一种改进的多传感器多目标跟踪联合概率数据关联算法研究

联合概率数据关联(JPDA)算法对单传感器多目标跟踪是一种良好的算法,但对于多传感器多目标跟踪的情况,特别是目标较为密集时,计算量剧增,会出现计算组合爆炸现象。因此,提出了一种改进算法,即对多传感器多目标量测进行同源划分,将多传感器对多目标的跟踪问题简化为单传感器对多目标的跟踪问题,然后将JPDA当作一种组合优化问题,采用连续型Hopfield神经网络求解关联概率。经仿真研究表明,该方法不仅克服了JPDA算法在多传感器多目标跟踪问题中的缺陷,还提高了跟踪精度。

Research of Improved Joint Probabilistic Data Association Algorithm for Multisensor-Multitarget Tracking

Joint probabilistic data association (JPDA) algorithm is a good method for the single sensor multitarget tracking. However, for the multisensor-multitarget (MSMT) tracking, especially in clutter, its calculation load became higher and it would cause the combination explosion problem of calculation. Therefore, the improved algorithm for the multisensor-multitarget (MSMT) tracking was proposed. The same source observations were classified into the same set. And then JPDA algorithm was treated as a combination problem of optimization. And continuous Hopfield-type neural network was introduced to solve the association probability. The simulation results show that the proposed method can overcome the limitation of JPDA algorithm in multisensor-multitarget (MSMT) tracking, and increase precision.