Discriminative Spectral-Spatial Attention-Aware Residual Network For Hyperspectral Image Classification

Abstract

Convolutional neural networks (CNNs) have been widely used in remote sensing image analysis, significantly improving the state-of-the-art. In this paper, we present a novel deep residual network based on spectral-spatial attention (DS <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> A-RN) for classification of hyperspectral images. First, we propose an efficient residual block allowing 3D cube inputs and consisting of spectral attention and spatial attention to simultaneously model the explicit relationship between spectral bands and neighboring pixels. Second, a center loss is introduced to combine with softmax loss to enable our model to learn discriminative features by encouraging inter-class separability and intra-class compactness. We evaluate our method for three real hyperspectral images and compare with many existing deep learning methods, showing that the proposed method can achieve state-of-the-art classification performance.