Abstract. Extracting road information from Lidar point cloud data is crucial for autonomous vehicle navigation, urban planning, and infrastructure management applications. Lidar technology provides detailed 3D representations of environments, making it an effective tool for capturing road and terrain features. Traditional setups, where Lidar sensors are mounted on vehicles or drones, can be limited in complex environments like narrow streets or areas with dense vegetation. This research introduces a novel approach by mounting a Velodyne VLP-16 Lidar sensor on a bicycle, offering increased manoeuvrability in restricted areas and enabling data collection in places inaccessible to vehicles or drones. This bicycle-mounted setup is also cost-effective, providing high-resolution data without expensive equipment. The study presents a methodology that begins with ground point extraction, filtering out non-ground elements to isolate the road surface. Further, specialised algorithms were developed to accurately identify and extract road and sidewalk points from the filtered data, accommodating the varying elevations and textures typical of urban environments. The Lidar data was supplemented with RGB images collected simultaneously during data acquisition, providing additional context for validation. Comparison with ground truth data yielded an 85% to 90% accuracy rate, demonstrating the reliability of the approach in identifying road and sidewalk features. The results of this study have broad applications, particularly in urban planning and autonomous navigation systems.
Bhatia et al. (Mon,) studied this question.