Superpixel-Based Semisupervised Active Learning for Hyperspectral Image Classification

Hyperspectral image (HSI) classification has become a hot topic in the field of remote sensing. In general, the complex characteristics of hyperspectral data make the accurate classification of such data challenging for traditional machine learning methods. In addition, hyperspectral imaging often deals with an inherently nonlinear relation between the captured spectral information and the corresponding materials. In recent years, deep learning has been recognized as a powerful feature-extraction tool to effectively address nonlinear problems and widely used in a number of image processing tasks. Motivated by those successful applications, deep learning has also been introduced to classify HSIs and demonstrated good performance. This survey paper presents a systematic review of deep learning-based HSI classification literatures and compares several strategies for this topic. Specifically, we first summarize the main challenges of HSI classification which cannot be effectively overcome by traditional machine learning methods, and also introduce the advantages of deep learning to handle these problems. Then, we build a framework that divides the corresponding works into spectral-feature networks, spatial-feature networks, and spectral-spatial-feature networks to systematically review the recent achievements in deep learning-based HSI classification. In addition, considering the fact that available training samples in the remote sensing field are usually very limited and training deep networks require a large number of samples, we include some strategies to improve classification performance, which can provide some guidelines for future studies on this topic. Finally, several representative deep learning-based classification methods are conducted on real HSIs in our experiments.

Few-shot learning (FSL) uses prior knowledge and supervised experience to effectively classify hyperspectral images (HSIs), thereby reducing the cost of large numbers of labeled samples. However, existing few-shot methods ignore the correlation between cross-domain feature channels, and the feature representation ability is insufficient. To address above issue, this paper proposes a novel Residual Channel Attention Based Sample Adaptation Few-Shot Learning for Hyperspectral Image Classification(RCASA-FSL) for hyperspectral image classification (HSIC), which can capture and enhance cross-domain dependencies through multi-layer residual connection and random-based feature recalibration. Specifically, a Deep Residual Feature Channel Attention Mechanism (DRFCAM) is designed to obtain cross-domain dependencies by residual concatenation, and further the residual structure is stacked for mining depth discrimination information. Furthermore, a new Random-based Feature Recalibration Module (RFRM) is proposed to reassign the feature weights via random matrix, which fully explore feature weight relationships to guide the sample adaptation process. Besides, we design a joint loss function with combining the FSL loss and domain adaptive loss for further optimization model. Experiments conducted on several standard hyperspectral datasets demonstrate that the proposed RCASA-FSL is superior to other FSL techniques in both quantitative and qualitative aspects.

Class incremental learning with analytic learning for hyperspectral image classification

This paper develops a new Neyman–Pearson detection approach, to be called band-specified virtual dimensionality (BSVD), to estimating the number of bands required by band selection (BS), $n_{\mathrm {BS}}$ , as well as finding desired bands at the same time. Its idea is derived from target-specified virtual dimensionality (TSVD) where targets under hypotheses as signal sources in TSVD are replaced with bands as signal sources and the test statistics derived for a Neyman–Pearson detector (NPD) is signal-to-noise ratio (SNR) that is used to derive orthogonal subspace projection (OSP) approach for hyperspectral image classification and dimensionality reduction. Accordingly, the resulting virtual dimensionality is referred to as OSP-based BSVD. Several benefits resulting from BSVD cannot be offered by the traditional BS methods. One is its direct approach to dealing with $n_{\mathrm {BS}}$ . Another is no-search strategy needed for finding optimal bands. Instead, it uses NPD to determine and rank desired bands for band prioritization. Most importantly, it determines $n_{\mathrm {BS}}$ and finds desired bands simultaneously and progressively.

