中文摘要：

基于内容图纸检索是图纸重用的关键技术，空间关系是图纸的内容的主要特征。由于空间关系描述复杂、特征向量的维数很高，在基于内容图纸检索中很难应用空间关系匹配。为了克服这些问题，提出了约简拓扑关系的方法，采用图谱方法提取拓扑特征向量，实现了特征向量的维数约减，从而通过多维索引技术有效地加速图纸检索。综合拓扑关系、形状和方位关系作为图纸内容，有效地提高了系统查准率和查全率，通过脱机方式提取拓扑关系和形状的特征向量，而联机方式提取方位关系特征向量，提高了系统检索效率。

英文摘要：

Content-based drawings retrieval is key technology in drawings reuse. Spatial relationships are major feature of drawings content. Owing to complexity of spatial relationships description and high dimension of feature vector, it＇s difficult to compare spatial relationships in content-based drawings retrieval. To overcome these drawbacks, a method to simplify topological relationship is proposed. Feature vector extracted from graph spectra can reduce dimension, which effectively accelerate retrieval through multi-dimensional indexing mechanisms. The system is based on combined content consisted of topological relationship, shape, and directional relationship, which results in improvement in the retrieval recall and precision. The system extracts the features of topological relationships and shape offiine, while directional relationships performed online, which results in improvement in efficiency.