ENGINEERING SEARCH BY COMBINING SHAPE AND ONTOLOGY

摘要

A significant amount of information generated during design and manufacturing contains textual and geometric descriptions such as CAD drawings. Solely text-based or geometry-based approaches have been prevalent in current engineering information retrieval techniques. In order to further enhance the retrieval performance and provide more cohesive engineering search environment, we developed a unified framework that combines text-based and geometry-based approaches. On the text side, we present a novel engineering ontology based retrieval that takes drawing annotations as input data. On the geometry side, we use a novel shape-based 2D drawing retrieval method with sketch query input. This multi-modal retrieval framework is applied on data obtained from an industry surgical robot design repository. The preliminary result is promising.