Efficient Deep Belief Network Based Hyperspectral Image Classification

Hyperspectral image (HSI) classification accuracy has been greatly improved by employing deep learning. The current research mainly focuses on how to build a deep network to improve the accuracy. However, these networks tend to be more complex and have more parameters, which makes the model difficult to train and easy to overfit. Therefore, we present a lightweight deep convolutional neural network (CNN) model called S2FEF-CNN. In this model, three S2FEF blocks are used for the joint spectral–spatial features extraction. Each S2FEF block uses 1D spectral convolution to extract spectral features and 2D spatial convolution to extract spatial features, respectively, and then fuses spectral and spatial features by multiplication. Instead of using the full connected layer, two pooling layers follow three blocks for dimension reduction, which further reduces the training parameters. We compared our method with some state-of-the-art HSI classification methods based on deep network on three commonly used hyperspectral datasets. The results show that our network can achieve a comparable classification accuracy with significantly reduced parameters compared to the above deep networks, which reflects its potential advantages in HSI classification.

In recent years, numerous deep learning (DL)-based frameworks have been proposed for hyperspectral image classification (HSIC). Considering a large number of spectral bands of hyperspectral images (HSIs), it is still challenging to effectively utilize the spectral information and achieve accurate classification when few training samples are available. To make full use of the spectral-spatial information in HSIs with few training samples, in this letter we propose a lightweight end-to-end attention-enhanced feature fusion network (AeF <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> N). The proposed AeF <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> N consists of four sequential stages, i.e., spectral feature augmentation, spatial contextual feature interaction, spectral feature augmentation, and classification. The first and third stages are used to capture and augment the discriminative spectral features, while the second stage is used to capture spatial information. Notably, two novel attention blocks, spectral augmentation attention (SAA) and spatial integration attention (SIA) are interactively introduced to capture significant spectral and spatial information, respectively. Based on the proposed spectral and spatial feature discrimination stages, the AeF <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">2</sup> N effectively identifies both spectrally significant (e.g., irregular small objects) and spatially significant (e.g., specific-shaped objects) land objects with high accuracy. Experimental results obtained on three benchmark hyperspectral datasets demonstrate the superiority of the proposed approach compared with six state-of-the-art DL-based methods in terms of higher classification accuracy and efficiency.

Adaptive spectral-spatial feature fusion network for hyperspectral image classification using limited training samples

Creating accurate land use and land cover maps using remote sensing images is one of the most important applications of remotely sensed data. Abundant spectral information in hyperspectral images (HSI) makes it possible to distinguish materials that would not be distinguishable by multi-spectral sensors. Spectral and spatial information from HSI is of primary importance for image classification. In this study, a hybrid stacked autoencoder (SAE) architecture and support vector machine (SVM) classifier was constructed to classify the HSI. The SAE architecture is constituted by stacking a multiple autoencoder (AE) deep learning network that consists in the encoder and decoder process. Spatial features in a neighbor region extracted from the principal component analysis (PCA) and the texture feature extracted from the gray-level cooccurrence matrix (GLCM) were fed into the classifier. It was found that the best result was from the combination of GLCM texture feature, PCA spatial feature, and spectral feature. Meanwhile, the representative features derived from SAE deep learning network were better than the original features. It reminded us that extracting the representative features from hyperspectral images is a key step of improving classification accuracy.

Designing active learning (AL) techniques to determine informative training samples for hyperspectral image (HSI) classification is an open research issue. In this chapter, several spectral–spatial AL techniques are designed by exploiting both, an interesting approach to fuse spectral and spatial information and different combinations of existing uncertainty and diversity criteria. In order to fuse spectral and spatial information, the dimensionality of HSI is reduced and mean filtering (for incorporation of spatial information) is applied to each component image in the reduced domain of HSI considering multiple windows of different sizes. The filtered images are concatenated with original component images to form an extended spatial profile for the HSI. These spectral–spatial features are used with different combinations of uncertainty and diversity criteria to design several spectral–spatial AL techniques for determining most informative pixels. The experiments carried out on two real HSI data sets show that the spectral–spatial AL methods are more robust than the AL methods based on spectral measurements only.

