Eyes from Eyes
We describe a family of new imaging systems, called Argus eyes, that consist of common video cameras arranged in some network. The system we built consists of six cameras arranged so that they sample different parts of the visual sphere. This system has the capability of very accurately estimating its own 3D motion and consequently estimating shape models from the individual videos. The reason is that inherent ambiguities of confusion between translation and rotation disappear in this case. We provide an algorithm and several experiments using real outdoor or indoor images demonstrating the superiority of the new sensor with regard to 3D motion estimation.
KeywordsProjection Matrice Unmanned Ground Vehicle Scene Point Epipolar Constraint World Point
Unable to display preview. Download preview PDF.
- 3.K. Daniilidis. On the Error Sensitivity in the Recovery of Object Descriptions. PhD thesis, Department of Informatics, University of Karlsruhe, Germany, 1992. In German.Google Scholar
- 4.K. Daniilidis and M. E. Spetsakis. Understanding noise sensitivity in structure from motion. In Y. Aloimonos, editor, Visual Navigation: From Biological Systems to Unmanned Ground Vehicles, Advances in Computer Vision, chapter 4. Lawrence Erlbaum Associates, Mahwah, NJ, 1997.Google Scholar
- 5.O. Faugeras and G. Toscani. The calibration problem for stereo. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, pages 364–374, Miami Beach, FL, 1986.Google Scholar
- 8.R. Y. Tsai. An efficient and accurate camera calibration technique for 3D machine vision. In Proc. IEEE Conference on Computer Vision and Pattern Recognition, pages 364–374, Miami Beach, FL, 1986.Google Scholar