Intelligent Video Analysis System In High-Voltage Test

Intelligent video imaging technologies have been widely used in varieties of applications because of its own advantages like being convenient and efficient. They’re one of the most important steps in the intelligent monitoring of high-voltage test, using which to make an analysis of the extent of icing insulator could provide effective data for the power equipment inspection. Traditional methods measuring high-voltage data have the disadvantages of not being intuitive and automatic. It’s meaningful for improving the efficiency of high-voltage test data measuring to make the analysis directly from the intelligent video images.In this paper, I propose an intelligent video analysis system for automatically computing the bridge percentage between insulators. The target research types of insulator including two types of double umbrella insulators, and the algorithm in the paper will conduct discussion according to their own characteristics. There are three main parts in this paper:image segmentation, classification of iced extent and the measurement based on image.In terms of image segmentation, aimed for the phenomenon of information loss when directly using traditional OTSU algorithm to make segmentation for insulator images, the paper come up with OTSU threshold compensation algorithm. After using OTSU algorithm in the HSI color space and get the original threshold, the new algorithm make compensation for threshold cooperating with image morphology and the largest connected domain method. The segmentation accuracy rate with OTSU threshold compensation algorithm improves by an average of30%than before. In terms of classification of iced extent, for solving the problem of different types of insulator couldn’t have the same solution, the paper extract features of different types of insulator and make classification. Firstly, make discussion about the extraction method of color, grayscale and texture features. The color feature is extracted from the component I and the classification method K-MEANS. The grayscale feature is extracted by computing image grayscale mean and variance value and finally the texture feature is represented by GLCM(Gray Level Co-occurrence Matrix). Secondly, make the classification with the feature mentioned above and SVM algorithm (Support Vector Machine).The experimental results show that the classification accuracy rate is about90%and more.In terms of measurement based on image, the paper make a discussion for two different types of insulator aimed for the requirement of bridge percentage computation. The main method is based on the measurement of the iced part contour of the insulator. Firstly, the paper proposes the automatic detection algorithm of disk-diameter and automatic pairing algorithm for disk-endpoint. Secondly, extract contour respectively according to each characteristics. Finally, the paper conducts contour tracing method based on eight connected region to do the bridge percentage computation. The experimental results show that the accuracy rate is about80%and more.