Detecting Lateral Offset Distance on Rural Road in Thailand by Using Point Cloud Data : A Case Study

Abstract: The lateral offset distance refers to the horizontal clearance that serves as an essential buffer for vehicle operation. Illegal permanent and temporary obstacles along roadsides, such as privately owned signs or building expansions, can obstruct visibility and increase the risk of collisions. Traditionally, road agencies have relied on manual inspections to estimate these distances, a process that is both time-consuming and often inaccurate. To overcome these challenges, this study introduces an automatic process for determining lateral offset distance using point cloud data. Approximately 10 kilometers of point cloud data were collected from rural roadways in Thailand using a Mobile Laser Scanning system. Clustering algorithms, specifically RANSAC and DBSCAN, were employed to identify road lines and walls within the point cloud data and calculate the lateral offset distance. The dataset encompasses a variety of road characteristics, including residential areas, high-traffic volume roads, and roads along rivers. Mean Absolute Error (MAE) was used to measure the accuracy of this algorithm by comparing the automated results to manual detection. The algorithm demonstrated the highest accuracy for high-traffic volume roads, achieving an MAE of 0.343 meters per kilometer. The study also reported average lateral offset distances across the different road classes, with roads along the river having the widest average lateral distance at 10.787 meters, and residential roads having the shortest average lateral distance at 4.911 meters. Suggestions for further improvements to the algorithm were also provided based on these findings.