Abstract: The present disclosure provides a target object tracking method. In this disclosure, the point cloud data sequence in the target scene is collected by lidars and the first frame of cloud point data is extracted; the bounding box of the target object in the first frame of cloud point data is determined using the preset three-dimensional target tracking algorithm. All the target point clouds in the bounding box are extracted to generate the first template point cloud. The template expansion processing is performed using the preset ATG algorithm to obtain the second template point cloud. The point cloud features in the search area of other frames of point cloud data are extracted, and are compared with the second template point cloud and the reference template point cloud to obtain the comparison result. Based on the comparison result, the tracking result of the target object is generated and displayed.