People Reacquisition across Multiple Cameras with Disjoint Views

This research consists in developing a system for discontinued tracking moving people across large sites with non-overlapping views. Firstly, in order to characterize the silhouette of an individual, we propose an appearance-based model which includes both colour and spatial feature of extracted silhouette. Once all the appearance information of each passage is extracted, a robust classification procedure is then presented to make the final decision of re-identification. The performance of the system is evaluated on two real datasets collected in difficult conditions. The experimental results show that the proposed system leads to very satisfactory results.