Abstract: Object tracking is an essential process for automating various video surveillance applications. In order to obtain the trajectories of every moving objects in a scene, the tracking algorithm has to equip with the ability in handling occlusion. Among the existing tracking algorithms, most of the researches used prediction model to estimate the object’s trajectory of the consecutive frames. The estimated position serves as a reference tool to detect and resolve occlusion. This study aims to analyze the performance of Kalman filter prediction model during occlusion incident. Although, Kalman filter is widely applied for object tracking, less effort is done on evaluating the parameter setting and its effect in long-term full occlusion. Experiments are conducted with tracking datasets of varying velocity and acceleration. The experimental result is compared with a conventional predicted motion model to verify the performance of Kalman filter during occlusion.