0606242M-估计的UKF算法及其在运动估计中的应用.pdfVIP

编号： 0606242 基于M-估计的UKF算法及其在运动估计中的应用* 1 1 1,2+ 周露平 王智灵 陈宗海 1(多媒体计算与通信教育部－微软重点实验室中国科学技术大学自动化系，安徽合肥 230027) 2(模式识别国家重点实验室（中国科学院 自动化研究所），北京 100080) 摘要： 针对运动模型的非线性和运动估计的鲁棒性问题，提出了一种结合 M-估计等价加权原理的鲁棒无迹卡尔曼滤波 (UKF)算法(称为 M-UKF) 。该方法首先利用 UKF 算法获得初步的运动参数的估计，然后利用 M-UKF 算法得到较为准确的 估计值。将 M-估计与 UKF 方法相结合，既解决了针对运动模型非线性的估计问题，又能较好的克服离群数据的干扰，大 大提高了估计的鲁棒性。通过模拟数据的仿真和实际图像序列的测试说明了该方法的有效性。 关键词：运动估计，卡尔曼滤波，无迹卡尔曼滤波，M-估计，鲁棒估计 中图分类号：TP391 文献标识码：A The UKF Algorithm Based on M-estimators and Its Application in Motion Estimation 1 1 1,2+ ZHOU Lu-ping , WANG Zhi-ling , CHEN Zong-hai 1(MOE-MS Key Laboratory of Multimedia Calculation and Communication, Department of Automation，University of Science and Technology of China, Hefei, 230027, P.R.China) 2(National Laboratory of Pattern Recognition (Institute of Automation, Chinese Academy of Sciences), Beijing 100080, P.R.China) +Corresponding author:Phn:+86-551-3606104, E-mail: chenzh@ustc.edu.cn Abstract: To solve the problems of nonlinear of motion model and robust of motion estimation, we present a new robust unscented kalman filter (UKF) method, called M-UKF, which is combined with the principle of the equivalent weight of M- estimate. In this method, we first use the UKF to get the initial estimation of the parameters of motion; then M-UKF is utilized to get more accurate estimations. The combination of M-estimators and UKF can not only solve the nonlinear problems but also conquer the inflations of the outliers, which has greatly improved the robust of estimation. Finally, the experiments demonstrate the validity of the method in this paper. Key words: motion estimation, kalman filter ，unscented kalman filter, M-estimators, robus