Unsupervised Change Detection (UCD) in Very High Resolution (VHR) Remote Sensing (RS) images remains to be a difficult challenge due to the inherent spatio-temporal complexity within data. Inspired by recent advancements in Visual Foundation Models (VFMs) and Contrastive Learning (CL), this research aims to develop CL methodologies to translate implicit knowledge in VFM into change representations, thus eliminating the need for explicit supervision. To this end, we introduce a Semantic-to-Change (S2C) learning framework for UCD in VHR RS images. Differently from existing CL methodologies that typically focus on learning multi-temporal similarities, we introduce a novel triplet learning strategy that explicitly models temporal differences, which are crucial to the CD task. Furthermore, random spatial and spectral perturbations are introduced during training to enhance robustness to temporal noise. In addition, a grid sparsity regularization is defined to suppress insignificant changes, and an IoU-matching algorithm is developed to refine the CD results. Experiments on three benchmark CD datasets demonstrate that the proposed S2C learning framework achieves significant improvements in accuracy, surpassing current state-of-the-art by over 31%, 9% and 23%, respectively. It also demonstrates robustness and sample efficiency, suitable for training and adaptation of various VFMs or backbone neural networks.

Implementing change detection (CD) with bi-temporal remote sensing images (RSIs) is essential for gaining insights into the dynamic evolution of the Earth’s surface. Recent advancements in deep learning methodologies have demonstrated considerable success in CD applications. However, the efficacy of supervised CD networks is significantly constrained by their dependence on high-quality change labels and precisely registered bi-temporal RSIs, which imposes significant limitations in real-world scenarios. To address this issue, we propose a novel single-temporal unsupervised CD method, STU-SAMI, which integrates the Segment Anything Model (SAM) with instance-level change generation. This method aims to facilitate the training of high-performance CD networks using unpaired and unlabeled single-temporal RSIs. The proposed approach comprises three main steps. First, the SAM is introduced to extract morphologically complete object instances from single-temporal RSIs. Second, an online instance-level change generation method is proposed, which transforms single-temporal RSIs into bi-temporal pseudo change samples based on the object instances obtained by SAM. Third, a streamlined and efficient deep Siamese CD network is constructed to support the training and inference processes. Extensive experiments conducted on three benchmark datasets demonstrate that the proposed method outperforms the existing state-of-the-art unsupervised CD methods. The code is available at https://github.com/IceStreams/STU-SAMI.

Recently, remote sensing images (RSIs) domain adaptation segmentation has been extensively studied. However, existing methods generally assume that source RSIs must be available, which is obviously an overly demanding condition and will increase unnecessary costs in practice. To this end, this letter takes the lead in exploring RSIs source-free adaptation segmentation, where only the offline model pretrained on the source domain and target RSIs are available. A novel method featuring prompt learning and vision foundation models is proposed, and the novelty design includes two aspects. First, to better adapt the general-purpose knowledge in the foundation model to different target RSIs, an attention-guided prompt tuning strategy is proposed, which can dynamically steer the knowledge at different layers and positions through prompts with different weights. Second, a feature alignment strategy with similarity distance is proposed for source-free domain adaptation by taking full advantage of the representation ability of the foundation model and the flexibility of prompt learning. Extensive experiments indicate that the performance of the proposed method is significantly superior to that of existing methods. Specifically, the mIoU of target RSIs has been improved by at least 3.14%~4.18%.

Multimodal land use/cover classification using optical and synthetic aperture radar (SAR) images has attracted significant attention because the unique radiation and geometric characteristics of these images provide complementary information regarding land properties. However, the significant differences between these modalities create a large semantic gap, posing challenges for effective feature fusion in multimodal learning. Moreover, missing modalities often occur in practical applications due to weather constraints or sensor malfunctions, posing challenges to achieving high performance in cross-modal learning. In this study, we proposed a multimodal online knowledge distillation (MMOKD) framework, designed for land use/cover classification of optical and SAR images using either full or missing modalities. This framework trains one modality-fusion network alongside two modality-specific networks in an end-to-end manner, facilitating both multimodal and cross-modal learning. More specifically, we developed a multimodal feature fusion (MFF) module for integrating heterogeneous features, and a single-modal feature generation (SFG) module for encapsulating cross-modal complementary information. Additionally, we proposed the joint distillation with multitype fusion knowledge (JD-MFK) method, guiding the modality-specific student networks to comprehensively learn the modality-fusion teacher network. Notably, we adopted an online distillation strategy for real-time feedback and synchronous updates of both modality-fusion and modality-specific networks. Finally, we conducted extensive experiments on two multimodal land use/classification datasets with advanced multimodal fusion, cross-modal distillation, and specific baseline networks for comparison. The results demonstrate the effectiveness of the proposed MMODD, which not only outperforms the other networks in both full- and missing-modality scenarios, but also significantly improves model training efficiency.

Hyperspectral image (HSI) classification has always been recognised as a difficult task. It is therefore a research hotspot in remote sensing image processing and analysis, and a number of studies have been conducted to better extract spectral and spatial features. This study aimed to track the variation of the spectrum in hyperspectral images from a sequential data perspective to obtain more distinguishable features. Based on the characteristics of optical flow, this study introduces an optical flow technique for the extraction of spectral flow that denotes the spectral variation and implements a dense optical flow extraction method based on deep matching. Lastly, the extracted spectral flow are combined with the original spectral features and input into a commonly used support vector machine (SVM) classifier to complete the classification. Extensive classification experiments on three benchmark HSI test sets show that the classification accuracy obtained by the spectral flow extracted in this study (SpectralFlow) is higher than traditional spatial feature extraction methods, texture feature extraction methods, and the latest deep-learning-based methods. Furthermore, the proposed method can produce finer classification thematic maps, thereby demonstrating strong practical application potential.

The existing hyperspectral image (HSI) classification encounters the obstacle of improving the classification accuracy with limited labeled samples. In this context, as a typical implementation of meta-learning, few-shot learning (FSL) makes the model learn by episodic training on source HSI, which has achieved significant improvements in small sample classification of target HSI. However, the existing FSL methods lack explicit consideration and exploration of the association between pixels, especially the intraclass association and interclass association between pixels in the support set and query set. To mitigate these issues, an FSL method based on edge-labeling graph neural network (FSL-EGNN) is proposed for small sample classification of HSI, which is the first attempt to explicitly quantify the associations between pixels by exploiting EGNN in HSI few-shot classification (FSC). Specifically, based on graph construction of HSI, episodic training is performed on the existing source HSI. During training, EGNN is used to predict the edge labels on the graph, thereby explicitly modeling the intraclass similarity and interclass dissimilarity between pixels of HSI. After the trained model is fine-tuned, it can realize FSC on the unseen target HSI. Experiments conducted on three benchmark HSI datasets demonstrate that the proposed FSL-EGNN outperforms the existing state-of-the-art methods with limited labeled samples.