Sketch And Semantic Based Retrieval For 3D CAD Models

In the process of product design, the reasonable reuse of existing CAD models can effectively shorten the product development cycle, and then reduce design costs. With the rapid growth of available 3D CAD models, easy, fast and efficient retrieval of suitable and reusable 3D models has become an urgent problem to be solved. There have been many related work about CAD model retrieval, but the current proposed methods are not convenient and accurate enough. Sketch-based retrieval method makes it more convenient to retrieval, but the retrieval results are still not accurate enough. Semantic-based retrieval methods take use of the users’ intent, which meet retrieval demands more; however, extracting the semantics is very difficult. To solve these problems, sketch and semantic based retrieval of 3D CAD models has been studied, with its main contents being as follows:(1) A semantic descriptor with fine-grained shape features is proposed, which is used to efficiently support sketch-based 3D CAD model retrieval. The semantic descriptor is easy to be extracted from the sketch, and has a high degree of discrimination, which can effectively improve the accuracy of retrieval.(2) A method of constructing semantic descriptor with fine-grained shape features from a sketch is put forward. We recognize 3D shape features from the CAD model corresponding to the sketch based on the 2D loops and its context. By dividing the sketch into rotational and non-rotational regions which are then dealt with separately, the method achieves the goal of improving the accuracy and efficiency of shape feature recognition.(3) A comprehensive method of similarity measurement is set forward. The method matches a sketch with a 3D CAD model based on both semantic similarity and the shape similarity, which improves the accuracy of retrieval.Based on these research results, a prototype system named SKEB-SMR is developed. The experimental results validate the main ideas of the proposed approaches.