Common electrical equipment includes transformers, switchgears, and circuit breakers, which are composed of multiple components. In this study, the identification of these components was realized via infrared thermal imaging of such devices. Based on the characteristics of infrared thermal imaging with less information, a variety of algorithms have been used for fusion. First, based on the Lab model, a combination of improved K-means clustering and morphology was used to extract the high-temperature region in the infrared image, which guaranteed efficiency and reliability. Second, a combination of improved SURF and perceptual hash algorithms was used to determine the three-phase components in the extracted area. The role of SURF was to compare the visible image of the known electrical device with all the images in the extracted area to determine the area with the most matching feature points in the infrared image. Compared with other infrared regions, we found two regions with the highest matching degree in other regions via the perceptual hash algorithm to locate the three-phase devices in the infrared image. This study is applicable to infrared image recognition and positioning without a large number of image data sets and provides ideas for the extraction of fault information of electrical equipment based on infrared imaging.