一种优化均值漂移模型的测站筛选方法  

A Station Selection Method for Optimizing Mean Shift Model

作　　者：范成成 张俊[2] FAN Chengcheng;ZHANG Jun(School of Mining Engineering,Guizhou University of Engineering Science,Bijie 551700,China;Mining College,Guizhou University,Guiyang 550025,China)

机构地区：[1]贵州工程应用技术学院矿业工程学院,贵州毕节551700 [2]贵州大学矿业学院,贵州贵阳550025

出　　处：《贵州大学学报（自然科学版）》2024年第5期26-31,共6页Journal of Guizhou University:Natural Sciences

基　　金：贵州省青年科技人才成长项目(黔教合KY字[2022]132号,黔教合KY字[2022]135号);贵州工程应用技术学院一流专业建设项目(ZY202104)。

摘　　要：针对均值漂移模型在异常测站剔除过程中出现的“回漂”现象导致粗差未能尽数剔除、使其拟合结果有偏于实际的问题,提出基于惩罚回归优化均值漂移模型对异常测站进行剔除的方法。利用优化后的模型方法对“中国内地构造环境监测网络”华南块体中东部观测的水平速度场数据进行拟合对比分析,并利用剔除后的有效测站进行形变特征反演。结果表明:利用惩罚回归优化的均值漂移模型能更有效地识别异常测站点,筛选后的测站数据能顾及块体整体运动的平滑性,同时比均值漂移模型剔除具有更高的拟合精度,利用有效数据反演的形变特征也符合该区域实际运动趋势。In view of the problem that the“backdrift”phenomenon of the mean drift model in the process of anomalous station exclusion leads to the failure to eliminate all the gross errors,and the fitting results are biased to reality,a method of eliminating abnormal stations based on penalty regression optimization mean drift model is proposed.The optimized model method was used to fit and compare the horizontal velocity field data observed in the central and eastern parts of the South China block by the“Chinese mainland tectonic environment monitoring network”,and the deformation characteristics were inverted by using the effective stations after elimination.The results show that the mean drift model optimized by penalty regression can identify abnormal stations more effectively,and the filtered station data can take into account the overall motion smoothness of the block and have a higher fitting accuracy than the mean drift model.The deformation characteristics inverted by using effective data also conform to the actual movement trend of the region.

关 键 词：地壳运动 异常测站 均值漂移模型 惩罚函数 应变特征 

分 类 号：P227[天文地球—大地测量学与测量工程] P313[天文地球—测绘科学与技术]