摘要
针对时间序列数据中存在的粗差问题,该文首先介绍了奇异谱分析法(SSA)和未确知滤波法(UF)的工作原理,考虑到SSA方法在识别趋势项和周期性信号方面及UF算法在区分粗差和异常值上的优势,在SSA准确提取信号的基础上结合UF算法提出了一种新的SSA-UF粗差探测法:首先利用SSA提取观测值序列的信号并获取残余分量;然后通过UF算法对残余分量进行分析确定粗差点的位置;最后确定粗差点并剔除。通过单因素和多因素主导变形的观测值序列两个实例的验证分析,结果表明,该文中提出的SSA-UF粗差探测法与SSA数据统计方法相比在监测数据处理中的粗差探测效果明显,可靠性更高,为后续监测数据分析处理奠定了较好的基础。
Aiming at the problem of gross errors detection of time series,firstly,the working principles of singular spectrum analysis(SSA)and unascertained filtering(UF)were introduced in this paper,with the consideration that SSA performs better in identifying trend items and periodic signals and UF algorithm has great ability in distinguishing the gross errors and outliers,thus SSA-UF gross error detection method was proposed.SSA was used to extract the signals of the observation series and obtain the residual components.Then UF algorithm was used to analyze the residual components to determine the location of the gross errors,and finally the gross errors were eliminated.Through experimental analysis of two real examples of single-factor and multi-factor dominant deformation,the results showed that the proposed SSA-UF gross error detection method was more effective and reliable than the SSA data statistics method in monitoring data processing,which lays a good foundation for subsequent monitoring data analysis and processing.
作者
张东华
李志娟
刘全明
黄磊
ZHANG Donghua;LI Zhijuan;LIU Quanming;HUANG Lei(Inner Mongolia Agricultural University,Hohhot 010018,China;Aeronautical Remote Sensing Surveying and Mapping Institute,Hohhot 010010,China)
出处
《测绘科学》
CSCD
北大核心
2020年第8期14-18,共5页
Science of Surveying and Mapping
基金
内蒙古自治区高等学校科学研究项目(NJZY20049,NJZY18064)
国家自然科学基金项目(51969023,51569018)
内蒙古自治区自然科学基金资助项目(2018MS05005)
内蒙古农业大学高层次人才科研启动项目(NDYB2018-60)。
关键词
奇异谱分析
变形监测时间序列
粗差探测
未确知滤波
singular spectrum analysis(SSA)
time series of deformation observation
gross error detection
unascertained filtering
