In recent years hyperspectral sensors have been developed with the ability to image hundreds of bands and are used in remote sensing. Hyperspectral data with the ability of spectral differentiation has the ability of recognition of arid land plants with low percentage of vegetation. Hyperion is a hyperspectral sensor which is a powerful tool for the identification of land coverage and has 240 bands on 400 nm to 2500 nm wavelength with the band width of 10 nm. Hyperion image is an unprocessed picture which is influenced by atmospheric effects and has noisy bands. The aim of this study is iection the potential of hyperspectral data level 1, Hyperion sensor in identifying endmembers of an entire Hyperion hyperspectral data. For this purpose subpixel SMA method is used for spectral differentiation of hyperspectral data and finally identifying endmembers. In this research, first bad pixels were omitted from hyperspectral image of Hyperion. Noisy and uncalibrated bands including bands (1-7, 58-76, 121-127, 129, 132, 165- 181, 183, 185- 187, 191, 205, 207- 210, 218- 223 and 225- 242) were omitted from the image and atmospheric correction of the image was done using FLAASH model in ENVI. The results showed specters of the image were improved after the atmospheric correction and absorbed bands of water were identified. After the atmospheric correction to extract the endmembers; MNF, PPI, multispectral space and SMA transformation methods were used. Spectral mixture analysis (SMA) is the partial coverage of different type of land coverage (for example green plant vegetation, litter, soil and shadow) that is simulated in a picture pixel. Finally, best bands for vegetation index were extracted using band ratio of hyperspectral Hyperion image and NDVI. The results of implementation of the mentioned methods on hyperspectral Hyperion image showed most data exists in the primary MNF images and in least data exists in the final images. PPI implementation with the threshold of 50000 led to separation of pure pixels of the image. In multi-dimensional space pixels were dir=ltr Key Words: Hyperxpectral, Endmember, SMA, MNF, PPI
