3D from Video

Principal Investigator:M. Pollefeys 
Funding Agency:Lockheed Martin 
Agency Number:RRMHS1792

Abstract
This project aims at the reconstruction of accurate, detailed, 3D models from video sequences using computer vision techniques. The focus will be mainly on the reconstruction of urban environments. A key enabling technology for 3D reconstruction is structure from motion. It is based on the fact that certain 3D properties of the world that have been lost during projection to the 2D images can be recovered if there was relative motion between the camera and the objects of the scene during the recording of the images or video streams. Dr. Marc Pollefeys and the Computer Vision group at the University of North Carolina at Chapel Hill (UNC) possess significant experience in structure from motion and 3D reconstruction and have presented cutting-edge results during the last several years. Here, we propose to build upon this experience to develop a fully automatic and robust processing pipeline for 3D model generation from video.