Hyperspectral unmixing is a crucial task for hyperspectral images (HSI) processing, which estimates the proportions of constituent materials of a mixed pixel. Usually, the mixed pixels can be approximated using a linear mixing model. Since each material only occurs in a few pixels in real HSI, sparse nonnegative matrix factorization (NMF) and its extensions are widely used as solutions. Some recent works assume that materials are distributed in certain structures, which can be added as constraints to sparse NMF model. However, they only consider the spatial distribution within a local neighborhood and define the distribution structure manually, while ignoring the real distribution of materials that is diverse in different images. In this paper, we propose a new unmixing method that learns a subspace structure from the original image and incorporate it into the sparse NMF framework to promote unmixing performance. Based on the self-representation property of data points lying in the same subspace, the learned subspace structure can indicate the global similar graph of pixels that represents the real distribution of materials. Then the similar graph is used as a robust global spatial prior which is expected to be maintained in the decomposed abundance matrix. The experiments conducted on both simulated and real-world HSI datasets demonstrate the superior performance of our proposed method.

Convolutional neural networks (CNN) have achieved excellent performance for the hyperspectral image (HSI) classification problem due to better extracting spectral and spatial information. However, CNN can only perform convolution calculations on Euclidean data sets. To solve this problem, recently, graph convolutional neural network (GCN) is proposed, which can be applied to the semi-supervised HSI classification problem. However, the GCN is a direct push learning method, which requires all nodes to participate in training process to get the node embedding. This may bring great computational for the hyperspectral data with large number of pixels. Therefore, in this paper, we propose a inductive learning method to solve the problem. It constructs the graph by sampling and aggregating (graphSAGE) feature from a node's local neighbourhood. This could greatly reduce the space complexity. Moreover, to enhance the classification performance, we also construct graph using spectral and spatial information (spectral-spatial GraphSAGE). Experiments on several hyperspectral image datasets show that the proposed method can achieve better classification performance compared with state-of-the-art HSI classification methods.

Hyperspectral classification is an important technique for remote sensing image analysis. For the current classification methods, limited training data affect the classification results. Recently, Conditional Variational Autoencoder Generative Adversarial Network (CVAEGAN) has been used to generate virtual samples to augment the training data, which could improve the classification performance. To further improve the classification performance, based on the CVAEGAN, we propose a Self-Attention-Based Conditional Variational Autoencoder Generative Adversarial Network (SACVAEGAN). Compared with CVAEGAN, we first use random latent vectors to obtain more enhanced virtual samples, which can improve the generalization performance. Then, we introduce the self-attention mechanism into our model to force the training process to pay more attention to global information, which can achieve better classification accuracy. Moreover, we explore model stability by incorporating the WGAN-GP loss function into our model to reduce the mode collapse probability. Experiments on three data sets and a comparison of the state-of-art methods show that SACVAEGAN has great advantages in accuracy compared with state-of-the-art HSI classification methods.

Recently, convolutional neural networks (CNNs) have demonstrated impressive capabilities in the representation and classification of hyperspectral remote sensing images. Traditional CNNs require massive data to sufficiently train the network. To tackle this problem, graph convolutional network (GCN) has been introduced for hyperspectral image classification. GCN methods usually construct the graph from either spectral or spatial domain, which has not adequately explored the information in the joint spectral–spatial domain. In this article, we propose a superpixel spectral–spatial feature fusion graph convolution network for hyperspectral image classification (S3FGCN). S3FGCN can comprehensively use information in spectral, spatial, and spectral–spatial domains with limited data. Moreover, to enhance the performance, we explore a shared weights’ GCN in the spectral–spatial domain. To further improve the efficiency, superpixels are used to construct the adjacency matrix. Finally, dynamic sampling is adopted to make the model focus more on difficult samples. In the experiments on four datasets, S3FGCN demonstrates better accuracy compared with the state-of-the-art hyperspectral image classification methods.

Due to its ability to capture long-range dependencies, self-attention mechanism based transformer models are introduced for hyperspectral image classification. However, the self-attention mechanism has only spatial adaptability but ignores channel adaptability, thus cannot well extract complex spectral-spatial information in hyperspectral images. To tackle this problem, in this paper, we propose a novel spectral-spatial large kernel attention network (SSLKA) for hyperspectral image classification. SSLKA consists of two consecutive cooperative spectral-spatial attention blocks with large convolution kernels, which can efficiently extract features in spectral and spatial domains simultaneously. In each cooperative spectral-spatial attention block, we employ the spectral attention branch and the spatial attention branch to generate the attention maps, respectively, and then fuse the extracted spatial features with the spectral features. With large kernel attention, we can enhance the classification performance by fully exploiting local contextual information, capturing long-range dependencies, as well as be adaptive in the channel dimension. Experimental results on widely used benchmark datasets show that our method achieves higher classification accuracy in terms of overall accuracy, average accuracy, and Kappa than several state-of-the-art methods.