角闪烁噪声下的高斯和容积卡尔曼滤波算法

开展角闪烁噪声下的目标跟踪问题研究对提高传感器的探测性能具有重要意义,其中角闪烁噪声具有的非高斯特性是一个长期困扰研究者的难点。针对该问题,首先通过理论分析指出了容积粒子滤波(cubature particle filter,CPF)在角闪烁噪声下的性能缺陷。其次,基于高斯和滤波(Gaussian sum filter,GSF)框架和容积卡尔曼滤波(cubature Kalman filter,CKF)算法,提出了适用于角闪烁下的高斯和容积卡尔曼滤波(Gaussian sum cubature Kalman filter,GSCKF)算法,该算法将目标后验概率密度用高斯密度加权求和近似,通过多路并行的CKF实现状态预测与量测更新,同时利用模型降阶算法限制高斯分量数目的增长,能应用于非线性、非高斯条件的状态估计。最后,设计了仿真实验对GSCKF和CPF的跟踪精度、鲁棒性和计算复杂度进行了对比。

Gaussian sum cubature Kalman tracking filter with angle glint noise

Research on target tracking with glint noise is important to improve the detection performance of sensor, in which the glint noise’s non Gaussian property has puzzled researchers for a long time. To overcome this problem, the cubature particle filter’s performance deficiency in target tracking with the glint noise is theoretically analyzed. Then, a algorithm called Gaussian sum cubature Kalman filter (GSCKF) is proposed. Based on the methodology of Gaussian sum filter (GSF) and the cubature Kalman filter(CKF), the proposed algorithm models the non Gaussian noise and the state posterior distribution as finite weighted Gaussian mixture, and a bank of CKF is running in parallel where the filtering and predictive distributions are updated by using the CKF equations. Moreover, the proposed algorithm utilizes the model reduction techniques to limit the number of Gaussian components, thus it is suitable for non linear and non Gaussian state estimation. In order to compare the performance of the two non Gaussian algorithms, comparative experiments between GSCKF and CPF from the three aspects of tracking accuracy, robustness and computational complexity are carried out.