Application of FastICA in mineral information extraction from hyperspectral remote sensing data

Abstract

Due to the lack of effective traditional hyperspectral image information extraction methods,this paper puts forward the hyperspectral remote sensing mineral information extraction technology based on the FastICA method. This technology first uses the virtual dimension( VD) method to determine the optimal number of features of hyperspectral remote sensing image data,and then employs the FastICA method to conduct dimensionality reduction and mixed pixel decomposition. Aimed at extracting mineral information,the technology uses the simulation plus noise hyperspectral remote sensing data as the experimental data,and the HyMap hyperspectral image is used as the end- member extraction accuracy evaluation data. The experimental results show that,after FastICA feature extraction,the precision of the hyperspectral simulation image remains higher than 90% in comparison with the classification of the spectral angle mapping( SAM),and the error of HyMap data end- member extraction control to 10- 3. All this demonstrates the feasibility and effectiveness of this technology in dimensionality reduction and mixed pixel decomposition of hyperspectral data.