In image processing, it is commonly assumed that the model ruling spectral mixture in a given hyperspectral pixel is linear. However, in many real life cases, the different objects and materials determining the observed spectral signatures overlap in the same scene, resulting in nonlinear mixture. This is particularly evident in volcanoes-related imagery, where both airborne plumes of effluents and surface deposit of volcanic ejecta can be mixed in the same observation line of sight. To tackle this intrinsic complexity, in this paper, we perform a pilot test using Nonlinear Principal Component Analysis (NLPCA) as a nonlinear transformation, that projects a hyperspectral image onto a reduced-dimensionality feature space. The use of NLPCA is twofold: (1) it is used to reduce the dimensionality of the original spectral data and (2) it performs a linearization of the information, thus allowing the effective use of successive linear approaches for spectral unmixing. The proposed method has been tested on two different hyperspectral datasets, dealing with active volcanoes at the time of the observation. The dimensionality of the spectroscopic problem is reduced of up to 95% (ratio of the elements of compressed nonlinear vectors and initial spectral inputs), by the use of NLPCA. The selective use of an atmospheric correction pre-processing is applied, demonstrating how individual plume and volcanic surface deposit components can be discriminated, paving the way to future application of this method.
This is an open access article distributed under the Creative Commons Attribution License
which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited