Human motion tracking using mean shift clustering and discrete cosine transform

Human motion tracking is an active area of research in computer vision and machine intelligence. It has many applications in video surveillance and human-computer interface. This thesis proposes two different methods to detect and track a specific person in a crowded environment. In the first method, the mean shift cluster procedure is used to get candidate clusters which converge within a few iterations. Discrete cosine transform (DCT) is applied to each cluster and to the known target to extract the features of the objects. To detect the target from a given image, Mahalanobis distance between each transformed candidate cluster and the target is measured. The cluster with the minimum distance is then considered as the desired target. Tracking is carried out by updating the cluster parameters over time using the mean shift procedure. In the second method, the person is identified in the first frame by analyzing the whole image at the sub-block level with the corresponding DCT results. Tracking in the subsequent frames are performed in a confined area defined by the initial position in the first frame where image color information is used for feature matching. The proposed human motion tracking algorithm has been investigated via extensive simulation results using real-life data and it shows excellent performance for detecting and tracking a prescribed person.