Integrating ground-level geospatial data with rich geographic context, like OpenStreetMap (OSM), into remote sensing (RS) foundation models (FMs) is essential for advancing geospatial intelligence and supporting a broad spectrum of tasks. However, modality gap between RS and OSM data, including differences in data structure, content, and spatial granularity, makes effective synergy highly challenging, and most existing RS FMs focus on imagery alone. To this end, this study presents GeoLink, a multimodal framework that leverages OSM data to enhance RS FM during both the pretraining and downstream task stages. Specifically, GeoLink enhances RS self-supervised pretraining using multi-granularity learning signals derived from OSM data, guided by cross-modal spatial correlations for information interaction and collaboration. It also introduces image mask-reconstruction to enable sparse input for efficient pretraining. For downstream tasks, GeoLink generates both unimodal and multimodal fine-grained encodings to support a wide range of applications, from common RS interpretation tasks like land cover classification to more comprehensive geographic tasks like urban function zone mapping. Extensive experiments show that incorporating OSM data during pretraining enhances the performance of the RS image encoder, while fusing RS and OSM data in downstream tasks improves the FM’s adaptability to complex geographic scenarios. These results underscore the potential of multimodal synergy in advancing high-level geospatial artificial intelligence. Moreover, we find that spatial correlation plays a crucial role in enabling effective multimodal geospatial data integration. Code, checkpoints, and using examples are released at [GitHub](https://github.com/bailubin/GeoLink_NeurIPS2025).

Building height is a key metric in transitioning urban analysis from two-dimensional to three-dimensional perspectives. It serves as a fundamental indicator for assessing urban development, population density, and energy consumption. Regardless of whether machine learning or deep learning methods are used, accurate height estimation inevitably relies on large quantities of labeled building height samples for model training. However, existing reference datasets often suffer from ambiguity due to floor-to-height conversions and temporal obsolescence. The advent of spaceborne light detection and ranging (LiDAR) provides a promising avenue to overcome these limitations, enabling the acquisition of large-scale, high-precision, and time-sensitive reference height data. Moreover, while buildings are inherently individual and discrete units, most existing height estimation studies operate at gridded scales, thereby obscuring inter-building height variability. In this study, we propose a novel approach to estimate individual building heights by integrating spaceborne LiDAR with multisource remote sensing data. First, leveraging building footprint data, we derive reference height samples by jointly retrieving building heights from the Ice, Cloud, and land Elevation Satellite-2 (ICESat-2) and the Global Ecosystem Dynamics Investigation (GEDI). Then, we construct a multimodal feature set by extracting seasonal features from a combination of multisource datasets, such as Sentinel-1, Sentinel-2, and SDGSAT-1. To model the spatially complex height distribution, we propose the Neighborhood-aware CNN-Transformer Network (NeCT-Net), which estimates building heights at the pixel level. To address the scarcity of tall building samples, we incorporate a semi-supervised learning paradigm. Specifically, pseudo-labels for high-rise buildings are generated using predictions from a teacher model and iteratively refined through self-training, thereby enhancing the student model’s learning capacity in tall-structure scenarios. Finally, we propose a post-processing approach that constrains height predictions using building morphological information, enabling the conversion from pixel-level predictions to vector-based building-level estimations. We apply the proposed method to two iconic urban areas—Beijing and Seattle. The results demonstrate strong predictive performance, with RMSE, MAE, and R values of 9.01 m, 6.08 m, and 0.93 for Beijing, and 8.72 m, 5.31 m, and 0.79 for Seattle. Compared to baseline methods, our approach reduces RMSE by 29.66% in Beijing and 7.43% in Seattle, while improving R by 47.62% and 46.06%. When benchmarked against existing individual building height products, our model achieves RMSE reductions of 31.43% and 21.16%, and R improvements of 57.63% and 23.44% in the two cities. These results highlight the methodological advancement and robustness of our approach. This work presents a new paradigm for large-scale, individual building height estimation through the integration of spaceborne LiDAR and multisource remote sensing data. The source code will be made available at https://github.com/Lunar-Elf/NeCT-Net.

