An improved EM algorithm for remote sensing classification

作     者：YANG HongLei PENG JunHuan XIA BaiRu ZHANG DingXuan 

作者机构：School of Land Science and TechnologyChina University of Geosciences School of Engineering and TechnologyChina University of Geosciences 

出 版 物：《Chinese Science Bulletin》 (中国科学通报)

年 卷 期：2013年第58卷第9期

页      面：1060-1071页

核心收录：

学科分类：083002[工学-环境工程] 0830[工学-环境科学与工程（可授工学、理学、农学学位）] 081802[工学-地球探测与信息技术] 08[工学] 0818[工学-地质资源与地质工程] 081602[工学-摄影测量与遥感] 0816[工学-测绘科学与技术] 

基　　金：supported by the National High-tech R&D Program of China(2007AA12Z226 and SS2012AA120804) the National Natural Science Foundation of China(40674015 and 41074009) the Doctoral Fund of Ministry of Education of China(20100022110008) the Fundamental Research Funds for the Central Universities(2-9-2011-227) the Open Research Fund of Key Laboratory of Digital Earth Science,Center for Earth Observation and Digital Earth,Chinese Academy of Sciences (2010LDE002) 

主　　题：EM算法 遥感分类 k-means算法 主成分变换 协方差矩阵 随机选择 多光谱图像 期望最大化 

摘      要：The use of a general EM(expectation-maximization) algorithm in multi-spectral image classification is known to cause two problems:singularity of the variance-covariance matrix and sensitivity of randomly selected initial *** former causes computation failure;the latter produces unstable classification *** paper proposes a modified approach to resolve these ***,a modification is proposed to determine reliable parameters for the EM algorithm based on a k-means algorithm with initial centers obtained from the density function of the first principal component,which avoids the selection of initial centers at random.A second modification uses the principal component transformation of the image to obtain a set of uncorrelated *** number of principal components as the input of the EM algorithm is determined by the principal contribution *** this way,the modification can not only remove singularity but also weaken *** results obtained from two sets of remote sensing images acquired by two different sensors confirm the validity of the proposed approach.