To tackle the semantic gap problem in 3D Computer Aided Design (CAD) model retrieval,an auto-tagging algorithm using local shape distributions and semantic probabilities was presented.The shape features of CAD model were extracted by using multi-scale local shape distributions method,and the shape similarity between different models was computed.According to the semantic information in a pre-classified training set,a probability-based tagging frame for tagging CAD models was constructed,so as to establish the correspondence between the low-level shape features and the high-level semantics of the models.The experimental results demonstrated the effectiveness of the proposed algorithm in auto-tagging and retrieving 3D CAD models.%针对三维CAD模型检索中的语义鸿沟问题,提出一种基于局部形状分布及语义概率统计的三维CAD模型自动语义标注算法.采用基于局部形状分布的多尺度特征提取方法获取CAD模型的形状信息,并计算不同模型之间的形状相似度；根据模型样本库中已知的语义分类信息,构建一个基于概率的标注框架对CAD模型进行语义标注,以建立模型形状信息和语义信息之间的联系.实验结果表明,该算法能够有效提高三维CAD模型检索的准确率,检索性能优于仅使用形状信息时的检索结果.
展开▼
