L2M-Reg: Building-level uncertainty-aware registration of outdoor LiDAR point clouds and semantic 3D city models

Abstract. This paper introduces an automated building extraction method combining CNN segmentation with multi-stage regularization. We address urban mapping challenges including boundary inaccuracies and topological errors through: (1) neighborhood matrix processing for local refinement, (2) spectral graph optimization (SBO) for global consistency, and (3) curvature-adaptive contour refinement (ACR) to preserve geometric features. The pipeline converts initial segmentations into precise polygons through hierarchical processing. Experiments show performance matching state-of-the-art methods like PolyWorld, with superior handling of complex geometries. Key innovations include integrated local-global artifact removal and topology-preserving regularization. The curvature-adaptive approach maintains critical architectural features while eliminating noise. Particularly effective for high-resolution imagery, our solution improves geometric fidelity in urban mapping applications. The framework demonstrates robust performance for 3D city modeling and GIS tasks, overcoming common segmentation limitations. Results confirm accurate building outline extraction from satellite/aerial data, advancing automated urban feature mapping.

Urban Digital Twins (UDTs) are emerging as a new paradigm in smart city strategies, enabling real-time interaction with urban environments and supporting data-driven decision-making. By expanding beyond traditional smart functions, UDTs facilitate the analysis and simulation of urban resilience and sustainability indicators within a virtual city ecosystem, addressing both immediate urban challenges and long-term planning goals. This paper introduces TwinCity, a city-scale Urban Digital Twin framework developed and validated through a case study of the Green City of Benguerir, Morocco. The framework incorporates a technical architecture based on semantic 3D city models, data integration, and simulation scenarios to analyse the solar energy potential of the rooftop, the energy consumption of the building and the morphological indicators. A user-friendly web interface was developed to visualise and interact with the UDT, ensuring its accessibility. By bridging the gap between technical challenges (such as data scarcity) and practical applications, this work offers a replicable model for cities in the Global South.

Abstract. This paper presents a novel methodology for the automated creation of 3D city models for Mexican cities using exclusively open data. In Mexico, while national topographic and elevation datasets exist, they lack crucial features like individual building footprints and road polygons, making it difficult to create 3D city models using the most common existing methodologies. The proposed method addresses these limitations by generating building footprints directly from high-resolution DSMs using a region-growing algorithm and deriving road polygons from the empty spaces between city blocks in the topographic data. These generated features, along with existing data for plant cover and water bodies, are then lifted to 3D using customisable rules. The methodology was implemented with Python and C++ scripts and tested in central Mexico City. Results show that the generated building footprints are often more accurate than those in global datasets (Microsoft, Google), particularly for non-rectilinear buildings, leading to recognisable city landmarks. However, the method has limitations, including missing approximately 30% of smaller buildings and occasionally misclassifying tall vegetation as buildings. Despite this, the work demonstrates the feasibility of creating useful 3D city models for the areas in Mexico with high-resolution elevation data.

Abstract. The increasing demand for accurate 3D urban models is driven by rapid urbanization and the need for advanced geospatial tools in smart cities and digital twin applications. However, integrating Digital Terrain Models (DTMs) with 3D city models remains a persistent challenge due to misalignment between building footprints and terrain surfaces. Existing methodologies suffer from elevation discrepancies, inefficient interpolation techniques, and computational limitations, leading to inaccuracies in urban simulations, particularly for flood risk assessment and infrastructure planning. This study proposes a dual-iteration morphometric algorithm to enhance DTM-3D building integration, ensuring precise alignment of urban structures with real-world topography. The methodology involves node-based terrain adjustments, vertex and planar interpolation, and an iterative refinement process including land lot and building footprints geometry that improves elevation conformity. The algorithm was applied to a case study in Section 13, Petaling Jaya, Malaysia, using LiDAR-derived elevation data for validation. The results demonstrate a reduction in Root Mean Square Error (0.387 to 0.112) and a 75.8% decrease in Mean Absolute Error (0.219 to 0.053) compared to conventional DTM models, indicating significantly improved terrain adherence. Hydrodynamic simulations further reveal that the refined DTM reduces flood overestimation errors by 205.49%, capturing localized elevation variations more accurately. These improvements facilitate more reliable flood modelling, infrastructure planning, and urban resilience strategies. This research advances geospatial modelling for smart cities and digital twins, offering a scalable, high-accuracy framework for urban planning, disaster risk management, and climate adaptation. The proposed algorithm enhances urban simulation fidelity, ensuring more precise decision-making in sustainable city development.

