Radar moving target detection technology is always a key technology in the field of radar signal processing. The traditional radar moving target detection technology is only suitable for uniformly moving targets, and the detection performance is limited. This paper proposes a radar Moving Target Detection(MTD) method based on Convolutional Neural Network(CNN) time-frequency processing. It extracts the Doppler shift information from the radar moving target echo, and then transforms it into time-frequency graph with short-time Fourier transform. After inputting the time-frequency graph into the CNN, the characteristic learning is performed to achieve the purpose of detection and classification. Simulation shows that this method is superior to traditional moving target detection methods.