多目标实时跟踪的一种数据关联算法

A Real-Time Algorithm of Data Association for Multitarget Tracking

联合概率数据关联是密集杂波环境下跟踪效果最理想的数据关联算法之一。文中在研究概率数据关联算法的基础上,提出了一种快速数据关联算法。通过不同阈值的选择,去除小概率事件,建立确认矩阵,再根据被跟踪目标跟踪门的相交情况,将跟踪空间划分成若干相互独立的区域。对同一区域内公共量测的概率密度值进行衰减,计算出关联概率。仿真实验结果表明,该算法能显著减少可行联合事件的搜索时间和数量,可以有效解决JPDA算法计算量过大的问题,且便于工程实现。

Under most circumstances, it is very difficult for Joint Probabilistic Data Association （JPDA） algorithm to achieve real-time multitarget tracking. We present an algorithm that we believe can do so in most circumstances. In the full paper, we explain our algorithm in some detail. In this abstract, we just add some pertinent remarks to listing the three topics of explanation. The first topic is ： the construction of confirmation matrix. In this topic, we propose an approximation method for constructing the confirmation matrix through removing the small probability events using the right threshold values. To obtain the confirmation matrix, we arrived at an approximation probability matrix through modifying the association probability. The second topic is： the division of independent tracking space. In this topic, we present the mathematical division of the confirmation matrix according to the intersection area of the tracking gate of the targets to be tracked. The third topic is： the computation of association probability. In this topic, we use eqs. （6） through （10） in the full paper to compute the association probability of targets through attenuating the public measurement probability density in the same intersection area. Finally we conduct the Monte Carlo simulation of our improved JPDA algorithm and that of the JPDA algorithm. The simulation results, summarized in Tables 1 through 3 in the full paper, show preliminarily that： （1） compared with JPDA algorithm, our improved JPDA algorithm saves the CPU computing time greatly, namely superior in real-time to JPDA algorithm; （2）the tracking accuracy of our algorithm approximates that of JPDA when targets are moving in various modes of motion.