Texture based classification of high resolution satellite data using gray level co occurrence matrix and semivariogram techniques

Abstract

Classification of land use land cover is the basic building block for spatial planning and management In recent years the advancement in satellite surveying has provided cost effectiveness and aerial coverage of collecting information It has improved the land use land cover analysis in identifying clear boundaries of regions On the other hand urban analysis including the identification and classification of urban features has become more complex due to detailed information No doubt that high resolution satellite data can provide better details due to improved spatial and spectral resolutions But it has made the features split in several pixels and thereby causing difficulty in feature delineation The spectral based methods that are commonly used are likely to fail due to the fact that each land use land cover type or feature is presented in several spatial adjacent pixels of different spectral values within class variations The features are much larger than the pixel resolution spatial and there is a need for additional information like shape size association and tonal textural relations as a group of the features