厚尾噪声条件下的学生t泊松多伯努利混合滤波器

A Student’s t Poisson multi-Bernoulli mixture filter in the presence of heavy-tailed noise

中文关键词  随机有限集  多目标跟踪  学生t混合  厚尾噪声  泊松多伯努利混合

英文关键词  random finite set  multi-target tracking  Student’s t mixture  heavy-tail noise  Poisson multi-Bernoulli mixture

中文摘要

      针对运动过程和观测过程均受到异常噪声干扰的复杂不确定性多目标跟踪问题, 本文创新性地提出了学生t混合泊松多伯努利混合滤波器. 首先, 直接将广域分布的异常噪声特性建模为学生t分布. 随后, 将泊松多伯努利混合滤波器的泊松点过程(PPP)和多伯努利混合(MBM)的概率密度参数合理的近似为学生t混合形式. 其次, 基于多目标概率密度的学生t混合模型, 详细推导了泊松多伯努利混合滤波器学生t混合共轭先验形式, 建立了学生t混合泊松多伯努利混合的闭式递推框架. 最后, 通过带显著拖尾分布特性的过程噪声和量测噪声共同干扰的复杂多目标跟踪仿真实验, 验证了所提滤波算法的有效性.

英文摘要

      Aiming at the complex uncertainty multi-target tracking where both the motion process and observation process are disturbed by anomalous noise, this paper innovatively proposes a Student’s t mixture Poisson multi-Bernoulli mixture filter. First, the anomalous noise characteristics of the wide-area distribution are directly modeled as the Student’s t distribution. Subsequently, the probability density parameters of the Poisson point process (PPP) and the multi-Bernoulli mixture (MBM) of the Poisson multi-Bernoulli mixture filter are reasonably approximated by the Student’s t mixture form. Moreover, based on the Student’s t mixture model which approximates the multi-target probability density, the Student’s t mixture conjugate prior form of Poisson multi-Bernoulli mixture filter is derived in detail and a closed-form recursive framework of Student’s t mixture Poisson multi-Bernoulli mixture is established. Finally, the effectiveness of the proposed filtering algorithm is verified by complex multi-target tracking simulation experiments under the joint interference of process noise and measurement noise with significant trailing distribution characteristics.