Endmember detection in remote sensing is one of the significant factors to enhance the accuracy of spectral unmixing procedures. Because of highly mixed pixels in the case of mineral detection studies, the ordinary methods based on pure pixel occurrence assumption will not y of Hyperion data using the ORASIS method in Agh-Dagh region of Ardebil, Iran and a comparison sield accurate results in all situations. This research aimed to detect the endmembers on a scenetudy to Pixel Purity Index approach. ORASIS is a collection of a number of algorithms that work together to produce a set of endmembers that are not necessarily from inside the data set, and therefore, the pixel purity assumption is not vital for accurate results. Applying the consecutive algorithms of the ORASIS method on the Hyperion data of the study area has been generated 23 virtual endmembers. Spectral plots of these unknown endmebers were then compared to the USGS spectral library using SAM and SFF techniques; which resulted in recognition of ten minerals including Alunite, Kaolinite, Hornblende, Arsenopyrite, Vermiculite, Gypsum, Lepidolite, Chert, Limonite and Malachite, while the PPI method represented only 5 mineralogical compositions including Kaolinite/Smectite, Clinoptilolite, Celsian, Muscovite and Montmorillonite. Abundances maps of the detected minerals were generated by Linear Spectral Unmixing (LSU) technique for both ORASIS and PPI methods. The results of ORASIS showed much better and reasonable coincidence with the geological settings according to the available exploration reports and direct field observations in the study area.