Abstract: As the gaussian radial basis function (RBF) is based on the Euclidean distance of two spectral vectors, it is not sensitive for variation of spectral curves of a material, which results in decrease of the performance of the RI3F based spectral clustering for hyperspectral imagery degenerate. In order to solve this problem, according to the spectral curves similarity description, a novel spectral similarity measurement based on spectral angle cosine was proposed, and the measurement was used to build the affinity matrix used by spectral clustering algorithms. Finally, the experiments carried on with several hyperspectral data. The results of the experiments prove the validity of the proposed method.

Key words: Hyperspectral image, Spectral clustering, Normalized cut, Spectral similarity measurement