K均值聚类进行多个粗差定位  

K-means Clustering Multiple Gross Errors

作　　者：宗琴 姜树辉 秦万英 

机构地区：[1]重庆建筑工程职业学院,重庆40072

出　　处：《北京测绘》2018年第1期37-39,共3页Beijing Surveying and Mapping

摘　　要：本文针对多个粗差探测与定位问题,应用智能信息处理技术中的K均值聚类算法,把对观测值的粗差定位转化为对该组观测值进行分类。依据粗差的出现约占观测总数的1%-10%,确定算法的一层终止条件为所有观测值分配为两类且其中对象较少的一类为粗差类;再引入方差比作为算法的二层终止条件,粗差类和非粗差类的方差比达到或超过试验倍数后,则接受初始聚类结果并最终结束算法。算法对早期的聚类分析方法定位粗差进行了改进和完善,其双重终止条件的应用,进一步提高了粗差探测的可靠性。In this paper,a plurality of gross error detection and localization issues,application of intelligent informa-tion proces狊 ng technology in the K-means clustering algorithm,the difference between the observed value of crude ：^n- to the set of positioning observations are classified. Based on gross errors occur abotion, determination algorithm layer termination conditons for all types of observations and ^n which signed to a class of less crude poor class; re--ntroduced as a variance ratio algorithm crude and non-crude poor class variance ratio reached or exceeded the test after multiple, accept ly ended the initial cluster algorithm. Cluster analysis algorithm for early positioning perfected,its dual application termination conditions,to further improve the gross error detection reliability.

关 键 词：K均值 多个粗差 定位 

分 类 号：P208[天文地球—地图制图学与地理信息工程]