基于扩展Meanshift电气设备发热故障区域提取方法

在红外电气故障监测中,设备过热故障区通常表现出高亮度等特点,然而现有的提取方法因红外图像固有的低对比度、边界模糊等特点容易引起提取区域范围增大.为此,本文提出了一种基于扩展的均值漂移聚类的红外图像故障过热区域提取方法,引入了基于邻域灰度的权重因子.同时,为了提升聚类效率,摒弃了传统均值漂移算法遍历整个红外图像平面进行迭代聚类的方法,融合了自高向低的聚类阈值分割机制,使得改进的均值漂移算法能快速地将故障区域像素进行聚类,实现红外图像中热故障区域的有效提取.实验结果表明本文区域提取性能优于目前现有的一些方法以及传统均值漂移聚类方法.

Fault Region Extraction of Electrical Equipment in Infrared Image by Using an Extended Mean Shift Method

In infrared images,fault regions in electrical equipment often appear at high brilliance.However,some existing methods to extract the region may be influenced by characteristics of infrared imaging,such as low contrast and boundary blur,thus obtaining a larger region than the real fault region.We propose an extended mean shift algorithm by introducing the weight factor associated with the neighboring pixels.To simultaneously improve the efficiency of the clustering method,the original mean shift clustering method to iterate in the image is abandoned.The threshold segmentation mechanism from top to down is merged in the mean shift algorithm.The speed of the clustering pixels is increased,and the fault region is extracted effectively.Experimental results show that the region extraction performance of our method is better than some existing methods and the traditional mean shift clustering method.