Abstract: The error of the target number estimated in the Gaussian mixture cardinalized probability hypothesis density filter is large at the presence of low detection rate. Addressing at this problem, a post-processing algorithm for the GM-CPHD filter based on track-estimate association is presented. Distance matrix between tracks and estimates is calculated, and track-estimate association is done via Hungarian assignment algorithm. A threshold for track continuity is engaged to prune some short tracks, and then some wrong estimates are deleted. Moreover, Lagrange interpolation method is employed to deal with the discontinuous tracks and the corresponding estimates of missing targets are compensated. The experimental results show that the post-processing algorithm can improve the accuracy of the estimated target number effectively.

Key words: target tracking, Gaussian mixture cardinalized probability hypothesis density filter, track-estimate association, Lagrange interpolation

摘要： 高斯势概率假设密度滤波算法在低检测率条件下目标数目估计会出现偏差。针对该问题，提出了一种基于航迹—估计关联的GM-CPHD后处理算法。计算航迹和估计之间的距离矩阵，利用匈牙利指派算法进行航迹—估计关联。通过设定航迹的连续性阈值对短航迹进行裁剪，并以此消除虚假目标估计。利用拉格朗日插值对各条不连续的航迹进行插值，以弥补由于低检测率而造成的遗漏估计。仿真实验结果表明，该处理算法能够有效地提高目标数目的估计精度。

关键词: 目标跟踪, 高斯混合势概率假设密度（GM-CPHD）滤波, 航迹&mdash, 估计关联, 拉格朗日插值