Real Time Multiple Object Tracking and OcclusionReasoning Using Adaptive Kalman Filters

Object tracking in image sequences is one of thefundamental steps in designing intelligent surveillance systems.The fact that Multiple Object Tracking (MOT) algorithmsrequires occlusion reasoning and data association, makes designof these algorithms much more complicated than Single ObjectTracking (SOT) algorithms. A new method for real time MOT isintroduced in this paper to efficiently solve the occlusion issue.Background subtraction has been employed for detecting objectsin this method. In order to computing data association betweenobject in current frame with previous tracks, a new distancefunction is introduced for implementing General NearestNeighbor (GNN) method. In the case in which objects are in adistance, Kalman filter with constant measurement noisecovariance has been used for tracking objects however whenocclusion happens, measurement noise covariance will beadapted by result of a local template matching in whichcorrelation coefficients method has been employed. Experimentalresults confirm the efficiency and robustness of proposed methodfor MOT and occlusion reasoning.