With the rapid development of coal mine enterprises and the increasing complication of coal mine power systems in our country, the fault location and prediction of coal mine power systems become more difficult. Therefore, the realization of intelligent diagnosis of faults in mine power systems is very important. Most of faults in mine power systems have an important relationship with temperature. This relationship is embodied in the nonlinear mapping relationship between the temperature and its grey level in infrared images. In view of the poor contrast and unapparent details of infrared images of mine power systems, an infrared image enhancement algorithm based on Nonsubsampled Contourlet Transform (NSCT) is put forward. In the algorithm model, a nonlinear matching function is constructed. The function can be used to implement certain enhancement processing of the edge of an infrared image and suppress the interference of noises. The experimental results show that this method can implement intelligent diagnosis and location of the overheating faults in mine power systems and provide solutions for check.