An adaptive method of model updating in MS tracking

Mean-Shift is an efficient algorithm in the visual tracking. The target model commonly is RGB of the object, which is composed by different components of target´s color. In the tracking process, the target model should change with the change of the backgrounds. So the updating of target model is demand in mean-shift tracking. Generally the popular updating algorithm was based on the integrity of the RGB model, which can´t satisfy the demands in real time. An adaptive model updating algorithm is proposed, in which the change of each component of target´s RGB model is computed respectively. The components with heavily weight are select and updated. The experimental result shows that algorithm proposed in this paper is more robust in tracking for a long time.