Automatic Reconstruction of Parametric, Volumetric Building Models from 3D Point Clouds

Dissertation, Universität Bonn, März 2019
 

Abstract

Planning, construction, modification, and analysis of buildings requires means of representing a building's physical structure and related semantics in a meaningful way. With the rise of novel technologies and increasing requirements in the architecture, engineering and construction (AEC) domain, two general concepts for representing buildings have gained particular attention in recent years. First, the concept of Building Information Modeling (BIM) is increasingly used as a modern means for representing and managing a building's as-planned state digitally, including not only a geometric model but also various additional semantic properties. Second, point cloud measurements are now widely used for capturing a building's as-built condition by means of laser scanning techniques. A particular challenge and topic of current research are methods for combining the strengths of both point cloud measurements and Building Information Modeling concepts to quickly obtain accurate building models from measured data. In this thesis, we present our recent approaches to tackle the intermeshed challenges of automated indoor point cloud interpretation using targeted segmentation methods, and the automatic reconstruction of high-level, parametric and volumetric building models as the basis for further usage in BIM scenarios. In contrast to most reconstruction methods available at the time, we fundamentally base our approaches on BIM principles and standards, and overcome critical limitations of previous approaches in order to reconstruct globally plausible, volumetric, and parametric models.

Download: http://hss.ulb.uni-bonn.de/2019/5419/5419.htm

Bilder

Bibtex

@PHDTHESIS{ochmann-2019-diss-reconstruction,
    author = {Ochmann, Sebastian},
     title = {Automatic Reconstruction of Parametric, Volumetric Building Models from 3D Point Clouds},
      type = {Dissertation},
      year = {2019},
     month = mar,
    school = {Universit{\"a}t Bonn},
  abstract = {Planning, construction, modification, and analysis of buildings requires means of representing a
              building's physical structure and related semantics in a meaningful way. With the rise of novel
              technologies and increasing requirements in the architecture, engineering and construction (AEC)
              domain, two general concepts for representing buildings have gained particular attention in recent
              years. First, the concept of Building Information Modeling (BIM) is increasingly used as a modern
              means for representing and managing a building's as-planned state digitally, including not only a
              geometric model but also various additional semantic properties. Second, point cloud measurements
              are now widely used for capturing a building's as-built condition by means of laser scanning
              techniques. A particular challenge and topic of current research are methods for combining the
              strengths of both point cloud measurements and Building Information Modeling concepts to quickly
              obtain accurate building models from measured data. In this thesis, we present our recent approaches
              to tackle the intermeshed challenges of automated indoor point cloud interpretation using targeted
              segmentation methods, and the automatic reconstruction of high-level, parametric and volumetric
              building models as the basis for further usage in BIM scenarios. In contrast to most reconstruction
              methods available at the time, we fundamentally base our approaches on BIM principles and standards,
              and overcome critical limitations of previous approaches in order to reconstruct globally plausible,
              volumetric, and parametric models.},
       url = {http://hss.ulb.uni-bonn.de/2019/5419/5419.htm}
}