基于最近邻的集中式多传感器多目标跟踪算法

Centralized Multisensor Multitarget Tracking Algorithms Based on Nearest Neighbor Standard Function

经典的集中式多传感器多目标跟踪算法通常计算量较大，经常难以满足系统的实时性要求，工程上实现起来比较困难，为进一步扩大集巾式多传感器的应用范嗣，使其在对算法实时性要求较高、跟踪精度要求较小的实际场合中广泛应用。文章基于最近邻域思想，研究了并行处理结构的集中式多传感器最近邻域算法，并从算法跟踪精度、实时性、有效跟踪率3个方面对其与经典的顺序多传感器联合概率数据互联算法进行了仿真比较。经仿真验证，并行处理结构的集中式多传感器最近邻域算法实时性提高了60％以上，且在跟踪背景杂波适中的情况下能够有效跟踪目标。

Because the calculation burden of the classical centralized multisensor multitarget tracking algorithms is usually heavy, the algorithms are hard to meet the requirement of the system real time performance and be carried out in practice. Aiming to expand the application of centralized multisensor, which could be used in the practical situation that has a high requirement of real time and a low requirement of tracking precision, based on nearest neighbor standard function, parallel centralized nearest neighbor standard function was proposed. The performance comparison and analysis of parallel centralized nearest neighbor standard function and sequential centralized multisensor joint probabilistic data association algorithm was presented from three aspects, namely tracking precision, realtime and effective tracking ratio. The simulation results show that compared with sequential centralized multisensor joint probabilistic data association algorithm, parallel centralized nearest neighbor standard function could track the targets effectively with the clutter of moderate density, of which the real time performance is improved more than 60%.