一种Camshift优化的粒子滤波跟踪算法

针对传统粒子滤波目标跟踪算法中用先验转移概率作分布函数时计算量大、粒子退化严重且未考虑最新观察信息等缺点,提出了一种Camshift优化的粒子滤波跟踪算法.算法首先在粒子滤波框架下,利用Camshift算法使粒子向目标状态的最大后验核密度估计方向移动.然后针对目标所处环境的不同,提出了适时调整参与Camshift算法优化的粒子数的方法,既考虑了跟踪算法的效率又考虑了粒子的多样性.跟踪结果表明,该算法的跟踪性能明显优于传统的粒子滤波算法,具有很好的实时性和鲁棒性.

A Camshift Optimized Particle Filter Tracking Algorithm

Weaknesses of large computational cost, degeneracy problems and current observation lackage are found in traditional particle filter tracking algorithm using prior transition probability as the proposal distribution. Hence, a Camshift optimized particle filter tracking algorithm is proposed. Firstly, particles based on the particle filter are moving towards estimated direction of maximal posterior kernel density of the target state. Secondly, a method of adjusting the number of particles in Camshift according to the environment of the target is proposed. Tracking results show that the tracking property of the proposed algorithm is obviously prior to the traditional one, which having excellent properties of real-time and robustness.