The invention discloses an identification-aided multi-target tracking method based on a depth neural network. Multi-target tracking refers to constant tracking of multiple targets of interest in an image captured by a camera. Tracking may easily fail for the existing multi-pedestrian tracking method when targets are totally blocked or disappear from view. According to the method of the invention, a multi-target tracking problem is converted into a problem of finding the corresponding relationship between targets detected inter frames, effective track recovery is realized based on a deep learning network after targets disappear and reappear, and thus, the effect of long-distance tracking is improved. The multi-target tracking method has good robustness, is very applicable to the field of security monitoring, and is of great practical value to assisting in maintenance of security in public places.