2 May 2025 in IEEE Geoscience and Remote Sensing Letters
Landslide extraction plays a critical role in disaster prevention and mitigation. However, acquiring sufficient landslide samples is often challenging, significantly limiting the performance of existing extraction methods. Self-supervised learning approaches, such as masked autoencoder (MAE) reconstruction and contrastive learning, offer a promising solution by reducing reliance on labeled data. Nonetheless, Transformer-based MAEs are computationally intensive, while CNN-based MAEs often struggle to capture global contextual information. To address these limitations, this letter proposes a CNN-Transformer hybrid masked autoencoder integrated with contrastive learning (CT-MAE) for pretraining a landslide extraction model. HRNet is employed for landslide extraction. An HRNet-based MAE, which has the c-apability to preserve the resolution and effectively integrate both shallow and deep features, is proposed for pretraining the HRNet. Moreover, the traditional Transformer-based MAE is applied as an auxiliary branch, and contrastive learning is utilized on the features extracted from the encoders of both MAEs. This auxiliary branch serves as a form of data augmentation, providing supplementary supervision, and enables the CNN-based MAE to capture richer global features without modifying the underlying CNN architecture. Experimental results on two publicly available datasets demonstrate that the proposed pretraining approach significantly enhances landslide extraction performance, highlighting its potential for large-scale applications.
2 May 2025 in IEEE Transactions on Geoscience and Remote Sensing
Employing multimodal remote sensing images (MRSIs) enhances ground object identification, yet the heterogeneity of MRSIs can increase the risk of model overfitting, particularly when training samples are limited. Self-supervised methods, such as masked autoencoder (MAE) reconstruction or contrastive learning, efficiently extract features from MRSIs with minimal reliance on labeled samples. However, current methods have not adequately considered both the shared and complementary features of MRSIs during the pretraining process, limiting feature representation and classification performance. To address these limitations, we propose a self-supervised pretraining framework called shared and complementary information interaction enhancement network (SCIIENet) for hyperspectral images (HSIs) and light detection and ranging (LiDAR)/synthetic aperture radar (SAR) data joint classification. Specifically, the framework incorporates an decoupling-interaction-enhancement approach through a triple-branch MAE architecture. The triple-branch architecture seperately captures the spectral features of HSIs that provide complementary information relative to LiDAR/SAR data, as well as the spatial features of HSIs and LiDAR/SAR data. We also introduce an interaction enhancement strategy that utilizes hierarchical contrastive learning (HCL) to bridge significant gaps in MRSI to emphasize the shared spatial information of MRSIs and incorporates a two-stage cross-modal fusion encoder (TCFE) to integrate the shared and complementary spatial information of different modalities. Moreover, we propose multi-stage skip connections (MSCs) for the reconstruction, which effectively preserve complementary spatial information. Extensive experiments on four datasets show that our method significantly improves classification accuracy compared to several state-of-the-art methods.
18 December 2025 in IEEE Geoscience and Remote Sensing Letters
Weakly supervised semantic segmentation (WSSS) based on image-level labels significantly reduces the labeling burden. However, current mainstream approaches optimize solely using single-image information, neglecting the rich semantic correlation among images and struggling to dynamically suppress interfering information. When confronted with complex backgrounds and multicategory remote sensing (RS) images, intraclass consistency and interclass discrimination pose significant challenges. To address these challenges, this letter proposes the cross-image class token constraints network (CICTC-Net). CICTC-Net establishes semantic correlations across multicategory RS images and implements two modules for targeted optimization. Specifically, the cross-image token enhancement (CITE) module constructs intraclass token relationship graphs and applies cross-image consistency constraints to enhance semantic consistency among objects of the same category. The class-patch interaction refinement (CPIR) module constructs a directed graph of class-patch relationships and employs a neighborhood selection mechanism to refine class tokens, thereby enhancing interclass discriminability. Experiments on two RS datasets demonstrate that this approach significantly outperforms existing state-of-the-art solutions.