基于TDOA测量的多目标P-GMPHD跟踪算法  

A New P-GMPHD Filter Algorithm for Multiple Target Localization Based on Passive Multilateral TDOAs

作　　者：徐志军 苗秀梅[2] 吴鑫辉[3] 

机构地区：[1]海军信息化部,北京100841 [2]海军装备部,北京100841 [3]海军工程大学电子工程学院,武汉430033

出　　处：《计算机与数字工程》2012年第9期12-15,共4页Computer & Digital Engineering

基　　金：国家自然科学基金项目(编号:60901069);湖北省自然科学基金(编号:2009CDB031)资助

摘　　要：针对传统的到达时差(Time difference of arrival,TDOA)多目标跟踪算法计算量大,估计精度低,虚警值较多等问题,提出了一种预关联高斯混合概率假设密度(P-GMPHD)多目标跟踪算法。该算法利用随机集理论对TDOA的目标状态和观测值进行建模,通过递推高斯混合来预测和更新各状态后验概率密度,避免了复杂的数据关联问题。为减轻高斯混合滤波的计算量,提出了将预测信息与观测值进行预关联的思想,剔除虚警值,从而显著地降低了计算量。仿真结果表明,该算法能在杂波环境下有效地利用TDOA测量值跟踪未知数目的多个运动目标,并且在不影响跟踪性能的情形下,其计算量比一般GMPHD有了较大的降低。According to the traditional multi-target tracking based on TDOA being of higher computational, lower estimated accuracy, and the presence of association uncertainty, a novel pre-association Gaussian mixture probability hypothesis density filter （P-GMPHD） is proposed. The approach involves modeling the targets and measurements as random finite sets and applying the Gaussian mixture to propagate the posterior density, which could avoid the difficult problem of data association. To alleviate the computation of GMPHD, a pre-association method which eliminates false measurements is introduced. Simulation results show that the P-GMPHD algorithm could deal with unknown number of emitters under the complex environment with clutter. Moreover, without loses tracking accuracy, the algorithm presents lower tracking computation.

关 键 词：多目标跟踪 随机有限集 TDOA P-GMPHD 

分 类 号：TN18[电子电信—物理电子学]