Archetypal analysis for endmember bundle extraction considering spectral variability

With the development of imaging technology, remote sensing images with a high spatial and spectral resolution have become available and have been used in various applications. Although many endmember extraction algorithms have been proposed for hyperspectral data sets which extract/select the standard endmember spectrum for each existing endmember class or scene component, there are still some problems in endmember extraction which lead to inaccurate unmixing. One of the important problems is that spectral variability is inevitable due to the different imaging conditions, especially in a hyperspectral image with a higher spatial resolution. In this article, to account for the spectral variability, an endmember bundle extraction algorithm based on archetypal analysis is proposed, and each material is represented with a few typical spectra. There are three steps in the proposed method of extracting endmember bundles: 1) Looking for pure pixels; 2) the first level archetypal analysis; and 3) the second level archetypal analysis. Experiments with both synthetic and real hyperspectral data sets indicate that, the proposed method could get a well unmixing result using the fewest typical spectra.