PCA and SVM as geo-computational methods for geological mapping in the southern of Tunisia, using ASTER remote sensing data set

The purpose of this study was to examine the efficiency of Advanced Space Borne Thermal Emission and Reflection Radiometer (ASTER) data in the discrimination of geological formations and the generation of geological map in the northern margin of the Tunisian desert. The nine ASTER bands covering the visible (VIS), near-infrared (NIR) and short-wave infrared (SWIR) spectral regions (wavelength range of 400–2500 nm) have been treated and analyzed. As a first step of data processing, crosstalk correction, resampling, orthorectification, atmospheric correction, and radiometric normalization have been applied to the ASTER radiance data. Then, to decrease the redundancy information in highly correlated bands, the principal component analysis (PCA) has been applied on the nine ASTER bands. The results of PCA allow the validation and the rectification of the lithological boundaries already published on the geologic map, and gives a new information for identifying new lithological units corresponding to superficial formations previously undiscovered. The application of a supervised classification on the principal components image using a support vector machine (SVM) algorithm shows good correlation with the reference geologic map. The overall classification accuracy is 73 % and the kappa coefficient equals to 0.71. The processing of ASTER remote sensing data set by PCA and SVM can be employed as an effective tool for geological mapping in arid regions.