UPF Tracking Algorithm Based on Color Distribution and Simulated Annealing
Jin Li, Hong Heather Yu, Lulu Zhou, Liang Hong, Lei Wang · 2007
In human body tracking, the quick and reliable position estimation is difficult and plays an important role. A new tracking algorithm for human body tracking, which is the unscented particle filter based on color distribution and simulated annealing (SACD-UPF), is proposed in this paper. Unscented particle filter (UPF) has proven successful for nonlinear and non-Gaussian estimation problem. However, the computational cost of the UPF is very great. Considering the veracity and reducing computational cost, the UPF is combined with color distribution and simulated annealing. Color distribution is applied, as it is robust to partial occlusion, is scale invariant and computationally efficient. And simulated annealing has the characteristics such as robustness and high efficiency, which can economize the computation time. The unscented particle filter based on the above two schemes, one is color distribution and the other is simulated annealing, which are used to track the human body movement. Experimental results indicate that the proposed algorithm is robust and practical.