中文摘要：

概率假设密度滤波器将目标的状态空间及观测空间描述为随机有限集合的形式，有效避免了多目标跟踪中复杂的数据关联问题。但对于不同类型的目标使用同样的全部观测数据集进行目标状态更新，未对观测数据进行合理分配，导致估计性能下降。该文提出一种观测最优分配的高斯混合概率假设密度多目标跟踪算法（MOA-GM-PHD），将目标分为已有目标和新生目标两类，推导极大似然门限来获得两类目标对应的最优观测数据，再分别进行目标状态更新。实验结果表明，该文方法目标跟踪效果优于传统GM-PHD滤波器。

英文摘要：

As describing target state space and sensor observation space by Random Finite Set （RFS）method,Probabil-ity Hypothesis Density （PHD）filter is a promising approach for multi-target tracking without consideration of measurement-to-track association.Due to the fact that classic GM-PHD filer has not distinguish the measurements of survival targets and birth targets,a Measurements Optimal Assigned GM-PHD filter （MOA-GM-PHD）is proposed in this paper,which distin-guishes the measurements set by using a maximum likelihood adaptive gate,and survival targets and birth targets update the PHD estimation using the optimal assigned measurements respectively.Simulation results show that the proposed algorithm obtained an improved performance.