Camshift object tracking method with unscented Kalman filtering motion estimation

In order to implement real-time tracking of motion object in complex scene of image sequence, a novel tracking method based on Unscented Kalman filtering motion estimation with Camshift theory was introduced. Above all, the motion state of selected object in present frame was estimated with Unscented Kalman filtering, and then the possible position of selected object in next frame was predicted. Finally, the tracked object center was search in adjacent area of the possible position and the object was oriented with histogram color matching principia of Camshift method. Experiment results show that the proposal algorithm can not only effectively solve object sheltered problem induced by sudden motion, but also the tracking performance owing to pose movement is improved. It is very useful for machine controller to provide more precise coordinate error indication.