AUTOMATIC ROAD CENTERLINE EXTRACTION FROM MOBILE MAPPING STEREO IMAGE SEQUENCES

摘要

MMS (Mobile Mapping System) provides an effective way to collect road data and how to process the massive data obtained by MMS automatically is currently a research hotspot. In this paper, an automatic method of extracting and calculating road centerlines from mobile mapping stereo image sequences is proposed. First, the road center segments in each image ROI (Region of Interest) can be extracted with image segmentation by color space transformation, and the road centeriine is then fitted by least square solution after classifying the centroids based on vanishing points. In the end, the homologous image points are matched under the constraint of epipolar line in stereo images, and the coordinates of each homologous point pair can be obtained. The experiment shows that 92.6% of the centerlines in all images are extracted and calculated with the proposed method.