Research On 3D Positioning And Reconstruction Technology For Tunnel Mobile Robot Based On Lidar

The 21 st century is a century for the exploration of the underground space for human beings.As an important part of the development of underground space,underground tunnel plays an important role.In fact,due to the backward construction standards,disrepair,neglect of maintenance or other factors,a large number of civil air defense tunnels built in the 1960 s and 1970 s were abandoned.In order to meet the requirements of follow-up planning and reuse of these underground space,it is necessary to measure their interior three-dimensional structure.In addition,for the tunnels which built for a long time,their structure charts,model structure and other data become inaccurate due to various reasons,and even have been lost,so it is necessary to remeasure and redraw the drawings.Therefore,it is of great significance to obtain the real three-dimensional structure information of the tunnel.The current TLS technology used for tunnel structure scanning has high accuracy,but the mapping cost is much too high,the field work time is long,the indoor work needs point cloud registration,and the single survey site cloud has blind spots.In this paper,a measurement system based on two Li DARs and one IMU is designed.One Li DAR coordinates with IMU to achieve positioning,and the other Li DAR vertically scans to obtain point cloud data.And a series of data processing from data acquisition,trajectory solution to the final point cloud processing and modeling is processed.According to the environmental characteristics of the tunnel,Li DAR and IMU are used as sensors to realize SLAM trajectory calculation of tightly coupled Li DAR/ IMU.Based on the framework via incremental smoothing optimization based on factor graph,scan context is introduced for key frame extraction and closed-loop detection to achieve high real-time and high-precision positioning.According to the characteristics of the tunnel point cloud model,a point cloud preprocessing process is designed.The SOR algorithm is used to remove discrete points,and then MLS is used to smooth the point cloud to supplement small holes,the uniform grid method is used to sparse the point cloud at last.The process can effectively make the large-scale point cloud data with a lot of noise into sparse point cloud with little noise.Aiming at the characteristics of the tunnel point cloud,such as narrow and long structure,obvious orientation and symmetrical cross-section structure,the idea of extracting the central axis of the tunnel is applied to the point cloud slice.Based on the adaptive method of the point cloud density,the thickness and interval of the slice are determined.The point cloud is sliced along the central axis of the tunnel.Then a series of NURBS curves are fitted by the obtained boundary lines,and finally a complete NURBS surface is obtained by skinning operation to realize the surface reconstruction from point line surface.Experiments show that the accuracy of the surveying and mapping data obtained by 3D reconstruction is within 5cm,which meets the requirements of relevant regulations.