A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
3D as-built modeling from incomplete point clouds using connectivity relations
Abstract As-built building information models (BIMs) based on the 3D point clouds of built environments need to be able to completely and automatically model building elements for various applications (e.g., structural analysis, facility maintenance, and environmental analysis). However, missing data during data acquisition can result in an inaccurate as-built BIM. This study thus proposes an automated as-built model generation method with complete geometry information extraction by exploiting the connectivity between the structural elements in a point cloud with missing data. We used a deep learning model to classify and segment the elements at the point level and employed a neighbor network to extract and model the exact geometry of elements. The experimental results demonstrate that the proposed method can automatically develop an as-built BIM from a point cloud with missing data by recognizing and modeling 99% of the individual elements from the structural elements. As a result, a complete BIM can be produced automatically by overcoming the limitations of missing data.
Highlights Automated as-built building information modeling for point cloud with missing data Deep learning model classifies and segments the components at the point level. Neighbor network extracts and models the exact geometry of all components. 99% of individual elements from structural components are recognized and modeled.
3D as-built modeling from incomplete point clouds using connectivity relations
Abstract As-built building information models (BIMs) based on the 3D point clouds of built environments need to be able to completely and automatically model building elements for various applications (e.g., structural analysis, facility maintenance, and environmental analysis). However, missing data during data acquisition can result in an inaccurate as-built BIM. This study thus proposes an automated as-built model generation method with complete geometry information extraction by exploiting the connectivity between the structural elements in a point cloud with missing data. We used a deep learning model to classify and segment the elements at the point level and employed a neighbor network to extract and model the exact geometry of elements. The experimental results demonstrate that the proposed method can automatically develop an as-built BIM from a point cloud with missing data by recognizing and modeling 99% of the individual elements from the structural elements. As a result, a complete BIM can be produced automatically by overcoming the limitations of missing data.
Highlights Automated as-built building information modeling for point cloud with missing data Deep learning model classifies and segments the components at the point level. Neighbor network extracts and models the exact geometry of all components. 99% of individual elements from structural components are recognized and modeled.
3D as-built modeling from incomplete point clouds using connectivity relations
Kim, Hyunsoo (author) / Kim, Changwan (author)
2021-07-26
Article (Journal)
Electronic Resource
English
Robust NURBS Surface Fitting from Unorganized 3D Point Clouds for Infrastructure As-Built Modeling
| British Library Conference Proceedings | 2014