A hybrid multi object tracker using mean-shift and background subtraction

Mean-shift tracking is an efficient method for tracking objects. In this paper, we propose a fully automatic static camera multiple object tracker based on mean shift algorithm. Foreground detection is used to initialize the object trackers. The bounding box of the object is used as a mask to decrease the number of iterations to find the new location of the object. To solve the potential problems due to the changes in objects' size, shape, to handle occlusion, split and to detect newly emerging objects as well as objects that leave the scene, trackers are updated. By using a shadow removal method, tracking accuracy is increased and possible false positives are overcome. As a result, an easy to implement, robust and efficient tracking method which can be used for automated video surveillance applications while solving the problems of standard mean shift tracking and being superior to this method is obtained.