核Direct LDA子空间高光谱影像地物分类  

Kernel Direct LDA Subspace Hyperspectral Image Terrain Classification

作　　者：刘敬[1] 

机构地区：[1]西安邮电学院电子工程学院,西安710121

出　　处：《计算机科学》2012年第6期274-277,共4页Computer Science

基　　金：国家自然科学基金(61003199);中央高校基本科研业务费专项资金(K50510020015);陕西省教育厅自然科学专项基金(2010JK821);西安邮电学院博士启动基金(000-1271)资助

摘　　要：为降低高光谱影像的数据维数,提高地物分类识别效率,提出了一种地物分类方法——核直接线性判别分析(Kernel Direct Linear Discriminant Analysis,KDLDA)子空间法;并推导出类先验概率的一般形式下KDLDA的解。KDLDA子空间法先采用KDLDA提取遥感影像的非线性可分特征,然后在KDLDA子空间采用最小距离分类器进行分类识别。机载可见光/红外成像光谱仪(Airborne Visible/Infrared Imaging Spectrometer,AVIRIS)的高光谱影像识别结果表明,相比原空间法、LDA子空间法、直接线性判别分析(Direct Linear Discriminant Analysis,DLDA)子空间法、核线性判别分析(Kernel Linear Discriminant Analysis,KLDA)子空间法,KDLDA子空间法可显著提高识别效率。In order to reduce the data dimensionality of hyperspectral image and improve recognition efficiency, a new terrain classification method, i. e. , KDLDA subspace method, was presented. Firstly, kernel direct linear discriminant analysis （KDLDA） was used to extract nonlinear discriminant features, and then shortest distance classifier was used to perform terrain classification in the KDLDA feature subspace. The solution of KDLDA under the ordinary form of class prior possibility was also deduced. Recognition results based on airborne visible/infrared imaging spectrometer （AVIRIS） hyperspectral image show that, comparing with original space method, LDA subspace method, direct linear discriminant analysis （DLDA） subspace method, and kernel linear discriminant analysis （KLDA） subspace method, the presented KDLDA subspace method can remarkably improve recognition efficiency.

关 键 词：地物分类 非线性可分性特征 核直接线性判别分析 高光谱影像 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术] TP751.1[自动化与计算机技术—计算机科学与技术]