Improving the Classification of LANDSAT Data using Standardized Principal Components Analysis

The standardized principal component analysis (SPCA) is the method that uses the correlation matrix instead of the covariance matrix, and then eigenvalues and eigenvectors from SPCA apply to image processing procedures. When each principal components column is compared between the eigenvector matric...

Full description

Saved in:
Bibliographic Details
Published inKSCE journal of civil engineering Vol. 7; no. 4; pp. 469 - 474
Main Authors Chang, Hoon, Yoon, Wan Seok
Format Journal Article
LanguageEnglish
Published Seoul 대한토목학회 01.07.2003
Springer Nature B.V
Subjects
Agricultural land
Analysis
Classification
Components
Correlation analysis
Covariance matrix
Eigenvalues
Eigenvectors
Image processing
Land cover
Land use
Landsat
Principal components analysis
Remote sensing
Spatial discrimination
Spatial resolution
Studies
Urban areas
Vegetation
Online AccessGet full text

Cover

More Information
Summary:The standardized principal component analysis (SPCA) is the method that uses the correlation matrix instead of the covariance matrix, and then eigenvalues and eigenvectors from SPCA apply to image processing procedures. When each principal components column is compared between the eigenvector matrices from two different time images, the sign of principal components indicates the possibility in land-use/land-cover changes. The Landsat ETM+2000 was obtained in Suwon with 30 meter ground spatial resolution. The principal component 2 explains the difference between urban and vegetation areas through all bands. Two bands, band 2 and band 7, show the change in sign, meaning that urban areas and vegetation areas might distinctively show the characteristics if band 2 and band 7 are used in classifications. The classification is applied to composite images, which are the PC2, and 2, and band 7 composite image and the band 1, band 3, and band 5 composite image from the signature separability analysis. The principal component analysis is a useful statistical measurement for selecting a band combination including the principal component images. The eigenvector matrix can provide the band determination from the sign changes. The SPCA also gives critical information on determining the appropriate principal component, and the classification results from the accuracy assessment matrix described the large improvement on agricultural lands and bare lands.
ISSN:1226-7988
1976-3808
DOI:10.1007/BF02895842