Abstract:

Feature spectral characteristics are the base of hyperspectral remote sensing technology. Based on rock spectral characteristics, for the purpose of classifying lithology by using Thermal Infrared Airborne Hyperspectral Imager (TASI) data, an improved lithological classification algorithm-spectral divergence energy-level matching (SDEM)-is presented in this paper. SDEM can identify tiny differences between any two different spectra. Also, this method takes both spectral band intensity and spectral waveform into account, and can effectively reduce the impact of image noises. Compared with the traditional lithological classification method-high spectral angle mapping (SAM), the improved algorithm can distinguish those similar but different spectra more precisely, and can identify those easily confused lithology. This method is also good at distinguishing the lithology known as “different features with similar spectra”. Using the TASI data of Liuyuan region in Gansu Province, we compared the lithological classification results of SDEM and SAM methods, and found that the SDEM method can identify the lithology that SAM can’t identify or wrongly identified. Based on our field validation work, the classification result by SDEM is more accordant with the actual distribution of rock, and is also more detailed.

Key words: thermal infrared hyperspectra, spectral characteristic, spectral divergence energy-level matching (SDEM), Liuyuan region in Gansu Province, lithological classification