一种改进的广义概率数据关联跟踪算法

Modified generalized probability data association algorithm

对于多目标跟踪问题,最近提出的全局次优的广义概率数据关联算法(GPDA)由于其新颖的可行性划分规则和较小计算存储需求而受到广泛关注。本文提出了一种基于广义联合事件分割组合的新关联算法。它通过引入目标的方向性信息,在基于新规则划分后,对进入有效域的传感器量测估计值权重系数进行调整,从而使最终的估计值更准确,关联精度得到进一步提高。利用该改进算法对杂波环境下多目标跟踪进行仿真实验,结果表明提出的关联算法继承了原有算法的优点,同时用较小的计算代价使得跟踪性能得到较大改善。

Recently, Generalized Probability Data Association （GPDA） algorithm for multi-target tracking is given more attention in virtue of its novel feasible rule and less computation burden and less computing memory. A new association algorithm is presented in this paper, which is based on the segmentation and combination of generalized joint event set. It incorporates the directional information and modifies weighting of state estimation of measurements in the validation region, and then makes the final estimation more exact and improves further performance. Some simulations are made to track multiple maneuvering targets in cluttered environment. The results show that, the proposed algorithm keeps the advantages of former algorithm, and achieves a more significant improvement at the cost of small computational increase.