Hyperspectral Image Classification Based On Deep Learning

Hyperspectral remote sensing technology is able to provide abundant information in the detection and recognition of ground objects,which has been widely used in the research of earth's surface science.At present,the hyperspectral image(HSI)classification has become one of the hotspots in the field of hyperspectral remote sensing.The traditional methods based on manual feature extraction only extract the superficial features of hyperspectral image,which will degenerate the feature representation capability.The classification methods based on deep learning are able to extract deeper and more comprehensive features than traditional ways.However,the existing methods based on deep learning neglect the correlation between the spatial and spectral information of hyperspectral image.Under this background,this paper mainly study the hyperspectral image classification methods based on deep learning.A novel joint spatial-spectral hyperspectral image classification method based on different-scale two-stream convolutional network and spatial enhancement strategy is proposed in this paper.This method is able to extract expressive spatial and spectrum features of hyperspectral image under limited training samples.First,the pixel blocks at different scales around the center pixel are selected as the basic units to be processed.Then,a spatial enhancement strategy is designed to obtain various spatial location information under the limited training samples by the spatial rotation and row-column transformation.Finally,the spatial-spectral features are learned by a different-scale two-stream convolutional network,and the classification result of the center pixel is obtained by a softmax layer.Experimental results demonstrate that the proposed method can improve the classification performance compared with other state-of-the-art methods.To further utilize the edge information of hyperspectral image,a novel hyperspectral image classification method based on edge-preserving two-stream convolutional network is proposed.This method sufficiently apply edge-preserving filter to the hyperspectral image classification,which respectively put the original hyperspectral image and the image processed by edge-preservation filtering to the two branches of two-stream convolutional network to get optimized spatial features and spectral features.In this way,the two kinds of information can be learned at the same time.Experimental results show that the proposed method can further improve the classification accuracy based on different-scale convolutional network.