期刊名称：International Journal of Engineering and Computer Science
印刷版ISSN：2319-7242
出版年度：2014
卷号：3
期号：7
页码：7221-7228
出版社：IJECS
摘要：Principal Component Analysis (PCA) technique is useful in reducing dimensionality of a data set in order to obtain a simpledataset where characteristics of the original dataset that contributes most to its variance are retained. This method is to transform theoriginal data set into a new dataset, which may better capture the essential information. Remote sensing images from orbiting satellites aregaining ground in recent years in inventory, mapping and monitoring of earth resources. These images are acquired in different wavelengthsof the electromagnetic spectrum and therefore there exist correlation between the bands. The developed algorithm can not only reduce thedimensionality of remote sensing image but also extract helpful information for differentiating the target feature from other vegetation typesmore effectively. In this paper the usefulness and innovative of PCA in processing of multispectral remote sensing images have been tinted.It has been observed that PCA effectively summarize the dominant modes of spatial, spectral and temporal variation in data in terms oflinear combinations of image frames. It provides maximum visual separability of image features thus improving the quality of ground truthcollection and also turn to improving the image classification accuracy. Here, we propose a fast alternative to iterative PCA that makes itsuitable for remote sensing applications while ensuring its theoretical convergence illustrated in the challenging problem of urbanmonitoring.
关键词：PCA; KPCA; HPC; Covariance Matrix and Classification.
