CEEMDAN和3σ组合的GNSS坐标时间序列粗差探测  

GNSS coordinate time series gross error detection with CEEMDAN and 3σ combination

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4期60-67,191,共9页Science of Surveying and Mapping

基　　金：国家自然科学基金项目(41905027)。

摘　　要：针对全球导航卫星系统(GNSS)坐标时间序列的异常值探测问题,该文提出了一种基于自适应完备集合经验模态分解(CEEMDAN)与3σ组合的异常值探测算法,先利用CEEMDAN方法获得提取信号后的残差时间序列,再采用3σ方法对残差时间序列进行异常值探测。利用模拟数据和实测数据进行实验,将新算法与EMD-3σ、SSA-3σ、CEEMDAN-IQR、EMD-IQR和SSA-IQR进行对比分析,模拟数据结果显示CEEMDAN-3σ算法对坐标时间序列的异常值探测具有很好的探测效果。实测数据结果表明,CEEMDAN-3σ算法探测率为0.8%,比其他5种算法至少提升了0.2%,说明新算法更能有效地探测坐标时间序列中的粗差。Aiming at the problem of outlier detection of global navigation satellite systems(GNSS) coordinate time series,this paper proposed an outlier detection algorithm based on complete ensemble empirical mode decomposition with adaptive noise(CEEMDAN) and 3σ combination was proposed.Firstly,the residual time series after signal extraction was obtained by CEEMDAN method,and then the 3σ method was used to detect the outliers of the residual time series.The new algorithm was compared with EMD-3σ,SSA-3σ,CEEMDAN-IQR,EMD-IQR and SSA-IQR by using simulated data and measured data.The simulation data results showed that the CEEMDAN-3σ algorithm had a good detection effect on the outlier detect ion of coordinate time series.The measured data results showed that the detection rate of CEEMDAN-3σ algorithm was 0.8%,which was at least 0.2 percentage points higher than the other five algorithms,indicating that the new algorithm could detect gross errors in coordinate time series more effectively.

关 键 词：GNSS坐标时间序列 自适应完备集合经验模态分解 3σ方法 粗差探测 

分 类 号：P228[天文地球—大地测量学与测量工程]