From stereo image sequences, the 3-D positions of the object points can be determined by stereo triangulation. Therefore, motion estimation can deal with 3-D coordinates directly. In this sense, one faces a less difficult problem here than in the monocular case we discussed before. However, stereo motion and structure analysis has its own characteristics, especially in dealing with uncertainty of the 3-D points. In this chapter, we will discuss stereo motion estimation using different weighting schemes, including unweighted, scalar-weighted, and matrix-weighted objective functions. We will investigate the performance limit in terms of the theoretical error bound. We will also address problems caused by outliers or gross errors in input data, and the use of robust statistics to cope with these problems.
KeywordsStereo Image Stereo Match Baseline Length Stereo Pair Error Covariance Matrix
Unable to display preview. Download preview PDF.
- 7.2N. Ayache, O. Faugeras: Building, registration, and fusing noisy visual maps, in Proc. 1st Int’l Conf. Computer Vision (IEEE Computer Soc. Press, Washington D.C 1987) pp. 73–82Google Scholar
- 7.6N. Cui, J. Weng, P. Cohen: Extended structure and motion analysis from monocular image sequences, in Proc. 3rd Int’l Conf. Computer Vision (IEEE Computer Soc. Press, Washington D.C. 1990) pp. 222–229Google Scholar
- 7.7N. Cui, J. Weng, P. Cohen: Motion and structure from long stereo image sequences, in Proc. IEEE Workshop on Visual Motion (IEEE Computer Soc. Press, Washington D.C. 1991) pp. 75–80Google Scholar
- 7.8O. D. Faugeras, M. Hebert: A 3-D recognition and positioning algorithm using geometrical matching between primitive surfaces, in Proc. 8th Int’l Joint Conf. Artif. Intell. (William Kaufmann, Los Angeles, LA 1983) pp. 996 - 1002Google Scholar
- 7.9T. S. Huang, S. D. Blostein, E. A. Margerum: Least-squares estimation of motion parameters from 3-D point correspondences, in Proc. IEEE Conf. Computer Vision and Pattern Recognition (IEEE Computer Soc. Press, Washington D.C. 1986) pp. 198–201Google Scholar
- 7.14S. M. Kiang, R. J. Chou, J. K. Aggarwal: Triangulation errors in stereo algorithms, in Proc. IEEE Workshop on Computer Vision (IEEE Computer Soc. Press, Washington D.C. 1987) pp. 72–78Google Scholar
- 7.15R. V. Kumar, A. Tirumalai, R. C. Jain: A non-linear optimization algorithm for the estimation of structure and motion parameters, in Proc. IEEE Conf. Computer Vision and Pattern Recognition (IEEE Computer Soc. Press, Washington D.C. 1989) pp. 136–143Google Scholar
- 7.21M. D. Shuster: Approximate algorithms for fast optimal attitude computation, in Proc. AIAA Guidance and Control Specialist Conf. (Palo Alto, 1978) pp. 88–95Google Scholar
- 7.22A. Tirumalai, B. Schunck, R. C. Jain: Robust dynamic stereo for incremental disparity map refinement (University of Washington, Seattle, WA 1990)Google Scholar
- 7.23J. Weng, N. Ahuja, T. S. Huang: Optimal motion and structure estimation, in Proc. IEEE Conf. Computer Vision and Pattern Recognition (IEEE Computer Soc. Press, Washington D.C. 1989) pp. 144–152Google Scholar
- 7.24J. Weng, P. Cohen, N. Rebibo: Fusion of stereo views in long image sequences, in Proc. First International Workshop on Multi-Sensor Fusion and Environment Modeling (Int’l Advanced Robotics Programme, Toulouse, France 1989). Also ref.: Motion and structure estimation from stereo image sequences, accepted and to appear in IEEE Trans. Robotics and Automation (1992)Google Scholar
- 7.25J. Weng, P. Cohen: Robust motion and structure estimation using stereo vision, in Proc. Int’l Workshop Robust Computer Vision (University of Washington, Seattle, WA 1990)Google Scholar
- 7.26J. Weng, P. Cohen, M. Herniou: Stereo camera calibration with nonlinear corrections, in Proc. 10th Int’l Conf. Pattern Recognition (IEEE Computer Soc. Press, Washington D.C. 1990) pp. 246–253. Also ref.: Camera calibration with distortion models and accuracy evaluation, accepted and to appear in IEEE Trans. Pattern Anal. Machine Intell., 14 (1992)CrossRefGoogle Scholar