目标机动和观测野值的双假设滤波辨识算法

摘要

针对目标跟踪观测过程中野值和机动互存的辨识问题,根据标准Kalman滤波算法不具有容错性的特征,提出利用Kalman滤波收敛时间辨识野值抑或机动导致观测值异常的方法,以减少测量值准确有用信息的丢失.然后,采用改进3σ准则对近似服从正态分布的小测量域内观测数据进行粗略预处理,并对异常值建立了残差扰动因子的双滤波器辨识,同时以设置同步并行动态的时间计时为判决条件,有效减小了后续目标跟踪的误差.仿真实例表明:所提出的双假设辨识算法能够实现实时辨识,且有效跟踪目标.%Problems exist in the identification of outliers and maneuvering data in the target tracking process because the standard Kalman filtering algorithm does not have the feature of fault tolerance.A method based on Kalman filtering convergence time is put forward to identify abnormal observed va-lues caused by outliers or maneuver.The method can reduce the loss of useful information in the measured values.An improved criterion with statistical characteristics is introduced in order to prepro-cess observation data in a small field that obey the normal distribution approximately;and then a de-tection algorithm about maneuver based on residual vector is established.At the same time,the syn-chronous parallel dynamic time threshold domain is set,so a series of abnormal points are identified as outliers or maneuver data,for Kalman filter tracks the estimated optimal target value.The experi-ments illustrate that the results are satisfactory with identification of continuous abnormal values, which proves that the two hypotheses filter identification is feasible in the target tracking process.