3D Model Retrieval System Based On Multi-view

With the advancement of science and technology,the performance of hardware devices such as computer basic equipment and graphic sensor is on the rise.The number of three-dimensional models is increasing explosively.Three-dimensional technology plays an increasingly important role in such fields as computer-aided design,medical imaging,3D games,virtual reality,geographic information system,biomimetic simulation and video animation.In actual production,most of the requirements can be satisfied by revising the existing three-dimensional model.Therefore,how to retrieve the three-dimensional model satisfying the requirements conveniently and quickly in an existing three-dimensional model library becomes a research hotspot.Text-based retrieval technology is simple,but requires a lot of manpower and resources,but also because of the differences in the person in charge will cause a great error.However,the content-based retrieval algorithm focuses on the visual characteristics of the model itself and has good performance in terms of efficiency and cost,which is currently the main research direction.This paper firstly introduces the field of 3D model retrieval from the research background and significance,the research status at home and abroad,etc.The 3D model retrieval method based on multi-view is mainly studied,the principle of common image feature extraction and common similarity measures are introduced.Then introduces the current popular open source search engine,in-depth study of the Lucene search engine theory and implementation methods,as well as the Lucene based image feature indexing tool LIRE toolkit,mastered its overall architecture,core indexing and search module.Then,a three-dimensional model retrieval system based on multi-view is realized by using the view-based 3D model retrieval algorithm and the open source search engine.The system Web-based SpringMVC framework has a good loose coupling,persistence layer using Mybatis framework makes the persistence layer more flexible.System has text,images and model search function.Finally,we verify the good retrieval performance of the retrieval system from two aspects: recall rate and precision.