Ke Wu

Deep learning has shown remarkable effectiveness in hyperspectral image change detection. Cross-domain few-shot learning has emerged as an effective solution to label scarcity in the target domain by exploiting labeled source domain data. However, existing methods often suffer from the interference of domain-sensitive channels and the instability of class prototypes. To address these issues, this article proposes a domain dropout-driven prototype enhancement network for cross-domain few-shot hyperspectral change detection. The method is built upon the metric-based meta-learning to learn transferable knowledge before adapting to new hyperspectral image change detection tasks. Specifically, a dual-branch domain dropout-driven feature extractor is constructed to adaptively dropout domain-sensitive channels, thereby extracting domain-invariant features to mitigate domain shift. Subsequently, a history-aware prototype enhancement module is designed to accumulate historical and current prototype information, thus improving the representativeness and stability of class prototypes. Finally, the trained model is transferred to the target domain to construct a change detection inference module under few-shot conditions. Experiments on several hyperspectral datasets illustrate the effectiveness of the proposed algorithm.

Hyperspectral image change detection (HCD) plays a vital role in remote sensing observation. Although deep learning-based HCD methods have achieved remarkable progress in recent years, several key challenges remain: first, mixed pixels in HSI hinder detection precision, resulting in coarse change results; second, existing methods that rely solely on abundance maps or coarse fusion of pixel-level and subpixel-level features fail to achieve effective matching between them in semantic description and information dimension, limiting the extraction of fine-grained spatial and spectral information; third, most HCD approaches pay limited attention to multiclass change detection, lacking mechanisms to emphasize minority classes at the feature level, which leads to unsatisfactory performance under class imbalance. To address these issues, a three-branch multiscale abundance feature fusion network (TMAFN) is proposed. The main contributions are as follows: first, abundance maps extracted with superpixel constraints is introduced to provide subpixel-level spatial feature consistency and explicit class proportion information; second, a spatial–channel–spatial module is developed to extract fine-grained spatial structures and highly discriminative semantic features, adaptively enhancing spatial and spectral distinctions; third, we specifically design a multihead cross attention mechanism combined with a multiscale residual fusion strategy based on abundance maps to guide the network's focus toward minority categories. Extensive experiments conducted on three benchmark hyperspectral datasets demonstrate that TMAFN consistently outperforms state-of-the-art deep learning models in overall accuracy and class-wise $F1$ score, effectively enhancing the detection of subtle and imbalanced change categories.

Landslides are geological hazards that endanger socioeconomic development and ecological security, with landslide susceptibility mapping (LSM) playing a critical role in risk management and spatial planning. Recently, ensemble learning (EL) models have gained attention for effectively addressing the limitations of individual deep learning (DL) models in LSM. However, EL models always built on single-pixel, multi-factor inputs struggle to capture the spatial structure features of terrain units, limiting their ability to depict complex disaster patterns. Moreover, the scarcity of landslide samples and high annotation costs constrain model performance in LSM. To overcome these challenges, we propose a Stacking model based on multidimensional feature collaboration and pseudo-labeling techniques, referred to as MFP_Stacking. A stacking EL model is first employed in MFP_Stacking to integrate global statistical attribute features extracted from one-dimensional vectors with multi-scale spatial topological features derived from three-dimensional vectors. This strategy of multidimensional feature collaborative modeling enhances the model’s ability to learn complex environmental patterns associated with landslides. Subsequently, pseudo-labeling techniques are adopted to incorporate unlabeled data into auxiliary training, thereby addressing the problem of sample scarcity. MFP_Stacking was applied to LSM in the Zigui–Badong section of the Yangtze River Basin and in Ya’an City, Sichuan Province. Experimental results demonstrate that the proposed model performs well in overcoming limitations in feature representation, alleviating sample scarcity, and enhancing the quality of LSM outcomes. It achieved an average improvement of 2.4% for the Zigui–Badong section and 2% for Ya’an City across various evaluation metrics compared to other models.

Many deep learning-based spatiotemporal fusion (STF) methods have been proven to achieve high accuracy and robustness. Due to the variable shapes and sizes of objects in remote sensing images, pyramid networks are generally introduced to extract multi-scale features. However, the down-sampling operation in the pyramid structure may lead to the loss of image detail information, affecting the model’s ability to reconstruct fine-grained targets. To address this issue, we propose a novel Dual-Stream Enhanced Pyramid based on Generative Adversarial Network (DSEPGAN) for the spatiotemporal fusion of remote sensing images. The network adopts a dual-stream architecture to separately process coarse and fine images, tailoring feature extraction to their respective characteristics: coarse images provide temporal dynamics, while fine images contain rich spatial details. A reversible feature transformation is embedded in the pyramid feature extraction stage to preserve high-frequency information, and a fusion module employing large-kernel and depthwise separable convolutions captures long-range dependencies across inputs. To further enhance realism and detail fidelity, adversarial training encourages the network to generate sharper and more visually convincing fusion results. The proposed DSEPGAN is compared with widely used and state-of-the-art STF models in three publicly available datasets. The results illustrate that DSEPGAN achieves superior performance across various evaluation metrics, highlighting its notable advantages for predicting seasonal variations in highly heterogeneous regions and abrupt changes in land use.

