多机动目标跟踪的BFG-GMPHD算法

针对高斯混合概率假设密度(GMPHD)滤波算法中的机动目标跟踪问题,提出BFG-GMPHD算法,扩展了GMPHD滤波算法的适用范围。算法利用最佳拟合高斯(BFG)分布来近似目标动态模型中的状态转移矩阵和过程噪声的协方差矩阵,实现了滤波器与不同动态模型的匹配;在对BFG分布进行递推时,引入了模型概率更新过程,解决了BFG仅依赖于先验信息的问题;并利用UKF算法对GMPHD的高斯分量进行递推,使得算法能处理量测方程为非线性的情况。仿真实验表明,BFG-GMPHD算法能快速匹配目标模型的变化,实现对多机动目标的有效跟踪,准确估计出目标的数目和状态。

A BFG-GMPHD Algorithm for Multiple Maneuvering Targets

In order to track maneuvering target with Gaussian mixture probability hypothesis density(GMPHD)filtering algorithm,the BFG-GMPHDalgorithm is proposed based on the best fitting Gaussian (BFG) algorithm which canrelax restrictions of the GMPHD filter algorithm. The proposed algorithm utilizes BFG distribution to approximate the state transition matrix and process noise covariance matrix oftarget kinematicmodel to match the filter with different kinematicmodel. The model probability update process is introduced into the recursion of BFG to solve the problem that therecursion of BFG is only determined by priori information;and the recursiveGMPHD Gaussian components is obtained by UKF algorithm which makes the algorithm have the ability to deal with nonlinear measurement equation. The simulation experiments show that the BFG-GMPHD algorithm can quickly match themodel probability and targetkinematic model, effectively track multiple maneuvering targets and accurately estimate targets number and state.