Online 3D Reconstruction Based on Lidar Point Cloud

This paper aims to propose an online 3D mesh reconstruction algorithm based on lidar sensors for large-scale scenes. In comparison to point clouds, meshes provide stable geometry and richer semantic information. To address the issues related to sparse point clouds, inaccurate normal estimation, and slow normal calculation, we introduce a new 3D reconstruction algorithm system. Our approach combines the robot's motion with the mesh reconstruction method, enabling us to generate high-quality meshes from the lidar point cloud in a frame-by-frame manner and then stitch them together to form a global mesh. To evaluate the proposed algorithm, we test it on two different datasets - the Kitti dataset obtained using the 64-line Velodyne lidar and the office park environment dataset obtained using the 32-line Ouster lidar. Our experimental results show that the algorithm is not only fast but also has a certain level of robustness towards the density of the laser point cloud.