The invention discloses a moving object trajectory tracking method. The method comprises the following steps of: firstly obtaining real-time monitoring video information, pre-processing the real-time monitoring video information, dividing each picture into S*S regions, detecting the regions through CNNs of deep learning, and obtaining object types and object coordinates of detected objects; obtaining trajectory information of each object through distance matching, and moving coordinate trajectory chain tables of objects which leave a tracking region through timeout; and finally extracting motion features of each target by utilizing the coordinate trajectory chain table of the target, and outputting the coordinate trajectory chain tables and the motion features. According to the method disclosed by the invention, features of to-be-detected objects do not need to be artificially designed, and good object detection recognition ability can be provided through machine training.