利用改进3σ法获得残差序列的共模误差提取  

The common mode error extraction of residual series was obtained by improved 3σmethod

作　　者：李爱国[1] 平书哑 郭敏[1] LI Aiguo;PING Shuya;GUO Min(School of Surveying and Land Information Engineering,Henan Polytechnic University,Jiaozuo,Henan 454000,China)

机构地区：[1]河南理工大学测绘与国土信息工程学院,河南焦作454000

出　　处：《测绘科学》2023年第7期72-83,共12页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41905027)。

摘　　要：针对GNSS坐标时间序列的异常值、缺失点插值和共模误差问题,该文采用一种经验模态分解(EMD)与3σ组合算法处理异常值,获取干净的残差时间序列,然后使用Matlab软件的fillmissing函数进行插补,获得连续的时间序列,最后利用独立成分分析(ICA)和主成分分析(PCA)方法对残差序列进行共模误差的提取,分析共模误差的影响。研究结果表明,与LS-3σ相比,EMD-3σ方法的探测率提高了0.2%,说明新算法的探测效果更好。PCA和ICA方法滤波后的RMS值分别平均减少了约22.01%、10.96%,说明PCA和ICA均能有效地提取残差时间序列的共模误差,提高坐标时间序列的精度,且PCA比ICA效果更好。Aiming at the problems of outliers,missing point interpolation and common mode error of GNSS coordinate time series,this paper used an empirical mode decomposition(EMD)and 3σcombination algorithm to process outliers and obtain clean residual time series.Then,the fillmissing function of Matlab software was used for interpolation to obtain continuous time series.Finally,the independent component analysis(ICA)and principal component analysis(PCA)methods were used to extract the common mode error of the residual sequence,and the influence of common mode error was analyzed and studied.The results showed that compared with LS-3σ,the detection rate of EMD-3σmethod was increased by 0.2%,indicating that the detection effect of the new algorithm was better.The RMS values of PCA and ICA methods were reduced by about 22.01%and 10.96%respectively.indicating that PCA and ICA could effectively extract the common mode error of residual time series and improve the accuracy of coordinate time series,and PCA was better than ICA.

关 键 词：坐标时间序列 EMD-3σ 共模误差 独立成分分析 主成分分析 

分 类 号：P237[天文地球—摄影测量与遥感]