论文部分内容阅读
为解决目标数未知或随时间变化的多目标跟踪问题,通常将多目标状态和观测数据表示成随机集形式,并通过递推计算目标状态联合分布的概率假设密度(PHD)来完成.然而,对于被动测角的非线性跟踪问题,PHD无法获得闭合解,为此提出一种新的高斯混合粒子PHD算法.该算法利用高斯混合近似PHD,以避免用聚类确定目标状态,并采用拟蒙特卡罗(QMC)积分方法计算目标状态的预测和更新分布.仿真结果验证了所提出算法的有效性.
In order to solve the problem of multi-target tracking with unknown number of targets or changing over time, multi-objective states and observations are usually expressed in random sets and recursively calculated by the probability hypothesis density (PHD) of joint distribution of target states.However, For the non-linear tracking problem of passive goniometry, PHD can not obtain the closed solution, so a new Gaussian mixture particle PHD algorithm is proposed.This algorithm uses Gaussian mixture approximation to PHD to avoid using cluster to determine the target state, Carlo (QMC) integral method is used to calculate the prediction and update distribution of the target states.The simulation results verify the effectiveness of the proposed algorithm.