The invention relates to a method for mapping mineral by taking use of the statistics characteristics of data without prior information for a high spectrum data, which comprises the following procedures: reading in the high spectrum data, centralizing the data, sphericalizing the data, setting up an analysis model of independent components, rating the independent components, extracting end members of mineral, min. noise separating and transforming, mixing, modulating, matching and filtering, mapping mineral. The method allows to extract end members of eroded mineral based on the analysis model of independent component by taking use the hi-order statistics feature of data themselves without unknowing any prior information, and to map eroded mineral by a way based on mixing, modulating, matching and filtering. The method avoids in the mineral identification the influence of uncertainness of spectrum feature of mineral caused by spectrum mixing, illumination, and mineralizing condition, combines the restricting condition of end-member content to be positive and 1 in the linear mixing decomposition theory, has effectively improves the detection limit for mineral and mapping mineral accuracy.