Research And Application On Ontology-based 3D Model Management Framework

The development of virtual reality and 3D virtualization technology leads to massive growth of 3D models, traditional method focuses on keyword matching to retrieve and manage models, which cannot meet the requirement of semantic retrieval and management of 3D models.To solve this problem, a 3D model management approach based on structural features is proposed. According to the core framework of the method, analysis of massive model instances for concept discovery by description words count and model structural analysis, build ontology in any specific field, design and implement a 3D model management system based on structural features, apply and validate the system at last.The main contribution of this paper is followed:(1) Based on the advantages of ontology semantic web on retrieving and management, analyzing 3D models’ structural features, mining and mapping concepts and properties of ontology, and proposing a 3D model management framework based on structural feature.(2) A 3D model management original ontology building method is proposed. First, upload massive 3D models, analyze visual information, and learn the normal concept, this procedure is called 3D model visual information extraction; Second, do words statistic algorithm to 3D model nodes’ description words, build candidate concept sets, complete 3D models’ concept classification, build 3D model management original ontology. Besides, Study ontology evolution and adjusting procedure.(3) Do model structural features analysis to 3D model instances of original ontology and study ontology optimization method based on 3D model structural features. Study AABB bounding box algorithm, analyze basic relationship between two models, do “skeleton” extraction to model instances, extract model structural level properties. Besides, study model instance’s automatic classification by ontology.(4) Develop a 3D model management platform based on structural features; Validate TMMA framework’s effectiveness which is proposed by this paper by applying the prototype system to an intelligent furniture sales and manufacturing enterprise.Results show that the method we proposed in this paper can solve the problems of inefficiency, redundancy, difficult to reuse in traditional method, achieve the goal of multi-dimension retrieval and management of 3D models.