SLAM Method Based On Fusion Of Dual Lidar

With the interdisciplinary development,Lidar SLAM technology has been introduced into the field of Surveying and mapping in recent years to carry out three-dimensional spatial mapping.Although the existing Lidar SLAM methods can better realize the 3D reconstruction of the scene,most of these frameworks are designed to realize the autonomous navigation and positioning task of the robot.They mainly deal with the Lidar scanning data of a single horizontal placement,emphasize the real-time operation of the algorithm,and downsample the point cloud data.However,high-precision 3D reconstruction in the field of Surveying and mapping puts forward higher requirements for the density,coverage of the point cloud map,while the real-time performance of the algorithm is not particularly emphasized.Therefore,in order to obtain a 3D point cloud map with higher coverage and higher accuracy at a lower cost,and to solve the problems of low coverage of a single laser scanner and sparse data of point cloud mapping results,a 3D Lidar SLAM reconstruction method integrating dual laser radars is proposed in this paper.A mobile Lidar SLAM measurement prototype system is designed and built,and the effectiveness of the algorithm is verified by experiments.The main research contents and innovations of this paper are as follows:(1)An external parameter calibration method of dual Lidar assisted by ground laser point cloud is proposed.The dense laser point cloud of the calibration scene obtained by the ground laser scanner is used as the medium,and then the laser point cloud frames obtained by the dual laser scanning head are unified into the unified coordinate system of the ground laser point cloud from their initial relative coordinate system,so as to calculate the external parameters of the dual laser scanning head and realize the calibration of the external parameters of the two Lidar.(2)A self calibration method of external parameters of dual Lidar without assistance is proposed.Firstly,SLAM is carried out with horizontal Lidar to obtain the sub map of horizontal point cloud map and the corresponding global pose data in the early stage of mapping;The same mapping algorithm is used for vertical Lidar to obtain the vertical point cloud sub image and vertical key frame data,and then the initial calibration parameters are obtained by matching the horizontal and vertical point cloud sub images.Then the vertical point cloud frame is registered to the horizontal point cloud sub image,and the optimized calibration parameters are calculated based on dynamically optimizing the parameters.Finally,the gross error of the calculated calibration parameters is eliminated and the arithmetic average is obtained to obtain the final calibration parameters,so as to realize the external parameter calibration of dual Lidar.(3)A SLAM method of dual Lidar scanning data fusion is proposed.Firstly,the SLAM mapping process based on pose map optimization is carried out for the horizontal laser scanning head to obtain the horizontal point cloud map and the globally optimized pose trajectory data.Then,the calibration parameters and horizontal global pose data is used to transform the vertical point cloud frame into the global coordinate system where the horizontal point cloud map is located,and then the dynamic parameter optimization is carried out to realize the integration of vertical point cloud data and horizontal point cloud map.In order to verify the method in this paper,a mobile measurement prototype system is designed and built,and the scanning data of two outdoor buildings and an indoor parking lot are collected for experiments.The quantitative accuracy is evaluated by using the plane fitting root mean square error of point cloud reconstruction results.The experimental results show that the proposed method integrating the scanning data of dual laser radar scanning head method can have more complete building elevation information compared to the traditional single horizontal laser scanning head mapping method.The plane fitting accuracy from the proposed method is up to 5cm,which can meet the accuracy requirements of point cloud mapping application.