多元假设检验GMPHD轨迹跟踪

由于在军事和民事领域逐步广泛的应用，数目不定的多目标跟踪技术正受到越来越多的关注。概率假设密度(PHD)滤波方法，特别是具有闭式递归的高斯混合概率假设密度(GMPHD)技术，在噪声和漏警等影响下仍能形成优越的群目标跟踪性能。然而PHD滤波器并不能实现多目标航迹跟踪，而其与传统数据互联的结合，复杂度高且跟踪效果不尽如人意。在该文中，各目标的航迹信息以假设形式表述，数据互联则是通过使用经典的多元假设检测方法判决假设矩阵实现。其与GMPHD的结合不仅实现了数据互联和轨迹管理，还因为积累时间信息大大降低了杂波干扰的影响。实验结果证明，该算法可以对多个目标所形成的轨迹实施正确跟踪，同时，计算量的大幅度降低带来了跟踪系统可实现性的提高。

Multiple Hypotheses Detection with Gaussian Mixture Probability Hypothesis Density Filter for Multi-target Trajectory Tracking

Multi-target tracking is becoming one of most focusing research topics because of the modern military affair requirements as well as civil developments. Among all the techniques, Probability Hypothesis Density(PHD) filtering approach, especially Gaussian Mixture PHD(GMPHD) filter, which has a closed form recursion, has shown its advantages in tracking unknown number of targets despite the impact of noise and missing detection etc. Existing PHD trajectory tracking methods combining PHD filter, which can not estimate the trajectories of multi-target alone, with traditional data association, are computationally expensive and almost intractable. In this paper, a brand new multi-target trajectory tracking algorithm based on random finite set theory is brought forward by adopting classical signal detection technique along with GMPHD filter. Using hypotheses representing the trajectory information, data association is accomplished through the hypothesis matrix judging on the same basement as track managing function. The simulation results suggest that this algorithm not only significantly alleviates the heavy computing load, but also performs multi-target trajectory tracking effectively in the meantime.