Research and Application of HOG Feature Based Power Grid Key Area Out of Bounds Detection

With the rapid development of electric power industry and the acceleration of the marketization process of electric power system reform, the importance of electric power safety production is more prominent. The traditional electronic fence mostly adopts radio frequency or infrared monitoring, which cannot be accurately identified. False positives will be generated when animals or inanimate objects enter the monitoring area. This paper aims to use interval capture method to extract feature through HOG, PCA and other feature extraction methods in real time, and then use SVM classifier to discriminate for the transgression detection system in key monitoring areas of power grid. In order to achieve the key areas of personnel crossing the precise monitoring.