Self-Constrained Baseline 3-D Correction Approach for Mobile Laser Scanning Point Cloud in Complex Urban Road Environments

Mobile laser scanning (MLS) can provide urban road spatial information, which gained increasing attention in various urban applications. However, the MLS platform position estimation is often inaccurate because the positioning observations tend to be interfered with or blocked out by surroundings in complex urban regions, resulting in the degraded quality of captured point cloud data. Instead of correcting inaccurate MLS platform positions directly, this article proposed a baseline to extract intrinsic characteristics from raw point clouds, which can be used for 3-D correction without additional reference information. Furthermore, this article designed a data-driven 3-D correction approach called self-constrained baseline correction model. First, baselines were generated from raw MLS data by extracting and connecting road markings. Next, intrinsic features information from raw data were extracted by calculating the baselines’ horizontal curvature and longitudinal gradient. Then, the problematic MLS point cloud can be located by abnormal feature information of baseline accordingly, dividing raw point cloud into reference and problematic data. Finally, nonrigid correction processing was performed by building a consistent expression and iteratively minimizing the discrepancy between problematic and reference data, enhancing the accuracy consistency of the MLS road point cloud. Experiments were conducted with six typical problematic scenes collected in Shanghai, China. We demonstrated that the 2-D and 3-D average deviation of problematic data was reduced by 1.06 and 1.10 dm. The accuracy inconsistency of corrected data was also evaluated by analyzing the standard deviation of feature information. The results showed that the data quality can be improved significantly.