箱粒子广义标签多伯努利滤波的目标跟踪算法
Target Tracking Method with Box？？Particle Generalized Label Multi？？Bernoulli Filtering

针对序列蒙特卡罗广义标签多伯努利滤波(SMC？？GLMB)算法计算效率低、实时性差的问题,提出了箱粒子广义标签多伯努利滤波的目标跟踪(Box？？GLMB)算法。该算法使用带标签的随机有限集描述多目标的状态,包括目标的位置和速度,并且对每个目标用互不相同的标签进行区分;然后利用箱粒子滤波算法近似单目标状态的概率密度,即用一组带权值的均匀分布拟合单目标状态概率密度;最后通过广义标签多伯努利滤波对多目标状态的概率密度进行预测与更新,从多目标状态后验概率密度中估计单目标的位置与速度,根据目标的标签可以实现航迹跟踪。Box？？GLMB算法结合了箱粒子滤波与GLMB算法的优势,能够跟踪目标航迹,同时提高计算效率。仿真结果表明,Box？？GLMB算法可以有效估计目标状态以及跟踪目标航迹,相比于SMC？？GLMB算法,计算效率提升了62％。
A target tracking method with box？？particle generalized label multi？？Bernoulli filtering (Box？？GLMB) is proposed to address the problem that the sequential Monte Carlo generalized label multi？？Bernoulli filter (SMC？？GLMB) has low computation efficiency and weak real？？time performance. A random finite set with labels is employed to describe states of multi？？targets, including positions and velocities of targets, and targets are distinguished by different labels. Then the box？？particle filter algorithm is used to approximate the probability density of single target, that is, the probability density of single target is approximated by weighted uniform distributions. Multi？？target densities are predicted and updated by using the generalized label multi？？Bernoulli filter (GLMB). Target states are estimated from posterior probability density and trajectories of targets are tracked by labels. The proposed method combines the advantages of the box？？particle filter and the generalized label multi？？Bernoulli filter. Box？？GLMB filter is able to track trajectories and to improve computational efficiency at the same time. Simulation results show that the proposed filter effectively estimate states of targets and track trajectories. A comparison with the SMC？？GLMB filter show that the proposed filter increases 62％ computational efficiency