ANALYSIS OF THE EFFICIENCY OF CLASSIFICATION OF HYPERSPECTRAL SATELLITE IMAGES OF NATURAL AND MAN-MADE AREAS
S. M. Borzov1, A. O. Potaturkin1, O. I. Potaturkin1,2, A. M. Fedotov2,3
1Institute of Automation and Electrometry, Siberian Branch, Russian Academy of Sciences, pr. Akademika Koptyuga 1, Novosibirsk, 630090 Russia
2Novosibirsk State University, ul. Pirogova 2, Novosibirsk, 630090 Russia
3Institute of Computational Technologies, Siberian Branch, Russian Academy of Sciences, pr. Akademika Lavrent’eva 6, Novosibirsk, 630090 Russia
Keywords: дистанционное зондирование Земли, гиперспектральные изображения, классификация типов поверхностей, спектр отражения, remote sensing, hyperspectral images, classification of surface types, reflection spectrum
Abstract
The efficiency of a number of the classical methods of supervised classification of hyperspectral data is estimated by an example of discriminating the types of the underlying surface in natural and man-made areas. The minimum distance, support vector machine, Mahalanobis, and maximum likelihood methods are considered. Particular attention is paid to studying the dependence of the data classification accuracy on the number of spectral features and the way of choosing them in the above-mentioned methods. Experimental results obtained by processing real hyperspectral images of landscapes of various types are reported.
|
