基于mean shift和粒子滤波的混合目标跟踪算法

考虑到处理非线性非高斯问题的粒子滤波方法在鲁棒性和速度方面的缺点,利用mean-shift算法找到后验概率的局部最优,用构成新的粒子集合来确定目标的最终位置,在不改变粒子滤波优点的同时提高了跟踪的速度。实验结果表明,这种改进的混合跟踪方法在保证准确性的同时,提高了系统的实时性和鲁棒性。

Tracking algrithm based on mean shift and particle filter

Considering the disadvantages of nonlinear and non-Gaussian particle filter method in the robustness and speed, the paper uses the mean shift algorithm to find the local optimal of posteriori probability, to determine the final goal set position with a new particles and not to change the advantages of the particle filter and improve tracking speed. The experimental results show that the improved mixed tracking method ensures the accuracy and improves the the real time character and robustness simultaneously in the system.