Volume XLII-5/W2
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-5/W2, 35–39, 2019
https://doi.org/10.5194/isprs-archives-XLII-5-W2-35-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.
Int. Arch. Photogramm. Remote Sens. Spatial Inf. Sci., XLII-5/W2, 35–39, 2019
https://doi.org/10.5194/isprs-archives-XLII-5-W2-35-2019
© Author(s) 2019. This work is distributed under
the Creative Commons Attribution 4.0 License.

  20 Sep 2019

20 Sep 2019

AUTOMATED ALIGNMENT OF LOCAL POINT CLOUDS IN DIGITAL BUILDING MODELS

T. Kaiser1, C. Clemen1, and H.-G. Maas2 T. Kaiser et al.
  • 1Faculty of Spatial Information, Dresden University of Applied Sciences, Friedrich-List-Platz 1, D-01069 Dresden, Germany
  • 2Institute of Photogrammetry and Remote Sensing, TU Dresden, Germany

Keywords: Structure from Motion, Point Cloud, Registration, Building Information Modeling

Abstract. For the correct usage and analysis within a BIM environment, image-based point clouds that were created with Structure from Motion (SfM) tools have to be transformed into the building coordinate system via a seven parameter Helmert Transformation. Usually control points are used for the estimation of the transformation parameters. In this paper we present a novel, highly automated approach to calculate these transformation parameters without the use of control points. The process relies on the relationship between wall respectively plane information of the BIM and three-dimensional line data that is extracted from the image data. In a first step, 3D lines are extracted from the oriented input images using the tool Line3D++. These lines are defined by the 3D coordinates of the start and end points. Afterwards the lines are matched to the planes originating from the BIM model representing the walls, floors and ceilings. Besides finding a suitable functional and stochastic model for the observation equations and the adjustment calculation, the most critical aspect is finding a correct match for the lines and the planes. We therefore developed a RANSAC-inspired matching algorithm to get a correct assignment between elements of the two data sources. Synthetic test data sets have been created for evaluating the methodology.