A novel auto-camshift algorithm used in object tracking

Camshift (continuous adaptive mean shift) algorithm uses color distribution information to track moving object, it can deal well with illumination changes, shadows, occlusions, but it suffers the need of initialization, and canpsilat deal with clutters well. To cope with these problems, auto-camshift (automatic continuous adaptive mean shift) algorithm was put forward to track moving object without manual intervention, it retrieved the initial track window by difference in frame. To deal with clutters due to the possible object-colored targets in the background, a novel face-body structure was utilized to track face-body simultaneously, overcoming the limitation of individual cue. At last, experimental results prove the validity and robustness of auto-camshift.