Improved mean shift target tracking method based on surf features

The invention belongs to the field of target tracking and particularly relates to an improved mean shift target tracking method based on surf features. The method includes that dynamic video is obtained; a piece of image of a target to be tracked is obtained from the dynamic video; surf features of the image of the target to be tracked are detected and stored in a database; the surf features of the target to be tracked in the database are matched with the surf features of each frame of image in the dynamic video, and when the target to be tracked is detected, the position of the target in the images is determined; a mean shift template is initialized according to a target position area; each frame of image in the dynamic video is subjected to iteration tracking by mean shift till the initialized template is updated again; the template updating frequency is determined according to the tracked target. By means of the improved mean shift target tracking method based on the surf features, the problems that target templates can not be updated by existing mean shift tracking methods and kernel bandwidths can not be changed are solved.