Integrating vision and laser point cloud data for shield tunnel digital twin modeling

Abstract

In shield tunnel digital twin modeling, constructing high-precision tunnel digital 3D models is essential but poses a challenging engineering problem that affects efficiency and accuracy. To address this, this paper presents a method based on 3D tunnel information models using image and point cloud data. A 3D Tunnel Information Model Construction Method (3DTIM) is proposed, based on vision and laser point cloud data, encompassing six core processing layers, and employing the YOLO-T tunnel inner wall image target recognition method alongside various point cloud data processing methods. Through this method, critical parameters can be rapidly and accurately extracted from images and point cloud data, enabling the direct creation of high-precision 3D tunnel models on the 3D modeling software with an accuracy better than 0.7 cm. The 3DTIM approach not only solves the problem of constructing high-precision tunnel 3D models, but also provides a valuable methodological reference for digital shield tunnel twin modeling, reducing the need for manual measurement and enhancing the efficiency of tunnel digital model construction. This research provides a new practical framework for future studies on tunnel 3D model construction and digital shield tunnel twin modeling based on image and point cloud data, forming a more valuable methodological reference.