Abstract
The spatial pixel resolution of common multispectral and hyperspectral sensors is generally not sufficient to avoid that multiple elementary materials contribute to the observed spectrum of a single pixel. To alleviate this limitation, spectral unmixing is a by-pass procedure which consists in decomposing the observed spectra associated with these mixed pixels into a set of component spectra, or endmembers, and a set of corresponding proportions, or abundances, that represent the proportion of each endmember in these pixels. In this study, a spectral unmixing technique is proposed to handle the challenging scenario of non-linear mixtures. This algorithm relies on a dedicated implementation of multiple-kernel learning using self-organising map proposed as a solver for the non-linear unmixing problem. Based on a priori knowledge of the endmember spectra, it aims at estimating their relative abundances without specifying the non-linear model under consideration. It is compared to state-of-the-art algorithms using synthetic yet realistic and real hyperspectral images. Results obtained from experiments conducted on synthetic and real hyperspectral images assess the potential and the effectiveness of this unmixing strategy. Finally, the relevance and potential parallel implementation of the proposed method is demonstrated.
Highlights
With the recent and rapid development of hyperspectral imaging technology, hyperspectral images have been widely used in various scientific fields, such as environmental mapping, risk prevention, urban planning, pollution monitoring and mining exploration [1]
This paper suggests replacing the usual solver by the neural network, the Kohonen's self-organising maps (SOMs) [27]
To validate the proposed MK-SOM-based unmixing algorithm, its performance has been compared to those obtained by the linear unmixing methods fully constrained least squares (FCLS) [13] and SUNSAL [14] and the nonlinear unmixing methods PPNM [16], rLMM [18], K-Hype and its generalised counterpart (SK-Hype) [41]
Summary
With the recent and rapid development of hyperspectral imaging technology, hyperspectral images have been widely used in various scientific fields, such as environmental mapping, risk prevention, urban planning, pollution monitoring and mining exploration [1]. Such a framework has several advantages: (i) it can simultaneously use numerous features (computed from the different bands of the hyperspectral images) to enrich the data similarity representations, (ii) it is an intermediate combination of data which means that the mixed pixel is preserved during the process without loss of information in its early combinations [22] and (iii) it can be parallelised to speed up the computation [23] To this end, Liu et al [24] have proposed a framework called multiple-kernel learning-based spectral mixture analysis (MKL-SMA) that integrates an MKL method into the training process of linear spectral mixture analysis.
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call