A Method of Swimming Rat Detection in Morris Water Maze by Using Image Processing

摘要

To test the effect of drug on the brain in psychology, brain science and medical science, the rat behavior analysis is a necessary experiment. Morris water maze is a popular neurological model to analyze rat behavior for evaluating rat's ability of study and memory. The ability is evaluated by the time and the distance. However, rat detection is principle process for rat behavior analyzation. In this paper, we proposed a method to detect rat's position in Morris water maze by image processing. Gaussian Mixture Models is used for foreground extraction. The Models is used to compute a multivariate distribution of image sequence. The complexity for automatic rat's behavior analyzation is that the rat has same color with the water. All are white color. To test the performance of the proposed method, five image sequence were used. It consists of 11,655 images. Detection accuracy rate was 83.53%. This method perform effectively even though the rat has same color with the water.