【作者】 胡振涛；

【导师】 刘先省；

【作者基本信息】 河南大学 ， 应用数学， 2006， 硕士

【摘要】 多传感器数据融合可以充分利用多传感器的量测资源,将多个传感器在时间和空间上的互补和冗余信息按照某种算法或准则进行综合,进而增强了对于机动目标状态估计的精确性、可靠性及对抗环境下的生存性。基于此,本文针对多传感器数据融合技术在机动目标状态估计中的应用进行了系统的研究,其主要工作如下:1、基于“当前”统计模型中加速度极限值的预先设定对于滤波效果影响的分析,利用目标机动状况与相邻采样时刻间位置估计量变化之间的函数关系实现噪声方差自适应,进而提出了一种修正的模型,并结合卡尔曼滤波递推算法,提出了一种改进的自适应滤波算法。2、针对Kalman滤波及其一些改进算法中量测方差预先设定对状态估计精度的影响,基于时空分析和最小二乘估计思想给出了一种在线估算量测方差的方法,以充分利用多个传感器单次采样数据和单个传感器多次采样数据中蕴含的冗余和互补信息。另外,依据Kalman递推滤波算法中滤波误差协方差阵具有表示目标当前状态不确定性的性质,并结合信息守恒原理,给出了一种分布式航迹权系数的动态分配方法。3、将Jerk模型与强跟踪滤波算法有机地结合,并利用时空综合分析和极大似然估计的思想推导出了一种系统量测方差自学习修正方法,以优化强跟踪滤波算法中次优渐消因子和滤波增益的在线选择,同时根据多传感器数据融合具有改善滤波精度的性质,进而给出一种基于Jerk模型的多传感器数据融合算法。4、根据模糊集合理论中隶属函数的特性,定义了一种量测数据间的支持度函数以有效地度量数据间相互支持度,该定义不仅继承了一致性数据融合算法充分提取量测信息中隐含的互补和冗余信息的优点,同时有效避免了门限预先设定的问题。另外,将该方法移植于基于时间序列分析的机动目标状态估计问题上,从而得到系统中各局部航迹估计权系数的合理分配。更多还原

【Abstract】 Multi-sensor data fusion can synthesize the spatio-temporal redundancy and complementary information according to some algorithms and criterions based on exploiting sufficiently multi-sensor measuring information, in the course of this synthesizing process, the accuracy and reliability and survivability of state estimation for maneuvering target are enhanced greatly. Based on above performances the applications of multi-sensor data fusion in state estimation for maneuvering target is studied systemically. The main work includes:Based on the analysis that the extreme value of acceleration presupposed causes influence in the“current”statistical model, a modified model is given, which utilizes the functional relationship between maneuvering status and estimation of the neighboring intersample position vector to carry out the self-adaptive of the process noise variance. Then combining with the recursive characteristic of Kalman filter, an improved self-adaptive filtering algorithm is presented.In some algorithms based on Kalman filter and its extension, the presupposition of measurement variance leads to the descent of state estimation accuracy. A self-adaptive estimation algorithm of measurement variance is presented based on spatio-temporal analysis and LSE, which make full use of redundancy and complementary information from the single sampled data of multi-sensors and the multi-times sampled data of single sensor. Furthermore, utilizing the characteristic that filtering error covariance expresses filtering precision and the principle of information conservation, the dynamic and reasonable distribution of distributed tracks weight coefficient is accomplished.Jerk model and strong tracking filter is organically assembled, and based on spatio-temporal synthetically analysis and LME, a self-learning estimation method of the system measurement variance is given. The method improves obviously the更多还原