It is necessary to track the rat in a Morris Water Maze for the research on spatial learning and memory of laboratory rats. In this paper, a video analysis system based on template matching and trajectory prediction is developed in order to effectively and accurately track the trajectories of the rats in the Morris Water Maze videos. The system can be automatically configured to any Morris Water Maze. Simple mouse operations are done to quickly and effectively locate the maze and the escape platform. Separation of the rat from the maze is based on the maximum interclass variance algorithm. A minimum area searching strategy based on template matching and trajectory prediction is developed to track and identify the swimming rat in the maze. Repeated experiments show that the system is reliable and can effectively identify and track the rat and obtain all the behavior parameters.