A Semantic Description Model And Retrieval Technology Research Of3D Virtual Human Based On The Metadata

As the new generation multimedia data type after images, sounds and videos,3D model has been widely used in different industries, such as virtual reality, industrial design, molecular biology and video animation, and has been shared with an increasing number of models. Therefore, currently it becomes a hot research topic to study model retrieval technology to help users quickly and accurately obtain3D model to reuse resources. Focusing on3D virtual human model, this paper initiates a research on knowledge representation techniques of3D model and on retrieval technology.According to the characteristics of metadata, this paper proposes a multi-level virtual human semantic description model. Dividing into feature layer, structure layer, design layer and role layer, the semantic description model can describe3D virtual human from different sides of the semantic description layer, can provide the definition of each description layer and can map between different layers. The semantic description model incorporates both the low-level model features which include shape and color characteristics and the upper semantic description of knowledge which includes topology, functional use and design that makes the model adaptable for virtual human multi-aspect model retrieval. In multi-level virtual human semantic description model, each layer complements with each other, guarantees information sharing and interoperability from different semantic level users.According to multi-level virtual human semantic description model, this paper constructed a3D virtual human ontology, combined with ontology technology and feature extraction techniques and designed a3D virtual human retrieval system. This paper also listed three retrieval strategies that include content-based retrieval strategy, semantic-based strategy and combination between content-based and semantic-based strategy. Based on the three retrieval strategies, this paper provides a varied retrieval interface including text,3D virtual human examples and an integration, constructs a3D virtual human model database that includes computer-aided instruction domain, games and entertainment domain and implements a3D virtual human retrieval prototype system. Experimental results show that this Multi Level3D Virtual Human Semantic Description Model Based on Metadata is reasonable and available.