Meanshift algorithm based on kernel bandwidth adaptive adjust

Object tracking algorithm based on Meanshift algorithm with fixed kernel bandwidth does not realize object tracking correctly with the scale of object changed. According to this, a scheme of kernel bandwidth adaptive adjustment and predictions of object cancroids based on Kalman filter is proposed in this paper. In this algorithm, Object location predicted based on Kalman filter is used to initialize the Meanshift algorithm. The variation tendency of the kernel bandwidth is also determined based on Kalman filter. Experiment results demonstrate that this algorithm can realize the kernel bandwidth adaptive adjustment and object location prediction. The robustness of the tracking algorithm is also enhanced.