Search

Technology approach for supervised classification of multi-band satellite images is used for analyses of soil degradation processes. It is based on the thematic soil mapping, based on automatic image interpretation and segmentation, which is utilizing the functionality of Remote sensing software system. Spatial resolutions of multiband images are increased by usage of "pan-sharpening" method and enhanced images are georeferenced and georectified. Training areas for supervised classification are created from terrain samples with examined soil types. Spectral signatures of classes generated in unsupervised and supervised modes are combined to obtain reliable classification by maximum likelihood method. An improvement of classification is achieved trough applying of estimation based on the chi-square method. The thematic raster images and vector layers from satellite images with medium and high resolution from Landsat 7 TM+, SPOT-5 and Kompsat-2 are compared and analysed. The distribution of land cover classes is evaluated and soil types discovered by image interpretation are compared with soil type properties. The obtained results prove the applicability of the suggested soil mapping technology. The produced digital soil maps are suitable for incorporation in GIS for the purposes of the sustainable land management. [ABSTRACT FROM AUTHOR]