Determining the 3-D pose of a flexible object by stereo matching of curvature representations

A 3-D pose determining method for flexible objects is developed by which the stereo matching problem of images is transformed into the matching of two curvature representations. A heuristic chain-coding method is presented for the chain-coding of self-overlapped skeletonized images and a wrapping method to cope with abrupt changes in curvature representations is described. Two curvature-based stereo matching methods are proposed to determine the 3-D poses of flexible objects. Experimental results with real flexible objects are presented. Copyright (C) 1996 Pattern Recognition Society. Published by Elsevier Science Ltd.