Research On Technologies Of Video Retrieval Supporting Semantic Analyzing

Facing the large amount of video, how to query and search the video fast and effectively to meet the users need had been a hot spot in the area of information retrieval. The traditional video retrieval technologies mainly include two types: the one is based on video meta-information and keywords matching, this technology can retrieve the useful information by keywords matching in text retrieval technologies based on video description. mainstream search engines such as Google and Baidu adopt this technology while it could not make use of videos contents and suffers from the low retrieval accuracy; the other is the retrieval based on contents of videos including their low features, but the low efficiency makes it impossible to support massive videos retrieval and highly concurrent of users. The video retrieval technology supporting semantic analyzing could solve problems of low precision of meta-information retrieval and low efficiency of retrieval based on contents of videos. The technique mainly deals with videos semantic information and builds indexes according to video semantic information. Combined with mature keywords matching technology, different users queries and demands could be met.Video retrieval supporting semantic analyzing mainly researches on how to acquire videos high feature semantic information and what kind of retrieval measures to take to search these semantic elements and finally reaches to high efficiency and high accuracy. Owing to the subjective understanding of videos high-level semantic information, people with different background won t share the same comprehension of the same segment of video, thus standard method to depict videos is urgent. Therefore, in this paper, we use MPEG-7 multimedia description schema to annotate video semantically and establish distributed full-text indexes. Under the help of search engine technologies, our video retrieval system-Xunet is born. The system can support keywords queries, semantic graph queries and natural language queries. When users submit sentences or semantic graph including semantic information, Xunet could return results of videos and video segments in demand and offer video online service in addition. The system integrates the technologies of natural language processing, search engine and the distributed retrieval and storage.We complete the development and testing of video retrieval system supporting semantic analyzing-Xunet and offer online service. The system could offer queries supporting semantic expression and realize the notable retrieval method supporting semantic analyzing.