图形相似的基本原理、方法及其在结构模式识别中的应用--《计算机学报》2002年09期

将相似的有关思想和方法引入到图形识别和理解中 ,系统地讨论了图形相似的基本原理和方法 .首先分析了图形的相似性特征、相似性的内涵以及相似性的性质 ,同时给出了图形相似度的计算方法 ;其次 ,借助图的方式对图形的各相似性特征进行描述 ,并提出了统一的图形特征编码方式 ,以此实现图形由“形”到“数”的转换 ;而且 ,提出了一种具有层次性的、以图形特征编码为基础、以相似度为依据的图形相似性判别算法 ;最后 ,将该方法应用于基于二维的结构模式识别中 ,以一个实际零件中的非相交基元体的识别为例说明了该方法的具体应用 .实验结果表明 ,与传统的识别方法相比较 ,该方法在模式的识别范围、模式的搜索以及模式的匹配速度等几个方面取得了较大进展 ,并且这种基于多个相似性特征的图形识别方法具有更大的灵活性和可靠性 .
【作者单位】：浙江大学CAD&CG国家重点实验室 杭州310027 (谭建荣;岳小莉);浙江大学CAD&CG国家重点实验室 杭州310027(陆国栋)
【关键词】：图形相似;相似性特征;相似度;图形编码;结构模式识别
【基金】：国家自然科学基金 (698780 3 8);国家“八六三”高技术研究发展计划 CIMS主题 (863 -5 11-9842 -0 0 6);浙江省自然科学基金 (6960 45 )资助
【分类号】：TP391.4
【DOI】：cnki:ISSN:0254-4164.0.2002-09-010
【正文快照】：
　　1　引　言图形相似是画法几何与计算机图形处理中的一个基本问题 ,许多与图形识别处理相关的研究都涉及到图形相似 ,如工程图形识别理解、特征识别、模式识别、分形、甚至制造领域中的成组技术等等 ,因此对它的研究具有非常重要的实际意义和广泛的应用价值 .鉴于不同领域的研究

　In this paper, the similarity idea and some corresponding method are introduced into 2D graphic recognition and understanding, and the basic principle and approaches of 2D graphic similarity are systematically provided. The premise of this paper is established on the following facts, that similarity among different graphic existed in many aspects, such as, topology structure similarity, geometry shape similarity, express function similarity, and so forth. As to the same similarity features, the similarity can be further separated out from each other according to their different similarity degrees. In this paper, the basic principle and methods of graphic similarity are discussed as follows. Firstly, the similarity of 2D graphic is analyzed, including similarity features, similarity connotation and similarity property, then the computation method of similarity degree is put forward. Secondly, the graphic features are described by means of graph, while the graphic coding mode is proposed at the same time, with which the graphic can be changed into digital. In the following, taken graphic code as the source and similarity degree as the basis, a hierarchy discrimination algorithm of graphic similarity is brought forward. At the end, more details are provided to illustrate the structure pattern recognition using this approach. The experimental results show that some improvements are obtained by using this method, such as the pattern recognition scope, the speed of pattern searching and matching, etc. Furthermore, the new algorithm has stronger flexibility and stability compared with those traditional ones.
【Keyword】：graphics similarity, similarity feature, similarity degree, graphic coding, structure pattern recognition