Abstract: Road traffic marking is an indispensable data in road navigation maps, which plays an important role in making high-precision maps, and navigation maps with higher accuracy, richer and clearer road detail information. The automatic extraction of road traffic marking information in vehicle-borne close-range stereo images is taken as the research target in this paper. For rapid and accurate collection of road traffic marking, an automatic extraction method is proposed based on geometric features. Firstly, the geometric features of road traffic marking are analyzed. Secondly, the extraction rules of traffic marking are constructed based on geometric features. Finally, combined with the foregoing rules, the automatic collection of three-dimensional spatial information of traffic marking is realized based on the seed points. The close-range stereo images collected by mobile measurement system of Nanjing normal university have been carried out to verify the experiment. The experimental results prove that this method compared with the traditional human-computer interaction collection method, has mere advantages in collection efficiency of traffic marking feature points and the geometric similarity of marking. This method approved it provides a technical support for the production of high precision navigation map for reference.

Key words: close-range photogrammetry, vehicle-borne stereo image, road traffic marking, marking characteristic rules, marking information extraction