Probabilistic Hypothesis Verification
We describe a method for hypothesis verification for view-based recognition using Bayes rule. For feature positions we assume conditional densities which are based on the observation that deviations in the relative position of feature points increase with the distance of feature points. We advocate the use of curve segments with constant sign of curvature instead of circular or elliptical arc segments. Our approach is validated with real and artificial data of mostly curved objects.
Unable to display preview. Download preview PDF.
- 2.Grimson, W.E.L.: Object Recognition by Computer, MIT Press, 1990.Google Scholar
- 4.Pope, A.R., Lowe, D.G.: Learning object recognition models from images. In Proc. ICCV, 296–301, 1993.Google Scholar
- 5.Pope, A.R.: Learning to recognize objects in images: acquiring and using probabilistic models of appearance. PhD thesis, University of British Columbia, 1995.Google Scholar
- 7.Wells, W.: MAP model matching. In Proc. CVPR, 486–492, 1991.Google Scholar