基于随机有限集的upf-cphd多目标跟踪randomfinitesetsbased.pdf

第33 卷第12 期 通 信 学 报 Vol.33 No. 12 2012 年12 月 Journa l on Communications December 2012 doi:10.3969/j .issn.1000-436x.2012.12.019 基于随机有限集的UPF-CPHD 多目标跟踪 王慧斌，陈哲，王鑫，马玉 （河海大学计算机与信息学院，江苏南京211100 ） 摘 要：提出一种基于随机有限集的无迹粒子基数概率假设密度滤波（UPF-CPHD, unscented particle fil- ter-cardinality probability hypothesis density ）的多目标跟踪方法。在粒子滤波框架下采用随机有限集（RFS , random finite sets ）对多目标状态和观测进行描述。在UPF 滤波框架下引入CPHD 算法同时递推目标状态和目标数目，并 计算必威体育精装版观测信息，估计结果更加精确，弥补PHD 估计目标数目不可靠的缺点。仿真实验表明，UPF -CPHD 多 目标跟踪方法能够降低超过50%的目标数目估计误差，并提高目标状态的估计精度。 关键词：随机有限集；多目标跟踪；无迹粒子滤波；基数概率假设密度滤波 中图分类号：TP71 文献标识码：A 文章编号：1000-436X(2012)12-0147-07 Random finite sets based UPF-CPHD multi-object tracking WANG Hui-bin, CHEN Zhe, WANG Xin, MA Yu (College of Computer and Information Engineering, Hohai University, Nanjing 211100,China) Abstract: A multiple tracking method based on UPF-CPHD was proposed, in which the state and observation of the ob- ject were both described by th e random finite sets (RSF). The CPHD algorithm was also introduced into the UPF frame- work to simultaneously deduce the object state and object number, making the estimation more precise. The experimental results show that the proposed UPF -CPHD algorithm is able to improve the estimation accuracy of the object number and state, as well as enhance the object tracking results. Key words: random finite sets ; multi-object tracking; unscented particle filter; cardinality probability hypothesis density [3]