Abstract
In this thesis, we propose a system to analyze and classify the video shot of baseball game. Our system classifies the shots of the baseball program into the eight categories: (1) The close-up shots for home team player, (2) The close-up shots for visit team player, (3) The close-up shots for the runner, (4) The pitching and battering shots, (5) The shots of a bird's eye view, (6) The fielding shots, and,(7) The shots of a home run or a foul ball, (8) The shots for player portrait, (9) others. Our system consists of top-down and bottom-up approach. The top-down approach generates the shot analysis graph representing the context of the shot, where the bottom-up approach provides the engines to extract the attributes for the operation of the graph. We have developed a generalized system that can be used for analyzing different types of video. Our system not only deals with the specific ideal shots, but also meets the two requirements: extensibility and flexibility.