A platform for research: civil engineering, architecture and urbanism
A platform for research: civil engineering, architecture and urbanism
Automatic structural mapping and semantic optimization from indoor point clouds
Abstract Indoor map plays an important role in the fields of indoor position and navigation, location-based services and emergency response. In this study, we address the issue of how to automatically generate indoor geometric maps from laser point clouds, in a manner that is not restricted to the Manhattan-world assumptions. The proposed method comprises two main contributions, namely, (i) indoor geometric structure extraction by the M-RSC (Modified Ring-Stepping Clustering) method and (ii) semantic-constrained optimization models. In the extraction phase, the extracted lines are usually irregular and incomplete because of missing data, object occlusions, and density change of point clouds, etc. Therefore, feature points of initial structural lines are extracted and introduced into the optimization models by different semantic constraints. Additionally, four indoor datasets are tested to demonstrate our approach. Experimental results show that the proposed method is effective and can make the final indoor maps more accurate in terms of the geometry.
Highlights Automatic indoor structure mapping to make up for the lack of indoor maps. Semantic constraints (e.g., perpendicular, collinear, and circular) are utilized to regularization. Curved buildings can be mapped and optimized without being restricted to the strict Manhattan world. The accuracy of the constructed indoor geometric map is at the centimeter level.
Automatic structural mapping and semantic optimization from indoor point clouds
Abstract Indoor map plays an important role in the fields of indoor position and navigation, location-based services and emergency response. In this study, we address the issue of how to automatically generate indoor geometric maps from laser point clouds, in a manner that is not restricted to the Manhattan-world assumptions. The proposed method comprises two main contributions, namely, (i) indoor geometric structure extraction by the M-RSC (Modified Ring-Stepping Clustering) method and (ii) semantic-constrained optimization models. In the extraction phase, the extracted lines are usually irregular and incomplete because of missing data, object occlusions, and density change of point clouds, etc. Therefore, feature points of initial structural lines are extracted and introduced into the optimization models by different semantic constraints. Additionally, four indoor datasets are tested to demonstrate our approach. Experimental results show that the proposed method is effective and can make the final indoor maps more accurate in terms of the geometry.
Highlights Automatic indoor structure mapping to make up for the lack of indoor maps. Semantic constraints (e.g., perpendicular, collinear, and circular) are utilized to regularization. Curved buildings can be mapped and optimized without being restricted to the strict Manhattan world. The accuracy of the constructed indoor geometric map is at the centimeter level.
Automatic structural mapping and semantic optimization from indoor point clouds
Wu, Hangbin (author) / Yue, Han (author) / Xu, Zeran (author) / Yang, Huimin (author) / Liu, Chun (author) / Chen, Long (author)
2020-11-02
Article (Journal)
Electronic Resource
English
Semantic Enrichment of Indoor Point Clouds - An Overview of Progress towards Digital Twinning
| TIBKAT | 2019