可见光图像颜色特征与支持向量机相结合的绝缘子污秽状态识别方法  

Identification of Insulator Contamination Grade Combining Color Features of Visual Image with Support Vector Machine

作　　者：金立军[1,2] 曹培[1] 胡娟[3] 

机构地区：[1]同济大学电子与信息工程学院,上海201804 [2]西安交通大学电力设备电气绝缘国家重点实验室,西安710049 [3]上海电气集团股份有限公司中央研究院,上海200000

出　　处：《高压电器》2015年第2期1-7,17,共8页High Voltage Apparatus

基　　金：国家自然科学基金项目(51177109);电力设备电气绝缘国家重点实验室资助(EIPE14211)~~

摘　　要：绝缘子污闪事故严重威胁着电力系统的安全运行,高压绝缘子污秽等级的准确识别是防污闪的重要研究内容。笔者提出了一种基于绝缘子可见光图像颜色特征和支持向量机技术,通过建立污秽等级与可见光图像颜色特征值的映射关系,实现绝缘子污秽状态识别的方法。以深圳供电局多个变电站不同污秽度红色陶瓷绝缘子数据为基础,采用改进种子区域生长法分割得到绝缘子盘面区域,提取RGB和HSV颜色空间36种特征,依据Fisher判据,筛选得到能表征绝缘子盘面污秽程度的S均值和S中值;设计支持向量机多值分类器,进行污秽等级划分。实验结果表明,该方法的绝缘子污秽等级准确率达96.67%,实现污秽状态的准确监测,为绝缘子污秽等级识别提供新思路。In respect that the high risk of flashover may cause power outage, it is urgent to realize the safe and accurate monitoring of insulator contamination severity to prevent flashover and ensure operation of power system. In this paper, a method based on visible image features of contaminated insulators and support vector machine is put forward by establishing a mapping between the contamination grade and the color features for the identification purpose of insulator contamination grade. Firstly, the improved image segmentation method based on the seed region growing method is adopted to achieve the discal surfaces of the red porcelain insulators with different contamination grades in the transformer substations in Shenzhen City. Then, thirty-six contamination features are extracted in RGB and HSV color space of the contaminated insulator images and the mean and median of S are selected as the feature values for the insulator contamination grade according to the Fisher criterion. Finally, a multi-class SVM for classification decision is designed. Experimental results show that the identification accuracy of the proposed method reaches 96.67%.

关 键 词：绝缘子 污秽等级 可见光图像 改进种子区域生长法 HSV 支持向量机 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]