Urban land use (ULU) undergoes continuous and dynamic changes, profoundly reshaping the spatial configuration of urban areas. Accurately quantifying these changes is critical for the precise planning and efficient management of limited urban land resources during urbanization. However, existing change detection methods struggle to address the unique challenges of ULU change detection, leading to a lack of reliable ULU change maps. Two major challenges hinder the study of ULU changes: (1) high internal heterogeneity, which obscures the main features of ULU, and (2) inaccurate boundary delineation. To overcome these challenges, we propose an object-based method for detecting ULU changes. Our approach employs a Siamese neural network equipped with a self-attention mechanism to extract global features, a Fusion module to integrate global and local features, and a multi-scale Amplifier to filter and enhance the main features. Subsequently, a hierarchical clustering framework aggregates the segmented objects with the main features into coherent ULU units. Experimental results demonstrate that the proposed method effectively addresses the two major challenges of ULU change detection. The proposed method achieves the best performance compared to representative methods in Beijing and Shanghai, two of China’s largest and most typical cities. Furthermore, the analysis of ULU change results reveals that both cities experienced rapid outward urban expansion between 2015 and 2020.

High-resolution (HR) satellite imagery over long time series is crucial for monitoring fine-scale land use and land cover changes. However, existing freely available Earth observation data are limited to medium to coarse resolutions or short time spans. Our goal is to reconstruct 10-m resolution Sentinel-2-like imagery from 1985 to 2015 using Landsat imagery. Unlike previous studies that heavily rely on HR reference images from the same period, we utilize the overlapping Landsat and Sentinel-2 observations to reconstruct 10-m Sentinel-2-like imagery from 1985 to 2015. Specifically, we propose the STARS-Net, which can effectively transfer multiple similar textures from Sentinel-2 to Landsat images, ensuring accurate reconstruction even with significant land changes. The innovation of our method lies in its ability to integrate any number of HR images, without limiting these images to similar times or the same locations, leading to more flexible and accurate reconstructions. As a result, a 10-m resolution dataset of five major China cities from 1985 to 2015 is available athttps://figshare.com/s/186e9b25426067e0dcf4with a five-year interval. Extensive comparative experiments demonstrate that STARS-Net outperforms state-of-the-art super-resolution (SR) methods and the results highlight that STARS-Net exhibits strong robustness in various scenarios, including cloud cover, image missing, and land cover changes. More importantly, our method produces the first historical image retrospect at 10-m resolution, which has a finer resolution compared to Landsat imagery and an important time extend compared with the Sentinel-2 data. Therefore, our dataset can help to facilitate a deeper understanding of terrestrial transformations over the past four decades.

As a method that can incorporate unlabeled data into model training, semi-supervised semantic segmentation (SSS) can mitigate the burden of manual annotation in geographic mapping tasks with very-high-resolution (VHR) images. Various SSS approaches have been proposed for VHR images, but most of them either rely on complex models or additional training procedures, increasing the computation cost. In this study, we propose a simple-yet-effective SSS approach named AP-semi for VHR images, which neither increases the number of learnable parameters nor introduces additional training steps. First, since data augmentation can influence the training data and process significantly, we refine the traditional CutMix transformation, an important data augmentation strategy, by generating adaptively sized cut boxes, optimizing the intensity of perturbation and making it better suited for the training phase. Second, in order to make all the unlabeled samples contribute to the model training, we design a prototype-based feature-level consistency loss, which can enforce the consistency of model predictions in the feature space. By combining the adaptive data augmentation strategy and prototype-based loss, AP-semi can achieve impressive results with a limited number of labeled samples, surpassing several baseline models in extensive experiments. Through experiments, we find that AP-semi can adapt to both convolutional neural networks (CNNs)- and transformer-based segmentation models. In the simulation experiment mimicking the routine operations in geographic mapping tasks, AP-semi demonstrates significant time-saving benefits.

