1999 年 19 巻 5 号 p. 407-422
We have developed the new effective data processing manner for ASTER VNIR-SWIR-TIR data. The disturbing factor in classifying rock type and mineral species using satellite data.is the product reflecting a mixture of various materials, which contains abundant noises. The manner estimates substances in each pixel by spectral pattern recognition with fuzzy inference, independently of channel number and end-member number. In addition, the results can be compared with other areas.
Then, applying the manner in the Cuprite Hill area, Nevada, U.S.A., mineral distribution maps of the specific minerals were prepared and the applicability of method was assessed correlating with the existing geological data. The image data used in this study were the adjusted ASTER simulation image prepared by air-borne AVIRIS, TMS and TIMS. The band to band registration was also adjusted.
In the case of Cuprite Hill area, five end-members were selected as follows according to the geological information. E/M1 (Hematite), E/M2 (Alunite), E/M3 (Kaolinite), E/M4 (Quartz), E/M5 (Buddingtonite). Results were compared with themineral assemblage of the alteration zone. Distributions of end-members showed a good correspondence with alteration zoning. A quantitative estimation was conducted using analysis results of ground survey that had been carried out by JGI (1990), MMAJ (1989, 1990), Abrams and Hook (1995) and Shipman and Adams (1987). Average coincidence between analysis results of ground survey and processed results was 73%.