Extracción automática de perfiles longitudinales del pavimento a partir de LiDAR móvil

Abstract

The functional performance of a pavement measures the level of comfort experienced by users traveling on a road. This comfort level is generally assessed by analyzing pavement regularity using standardized indices such as the International Roughness Index (IRI) or the Ride Number (RN). The fundamental input for calculating these indices is the longitudinal profile of the pavement ruts. Traditionally, these profiles are reconstructed using methods such as precision leveling or inertial profilometers. Although the latter are the most widely used and have proven effective, they have significant limitations, such as sensitivity to variations in vehicle speed and a lack of repeatability in measurements, which compromises the reliability of their use in managing large pavement networks. With the advancement of mobile LiDAR technology, an opportunity arises to overcome these limitations by leveraging high-resolution, three-dimensional spatial data. This research proposes an efficient and automated methodology for processing mobile LiDAR point clouds to obtain longitudinal pavement profiles, achieving high computational performance. The methodology is organized into two phases: the first addresses the development of a strategy based on artificial neural networks for the automatic extraction of the road surface, while the second focuses on the automated identification of the edges and centerline of the roadway in order to trace and calculate the longitudinal pavement profiles. Advanced techniques such as concave envelopes, topological skeletons, and KD trees are used for this latter phase. The methodology was implemented in Python and tested with open-source LiDAR point clouds. Its viability was demonstrated by achieving reduced times in the automatic extraction of profiles and products suitable for calculating standardized indices such as the IRI.