Detection and tracking of objects and persons by using multiple cameras is increasing rapidly. There are a lot of algorithms in literature able to track moving objects in a video. These algorithms are sufficient to detect fragments of the trajectory and verify that the objects have a coherent motion.
Unfortunately, these algorithms are not sufficiently robust to occlusions, merging and separations. The inadequacy of existing algorithms is a concern. Indeed, as the problem of tracking moving objects on a camera will not be resolved, the multi cameras tracking may not give satisfactory results. The first part of thesis work will be to improve the tracking algorithms in order to make them more robust.
The second part of the thesis will concern multi cameras tracking itself. It will be based on tracking algorithms on single camera, the goal being to transmit relevant data from a camera to another in order to automatically track moving objects.