摘要
针对自然光照条件下复合绝缘子表面存在阴影或暗区域不能准确分割的情况,提出一种基于改进色差法的复合绝缘子图像分割方法。首先,将采集的图像中复合绝缘子所在区域划分为亮、正常、暗三个部分,分别对亮、暗区域进行光照补偿后,重新合成绝缘子图像;其次,对该合成图像进行颜色分量提取和分析,提出一种改进的色差方法获取最优色差灰度图像;最后,采用Otsu自适应阈值算法结合区域标记和几何形状分析进行分割,从而实现输电线路光照不均复合绝缘子目标的准确识别。实验结果表明：该改进方法可以比较完整且准确分割出绝缘子,为后续绝缘子特征提取、故障分析提供了有利参考;将该改进方法分别与传统阈值分割方法、2R-G-B（红色分量Red、绿色分量Green、蓝色分量Blue）色差法进行比较,三者平均分割误差分别为8.18%,36.12%、27.88%,降低了27.94%和19.70%,尤其对弱光、强光下绝缘子分割效率平均提高30%以上。
In view of shadows or dark regions of composite insulator surface under the condition of natural light, insulator cannot be accurately segmented. Consequently, we put forward a kind of composite insulator images segmentation method based on improved color differences. Firstly, the area of the composite insulator in the collected image is divided into three parts, namely, the bright, normal, and dark. After the light and dark areas are compensated separately, the insulator image is re-synthesized. Secondly, according to color component extraction and analysis, an improved color difference method is presented to obtain the optimal color difference image. Finally, the Otsu adaptive threshold algorithm combined with the region mark and geometric shape analysis is used to segment insulator, so as to realize the accurate identification of the non-uniform illumination composite insulator of the transmission line. The experimental results show that the proposed method can provide a complete and accurate segmentation of the insulator, which provides a useful reference for the subsequent insulator feature extraction and fault analysis. By comparing the method with the traditional threshold segmentation method and the 2 R-G-B（Red, Green and Blue） color difference method, the average segmentation error is 8.18%, 36.12%, 27.88%, respectively, and the segmentation error is reduced by 27.94% and 19.70%, especially for low light and strong light, the insulator segmentation efficiency is increased by more than 30%.
作者
黄新波
张慧莹
张烨
刘新慧
邢晓强
杨璐雅
HUANG Xinbo;ZHANG Huiying;ZHANG Ye;LIU Xinhui;XING Xiaoqiang;YANG Luya(School of Electronics and Information,Xi＇an Polytechnic University,Xi＇an 710048,China;School of Electro-Mechanical Engineering,Xidian University,Xi＇an 710071,China)
出处
《高电压技术》
EI
CAS
CSCD
北大核心
2018年第8期2493-2500,共8页
High Voltage Engineering
基金
陕西省重点科技创新团队计划(2014KCT–16)
陕西省协同创新计划项目(2014XT–07)
陕西省工业科技攻关项目(2016GY–052)
陕西省自然科学基础研究计划(2017JQ6054)
西安工程大学博士启动基金(BS1505)~~
关键词
复合绝缘子
光照不均
同态滤波
色差法
图像分割
composite insulator
non-uniform illumination
homomorphic filter
color difference
image segmentation