Global contextual information is crucial for the semantic segmentation of remote sensing (RS) images. However, the majority of current approaches depend on convolutional neural networks (CNNs). Due to the local receptive fields inherent in convolutional operations, these networks typically capture image features within limited areas and struggle to comprehend broader contextual information in the images. In this study, a dual-branch encoding approach, DESformer, is proposed, integrating transformers with CNN, to effectively capture global multiscale context information and enhance edge feature extraction. In addition, DESformer incorporates a feature interaction module (FIM) to combine local features with global representations extracted by transformers and CNN, respectively, across different resolutions. This approach enhances the capability to capture local features in RS images and improves the understanding of extensive spatial relationships. Subsequently, we employ a novel top-down approach for global supervision of the traditional feature pyramid multilevel visual integration (MVI) module, by harnessing the clear visual center information obtained from the deepest internal features. To successfully concentrate on important information and preserve sensitivity to features at various scales, the preceding shallow features are muted. In addition, FIAB-Loss, a loss function is introduced, combining a focal loss with IOU and active boundary loss (ABL). This composite loss function strengthens the model’s focus on challenging-to-distinguish categories. Extensive experiments conducted on three datasets, including the semantic segmentation of lakes in the Tibetan Plateau and the ISPRS’s Vaihingen benchmark, validate the efficacy of the proposed method. The experimental results indicate that the network exhibits exceptional performance in processing VHR images and accurately extracting edge features.

Automatic change detection of open-pit mines from high-resolution remote sensing images is of great significance for the mining and management of mineral resources. For this purpose, we propose a siamese multiscale change detection network (SMCDNet) with an encoder-decoder structure. First, the multiscale low-level and high-level features of the bi-temporal image are extracted by a siamese network. Second, a multilevel feature absolute difference (MFAD) module is proposed to fuse the low-level and high-level change features. Finally, convolution and up-sampling operations are used to recover the details of the changed areas. A self-made open-pit mine change detection (OMCD) dataset is employed to conduct experiments. Experimental results have demonstrated that the proposed method is superior to the comparison networks.$F1$- score of 88.13% is achieved by the proposed SMCDNet. The OMCD dataset produced in this study has been made public at the following link:https://figshare.com/s/ae4e8c808b67543d41e9.

Land uses (e.g., commercial, residential, and industrial lands) and functional spaces (e.g., living, productive, and ecological spaces) are two-level landscape patches and totally work as basic units for urban planning. The two-level patches are interrelated and mutually binding, but existing mapping methods extracted them separately, leading to substantial conflicts and errors in their mapping results. Accordingly, this study proposes a synergistic classification of multilevel land patches (SC-MLPs). It considers a multitask learning strategy and proposes a novel correlation loss function to measure the correlations between land uses and functional spaces, which is expected to resolve conflicts and improve the accuracy of the two-level land patch mapping results. Consequently, land-use and functional-space maps of three major Chinese cities are generated, which generally have a high resolution of 2 m and high overall accuracies of 90.1% for land uses and 93.8% for functional spaces. Compared to state-of-the-art land-use and functional-space mapping methods, our results have not only higher accuracies but also a better consistency which is improved by 36%. Accordingly, the proposed SC-MLP can generate not only accurate but also consistent maps of land uses and functional spaces, which plays a fundamental role in land system research and urban planning.

Although semantic segmentation models based on deep neural networks (DNNs) have achieved excellent results, generalizing well from one remote sensing dataset (source domain) to another dataset with different acquisition conditions (target domain) remains a major challenge. Many domain adaptation (DA) approaches have been proposed to address this problem. DA aims to help DNNs learn a generalizable representation space in which source and target domains have similar feature distributions, but most of the existing DA approaches have difficulty in aligning the high-dimensional image representations of two domains directly. In this study, we proposed a model integrating contrastive learning and adversarial learning in a unified framework for aligning two domains in both representation space and spatial layout. Specifically, the model consists of a semantic segmentation network for feature extraction and two branches for DA. The first branch is used for adaptation in representation space directly by a proposed pixelwise contrastive loss, while the second branch is used for adaptation in predicted results to help two domains have similar spatial layouts through a novel but simple entropy-based similarity discriminator. Additionally, a training strategy called category similarity matching sampling was proposed to provide source and target image pairs with similar category composition for each training iteration, which can help the two branches work better. Extensive experiments indicated that the two branches can benefit each other to gain a superior performance and DA pretraining by our methods can achieve impressive results with only a small number of target labeled samples.

