28 mai 2015
Ce document est lié à :
info:eu-repo/semantics/altIdentifier/doi/10.6084/m9.figshare.7211627
http://creativecommons.org/publicdomain/zero/1.0/ , info:eu-repo/semantics/OpenAccess
Polina Lemenkova, « Innovations in the Geoscience Research: Classification of the Landsat TM Image Using ILWIS GIS for Geographic Studies », HAL-SHS : géographie, ID : 10.6084/m9.figshare.7211627
Environmental mapping is a necessary tool for the geoscience research in the university classes of geography, GIS cartography and mapping. However, GIS methods of processing remote sensing data are often being discussed, and the optimal approaches are disputed. This work reports innovative approach of the processing Landsat TM satellite image in ILWIS GIS software using unsupervised and supervised classification methods. The methods of ILWIS GIS are compared and the results described in this paper. The methodology is based on the use of supervised classification of the satellite image, an innovative method in geographic research. Supervised classification of the raster imagery aimes at the recognizing of the class membership for each pixel during image analysis. The supervised classification of the multi-spectral imagery has been performed using 'Classify' operator in ILWIS applied to Landsat TM 1988. The classification process included following steps. First, the models of the classes were defined by creating a "sample sat" in ILWIS GIS. Namely, the training pixels with similar spectral values were defined and selected as representations for various classes. These pixels have contrasting colors, visually visible and distinguishable on the image, which serve as training areas for diverse classes. The sample pixels were defined in the Sample Set Editor in ILWIS, which was initially created in the main menu. A created Sample Set has a reference to the set of Landsat bands (1-7), which are needed to create sample statistics. After assigning pixel sets, a raster polygon map was automatically created with .mpr file extension. It contained sample pixels, location and legend, i.e. the names of the classes allocated to pixels. Easy interpretation of the image strongly depends on the optimal color composite map. The classification has been completed in interactive way, using several attempts of creating training samples, selecting various sample sets and respective pixels selected in Sample Set Editor using domain for classification ―Landclasses. The classification was repeated the until the final results were achieved. Shadows of green colors represent grass, shrub vegetation coverage and forest canopy. The results demonstrate successful application of the innovative approach of technical satellite image processing for the studies of the environment. The presented methods can be used by students of geography.