Research For The Video Semantic Retrieval Technology Based On Multimodality Information

With the rapid development of multimedia and network technology, when people enjoy the convenience of information, they also face the problem which long-term existence and should not be ignored, which is how to retrieval or locate the information you need from the network video media database quickly and conveniently. When the increasingly resources in the network have greatly enriched people’s life and work, video retrieval has gradually became a challenging problem for researchers in video analysis and retrieval. In order to solve the problems, we usually need the method of semantic classification of video information to achieve full mining the object on video information concepts existing between these factors, while the final goal was to provide users and Internet base on keywords retrieval way. Meanwhile, video content according to the appropriate and flexible way, for users to search on the web to provide thoughtful assistance on the terminal server, made the convenient on the video information retrieval as booked. which made the users found the most efficient way to find the contents they interested. The retrieval technology of video semantic need to solve hard technical problems in all aspects.Because much of video data on Internet, and video data are not consistent in many aspects with text and images, the biggest characteristic on video was a great space redundant, so video media in the information storage need to compressed redundancy. Using the compressed domain information for feature analysis and extraction of video information, which required a lot of research. After finished feature extraction analysis, also needed to classify videos shot based on semantic, the semantic shot type information and corresponding audio data type information to effectively carry out video story theme understanding. And then we need to solve the classification of text on information caption detection, localization and video and so on. In this paper, we designed a multimodality information-based semantic retrieval platform of video film and television and news to process structured analysis, feature extraction, parameter setting, classification of audio information in video, similarity retrieval of video information based on VC++ language. Finally realized the division of decision-making on video affective, to improve the final retrieval results. The main work of this paper was to research for the video semantic retrieval technology based on multimodality information.