针对传统的粒子滤波采用系统转移概率作为建议分布,不能利用当前观测信息.提出了一种结合集合卡尔曼滤波的粒子滤波跟踪方法.对每个粒子产生一个采样子集,使用集合卡尔曼滤波结合当前的观测信息构造建议分布,依据新的建议分布对粒子进行采样.同时在跟踪过程中对于遮挡现象给出了判断和解决方法.实验结果证明该方法提高了粒子滤波估计的准确性,相对于传统粒子滤波和其他粒子滤波方法有更好的稳定性.
The conventional particle filter uses system transition as the proposal distribution. In order to improve the performance of particle filters for target tracking, a sub-set of each particle is first sampled and then the Ensemble kalman filter is proposed to construct proposal distribution. If the target is occluded, the method based on similarity of the sub-block is used to judge occlusion and keep tracking. Experimental results show that the proposed algorithm improves the stability of object tracking and increases the estimation accuracy.