高斯过程回归模型多扩展目标多伯努利滤波器  

A multiple extended target multi-Bernouli filter based on Gaussian process regression model

作　　者：陈辉[1] 李国财 韩崇昭[2] 杜金瑞 CHEN Hui;LI Guo-cai;HAN Chong-zhao;DU Jin-rui(School of Electrical and Information Engineering,Lanzhou University of Technology,Lanzhou Gansu 730050,China;Institute of Integrated Automation,School of Electronic and Information Engineering,Xi’an Jiaotong University,Xi’an Shaanxi 710049,China)

机构地区：[1]兰州理工大学电气工程与信息工程学院,甘肃兰州730050 [2]西安交通大学电子与信息工程学院综合自动化研究所,陕西西安710049

出　　处：《控制理论与应用》2020年第9期1931-1943,共13页Control Theory & Applications

基　　金：国家国防基础科研项目(JCKY2018427C002);国家自然科学基金项目(61873116,51668039,61763029);甘肃省科技计划项目(18JR3RA137)资助。

摘　　要：针对复杂不确定性环境下不规则形状的多扩展目标跟踪问题,本文提出了一种基于高斯过程回归(GPR)模型的多扩展目标多伯努利(GPR–ETCBMeMBer)滤波算法.首先,在利用有限集统计理论(FISST)将多扩展目标的状态集与量测集分别建模为多伯努利随机有限集(MBer RFS)和泊松随机有限集(Poisson RFS)的基础上,通过GPR方法建立多扩展目标随机超曲面的跟踪滤波模型.然后,基于容积卡尔曼滤波器(CKF)详细推导并提出GPR多扩展目标多伯努利滤波算法的高斯混合(GM)实现.最后,通过构造对星凸形多扩展目标和多群目标跟踪的仿真实验验证了本文所提算法的有效性.In view of the tracking problem of multiple extended target with irregular shape in the complicated and uncertain environment,a Gaussian process regression(GPR)based multiple extended target multi-Bernoulli filter(GPR–ETCBMeMBer)algorithm is proposed in this article.Firstly,on the basis of modeling state set and measurement set of multiple extended target as multi-Bernoulli random finite set(MBer RFS)and Poisson RFS respectively by using finite set statistics(FISST),This article models the random hypersurface based filtering algorithm of multiple extended target via GPR approach.Then,this article derives in detail and proposes a Gaussian mixture(GM)implementation of the GPR–ETCBMeMBer filter via the cubature Kalman filter(CKF).Finally,the effectiveness of the proposed method is verified by the simulations of star-convex shape multiple extended target tracking and multiple group target tracking.

关 键 词：多扩展目标跟踪 随机超曲面 高斯过程回归 随机有限集 多伯努利滤波器 

分 类 号：TN713[电子电信—电路与系统]