Automatic Video Annotation Based On Fuzzy Graph

Automatic video annotation in Content-Based Image Retrieval (CBIR) is very important and challenging task. As the difference understanding of image information between computers' "visual similarity" and people's of "semantic similarity", the "semantic gap (Semantic gap)" is inevitable existence. "semantic gap"is the main reason for the users to use the CBIR system. How to jump the "semantic gap" is a currently urgent technical problem needs to solve. Semantic tagging is an intuitive idea to effectively supply the gap and a long-term research topics of concern. Unfortunately, manual semantic tagging takes a huge human and material resources. It is hoped that statistical model and the machine learning method can be used to semantic tagging in order to support semantic-based video retrieval.In this paper, the status and existing problems of the video semantic tagging are anatomized. A number of classical machine learning algorithms, such as the K nearest neighbor algorithm, Bayesian algorithm and Support vector machine algorithm, are introduced.On this basis, automatic video annotation based on fuzzy graph is put forward to solve the problems. Fuzzy graph is an extension of the Graph theory. Fuzzy operators are applied to Graph to achieve fuzzy reasoning using fuzzy semantic. The experimental results show that fuzzy graph can be used to establish good links between the semantics in some specific fields. Once the first semantic is determinate, the relational semantics could be found accurately. Finally, automatic video annotation comes true.