Lane-level digital maps with accurate lane boundaries and centerlines are critical for autonomous driving and high-definition map services. However, constructing such map with LiDAR data is labor-intensive and costly, especially over large arears. In this study, we proposed an efficient cross-modal approach for lane-level map extraction by integrating aerial image with roadside LiDAR data. Our method leverages aerial images to provide building edge and lane centerline annotations, while the LiDAR point cloud is processed to extract building edge contours and vehicle trajectories. By applying automatic rotation alignment and an affine transformation between two modalities, we avoid the need for continuous LiDAR-based scanning of lane boundaries. This pipeline supports both intersection and non-intersection areas and enables scalable generation of lane-level maps across larger region. Our approach provides a cost-effective solution for the generation of lane-level LiDAR maps, facilitating the development of autonomous vehicle systems and advanced driver-assistance technologies.

Annual intersection screening determines which locations enter safety evaluation when severe crashes are sparse and review budgets are limited. Severity weighted crash frequency gives agencies a traceable additive score, but each crash enters that score through a fixed severity class value. This study tests whether calibrated crash level probabilities of fatal or suspected serious injury can replace those fixed values without losing the audit structure of an additive site score. Crash Records Information System records from two Texas urban networks were processed through separate regional pipelines. Records were collapsed to the crash level, assigned deterministic intersection keys, screened for outcome leakage, modeled with balanced logistic regression, calibrated with sigmoid calibration using the last screening year, and aggregated to intersection scores. Scores from 2016 through 2021 crash histories were evaluated against fatal or suspected serious injury crashes observed from 2022 through 2025. At the 5% screening threshold, the probability weighted ranking captured 61 future severe crashes within the midsized network compared with 47 under fixed severity weighting, and 515 within the larger metropolitan network compared with 445. The same ordering held at the 10% and 15% thresholds. Probability weighted rankings also produced higher adjacent year Spearman rank stability in every evaluated base year. The evidence supports calibrated crash level probability as a consequence contribution for intersection network screening when calibration and leakage controls are maintained at the local data scale.

Roadside LiDAR is a key sensing technology for intelligent transportation systems (ITSs) due to its high-precision spatial information and reliable monitoring of traffic environments. However, extracting traffic information from LiDAR point cloud data remains challenging because measurements are produced through angular sampling, causing the spacing between adjacent points to depend on radius and beam distribution. This study proposes a geometry-aware framework that incorporates LiDAR sampling geometry into the neighborhood criterion used to determine point-to-point association. The formulation defines neighborhood tolerance as a function of radial distance and vertical angular separation, enabling clustering decisions that are consistent with the sensing mechanism. In addition, the approach integrates deployment constraints based on sensor mounting height and region-of-interest limits to maintain physically meaningful connectivity under roadside sensing conditions. A systematic calibration procedure is conducted to estimate the scaling factor and radial spacing parameters and evaluate the method using both controlled and real-world datasets. Experimental results reveal that the proposed approach improves clustering accuracy and stability by reducing false negatives in sparse regions while avoiding excessive cluster merging in dense areas. The method demonstrates robust performance across varying sensing conditions and achieves higher accuracy than baseline approaches without parameter retuning, while introducing negligible computational overhead.