A Fast Classification Method of Isomorphic Polygons in Design Data of Large Scale Integrated Circuit

Abstract

A large number of isomorphic polygons are included in design data of large scale integrated circuit. Fast search and classification for isomorphic polygons in these design data are expected to apply to compactification of design data and fast pattern search. A pattern matching method was proposed to reduce a large amounts of design data efficiently. In this method, feature of polygon shape is expressed by an angle-segment sequence called contour signature. So, this method realizes fast processing by efficient comparison using the contour signature. Furthermore, this method realizes compactification of the design data by hashing of classification results. However, there is a problem that feature amounts increase with increase of the number of vertices of polygons. As a result, computation amounts and memory usage amounts are increased. To solve this problem, in this paper, we propose a high speed and high accuracy classification method of isomorphic polygons. The proposed method can reduce a shape comparison process significantly by using a geometric invariant feature value for a polygon. The feature value is calculated by summation of distance from each vertex to center of gravity, and a polygon contour length. This value does not be affected by parallel shift, rotation and reduced scale. The feature value is same for isomorphic polygons and it is possible to fast access an isomorphic polygon group. From evaluation results of the proposed method, we verified that the proposed method can realize fast classification for isomorphic polygons, and compactification of design data.