Tunnel 3D Reconstruction Method Based on Structure from Motion and Multi-View Stereo

Abstract

Tunnel morphology measurement is crucial to the assessment of tunnel construction quality and health monitoring during service. However, the traditional measurement techniques relying on profilers and total stations are limited in measurement points and efficiency. Among the emerging technologies, ground radar is seriously interfered with by electromagnetic waves, whereas the laser scanning technique is high in cost and complicated in operation. To this end, this article proposes a 3D reconstruction method for tunnels based on Structure from Motion (SfM) and Multi-View Stereo (MVS). Firstly, we use the SfM algorithm to obtain a sparse point cloud. Secondly, we use the MVS algorithm to obtain a dense point cloud based on the SfM algorithm. On this basis, we enhance the quality of tunnel reconstruction based on Poisson reconstruction and statistical filtration. The proposed method enables the generation of high-quality three-dimensional models of tunnels from just a single video segment, thus eliminating the previous reliance on high-quality static images. Finally, we use real-world tunnel data to verify the effectiveness of the proposed method based on the evaluation protocol we introduced, and the results show that the proposed method is significantly better than the benchmark methods.