基于核密度的动态初始化重置粒子滤波  

Dynamic initialization reset algorithm for particle filtering based on kernel density

作　　者：白剑锋[1] 南建国[1] 邬蒙[1] 查翔[1] BAI Jian-feng;NAN Jian-guo;WU Meng;ZHA Xiang(College of Engineering,Air Force Engineering University,Xi'an Shaanxi 710038,China)

机构地区：[1]空军工程大学工程学院,西安710038

出　　处：《计算机应用》2012年第1期295-298,共4页journal of Computer Applications

摘　　要：针对粒子滤波过程中,长时间的重采样造成的粒子多样性枯竭,由此导致目标跟踪中出现的精度下降及跟踪轨迹大幅振荡的现象,通过对采样粒子分布规律的研究,根据粒子枯竭的程度设置重置门限,在滤波过程中实时地检测粒子枯竭参数,当粒子的枯竭超过设置门限时,采用重置初始化粒子的方法来缓解采样粒子的枯竭趋势,有效地增加了长时间大量重采样后粒子的多样性,避免了粒子所含信息过多的丢失,显著地提高了粒子滤波的精度,在二维目标跟踪模型中应用所提算法并进行仿真实验,仿真结果证明了算法的可行有效。It has been found that the accuracy of particle filtering is much lower when the maneuvering target tracking process has been executed for a long time.The reason for this problem is that the diversity of the sampled particles is rapidly lost because of the excessive resampling.Therefore,the track of the maneuvering target estimated by the particle filtering will be widely wiggly from the true one.Through the research of the distribution of the sampled particles,a new algorithm was proposed.And a detected threshold was set to detect if the particle was dried up badly.When the particle was dried up badly,the particles of the state-space would be reset to relax the degree,so the new particles could contain more distribution information.The new algorithm has a high capability in the simulation of the 2-D maneuvering target tracking.

关 键 词：核密度 粒子滤波 采样粒子 机动目标跟踪 重采样 

分 类 号：TP391.413[自动化与计算机技术—计算机应用技术] TP183[自动化与计算机技术—计算机科学与技术]