Probabilistic road estimation and lane association using radar detections

Knowledge about the course of a road is an important prerequisite for numerous advanced driver assistance systems. Current approaches for lane recognition and estimation are usually utilizing camera sensors in order to detect and process lane markings on the road surface. In this paper, it is investigated if it is feasible to estimate the course of a road using an automotive radar sensor, only. For that purpose, a data fusion approach is proposed which processes stationary radar detections from the road border as well as tracked vehicles. Furthermore, a probabilistic algorithm for associating vehicles to particular lanes is presented. Both techniques are empirically evaluated against a classical camera-based approach.