Geometric Feature Extraction of Road from UAV Based Point Cloud Data

This study presents a new approach to achieving the high accuracy geometric feature extraction of road surface automatically from UAV based images. The proposed methodology begins with the automatic extraction of road surface from point cloud. The extraction of road is based on point clouds and machine learning classification algorithm. Then, road boundaries are derived from extracted road surface points and are used to estimate the road centerline. The point clouds are then used to create digital elevation models to extract profile and cross-section elevations at specified intervals by referenced the estimated smooth road centerline. The accuracy of the road surface classification is evaluated by comparing manual classified points. According to the results, precise road extraction, road centerline, profile, and cross-sections are produced with high accuracy using the proposed approach.