Remote sensing images (RSIs) are frequently characterized by multi-scale inter-class objects and inconsistently distributed objects due to scene limitations, which would cause a significant data imbalance challenging the corresponding semantic segmentation. Recent methods have leveraged various deep learning techniques to capture high-quality representations for RSI semantic segmentation, but are hardly capable of addressing the afore-mentioned challenge given their limited explorations towards the mechanisms behind the representations. The recently discovered Neural Collapse (NC) phenomenon in computer vision models suggests the simplex equiangular tight frame (ETF) as the optimal representation structure, which has motivated us to observe that the optimal structure of last-layer representations is disrupted and inter-class representations for minor classes tend to become closer to each other beacuse of data imbalance. To address these issues, we propose Inter-class and Intra-class Neural Collapse Tuning (In2NeCT) to optimize the representations that satisfy the simplex ETF, which facilitates the discrimination of inter-class representations and the coherence of intra-class representations. Extensive experiments on three datasets demonstrate that our In2NeCT consistently leads to significant improvements in performance and outperforms the state-of-the-art methods.

Geographically weighted regression (GWR) facilitates spatial modeling by providing location-specific coefficients to capture spatial non-stationarity. GWR incorporates a distance decay effect, assigning greater weights to proximal observations under the assumption they exert more influence on the regression parameters. However, distant observations may share significant context similarities, such as socioeconomic or environmental factors, which can influence the regression model. This study introduces an attention-based architecture to address context similarity between samples. A deep learning model termed Context-Attention Geographically Weighted Regression (CatGWR) is proposed to integrate context similarity with distance-based proximity to enhance the estimation of spatial non-stationarity in spatial regression models. Such an integration results in contextualized spatial weights for CatGWR to identify the varying patterns of nonstationary relationships across different spatial locations and context conditions. Validation through simulation experiments and an empirical study on housing prices in Shenzhen, China, shows the superior predictive accuracy and robustness of CatGWR in modeling complex spatial interactions, especially under contextual influences, in which CatGWR improves the R2 of fit and prediction results by at least 6% compared to existing models. Future work will focus on optimizing bandwidth selection and exploring additional attention mechanisms to enhance model performance.

In recent years, the rapid advancement of multi-sensory platforms has significantly increased the availability of multisource remote sensing data, facilitating its systematic application to various tasks. The joint classification of hyperspectral images (HSIs) and light detection and ranging (LiDAR) data remains a critical research topic, with a key challenge being the effective extraction and integration of complementary information from multi-source remote sensing data. However, existing graph convolutional networks (GCNs)-based methods often fail to account for the heterogeneous topological relationships between HSI and LiDAR. Moreover, the discriminative power of HSI and LiDAR features extracted by existing methods is insufficient. In addition, existing methods are unable to fully exploit the rich self-supervised information present in local neighborhood. To address these limitations, we propose a heterogeneous contrastive graph fusion network (HCGFN) for the joint classification of HSI and LiDAR data. First, we propose a branch enhancement module to enhance the discriminative power of HSI and LiDAR. Second, a contrastive learning module is introduced to effectively align HSI and LiDAR representations. Finally, we propose a dynamic heterogeneous graph structure learning module to model heterogeneous relationship and achieve efficient interaction and effective fusion between HSI and LiDAR. The extensive experimental results on three benchmark datasets indicate the effectiveness of the proposed HCGFN compared with other state-of-the-art methods. Specifically, under limited training samples, the proposed HCGFN outperformed state-of-the-art methods in overall accuracy by 5.10%, 2.46%, and 8.79% on datasets Trento, MUUFL, and Houston2013, respectively.

