摘要
针对绝缘子表面热场分布与湿污条件的紧密联系,提出一种基于绝缘子红外热像处理的现场污秽等级检测新方法。污秽检测的基本过程为:对红外图像进行平稳小波域局部自适应去噪抑制成像过程中的白噪声干扰;在对数变换的基础上利用最大类间方差法进行图像分割;利用以二值图像重心坐标为起点的不同角度散射直线来采样绝缘子盘面边缘点;应用最小二乘拟合盘面边缘椭圆方程;提取椭圆内长轴以上图像区域;规格化,统计灰度直方图。按上述过程建立不同湿度条件下各污秽等级绝缘子热像灰度直方图特征库,采用与最近邻湿度样本的灰色综合关联度最大相似准则评定现场污秽等级。实验结果表明了该方法用于绝缘子现场污秽等级检测的可行性和有效性。
In view of the close relationship between thermal distribution and conditions about contamination and wetness on insulators＇ surface, a detection method for on-site pollution level of insulators is proposed based on infrared- thermal-image processing. The procedures of detection were as follows： Firstly, infrared thermal images were denoised with the stationary wavelet-domain local-adaptive method to suppress white-noise interference in irtfrared imaging. Secondly, Otsu＇s method was adopted for image segmentation on the basis of logarithmic transformation to the denoised images. Thirdly, edge points of disc surface of insulators were sampled through scattering lines at different angles from the barycentric coordinates, which were computed from binarized images. Fourthly, the elliptic equation for disc surface edge was fitted by the least square method. Fifthly, the ellipse image region above its long axis was abstracted and normalized.＇ Finally, the gray histograms of the normalized images were calculated. According to above procedures, a feature library was established, including massive infrared-thermal-image gray histograms of various pollution-level insulators under different humidity. On-site pollution level was evaluated by maximum comparability criteria of gray synthetically relational degrees with samples under the nearest humidity condition. Experimental results proved the method＇s feasibility and effectiveness in detection of on-site pollution level for insulators.
出处
《中国电机工程学报》
EI
CSCD
北大核心
2010年第4期132-138,共7页
Proceedings of the CSEE
基金
湖南省科技计划项目(2006GK3043)
长沙市科技计划项目(20077140619)~~
关键词
绝缘子
红外热像
现场污秽等级
平稳小波变换
图像去噪
最小二乘
直方图
灰色综合关联度
insulator
infrared thermal image
on-site pollution level
stationary wavelet transform
image denoising
least square
histogram
gray synthetically relational degree
