Matching oversegmented 3D images to models using association graphs

摘要

A new enhancement to the use of association graphs for model matching is introduced. The specific problem addressed is matching when, due to noise, errors occur in the segmentation process, and these errors result in oversegmentation: that is, a single surface in the model is reported as two surfaces by the segmenter. The association graph is augmented by ‘merge nodes’ which possess the appropriate properties. The algorithm is described and examples are presented.