Remote sensing images are usually characterized by complex backgrounds, scale and orientation variations, and large intraclass variance. General semantic segmentation methods usually fail to fully investigate the above issues, and thus their performances on remote sensing image segmentation are limited. In this article, we propose our LOGCAN++, a semantic segmentation model customized for remote sensing images, which is made up of a global class-aware (GCA) module and several local class-aware (LCA) modules. The GCA module captures global representations for class-level context modeling to reduce the interference of background noise. The LCA module generates local class representations as intermediate perceptual elements to indirectly associate pixels with the global class representations, targeting dealing with the large intraclass variance problem. In particular, we introduce affine transformations in the LCA module for adaptive extraction of local class representations to effectively tolerate scale and orientation variations in remote sensing images. Extensive experiments on three benchmark datasets show that our LOGCAN++ outperforms current mainstream general and remote sensing semantic segmentation methods and achieves a better trade-off between speed and accuracy.

Brightness temperature (TB) derived from micro- wave radiometers (MRMs) onboard China’s Chang’e (CE) satellites has provided significant insights into the Moon’s subsurface thermal conditions and evolution. However, conventional TB mapping techniques emphasize spatial correlations among observational data points while largely neglecting the influence of inherent lunar surface factors. In this study, we propose a novel TB estimation approach utilizing geographically neural network weighted regression (GNNWR) combined with multisource lunar remote sensing data to generate TB maps at a higher spatial resolution of$0.0625^{\circ } \times 0.0625^{\circ }$. This method integrates crucial lunar surface parameters in heat conduction and radiation transfer models in a new framework, thereby reducing the risk of overestimation associated with high-resolution targets in sparsely distributed samples. In addition, by replacing the traditional geographically weighted regression (GWR) kernel with a spatially weighted neural network (SWNN), the model effectively addresses spatial nonstationarity and heterogeneity present in TB data and microwave radiative transfer. Comparative analyses demonstrate that the GNNWR approach achieves superior performance, as evidenced by the highest$R^{2}$and the lowest mean absolute error (MAE), the mean absolute percentage error (MAPE), and the root mean square error (RMSE). Furthermore, the generated TB maps demonstrate strong alignment with observed spatial trends. These maps also reveal fine-scale thermal features typically obscured by conventional interpolation methods, enhancing their utility for microwave thermal emission analysis and geological studies.

Wireframe parsing aims to identify vectorized line segments as pairs of endpoints from an image. Conventional methods usually require field-specific knowledge for manual introduction of auxiliary processes or auxiliary learning tasks towards satisfactory performances. Such pipelines are, however, characterized by high complexity, insignificant efficiency, and limited space for further performance improvement. To address these issues, we propose WirePAuS, a novel Wireframe Parser with an Auxiliary-free Single-shot pipeline, which requires no auxiliary processes or auxiliary learning tasks. This is based on its capability to generate appropriate prior information from a prior-informed feature extractor, which incorporates frequency-domain and Hough-domain prior information on line segments in the backbone. Meanwhile, we devise a structurally compact pipeline that enables the parser to directly predict the focal midpoints as line objects and exploit their displacements for the corresponding endpoints. Our end-to-end trainable WirePAuS is capable to capture rich structural details during single-shot inference. Extensive experiments suggest that the proposed method reaches significantly improved performances for wireframe parsing and outperforms a series of state-of-the-art methods.

The estimation of spatial heterogeneity within real estate markets holds significant importance in house price modelling. However, employing a single or straightforward distance to measure spatial proximity is probably insufficient in complex urban areas, thereby resulting in an inadequate modelling of spatial heterogeneity. To address this issue, this paper incorporates multiple distance measures within a neural network framework to achieve an optimized measure of spatial proximity (OSP). Consequently, a geographically neural network weighted regression model with optimized measure of spatial proximity (osp-GNNWR) is devised for the purpose of spatially heterogeneous modeling. Trained as a unified model, osp-GNNWR obviates the need for separate pretraining of OSP. This enables OSP to delineate the modeled spatial process through a post hoc calculated value. Through simulation experiments and a real-world case study on house prices, the proposed model reaches more accurate descriptions of diverse spatial processes and exhibits better overall performance. The interpretable results of the case study in Wuhan demonstrate the efficacy of the osp-GNNWR model in addressing spatial heterogeneity within real estate markets, suggesting its potential for modelling and predicting complex geographical phenomena.

