两阶段混合粒子滤波的目标跟踪

基于粒子滤波的目标跟踪,跟踪的成功率和精度与目标运动速度和算法的粒子数密切相关.较大的粒子数能够跟踪速度更快的目标,同时提高跟踪的精度,但会降低算法的实时性.为了解决这个问题,提出一种两阶段混合粒子滤波算法,在第一阶段中,利用少量粒子基于距离角度模型对目标的位置进行粗略估计.在第二阶段中,利用均值偏移算法对目标位置进行精确估计,同时利用粒子滤波对均值偏移的窗口进行自适应调整.实验表明,提出的两阶段混合粒子滤波算法,不仅能够实时地跟踪尺寸变化的目标,而且能够跟踪运动速度快的目标.

Two-Step Mixed Particle Filtering for Target Tracking

The success rate and accuracy of target tracking via particle filtering are closely related to target speed and number of particles. More particles will enable the algorithm to track target faster, while the real-time performance drops. To solve the problem, a new two-step mixed particle filtering algorithm was proposed. During the first step, based on the distance-angle model, some particles are used to estimate the target position roughly. During the second step, the mean shift algorithm is used to estimate the target position accurately and, meanwhile, the particle filter is used to regulate the windows' size of mean shift adaptively. Experimental results showed that the proposed algorithm can track deformed target in real time at high speed.