系统工程与电子技术 ›› 2022, Vol. 44 ›› Issue (12): 3603-3613.doi: 10.12305/j.issn.1001-506X.2022.12.03

• 电子技术 • 上一篇    

多模型广义标签多伯努利滤波器

辛怀声*, 宋鹏汉, 曹晨   

  1. 中国电子科技集团公司电子科学研究院, 北京 100041
  • 收稿日期:2021-09-05 出版日期:2022-11-14 发布日期:2022-11-24
  • 通讯作者: 辛怀声
  • 作者简介:辛怀声 (1982—), 男, 高级工程师, 博士研究生, 主要研究方向为情报融合、多目标跟踪|宋鹏汉 (1989—), 男, 高级工程师, 博士, 主要研究方向为情报融合、目标识别|曹晨 (1974—), 男, 研究员, 博士, 主要研究方向为雷达系统

Multiple model based generalized labeled multi-Bernoulli filter

Huaisheng XIN*, Penghan SONG, Chen CAO   

  1. China Academy of Electronics and Information Technology, Beijing 100041, China
  • Received:2021-09-05 Online:2022-11-14 Published:2022-11-24
  • Contact: Huaisheng XIN

摘要:

标准广义标签多伯努利算法没有对目标状态转移密度进行深入分析, 在带入确定运动模型的情况下无法对机动目标进行跟踪。针对这个问题, 参考基于马尔可夫跳变分支合并策略的多模型算法, 提出了交互多模型广义标签多伯努利算法、一阶广义伪贝叶斯广义标签多伯努利算法, 以及二阶广义伪贝叶斯广义标签多伯努利算法, 并将这三种多模型算法与同样针对机动多目标的马尔可夫跳变系统广义标签多伯努利算法进行比较。仿真结果表明, 与马尔可夫跳变系统广义标签多伯努利算法相比, 所提三种算法具有更低的计算时间消耗和更高的跟踪精度。其中, 一阶广义伪贝叶斯广义标签多伯努利算法计算时间消耗最低, 二阶广义伪贝叶斯广义标签多伯努利算法跟踪精度最高, 交互多模型广义标签多伯努利算法综合性能最好。

关键词: 广义标签多伯努利, 多模型, 广义伪贝叶斯, 马尔可夫跳变, 目标跟踪

Abstract:

In standard generalized labeled multi-Bernoulli (GLMB), the target state transfer density is not discussed in detail. In maneuvering target tracking scenario, GLMB will not work properly when bringing into the case of determining the motion model. To solve this problem, the theory of merging multiple jump Markov chains in multi-model (MM) maneuvering target tracking algorithm is introduced. Three MM-GLMB filters, the interacting multiple model based GLMB (IMM-GLMB), the generalized pseudo Bayes1 based GLMB (GPB1-GLMB) and the generalized pseudo Bayes2 based GLMB (GPB2-GLMB) are presented in this paper. Three filters are compared with the jump Markov system based GLMB (JMS-GLMB) which is also designed to solve the multi-maneuvering targets tracking problem. Simulation results show that the proposed three filters have lower computational cost and higher tracking accuracy compared with JMS-GLMB. Among them, GPB1-GLMB has the lowest computational cost, GPB2-GLMB has the highest tracking accuracy, while IMM-GLMB has the best overall performance.

Key words: generalized labeled multi-Bernoulli (GLMB), multiple model (MM), generalized pseudo Bayes (GPB), jump Markov, target tracking

中图分类号: