Automated extraction of geometric information from LiDAR point clouds on curved ramps

Several approaches have been implemented to extract road geometric information from point clouds originating from different LiDAR systems. However, they are unsuitable for scenarios lacking trajectory data and involving road widening and complex alignment combinations, particularly in the case of curved ramps. This article proposes an automated framework to process discrete LiDAR point clouds and extract geometric information for these ramps. The framework primarily contributes in three key areas: 1) A node identification method is proposed to accurately segment the horizontal and vertical alignments, especially for fluctuating curvature and varying longitudinal grade; 2) By determining road axis points using road markings and boundaries, the framework supports road widening and all types of ramp cross sections; 3) Cross sections are extracted without slicing and rotating, allowing width calculation within each section. Test results show that the framework achieves geometric extraction accuracies between 90.79 % and 100 %, demonstrating its effectiveness for curved ramps.