Abstract
Brain injury and its related diseases is the major health problem in Taiwan as well as around the world. For biological purpose, biologists may want to understand the corresponding behavior change of a fish between brain lesion and reconstruction. So in this thesis we present a computerized video-tracking tool, by using the image processing techniques to tracking multiple zebrafish, recording their swimming path and quantify their behavior.In the viewpoint of video processing, multiple object tracking have been studied for a long time and handling the occlusion also has been one of the major problems. Once occlusion occurs, objects are occluded by others, it may cause tracking errors. So separating occlusion objects is an important task for multiple object tracking. In this thesis, in order to solve occlusion problems, we propose two structural separating methods (i) extrapolation method and (ii) skeleton analysis method. Both methods are based on analyzing the fish skeleton which is obtained by morphological image processing. And then by matching the associations which are extracted from skeleton between consecutive frames, we can track fish individually. At the final part of this thesis, the experiment results show our separating methods are efficient and having acceptable accuracy.