The views from three cameras are combined to extract accurately the position on the floor and thus the trajectory. Simultaneously the human short term behavior is extracted (walking, running, abrupt, standing etc).
