Abstract: This paper presents an innovative architecture for multisensor video surveillance characterized by two pan tilt video cameras. The aim of the system is to track and to characterize moving objects in outdoor environments in real time, with a robust behaviour. In particular, the system here presented exhibits the capacity of extracting video shots by means of a mobile camera operating at multiple zoom levels. The mobile camera collects successive frames of face oval, segments them and track them over time automatically, starting from a cooperative initialisation of a static widefield camera performing automatic object tracking and classification.