COMPLEX APPROACH FOR TEXTURE AND MULTISPECTRAL SEGMENTATION OF HIGH RESOLUTION SATTELITE IMAGES

11th International Multidisciplinary Scientific GeoConference SGEM2011, www.sgem.org, SGEM2011 Conference Proceedings/ ISSN 1314-2704, June 20-25, 2011, Vol. 2, 321-328 pp

ABSTRACT

It is introduced a complex approach for texture and multispectral segmentation, based
on texture analyses. The data fusion method is used for combining the multispectral and
texture features of classes in the areas which is investigated. The texture features are
estimated by the calculation of variance and skewness statistics over the windows with
different sizes. The results are integrated with spectral channels of satellite images. The
images from KOMPSAT-2 satellite are used for testing purposes. The spectral data are
enhanced by "pan-sharpening" procedure that uses panchromatic channel. Therefore
final multispectral data has the terrain element size of 1m. The analyses are provided by
usage of EDRAS Imagine program package. Different combinations of spectral
channels and channels with texture properties are combined and compared. The results
of unsupervised classification are presented and analysed.

Keywords: Image classification, Feature extraction, Texture analyses, Data fusion, High Resolution Satellite images