The analysis of hyperspectral data using linear unmixing requires the determination of endmembers consisting in hyperspectral data.All endmember determination methods at present need complex calculation and can not be proved universal for all the hyperspectral data.According to the effective analysis of the information presented in hyperspectral data,a new method based on feature space analysis and spectral correlation mapping(FSASCM) was proposed.The method mentioned is simple and universal for almost all the hyperspectral data.