CSIC Research Talk - Dr. Minglinag Zhou (Tongji University), Title - Streamlining Tunnel Lining Safety Assessment: Automated 3D Defect Inspection and Evaluation

Abstract:
Lining safety is a critical concern for aging tunnel infrastructure. Manual inspection for tunnel lining defects can be time-consuming, inefficient, and error-prone. However, computer vision-based inspection technology provides a highly accurate and efficient solution for detecting tunnel lining defects. In this seminar, an integrated approach for tunnel defect inspection and evaluation will be introduced, utilizing mobile laser scanning (MLS) and linear camera systems (LCS). This approach involves denoising point clouds and fitting ellipses to derive tunnel lining deformations, a deep learning model to detect and quantify leakage and spalling defects, and extracting fine crack defects from high-resolution images obtained from the LCS. By using processed point cloud data and matched defect features in image data, a 3D tunnel model can be reconstructed, providing accurate spatial locations of deformations and lining defects. Finally, a Tunnel Defects Index system will be introduced, which allows for the efficient evaluation of the safety state of shield tunnels, providing a reference for decision-making by tunnel maintenance engineers.