Vision Based Motion Tracking in Real Time Videos

Motion tracking in a video involves the process of following the transition of an object from its initial position to the displaced position. An object can be anything such as a human, property, vehicle etc. Motion tracking has gained significant importance especially with the increase in the desire to develop computer vision based intelligent systems for video surveillance, human tracking, gestures and behavior study, anomaly detection, traffic monitoring etc. Optical flow is a common approach used in vision based motion tracking. This paper discusses in detail, the taxonomy of different algorithms in optical flow and strikes a comparison among the different optical flow algorithms. Also, the experimental observations made by tracking the human motion using two optical flow approaches have been presented and it could be found that Lucas-Kanade algorithm produced better tracking vectors when compared to Horn-Schunck algorithm.