A sparse multiple endmember spectral mixture analysis algorithm of hyperspectral image

In the traditional linear spectral mixture model, a class is represented by a single endmember. However, the intra-class spectral variability is usually large, so an endmember is difficult to portray a category accurately, leading to incorrect unmixing results. Some algorithms play a positive role in overcoming the endmember variability, but there are shortcomings on computation intensive, unsatisfactory unmixing results and so on. For these issues, we have proposed a sparse multiple endmember spectral mixture analysis algorithm (SMESMA). First determine the intra-class spectra of all the feature classes for each pixel using orthogonal matching pursuit algorithm (OMP), then find the optimal number of endmember combinations according to the relative increase in root-mean-square error to avoid over-fitting. Synthetic and real data experiments show that the SMESMA unmixing results are ideal comparatively and the abundance error is the lowest among the five methods and multiple endmember spectral mixture analysis is more reasonable.