δ-广义标签多伯努利滤波算法的非线性扩展  

Nonlinear extension of δ-generalized labeled multi-Bernoulli filtering algorithm

作　　者：齐美彬[1] 胡晶晶 程佩琳 靳学明[2] QI Meibin;HU Jingjing;CHENG Peilin;JIN Xueming(School of Computer and Information,Hefei University of Technology,Hefei 230009,China;China Electronics Technology Group Corporation 38th Research Institute,Hefei 230088,China)

机构地区：[1]合肥工业大学计算机与信息学院,安徽合肥230009 [2]中国电子科技集团第38研究所,安徽合肥230088

出　　处：《系统工程与电子技术》2021年第12期3571-3578,共8页Systems Engineering and Electronics

基　　金：国家自然科学基金(61771180)资助课题。

摘　　要：针对高斯混合(Gaussian mixture,GM)实现的变分贝叶斯-δ-广义标签多伯努利(variational Bayesian-δ-generalized labeled multi-Bernoulli,VB-δ-GLMB)滤波算法在非线性场景下跟踪性能较低这一问题,结合基于临近点算法(proximal point algorithm,PPA)和变分贝叶斯(variational Bayesian,VB)的迭代优化与容积卡尔曼滤波(cubature Kalman filtering,CKF),提出一种适用于非线性模型的机动多目标跟踪算法。该算法在GM-VB-δ-GLMB的基础上采用逆伽马(inverse-Gamma,IG)和高斯乘积混合分布近似量测噪声协方差和状态联合后验分布;利用PPA-CKF-VB(PCKF-VB)方法对传递过程中的高斯项参数进行预测更新;最后为提高滤波精度进行变分贝叶斯容积RTS(VB cubature Rauch-Tung-Striebel,VB-CRTS)平滑。仿真结果表明,对于量测噪声未知的非线性系统,所提的算法与现有的VB-δ-GLMB算法相比目标跟踪精度有显著提高。Aiming at the low tracking performance of the Gaussian mixture(GM)variational Bayesian-δ-generalized labeled multi-Bernoulli(VB-δ-GLMB)filtering algorithm in nonlinear scenes,a maneuvering multi-target tracking algorithm is proposed,which combined with the iterative optimization based on proximal point algorithm(PPA)and(variational Bayesian,VB)approximate and cubature Kalman filter(CKF),to suit for nonlinear models.Based on GM-VB-δ-GLMB,the proposed algorithm uses inverse Gamma(IG)and Gaussian product mixture distribution to approximate the joint posterior density of the measurement noise covariance and state.Then,PPA-CKF-VB(PCKF-VB)method is used to predict and update the Gaussian term parameters in the transfer process.Finally,smoothing of variational Bayesian cubature Rauch-Tung-Striebel(VB-CRTS)is utilized to improve the filtering accuracy.Simulation results indicate that the target tracking accuracy of the proposed algorithm is obviously improved compared with the existing VB-δ-GLMB algorithm for the nonlinear system with unknown measurement noise.

关 键 词：δ-广义标签多伯努利算法 非线性模型 容积卡尔曼滤波 临近点算法 变分贝叶斯近似 

分 类 号：TN953[电子电信—信号与信息处理]