基于IMM-UKF的转弯机动目标跟踪

孙 宇,薛斌党

摘 要:针对目标作转弯机动时产生运动模式的不确定性和运动模型的非线性问题,提出基于Unscented卡尔曼滤波器的交互多模型算法。该算法采用带有极坐标系速度的转弯模型和二维Singer模型作为模型集,将Unscented卡尔曼滤波取带传统的扩展卡尔曼滤波解决转弯模型的非线性,同时在模型交互时使用Unscented变换取代雅可比矩阵解决目标状态转换时的非线性。通过Monte-carlo仿真表明,与标准交互多模型方法相比,基于Unscented卡尔曼滤波器的跟踪算法具有很好的跟踪性能。

关键词:Unscented卡尔曼滤波;交互多模型;目标跟踪;非线性

中图分类号:TN953文献标识码:A

文章编号:1004-373X(2009)12-081-04

Maneuvering Target Tracking with Coordinated Turn Motion Based on IMM-UKF

SUN Yu1,XUE Bindang2

(1.School of Instrument Science and Opto-electronics Engineering,Beihang University,Beijing,100083,China;

2.School of Astronautics,Beihang University,Beijing,100083,China)

Abstract: Aiming at the target motion uncertainty and dynamic model nonlinear when target has a coordinated turn maneuver,Interacting Multiple Model(IMM)algorithm based on Unscented Kalman filter is proposed.The coordinate turn model with polar velocity and the 2D Singer model are chosen as the model set,and the Unscented Kalman filter is proposed to handle nonlinear of the state model.Meanwhile,Unscented transform,instead of Jacobi Matrix is taken in model interaction to solve the linearized loss problem when the states are transformed between the different models.The result of Monte-Carlo simulation indicates that this algorithm works better than the traditional IMM.

Keywords:Unscented Kalman filter;interacting multiple models;target tracking;nonlinear

0 引 言

转弯模型是机动目标运动模型中的重要模型之一。目前已提出了多种转弯模型,在未知角速度的模型中通常为带有直角坐标系速度的转弯模型(CT)和带有极坐标速度的转弯模型(HT)[1]。这两种模型均为非线性模型,一般认为后者的性能要强于前者,但是非线性程度也较高[2]。交互多模型(Interacting Multiple Mode,IMM)方法是一种有效跟踪转弯机动目标的方法 [3],该方法的性能在很大程度上取决于所选的模型集是否能描述不同的运动状态。由于转弯模型中目标的状态为位置、速度和角速度,与其他的模型,例如Singer模型的状态不同,所以基于IMM的转弯机动目标跟踪算法中,模型状态之间必须要进行转换,例如角速度转换为加速度等,通常转换函数都为非线性函数。传统转换方法为计算转换函数的雅可比矩阵(ET)[4],而雅可比矩阵的计算量较大,同时精度只能达到一阶的精度。……