In recent years, graph convolutional networks (GCNs) have been introduced for hyperspectral image (HSI) classification due to their ability to effectively process the inherent graph structure of HSI data. However, existing GCN-based methods heavily rely on manually selecting superpixel segmentation scales, which limits their ability to capture multi-scale contextual relationships adaptively. Moreover, the scarcity of labeled samples in HSI data further degrades the model's robustness. To address these challenges, we propose a novel semi-supervised framework - multiscale feature search graph convolutional network with meta pseudo labels (MFSGCN-MPL). Firstly, the hyperspectral images are segmented into superpixels of different scales, transforming them into a graph structure. After that, a neural network search algorithm is used to optimize the combination of superpixel feature weights, improving the discriminability of feature expression. Finally, meta pseudo labels are generated based on a semi-supervised teacher-student model that shares the same graph convolutional network, and the student network is fine-tuned to enhance its robustness. The proposed MFSGCN-MPL model is implemented on three commonly used HSI datasets and compared with some semi-supervised and supervised classification methods. The results confirmed that the proposed model automatically captured features and achieved higher classification accuracies under small sample conditions.

Remote Sensing Scene Classification (RSSC) is an important and challenging research topic. Transformer-based methods have shown encouraging performance in capturing global dependencies. However, recent studies have revealed that Transformers perform poorly in capturing high frequencies that mainly convey local information. To solve this problem, we propose a novel method based on High-Frequency Enhanced Vision Transformer and Multi-Layer Context Learning (HETMCL), which can effectively learn the comprehensive features of high-frequency and low-frequency information in visual data. First, Convolutional Neural Networks (CNNs) extract low-level spatial structures, and the Adjacent Layer Feature Fusion Module (AFFM) reduces semantic gaps between layers to enhance spatial context. Second, the High-Frequency Information Enhancement Vision Transformer (HFIE) includes a High-to-Low-Frequency Token Mixer (HLFTM), which captures high-frequency details. Finally, the Multi-Layer Context Alignment Attention (MCAA) integrates multi-layer features and contextual relationships. On UCM, AID, and NWPU datasets, HETMCL achieves state-of-the-art OA of 99.76%, 97.32%, and 95.02%, respectively, outperforming existing methods by up to 0.38%.

Existing deep learning-based spatiotemporal fusion (STF) methods for remote sensing imagery often focus exclusively on capturing temporal changes or enhancing spatial details while failing to fully leverage spectral information from coarse images. To address these limitations, we propose a Bidirectional Cross Spatiotemporal Fusion Network with Spectral Restoration (BCSR-STF). The network integrates temporal and spatial information using a Bidirectional Cross Fusion (BCF) module and restores spectral fidelity through a Global Spectral Restoration and Feature Enhancement (GSRFE) module, which combines Adaptive Instance Normalization and spatial attention mechanisms. Additionally, a Progressive Spatiotemporal Feature Fusion and Restoration (PSTFR) module employs multi-scale iterative optimization to enhance the interaction between high- and low-level features. Experiments on three datasets demonstrate the superiority of BCSR-STF, achieving significant improvements in capturing seasonal variations and handling abrupt land cover changes compared to state-of-the-art methods.

High-resolution and thermal infrared hyperspectral data acquired from the Thermal Infrared Airborne Spectrographic Imager (TASI) have been recognized as efficient tools in geology, demonstrating significant potential for rock discernment. Deep learning (DL), as an advanced technology, has driven substantial advancements in lithological mapping by automatically extracting high-level semantic features from images to enhance recognition accuracy. However, gathering sufficient high-quality lithological samples for model training is challenging in many scenarios, posing limitations for data-driven DL approaches. Moreover, existing sample collection approaches are plagued by limited verifiability, subjective bias, and variation in the spectra of the same class at different locations. To tackle these challenges, a novel sample generation method called multi-lithology spectra sample selection (MLS3) is first employed. This method involves multiple steps: multiple spectra extraction, spectra combination and optimization, lithological type identification, and sample selection. In this study, the TASI hyperspectral data collected from the Liuyuan area in Gansu Province, China, were used as experimental data. Samples generated based on MLS3 were fed into five typical DL models, including two-dimensional convolutional neural network (2D-CNN), hybrid spectral CNN (HybridSN), multiscale residual network (MSRN), spectral-spatial residual network (SSRN), and spectral partitioning residual network (SPRN) for lithological mapping. Among these models, the accuracy of the SPRN reaches 84.03%, outperforming the other algorithms. Furthermore, MLS3 demonstrates superior performance, achieving an overall accuracy of 2.25–6.96% higher than other sample collection methods when SPRN is used as the DL framework. In general, MLS3 enables both the quantity and quality of samples, providing inspiration for the application of DL to hyperspectral lithological mapping.

Coral reef bleaching events have become more frequent all over the world and pose a serious threat to coral reef ecosystems. Therefore, there is an urgent need for better detection of coral reef bleaching in a time- and cost-saving manner. In recent years, remote sensing technology has often been utilized and gained recognition for coral reef bleaching detection. However, bleaching corals in the water always have weak spectral change signals, causing difficulties in using remote sensing data. Additionally, uneven change samples make it challenging to adequately capture the details of coral reef bleaching detection and produce thematic maps. To resolve these problems, a novel method named coral reef bleaching detection by positive-unlabeled bagging (CBD-PUB) is proposed in this paper. To test the capacity of the method, a series of multi-temporal Sentinel-2 remote sensing images are utilized, and Lizard Island in Australia is taken as a case study area. The pseudo-invariant feature atmospheric correction (PIF) algorithm is adopted to improve coral reef bleaching spectral signals. After that, CBD-PUB is employed to effectively explore coral reef bleaching variation and its corresponding influence relations. The experimental results show that the overall accuracy of bleaching detection by the proposed algorithm reaches 92.1% and outperforms the traditional method. It fully demonstrates the feasibility of the model for the field of coral reef bleaching detection and provides assistance in the monitoring and protection of coral environments.