电子学报 ›› 2022, Vol. 50 ›› Issue (7): 1722-1734.DOI: 10.12263/DZXB.20201274

• 学术论文 • 上一篇    下一篇

被动声纳粒子滤波检测前跟踪的检测门限设置方法

奚畅, 蔡志明, 袁骏   

  1. 海军工程大学电子工程学院,湖北 武汉 430000
  • 收稿日期:2020-11-12 修回日期:2021-04-12 出版日期:2022-07-25
    • 作者简介:
    • 奚畅 男,1992 年2 月出生,河北保定人. 现为海军工程大学电子工程学院博士研究生. 主要研究方向为水声信号处理.E-mail: xichangwxx@163.com
      蔡志明 男,1962 年11 月出生,浙江湖州人. 2002年毕业于哈尔滨工程大学,现为海军工程大学教授、博士生导师. 主要研究方向为信号处理、水声物理及声纳技术.
      袁骏 男,1979 年11 月出生,江苏吴江人. 2017年毕业于海军工程大学,获博士学位.现为海军工程大学讲师. 主要研究方向为水声信号处理.E-mail:yjun_hg@163.com

A Method of Determining Detection Threshold for Particle Filter Track-Before-Detect in Passive Sonar

XI Chang, CAI Zhi-ming, YUAN Jun   

  1. College of Electronic Engineering, Naval University of Engineering, Wuhan, Hubei 430000, China
  • Received:2020-11-12 Revised:2021-04-12 Online:2022-07-25 Published:2022-07-30

摘要:

对于粒子滤波检测前跟踪算法,在被动声纳观测量满足指数分布的情况下,本文提出一种按照系统要求的虚警概率设置检测门限的方法.首先,利用贝叶斯定理推导了后验概率比表达式;然后,设置对数后验概率比为检验统计量,通过积分运算离散化处理和Jacobian定理对其进行简化;最后,在不存在目标的情况下分析检验统计量的概率分布特性,得到一定虚警概率要求下检测门限的解析表达式,利用有目标和没有目标情况下观测单元的统计特性差异,证明了在控制虚警概率的同时可以实现对目标的检测.蒙特卡洛仿真结果表明,利用所提的门限设置方法得到的虚警概率与系统要求的虚警概率之间偏差小于0.01,检测概率大于0.95.

关键词: 检测前跟踪, 粒子滤波, 检测门限, 指数分布

Abstract:

This paper proposes a method of determining detection threshold according to the demanded false alarm probability for particle filter track-before-detect, in the case of the passive sonar observables are all exponential distributions. Firstly, the expression of posterior probability ratio is derived based on the Bayesian theorem. Secondly, the logarithmic posterior probability ratio is set as the test statistic, and it is simplified by the discretization of the integral operation and the Jacobian logarithm. Finally, the probability distribution characteristics of the test statistic is analysed in the absence of a target, and the analytical expression of the detection threshold is obtained according to the demanded false alarm probability. Using the difference in statistical characteristics of the observables with and without a target, it is proved that the detection of the target can be achieved while controlling the false alarm probability. The Monte-Carlo simulation results show that using the proposed method, the deviation between the calculated value of false alarm probability and the required value of false alarm probability is less than 0.01 and the detection probability is greater than 0.95.

Key words: track-before-detect, particle filter, detection threshold, exponential distributions

中图分类号: