A Fully Automated Approach to Extract and Assess Road Cross Sections From Mobile LiDAR Data

Road cross sections are designed to ensure safe operation of highways. Tangent segments are typically designed with cross slopes to ensure efficient drainage of water off the road's surface, likewise, on horizontal curves the cross section is superelevated (tilted) to help vehicles counteract centrifugal forces. In both cases, ineffective slopes that do not meet design requirements, put vehicles at risk of overturning and skidding. Similarly, if deficiencies exist in side slopes, the chance of recovery for vehicles that run-of-the-road decreases substantially. Thus, transportation agencies must constantly assess elements of a road's cross section to ensure that they meet current design standards throughout their service life. The microscopic nature of cross sectional elements makes measuring such information time consuming, highly disruptive to traffic, and resource intensive. To facilitate more efficient assessments of such features, this paper proposes a novel algorithm to extract road cross sections from light detection and ranging data. The algorithm involves estimating vectors which intersect the road's axis, whereby points within proximity to the vectors are retained and extracted. Slope information is then measured off the retained points. The proposed algorithm is fully automated and employs multivariate adaptive regression splines to identify locations of change in slope. The algorithm was tested on two highway segments in Alberta. The high efficiency and precise manner in which the slope data was extracted, demonstrates the value of using the proposed algorithm in performing network-level assessment of road cross sections.