Parallelepipedic shape modeling for moving objects in video surveillance systems

The major challenge of video surveillance systems is the automated detection and interpretation of events of interest. In case of an abnormal event taking place, an alert should be delivered. Generally, a video surveillance system's framework combines three main phases: moving objects extraction, moving objects classification and tracking, scenario recognition. The last stage depends on the aimed application such as moving objects counting or moving objects behavior understanding. The extraction of moving objects, followed by the object tracking and recognition, are very crucial stages that affect the whole subsequent process. In this paper, we propose a geometric 3D shape modeling method to represent the moving object in a parallelepiped shape aiming to simplify the object representation. At the same time, the required features for objects tracking are maintained. Experimental results are given to demonstrate that the proposed technique is effective and efficient for road traffic surveillance.