Feature extraction from hyperspectral data under the support of fractal signal algorithm.

Hyperspectral data have been widely and successfully used in lithological mapping and mineral exploration for decades. However, different lithologies always show quite similar spectral shape and it is hard to distinguish them by conventional algorithms. This paper gives a new method for lithology feature extraction from hyperspectral data. The fractal signal image of airborne hyperspectral data of compact airborne spectrographic imager is calculated and used for lithology feature extraction. The results indicate that fractal signal analysis enable enhancement of lithologic information in hyperspectral images and eventually allow lithologic mapping by fractal signal algorithms. The influencing factor of the fractal signal and characterized scale are also discussed, it indicates that the iteration step size have influence on both the fractal signal and characterized scale. Little research has been made on the fractal signal features of spectral curves, and their physical implications and quantification are yet to be further studied. [ABSTRACT FROM PUBLISHER]