A road edge detection approach for marked and unmarked lanes based on video and radar

In this paper we present a reliable and real-time perception approach for detecting road boundaries by fusing video and radar. The road boundary is defined as the change from road surface to non-road area. We show the integration of a multi-lane recognition system into the detection algorithm, which makes the approach independent of the number of lanes and the visibility of lane markings. The performance of the system is evaluated by means of manually labeled reference data. Information on the road geometry like the curvature and the relative position of the ego vehicle is crucial for ADAS. The road boundary detection acts as main component for postponed functions like a run-off-road prevention, which keeps a vehicle on the drivable area. This work is part of the European project interactIVe, which addresses new technologies and approaches to increase vehicle safety through an integrated platform.