Algorithm to Unmixing Hyperspectral Images Based on APSO-GMM

The mixed pixels of hyperspectral images can be described effectively through Gaussian Mixture Model, this paper presents a new algorithm for unsupervised unmixing from hyperspectral data, term Adaptive Particle Swarm Optimization Gaussian Mixture Model(APSO-GMM). The algorithm employ hybrid of APSO and EM to find the most advantageous parameters of GMM, the search process of the best particle exploited the parameters estimatation of multiobjective GMM, the algorithm can extract end member and decompose mixed pixels together. Experimental on synthetic and real hyperspectral data demonstrate the proposed algorithm has better unmixing result.