Consideration of Road Feature Extraction Using Low-Cost LiDAR Mounted in a Vehicle

Surveying by mobile mapping systems (MMS), terrestrial laser scanners (TLS), and portable laser devices is widely used in public works projects in Japan, resulting in the accumulation of vast amounts of point cloud data in various locations. A point cloud is a collection of three-dimensional points and is increasingly being used as the base data for the digital twinning of roads and urban spaces. However, since each point does not retain the attributes of the road features, a structured semantic point cloud for each road feature is required. Therefore, the authors have developed semantic point clouds that can be used for the maintenance and inspection of road features using road drawings, road maps for autonomous driving, and deep learning techniques. Conversely, continuous measurement of roads and urban spaces by laser scanning is necessary to monitor changes in the road features. While MMS, the mainstream laser measurement method, can capture highdensity point clouds, it is expensive to deploy and requires frequent measurements. In this study, we employed a low-cost LiDAR, which can easily measure point clouds, to generate area data of road features and extract these features from the point clouds. Consequently, we were able to extract road features with an average F-measure of 0.800 or higher, indicating that the point cloud data measured by the low-cost LiDAR can be effectively structured according to the road features.