News Videos Retrieval System Based On Visual And Text

With the development of the Internet and multimedia technology, digital video information represents an explosive growth tendency, a large amount of video multimedia data is generated. How to retrieve the video data that users are interested in is one of the most popular directions in the computer retrieval technique. Except traditional text-based retrieval, content-based video retrieval has become an important issue in the field of computer science. In the paper, CCTV News video as the research object, based on its organization, structure and retrieval methods, by combining the visual characteristics of video image(for example the color, texture, local feature points and NDK etc.) with the traditional content-based video retrieval, the CCTV News video retrieval system based on vision and text is proposed. For the news videos retrieval process, story segmentation is its emphasis and difficulty. On the basis of previous work, story segmentation is based on distinguishing the caption frame from the anchor frames in key frames and supplemented with audio detection. This method can greatly improve the accuracy and completeness of the story segmentation. Thereby the reliability and effectiveness of the news video retrieval have been highly improved.The work of this paper is as follows:(1) Realize the data mining of news videos from the internet. Multimedia information retrieval systems firstly obtained the retrieved data, and then analyze the background data of the CNTV, by resolving the IP addresses, the URL link of the CNTV news videos and the id, data and other information of this video is obtained.(2) Analyze the key frame extraction technology. This paper comprehensively analyzes the current existing key-frame extraction technology, and select the key-frame extraction method that suit news video.(3) Present the method of news videos segmentation that combine the caption frame, anchor frame and mute point detection. The identification and extraction of caption frame and anchor frame is realized, as well as the detection of mute point. And the news video are segmented through them.(4) Construct the retrieval platform of news videos retrieval system. Users can retrieval news videos through the color, the NDK, the story, texture, time, et al on the retrieval platform.