A shotcreting robot needs to reconstruct the arch surface in three dimensions (3D) during the process of spraying a tunnel. To solve this problem, we propose an improved marching cube (MC) reconstruction method based on a point cloud splice and normal re-orient. First, we use the explosion-proof LIDAR to acquire the point cloud data of the tunnel arch, followed by the use of the iterative closest point algorithm, a PassThrough filter, and a StatisticalOutlierRemoval filter for point cloud splicing, data segmentation, and simplification, respectively. In order to improve the reconstruction accuracy, we adjusted the estimated point cloud normal for normal consistency and obtained the geometric features of the complex point cloud surface. Furthermore, combined with the improved MC algorithm, the 3D reconstruction of the tunnel arch is realized. The experimental results show that the proposed method can reconstruct the 3D model of the tunnel arch surface quickly and accurately, which lays a foundation for further research on a trajectory plan, spraying status monitors, and control strategies.