Investigation of Chaotic-Type Features in Hyperspectral Satellite Data


Osman Kocal and Mufit Cetin, Yalova University, Turkey


Hyperspectral images provide detailed spectral information with more than several hundred channels. On the other hand, the high dimensionality in hyperspectral images also causes to classification problems due to the huge ratio between the number of training samples and the features. In this paper, Lyapunov Exponents (LEs) are used to determine chaotic-type structure of EO- 1 Hyperion hyperspectral image, a mixed forest site in Turkey. Experimental results demonstrate that EO-1 Hyperion image has a chaotic structure by checking distribution of Lyapunov Exponents (LEs) and they can be used as discriminative features to improve classification accuracy for hyperspectral images.


Hyperspectral Image, EO-1 Hyperion, Phase Space, Chaos, Lyapunov Exponents

Full Text  Volume 5, Number 10