| Title |
Effective 3D Plane Extraction Method from Point Cloud Data of Various Indoor Spaces |
| Authors |
오유진(Youjin Oh) ; 도낙주(Nakju Doh) |
| DOI |
https://doi.org/10.5573/ieie.2020.57.10.72 |
| Keywords |
Plane extraction; Point cloud data; 3D indoor modeling; Curved surface estimation |
| Abstract |
We propose the method to extract three dimensional planes from point cloud data of indoor spaces and represent indoor spaces with those 3D planes. The proposed method extracts planes from not only planar architectural components but also non-planar architectural components like curved surfaces. This algorithm is robust to various indoor spaces with different sizes and complexities. To extract accurate planes, we utilize a hierarchical plane extraction method to extract initial planes and then use normal vectors of the point cloud data to determine and refine those planes. In the case of non-planar architectural components, we reduce three dimensional data into two dimensional data by projection to reduce the complexity. Then, we divide the data into several line segments, recover two dimensional points that compose the line segments into three dimensional points, and extract planes from the recovered points. The algorithm is verified by six dataset with various sizes and complexities. The suggested algorithm has higher extraction rate, precision, and recall than the other plane extraction method. Also, the proposed method effectively estimates non-planar components with several planes. |