基于星凸RHM的扩展目标SMC-PHD滤波  

SMC-PHD filter for extended target tracking based on star-convex random hypersurface models

作　　者：王雪[1] 李鸿艳[2] 孔云波[2] 蒲磊[2] 樊鹏飞[2] 

机构地区：[1]中国人民解放军空军试飞局,西安710089 [2]空军工程大学信息与导航学院,西安710077

出　　处：《计算机应用研究》2017年第7期2144-2147,共4页Application Research of Computers

基　　金：陕西省自然科学基金资助项目(2015JM6332)

摘　　要：扩展目标跟踪与传统目标跟踪不同,不仅需要对目标的运动状态进行跟踪,同时对于目标的外形特征也不能忽略。针对扩展目标跟踪过程中存在的外形拟合和非线性的问题,提出一种基于星凸随机超曲面模型(RHM)的扩展目标序贯蒙特卡洛概率假设密度滤波(SMC-PHD)算法。该算法运用星凸RHM对扩展目标量测源建模,在SMC-PHD的框架下,推导出非线性滤波算法的量测似然表达式和更新方程,实现扩展目标跟踪。仿真结果证明,所提算法的跟踪性能较其他滤波对于目标扩散程度和质心估计均有提高。Extended object tracking is different from the traditional object tracking technology, it does not ignore the target’s shape.Extended object tracking simultaneously had both the centroid’s kinematical state and the shape of target, this paper proposed a sequential Monte Carlo probability hypothesis density （SMC-PHD） filter based on star-convex random hypersurface models（RHM） for target’s shape and nonlinear extended target tracking. The proposed algorithm described the extension of measurements by the star-convex random hypersurface model, and then it was embedded into the SMC-PHD, derived nonlinear estimator’s likelihood function and measurement update to track extended target. Simulation results show that the proposed method outperforms the other PHD filter in extension estimation as well as centroid state estimation.

关 键 词：星凸形 随机超曲面模型 扩展目标 序贯蒙特卡洛概率假设密度 

分 类 号：TN911.7[电子电信—通信与信息系统]