Mobile Laser Point Cloud Position Consistency Correction In Urban Scene

Mobile laser scanning system collects high-density and high-precision threedimensional information of road environment by non-contact laser ranging efficiently.It plays an important role in fields such as urban road assets management,agriculture and forestry,high-definition map and driverless car,and it is of great significance to the realization and construction of “smart city” and “digital traffic”.However,due to the accumulation of measurement errors of GPS and IMU in the vehicle laser scanning system,there is a significant position deviation in mobile laser scanning point clouds,which seriously affects the subsequent data processing and applications.In engineering applications,control points are usually laid along the vehicle trajectory,and the relative and absolute accuracy of the mobile laser scanning point clouds is corrected according to the ground control points.However,the placement of control points is difficult,timeconsuming and costly.The existing mobile laser scanning point cloud data quality improvement algorithms can be divided into two categories: data-driven and sensordriven.The data-driven methods use the geometric features extracted from point clouds to calculate the transformation matrix,and the sensor-driven method first corrects the position and attitude of the vehicle trajectory and then resolves the point cloud data.In order to improve the quality of point cloud data,and the existing methods have their own shortcomings,it's difficult to apply to complex mobile laser scanning point cloud data acquisition scenarios.Therefore,it's necessary to study the position consistency correction algorithm for mobile laser scanning point cloud data.In this paper,the following research work is carried out based on pairwise registration and graph optimization using vehicle trajectory:(1)The research background,significance and current situation of correcting the position consistency of mobile laser scanning point clouds are described in detail.(2)The composition of the mobile laser scanning system is introduced in detail.The characteristics and error sources of the mobile laser scanning point cloud data are analyzed.The applications of the system in various fields are briefly summarized.(3)According to the acquisition characteristics of the mobile laser scanning systems,a multi-level segmentation method of mobile laser scanning point cloud data based on trajectory is proposed,and its experimental verification is carried out.(4)A method of feature matching is developed,which improves the accuracy and efficiency of feature matching in binary shape context.Based on a binary shape context feature,a pairwise registration method for mobile laser scanning point clouds from coarse to fine is proposed.The accuracy and efficiency of feature matching in binary shape context and the accuracy and robustness of pairwise registration results are verified by experiments.(5)On the basis of pairwise registration,the global position consistency correction of mobile laser scanning point cloud is carried out based on graph optimization.It shows that: mobile laser scanning point cloud data are firstly subdivided into sub-regions according to the trajectory,then registered by three-dimensional feature points,and finally optimized globally according to graph optimization,which can correct the position inconsistency.The proposed method is efficient and robust for mobile laser scanning point cloud data in urban scenes with different deviation levels.