Using unmixing methods to classify lithological and alteration units based on hyperspectral images

Nowadays providing geologic maps using satellite images has been developed. By using hyperspectral images more accurate studies in a wide range have been done. Since it is so expensive to provide these maps with field measurements therefore it is better to use new methods. This study provides a lithological and alteration mapping units with dominant minerals based on hyperspectral images of EO1-Hyperion satellite. To do so, two different zones were investigated: the Cuprite-Nevada and Mozahem volcano in Iran which have suitable conditions for our study. Five methods with different structures have been used: SAM, ACE, CEM, OSP, and LSU to evaluate their ability of geological unit separation. The results show that the differences and separability level in spectral signatures of training data are main factors in affecting the results in covariance base methods but it is low in the linear methods. This study revealed the accuracy of 86.45% for LSU in mineral mapping of Cuprite area and 69.54% for ACE in alteration mapping for Mozahem volcano which displays more efficiency than the other methods.