Research Of 3D Model Geometry Shape Similarity Evaluation Based On 2D Dimensional View

Compared with the two-dimensional scene, performance characteristics and three-dimensional scene of three-dimensional models are with a more visual perception in detail. And 3D models are the fourth multimedia data types following image, sound, video. As three-dimensional models are widely used in mechanical engineering, medicine, chemistry, molecular biology, and other fields of entertainment, there are about two billion of three-dimensional CAD model with a lot of three-dimensional models generated everyday. How to retrieve from the existing three-dimensional models quickly and accurately to reuse or improve is significant.In view of the limitations of text-based retrieval methods, the retrieval methods based on content of models has become a hot research in various fields in recent years, and the most critical technology is characterized by feature-extraction of three-dimensional models. The algorithm based on 2D view of 3D model feature-extraction is one of algorithms based on content. The key technologies of 3D model geometry shape similarity evaluation based on 2D dimensional view are studied and the main work includes the following aspects:(1) This paper describes the background and current situation of the three-dimensional model retrieval technology, and summarizes and analyzes the key technology of three-dimensional model retrieval (model preprocessing, feature extraction, similarity matching).(2) In the early times of product design, designers are accustomed to design sketches to express their thoughts. In this paper, sketch-based retrieval methods are studied, and distributed on the shape of three-dimensional view similarity evaluation algorithm is proposed. Before retrieving, we extract three dimensional view of the model because model features are stored in the database through D2-shape functions; when retrieving, functions extract relevant features from three main drafts by the user and corresponding algorithms carry out the matches with features stored in the database to achieve the purposes of retrieving.(3) A three-dimensional model retrieval algorithm based on weight is proposed for three-dimensional models input by the user. Three main drafts represent "Details" view, reflecting the characteristics of the boundary of and the complete internal information of a model; three-dimensional model projection profile image represents the "rough" shape of the model, reflecting projection features only. With two different but complementary features, we get a projected image invariant matrix. After that, weigh them with the D2-shape functions with a joint descriptor gotten and three-dimensional model retrieval is achieved by distance metric.(4) The combination of these algorithms, with Windows XP Professional/Visual Studio.Net2008 development environment and OpenCASCADE geometric platform adopted, we develop a three-dimensional model retrieval system with standard Windows programming interface style. Finally, we finish two search methods to search:user sketches and three-dimensional models input.