Abstract: To extract the linear structures from of large data sets, the mathematical model of extracting the linear structures based on weighted total least absolute deviations criterion is proposed, and a new robust linear structures extraction method—BIRCH-WTALD-LGA is presented. It is shows that the method can effectively extract linear structures from large data set with contaminated data by means of computer simulation and application in earthquake distribution data clustering.

Key words: BIRCH algorithm, weighted total least absolute deviations, linear structures

摘要： 针对大规模数据集的线性结构提取问题，建立加权全最小一乘准则（WTALD）下的线性结构提取模型，给出了BIRCH-WTALD-LGA稳健线性结构提取新方法，数值模拟和在地震分布数据聚类分析中的应用说明该方法可以有效地提取大规模污染数据集中的线性结构。

关键词: BIRCH算法, 加权全最小一乘, 线性结构