基于交互式多模型的不敏卡尔曼概率假设密度滤波算法  

Unscented Kalman probability hypothesis density filter based on interacting multiple model

作　　者：胡振涛[1] 张谨[1] 郭振[1] 

机构地区：[1]河南大学图像处理与模式识别研究所,河南开封475004

出　　处：《控制与决策》2016年第12期2163-2169,共7页Control and Decision

基　　金：国家自然科学基金项目(61300214);中国博士后科学基金项目(2014M551999);河南省高校科技创新团队支持计划项目(13IRTSTHN021)

摘　　要：针对非线性高斯场景下目标数目未知或随时间变化的机动多目标跟踪问题,提出一种基于交互式多模型的不敏卡尔曼概率假设密度滤波算法.首先,在高斯混合概率假设密度滤波框架下,结合不敏卡尔曼滤波中状态预测和量测更新的实现机理,构建一种不敏卡尔曼概率假设密度滤波器;然后,通过引入交互式多模型方法中状态模型软判决机制,实现对目标机动过程中运动模式不确定的处理;最后,通过理论分析和仿真结果验证了所提出算法的可行性和有效性.Aiming at the maneuvering multi-target tracking problem with unknown or time-varying number of targets in the nonlinear Gaussian condition, an unscented Kalman probability hypothesis density filter based on the interactive multiple model is proposed. Firstly, combining with the implementation mechanism of state prediction and measurement update in the unscented Kalman filter, the unscented Kalman probability hypothesis density filter for the nonlinear Gaussian system is constructed in the framework of the Gaussian mixture probability hypothesis density filter. On this basis, the motion pattern uncertainty in the target maneuvering system is solved by utilizing the soft decision mechanism of the state model in the interactive multiple model algorithm. Theoretical analysis and simulation results verify the feasibility and effectiveness of the proposed algorithm.

关 键 词：机动多目标跟踪 概率假设密度滤波 交互式多模型 不敏卡尔曼滤波 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]