Research on Extraction of Road Markings Using Mobile Laser Scanning Data

This paper aims to introduce a method of roadway and road markings extraction using mobile laser scanning (MLS) data to solve the problems of extraction inefficiency and incompleteness with the existing methods. Firstly, the method spatially reconstructs the roadway, extracts the ground point cloud by Euclidean clustering, and filters out vehicle noises based on K-Means++ clustering algorithm. Then, the framework of road marking extraction and semantic identification is developed. The corrected road surface point cloud is further segmented by the improved dynamic threshold method. The minimum bounding rectangle and shape are employed for semantic identification and classification. Finally, a case study is also presented to show the high accuracy and completeness (above 90%). Compared with the existing methods, a more efficient and complete solution is provided, which can facilitate the inventory of road assets and the evaluation of traffic safety for practical engineering projects.