Shape Knowledge Indexing Using Invariant Shape Code

Shape knowledge indexing is crucial in design reuse and knowledge engineering, in which the pivot issue is to establish the unique representation of the invariant shape properties. In this paper we present a local shape-indexing scheme defined on the Fourier spectrum of the shape distribution. We theoretically prove that our shape code is strictly invariant of affine transformations, and be suitable as an index of the shape knowledge of a form feature. A framework applying the invariant shape code in shape knowledge indexing is presented. Associated empirical evaluations are given to demonstrate the robustness of the invariant shape code regarding input noises; in addition, the quantity analysis of the results is provided. Eventually, our approach is promiseful to develop the unambiguous shape index in large shape knowledge repository.