The objective of this work was to evaluate the potential of hyperspectral classification techniques in the Advanced Spaceborne Thermal Emission and Reflection Radiometer (ASTER) reflectance data (visible to short wave infrared region) and spatial resolution (15 and 30 m), to map volcanic rocks in Ametista do Sul Region, Rio Grande do Sul State, Brazil. This region is one of the most important amethyst mineralization of the World. The spectral behavior of these rocks is similar to shadows and soils when interpreted with traditional digital classification techniques and multispectral sensors, like TM-Landsat, CCD-CBERS, etc. As an alternative was applied hyperspectral image processing technique (Spectral Angle Mapper - SAM) to identify and discriminate basalt rocks occurrence in mixed pixels. Due to vegetation around and covering some outcrops and the pixel spatial resolution, it was not possible to extract a basalt endmember directly into the ASTER image, being used so an endmember from NASA spectral library. To compare SAM results with traditional classification techniques were applied the Maximum Likelihood (MaxVer) algorithm. The SAM technique produced better results than MaxVer, but the error persisted, even in a lesser proportion, in mixed pixels with “Shadows”, "Soils" and “Basalt” classes.