3D CAD solid model retrieval based on region segmentation

Abstract

3D shape retrieval may find the existing models as reference for design reuse. 3D segmentation decomposes models into new elements with large granularity and salient shapes to replace the faces in a solid model. In this way, it may reduce the complexity of a CAD model and make a local salient shape more prominent. Therefore, a retrieval method for 3D CAD solid models based on region segmentation is proposed in this paper. To deal with the problems of poor efficiency and uncertain results, a three-step segmentation method for CAD solid models is introduced. First, face adjacency graph (FAG) descriptions for query models and data models are created from their B-rep models. Second, the FAGs are segmented into a set of convex, concave and planar regions, and the relations among the regions are represented with a region graph. Finally, the sub-graphs are combined recursively to form optimal region sub-graphs with respect to an objective function through an optimal procedure. To avoid using complex graph matching or sub-graph matching for model shape comparison, region property codes are introduced to represent face regions in a CAD model. The similarity between the two compared models is evaluated by comparing their region property codes. The experiments show that the proposed method supports 3D CAD solid model retrieval.