For the efficient manipulation of information in support of design and manufacturing one would need to access large active databases. In this context, it is envisioned that 3-D object models must be easily retrieved for modification, or for simply reviewing related information. This paper addresses the development of a classification system for the efficient retrieval of 3-D object CAD models. For the design of a new object the user would access the database to locate the most similar object model and, if worthwhile, modify it to attain a model for the new object under consideration.
The proposed classification procedure comprises three main stages: First, 3-D geometrical data, sufficient for classification, is determined and extracted from a selected model of the object, and a primary representation domain is determined. Contours, were selected as the basic descriptive units. Based on this selection, relationships defined by descriptor elements were established between (a) two solid contours; (b) a solid and a hollow contour; and, (c) two hollow contours. In order to facilitate the use of the descriptor elements, a simple alpha-numeric coding system was developed.
In the second stage, the description of the object is mapped from the geometrical domain into a numerical-factor domain, where each contour is represented by a numerical value.
In the final stage, a code is generated based on the factor-domain representation in the form of a Fourier signature. Classification is then carried out according to the similarity of the generated Fourier signatures.