Considering that it is difficult to collect vibration signal of the running gearbox bearings for condition monitoring and fault diagnosis,we used infrared thermal image to carry out the jobs.Firstly,infrared thermal images of the two end bearings of the gearbox were collected by infrared thermography,and the infrared thermal image was decomposed into intrinsic mode function through bi-dimensional empirical mode decomposition.Then,the image was processed for image fusion with the principal component analysis method,and the image with enhanced contrast was obtained.Finally,feature parameters of the original image and the enhanced image were extracted,and cluster analysis was used to isolate the bearings of different states.It is proved that the infrared thermal image can accurately diagnose the faults of bearings.