Hyperspectral Image Classification Based on 3D-2D Hybrid Convolution and Graph Attention Mechanism

Abstract

Convolutional neural networks and graph convolutional neural networks are two classical deep learning models that have been widely used in hyperspectral image classification tasks with remarkable achievements. However, hyperspectral image classification models based on graph convolutional neural networks using only shallow spectral or spatial features are insufficient to provide reliable similarity measures for constructing graph structures, limiting their classification performance. To address this problem, we propose a hyperspectral image classification model combining 3D-2D hybrid convolution and a graph attention mechanism. First, a 3D-2D hybrid convolutional network is constructed and used to rapidly extract deep features that express spatial and spectral associations. Then, the graph is built based on deep spatial-spectral features to enhance the feature representation of the graph. Finally, a network of graph attention mechanisms is adopted to learn long-range spatial connections and to classify them using the extracted spatial features. The experimental results on two datasets, Indian Pine and the University of Pavia, show that the proposed method can achieve higher classification accuracy compared with other advanced methods.