Due to the spectral diversity of objects within the same classes and the spectral similarity between different classes, classical fuzzy classification methods perform poorly in multispectral remote sensing image classification. The basic reason is that they often use a single spectral curve to represent a land cover type while ignoring spectral diversity characteristics. To solve this issue, this article proposes a novel multicenter supervised fuzzy classification (MCSFC) method for modeling spectral diversity in multispectral remote sensing data. Images are first split into clusters with similar sizes or volumes, namely, granularities, by a hierarchical clustering process. Second, the granularities are labeled by training samples. As a result, the centers of the granularities corresponding to one type of sample can represent the spectral diversity within one class and, thus, are treated as the multiple centers of a land cover type. The membership degree of each unlabeled sample belonging to any land cover type is determined by the shortest distance to the centers of each type, and the spectral diversity is considered in this stage. A comparison reveals that the proposed method clearly improves the classification performance of various single-center fuzzy semisupervised clustering (FSSC) methods. Two case studies indicate that the granularity volume parameter notably affects the classification performance, and the overall accuracy (OA) decreases with increasing granularity volume.

This study proposes a land cover classification method using remote sensing images and LiDAR data. In this method, deep learning and conditional random fields (CRF) optimization is applied for accurately classifying land covers. To address the issue that pixel-based deep learning methods are difficult to capture the precise outline of ground objects, we combine deep feature learning strategy with image objects for accurately interpreting remote sensing images. Context information revealing relationships between image objects are explored for optimizing the classification result by using object-based CRF, where height information derived from LiDAR data is considered. Vaihingen dataset is used to validate the proposed method and an overall classification accuracy of 92.4% is achieved.

Satellite image time series has been broadly used to map land covers due to the repetitive measurements ability and rich spatiotemporal information it contained. However, existing studies in SITS tended to only express and employ one or two aspects of spectral, spatial, and temporal features without fully utilizing all the three information, leading to the unsatisfactory results of SITS classification. Accordingly, this study aim at obtaining land cover information base on classification of SITS with integratively expressing and employing spectral, spatial and temporal features. Moreover, results of different combinations of the three kinds of features are evaluated and compared to infer the effect and importance of features, and find the optimal features for SITS classification. Experiments carried out with this methodology show the effectiveness of this approach, potentially contributing to the feature extracted and classification patterns in SITS.

Urban functional zones (e.g. commercial, residential, and industrial) are basic units for city planning and management, and play an important role in city studies. However, functional zones are difficult to extract from very-high-resolution (VHR) remote sensing images, as they are various in components, sizes, and heterogeneities, leading to different segmentation scales. To resolve this issue, this study uses selfhood scale, a local optimum scale, to extract functional zones. Firstly, geoscene segmentation is used to delineate functional zones at multiple scales. Then, selfhood scales are calculated to measure the local optimum scales of segmenting functional zones, based on which multiscale segmentation results can be finally assembled into one layer to generate functional-zone boundaries. The experimental results indicate this method that is effective to delineate functional zones in Beijing, adapting to local built environments.

For hyperspectral remote sensing image (HSI) classification, the learning process of deep neural networks has been progressively advanced in depth, but the fine features are often largely lost or even disappear in the process of depth transfer. With the increase in feature aggregation and connectivity, the complexity of the network and the training parameters increases greatly, requiring more training time. This paper proposed a multi-scale dense network (MSDN) for HSI classification that made full use of different scale information in the network structure and combined scale information throughout the network. It implemented feature extraction of HSIs in two dimensions, including the features at fine and coarse levels. In the horizontal direction, it considered the deep extraction of HSI features, and the 3-D dense connection structure was used for aggregating features at different levels. In the vertical direction, scale information was considered, and three-scale feature maps at low, middle, and high levels were generated based on the first layer of the network. The MSDN used stride convolution for downsampling and combined feature information at different scale levels. The MSDN extracted features along the diagonal line. The network implemented the reconstruction of deep feature extraction and multi-scale fusion for HSI classification. The MSDN model performed well on representative HSI datasets, namely, the Indian Pines, Pavia University, Salinas, Botswana, and Kennedy Space Center datasets. It improved the training speed and accuracy for HSI classification and especially improved the convergence speed, which effectively saved computing resources and had high stability.

