改进卡尔曼滤波的目标跟踪研究  

Study on Target Tracking Based on Improved Unscented Transform Kalman Filtering

作　　者：杨柳[1] 

机构地区：[1]牡丹江师范学院计算机系,黑龙江牡丹江157012

出　　处：《计算机仿真》2010年第9期351-355,共5页Computer Simulation

基　　金：黑龙江省教育厅科学技术研究项目(11531390);牡丹江师范学院科学技术研究项目(KY2008001)

摘　　要：根据扩展卡尔曼滤波(EKF)是目标跟踪中的常用算法,但方法涉及Taylor级数的展开,引起了运算量较大,运算结果的均值和协方差只精确到一阶,使得滤波精度不高。为提高滤波速度和精度,将Unscented变换与卡尔曼滤波相结合,建立了Unscented卡尔曼滤波(UKF)数学模型。Unscented变换是基于高斯分布理论,通过Sigma点能够获取精确到三阶矩均值和协方差,提高了滤波精度。计算仅涉及标准的向量和矩阵操作,不需要计算非线性函数的Jacobian或者Hessians矩阵,提高了滤波速度。通过运动实验进行仿真对比,结果表明对于非线性目标跟踪系统,UKF算法具有更高的滤波精度和稳定性。Extended Kalman Filtering is in common usage in target tracking currently,but this method involves Taylor series expansion,so it has high computational cost,the typical value and covariance only has first derivative precise,so the precision is lower. To aim at raising the speed and precision,Unscented transform has been drawn into Kalman Filtering in this paper,and the model of UKF has been published. Because Unscented transform is based on the theory of Gaussian Distribution,so the typical value and covariance has Third-order precise through Sigma spots in this method,the precision of filtering has been raised. It only involves the operations of standard vector and matrix,and has no use for the calculation of non-liner matrix of Jacobian and Hessians,so the speed of filtering has been increased. Through movement emulation,it shows that for nonlinearity target tracking system,UKF gives better accuracy and stability.

关 键 词：目标跟踪 卡尔曼滤波 高斯分布 滤波精度 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]