Improved object classification of laserscanner measurements at intersections using precise high level maps

This paper deals with real-time object classification at intersection scenarios. Objects are observed using a multilayer laserscanner. The classification is performed using well-known methods of statistical learning. The statistical classification is corrected by rule based a priori knowledge. Precise high level maps provide the possibility to additionally improve the classification by using infrastructure information and the position of the objects in the scene. Classification results of several neural networks and support vector machines are described. Finally, the improvement by high level maps and the final system performance are presented.