基于扩展卡尔曼滤波的交互式多模型跟踪算法研究  

Research on interacting multiple model tracking algorithm based on extended Kalman filter

作　　者：陈晓楠[1] 张子阔 索继东[1] 罗超发 杜振邦 CHEN Xiaonan;ZHANG Zikuo;SUO Jidong;LUO Chaofa;DU Zhenbang(School of Information Science and Technology,Dalian Maritime University,Dalian 116026,China;State Power Investment Group Shandong Marine Energy Development Co.,Ltd.,Qingdao 266034,China)

机构地区：[1]大连海事大学信息科学技术学院,辽宁大连116026 [2]国家电投集团山东海洋能源发展有限公司,山东青岛266034

出　　处：《现代电子技术》2024年第13期71-76,共6页Modern Electronics Technique

摘　　要：在辅助驾驶系统中,行人轨迹跟踪一直是一项有挑战性的任务,因为行人的回波信号中往往存在着许多干扰噪声。此外,行人在运动过程中可能会做出突然转身或其他改变方向的行为,这将直接导致行人运动轨迹呈现出非线性特征。针对上述问题,文中提出一种基于扩展卡尔曼滤波的交互式多模型跟踪(IMM-EKF)方法,适用于毫米波雷达对行人进行轨迹跟踪。首先,在扩展卡尔曼滤波算法(EKF)的基础上重构状态预测协方差矩阵,来补偿EKF非线性化过程中引入的误差;然后将改进的EKF作为交互式多模型算法(IMM)中的滤波器,根据行人运动特性选择匀速模型和协调转弯模型作为跟踪模型,利用所提出的IMM-EKF算法进行轨迹跟踪。实验结果表明,所提出的滤波算法较典型的EKF和改进的EKF算法,在跟踪滤波精度方面均有所提升,同时具备更优的跟踪鲁棒性。Pedestrian track tracking is always a challenging task in the assisted driving systems,because there is often a lot of interference noise in the pedestrian echo signal.In addition,pedestrians may make sudden turns or other directional changes during movements,which will directly lead to nonlinear features of pedestrian movement trajectory.In view of the above,an interacting multiple model tracking method based on extended Kalman filter(IMM-EKF)is proposed,which is suitable for millimeter-wave radar to track pedestrians.On the basis of the extended Kalman filter(EKF)algorithm,the state prediction covariance matrix is reconstructed to compensate the errors brought in the nonlinear process of EKF,and then the improved EKF is used as the filter in the interacting multiple model algorithm(IMM).According to pedestrian movement characteristics,uniform speed model and coordinated turn model are selected as tracking models,and the proposed IMM-EKF algorithm is used for track tracking.The experimental results show that the proposed filtering algorithm can improve the accuracy of tracking filter to some extent and has better tracking robustness in comparison with the typical EKF algorithm and the improved EKF algorithm.

关 键 词：行人轨迹跟踪 扩展卡尔曼滤波 交互式多模型 毫米波雷达 状态预测协方差矩阵 辅助驾驶 

分 类 号：TN953.6-34[电子电信—信号与信息处理]