数据关联算法的研究

Research on data association algorithms

在现代目标跟踪系统中,数据关联是一个重要组成部分.由于测量来源的不确定性,密集环境中的目标跟踪比较困难.概率数据关联算法和联合概率数据关联算法能很好地处理目标跟踪并在很多领域得到了广泛的应用.但是一旦出现某种干扰或是故障,通过概率数据关联算法得到的滤波值也会偏离真实值很多,造成滤波发散,严重影响性能.针对这一不足,基于概率数据关联算法中的组合新息,提出了修正概率数据关联算法.最后用一些实例来评价该跟踪算法的跟踪性能,仿真结果表明了修正概率数据关联算法和修正联合概率数据关联算法的有效性.

Data association has a significant role in a modern target tracking system. Target tracking in dense environments is very difficult because of uncertainties in measurement. Probabilistic data association （PDA） and joint probabilistic data association （JPDA） are suitable for target tracking and have found widespread applications in many areas. But if interference is present, the estimated values with PDA will differ a lot from the true values and performance will deteriorate dramatically because of this filtering divergence. In order to overcome this disadvantage, a modified PDA （MPDA） is proposed based on the combined information of the PDA. Some simulations were done to evaluate its tracking performance. The results demonstrate the effectiveness of MPDA and MJPDA.