Abstract: In the field of hyperspectral remote sensing image classification, spatial and spectral feature fusion can improve the effect of classification. A multifeature spectral and spatial classification method for hyperspectral images was proposed. In the postprocessing procedure, the saltandpepper noises were removed by using superpixel information. It was also used in the preprocessing procedure, and the feature vector of pixels was weighted by superpixel information. The experiment results show that the results of proposed method are better than the present methods.

Key words: hyperspectral remote sensing image classification, spatial feature, spectral feature, superpixel, linear weighted fusion