An adaptive target tracking method based on motion vector

An adaptive target tracking method based on motion vector is proposed to enhance the veracity of dynamical model which use Brownian motion to model the dynamics of the state of the target. At first, motion vectors given by motion estimation were calculated using statistic method. Then the statistics result was used to adjust the parameter of the dynamical model adaptively. Lastly the state of the target was inferred using dynamical model and particle filter. The method proposed in this paper adjust the parameter of dynamical model adaptively to predict the dynamic of the target state more exactly from the compressed code directly, and can improve the veracity and stability of the tracker without increasing the computation complexity. Numerous experiments demonstrate the effectiveness of the proposed method which can model the dynamic of the target state exactly and can make the tracker more robust to the change of the target movement.