Extended Abstract
This paper presents an approach for digitally modeling road bridges in Iran using the CityGML standard. The goal of this study is to spatially model bridge data by leveraging CityGML standards and GIS. The recommended method enabled modeling bridges with at least 4m length and 2m free height to be modeled with 5cm precision.
Introduction
The demand for precise 3D modeling of urban infrastructure is rising due to its applications in urban planning, infrastructure management, and disaster mitigation. Bridges, as crucial components of transportation networks, play a significant role in facilitating traffic flow and connecting regions. However, their complex 3D structures and interactions with environmental features, such as rivers and road networks, pose challenges in modeling and analysis. Accurate 3D models are necessary for structural analysis, maintenance planning, and traffic management.
CityGML, an international standard for 3D urban modeling, enables the integration of semantic, geometric, and topological data into a unified framework. Unlike traditional modeling methods that often lack standardization and interoperability, CityGML, with its modular data structures, covers various urban features, including bridges. This study focuses on using CityGML to develop a precise 3D model of road bridges in Iran at Level of Detail 3 (LOD3). The objective is to create a semantically rich and geometrically accurate model that supports advanced analyses and decision-making in bridge management.
Materials & Methods
Data Collection
The study was conducted on the Kheiroudkenar Bridge, located in Nowshahr, Iran, at geographic coordinates 36.62798°N and 51.58104°E. This bridge, with a length of 70 meters, width of 12 meters, and free height of 5 meters, serves as a critical transportation link. A UAV equipped with a high-resolution camera (focal length: 8.8 mm; spatial resolution: 0.0024 mm) was used to capture 810 overlapping images. Flights were conducted at four different altitudes (4–15 meters) and camera angles between 30° and 90° to ensure complete coverage of the bridge and its surroundings. To ensure precise georeferencing of the data, ground control points (GCPs) were collected using multi-frequency GPS devices with 3cm precision and located where they had high visibility in the images.
Data Processing
Images were processed using Agisoft Metashape and the Structure-from-Motion (SfM) algorithm. This technique automatically aligns overlapping images and generates a dense point cloud with over six million points. SfM's capability to estimate camera parameters, such as focal length and lens distortion, makes it suitable for creating accurate 3D models from aerial imagery.
The initial point clouds were de-noised to remove outliers and then segmented into distinct bridge components, including the deck, supports, railings, and ramps. This segmentation enabled the creation of separate models for each structural element.
Modeling
3D models of the bridge were developed at multiple Levels of Detail (LOD), ranging from LOD1 (basic geometry) to LOD3 (detailed geometry). Each LOD represents varying levels of precision and detail, catering to different analytical needs. At LOD1, an overview of the bridge structure is provided, while LOD2 and LOD3 offer more precise structural features.
CityGML facilitated the classification of bridge components into groups such as BridgePart, BridgeInstallation, and BridgeConstructionElement, each with specific geometric and semantic attributes. This classification supports detailed analysis and management of bridge components, ensuring comprehensive coverage of the structural and functional aspects of the bridge.
Results & Discussion
This study produced a georeferenced 3D model based on the CityGML standard, suitable for integration into 3D modeling systems and Geographic Information Systems (GIS). The model includes detailed representations of bridge components classified under appropriate CityGML categories, enabling precise analysis and management.
The parametric modeling process ensured high accuracy, with the generated point clouds having a mean error of approximately 0.040873 meters. The feasibility of using CityGML for precise bridge modeling was demonstrated, as was its ability to integrate bridge data with other urban elements.
CityGML’s multiple levels of detail provide flexibility in modeling, allowing for the presentation of varying levels of detail depending on the required analysis. LOD1 depicts the basic geometry of the bridge, suitable for general visualization and preliminary assessments. LOD2 includes more detailed geometries, ideal for structural analysis and more precise visualizations. LOD3 provides highly detailed features, including specific structural elements and components, essential for in-depth engineering analyses and maintenance planning.
Conclusion
The research concludes that CityGML’s data structures are highly effective for bridge modeling, meeting the needs of 3D modeling systems and GIS applications. The developed models provide precise and detailed information about bridge structures, facilitating better management, maintenance, and analysis.
CityGML enables the integration of new bridge elements and their dynamic relationships, expanding the model’s applicability for various use cases. Future studies should focus on extending the model to include different types of bridges and developing dynamic relationships among bridge components. This will enhance the usability and robustness of the model for detailed urban analyses and provide a comprehensive tool for urban planners, engineers, and emergency responders.