面向应用的GM-CPHD雷达多目标跟踪技术

摘要

This paper studies Gaussian mixture cardinalized probability hypothesis density (GM-CPHD) algorithm.Firstly,the recursive form of the cardinalized probability hypothesis density (CPHD) algorithm is presented using the theory of random finite set and optimal Bayes estimation.Then,under the linear and Gaussian assumption,the analytic close-form solution to CPHD is derived to decrease the computation complexity and meet the real-time requirement.At last,simulation results show that GM-CPHD performs better than GM-PHD at target number estimation.%针对雷达密集多目标跟踪数据关联的难题,为了进一步降低目标个数估计误差,研究高斯混合-势概率假设密度方法(GM-CPHD).首先,在随机集框架和最优贝叶斯理论下,给出了CPHD递归形式;然后,在线性高斯假设条件下,详细给出了GM-CPHD强度和势预测和更新的递归闭合解,降低了计算复杂度,满足跟踪实时性要求;最后,仿真实验结果显示,GM-CPHD目标个数估计精度比GM-PHD更高; 雷达实验数据测试结果显示,GM-CPHD在不需要数据关联的情况下,能够有效抑制大量杂波,稳定地估计目标个数和目标状态.