Share Email Print

Proceedings Paper

Fast similarity measure of 3D planar shapes in canonical frames
Author(s): Stephen King Wah Kwok; Ken K.C. Lo
Format Member Price Non-Member Price
PDF $14.40 $18.00

Paper Abstract

An efficient algorithm for recognizing general curved shapes by their contours is proposed. In this algorithm, a convex hull is constructed on the test shape so as to extract the extreme points as the correspondence points between the image and the model. Criteria are devised for choosing four of these extreme points to form a canonical frame onto which all other points are projected. By so doing, invariant curves of a shape, which are robust to noise, can be attained. For saving processing time, we propose a specific distance measure on the extreme points of the curves from a major axis of the canonical frame. These invariant distances can be used to measure the similarity between the shapes of interest. The advantage of this strategy is that there is no requirement for re-parameterization, which is time consuming. Compared to backprojection of the whole shape, our proposed method is more efficient and more flexible. Results show that different resolutions of the sampling points on a contour do not affect our distance measure. Recognition rate can reach as high as 99%.

Paper Details

Date Published: 23 September 1999
PDF: 8 pages
Proc. SPIE 3811, Vision Geometry VIII, (23 September 1999); doi: 10.1117/12.364099
Show Author Affiliations
Stephen King Wah Kwok, Hong Kong Polytechnic Univ. (Hong Kong)
Ken K.C. Lo, Hong Kong Polytechnic Univ. (Hong Kong)

Published in SPIE Proceedings Vol. 3811:
Vision Geometry VIII
Longin Jan Latecki; Robert A. Melter; David M. Mount; Angela Y. Wu, Editor(s)

© SPIE. Terms of Use
Back to Top