基于谱系聚类的粗糙集离散化方法在胶合板缺陷检测中的应用  

Discretization Method in Rough Sets Based on Hiearachical Clustering and Its Application for Defect Inspection of Wood Veneer

作　　者：李孟歆[1,2] 吴成东[2] 韩中华[2] 

机构地区：[1]沈阳建筑大学,辽宁沈阳110168 [2]东北大学,辽宁沈阳110004

出　　处：《计算机与现代化》2008年第11期54-57,共4页Computer and Modernization

基　　金：科技部国际合作重点资助项目(2003DF020009)

摘　　要：采用谱系聚类方法对大量数据进行离散化处理,通过散点图和树状图自动寻求最优分类数目。胶合板缺陷检测样本数据离散化实验结果表明,该方法根据数据内部结构特点,进行了可变的类别划分,与可k均值聚类和模糊聚类相比,该方法使约简的条件属性进一步减少,有利于数据的后期处理,并且进一步减少了系统的运行时间。Hierarchical clustering method is used to deal with a large amount of data for attribute discretization of rough sets. The optimal nunmbers of clusters are analyzed and decided from the static plots and the dendrogram. The experimental results using defect data of wood veneers show that the number of clusters is variable based on the internal features of data. Comparative research is carried out using the K-means clustering fuzzy C-clustering, which shows the method can be used to fuaher reduce the number of attributes, and also be good for post processing. It can shorten the running time of the system as well.

关 键 词：聚类分析 谱系聚类 离散化 胶合板 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]