Three Dimensional Reconstruction Algorithm Using Portable Lidar

Three-dimensional reconstruction is the generation of mathematical models for threedimensional objects that correspond to computer logical expressions.The technology of three-dimensional reconstruction is widely employed in unmanned driving,surveying,logistics,and other areas.Depth information is obtained using an environment-aware sensor,which combines depth data collected multiple times into a single coordinate system to form a three-dimensional model.In this paper,lidar is used as the primary sensor for three-dimensional environment reconstruction.However,due to lidar’s limited field of view,it is sometimes difficult for a single lidar to completely cover the environment.For this reason,a portable lidar device is being developed that incorporates a multiline lidar,an IMU(inertial measurement unit),and a brushless direct current motor.The lidar rotates continuously around its Y axis during mobile reconstruction to obtain a larger scanning field of view.The SLAM algorithm,which is tightly coupled with lidar and IMU,is used in this paper to reconstruct the environment,allowing for accurate motion estimation and real-time performance.The main work is as follows:1)The hardware and software systems are designed based on the functional requirements of the three-dimensional reconstruction system of portable lidar,the fundamentals of lidar,IMU,and other sensors are introduced,the parameters are calibrated,the other hardware is reasonably selected,and the algorithm framework,including the front-end odometer and back-end optimization,is designed under the ROS environment.2)A front-end odometer is built using the integration of multiple-line lidar and an IMU to address the issue of the lidar odometer’s poor robustness.Correct point cloud distortion induced by lidar motion with the IMU.The feature points are extracted based on the curvature information,the keyframe selection method is established,and the local map is built.To achieve real-time local position estimation,the IMU pre-integral provides the initial position of point cloud matching.3)A global position optimization and mapping algorithm is intended to account for the accumulated error of the front-end odometer.To establish global data association in largescale scenes,a loopback detection algorithm based on the Scan-Context descriptor is used.Create a front-end odometer factor and a loop factor,optimize global position and posture using a factor diagram,and validate with open data.4)To post-process the reconstructed data in portable lidar application scenarios,a dynamic object filtering approach is applied.The scanning field properties of portable lidar and the influence of the reconstruction algorithm are validated using indoor and outdoor reconstruction tests.Lastly,to validate the reconstruction effect of portable lidar,a three-dimensional reconstruction experiment was carried out utilizing open-source data.According to the experimental results,the average absolute error of the pose of the algorithm in this paper is0.5172 m,which is 58.9% of the Lego-OAM algorithm and 59.4% of the LIO-SAM algorithm.The root mean square of the algorithm in this paper is 0.5604 m,which is less than the 0.9661 m of the Lego-OAM and the 0.9603 m of the LIO-SAM.With achieving a greater scanning and reconstruction range,it can ensure reconstruction accuracy and realtime performance.