3D Gaussian Splatting (3D GS), known for its efficient and explicit radiance field representation, demonstrates considerable potential for modeling complex 3D scenes. However, its geospatial applicability remains limited, especially for areas such as multi-level scene parsing, heterogeneous 3D data extraction and fusion, and task-driven structured representations. As a preliminary step to address these challenges, this paper proposes a novel 3D GS framework that jointly optimizes geometry and semantics. First, a radiance field optimization mechanism that integrates multi-view 2D semantic labels with monocular depth priors is developed. This mechanism generates Gaussian representations with rich geospatial semantic attributes while substantially improving geometric accuracy. For complex urban environments, a geometry–semantics co-optimization module, comprising a semantics-guided adaptive densification strategy and a depth-weighted semantic propagation method, is further introduced. These strategies effectively suppress cross-view semantic noise and optimize memory efficiency. Experimental results demonstrate that, compared to Light Detection and Ranging (LiDAR)-scanned ground truth, the proposed framework achieves a mean Intersection over Union (mIoU) of 81.2% for semantic segmentation and a mean geometric error of 0.093 m, all while preserving high-fidelity rendering quality. Overall, this work provides a practical pathway for integrating 3D GS into GIS ecosystems and establishes the groundwork for advanced geospatial applications.

VoxCity: A seamless framework for open geospatial data integration, grid-based semantic 3D city model generation, and urban environment simulation

High resolution 3D City Models are essential for the Smart City paradigm, yet acquiring accurate spatial data in dense tropical urban environments remains challenging due to the limitations of passive optical sensors. This study addresses these issues by validating a semi-automated workflow for generating Level of Detail 2 (LOD2) building models using UAV LiDAR data, specifically focusing on mitigating systematic strip misalignment errors. Using Surabaya’s Tunjungan Street as a case study, the research implements a rigorous strip adjustment method followed by nDSM-based extrusion guided by 2D footprints. Results demonstrate that strip adjustment is indispensable, improving Z-axis consistency by nearly 50% (from 1.23 mm to 0.63 mm). Crucially, this improvement minimizes vertical discrepancies and outliers, ensuring the high data cohesion required for precise architectural reconstruction and preventing geometric misinterpretations. The final products achieved a Horizontal Accuracy (CE90) of 0.079 m and Vertical Accuracy (LE90) of 0.385 m, surpassing Indonesian Class 1 standards for 1:5,000 scale mapping. Furthermore, the LOD2 models exhibited a vertical RMSE of 0.268 m, confirming the workflow's reliability for precision critical urban planning.

Abstract. The demand for accurate, lightweight 3D building models is rapidly growing in urban analysis, digital twins, and geospatial information systems. Single-source airborne point clouds, such as airborne laser scanning (ALS) or dense image matching (DIM), often suffer from geometric incompleteness, uneven density, and misalignments, limiting the reliability of Level of Detail (LOD) building reconstructions. While substantial progress has been made in single-source building reconstruction and multi-source fusion, fully automated LOD generation pipelines that effectively exploit cross-source airborne data remain limited. This paper presents an automated workflow for generating precise LOD building models from cross-source fused point clouds, leveraging the precision of ALS and the high resolution of DIM to improve model fidelity. Using point clouds obtained from a slice-to-slice fusion approach, experiments on Luxembourg datasets demonstrate a reduced model standard deviation of 0.17m compared to 0.20m for ALS, 0.29m for DIM, and 0.27m for conventional ICP-based fused point clouds. The results show that our workflow, combined with a polygon fitting algorithm and cross-source fused data, significantly enhances building model accuracy and geometric completeness, highlighting the value of multi-source integration for automated 3D city modeling.

