Improving lithological mapping by support vector machine classification using integrated visible-near infrared, shortwave infrared, and thermal infrared of ASTER data sets

This study aims to enhance lithological mapping by employing Support Vector Machine (SVM) classification on integrated visible-near infrared (VNIR)-shortwave infrared (SWIR), and thermal infrared (TIR) datasets from the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER). The study focuses on a part of Kerman Province, the Dehsard area, which was selected due to its diverse lithological units, including igneous, sedimentary, and metamorphic rocks, as represented in the 1: 100,000 geological map of Dehsard. SVM classification was applied on the three datasets using training areas derived from the geological map, along with principal component analysis (PCA) and minimum noise fraction (MNF). The results indicated that the SVM classification on the 14-band ASTER data yielded more accurate results compared to the 9 and 5 band datasets. This study underscores the effectiveness of integrating multiple spectral bands in improving the precision of lithological mapping, which is essential for mineral exploration and geological studies.