δ-广义标记多伯努利滤波器的非线性应用扩展  

Nonlinear Application Extension for δ-Generalized Labeled Multi-Bernoulli Filter

作　　者：侯利明 连峰[1,2] 王伟 HOU Liming;LIAN Feng;WANG Wei(Ministry of Education Key Lab for Intelligent Networks and Network Security,Xi’an Jiaotong University,Xi’an 710049,China;Shaanxi Key Laboratory of Integrated and Intelligent Navigation,Xi'an 710068,China;The Navigation Key Laboratory of the 20th Research Institute of China Electronics Technology Corporation,Xi'an 710068,China)

机构地区：[1]西安交通大学智能网络与网络安全教育部重点实验室,西安710049 [2]陕西省组合与智能导航重点实验室,西安710068 [3]中国电子科技集团公司第二十研究所导航重点实验室,西安710068

出　　处：《西安交通大学学报》2019年第6期109-116,共8页Journal of Xi'an Jiaotong University

基　　金：国家自然科学基金资助项目(61473217);陕西省组合与智能导航重点实验室开放基金资助项目(SKLIIN-20180110)

摘　　要：针对非线性模型下δ-广义标记多伯努利(δ-GLMB)滤波器的序贯蒙特卡洛(SMC)实现过程计算复杂度过高、难以实现快速准确滤波的问题,给出了δ-GLMB滤波器的积分卡尔曼高斯混合(QK-GM)实现过程。该算法基于Gauss-Hermite数值积分规则获取一组带权重的积分点,利用这些积分点求取多目标密度函数的均值和协方差矩阵。将该算法与已有的扩展卡尔曼高斯混合(EK-GM)实现、无味卡尔曼高斯混合(UK-GM)实现和SMC实现在不同的杂波强度和检测概率条件下就多目标跟踪精度和时间消耗等方面做了较为详细的对比,结果表明,与SMC实现方法相比,QK-GM-δ-GLMB算法能以完全可接受的时间开销为代价,将多目标跟踪精度提高10%以上。该算法为δ-GLMB滤波器在非线性场景中的应用提供了一种新的实现方法。Quadrature Kalman Gaussian mixture (QK-GM) implementation for δ-generalized labeled multi-Bernoulli (δ-GLMB) filter is proposed to solve the problem that the sequential Monte Carlo (SMC) implementation for δ-GLMB filter in nonlinear model is too complex to realize fast and accurate filtering. The algorithm is based on the Gauss-Hermite quadrature rule to obtain a set of weighted quadrature points, and these points are then utilized to calculate mean values and covariance matrixes of the multiobjective density functions. The proposed QK-GM implementation is compared with the existing algorithms, including extended Kalman Gaussian mixture (EK-GM), unscented Kalman Gaussian mixture (UK-GM) and SMC implementation, in detail in terms of tracking accuracy and time consumption for multi-target tracking in the presence of different clutter densities and detection probabilities. Simulation results and a comparison with SMC method show that the QK-GM implementation of δ-GLMB filter improves the multi-target tracking accuracy by more than 10% at the expense of a completely acceptable time cost. The proposed algorithm may provide a new method for applications of δ-GLMB filter in nonlinear scenes.

关 键 词：δ-广义标记多伯努利 积分卡尔曼 高斯混合 多目标跟踪 

分 类 号：TP393[自动化与计算机技术—计算机应用技术]