递推样条滤波的工程化应用研究

Research of Recursive Spline Filter for Engineering Applications

递推样条模型是目标跟踪中的一类高效建模方法,理论研究与仿真实验表明它有很高的精度和稳健性,但鉴于工程应用问题的复杂性,还需根据实测数据对模型进行验证和改造。以全测速体制下的弹道目标跟踪为背景,在分析实测数据复杂性的基础上,进行了模型工程化应用研究:确定了状态方程和模型更新策略;基于"当前"统计思想,提出了机动适应性更强的"当前"样条模型;引入面向状态的野值检择方法,提高了观测数据质量低时的弹道估计精度;给出了基于不完全预测时间阈值判断滤波发散趋势的方法,和重新初始化的发散应急机制。定性分析和计算实例验证了方法的可行性和有效性。

According to theoretic analysis and simulation result,the recursive spline modeling method is considered to possess high accuracy and robustness for real-time target tracking.However,in view of the complexity in engineering application,practical validation and potential improvement to the model should be desiderated according to actual test data.On the basis of the complexity analysis of actual test data,several strategies for improving practicability are proposed in detail.Firstly,state transfer equations and model updating rules are chosen from several candidates.Secondly,a new ＂current＂ spline model more adaptive for strong maneuvers is suggested.Thirdly,based on the redundancy of the observations,a new state-oriented outlier detection method is introduced.Finally,utilizing the time threshold of incomplete prediction,a criterion for judging that whether divergences occur is given in combination with a re-initialization technique for dealing with these divergences.Feasibility and validity of these strategies are proved by qualitative analysis and calculation of actual test data.