Spatial ray features for real-time ego-lane extraction

In order to support driver assistance systems in unconstrained environments, we propose to extend local appearance-based road classification with a spatial feature generation and classification. Therefore, a hierarchical approach consisting of multiple low level base classifiers, the novel spatial feature generation, as well as a final road terrain classification, is used. The system perceives a variety of local properties of the environment by means of base classifiers operating on patches extracted from monocular camera images, each represented in a metric confidence map. The core of the proposed approach is the computation of spatial ray features (SPRAY) from these confidence maps. With this, the road-terrain classifier can decide based on local visual properties and their spatial layout in the scene. In order to show the feasibility of the approach, the extraction and evaluation of the metric ego-lane driving corridor on an inner city stream is demonstrated. This is a challenging task because on a local appearance level, ego-lane is not distinguishable from other asphalt parts on the road. However, by incorporating the proposed SPRAY features the distinction is possible without requiring an explicit lane model. Due to the parallel structure of this bottom-up approach, the implemented system operates in real-time with approximately 25 Hz on a GPU.