电子学报 ›› 2021, Vol. 49 ›› Issue (7): 1346-1353.DOI: 10.12263/DZXB.20200960

• 学术论文 • 上一篇    下一篇

闪烁噪声统计特性未知情况下的鲁棒广义标签多伯努利滤波器

侯利明1, 连峰1, 谭顺成2,3, 徐从安3   

  1. 1.西安交通大学自动化科学与工程学院综合自动化研究所,陕西 西安 710049
    2.南京电子技术研究所,江苏 南京 210039
    3.海军航空大学信息融合研究所,山东 烟台 264001
  • 收稿日期:2020-08-31 修回日期:2021-03-03 出版日期:2021-07-25 发布日期:2021-08-11
  • 作者简介:侯利明 男,1987年9月出生,河南延津人,博士研究生. 主要研究方向为目标跟踪、信息融合与传感器管理. E‑mail:hliming2017@stu.xjtu.edu.cn
    连 峰 男,1981年9月出生,陕西宝鸡人,博士,教授,博士生导师. 主要研究方向为目标跟踪、信息融合与传感器管理. E‑mail: lianfeng1981@xjtu.edu.cn
    谭顺成(通讯作者) 男,1985年12月出生,湖南湘潭人,南京电子技术研究所博士后,海军航空大学讲师. 主要研究方向为信息融合、雷达数据处理. E‑mail: tanshuncheng85@sina.com
  • 基金资助:
    国家自然科学基金(61671462);国家公派留学基金(201906280272);陕西省自然科学基础研究计划(2020JQ?073)

Robust Generalized Labeled Multi‑Bernoulli Filter for Multi‑Target Tracking with Unknown Statistical Characteristics of Glint Noise

Li-ming HOU1, Feng LIAN1, Shun-cheng TAN2,3, Cong-an XU3   

  1. 1.Institute of Integrated Automation,School of Automation Science and Engineering,Xi’an Jiaotong University,Xi’an,Shaanxi 710049,China
    2.Nanjing Research Institute of Electronics Technology,Nanjing,Jiangsu 210039,China
    3.Institute of Information Fusion,Naval Aviation University,Yantai,Shandong 264001,China
  • Received:2020-08-31 Revised:2021-03-03 Online:2021-07-25 Published:2021-08-11

摘要:

为了解决闪烁噪声统计特性未知情况下的多目标跟踪问题,提出一种鲁棒广义标签多伯努利(Generalized Labeled Multi?Bernoulli, GLMB)滤波器.该滤波器采用均值未知且时变的多维Student’s t分布对统计特性未知的闪烁噪声进行建模.它放宽了闪烁噪声均值为零的限制性假设,可以自适应地处理闪烁噪声均值未知且时变条件下的多目标跟踪问题.本文在GLMB滤波框架下,利用变分贝叶斯方法对增广状态中的参数进行变分迭代,并通过最小化Kullback?Leibler散度得到边缘似然函数的近似解.仿真结果表明,在闪烁噪声统计特性未知的情况下,所提滤波器能有效地对多目标进行跟踪.

关键词: 随机有限集, 多目标跟踪, 闪烁噪声, 统计特性未知, 变分贝叶斯推断, 广义标签多伯努利滤波器

Abstract:

A robust generalized labeled multi?Bernoulli (GLMB) filter is presented to perform multi?target tracking (MTT) with unknown statistical characteristics of glint noise. The glint noise with unknown statistical characteristics is modeled as a multivariate Student’s t distribution with unknown and time?varying mean. The proposed filter relaxes the restrictive assumption that the mean of glint noise is zero, and can effectively deal with the problem of MTT under the condition that the mean of glint noise is unknown and time?varying. The variational Bayesian approximation is applied in the GLMB filtering framework with the augmented state. The approximate solution of the marginal likelihood function can be obtained by minimizing the Kullback?Leibler divergence. The simulation results demonstrate that the proposed filter can effectively track multi?target when the statistics of glint noise is unknown.

Key words: random finite set, multi?target tracking, glint noise, unknown statistical characteristics, variational Bayesian inference, generalized labeled multi?Bernoulli filter

中图分类号: