Laserscanner based road curb feature detection and efficient mapping using local curb descriptions

In this paper a new approach for road curb detection is presented. Different segmentation and classification methods for the extraction of road curb features from a 3d point cloud derived from a laserscanner are described and compared. Further we introduce a novel grid map containing local linear descriptions of the road curbs. In each cell of the map the parameters of a linear model are estimated. These model parameters are directly used for a robust chaining process of road curb features to road curbs. Using local linear descriptions of road curbs allow the robust detection of road curbs with a high position accuracy and low computational efforts regarding computation time and memory consumption.