主导概率数据关联及算法

MULTITARGET TRACKING USING DOMINANT PROBABILITY DATA ASSOCIATION

本文给出主导概率数据关联（ＤＰＤＡ）及算法。定义主导联合事件和主导概率，考虑实际目标环境中可行联合事件出现为非等概率，导出主导概率数据关联计算方法。它无需计算所有可行联合事件的联合概率，从而克服了联合概率数据关联（ＪＰＤＡ）的组合问题。本文证明了ＤＰＤＡ性能上限不低于ＪＰＤＡ，下限不低于ＰＤＡ，而下限出现的概率极小。本文给出的ＭｏｎｔｅＣａｒｌｏ仿真表明ＤＰＤＡ算法具有很好的性能和较高的实用价值。

A new algorithm of probability data association for multitarget tracking, the Dominant Probability Data Association (DPDA), is presented in this paper. In DPDA, we consider the fact that the probabilities of the joint events occuring in practical enviroment are not equated. We define the joint events with great occuring probabilities are dominant joint events and correspondently the probabilities are dominant joint probabilities. Then,using Bayes rule, we can deduce the formula of marginal probability of which the return association to the target. In DPDA, we are unnecessary to calculate the all feasible joint probabilities of feasible joint events to get the marginal probability and avoid the NP problem such as in joint probability data association (JPDA). the performance we proved in DPDA is that the top limited is no less than JPDA and the low limited is no less than PDA, and the probabilities of the events occuring with low limited performance are very small.