The exponential growth of natural language text data in social media has contributed a rich data source for geographic information. However, incorporating such data source for GIS analysis faces tremendous challenges as existing GIS data tend to be geometry based while natural language text data tend to rely on natural language spatial relation (NLSR) terms. To alleviate this problem, one critical step is to translate geometric configurations into NLSR terms, but existing methods to date (e.g. mean value or decision tree algorithm) are insufficient to obtain a precise translation. This study addresses this issue by adopting the random forest (RF) algorithm to automatically learn a robust mapping model from a large number of samples and to evaluate the importance of each variable for each NLSR term. Because the semantic similarity of the collected terms reduces the classification accuracy, different grouping schemes of NLSR terms are used, with their influences on classification results being evaluated. The experiment results demonstrate that the learned model can accurately transform geometric configurations into NLSR terms, and that recognizing different groups of terms require different sets of variables. More importantly, the results of variable importance evaluation indicate that the importance of topology types determined by the 9-intersection model is weaker than metric variables in defining NLSR terms, which contrasts to the assertion of ‘topology matters, metric refines’ in existing studies.

Building patterns exhibited collectively by a group of buildings are fundamental to understanding urban forms, classifying urban scenes, analyzing urban landscapes, and generalizing maps. The existing studies have used geometric homogeneity or regularity to represent and discover limited patterns for map generalization, or used interval and rectangle algebra to represent relations between spatial objects. These approaches, however, cannot illustrate how patterns are produced by using syntax or grammar (i.e. relations between buildings) to link words (i.e. buildings) into sentences (i.e. building patterns), making it impossible to represent and discover building patterns with diverse structures. This study presents a relation-based approach to formalize and discover arbitrary building patterns at three abstract levels. At the bottom level, a relative and local frame of reference is defined, and 169 basic relations are derived to represent relative positions between buildings. At the middle level, the 169 relations, qualitative angle description, and qualitative size are combined to formalize important semantic relations between two buildings, which include collinear, perpendicular, and parallel relations. At the top level, the relations at the bottom and middle levels are used to formalize three types of building patterns, including collinear patterns, the structured patterns with acceptable names, and other patterns of interest. Algorithms implementing the three levels of relations are presented and applied to demonstrate the effectiveness of the proposed approach in discovering building patterns from databases and querying building patterns. The results indicate that the relational approach is generic to effectively represent and discover building patterns with arbitrary structures. In addition, it complements the existing geometric methods for recognizing building patterns, and the interval and rectangle algebra for representing building relations.

In this paper, we propose a spectral–spatial feature based classification (SSFC) framework that jointly uses dimension reduction and deep learning techniques for spectral and spatial feature extraction, respectively. In this framework, a balanced local discriminant embedding algorithm is proposed for spectral feature extraction from high-dimensional hyperspectral data sets. In the meantime, convolutional neural network is utilized to automatically find spatial-related features at high levels. Then, the fusion feature is extracted by stacking spectral and spatial features together. Finally, the multiple-feature-based classifier is trained for image classification. Experimental results on well-known hyperspectral data sets show that the proposed SSFC method outperforms other commonly used methods for hyperspectral image classification.

Qualitative knowledge representation of spatial locations and relations is popular in many text-based media, for example, postings on social networks, news reports, and encyclopedia, as representing qualitative spatial locations is indispensable to infer spatial knowledge from them. However, an integrative model capable of handling direction-based locations of various spatial objects is missing. This study presents an integrative representation and inference framework about direction-based qualitative locations for points, lines, and polygons. In the framework, direction partitions of different types of reference objects are first unified to create a partition consisting of cells, segments, and corners. They serve as a frame of reference to locate spatial objects (e.g., points, lines, and polygons). Qualitative relations are then defined to relate spatial objects to the elements in a cell partition, and to form the model of qualitative locations. Last, based on the integrative representation, location-based reasoning mechanism is presented to derive topological relations between objects from their locations, such as point–point, line–line, point–line, point–polygon, line–polygon, and polygon–polygon relations. The presented model can locate any type of spatial objects in a frame of reference composed of points, lines, and polygons, and derive topological relations between any pairs of objects from the locations in a unified method.

