航空学报 > 2021, Vol. 42 Issue (11): 524851-524851   doi: 10.7527/S1000-6893.2020.24851

一种改进的直方图概率多假设多目标跟踪方法

张奕群, 尹立凡, 王硕, 孙承钢   

  1. 北京电子工程总体研究所, 北京 100854
  • 收稿日期:2020-10-09 修回日期:2020-12-12 发布日期:2021-02-02
  • 通讯作者: 张奕群 E-mail:yiqunzhang@hotmail.com
  • 基金资助:
    国家级项目

An improved histogram PMHT multi-target tracking method

ZHANG Yiqun, YIN Lifan, WANG Shuo, SUN Chenggang   

  1. Beijing Institute of Electronic System Engineering, Beijing 100854, China
  • Received:2020-10-09 Revised:2020-12-12 Published:2021-02-02
  • Supported by:
    National Level Project

摘要: 直方图概率多假设跟踪(H-PMHT)方法及其变形泊松分布直方图概率多假设跟踪(P-HPMHT)方法的一个主要缺点是其量测模型仅考虑了背景杂波而没有考虑传感器噪声,从而导致在低信噪比条件下检测概率较低。针对这一问题,提出了一种带传感器噪声模型的H-PMHT方法,通过将传感器噪声引入量测模型,从而明显提高了对低信噪比目标的跟踪检测能力。该方法的计算量与目标数保持线性关系,仍然适用于目标数目较多的情况。仿真实验表明:该方法在误跟踪比率为1‰,信噪比为6 dB时,检测比率可提升近20%,信噪比为3 dB时,可提升近10%。

关键词: 直方图概率多假设跟踪, 低信噪比, 多目标, 检测前跟踪, 泊松分布, 期望极大化方法

Abstract: One of the main disadvantages of the Histogram Probability Multi-Hypothesis Tracking (H-PMHT) method and its variant Poisson distribution Histogram Probability Multi-Hypothesis Tracking (P-HPMHT) method is that their measurement model only considers the background clutter and does not consider the sensor noise, resulting in lower detection probability under low signal-to-noise ratio conditions. To overcome this problem, an improved H-PMHT with the sensor noise model is proposed. By introducing sensor noise into the measurement model, the ability to track and detect targets with low signal-to-noise ratio is significantly improved. The calculation amount of the method proposed maintains a linear relationship with the target number, and it is still suitable for the situation with many targets. Simulation experiments show that when the false tracking ratio is 1/1000, the detection rate of this method can be increased by nearly 20% when the signal-to-noise ratio is 6 dB, and by nearly 10% when it is 3 dB.

Key words: histogram probability multi-hypothesis tracking, low signal-to-noise ratio, multi-target, track before detect, Poisson distribution, expectation maximization

中图分类号: