Particle filter is widely used for visual tracking with superior performance in terms of accuracy and robustness, but it suffers from the heavy computational load, and the calculation complexity increases quickly with the state dimension and the number of particles. In this paper, the tracking problem was considered as a coarse-to-fine process to find the optimal state, thus, a hierarchical Kalman-particle filter (HKPF) with precise motion model, called improved hierarchical Kalman-particle filter (IHKPF), was proposed, in which Kalman filter with Jerk model was used to predict a local region around the estimation of global linear motion, and then particles were generated in the local region. The reason for introducing Jerk model was that the inadequate tracking performance of current models with the higher order derivating in the case of very highly maneuvering targets were not tracked therefore, Jerk was added. The high order state variable Jerk was applied in motion model of IHKPF. The HKPF, PF and proposed in the paper were used to compelete track experiment. The experimental results among the proposed algorithm HKPF and PF indicate that Jerk model provides higher accuracy prediction, resulting in well- behaved tracking in complex environment.