School of Automation, Hangzhou Dianzi University, Xiasha Higher Education Zone, 2nd Street, Hangzhou 310018, China.
School of Computer Science and Engineering, Tianjin University of Technology, 391 Bingshuixi Road, Xiqing District, Tianjin 300384, China.
Sensors (Basel). 2018 Jun 1;18(6):1772. doi: 10.3390/s18061772.
Multitarget tracking in clutter using bearings-only measurements is a challenging problem. In this paper, a performance improved nonlinear filter is proposed on the basis of the Random Finite Set (RFS) theory and is named as Gaussian mixture measurements-based cardinality probability hypothesis density (GMMbCPHD) filter. The GMMbCPHD filter enables to address two main issues: measurement-origin-uncertainty and measurement nonlinearity, which constitutes the key problems in bearings-only multitarget tracking in clutter. For the measurement-origin-uncertainty issue, the proposed filter estimates the intensity of RFS of multiple targets as well as propagates the posterior cardinality distribution. For the measurement-origin-nonlinearity issue, the GMMbCPHD approximates the measurement likelihood function using a Gaussian mixture rather than a single Gaussian distribution as used in extended Kalman filter (EKF). The superiority of the proposed GMMbCPHD are validated by comparing with several state-of-the-art algorithms via intensive simulation studies.
基于量测的多目标杂波环境下的跟踪是一个具有挑战性的问题。本文在随机有限集（RFS）理论的基础上提出了一种性能改进的非线性滤波器，称为基于高斯混合量测的基数概率假设密度（GMMbCPHD）滤波器。该滤波器能够解决两个主要问题：量测起源不确定性和量测非线性，这构成了基于量测的杂波环境下多目标跟踪的关键问题。对于量测起源不确定性问题，所提出的滤波器估计多个目标的 RFS 强度并传播后验基数分布。对于量测起源非线性问题，GMMbCPHD 使用高斯混合而不是扩展卡尔曼滤波器（EKF）中使用的单个高斯分布来近似量测似然函数。通过与几种最先进的算法进行密集的仿真研究，验证了所提出的 GMMbCPHD 的优越性。