Deep learning methods, especially convolutional neural networks(CNN), have been widely used in hyperspectral image(HSI) classification. Recently, graph convolutional networks (GCN) have shown great potential in HSI classification problem. However, the existing GCN-based methods have several problems. First, the existing methods rely too much on the adjacency matrix, which cannot be changed during training. Furthermore, most of them can only use a single kind of feature, and fail to extract the spectral-spatial information from the HSI. Finally, for the existing GCN-based methods, it is difficult to achieve the same accuracy as the mature CNN methods. In this paper, we propose a spectral-spatial hypergraph convolutional neural network (S²HCN) for HSI classification. Compared with the existing GCN-based methods, S²HCN has the following advantages. Different from the adjacency matrix that is fixed during training of GCN, S²HCN can dynamically update the weight of the hyperedge during training, which reduces the reliance on prior information to a certain extent. In addition, S²HCN generates hyperedges from the spectral and spatial features independently, and adopts the incidence matrix composed of all hyperedges to replace the adjacency matrix in GCN. In this way, the spectral and spatial features can be better integrated. Finally, compared to a simple graph structure, the hypergraph structure can express the high-dimensional relationships in the data, which is beneficial to classification problems. Sufficient experiments on two popular HSI datasets have proved the effectiveness of S²HCN.

Hyperspectral image (HSI) classification has become one of the most significant tasks in the field of hyperspectral analysis. However, classifying each pixel in HSI accurately is challenging due to the curse of dimensionality and limited training samples. In this paper, we present an HSI classification architecture called camera spectral response network (CSR-Net), which can learn the optimal camera spectral response (CSR) function for HSI classification problems and effectively reduce the spectral dimensions of HSI. Specifically, we design a convolutional layer to simulate the capturing process of cameras, which learns the optimal CSR function for HSI classification. Then, spectral and spatial features are further extracted by spectral and spatial attention modules. On one hand, the learned CSR can be implemented physically and directly used to capture scenes, which makes the image acquisition process more convenient. On the other hand, compared with ordinary HSIs, we only need images with far fewer bands, without sacrificing the classification precision and avoiding the curse of dimensionality. The experimental results of four popular public hyperspectral datasets show that our method, with only a few image bands, outperforms state-of-the-art HSI classification methods which utilize the full spectral bands of images.

To overcome the difficulty of automating and intelligently classifying the ground features in remote-sensing hyperspectral images, machine learning methods are gradually introduced into the process of remote-sensing imaging. First, the PaviaU, Botswana, and Cuprite hyperspectral datasets are selected as research subjects in this study, and the objective is to process remote-sensing hyperspectral images via machine learning to realize the automatic and intelligent classification of features. Then, the basic principles of the support vector machine (SVM) and extreme learning machine (ELM) classification algorithms are introduced, and they are applied to the datasets. Next, by adjusting the parameter estimates using a restricted Boltzmann machine (RBM), a new terrain classification model of hyperspectral images that is based on a deep belief network (DBN) is constructed. Next, the SVM, ELM, and DBN classification algorithms for hyperspectral image terrain classification are analysed and compared in terms of accuracy and consistency. The results demonstrate that the average detection accuracies of ELM on the three datasets are 89.54%, 96.14%, and 96.28%, and the Kappa coefficient values are 0.832, 0.963, and 0.924; the average detection accuracies of SVM are 88.90%, 92.11%, and 91.68%, and the Kappa coefficient values are 0.768, 0.913, and 0.944; the average detection accuracies of the DBN classification model are 92.36%, 97.31%, and 98.84%, and the Kappa coefficient values are 0.883, 0.944, and 0.972. The results also demonstrate that the classification accuracy of the DBN algorithm exceeds those of the previous two methods because it fully utilizes the spatial and spectral information of hyperspectral remote-sensing images. In summary, the DBN algorithm that is proposed in this study has high application value in object classification for remote-sensing hyperspectral images.

