POLE INCLINATION ANGLE DETERMINATION FROM MOBILE LASER SCANNING DATA WITH THE ASSISTANCE OF HIGH-DEFINITION MAP

Among all kinds of road infrastructure, pole-like object is one of the common types established along the road. Pole-like objects, e.g., lamp posts and traffic sign poles, provide basic traffic functions that maintain driving order and safety. However, poles may incline due to the strike of external force or accumulative deformation over time, endangering both the drivers and pedestrians. Therefore, this research aims at developing an algorithm to automatically calculate the pole inclination angle with the assistance of High-Definition map (HD map). In this study, mobile laser scanning data is adopted for the computation of pole-like object inclination. The mobile laser scanners used are two Velodyne VLP-16s, and the point cloud data is stitched through GNSS and IMU. In addition, the proposed method combines the pole locations provided by HD map, so the step of pole detection can be omitted. And the advantage is that it not only reduces data processing time but also avoids the possibility of pole detection failure. Based on the pole object locations provided by HD map, smaller point sets are cut out from the raw LiDAR point clouds first. By processing these point sets with CANUPO algorithm, the points represent pole-like structure are remained. The next step is to extract the real pole points through Random Sample Consensus (RANSAC) and Density-based spatial clustering of applications with noise (DBSCAN). In the last step, the pole inclination angle can be calculated by fitting the pole points and setting the threshold to determine the effective center of the circle. The experiment of this research is mainly divided into two parts. The first part collects data for about 30 poles with two 16-line LiDAR Velodyne VLP-16 and terrestrial laser scanner (TLS) RIEGL VZ-400 in experimental area. The TLS data will be processed manually and only the pole point cloud will be left as the reference data. The purpose is to use higher precision instruments and manual processed data to calculate the RMSE between the MLS data and TLS data. And the final calculated RMSE value is 1.0 degrees. In addition, since the point cloud data of the skewed poles are difficult to obtain, we use the Heidelberg LiDAR Operations Simulator (HELIOS++) to simulate the vehicle point cloud data in the second part of the experiment. Three different types of pole models were included in the experiment, lampposts, traffic light poles and street sign poles. And the heights of the three types of poles are 2 meters, 5.5 meters and 10 meters respectively. And we tested the influence of different factors on the calculation results of pole inclination. The final results show that the proposed method can effectively extract the features of poles, and it can be applied to calculate the inclination angles of most poles.