利用P-EDMP与光谱进行高光谱遥感影像分类  

Classifying Hyperspectral Image Based on P-EDMP and Spectrum

作　　者：李祖传[1,3] 马建文[1] 张睿[2,3] 李利伟[1] 

机构地区：[1]中国科学院对地观测与数字地球科学中心,北京市100190 [2]中国科学院遥感应用研究所,北京市100101 [3]中国科学院研究生院,北京市100049

出　　处：《武汉大学学报（信息科学版）》2010年第12期1449-1452,共4页Geomatics and Information Science of Wuhan University

基　　金：国家863计划资助项目(2007AA12Z157);国家自然科学基金资助项目(40901234);中国科学院知识创新工程青年人才领域前沿专项资助项目(O8S01100CX)

摘　　要：提出了一种改进的扩展形态剖面导数(P-EDMP)以及一种融合P-EDMP与光谱的分类方法。采用AVIRIS高光谱遥感数据,与融合光谱和扩展形态剖面(EMP)的方法进行对比实验,结果表明,在描述高光谱遥感影像的形态特征上,P-EDMP与EDMP相当,但是P-EDMP的时间复杂度要小;在分类精度上,所提方法要优于融合光谱与EMP的方法。Among classification of hyperspectral imagery,there are some advantages by combining spectral and morphological information,which is a hot research hop.Extended differential of morphological profiles（EDMP） is a kind of feature describing morphological information of multi-channel imagery.The methods adopts a vector ordering method based on distance to extend grey-level mathematical morphology to multivariate morphology,then constructs the corresponding morphological feature.EDMP has already obtained satisfied results in classification of hyperspectral imagery.However,EDMP is time-consuming and ignores spectral information.To overcome these problems,an improved extended differential of morphological profiles（P-EDMP） and a classification method based on fusing P-EDMP and spectrum are proposed.Using AVIRIS hyperspectral imagery as input,comparative experiments between the proposed method and that of fusing spectrum and extended morphological profiles（EMP） are carried out.Experimental results showed that P-EDMP is competitive with EDMP in describing morphological information of hyperspectral imagery,but is less time-consuming.The proposed method is superior to that of fusing spectrum and EMP in terms of classification accuracies.

关 键 词：形态剖面导数 遥感影像 融合 分类 

分 类 号：P237.4[天文地球—摄影测量与遥感] TP753[天文地球—测绘科学与技术]