A Mechanical CAD Drawings Retrieval Method Based on Text and Image Information

In the field of industrial production, a lot of engineering drawings are drawn every day. For the purpose of reducing processing costs and improving production efficiency, reusing the used engineering drawings is necessary. In order to effectively reuse engineering drawings, we propose a drawings retrieval method based on text and image information. Firstly, a text parser is used to extract the text information about the manufacturing process features in the drawing and to identify drawings with the same features in the database. Then, the histogram similarity algorithm is used to roughly filter the drawings. Finally, calculating the similarity of the image geometry information of mechanical parts based on the multi-hash algorithm. Through experimental tests, the average recall rate of our proposed drawing retrieval model is about 87% and the average error retrieval rate is less than 20%. The experimental results show that our method meets the actual market demand and has good practical application capability.