Real-time road-scene classification based on a multiple-lane tracker

A vision-guided autonomous vehicle is taken as an example for studying problems related to real-time recognition and classification of well-marked highway scenes. An algorithm is proposed, that is able to detect and to track the vehicle´s lane, to search for additional lanes, and to classify the localized lane markers as solid or broken. On this basis a symbolic scene description is generated including the number of available lanes, which is visualized as a bird´s eye view. Also, results from real-time tests of the algorithm, implemented in the vision system BVV 3, with several different highway-scenes are reported