June 4th afternoon, a half-day tutorial on :
Spectral Unmixing of Hyperspectral Data
Professor Antonio J. Plaza, Department of Technology of Computers and Communications, University of Extremadura, Spain
Spectral mixture analysis, also called spectral unmixing, has been widely used in remote sensing image analysis. It involves the separation of a mixed spectral signature into its pure components or spectra (called endmembers), and the estimation of the abundance value for each endmember. A standard technique for spectral unmixing is the linear mixture model (LMM), which assumes that the collected spectra at the spectrometer can be expressed in the form of a linear combination of endmembers weighted by their corresponding abundances. Although the linear model has practical advantages such as ease of implementation and flexibility in different applications, there are many naturally occurring situations where a nonlinear mixture model (NMM) may best characterize the resultant mixed spectra for certain endmember distributions. In particular, nonlinear mixtures generally occur in situations where endmember components are randomly distributed throughout the field of view of the instrument. In those cases, the mixed spectra collected at the imaging instrument is better described by assuming that part of the source radiation is multiply scattered before being collected at the sensor.
This tutorial will be specifically focused on illustrating standard techniques and recent advances in spectral unmixing for hyperspectral image analysis using either LMMs or NMMs. For a hyperspectral image, its high data dimensionality relaxes the limitation imposed on the number of endmembers (and their abundance maps) that can be retrieved; however, this high dimensionality and data complexity bring about additional challenges for spectral unmixing. Techniques discussed in this tutorial will cover important approaches, including both semi-supervised and fully automatic endmember extraction algorithms, unconstrained and fully constrained abundance estimation techniques, estimation of the number of endmembers, and other recent advances in spectral unmixing, with the consideration of hyperspectral image specialty. It is for middle-level researchers with remote sensing image processing background.