An MPEG-7 compatible video retrieval system with support for semantic queries

In this paper, we design and implement a distributed video retrieval system with support for semantic queries named XUNET. It firstly uses a self-made video semantic processing tool, which is based on a video analysis tool and added with semantic graph annotation and natural language processing functions, to parse videos and generate corresponding MPEG-7 description files. Subsequently, it establishes distributed index of the MPEG-7 files and distributed storage of video files separately. The system provides numerous web query interfaces, including keywords semantic expansion query, semantic graph query and natural language query, when clients´ query intentions have been submitted, specific interested videos or segments will be quickly exhibited through web browsing or VOD as query results. Because of adopting a distributed architecture, XUNET is highly scalable and supports mass video information index and retrieval.