety of insulators.