The relative temperature difference method is a quantitative analysis method for infrared image defect diagnosis of power equipment, but it is easy to misjudge or omit the defect of voltage overheat type. Therefore, this paper proposes a sliding window technology combined with the relative temperature difference method and the same kind of comparison method to diagnose and classify different types of faults. The process is that the window slides in the three-phase equipment A, B and C, and the maximum temperature value is selected respectively. The lowest temperature in the same position is taken as the normal temperature of the equipment, while the rest are taken as the hot spot temperature. According to the position of the window, the overheat type of the hot spot temperature is determined. In the process of defect diagnosis, the current transformer is taken as the research object, and the relative temperature difference of hot spot temperature or the temperature difference of the same part of the equipment is calculated according to the corresponding overheat type to diagnose the equipment. Experiments show that this method can automatically identify the defect location, defect type and defect level of CT, and realize the intelligent and efficient defect diagnosis of CT.