摘要： 三维模型检索领域中基于内容的检索方法不能充分表达模型语义信息。针对该问题,提出一种包含语义分类信息的三维模型检索方法。采用人工分类信息、有限的语义标准信息等构建异构语义信息网络,并将其转换为三维模型的异构语义特征,在此基础上使用包含模型语义特征的主题分类方法,并将其应用于模型检索中。实验结果表明,与基于内容的三维模型检索方法相比,该方法能提高三维模型检索的准确性。

关键词: 异构语义网络, 异构特征, 统一关系矩阵, 隐含狄利克雷分配, 三维模型, 检索模型

Abstract: For the problem that the method of content-based retrieval can not fully express the semantic information in the field of 3D model retrieval,a 3D model retrieval method containing semantic classification information is proposed.Artificial classification information and limited sematic annotation information,etc are used to build a heterogeneous semantic information network,and it is converted to a 3D model heterogeneous semantic features.Based on that,a subject classification method containing model semantic feature is used,and it is applied to the model retrieval.Experimental results show that compared with the conventional method of content-based 3D model retrieval,the method can improve the accuracy of 3D model retrieval.

Key words: heterogeneous semantic network, heterogeneous characteristics, unified relation matrix, Latent Dirichlet Allocation(LDA), 3D model, retrieval model

中图分类号: 

• TP391