3D CAD Model Similarity Estimation Method Based On Dilation

With the rapid development of CAD technology, 3D CAD model similarity estimation and retrieval become more and more important for industries to reduce the cost and improve the efficiency of design process. Due to the complexity and diversity of the CAD models, the generation of proper shape descriptor of CAD models is a big challenge. In this thesis, the researches are mainly focused on the approach to constructing proper shape descriptors for 3D CAD model similarity assessment.The main contents of the thesis are as follows:A new approach to generating dilation based skeleton (DBS) is presented. In this approach, a initial DBS is generated based on the dilation units first, then the DBS is refined as needed, and graph matching algorithm is adopted to estimate the ratio of similarity according to the DBS graph which is constructed from the DBS directly. As a shape descriptor, the DBS graph contains the gross topology and geometric features of a 3D CAD model, and thus it is stable enough and can effectively support 3D CAD model similarity assessment.The DBS is further extended to a multi-scale shape descriptor called dilation based multi-resolutional skeleton (DBMS). In order to generate the DBMS of a CAD model, we get the DBS of each layer first according to the multi-layer dilation units, then these DBSs are well organized to form the whole DBMS and the DBMS graph, finally a multi-layer matching method is also given to support the efficient matching among different DBMS graphs. Utilizing this new shape descriptor, details of the model can be easily caught using high-level layer of the DBMS and the stability of the whole descriptor is guaranteed by each layer's stability at the same time.Based on the above results, a prototype system ZD- 3DSES is developed. Meanwhile, some tests on this prototype with some CAD models are conducted, which shows the validity of the proposed approaches.