角闪烁下基于变分贝叶斯-交互式多模型的目标跟踪

开展角闪烁噪声下的目标跟踪研究对提升传感器的探测性能具有重要意义，其中角闪烁噪声具有的分布未知和非平稳特性是长期困扰研究者的难点。针对该问题，该文首先给出角闪烁下基于变分贝叶斯参数学习的跟踪滤波理论框架。其次，提出一种联合估计运动状态和闪烁噪声分布的变分贝叶斯-交互式多模型(VB-IMM)算法，该算法通过设计多个并行的跟踪模型处理角闪烁的跟踪问题，同时利用变分贝叶斯方法实现闪烁噪声分布参数的在线学习，并反馈给跟踪模型，实时调整跟踪模型参数。最后，设计了仿真实验对算法在闪烁噪声分布未知和非平稳条件下的跟踪性能进行了验证，同时对算法的计算复杂度进行了仿真分析。仿真结果表明，在量测噪声分布未知和非平稳条件下，VB-IMM具有较高的跟踪精度，且算法复杂度较小，易于实现。

Variational Bayesian-interacting Multiple Model Tracking Filter with Angle Glint Noise

Research on target tracking with glint noise is important to improve detection performance of sensor, in which the glint noise’s unknown distribution and non-stationary property puzzle researchers for a long time. In order to solve this problem, the tracking theoretical framework of variational Bayesian parameter learning with glint noise is firstly introduced. Then, a novel algorithm called Variational Bayesian-Interacting Multiple Model (VB-IMM) is proposed to estimate the system states as well as the unknown glint noise’s distribution. The proposed algorithm designs a bank of tracking filters in parallel with different measurement noise. Moreover, the algorithm utilizes variational Bayesian method to learn distribution parameters of the glint noise online and feed these parameters back to the tracking filters to revise the filters. In order to validate the performance of this algorithm, comparative experiments are carried out from two aspects of tracking accuracy and computational complexity. Simulation results verify good performance of tracking error and low computational complexity of the proposed algorithm.