Hyperspectral Image Classification Using Spatial and Spectral Features Based on Deep Learning

In recent years, deep learning has been widely applied to the classification of hyperspectral images and has achieved good results. Most of the hyperspectral image classification experiments use spatial features, and it is easy to ignore the spectral features of the image. In order to make full use of the advantages of convolutional neural network (CNN), We use the MNF(Minimum Noise Fraction) and GABS(Gray value with adaptive band selection) methods to obtain the bands required for classification, and then use 3DCNN to extract the spatial features and spectral features of hyperspectral images. The combination of spatial spectral features not only adds useful information, but also effectively eliminates some of the noise. In this paper, a three-dimensional convolutional neural network is used to simultaneously extract the features of the fused spatial spectral information and classify the hyperspectral image. The proposed method is tested on three publicly available data sets: University of Pavia, Salinas and Indian Pine. The results of the competition show that our method can achieve the state-of-the- art classification accuracy in the three data sets.