Three-Dimensional (3-D) by Videotaping

Vexcel Corporation, in collaboration with Stanford Research Institute International, proposes to design and demonstrate techniques to reconstruct 3-D scene and object structure from an image sequence, taken freely by an ordinary video camera. The techniques proposed are based upon algorithms developed as part of Vexcel's photogrammetric workstation that is used to manually build 3-D models from digital photography. Currently available algorithms will be automated through the use of new image processing and photogrammetric techniques invented at Vexcel, SRI, and elsewhere. Anticipated results of this Phase I effort are a proof of feasibility of quickly and automatically extracting 3-D scene information from a video sequence. Also, characterization of the completeness and accuracy of the 3-D models and their sensitivity to errors in the measurement of image match points resulting from the reduced resolution of video quality images.