Variations in atmosphere, lighting, and imaging systems result in diverse category distributions in hyperspectral imagery, impacting the accuracy of cross-domain hyperspectral image classification (HSIC). Unsupervised domain adaptation (UDA) aims to address this issue by learning a model that generalizes effectively across domains, leveraging labels only from source domain (SD). Most existing UDA methods focus on aligning distributions between domains without fully considering the valuable information within individual domains. To fill this gap, this letter proposes a domain-collaborative contrastive learning (DCCL) method. DCCL integrates a novel pseudo-labeling strategy with a cross-domain contrastive learning (CL) framework. Specifically, in the pseudo-labeling phase, the confident examples in target domain (TD) are collaboratively labeled according to the labeled examples in SD and the class centers in TD. Then, the CL phase simultaneously minimizes in-domain and cross-domain contrastive loss to promote the aggregation of examples from the same category in both domains. Experimental results demonstrate that the DCCL achieves the accuracy rates of 93.47% and 54.59% on Pavia and Indiana datasets, respectively, surpassing the performance of other state-of-the-art UDA methods. Our source code is available at <uri xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">https://github.com/Leap-luohaiyang/DCCL-2024</uri>.

Multiclass feature learning for hyperspectral image classification: Sparse and hierarchical solutions

The vision transformer (ViT) has become a hot topic in image processing due to its global feature extraction capabilities. However, the ViT suffers from over-smoothing in feature extraction and over-fitting in the training procedure, so it is hard to achieve satisfactory performance in hyperspectral image (HSI) classification. To address these issues, we propose a ViT with contrastive learning (CViT). The network architecture includes a patch embedding module, transformer blocks, and a classifier. The training of CViT can be considered as an optimization problem with a supervised contrastive loss, an unsupervised contrastive loss, and an ℓ <sub xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sub> -regularizer with respect to linear self-attention weights. Specifically, the supervised contrastive loss is proposed to alleviate the negative effects of HSI features’ spectral variability and spatial diversity by increasing intra-class consistency. On the other hand, the unsupervised contrastive loss is exploited to reduce redundancy by reconstructing global structural information. In particular, regularized linear self-attention weights reduce the over-smoothing issue. Extensive experimental results on three HSI datasets demonstrate that the proposed CViT achieves competitive performance.

Despite the rapid development of deep learning in hyperspectral image classification (HSIC), most models require a large amount of labeled data, which are both time-consuming and laborious to obtain. However, contrastive learning can extract spatial–spectral features from samples without labels, which helps to solve the above problem. Our focus is on optimizing the contrastive learning process and improving feature extraction from all samples. In this study, we propose the Unlocking-the-Potential-of-Data-Augmentation (UPDA) strategy, which involves adding superior data augmentation methods to enhance the representation of features extracted by contrastive learning. Specifically, we introduce three augmentation methods—band erasure, gradient mask, and random occlusion—to the Bootstrap-Your-Own-Latent (BYOL) structure. Our experimental results demonstrate that our method can effectively improve feature representation and thus improve classification accuracy. Additionally, we conduct ablation experiments to explore the effectiveness of different data augmentation methods.

Hyperspectral image classification is an important topic in the field of remote sensing. However, the high dimensionality and high spatial-spectral correlation of hyperspectral image will easily lead to poor classification performance due to the Hughes phenomenon. In this paper, we proposed a adaptive group local Riemannian embedding, called AGLRE, to extract the spatial and spectral features from hyperspectral image. It firstly mapped original data into a Riemannian manifold by constructing region covariance matrices for each pixel of hyperspectral image. And the multiple local tangent space on Riemannian manifold were learned by adaptive neighbourhood strategy. Lastly, the local linear embedding was applied to reduce the dimensionality and merge multiple tangent space into a global coordinate. Experimental results on public hyperspectral data set showed that the proposed method can achieve higher classification performance than other competing algorithms.

In Hyperspectral Image (HSI) classification, acquiring large quantities of high-quality labeled samples is typically costly and impractical. Traditional deep learning methods are limited in such scenarios due to their dependence on sample quantities. To address this challenge, researchers have turned to Few-Shot Learning (FSL). Although existing FSL methods improve classification performance by enhancing domain invariance through domain adaptation, they often overlook the critical issue of high inter-class similarity and large intra-class variability. Moreover, during domain alignment, features of different categories within the same domain may become confused. To address these issues, this paper proposes a novel Domain-Invariant Few-Shot Contrastive Learning (DIFSCL) method, which combines domain adaptation and contrastive learning strategies to not only learn domain-invariant features but also significantly enhance inter-class discriminability. Based on this, we further design a multi-scale adaptive attention mechanism for a hyperspectral feature extraction network to more effectively extract and optimize generalized features within the DIFSCL framework, significantly improving intra-class consistency and inter-class discriminative capability of features. Experimental results on three widely used HSI datasets demonstrate that our method significantly outperforms existing techniques in few-shot classification tasks.

