Contamination Identification and Classification on Composite Insulator by Visible Light Images

摘要

Pollution flashover occurs more easily on contaminated transmission line insulators, which causes huge losses to the power system. In order to prevent the pollution flashover, the first step is to identify the contamination grade of insulator. This paper identifies the pollution grade through the color characteristic of visible light image. Firstly, the image segmentation method based on the randomized Hough transform method is adopted to achieve the surfaces of the red composite insulators with different contamination grades in the transformer substations of Heilongjiang Power Grid. Then, a total of 36 kinds of characteristic of $R, G, B, H, S$ and $V$ component images are calculated. According to Fisher criterion, the mean and median of S component, which can significantly represent the contamination grade, are selected. Finally, a support vector machine for classification decision is designed. Experimental results show that the identification accuracy of the composite insulators reaches 97.5%, which proves this method can be used for identification of composite insulator contamination grade.