The Effect of Spectral Unmixing of Hyperspectral Imagery for Mapping of Soil Properties

摘要

In this study, a method using the normalized reflectance was employed to remove the effects of topography and surface roughness to observed spectrum and the normalized reflectance was used to perform subsequent correction and analysis. Then, we examined the removal of the vegetation effects (such as green vegetation and dry vegetation), because the ground surface shows the mixture of soil and vegetation. Comparison between the normalized pure soil spectrum and the measured soil spectrum using an ASD FieldsSpec showed a generally good agreement. This relatively simple pre-processing allowed effective unmixing of both green and dry vegetation effects and extraction of the soil normalized-reflectance. Using the soil spectrum derived from spectral unmixing, we estimated the fundamental soil properties such as clay mineral content. The results showed higher accuracy for soil data obtained from spectral unmixing than reflectance data, which demonstrated the effectiveness of our correction method.