This research describes the Sarajevo Urban Digital Twin (UDT) as an applied tool for city planning and for presenting complex simulation results in an adequate way for a broad audience. Developed within the Urban Transformation Project Sarajevo (UTPS), it explains how traffic simulations from the ETH Zurich (ETHZ) software EnerPol are transformed from raw binary files into time-based scenes in ArcGIS Pro (ESRI). In areas with limited local data, satellite imagery is used to create the base 3D city model. The following sections describe how building footprints and heights are extracted not only to generate Level of Detail 2 (LOD2) models with roof shapes and precise outlines, but also to capture land-cover information essential for the UDT. The automatized workflow converts simulation events into georeferenced features and matches them with the street network, so planners can watch traffic patterns shift throughout the day based on different assumptions regarding population changes. A key purpose of the UDT is to make these data and scenarios visible and understandable for people with different professional backgrounds. By offering a shared visual reference, the model supports planning processes where citizens, experts, and officials need to discuss challenges and agree on policy strategies. The study shows how linking data collection, processing, and clear visual output creates a decision-support tool that helps turn complex urban information into knowledge that can guide real interventions with real stakeholders.

Abstract. Rooftop type classification refers to the process of identifying and categorizing the structural geometry of building roofs using geospatial data. It plays a critical role in urban analysis, aiding applications such as 3D city modeling, solar potential estimation, infrastructure planning and post-disaster damage assessment. This research proposes a transformer-based deep learning model, the Swin Transformer, for rooftop classification to handle the issues of complex roof shapes, small or similar-looking structures, and variations in roof types, especially in areas with unplanned construction. The model is trained on an orthophoto-derived GeoTIFF images having four roof type such as flat, gable, complex and bug. Images were resized to 256×256 pixels and processed in batches of 128. This dataset is split into 2528 training images, 544 testing images, and 545 validation images. The Transformer architecture achieves overall performance with a test accuracy of 75%, showing excellent results for gable classes having F1-score of 85.23% and complex classes achieving F1-score of 70.75%, while flat and bugs classes show moderate performance due to lower recall. With the integration of early stopping and a learning rate scheduler, the Swin Transformer showed improved precision for bugs from 60.94% to 66.67% and flat from 78.41% to 66.93% classes, while maintaining a comparable overall accuracy 75.00% to 74.63% and enhancing class balance in predictions. The proposed architecture is also compared with other state-of-the art models and can be used in future applications such as distinguishing roofs from other urban and roof improved geospatial analysis and smart city development. Contact author first name:

Urban scene segmentation is essential for 3D city modeling and plays a crucial role in various remote sensing applications, including urban planning and environmental monitoring. While integrating knowledge graphs with scene segmentation has improved accuracy, existing methods often depend on dataset-specific knowledge graphs, limiting their generalizability across diverse remote sensing data. To address this, we propose a novel framework that leverages large language models (LLMs) to construct a universal knowledge graph from multi-source geospatial data and incorporate it into remote sensing semantic segmentation tasks, enhancing adaptability and robustness in urban scene understanding. Specifically, the framework comprises two key components: (1) a Graph Construction module that employs LLMs to extract cross-domain semantic relationships and build a universal knowledge graph, and (2) a Knowledge Graph Fusion module (KGFusion) that incorporates the graph into a semantic segmentation network to enhance semantic understanding. To evaluate the adaptability of our method across diverse domains, we curated a mixed dataset encompassing urban, rural, and port scenes. Experimental findings validate the efficiency and adaptability of our method, achieving 70.94% mIoU on the UAVid dataset and 63.23% on the Mixed dataset, outperforming the baseline by 0.43% and 1.04%, respectively. These results validate the robustness of our method in cross-domain scenarios and highlight its potential for broader applications in complex urban environments.

