Abstract
We consider in this chapter the problem of recognition and localization of polyhedral objects from range images. Of particular interest are scenes that contain multiple polyhedral objects with objects partially occluding each other. As we have pointed out in the previous chapter, global shape descriptors are ineffective in the face of occlusion and clutter. Recognition-via-localization using only local shape descriptors in conjunction with a proper constraint propagation technique proves to be effective for solving this particular problem. We consider both the interpretation tree (IT) search and the generalized Hough transform in this chapter. In particular, we show that the use of qualitative features can achieve a significant reduction in the combinatorial complexity of the search space of scene interpretations, resulting in fewer spurious scene interpretation hypotheses being generated and greater robustness and efficiency for the recognition and localization process.KeywordsRange ImageStep EdgeModel FacetMatch QualityPlanar FacetThese keywords were added by machine and not by the authors. This process is experimental and the keywords may be updated as the learning algorithm improves.
Published Version
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call