| Nowadays, with the rapid development of computer networks and the multimediatechnology, there comes more and more multimedia information. As a major multimedia,video plays the most important role in our daily life. News video as a video whichprovides most information, its content semantic analysis attracts attentions from moreand more researchers. News video content structure analysis is the most importantunderlying technology for news video semantic analysis, and it is also the basis forvideo content security analysis.A framework of news video content structure analysis based on anchor persondetection is proposed in this thesis. On the basis of news video special structure, wedesign the framework to segment news video into some story units. Firstly, do the shotsegmentation for news video and extract some keyframes from the shots. Secondly dothe analysis of the keyframes. Find out all the anchor person units in the news video viathe fusion of face detection, YUV color histogram, PCA and SVM technology and so on.Finally, cluster the anchor person units and get the boundaries of news video story unitsto make the analysis for video content structure.Experiments on a medium-sized news video from CCTV9, the results show thatthe framework can analyse the news video content structure effectively and efficiently,and the precision of news video story segmentation is about 90%. |
PDF Full Text Request