Abstract: By introducing the image analysis method,this paper proposed a new method of directly extracting the image text features from scalogram of continuous wavelet transform of rotor faults signals.Firstly,the rotor fault experimental rig was used to simulate unbalance,misalignment,rubbing and oil whirling faults,and faults samples were obtained;secondly,the scalograms of typical faults was analyzed,and the sensitivity of digital features of scalograms to faults was studied;finally,the integrated back propagation(BP) neural network was used to carry out the diagnosis based on digital features of scalogram.The results fully show the effectivity of the new digital features of scalogram put forward in this paper.