The past few decades have witnessed steady innovations in remote sensing technologies; however, elevation data needed for creating 3D city models are not reachable for several regions in all over the world. Many developed states still without proper nationwide elevation measurements dataset for developing sufficient 3D city models. The current paper addresses the possibility of producing 3D models for areas without elevation data but with footprints, measurements collected from government departments and volunteered individuals. The study aims to investigate and evaluate a different approach to create three-dimensional city models based on data that existed in open-source maps when elevation measurements are not available. The proposed approach can be divided into two stages: footprint and shadow data collection, and height estimation. At first, the footprint information and shadow area are manually gathered from satellite images, then the building height is predicted based on rooftop and shadow data. SketchUp, a 3D design software, is employed as an efficient tool for creating the 3D virtual city model. To develop such a model, the software utilizes procedural modeling in addition to an image-based approach. The developed model can produce a satisfactory and realistic virtual scene within a short time and for a large area. The 3D city modeling resulted from estimated heights is considered as a rational provisional solution at areas where elevation data are not available or are out-dated.

This paper presents an implemented methodology to convert highly detailed building information models (BIMs) into geospatial 3D city models (Geos) at multiple levels of detail (LoDs). As BIM models contain highly detailed and complex geometries that differ significantly from city model standards, abstraction and conversion methods are required to generate usable outputs. Our study addresses this by developing a methodology that generates nine different LoDs from a single IFC input. These LoDs include both volumetric and surface-based abstractions for exterior and interior representations. The methodology involves voxelisation, filtering and simplification of surfaces, footprint derivation, storey abstraction, and interior geometry extraction. Together, these approaches allow flexible conversion tailored to specific applications, balancing accuracy, complexity, and computational efficiency. The methodology is implemented in a prototype tool named IfcEnvelopeExtractor. It automates IFC-to-CityGML/CityJSON conversion with minimal user input. The methodology was tested on a variety of models ranging from small houses to multistorey buildings. The evaluation covered geometric accuracy, semantic accuracy, and model complexity. Results show that non-volumetric abstractions and interior abstractions performed very well, producing robust and accurate results. However, the accuracy decreased for volumetric and complex abstractions, particularly at higher LoDs. Problems included missing or incorrectly trimmed surfaces, and modelling gaps and tolerance issues in the input IFC models. These limitations reveal that the quality of the input BIM models significantly affects the reliability of conversions. Overall, the methodology demonstrates that automated, flexible, and open-source solutions can effectively bridge the gap between BIM and geospatial domains, contributing to scalable GeoBIM integration in practice.

This paper presents a high-resolution urban-scale evaluation of the impact of retro-reflective (RR) façade materials on building thermal load. Unlike earlier studies limited to isolated buildings or simplified geometries, we integrate LiDAR-derived 3D city models, local meteorological data, and per-triangle thermal load simulation to quantify seasonal thermal load impacts on an urban scale. The triangle-based framework enables detailed estimation of shading, orientation, and vegetation effects under realistic urban configurations. The methodology was applied to 914 buildings in Celje, Slovenia, represented by more than seven million building surface triangles. Results show that Prism RR material increased annual heating demands by 2.1% and reduced cooling demands by 0.76%, while Glass bead material increased heating by 1.6% and reduced cooling by 0.65%. On the days of maximum city-aggregate cooling demand reduction, Prism and Glass bead materials reduced cooling demands by up to 19.31% and 15.39%, respectively. These location-specific results demonstrate a seasonal trade-off, where reduced summer cooling demand is counterbalanced by increased heating demand. The analysis also identifies a previously underreported seasonal asymmetry, with the highest heating demand increases occurring in spring when solar irradiation is high yet heating demand remains.

ABSTRACT Efficient management of digital assets throughout their lifecycle is increasingly challenging due to fragmented workflows, insufficient traceability, and limited scalability in traditional approaches. This study highlights the transformative potential of integrating versioning strategies into asset lifecycle management frameworks, emphasizing the benefits of iterative change tracking, enhanced traceability, and improved collaboration. By embedding metadata and versioning information into CityJSON files, this research demonstrates how changes in geometry and attributes can be systematically documented and visualized. Git-inspired workflows, including branching and merging, are applied to showcase effective management of 3D city models, ensuring that updates align with lifecycle stages. The absence of integrated versioning systems in traditional frameworks often leads to inefficiencies, such as data redundancy, inconsistency, and loss of historical context, which can compromise decision-making and resource allocation. Visualization experiments in this study underscore the value of versioning in enabling stakeholders to analyze asset evolution and make informed decisions in real-time. The findings reveal that incorporating versioning into lifecycle management frameworks significantly enhances scalability, data consistency, and operational efficiency in dynamic urban environments. This framework offers a robust, future-proof solution for organizations seeking to optimize digital asset management and maximize productivity.