In this research, a deep learning approach for hyperspectral image (HSI) classification was developed, incorporating attention mechanisms, multiscale feature learning, and utilization of unsampled pixels. The proposed model, multiscale attention-based hybrid spectral network and UNet (MSA-HybridSN-U), was evaluated on three benchmark datasets: Indian Pines, University of Pavia, and University of Houston using Overall Accuracy (OA), Average Accuracy (AA), and Kappa index (K). The proposed model MSA-HybridSN-U had high accuracy on Indian Pines (IP), Pavia University (PU), and Houston University (HU) datasets. It achieved the highest average accuracy compared with other models like 3D CNN, HybridSN, M3D-DCNN, DBDA, and ACA-HybridSN and higher than 2D CNN and SVM. For the IP dataset, the OA and AA scores are both 99.71% and 99.65% respectively, indicating a high level of accuracy in classifying the samples. The Kappa statistic, which measures inter-annotator agreement, is also high at 0.997, suggesting that the method is consistent in its predictions. For the PU dataset, the results are even higher with OA and AA scores of 99.97% and 99.89% respectively, and a Kappa statistic of 0.999, indicating even higher accuracy and consistency. Finally, for the HU dataset, the results are similarly high with OA and AA scores of 99.47% and 99.92% respectively, and a Kappa statistic of 0.994. The use of attention mechanisms, multiscale features, and unsampled pixels has improved the classification performance of the model. The combination of spectral and spatial attention modules improved the accuracy the most, with the highest accuracy of 99.97% on the PU dataset. Utilizing unsampled pixels in the classification process resulted in a noticeable improvement in accuracy, particularly on the HU dataset with an improvement of 0.29%. The results show the effectiveness of combining multiscale features and attention modules in improving the accuracy of HSI classification.

In recent years, the interest in semi-supervised learning has increased, combining supervised and unsupervised learning approaches. This is especially valid for classification applications in remote sensing, while the data acquisition rate in current systems has become fairly large considering high- and very-high resolution data; yet on the other hand, the process of obtaining the ground truth data may be cumbersome for such large repositories. In this paper, we investigate the application of semi-supervised learning approaches and particularly focus on the small sample size problem. To that extend, we consider two basic unsupervised approaches by enlarging the initial labeled training set as well as an ensemble-based self-training method. We propose different strategies within self-training on how to select more reliable candidates from the pool of unlabeled samples to speed-up the learning process and to improve the classification performance of the underlying classifier ensemble. We evaluate the effectiveness of the proposed semi-supervised learning approach over polarimetric SAR data. Results show that the proposed self-training approach using an ensemble-based classifier that is initially trained over a small training set can achieve a similar performance level of a fully supervised learning approach where the training is performed over significantly larger labeled data. Considering the difficulties of the manual data labeling in such massive volumes of SAR repositories, this is indeed a promising accomplishment for semi-supervised SAR classification.

A semi-supervised deep learning approach for cropped image detection

AN ASSESSMENT OF SPATIAL VARIATION OF LAND SURFACE CHARACTERISTICS OF MINNA, NIGER STATE NIGERIA FOR SUSTAINABLE URBANIZATION USING GEOSPATIAL TECHNIQUES

Deep neural network has been extensively applied to hyperspectral image (HSI) classification recently. However, its success is greatly attributed to numerous labeled samples, whose acquisition costs a large amount of time and money. In order to improve the classification performance while reducing the labeling cost, this article presents an active deep learning approach for HSI classification, which integrates both active learning and deep learning into a unified framework. First, we train a convolutional neural network (CNN) with a limited number of labeled pixels. Next, we actively select the most informative pixels from the candidate pool for labeling. Then, the CNN is fine-tuned with the new training set constructed by incorporating the newly labeled pixels. This step together with the previous step is iteratively conducted. Finally, Markov random field (MRF) is utilized to enforce class label smoothness to further boost the classification performance. Compared with the other state-of-the-art traditional and deep learning-based HSI classification methods, our proposed approach achieves better performance on three benchmark HSI data sets with significantly fewer labeled samples.