一种基于K-Means均值聚类算法的大地测量数据粗差探测方法  

A Outlier Detection Method of Geodetic Data Based on K-Means Clustering Algorithm

作　　者：杨玉 吴云龙 单维锋[1] 石江浩[3] Yang Yu;Wu Yunlong;Shan Weifeng;Shi Jianghao(Institute of Disaster Prevention,Hebei Key Laboratory of Earthquake Dynamics,Sanhe 065201,China;Key Laboratory of Earthquake Geodesy,Institute of Seismology of China Earthquake Administration,Wuhan 430071,China;Beihang University,Beijing 100191,China)

机构地区：[1]防灾科技学院,河北省地震动力学重点实验室,河北三河065201 [2]中国地震局地震研究所,中国地震局地震大地测量重点实验室,湖北武汉430071 [3]北京航空航天大学,北京100191

出　　处：《城市勘测》2021年第4期127-131,共5页Urban Geotechnical Investigation & Surveying

基　　金：国家自然科学基金项目(41931074,41974096)。

摘　　要：粗差探测是大地测量数据处理过程中的重要环节。通过实验模拟生成一类典型大地测量观测数据的粗差,将生成含有白噪声和粗差的观测数据作为数据集,通过K-Means均值聚类算法对数据进行聚类,依据粗差数量约占观测值总数的1%,实验中选取参数5作为该算法的聚类中心个数,远离聚类中心的数据可被认为是粗差,因此粗差的探测成为研究重点。研究结果显示,该粗差探测方法的准确率为88.17%,对海量大地测量观测数据的预处理具有应用价值和科学意义。Outlier detection is an important part of geodetic data processing.The outlier detection of a class of typical geodetic observation data are generated by simulation in the experiment,and the observation data with white noise and ouliers are generated as data sets.The data are clustered by K-Means clustering algorithm.According to the number of outliers accounting for about 1%of the total observations values,parameter 5 is selected as the number of clustering centers of the algorithm in the experiment,and the data far away from the clustering center can be considered as outliers,so the detection of outlier has become the focus of research.The results show that the accuracy of the outlier detection method is 88.17%,which has application value and scientific significance for the preprocessing of massive geodetic observation data.

关 键 词：卫星重力 粗差探测 K-Means均值聚类算法 聚类中心个数 

分 类 号：P207[天文地球—测绘科学与技术] P223