针对粒子滤波跟踪算法计算代价大以及Meanshift跟踪算法容易陷入局部极值等问题,提出一种嵌入均值优化的粒子滤波跟踪算法.该算法根据粒子滤波的运动模型估计目标区域位置,利用Bhattacharyya距离度量粒子区域和目标模型的相似性,并根据相似性来更新粒子权值,使用Meanshift优化算法改善粒子的估计位置,使得这些粒子的候选区域能更加接近目标模板,极大提高了粒子的使用效率.实验结果表明,该算法能够有效进行人的跟踪,处理人的短暂遮挡问题,性能优于粒子滤波算法,有较好的实用性.%Since the particle filter tracking algorithm has huge computation cost and Meanshift tracking algorithm always falls into local extreme value, a new target tracking algorithm based on Meanshift optimization embedded particle filter is proposed. The algorithm estimates the object' s location by motion model, the similarities of these particles is measured by the Bhattacharyya distance , update the weights of these particles, and then Meanshift optimization improves the estimation of these particles, makes these candidate locations of particles more closer to real location of tracking object, the efficiency of particles is greatly increased. Experimental results show that the algorithm can efficiently track the moving gargets and treat the problem of short cover, so it is better than the particle filter and of high practicability.
展开▼
