Collision prediction between vehicles in an unstructured environment

In this paper we present a method for prediction and trajectory tracking of a set of vehicles operating in a port. The objective is to enable the prediction of risk situations in order to avoid collisions. The system is designed for high traffic areas located in ports or civil engineering projects, environments in which traffic occurs without structured lanes of movement. While many studies have been made on detection of collision risks in a structured environment (traffic lanes, roads, intersection, urban, etc.), few works concern open areas. The proposed system is static and positioned at critical points. The vehicles have no specific sensor, and in order to predict sufficiently in advance the risk of collision, we use cameras with wide field of view. From the images acquired by cameras with a wide field of view, position and velocity of the moving objects are extrated and tracked in order to prevent collisions. Highlighting of critical situations is done with an analytical approach in the calculation of the intersections of ellipses uncertainties.