The spatial attention mechanism has been frequently employed for the semantic segmentation of remote sensing images, given its renowned capability to model long-range dependencies. As remote sensing images often exhibit intricate backgrounds, significant intraclass variability, and a foreground-background imbalance, spatial attention mechanism-based methods somehow tend to introduce an extensive amount of background context through intensive affinity operations, causing unsatisfactory segmentation outcomes. While several class-aware methods attempt to attenuate the interference of background context by generating class representations as representative features, they still encounter challenges related to independent correlation calculation and single-confidence scale class representations. We introduce a dual-domain optimized class-aware network designed to address these challenges. In the semantic domain, we use category confidence as a scaling criterion to derive class representations at multiple confidence scales, effectively modeling pixel-class relationships. In the spatial domain, we leverage pixel-class relationships and their consensus to enhance relevant correlations while suppressing erroneous ones. Experimental results on three datasets demonstrate that the proposed method surpasses previous state-of-the-art ones for remote sensing image segmentation. Code is available athttps://github.com/xwmaxwma/rssegmentation.

Graph convolutional networks (GCNs) have recently gained prominence in hyperspectral images (HSIs) classification tasks given their superior performance on non-Euclidean data. However, GCN-based methods are heavily reliant on complete graph structural information, which can cause the aggregation and transmission of information across nodes from differing classes, thereby compromising the classification performance. Furthermore, the scarcity of labeled pixels in HSIs often limits the representational capability of such methods. To address these issues, we propose an aggregative and contrastive dual-view graph attention network (ACoD-GAT) for HSI classification. Specifically, we present a progressive aggregation module, including a pixel clustering submodule and a node aggregation submodule to exploit semantic information at various levels. Besides, we integrate multiscale manipulation with a diffusion matrix to construct the dual view to further extract semantic information from both local and global perspectives. Moreover, we design an unsupervised contrastive loss function and a supervised contrastive loss function to facilitate contrastive learning on the dual view, improving the representational capabilities of ACoD-GAT with very few labeled samples. The extensive experimental results on four benchmark datasets demonstrate the superiority of the proposed ACoD-GAT compared with other state-of-the-art methods.

Conventional deep learning-based methods for single remote sensing image super-resolution (SRSISR) have made remarkable progress. However, the super-resolution (SR) outputs of these methods are yet to become sufficiently satisfactory in visual quality. Recent diffusion model-based generative deep learning models are capable to enhance the visual quality of output images, but this capability is limited due to their sampling efficiency. In this article, we propose FastDiffSR, an SRSISR method based on a conditional diffusion model. Specifically, we devise a novel sampling strategy to reduce the number of sampling steps required by the diffusion model while ensuring the sampling quality. Meanwhile, the residual image is adopted to reduce computational costs, demonstrating that integrating channel attention and spatial attention begets a further improvement in the visual quality of output images. Compared to the state-of-the-art (SOTA) convolutional neural network (CNN)-based, GAN-based, and Transformer-based SR methods, our FastDiffSR improves the learned perceptual image patch similarity (LPIPS) by 0.1–0.2 and achieves better visual results in some real-world scenes. Compared with existing diffusion-based SR methods, our FastDiffSR achieves significant improvements in pixel-level evaluation metric peak signal-noise ratio (PSNR) while having smaller model parameters and obtaining better SR results on Vaihingen data with faster inference time by 2.8–28 times, showing excellent generalization ability and time efficiency. Our code will be open source athttps://github.com/Meng-333/FastDiffSR.

