| 引用本文: | 罗少华,徐 晖,安 玮,等.粒子PHD滤波存活粒子采样新方法.[J].国防科技大学学报,2012,34(2):102-106.[点击复制] |
| LUO Shaohua,XU Hui,AN Wei,et al.New method for survive particle sampling of particle PHD filter[J].Journal of National University of Defense Technology,2012,34(2):102-106[点击复制] |
|
| |
|
|
| 本文已被：浏览 7082次 下载 6057次 |
| 粒子PHD滤波存活粒子采样新方法 |
|
|
(1.国防科技大学 电子科学与工程学院，湖南 长沙 410073;2.93046部队，辽宁 沈阳 110000)
|
| 摘要: |
| 针对多目标跟踪粒子概率假设密度滤波算法中存活粒子的重要性密度采样问题，给出一种结合最新量测信息的存活粒子重要性密度采样新方法。该方法根据最新量测集中的各个量测与目标粒子的单步预测状态的似然值，以概率选取量测值，利用无迹变换获得粒子的重要性密度函数，并对其进行采样实现粒子概率假设密度滤波中存活粒子的采样，有效地减轻了粒子的退化现象。3目标跟踪仿真试验中，当目标模型与跟踪算法使用的目标模型不匹配时，采用所提出的存活粒子采样方法的粒子概率假设密度滤波算法最优子模式分配距离下降约70km。论文给出的存活粒子采样新方法显著地提高了多目标跟踪粒子概率假设密度滤波算法的鲁棒性。 |
| 关键词: 概率假设密度滤波 粒子采样 多目标跟踪 无迹变换 |
| DOI： |
| 投稿日期：2011-05-20 |
| 基金项目:国家部委资助项目 |
|
| New method for survive particle sampling of particle PHD filter |
| LUO Shaohua1, XU Hui1, AN Wei2, XU Dan1, ZHANG Zhiheng3 |
|
(1.College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China;2.1 College of Electronic Science and Engineering, National University of Defense Technology, Changsha 410073, China;3.2.Unit 93046,Shenyang 110000,China)
|
| Abstract: |
| For the problem of survive particle importance sampling in multitarget tracking Probability Hypothesis Density particle filter, a new algorithm of survive particle importance sampling is presented. For every particle, the algorithm exploits likelihood between latest measurements, and the particle chooses one measurement of the set of measurement to obtain importance distribution by update particle, and draws samples of survive particle from the importance distribution. The presented algorithm reduced degeneration of particle efficiently. In simulation scenario of 3 targets tracking, the optimal sub-pattern assignment metric of particle probability hypothesis density filter, which adopted the presented survive particle importance sample method, decreases about 70Km when targets model used in target tracking method is different from actual targets model. The proposed method enhances the robustness of multitarget tracking of particle probability hypothesis density filter remarkably. |
| Keywords: particle probability hypothesis density particle sample multi-target tracking unscented transformation |
|
|
|
|
|
