A spectral feature extraction algorithm for hyperspectral RS oil spill images

The spectral curve´s feature is important for the nature analysis of the surface in hyperspectral remote sensing process. In this paper, a spectral curve feature extraction algorithm is proposed to deal with the classification of surface features. Several existing methods are compared with the new algorithm by testing the hyperspectral data sets which are obtained by experiments. The results show that the proposed algorithm can better describe the shape of the curve. The experiment on the actual hyperspectral images also shows the effectiveness of the algorithm.