电网巡检图像中绝缘子的识别

为提高电网巡检时绝缘子故障诊断的效率,提出一种从巡检图像中识别绝缘子的方法。首先将绝缘子彩色图像从RGB空间转换到HSI空间,提取出S分量,然后基于形态学算法改进最佳熵阈值分割算法分割S分量图,再利用图像的灰度信息复原图像并滤波,最后分别计算绝缘子和背景区域的形状特征值,设计分类决策条件。结果表明,该方法能准确识别绝缘子。

Recognition of Insulator String in Power Grid Patrol Images

In order to increase the efficiency of the insulator faults detection,a new method to recognize the insulators in the power grid patrol images was proposed.Firstly,this method converted the insulator images from RGB color space into HSI color space and extracted the S component.Then,the optimal entropic threshold segmentation algorithm was improved based on morphology algorithm to segment the insulator from backgrounds more accurately and the segmentation result was filtered by using the gray value information and the area value of every region.Finally,selected shape features of the insulator and backgrounds were calculated and classified to design the classification decision standards.The example and testing result shows that this method can recognize insulators from patrol images accurately.