Incremental free-space carving for real-time 3D reconstruction

  • Author / Creator
    Lovi, David Israel
  • This thesis addresses the problem of automatic real-time 3D reconstruction of general scenes from monocular video. Whereas many impressively accurate reconstruction techniques exist in the multi-view stereo literature, most are slow offline batch methods designed to work in highly calibrated settings. Real-time reconstruction opens doors for real-time applications. This thesis presents a swift but approximate incremental method based on visibility and free-space constraints. "Free space" refers to the observation that lines of sight between photographed surfaces and their observing cameras must be empty; otherwise the surfaces would be occluded. Our approach begins with a sparse reconstruction from online Structure from Motion and interpolates the resulting points in a free-space aware manner to produce a physically consistent dense 3D model. We validate our algorithms on real and synthetic data, perform complexity analysis proving the real-time quality, and demonstrate the algorithms' usefulness for improving visualization in a tele-robotics context.

  • Subjects / Keywords
  • Graduation date
    Fall 2011
  • Type of Item
  • Degree
    Master of Science
  • DOI
  • License
    This thesis is made available by the University of Alberta Libraries with permission of the copyright owner solely for non-commercial purposes. This thesis, or any portion thereof, may not otherwise be copied or reproduced without the written consent of the copyright owner, except to the extent permitted by Canadian copyright law.