Imaging sensor technologies are capable of generating massive amounts of spatio-temporal 3D
point and image data, which facilitate the capturing of highly detailed of vast environments.
Driven by the development in digital 3D imaging and scanning technology, the efficient capturing,
processing and reconstruction of 3D objects and entire architectural structures have become
critical tasks in the context of computer aided architectural design and planning, virtual
building construction and maintenance, as well as industrial engineering. Despite previous work on
urban capturing and modeling, complete solutions have not yet been achieved to address the accurate
structural reconstruction of 3D interiors. Proposed prior solutions suffer from limiting
assumptions on the geometry, missing data, partially occluded objects, or large structural
artifacts. In this research project we develop new methods and models for the effective
geometric and semantic structural reconstruction of 3D interior environments from massive scanned
3D point cloud data. In this talk we will review our solutions to handle large structural artifacts
based on robust local geometric planar patch analysis, and the detection of the major wall
structures to label the architectural components of the scene.
Professor of Computer Science, University of Zürich
Computer graphics, scientific visualization, parallel rendering, multiresolution modeling,