Fast 3D Scene Reconstruction from Video and Applications
Jan Michael Frahm
University of North Carolina at Chapel Hill
AG4 Group Meeting
Jan-Michael Frahm is working as a post doctoral research associate
in the Computer Vision group of Prof. Marc Pollefeys at the
University of North Carolina at Chapel Hill. The main focus of his
research is fast and reliable 3D reconstruction from video in
real-time.
Recent developments in computer vision have improved the performance
of structure from motion techniques for 3D model reconstruction from
video. The current research challenge is to achieve reconstructions
at interactive frame rates and to enlarge the numbers of images used
from a few thousand frames at most to several millions of frames.
This opens a wide field of applications for these techniques as well
as it boosts the research in computer vision to improve the
reliability of the methods. The talk focusses on the work on fast
methods for reliable automatic geo-registered large scale 3D
scene-reconstructions from many images for a sensor-augmented
multi-camera system. Besides high quality in terms of both geometry
and appearance, the used techniques can achieve fast performance on
multiple CPUs and GPUs aiming at real-time performance in the near
future. The discussed algorithms of the processing system are
real-time 2D tracking on GPU, a robust estimator for camera poses
which efficiently handles (quasi-)degenerate data, a new fast stereo
approach on the GPU, and a fast stereo fusion approach for error
corrections in the stereo estimations. Furthermore, the talk briefly
introduces applications of these techniques for large scale urban
modeling from millions of images and for augmented reality in TV
broadcasting.