GLOBAL SHAPE INFORMATION MODELING AND CLASSIFICATION OF 2D WORKPIECES

Abstract

This paper presents a representation scheme for modelling the global shape information of 2D workpieces and uses this representation scheme to classify workpieces. The global shape information of a workpiece could be seen as its topological structure. Retrieving workpieces with similar global shape would facilitate some design and manufacturing activities. To derive the proposed representation scheme, a 2D workpiece first has to be approximately enclosed by a rectilinear polygon. Then, the simplified skeleton of the rectilinear polygon, a tree structure consisting of a set of line segments, is derived. The line segments on the simplified skeleton involve two types: real links and virtual links. In this paper, we prove that each real link models a rectangular region, and the union of the modelling regions of all real links is equal to the region enclosed by the rectilinear polygon. Briefly, the simplified skeleton of a rectilinear polygon denotes a global decomposition of the polygon, and is topologically similar to the polygon. On the basis of simplified skeletons, a neural network approach is developed for classifying 2D workpieces.