An Automatic Extraction Method of Road from Vehicle-Borne Laser Scanning Point Clouds

Laser scanning lines index are built from original vehicle-borne laser scanning data. An classification method for automatic extraction of road pavement and side is proposed. Firstly, through the analysis of the spatial distribution characteristics of different objects in scanning lines, a clustering of objects profile points are applied. Then, according to the geometric features of the point set, the type of point set is determined. Finally, the distribution of the edge points of the adjacent multiple scanning lines is used to de-noise. Two point cloud data provided by Vehicle Survey System are used in the experiment. The average integrity rate of road pavement and side extraction are 94.4%, 86%, the average accuracy rate are 98.9%, 99.1%.The experiment shows that this method can effectively decrease the error classification of pavement points, reduce the objects interference to the roadside extraction, and adapt to different road conditions of urban street.