一种快速的多目标跟踪非线性滤波算法

多机动目标跟踪问题是目前目标跟踪领域的一个重要研究方向,而数据关联与跟踪维持是多目标跟踪的核心部分。利用支持向量机在分类识别方面的优势,研究了基于支持向量机的数据关联方法。在此基础上,采用交互式多模型算法和无味卡尔曼滤波相结合的方法研究了多机动目标的跟踪问题。在该方法中,目标的运动状态和方位误差由选定的采样点来近似,在每个更新过程中,采样点随着状态方程传播并随非线性测量方程变换,得到目标的运动状态和方位误差的均值,避免了对非线性方程的线性化,至少给出最佳估计的二阶近似。与传统的扩展卡尔曼(EKF)方法进行了仿真比较,仿真结果表明了该算法的有效性。

A fast nonlinear Kalman filter algorithm for multiple target tracking

Multiple maneuvering targets tracking is a research focus of target tracking at present.Data association and tracking maintaining are cores of multi-target tracking.Considering the advantage of Support Vector Machine(SVM) in classification and identification,we studied the method of data association of multiple maneuvering targets based on SVM.Then we studied the tracking problem of multiple maneuvering targets by using the combination of interacting multi-model and unscented Kalman filter algorithm.In the algorithm,the state and the biased error of the target are approximated by specified sample points.During each updating process,the sample points are propagated with the state equation and transformed with the nonlinear measurement equation.From these sample points,the posterior mean and covariance of the target state are obtained,and the biased error are accurately computed to the second order.Compared with the extended Kalman filter,the simulation results demonstrate the availability of the studied tracking algorithm.