Power equipment identification in infrared image based on Hu invariant moments

CHEN Jun-you1， JIN Li-jun1， DUAN Shao-hui2， YAO Sen-jing2， ZHAO Ling2
（1. College of Electronic and Information Engineering， Tongji University， Shanghai 201804， China；
2. Shenzhen Power Supply Co.， Ltd.， Shenzhen 518010， China）

Abstract： Aiming at the problems of low efficiency and poor real-time by manual of power equipment recognition in power system infrared detection，digital image processing was proposed to realize the efficient and accurate recognition based on the images obtained from the Infrared image. Firstly，the high temperature point was found as seed in power equipment from the message of infrared temperature. The background was removed effectively by region growing method to obtain the binary image of entire equipment. Secondly，Hu invariant moments and its improved algorithm were selected as the methods of feature extraction. Hu invariant moments of binary images were calculated， and feature vectors of power equipment were obtained. Finally，classifier based on BP neural network was designed to achieve different power equipment recognition，which will be used in fault diagnosis with temperature message. The research results indicate that，this method can receive a high recognition rate for different equipment and has less time-consuming，so that it will get a good prospect.
Key words： power equipment identification； infrared image； Hu invariant moments； region growing； BP neural network