Volume XLII-2/W15
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2/W15, 1083–1087, 2019
https://doi.org/10.5194/isprs-archives-XLII-2-W15-1083-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-2/W15, 1083–1087, 2019
https://doi.org/10.5194/isprs-archives-XLII-2-W15-1083-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  26 Aug 2019

26 Aug 2019

REGISTRATION OF 2D DRAWINGS ON A 3D POINT CLOUD AS A SUPPORT FOR THE MODELING OF COMPLEX ARCHITECTURES

Q. Semler1, D. Suwardhi2, E. Alby1, A. Murtiyoso1, and H. Macher1 Q. Semler et al.
  • 1Photogrammetry and Geomatics Group, ICube Laboratory UMR 7357 INSA Strasbourg, France
  • 2Remote Sensing and GIS Group, Bandung Institute of Technology, Indonesia

Keywords: Architectural drawings, semantics, discretization, segmentation by slices, ICP, registration, buffer zone, classification

Abstract. Laser scanning and photogrammetry methods have seen immense development in the last years. From bulky inaccessible systems, these two 3D recording systems has become more or less ubiquitous, which is also the case in the heritage domain. However, automation in point cloud classification and semantic annotation remains a much studied topic. In this paper, an approach to help the classification of point cloud is presented using the help of existing 2D drawings. The 2D drawings are registered unto the 3D data, to then be used as a support in the 3D modeling step. The developed approach includes the computation of the point cloud cross section and detection of feature points. This is then used in a 3D transformation followed by ICP refinement to properly register the vectorized 2D drawing on the 3D point cloud. Results show that the developed algorithm manages to register the 2D drawing automatically and with promising results. The automatically registered 2D drawing, which often times already includes semantic information, was then used to help classify the point cloud into several architectural classes.