一种简化的联合概率数据关联算法

Improving JPDA(Joint Probabilistic Data Association) Algorithms

从分析联合概率数据关联(JPDA)算法的确认矩阵入手,建立了计算聚概率矩阵的近似公式,并根据聚概率矩阵中元素的大小,重新定义了新的确认矩阵,使可行联合事件的数目显著减少,有效地解决了联合概率数据关联算法可行联合事件过多、计算负荷过大以及实时性能差的问题。理论分析和MonteCarlo仿真表明,该算法具有较大的工程应用价值。

JPDA algorithm and its several improved versions[1-3] suffer, the serious shortcoming that feasible joint events are still too many and that the calculation load is still too big and cannot be completed in real-time system. We present our improved version of JPDA, SJPDA (simplified JPDA), which appears to be much better than previous JPDA algorithms. We explain in much detail the following topics: (1) the approximate calculation of cluster probability matrix; (2) the method of establishing the validation matrix; (3) the calculation of the probability of joint events. We performed for 100 times Monte Carlo simulations for each of three algorithms-Fitzgerald JPDA[1], Cheng JPDA[3], SJPDA-and made statistical analysis for each algorithm. The statistical averages of correct association probability for Fitzgerald JPDA, Cheng JPDA, and SJPDA are respectively 86%, 93%, and 92%, showing that the tracking precision of SJPDA is nearly the same as that of Cheng JPDA but the tracking precision of Fitzgerald JPDA is worst. The statistical averages of calculation time for Fitzgerald JPDA, Cheng JPDA, and SJPDA are respectively 7.2 s, 13.5 s, and 7.3 s, showing that the calculation time of SJPDA is a little higher than that of Fitzgerald JPDA but much lower than that of Cheng JPDA.