电子学报 ›› 2022, Vol. 50 ›› Issue (5): 1089-1097.DOI: 10.12263/DZXB.20210374

• 学术论文 • 上一篇    下一篇

强杂波背景下基于变分贝叶斯推理的机载雷达目标跟踪算法

李淑慧1, 邓志红1, 冯肖雪1, 潘峰1,2   

  1. 1.北京理工大学自动化学院,北京 100081
    2.昆明北理工产业技术研究院有限公司,云南 昆明 650000
  • 收稿日期:2021-03-21 修回日期:2021-06-07 出版日期:2022-05-25 发布日期:2022-06-18
  • 作者简介:李淑慧 女,1991年生于河南省商丘市.博士研究生.主要研究方向为目标跟踪、多传感器融合、状态估计. E‑mail: 2579371789@qq.com
    邓志红 女,1974年出生于辽宁省昌图县.现为北京理工大学自动化学院教授,博士生导师.主要从事导航制导与控制领域的研究工作.E‑mail: dzh_deng@bit.edu.cn
    冯肖雪 女,1988年出生于河北省邢台市.现为北京理工大学自动化学院讲师,从事多源信息融合、状态估计方面的研究工作. E‑mail: fengxiaoxue@bit.edu.cn
  • 基金资助:
    国家自然科学基金(61433003);广东省科技创新战略专项基金(skjtdzxrwqd2018001);云南省科技厅重点研究项目(2018BA070);云南省应用基础研究项目(201701CF00037)

Variational Bayesian Inference‑Based Airborne Radar Target Tracking Algorithm in Strong Clutter

LI Shu-hui1, DENG Zhi-hong1, FENG Xiao-xue1, PAN Feng1,2   

  1. 1.School of Automation,Beijing Institute of Technology,Beijing 100081,China
    2.Kunming?BIT Industry Technology Research Institute INC,Kunming,Yunnan 650000,China
  • Received:2021-03-21 Revised:2021-06-07 Online:2022-05-25 Published:2022-06-18

摘要:

机载雷达遭受的强杂波干扰以及目标的强机动使噪声呈现长拖尾的非高斯特性. 此外,载机的运动导致杂波淹没目标的航迹,使雷达无法检测到目标,出现随机的量测丢失现象. 为此,设计了强杂波背景下含量测丢失的目标跟踪算法. 该算法采用学生t分布来模拟非高斯噪声的长拖尾特性. 通过引入伯努利随机变量,将求和形式的后验概率密度函数转换成乘积形式的概率质量函数,并构建了分层状态空间模型. 在此基础上,设计了用于量测丢失的鲁棒变分贝叶斯平滑器. 以机载雷达跟踪空中目标为例验证了本文算法的有效性.

关键词: 机载雷达, 杂波, 量测丢失, 概率图模型, 变分贝叶斯推理, 多变量学生t分布

Abstract:

The strong clutter interference suffered by the airborne radar and the strong maneuvering of the target make noise non-Gaussian and heavy-tailed. Besides, the movement of the carrier aircraft induces the target is totally submerged by the clutter, so that the radar cannot detect the target. To this end, a target tracking algorithm for missing measurements in strong clutter is designed. Student t distribution is utilized to model the heavy-tailed property of non-Gaussian noise. The posterior probability density function(PDF) of the summation form is converted into the probability mass function(PMF) of the product form by introducing Bernoulli random variables. And a hierarchical state space model is further devised. Based on this model, a robust variational Bayesian smoother for measurement dropouts(RVBSD) is designed. An example that the airborne radar tracks an airborne target is given to verify the effectiveness of the proposed algorithm.

Key words: airborne radar, clutter, missing measurement, probabilistic graph model, variational Bayesian inference, multivariable Student t distribution

中图分类号: