公共交通安全下基于变分贝叶斯期望最大化的扩展目标跟踪方法

Extended Target Tracking Method Based on Variational Bayesian Expectation Maximization in Public Transportation Safety

公共交通安全场景中,在量测噪声是闪烁噪声的情况下,扩展目标的跟踪性能将会急剧下降。针对这一问题,提出一种闪烁噪声下基于变分贝叶斯期望最大化(VBEM)的扩展目标跟踪算法。该算法将目标的扩展形状建模为椭圆随机超曲面模型,并将其嵌入CPHD滤波器,通过变分贝叶斯期望(VBE)和变分贝叶斯最大化(VBM)两个迭代步骤,对目标状态和联合概率密度中的相关参数进行预测和更新。仿真实验表明,所提算法可以自适应地跟踪未知数目,且量测噪声为闪烁噪声条件下的多扩展目标,同时具有较高的跟踪精度。

In public traffic safety scenarios,when the measurement noise is flickering noise,the tracking performance of extended target will decrease dramatically.In order to solve this problem,an extended target tracking algorithm based on variational Bayesian expectation maximization(VBEM)in flickering noise is proposed.The extended shape of the target is modeled as an elliptic stochastic hypersurface model and embedded in CPHD filter.The parameters of the target state and joint probability density are predicted and updated by two iteration steps of variational Bayesian expectation(VBE)and variational Bayesian maximization(VBM).The simulation results show that the proposed algorithm can track the unknown number adaptively,and the measurement noise is a multi-extended target with high tracking accuracy under the condition of flicker noise.