Most hyperspectral image (HSI) classification methods assume that all classes in the test set are present during training. However, in real-world applications, acquiring labeled training samples is challenging. As a result, it is difficult for the training dataset to cover all possible land cover types, leading to the generalized zero-shot learning (GZSL) problem. Recently, vision-language models (VLMs) have provided rich semantic priors for land cover classes, offering promising potential for GZSL. However, two fundamental gaps hinder their application to HSI classification: the task paradigm gap, arising from the difference between image-level VLMs and the pixel-level HSI classification task; and the knowledge gap, due to the inconsistency between VLM features and HSI spectral–spatial representations. To bridge both gaps, a novel framework leveraging VLM semantic priors for GZSL in HSI classification is proposed, primarily using pseudo-labeling technique to provide knowledge for unseen classes. Specifically, a pseudo-label generation and enhancement module enables a paradigm transition from image-level understanding to pixel-level classification by incorporating HSI’s spatial information. A pseudo-label correction module then refines noisy labels using spectral cues to address the knowledge gap. Finally, a global learning strategy integrates pseudo-label distillation, supervised learning, and feature regularization to classify seen classes while enabling generalization to unseen ones. Experiments on benchmark HSI datasets demonstrate the proposed method’s superiority in generalized zero-shot classification. This work highlights the potential of VLMs in advancing HSI classification in practical applications.

In this paper, we propose a hyperspectral image (HSI) classification method using spectral-spatial long short term memory (LSTM) networks. Specifically, for each pixel, we feed its spectral values in different channels into Spectral LSTM one by one to learn the spectral feature. Meanwhile, we firstly use principle component analysis (PCA) to extract the first principle component from a HSI, and then select local image patches centered at each pixel from it. After that, we feed the row vectors of each image patch into Spatial LSTM one by one to learn the spatial feature for the center pixel. In the classification stage, the spectral and spatial features of each pixel are fed into softmax classifiers respectively to derive two different results, and a decision fusion strategy is further used to obtain a joint spectral-spatial results. Experiments are conducted on two widely used HSIs, and the results show that our method can achieve higher performance than other state-of-the-art methods.

Hyperspectral Image (HSI) classification is one of the most pervasive issue in hyperspectral remote sensing field. Deep learning is an efficient learning algorithm that has been recently applied to HSI classification. This paper proposes a new spectralspatial HSI classification method based on the deep features extraction using stacked-auto-encoders (SAE) and unsupervised HIS segmentation. Specifically, first the SAE model is exploited as a classical spectral information-based classifier to extract the deep features. Second, spatial dominated information is extracted by using effective boundary adjustment based segmentation technique. Finally, maximum voting criteria is used to merge the extracted spectral and spatial features, which results into the accurate spectral-spatial HSI classification. Experimental results with widely-used hyperspectral data confirms that the new spectral and spatial classification approach is able to improve results significantly in terms of classification accuracies.

Deep learning has become a mainstream method of hyperspectral image (HSI) classification. Many DL-based methods exploit spatial-spectral features to achieve better classification results. However, due to the complex backgrounds in HSIs, existing methods usually show unsatisfactory performance for the class pixels located on the land-cover category boundary area. In large part, this is because the network is susceptible to interference by the irrelevant information around the target pixel in the training stage, resulting in inaccurate feature extraction. In this paper, a new multibranch transformer architecture (SST-M) that assembles spatial attention and extracts spectral features is proposed to address this problem. The transformer model has a global receptive field and thus can integrate global spatial position information in the HSI cube. Meanwhile, we design a spatial sequence attention model to enhance the useful spatial location features and weaken invalid information. Considering that HSIs contain considerable spectral information, a spectral feature extraction model is designed to extract discriminative spectral features, replacing the widely used PCA method and obtaining better classification results than it. Finally, inspired by semantic segmentation, a mask prediction model is designed to classify all of the pixels in the HSI cube; this guides the neural network to learn precise pixel characteristics and spatial distributions. To verify the effectiveness of our algorithm (SST-M), quantitative experiments were conducted in three well-known datasets, namely, IP, PU, and KSC. The experimental results demonstrate that the proposed model achieves better performance than the other state-of-the-art methods.

