面向多目标跟踪的PHD滤波多传感器数据融合算法

Algorithm of Multi-sensor Data Fusion of PHD Filtering for Multi-target Tracking

针对密集杂波环境下单传感器应用高斯混合PHD算法进行多目标跟踪时性能下降的问题,提出一种面向多目标跟踪的PHD滤波多传感器数据融合算法。首先构建了基于高斯混合PHD滤波的多传感器数据融合系统框架,各传感器利用高斯混合PHD滤波算法进行局部状态估计,然后对各传感器的状态估计结果进行关联度计算,最后通过构建自适应混合参数,引入协方差交叉算法对关联状态进行融合。仿真实验表明,与单传感器高斯混合PHD多目标跟踪算法相比,所提算法有效提高了目标数量和状态的估计精度。

For the problem of tracking performance degradation of the single sensor in the denseclutter environment using the Gaussian mixture PHD algorithm,a multi-sensor data fusion algorithmfor multi-target tracking based on PHD filters is proposed.First,a multi-sensor data fusion frameworkbased on the Gaussian mixture PHD filters is constructed.Then,each sensor processes local state isestimated by a Gaussian mixture PHD filter algorithm.After that,calculatethe correlated values of thestatesfrommultiple sensors is calculated.Finally,the associated states is fused through adopting acovariance intersection algorithm with an adaptive parameter modification.Simulation results show that,compared to the single sensor Gaussian Mixture PHD filter,the proposed algorithm can effectivelyimprove the estimation accuracy of the target number and state.