GNSS Precise Point Positioning Enhanced Mobile Laser Scanning Railway Point Cloud Registration
Abstract
Mobile laser scanning (MLS) is widely used for high-precision measurement of linear infrastructure such as railways. The fusion of Global Navigation Satellite System (GNSS) data with MLS is highly valuable for obtaining high-precision point cloud results. Leveraging the advantages of GNSS precise point positioning (PPP), which estimates position using single-receiver observations, we propose a GNSS PPP-enhanced MLS point cloud registration method that operates without ground base stations. The method integrates PPP positions with local semantic features extracted from railway environments, including ballast bed planes, barrier planes, utility poles, and power lines. A coarse-to-fine registration framework is developed, where PPP provides both initial values for interframe translation and global constraints to suppress drift, while semantic features are used to construct rotation matrices, followed by iterative closest point (ICP) for fine registration. Experiments were conducted on a 3-km railway section, where a high-precision laser scanner mounted on a track inspection trolley captured dense point clouds at an effective rate of up to 1 million points per second. Results show that PPP constraints effectively suppress cumulative drift, achieving centimeter-level registration accuracy. The proposed method yields a 2-D root mean square error (RMSE) of 0.048 m and a 3-D RMSE of 0.079 m. Moreover, semantic geometric features significantly improve rotation accuracy, reducing roll and pitch RMSE by 65.5% and 77.4%, respectively. The robustness experiment achieved 2-D and 3-D RMSE values of 0.034 m and 0.069 m on an approximately 5-km railway section. These results demonstrate that the PPP-based strategy combined with semantic geometric constraints can achieve high-precision point cloud registration without ground base stations, offering a cost-effective solution for railway surveying and similar linear infrastructure applications.