Soccer Video Crawling And Copy Detection

With the rapid development of Internet and multimedia technology, videoinformation is expanding. Common crawler can’t meet the specialized needs of people,such as google, baidu and so on. Thus as to soccer video search, it becomes an urgentproblem that how to crawl the soccer video. At the same time more and more people areused to distributing video in the form of reproducing and sharing, leading to lots of similaror identical video on the Internet, lots of identical videos not only increase the redundantof database, but also reduce the retrieval efficiency. Therefore, content-base videodetection has become an important topic in international research.Based on in-depth analyzing the calculation of web pages relevance to the topic andprediction algorithm, A focus crawler which is based on thesaurus classification isproposed, the algorithm pre-judgment the relevance to the topic through the analysis ofpage title, the pages which have high priority to the topic are first crawled, and theprecision rate is improved. In the videos which are crawled, lots of copy videos are exist,By analyzing the respective advantages and disadvantages of different feature selectionand matching algorithm in video copy detection, a new video copy detection algorithm atshot level which combines with the knowledge of soccer has been proposed, at the sametime the copy type is also determined through the analysis of copy shots. In order toimprove the precision rate of copy detection, the Scale-invariant feature transform whichhas good robustness is selected, the Nearest and the Next Distance Ratio algorithm whichis used to high-dimensional feature is selected, the algorithm can effectively solve theproblem of matching local features. In order to improve the matching speed of shotssequence, soccer shot type is fully used, and shot content is quantified, thus the matcheswhich are unnecessary are avoid.Experiment results show that the precision rate by the crawler which is based onthesaurus classification is improved, and the web pages which are relative to topic is gooddetermined. In the videos which are crawled, content-based soccer video copy detection atshot level has good results on precision rate, the type of shots are used to filter in thematching process, it also achieves good results. However, the exact match used in soccervideo crawler method is to some extent limit the crawl coverage, it needs to improve, andin order to improve the precision rate, the algorithm which is used to feature matching forkey frame also still needs further analysis.