An Innovative Method of Endmember Extraction of Hyperspectral Remote Sensing Image Using Elitist Ant System (EAS)

Spectrum unmixing is an important content of Hyperspectral Remote Sensing (HRS) image processing, and end member extraction is a key step of spectrum unmixing. Most of the available end member extraction algorithms are based on linear spectral mixture model, and algorithms´ accuracy is closely related to data´s quality. According to linear spectral mixture model, this paper redefined end member extraction as a combinatorial optimization problem (COP), then introduced Elitist Ant System (EAS), a classical type of Ant Colony Optimization (ACO), to solve COP. The experiments revealed that a better end member extraction result could be obtained from HRS data of the same quality by using EAS than N-finder and VCA.