Abstract:The identification of power equipment and its key components based on infrared image processing is significant to infrared diagnostic technology,of which the characteristic parameters are difficult to extract as the result of image tilt,image zoom and shape similarity of the device. To solve this problem,the infrared images of five devices with similar shape are studied in this paper,including current transformer,voltage transformer,lightning arrester,isolating switch and circuit breaker. For the device classification and identification,Zernike moments,with the advantage of rotation and scaling invariance,are used as the characteristic features,which are later identified by Relevant Vector Machine (RVM). The experimental results show that the method proposed will not be affected by the target position and tilt angle in the image. Meanwhile,it can customize a large number of high-quality samples and effectively distinguish different devices,whose recognition accuracy reaches 94.7%,verifying the effectiveness and practicability of this method.