On Compositional Hierarchical Models for holistic Lane and Road Perception in Intelligent Vehicles

This work is a contribution to the vision based perception of multi lane roads of urban intersections. Given multiple input features the proposed probabilistic hierarchical model infers the lane structure as well as the location of stoplines and the turn directions of individual lanes. Thereby, it expresses prior expectations on the road topology using weak probabilistic constraints which allows for the detection of parallel lanes as well as splitting and merging lanes.