Skip to Main Content
Many systems have been designed for pedestrian detection and tracking. However, pedestrians may or may not be in dangerous situations. Signaling every detected pedestrian generates too many false alarms, which could lead the driver not to pay attention to the warnings. The contribution of this paper is to develop a novel approach to assessing the risk of collision with a pedestrian based on the scenario and the behavior of the pedestrian. The risk assessment is based on extensive offline Monte Carlo simulations relying on a simple, yet representative, stochastic model of the pedestrian. The approach has been applied for the design of the Risk Assessment Unit of the PROTECTOR project dealing with pedestrian detection and classification funded by the European Commission.