Multiple kernel learning for representation-based classification of hyperspectral images

In the paper we propose a multiple kernel learning framework for representation-based classification (MKL_RC) of hyperspectral images. Unlike the existing methods that often exploit the single feature extraction method or the single kernel method; moreover, the single feature representation and kernelized RC is biased and less stable due to the high coherence of the training samples. The proposed approach is different from traditional kernelized methods and characterized by multiple features and multiple kernel learning in a representation-based classification manner. Experimental results on several real HSI datasets demonstrate that the proposed method can achieve superior performance than the state-of-the-art classification methods.