海洋地磁参量实时测量野值的在线辨识与改正  

On-line Identification and Correction of the Outlier for the Oceanic Geomagnetic Parameter Measurement

作　　者：徐遵义[1,2] 魏东[1] 晏磊[2] 

机构地区：[1]山东建筑大学计算机科学与技术学院,济南250101 [2]北京大学地球与空间科学学院,空间信息集成与3S工程应用北京市重点实验室,北京100871

出　　处：《地球物理学进展》2012年第3期1034-1039,共6页Progress in Geophysics

摘　　要：地磁参量实时测量野值的在线辨识与改正是影响地磁导航算法定位概率和精度的重要因素.地磁导航定位的研究目前主要集中于匹配定位算法,地磁参量实时测量数据处理的研究较少.本文将多层递阶非平稳时间序列预测模型引入海洋地磁参量实时测量野值在线辨识与改正,利用多层递阶模型对实时测量值进行一步预测,根据迪克松准则进行野值辨识,利用中值滤波和一步预测值对野值进行改正.仿真数据和实测数据的实验室仿真结果表明,本文提出的野值在线辨识与改正算法不仅可以检测全部孤立型野值而且对多数斑点型野值也有较好的辨识效果,对产生野值处信号的复原误差小于5%.The outlier identification and correction of the geomagnetic parameter measurement is an important influencing factor for the matching probability and precision of the geomagnetic navigation and positioning algorithm. The present search of the geomagnetic navigation and positioning mainly focuses on the matching algorithm; there is less study on the real-time processing for the measured value of magnetic parameter. In this paper, the multi-level recursive method of the non-stationary time series has been introduce into the on-line identification and correction of the outliers for the oceanic geomagnetic parameter measurement. The one-step prediction of the real-time measured value is made by using the multi-level recursive method, the outlier identification is done by the Dixon criterion, and the correction is made by using the median filter and the one-step predictive value. The experiment has been made by using the emulation data and actual data in the laboratory, the results show that not only the all isolated type but also the most spot-type outlier value can be detected well and the signal reconstruction error is less than 5%.

关 键 词：关键词地磁测量 野值辨识 多层递阶方法 ARMA模型 狄克松准则 

分 类 号：TP39[自动化与计算机技术—计算机应用技术] P738[自动化与计算机技术—计算机科学与技术]