Journal of Computer Applications ›› 2020, Vol. 40 ›› Issue (1): 299-303.DOI: 10.11772/j.issn.1001-9081.2019061128

• Frontier & interdisciplinary applications • Previous Articles     Next Articles

Solution method to anomalous smoothing problem in particle probability hypothesis density smoother

HE Xiangyu1, YU Bin2, XIA Yujie1   

  1. 1. School of Physics and Electronic Information, Luoyang Normal University, Luoyang Henan 471934, China;
    2. Department of Power Generation and Operation, SPIC Nanyang Thermoelectricity Company Limited, Nanyang Henan 473000, China
  • Received:2019-07-01 Revised:2019-09-06 Online:2020-01-10 Published:2019-10-10
  • Contact: 何祥宇
  • Supported by:
    This work is partially supported by the Foundation for Young Backbone Teachers in Higher Education of Henan Province (2018GGJS126).

粒子概率假设密度平滑器异常平滑问题的解决方法

何祥宇1, 于斌2, 夏玉杰1   

  1. 1. 洛阳师范学院 物理与电子信息学院, 河南 洛阳 471934;
    2. 国家电投集团南阳热电有限责任公司 发电运行部, 河南 南阳 473000
  • 作者简介:何祥宇(1982-),男,河南泌阳人,讲师,博士,主要研究方向:多目标跟踪;于斌(1982-),男,河南开封人,助理工程师,主要研究方向:发电厂生产运行;夏玉杰(1978-),男,河南洛阳人,副教授,博士,主要研究方向:通信信号处理。
  • 基金资助:
    河南省高等学校青年骨干教师培养计划项目(2018GGJS126)。

Abstract: To solve the anomalous smoothing problems caused by the missed detection or target disappearance in the particle Probability Hypothesis Density (PHD) smoother, an improved method based on the modified target survival probability was proposed. Firstly, the prediction and update formulas of forward filtering were modified to obtain the target intensity function of filtering and estimate the number of survival targets in filtering process. On this basis, using the estimated value changes of forward filtering of survival number to judge whether targets disappearance or missed detection occurring, and the survival probability used in backward smoothing calculation was defined. Then, the iterative calculating formula for backward smoothing was improved with the obtained survival probability, and the particle weights were obtained on this basis. The simulation results show that the proposed method can solve the anomalous smoothing problems in PHD smoother effectively, its time averaged Optimal SubPattern Assignment (OSPA) distance error is decreased from 7.75 m to 1.05 m compared with standard algorithm, which indicates that the tracking performance of the proposed method is improved significantly.

Key words: multi-target tracking, random finite set, Probability Hypothesis Density (PHD), filtering, smoothing

摘要: 针对粒子概率假设密度(PHD)平滑器中由漏检或目标消失现象引起的异常后向平滑估计问题,提出一种基于目标存活概率修正的改进方法。首先,修正前向滤波的预测与更新计算公式以获取滤波的目标强度函数和估计滤波过程的存活目标个数。在此基础上根据存活目标个数的前向滤波估计值的变化情况,判断跟踪过程中是否存在目标消失或漏检现象,确定后向平滑计算用到的目标存活概率值,然后采用此确定的存活概率值来改进后向平滑迭代计算公式,据此计算PHD分布中的粒子权值。仿真结果表明,所提方法能有效地解决PHD平滑器的异常平滑问题,其时间平均的最优子模式分配(OSPA)距离误差相对于标准算法由7.75 m减小至1.05 m,目标跟踪性能有了明显提升。

关键词: 多目标跟踪, 随机有限集, 概率假设密度, 滤波, 平滑

CLC Number: