Modeling bridges using Point Cloud Data (PCD) has become essential, as design drawings for many existing bridges are often unavailable. Challenges remain in PCD-based modeling, including accurate and efficient decomposition of PCD into individual components, as well as reliable and robust Geometric Digital Twinning (GDT) of components with complex geometries and data incompleteness. In addition, digital models must extend beyond pure geometric representations and incorporate semantic and structural information in a platform-neutral format to ensure interoperability. To address these challenges, this paper proposes a methodology for generating Industry Foundation Classes (IFC4x3)-compliant Bridge Information Models (BrIMs) from raw PCD, where the BrIMs provide a structured description of bridge components and their relationships, and the IFC standard enables open and interoperable data exchange. First, a deep learning–based semantic-to-instance segmentation pipeline is developed to achieve panoptic segmentation of bridge components. Second, GDT of individual components is achieved by automated element-level dimension estimation, followed by cross-sectional geometry extraction using Parametric Prototype Models (PPMs). Finally, an automated framework is introduced to generate IFC-compliant BrIMs by converting tabular and parameterized bridge information into semantically rich and geometrically accurate IFC models. An experiment conducted on a real-world bridge confirmed the effectiveness of the proposed methodology. Moreover, the use of PPMs together with the structured IFC generation workflow provides high editability, supporting flexible model updates and refinements. The results indicate that the framework significantly enhances the automation, accuracy, and interoperability of bridge modeling workflows, thereby supporting lifecycle-oriented bridge management and digital twin–based applications.
Lin et al. (2026) studied this question.