Super-resolution (SR) methods have significantly contributed to the improvement of the spatial resolution of remote sensing (RS) images. The development of deep learning empowers novel methods to learn informative feature representation from massive low-resolution (LR) and high-resolution (HR) image pairs. Conventional RS image SR methods, however, may fail in large-scale ($\times 8$and$\times 9$) SR tasks. Specifically, a larger scale factor corresponds to less information in LR images, which is a considerable challenge to SR. To address the issue, we propose a novel method for single RS image SR (SRSISR) based on stratified dense sampling to effectively extract image features. Specifically, the proposed SR dense-sampling residual attention network (SRDSRAN) combines dense sampling and residual learning to improve multilevel feature fusion and gradient propagation and employs local and global attentions to learn important features and long-range interdependence in the channel and spatial dimensions. Meanwhile, we also devise a discriminator model using local and global attentions and with the loss function integrating${L}_{1}$pixel loss,${L}_{1}$perceptual loss, and relativistic adversarial loss to obtain the perceptually realistic images. Besides, we introduce a chain training to promote performance and expedite the training process for large-scale SR. Experimental results on UC Merced image and other multispectral data demonstrated that our SRDSRAN outperformed the current state-of-the-art methods quantitatively and in visual quality and obtained a higher classification accuracy in scene classification, proving its potential for applications with other downstream tasks. The code of SRADSGAN will be available athttps://github.com/Meng-333/SRADSGAN.

Semantic segmentation is one of the most significant tasks in remote sensing (RS) image interpretation, which focuses on learning global and local information to infer the semantic label of each pixel. Previous studies devise encoder-decoder structured deep learning (DL) models to extract global and local features from RS images with the help of pretraining knowledge to predict semantic labels. However, due to the common heterogeneity between the data for pretraining and the data to be semantically segmented, these models fail to learn general features appropriate to RS datasets. In this article, we propose a novel formulation of the above problem from a causal perspective, where the learned features from pretrained models result from causality and spurious correlations, and only the former carries general information that remains invariant regardless of the exact task and dataset. Based on the above formulation, we propose stepwise intervention and reweighting (SIR). It can reduce the confounding bias introduced by the pretraining knowledge and improve the model’s ability to learn general features, making semantic segmentation of RS images benefit more from pretraining. Besides, we conduct a detailed theoretical analysis of our methods and conduct extensive experiments on two widely used public RS datasets. Experimental results demonstrate that applying SIR to encoder-decoder semantic segmentation models achieves performance improvements, proving the effectiveness and application values of the proposed method.

Downscaling nighttime light (NTL) from satellite imagery presents valuable applications at a more detailed spatial scale, especially in the realms of urban expansion and socio-economic assessment. Nevertheless, due to the complexity of geographical conditions and uncertainties in the relationships among multiple factors, the precision of NTL downscaling often encounters constraints. In this work, an incorporated multifactor geographically neural network weighted regression (MF-GNNWR) NTL downscaling framework is proposed to solve the spatial nonstationarity in high-heterogeneous urban areas, which mainly uses geographically neural network weighted regression (GNNWR) combined with multiple factors including surface physical characteristics, socio-economic attributes, and human activities to improve the accuracy of NTL, particularly in urban regions with complicated land cover. The findings illustrate that the MF-GNNWR framework displays finer downscaling accuracy on different land cover, effectively enhancing data quality. Notably, our findings underscore the pronounced influence of socio-economic and human activity factors on NTL downscaling. Comparative analysis against several alternative downscaling methodologies reveals that the MF-GNNWR framework outperforms them, exhibiting a remarkable 23.10% improvement in the Pearson correlation coefficient (r) and achieving a root-mean-square error (RMSE) of$16.95~\text {{nW/c}{m}}^{2}{/\text {sr}}$, and after residual compensation, r continue s to increase by 1.5%, while RMSE decreases by$0.157~\text {nW/cm}^{2}{/\text {sr}}$. These findings highlight the efficacy of the proposed framework in downscaling NTL, underscoring its advantages and practical utility.

