一种装配体模型的离散量化与相似性分析方法

A method for discrete quantization and similarity analysis of assembly model

装配体模型相似性分析作为产品信息重用领域的重点问题而得到广泛关注,目前一些方法利用图论的相关知识,能够实现结构信息的发掘但通常较为复杂,也有一些方法从向量集合的角度入手以获得更好的计算效率,但忽略了零件间的连接关系。针对这些方法的优势与不足,提出了一种基于结构离散的装配体模型信息量化方法,将装配体结构信息融入向量化描述符中,并建立相应的索引结构与相似性度量方法。该方法首先提取装配体中结构特征形成连接图,并将装配体分解为若干个相互连接的零件构成的结构单元;然后,利用结构-形状距离函数对各个结构单元进行量化表征,以此为基础构建基于高维向量的装配体描述符;最后,基于词袋(BOW)模型和超球体软分配策略建立倒排索引与过滤机制,通过求解查询装配体模型与库模型间的最优匹配,最终实现装配体模型的相似性分析。

Similarity analysis of the assembly model is widely concerned in the field of product information reuse.At present,some methods utilize the graph theory to excavate structural information,but they are usually complicated.There are also some methods considering the vector collection to get better computing efficiency,but they ignore the connection between parts.Based on the advantages and disadvantages of these methods,this paper proposes an assembly model information quantization method,which integrates structural information into vectorization descriptor and establishes a corresponding index structure and similarity measurement method.First,a connection graph is used to represent the structure features of the assembly,and interconnected part models are divided into several structural units.Then,each structural unit is quantified by the structure-shape distribution function.On this basis,the point set-based assembly descriptor is constructed.Finally,an inverted index and filtering mechanism is established based on the Bag of Word(BOW)algorithm and hypersphere soft allocation strategy.By solving the optimal matching between the query assembly model and the library model,similarity analysis of the assembly model is ultimately realized.