This article aims to develop an improved algorithm for classification of point cloud data. The primary component of this algorithm is determination of the classification thresholds for different geographical objects, which helps in the automatic classification of the LiDAR point cloud data. The algorithm was tested to classify the point cloud of three different areas of Ha Long city in Quang Ninh province. The results from the three areas show that for the ground points our algorithm is on average 6.4% more accurate than the traditional progressive TIN densification (PTD) algorithm. Further, with the proposed point cloud classification algorithms the average accuracy for asphalt roads is 87.77%, 98.09% for vegetation, and 96.86% for roof objects. The classified roof objects were further processed for house digitization, which provided an average accuracy of 92.07%. The whole dataset was used to develop 3D city models of the three areas (A1, A2 and A3 in Figure 7) in Hon Gai ward, Ha Long city with Level of Detail (LoD) 2.

Corrigendum to “A framework for road space extraction from point clouds and integration into 3D city models.” [Int. J. Appl. Earth Obs. Geoinf. 143 (2025) 104803

The widespread use of three-dimensional (3D) city data plays a significant role in various applications, such as mixed reality, infrastructure facility management, solar potential analysis, navigation, and so on. Ensuring high spatial and semantic quality in these endeavours is crucial to gathering proper results. Ensuring quality means verifying that the data adheres to relevant standards. Although these relevant standards are openly published, there are issues with the names of interoperability and reusability in academic studies and software development efforts. In this study, these issues are addressed using semantic web technologies. Most 3D city models (3DCMs) are treated as knowledge graphs (KG) with this approach. The main contribution of the study is a web-based interoperable tool for validation of CityGML Level of Detail 2 (LOD2) 3DCMs, which is compatible with relevant standards. Besides, an open-source 3DCM-to-KG converter and an open validation ontology are published as by-products while accomplishing the main goal. By virtue of the KG approach, the 3DCM KG becomes capable of carrying its own validation constraints, which come from the validation ontology. With these efforts, this study provides a practical, interoperable solution to improve the quality and usability of 3DCMs and validation plans, fostering consistency across applications while aligning with established standards in the field.

Abstract. Japan's geospatial information policy has evolved significantly since the 1995 Great Hanshin-Awaji Earthquake, marked by the enactment of the Basic Act on the Advancement of Utilization of Geospatial Information in 2007 and the recent "Project PLATEAU," a 3D city model initiative launched in 2020. This paper analyses the 30-year history of Japan's geospatial information policy and focuses on the impact of PLATEAU on administration and industry. The policy's journey, from the initial NSDI definition in 1999 to the contemporary PLATEAU Vision, showcases a shift from infrastructure establishment to broader social implementation. Key milestones include the 2007 Basic Act, which formalized geospatial information, and the Quasi-Zenith Satellite System (QZSS). PLATEAU, a nationwide 3D city model project, stands out for its rapid data creation, open data approach, and diverse use case development across sectors like urban planning, disaster prevention, and mobility. Analysis reveals a transition from infrastructure-centric policies to user-oriented strategies, with standardization efforts evolving from domestic rules to open standards like CityGML. PLATEAU's success stems from its "StandardsOps" methodology, emphasizing agile specification revisions and open community engagement. This approach, which balances open discussions with strict description rules, has fostered a dynamic standardization ecosystem. PLATEAU's impact extends beyond data standardization, influencing business model innovation and industry productivity. Its adoption of open standards and agile methods sets a precedent for future geospatial information policies in Japan and globally, demonstrating the potential for rapid innovation through collaborative standardization.