论文部分内容阅读
概率假设密度(PHD)滤波算法已被证明是实时多目标跟踪的有效方法,但现有这些基于PHD滤波的方法假设量测噪声协方差先验已知,而实际中量测噪声协方差可能是未知或随着环境改变而变化。针对这一问题,提出了一种适用于非线性量测模型的自适应噪声协方差多目标跟踪算法。该算法以PHD滤波为基础,采用容积卡尔曼(CK)技术近似非线性量测模型,利用逆威沙特(IW)分布描述量测噪声协方差分布,通过变分贝叶斯(VB)近似技术迭代估计量测噪声协方差和多目标状态联合后验密度。仿真结果表明,本文所提算法可有效估计量测噪声协方差,同时实现准确的目标数和目标状态估计。
Probability hypothesis density (PHD) filtering algorithms have proven to be effective methods for real-time multi-target tracking, but existing methods based on PHD filtering assume that the covariance of measurement noise is known a priori and that the covariance of the actual measurement noise may be Unknown or as the environment changes. To solve this problem, an adaptive noise covariance multi-target tracking algorithm suitable for non-linear measurement models is proposed. Based on the PHD filter, this algorithm uses the volumetric Kalman (CK) technique to approximate the nonlinear model and describes the noise covariance distribution by using the inverse Wishart distribution. By using the Variational Bayesian (VB) approximation Iterative estimation of covariance and multi-target state covariance post-test density estimation. Simulation results show that the proposed algorithm can effectively estimate covariance of measurement noise and achieve accurate target number and target state estimation at the same time.