Research Of CDB-Multimedia Datebase System

The multi-media database technology is a new database technology. Content-Based Retrieval (CBR) as a current seaching means of the urgent demand in multi-media database system has been a subject needed to be researched. Several key points of CBR will be researched and discussed in the paper for providing theoretic guidance when we build an applied CBR multi-media database in the future. Main contents including:(1) The research of setting up multi-media database system model. For supporting the user searching multi-media information by CBR,The multi-media database system must can set up the model about the semantic characteristic,medium physics characteristic of the medium object and the timespace relation of the medium objects expediently. The paper puts forward a kind of unifed multi-media object semantic data model (UMOSDM) based on analyzing multi-media information characteristics for setting up the model to multi-media information effectively. The model combines object-oriented data model with semantic data model and conceptual model together,and leads some concepts of artificial intelligence for the expression of the feilcl knowledge.(2) The research of multi-media database system storage and organization. The paper puts forward a Combined-Allocation-Scheme based on the idea of Constrained-Block-Allocation at the storing allocation strategy and gives Disk -Striping about the storing of big capacity disk array. In the study about the access path of the multi-media data,We introduce several kinds of representative multi-dimensions data index structure including K-d tree,R tree,VP tree and so on.(3) The research of multi-media database system architecture.. It introduces the multi-media database system architecture of CBR in the text and does explain for each part within system architecture.. Among them we put empthases on the contents of field knowledge' sintroduction and expression in the knowledge assistance module.(4) The research of searching multi-media data by CBR. It includes the analysis to multi-media searching type and customer context' s support to the likeness and vagueness searching. While the searching-handling process and characteristic-index strategy and the likeness-matches algorithm will be introduced.(5) The research of information handling and characteristic extracting about image and video. The process of an image object' s creation and characteristic extracting of video will be explained detailedly.