2026· IEEE Transactions on Geoscience and Remote Sensing· Vol 64, pp. 4415716-4415716· 0 citations· 47 references
Abstract
Multimodal fusion methods have shown great potential in remote sensing image analysis, but existing approaches rely heavily on massive amounts of annotated data. This is not only costly and time-consuming but also prone to subjective bias. To address this issue, we propose a category-prior-based self-supervised framework, CGSNet, which uses category prior maps extracted from multispectral images as supervisory signals for end-to-end training. An adaptive confidence-weighted pseudo-label generation mechanism is designed to alleviate noise and errors in prior maps by replacing binary labels with continuous confidence maps, enabling the learning of uncertain interclass features. In addition, a multispectral feature-guided refinement strategy utilizes color and texture information to calibrate class transition regions and enhance the discriminative power of pseudo-labels in complex scenes. A dynamic mask selection strategy further enhances the model’s robustness and generalization capabilities through progressive learning. Experiments demonstrate that CGSNet achieves state-of-the-art performance without the need for human annotation, achieving an Mean Intersection over Union (mIoU) score of 78.46% on the Gaofen image dataset (GID) (vegetation) dataset and 79.58% on the Zurich (vegetation) dataset—12.22% and 15.07% higher than existing methods, respectively—while exhibiting strong cross-dataset zero-shot generalization capabilities. The code will be available at https://github.com/NUAALISILab
Semantic segmentation of high-resolution remote sensing images remains challenging due to complex spatial structures, multiscale object variations, fine-grained category differences, and high interclass similarities. Conventional segmentation methods usually rely on fixed convolutional heads or single feature representations, which makes it difficult to effectively model both intraclass appearance variations and interclass texture similarities, often leading to category confusion, missed objects, and incomplete segmentation in complex scenes. To address these challenges, we propose a state-aware prototype learning network, termed SAPLNet. Specifically, a cross-stage state refiner is introduced to progressively refine multilevel features by integrating the input features with the outputs of different stages through state-aware gated normalization. Then, a weighted feature pyramid decoder performs top-down fusion of the refined hierarchical features, combining high-level semantic information with low-level spatial details. Furthermore, a state-aware multiprototype classifier is designed to construct multiple semantic prototypes for each class via ground-truth-guided local class-center extraction and momentum-based prototype memory updating. A global state vector derived from the refined cross-stage features is used to adaptively modulate decoder features, improving the matching reliability between pixel features and class prototypes. In addition, prototype compactness loss, prototype diversity loss, and lightweight boundary loss are employed to enhance intraclass consistency, prototype discriminability, and boundary awareness. Experimental results demonstrate the effectiveness and superiority of SAPLNet.
Zeyu Zhao, Zhaolong Gao, Jun Feng· IEEE Journal of Selected Top...· 0 citations
Pixel-level annotation of remote sensing imagery is costly, motivating weakly supervised semantic segmentation (WSSS) using only image-level labels. However, class activation maps (CAMs) often highlight only discriminative sub-regions and fail to separate adjacent land-cover regions, particularly in remote sensing scenes characterized by densely co-occurring land-cover classes and substantial variations in object scale. To address these limitations, we propose a three-stage framework that integrates complementary priors from Contrastive Language–Image Pre-training (CLIP), Self-Distillation with No Labels version 2 (DINOv2), and the Segment Anything Model (SAM). First, a lightweight CLIP adapter aligns vision–language priors with remote sensing imagery, while sigmoid-based multi-label decoupled distillation replaces class-competitive distillation with independent class-wise supervision, producing more complete CAMs. Second, DINOv2-guided feature clustering decomposes large merged regions before SAM prompt generation, while Spatial–Semantic Constraints are used to construct confidence-guided point-and-box prompts and reject excessively expanded or semantically inconsistent masks, thereby generating reliable pseudo-labels. Finally, a compact segmentation network is initialized with the weights learned in Stage 1 and retrained using the refined pseudo-labels generated in Stage 2, eliminating the need for foundation models during inference. Experiments on the Potsdam, LoveDA, and DeepGlobe datasets show that the proposed method achieves mean intersection over union (mIoU) scores of 53.16%, 52.66%, and 62.98%, respectively, outperforming state-of-the-art WSSS baselines by 6.55, 1.16, and 1.27 percentage points, respectively. These results demonstrate the effectiveness and generalizability of the proposed framework across diverse remote sensing scenarios under image-level supervision.
A lightweight and efficient framework that integrates CLIP and DINO foundation models to address three challenges: semantic misalignment between generic text prompts and RSI-specific visuals; static CAM quality; and incomplete object coverage is proposed.
Xin Li, Nicola Genzano, M. Gianinetto et al.· ISPRS Annals of the Photogra...· 0 citations
This work advances WSSS for remote sensing images by addressing key limitations of CAM-based pseudo-labels generation, offering a cost-effective alternative to fully supervised approaches and facilitating broader applications in geographic information science and earth observation.
Jiaming Fan, Dali Chen, Yang Liu et al.· International Journal of Rem...· 0 citations
The remote sensing scene classification (RSSC) task plays a pivotal role in Earth observation missions, yet its progress remains constrained by the scarcity of high-quality labeled imagery. This article introduces a self-supervised learning (SSL) paradigm to address this challenge. First, for pseudo-label construction, a large set of long-interval satellite revisit imagery is collected and processed with pixel-level registration. The SIFT inliers retained during registration serve as saliency priors to guide asymmetric masking across views. This produces positive pairs that preserve global scene consistency while introducing controlled object-level ambiguities. Second, we propose a progressive layer-wise contrastive learning framework (MTC-Net) that couples the pseudo-label with the network’s representational hierarchy, forming a curriculum from local texture robustness to global semantic invariance. A dual-attention module with spatial–channel branches is further embedded to recalibrate intermediate features. The learning paradigm encourages the model to perform cross-view contextual reasoning rather than relying on pixel-wise correspondences. Experiments on three widely used datasets demonstrate that MTC-Net achieves competitive classification accuracy under limited-label settings, while ablation and visualization studies validate the effectiveness of establishing scene-level invariance through multi-temporal contrastive alignment.
Xiao Xiao, Han Zhang, Kenan Cheng et al.· Remote Sensing· 0 citations
Infrared imaging has unique advantages in remote sensing observation and non-contact measurement, but its inherent low contrast and blurred structural details limit the reliability of scene interpretation by human observers. Unlike deep learning-based approaches that rely on data-driven training and substantial computational resources, we propose a Hybrid Prior Enhanced Decomposition (HPED) model, a training-free, model-driven framework that incorporates structural and luminance priors into a multi-stage enhancement pipeline. An l1–l0-regularized decomposition separates the input into a base layer that preserves global structures and salient edges and a detail layer in which low-amplitude fluctuations and noise are suppressed. A prior-preserving bi-gamma correction method enhances base-layer contrast through prior-guided histogram segmentation and adaptive gray-level redistribution. An improved grayscale mapping strategy further enhances global contrast while maintaining interframe consistency. Experiments on real SWIR, MWIR, and LWIR images show that HPED ranks first among evaluated traditional and deep learning-based methods on key perceptual quality metrics (SSIM, VIF, LIF), while achieving over 25 fps on a CPU-only platform, sufficient for smooth real-time visual display. Task-oriented evaluation further shows that the HPED improves CNR and SCR by 174.7 ± 11.4% and 298.5 ± 52.1% on average over the raw input, outperforming all competing methods and suggesting potential applicability in downstream machine perception tasks such as detection and tracking.
Jie Li, Cheng Wang, Xiangyu Li et al.· Remote Sensing· 0 citations