Translator Disclaimer
3 October 1996 Free-form registration for 3D shape comparison
Author Affiliations +
This paper presents a new algorithm for matching two sets of 3D data points. It has a better polynomial time complexity, O(N2.5), over previous techniques. There is not restriction on the amount of missing point-pair correspondence between the two sets of data, provided that a majority of the data points between the two images are consistent in terms of rigidity constraints. The algorithm makes use of the Hough bucketing technique for determining the optimal solution, with a modest 1D Hough space for the rotational space search. The search in the translational space is carried out sequentially. No restrictions on al six transformation parameters are imposed and any arbitrarily shaped surface can be registered provided a unique solution exists. The only assumption made is that the surface be piece-wise smooth to allow surface normal calculations on most regions except at discontinuity boundaries. Experimental results showed successful registrations of a simple bottle and a couple of anatomical surfaces captured in 64 by 64 range images with the execution time of about half a minute per registration on a personal computer.
© (1996) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
K. H. Lau and Kim Chew Ng "Free-form registration for 3D shape comparison", Proc. SPIE 2899, Automated Optical Inspection for Industry, (3 October 1996);


An improved silhouette for human pose estimation
Proceedings of SPIE (September 12 2017)
Hough-based recognition of complex 3-D road scenes
Proceedings of SPIE (February 14 1992)
Real-time classification of wooden boards
Proceedings of SPIE (February 01 1991)

Back to Top