变结构多模型估计单Kalman滤波跟踪机动目标算法

在机动目标跟踪中，如果我们估计出目标的机动量，就可以使用单Kalman滤波来跟踪目标且跟踪精度和IMM滤波方法接近。文献l]中提出用N个离散加速度Ui覆盖目标机动量，然后用它们加权的方法来得到机动量的估值。这些离散加速度量Ui的选择是个难点。本文提出使用变结构多模型方法来估计目标的机动量，该方法可以实时改变参与估计机动量的模型数目和参数，排除一些不必要的模型来减少模型数目和竞争。仿真实验结果表明，在大大缩短计算时间的前提下，本算法依然能够获得与原算法一样的跟踪精度。

A Single Kalman Filtering Algorithm with Variable Structure Multiple-Model Estimation for Maneuvering Target Tracking

For maneuvering target tracking, after estimating maneuver of the target, single Kalman filtering algorithm can be used to obtain tracking accuracy similar to Interacting Multi-Model (IMM) algorithm. Some researchers chose to use N discrete acceleration values Ui to cover the possible range of maneuver and estimate the maneuver by weighting1]. In such process, It is difficult to choose Ui properly. In this paper, We propose to use variable structure multiple-model to solve this problem. By this approach, the number of models and parameters can be adjusted in real-time. Unnecessary models can be excluded from the action of estimating as to reduce excessive competition. Simulation results show that the time for estimation can be shorten, while the accuracy can be maintained as good as the original algorithm..