This study wants to demonstrate that two different polarimetric target decomposition methods can improve SAR data accuracy for estimating the parameters of bare soil surface. To achieve this goal, two experiments are conducted: (1) both Freeman and Cloude decomposition methods are performed on JPL/AIRSAR L-band fully polarimetric data; and (2) Advanced Integral Equation Model (AIEM) is used to simulate backscatting coefficients. The root mean square errors (RMSEs) of σ0hh, σ0vvbetween original data and AIEM simulated data are 1.96 and 1.25 dB. However, if Cloude method is used to decompose original data, the RMSEs will be reduced to 1.45 and 1.14dB, respectively; for Freeman method, the RMSEs are 1.64 and 1.35 dB. Therefore, polarimetric target decomposition compensation, especially Cloude method, can help to improve the accuracy of SAR data for estimating the parameters of bare soil surface.

Multi-resolution spatial data always contain the inconsistencies of topological, directional, and metric relations due to measurement methods, data acquisition approaches, and map generalization algorithms. Therefore, checking these inconsistencies is critical for maintaining the integrity of multi-resolution or multi-source spatial data. To date, research has focused on the topological consistency, while the directional consistency at different resolutions has been largely overlooked. In this study we developed computation methods to derive the direction relations between coarse spatial objects from the relations between detailed objects. Then, the consistency of direction relations at different resolutions can be evaluated by checking whether the derived relations are compatible with the relations computed from the coarse objects in multi-resolution spatial data. The methods in this study modeled explicitly the scale effects of direction relations induced by the map generalization operator – merging, thus they are efficient for evaluating consistency. The directional consistency is an essential complement to topological and object-based consistencies.

The scale dependences of topological relations are caused by the changes of spatial objects at different scales, which are induced by the reduction of attributes. Generally, the detailed partitions and multi-scale attributes are stored in spatial databases, while the coarse partitions are not. Consequently, the detailed topological relations can be computed and regarded as known information, while the coarse relations stay unknown. However, many applications (e.g., multi-scale spatial data query) need to deal with the topological relations at multiple scales. In this study new methods are proposed to model and derive the scale dependences of topological relations between lines and multi-scale region partitions. The scale dependences of topological relations are modeled and used to derive the relations between lines and coarse partitions from the relations about the detailed partitions. The derivation can be performed in two steps. At the first step, the topological dependences between a line and two meeting, covered and contained regions are computed and stored into composition tables, respectively. At the second step, a graph is used to represent the neighboring relations among the regions in a detailed partition. The scale dependences and detailed relations are then used to derive topological relations at the coarse level. Our methods can also be extended to handle the scale dependences of relations about disconnected regions, or the combinations of connected and disconnected regions. Because our methods use the scale dependences to derive relations at the coarse level, rather than generating coarse partition and computing the relations with geometric information, they are more efficient to support scale-dependent applications.

Research on road extraction from digital imagery is motivated by the need for data acquisition and update for geographic information systems (GIS). Roads usually have parallelism of road sides, and on the images, especially the edge map, there are dual edges for each road. In this paper, we propose an approach for automatically extracting road from ETM panchromatic image with a resolution of 15 meters based on Dual-Edge Following. Our approach uses the edge detector with embedded confidence (EDEC, Peter Meer, 2001) to detect road edge, then traces road to generate road candidates by Dual-Edge Following, next exploits the perceptual organization based on probability to link the road segments. Dual-Ddge Following use edge information of both road sides to search for road segments which can improve the precision of road segments. The experiment with ETM panchromatic image in Xinjiang, China shows the validity of the approach.