3D Tunnel Surface Reconstruction Method in Complex Scenarios

Three-dimensional reconstruction of a tunnel is an indispensable process to the understanding of tunnel environment. A 3D surface reconstruction algorithm based on an improved Iterative Closest Points (ICP) algorithm and Greedy Projection Triangulation (GPT) algorithm is proposed to obtain the information of tunnel surface, which is the critical foundation of an automation tunnel shotcrete construction system. Firstly, tunnel arched surface features are extracted by an improved Density-Based Spatial Clustering of Applications with Noise (DBSCAN) algorithm, whose key parameters are selected by a proposed strategy. Then, we have improved the ICP algorithm on the basis of geometric features of tunnel to register the tunnel point cloud, specifically, the error metric is reformed by the non-linear least squares and the change of normal angles between a point and its neighboring points. With the output result of registration, a GPT algorithm is used to reconstruct the tunnel surface. Finally, we compared the time and fineness of the reconstructed surface with existing methods and our algorithm performs more balanced between running time and the fineness of the reconstructed surface.