Geographically weighted regression (GWR) is a classical method of modeling spatially non-stationary relationships. The geographically neural network weighted regression (GNNWR) model solves the problem of the inaccurate construction of spatial weight kernels using a spatially weighted neural network. However, when the spatial distribution of observations is uneven, the spatial proximity expression in the input of GWR and GNNWR models does not fully represent the impact of the whole research space on the estimating point. Therefore, we established a global spatial proximity grid (GSPG) to express the spatial proximity of each estimating point and proposed a spatially weighted convolutional neural network (SWCNN) to extract the relationship between the GSPG and spatial weights. Finally, we proposed a geographically convolutional neural network weighted regression (GCNNWR) model combining SWCNN and ordinary linear regression (OLR) model to estimate spatial non-stationarity. We used two case studies of simulated data and real environment data to demonstrate the advancements of the GCNNWR model. The GCNNWR model achieved higher estimation accuracy and greater predictive power than the OLR, GWR, multi-scale GWR (MGWR), and GNNWR models. Moreover, the GCNNWR model maintained its better stability and accuracy in estimating spatially non-stationary relationships when the distribution of observations was uneven.

Geographically weighted regression (GWR) and geographically and temporally weighted regression (GTWR) are classic methods for estimating non-stationary relationships. Although these methods have been widely used in geographical modeling and spatiotemporal analysis, they face challenges in adequately expressing space-time proximity and constructing a kernel with optimal weights. This probably results in an insufficient estimation of spatiotemporal non-stationarity. To address complex non-linear interactions between time and space, a spatiotemporal proximity neural network (STPNN) is proposed in this paper to accurately generate space-time distance. A geographically and temporally neural network weighted regression (GTNNWR) model that extends geographically neural network weighted regression (GNNWR) with the proposed STPNN is then developed to effectively model spatiotemporal non-stationary relationships. To examine its performance, we conducted two case studies of simulated datasets and environmental modeling in coastal areas of Zhejiang, China. The GTNNWR model was fully evaluated by comparing with ordinary linear regression (OLR), GWR, GNNWR, and GTWR models. The results demonstrated that GTNNWR not only achieved the best fitting and prediction performance but also exactly quantified spatiotemporal non-stationary relationships. Further, GTNNWR has the potential to handle complex spatiotemporal non-stationarity in various geographical processes and environmental phenomena.

Geographically weighted regression (GWR) is a classic and widely used approach to model spatial non-stationarity. However, the approach makes no precise expressions of its weighting kernels and is insufficient to estimate complex geographical processes. To resolve these problems, we proposed a geographically neural network weighted regression (GNNWR) model that combines ordinary least squares (OLS) and neural networks to estimate spatial non-stationarity based on a concept similar to GWR. Specifically, we designed a spatially weighted neural network (SWNN) to represent the nonstationary weight matrix in GNNWR and developed two case studies to examine the effectiveness of GNNWR. The first case used simulated datasets, and the second case, environmental observations from the coastal areas of Zhejiang. The results showed that GNNWR achieved better fitting accuracy and more adequate prediction than OLS and GWR. In addition, GNNWR is applicable to addressing spatial non-stationarity in various domains with complex geographical processes.

Spatiotemporal kriging (STK) is recognized as a fundamental space-time prediction method in geo-statistics. Spatiotemporal regression kriging (STRK), which combines space-time regression with STK of the regression residuals, is widely used in various fields, due to its ability to take into account both the external covariate information and spatiotemporal autocorrelation in the sample data. To handle the spatiotemporal non-stationary relationship in the trend component of STRK, this paper extends conventional STRK to incorporate it with an improved geographically and temporally weighted regression (I-GTWR) model. A new geo-statistical model, named geographically and temporally weighted regression spatiotemporal kriging (GTWR-STK), is proposed based on the decomposition of deterministic trend and stochastic residual components. To assess the efficacy of our method, a case study of chlorophyll-a (Chl-a) prediction in the coastal areas of Zhejiang, China, for the years 2002 to 2015 was carried out. The results show that the presented method generated reliable results that outperform the GTWR, geographically and temporally weighted regression kriging (GTWR-K) and spatiotemporal ordinary kriging (STOK) models. In addition, employing the optimal spatiotemporal distance obtained by I-GTWR calibration to fit the spatiotemporal variograms of residual mapping is confirmed to be feasible, and it considerably simplifies the residual estimation of STK interpolation.