期刊名称：ISPRS Annals of the Photogrammetry, Remote Sensing and Spatial Information Sciences
印刷版ISSN：2194-9042
电子版ISSN：2194-9050
出版年度：2006
卷号：XXXVI Part 7
出版社：Copernicus Publications
摘要：High resolution images are rich sources of spatial information and details as well as noises. Because of the spectral similarities between different objects, pixel-based classifications of these images do not provide very accurate results. Incorporation of spatial data can lead to significant improvements in the accuracy of classifications. Several methods have been proposed for extraction of this information and generation of useful features for use in the classification process. In this study, different techniques for quantification of texture data including statistical, geostatistical, Fourier-based methods and gray level co-occurrence matrix have been investigated. New features have been generated by using spectral bands and the first PC resulting from the PCA transformation. The results have shown that the improvement of classification accuracy is significant and varies for different classes and features
