Abstract: In the context of power big data, relying on deep learning and machine vision technology to realize power equipment identification and thermal fault diagnosis from massive infrared image data has become an urgent problem for operation and maintenance work in this field. In this paper, infrared image is taken as the research object, and the research progress of power equipment identification and thermal fault diagnosis in infrared images based on the methods of traditional image processing and deep learning is reviewed. First, the development and technical process of infrared image recognition and thermal fault diagnosis of power equipment are described briefly. Then, the traditional image processing methods for power equipment identification and thermal fault diagnosis are introduced, which are summarized from five aspects: image preprocessing, image registration, image segmentation, feature extraction and classification, and thermal fault diagnosis. The deep learning-based identification and thermal fault diagnosis methods for substation equipment and transmission lines equipment are emphasized. Compared with the traditional image processing method, the deep learning method can identify the thermal fault of power equipment quickly and accurately. Finally, the problems existing in the application of vision technology based on deep learning in power equipment identification and thermal fault diagnosis are pointed out. Based on the existing deep learning technology and recent research ideas, the future research work is prospected.