Hyperspectral image (HSI) classification has been extensively applied for analyzing remotely sensed images. HSI data consist of multiple bands that provide abundant spatial information. Convolutional neural networks (CNNs) have emerged as powerful deep learning methods for processing visual data. In recent work, CNNs have shown impressive results in HSI classification. In this paper, we propose a hierarchical neural network architecture called feature extraction with hybrid spectral CNN (FE-HybridSN) to extract superior spectral–spatial features. FE-HybridSN effectively captures more spectral–spatial information while reducing computational complexity. Considering the prevalent issue of class imbalance in experimental datasets (IP, UP, SV) and real-world hyperspectral datasets, we apply the focal loss to mitigate these problems. The focal loss reconstructs the loss function and facilitates effective achievement of the aforementioned goals. We propose a framework (FEHN-FL) that combines FE-HybridSN and the focal loss for HSI classification and then conduct extensive HSI classification experiments using three remote sensing datasets: Indian Pines (IP), University of Pavia (UP), and Salinas Scene (SV). Using cross-entropy loss as a baseline, we assess the hyperspectral classification performance of various backbone networks and examine the influence of different spatial sizes on classification accuracy. After incorporating focal loss as our loss function, we not only compare the classification performance of the FE-HybridSN backbone network under different loss functions but also evaluate their convergence rates during training. The proposed classification framework demonstrates satisfactory performance compared to state-of-the-art end-to-end deep-learning-based methods, such as 2D-CNN, 3D-CNN, etc.

Local covariance matrix descriptor is a new spatial-spectral feature generation method. It has been successfully applied for remote sensing image classification. Meanwhile, there are some critiques of it because it neglects nonlinear relationships between features, which are serious when applied to hyperspectral images (HSIs). So, the present paper aims to develop weighted local kernel matrix (WLKM) descriptors for the spatial-spectral classification of HSI. The developed weighted local kernel matrix features, including spectral-textural-geometrical aspects, have been used in two classification schemes proposed. In the first approach, called ‘early fusion’, the weighted sum of WLKM descriptors derived from spectral and spatial features is classified using the log-Euclidean kernel SVM. In the second approach, called ‘late fusion’, a multiple log-Euclidean kernel SVM strategy based on the WLKM descriptors of spatial and spectral features is developed for HSI classification. Experiments on three widely used HSI datasets have proved the superiority of the proposed approaches over some recent spatial-spectral HSI classification techniques.

For hyperspectral image (HSI) classification, two branch networks generally use the convolution neural networks (CNNs) to extract the spatial features and the long short-term memory (LSTM) to learn the spectral features. However, CNN with a local kernel neglects the global properties of the whole HSI. LSTM doesn't consider the macroscopic and detailed information of spectra. In this paper, we propose a dual-view spectral and global spatial feature fusion network (DSGSF) to extract the spatial-spectral features for HSI classification, including a spatial subnetwork and a spectral subnetwork. In the spatial subnetwork, we propose a global spatial feature representation model based on the encoder-decoder structure with channel attention and spatial attention to learn the global spatial features. In the spectral subnetwork, we design a dual-view spectral feature aggregation model with view attention to learn the diversity of spectral features. By fusing the two subnetworks, we construct DSGSF to extract the spatial-spectral features of HSI with strong discriminating performance. Experimental results on three public datasets illustrate that the proposed method can achieve competitive results compared with the state-of-the-art methods. Code: https://github.com/RZWang-WH/DSGSF.