Potential of Using Intensity Data from Mobile Multibeam Spinning LiDAR for Predicting Standard Retroreflectivity of Lane Markings

The evaluation of lane marking retroreflectivity is crucial for maintaining safety, especially for connected and autonomous vehicles. Handheld and mobile retroreflectometer units are the accepted technologies for checking the visibility of lane markings. Research has shown the feasibility of using mobile light detection and ranging (LiDAR) to infer lane marking retroreflectivity through the LiDAR intensity. Prior research mainly focused on survey-grade profiler LiDAR. This study focused on investigating the potential of using lower-cost mapping-grade mobile LiDAR equipped with multibeam spinning laser scanners. The evaluation strategy is based on: (1) investigating the performance of profiler and multibeam spinning LiDAR in terms of homogeneous coverage of the road surface and introduced occlusions by neighboring traffic; (2) examining the impact of the laser beam viewing/entrance angles on observed LiDAR intensity from lane markings; (3) checking the repeatability of LiDAR intensity measurements from different drive-runs in the same/opposite direction; (4) conducting correlation/regression analysis between LiDAR intensity and measured retroreflectivity by mobile systems; and (5) studying the impact of the number of used LiDAR units on the prediction quality of standard retroreflectivity. The research introduces a web-visualization portal for the communication of the findings of the correlation/regression analysis as well as investigating anomalies in LiDAR intensity and retroreflectivity measurements. The processing and analysis strategies are illustrated using data collected at various sites, including local roads and interstate/US highways. The experimental results demonstrate superior performance of multibeam spinning LiDAR over profiler systems. A positive correlation reaching up to 0.90 between LiDAR intensity from the former and retroreflectivity has been observed. Moreover, the research established regression functions showing a prediction accuracy for retroreflectivity within the range of 200–300 mcd/lx/m2 (5%–10% of the maximum measured retroreflectivity). Additionally, using the same models also revealed that increasing the number of LiDAR sensors enhanced the prediction accuracy, with a coefficient of determination (R2) reaching 0.83 when using four sensors. Further, the models showed reliable classification of lane markings into “good” or “substandard” retroreflectivity categories, even with a single sensor, offering cost-effective scalability.