Rutting extraction from vehicle-borne laser point clouds

Abstract

Rutting is a type of structural road damage that seriously affects traffic safety, and rutting conditions are typically analyzed only from a two-dimensional cross-sectional perspective. Rutting detection currently lacks directional features and trends along the traveling direction. To address this issue, this paper develops a rutting extraction methodology from vehicle-borne laser point clouds to reflect the actual rutting conditions. The proposed method locates rutting points from cross-sectional data and further integrates the spatial correlation information of continuous cross sections to accurately extract dangerous rutting regions and longitudinal feature lines. Comprehensive experiments show that the Recall and Precision of rutting extraction are higher than 85 % and 90 % respectively, while also exhibiting higher robustness compared to other methods. These results demonstrate the effectiveness and accuracy of the proposed method for rutting extraction in large-scale road scenes. Future research will focus on deep learning-based road damage monitoring and provide valuable references for traffic management, road maintenance, and safety.