Application of Hyperion hyperspectral imagery on geological mapping in Central Mountain Range of Taiwan: preliminary results

摘要

The characteristics of geological material, such as the lithology and mineral composition, provide important information in geohazard assessment and mineral exploration. The lofty Central Mountain Range located in the mid east of Taiwan was formed by the slow collision of the Asian continental plate and the Philippine plate, resulting in a 340-km × 80-km area with average height of 2000-m where implementation of the traditional approach of geological survey is impractical. Up to now, very little is known about the characteristics of geological material in this area. The hyperspectral imagery obtained from the airborne platform had been applied in lithology mapping since 1980s. All ideal results were reported in the sparse-vegetation and well-exposed outcrop area. The space-borne Hyperion sensor onboarded the EO-1 satellite was successful deployed in November 2000. Its advanced spectral range and resolution motivates us to apply the hyperspectral imagery to the work of geological mapping in the Central Mountain Range of Taiwan. The challenge is that the coverage of vegetation is dense while the useful outcrop is relatively sparse in this area. To apply the hyperspectral imagery to the work of geological mapping in dense-vegetation and sparse-outcrop region, we investigate the feasibility by conducting a numerical experiment with simulated hyperspectral imagery. The spectra of different surface features are selected and mixed under various controlled conditions. A series of hyperspectral unmixing processing was taken and the results were compared to the simulated conditions. Even though the outcrop area is too small to constitute a pure pixel, results from this experiment indicate that the outcrop still can be identified, as long as the spectra of that particular outcrop is measured and provided as an endmember in the processing. We took a further step to process and analyze the Hyperion image taken on 5 June 2005 in this area. The commercial package, FLAASH 4.1, was employed to correct the atmospheric effect. The ground truths were collected during several field works to the Pao-Lai and He-Ping sites. Various samples of rocks were brought back to the lab to measure the spectral reflectance using a hyperspectral spectroradiometer (GER 2600). These spectra were supplied as endmembers in the processing of Hyperion image. The standard approaches of mixed-pixel analysis were followed and the results of classification were compared to the existing geological map. Results of this research encourage the application of Hyperion hyperspectral imagery on geological mapping in Central Mountain Range of Taiwan.