Three-Dimensional Geometric Modeling by Initial Model Modification

A new and versatile method of geometric modeling is proposed. Methods of geometric modeling for the object shape approximation proposed so far are “modeling-after-measurement,” but this proposal is “measurement-after-modeling.” A three-dimensional geometric model is transferred to another geometric model, by modifying the geometry of each mesh point, based on multi-directional images of the new object. The object is set on a precisely indexed turn table and rotated in a certain angle around the table axis. The initial model is also rotated in the same angle in the computer. Geometric coordinate values of each mesh point, which appears on the silhouette at the angle, are modified by using the silhouette image of the object or by measuring the mesh position. This method is not limited to the silhouette image but the position measurement is also applicable. Local modification is applicable to cope with the occluded difficulties inherent in the silhouette detection method. Transformations of a male model to another model of a female and a cylindrical model to a mannequin dummy model are demonstrated. Shape of the mesh and the vertex connectivity in the facet mesh of the initial model is conserved throughout the modification process and passes on the new geometric model. Desired meshing of the initial model transfers to the new. This is convenient to the use of numerical analyses such as FEM (Finite Element Method) and BEM (Boundary Element Method) in object approximate modeling and no further mesh generation required. Accuracy depends on the quality of the multi-directional imaging and/or position measurement. This modification only of the geometry and not of the topology will bring together the reuse of the initial model and the ease of modeling, starting from a simple geometric model with desired meshing. Key-Words: 3D geometric modeling, 3D model modification, Reuse of 3D model, Topology conserved model modification, Modeling of natural object, 3D machine vision.