一种基于随机集的模糊观测的多目标跟踪算法

为解决目标数未知或随时间变化时模糊观测的多目标跟踪问题,将多目标状态和模糊观测数据表示为随机集形式,利用模糊观测的似然函数融合模糊数据,建立了模糊观测的概率假设密度(probability hypothesis den-sity,PHD)粒子滤波方法。这种方法首先利用粒子滤波预测和更新随机集的PHD,然后估计目标数N,最后找出N个PHD最大的点就是多目标的状态估计。在相同的仿真环境中,用这种方法与用重心去模糊器进行去模糊处理后的观测数据同时跟踪目标数变化情况下的多目标,并进行了比较,结果表明,模糊观测多目标跟踪的PHD粒子滤波能稳健跟踪目标数未知或随时间变化时的目标状态和目标数,性能好于去模糊情况。

Random Sets-Based Multi-target Tracking with Ambiguous Measurements

When the number of targets is unknown or varied with time,we represented multi-target state and ambiguous measurements as random sets,used ambiguous likelihood to fuse ambiguous data for tracking multiple targets,and implemented multi-target tracking with probability hypothesis density(PHD) particle filter.Firstly,we used particle filter to predict and update the PHD of Random Sets.Secondly,estimated the number of targets N.Finally,extracted N peaks of PHD.Simulations show that in the same circumstance,our method can robustly tracking multiple targets,performance is better than PHD particle filter with non-fuzzy measurements.