Construction of city video database based on automatic scene detection and recognition

In this paper, a database called “city video database system” (CVDS), which represents situations in the real-world by maps and shots of news videos is discussed. Using the database, users can access shots of news video by maps, or obtain the relevant map information while news video is being watched over the Internet. Here the key issue is that the links between shots of news videos and maps can be realized automatically in real-time. In news video, the information relating to the real world have many kinds of forms, however the information on building an object holds an important position. Therefore, in this paper, problems in the detection and recognition of information relating to building an object in news video is discussed. Here, at first the shots of news video including the information on real-world objects (buildings) are detected. Then, objects relating to the real-world are recognized, and the links between the shots of news videos and maps are constructed automatically. Experimental results will reveal the effectiveness of the methods proposed in this paper