Extended path prediction using camera and map data for lane keeping support

Lane departure or lane keeping systems should detect in advance not only if a road/lane departure would occur, but also when it would happen. The goal of the paper is twofold; on one hand achieves accurate detection of such incidents decreasing system misses; on the other hand it extends the electronic horizon of the supervised area to 500 m minimizing false alarms caused by intentional lane changes. To achieve the goals, the paper proposes an extended lane tracking scheme based on an information fusion system, map and camera data, and a path prediction algorithm that incorporates curvilinear vehicle models for the short term prediction and on-line learning for a wider time perspective. The presented algorithm is tested using simulated and real map, camera and vehicle data.