Computer Vision–Based Conformity Inspection and Contour Analysis of Ultradeep Diaphragm Wall Trenches

As underground engineering advances rapidly, the increasing depth of foundation pits has led to the emergence of ultradeep diaphragm walls. During the construction of diaphragm walls, the quality of the trench directly impacts the success and safety of subsequent construction. To shift diaphragm wall trench excavation from its current overreliance on experience to a more data-driven methodology, while ensuring construction safety, this paper presents a new computer vision (CV)-based conformity inspection method for ultradeep diaphragm wall trenches. A new method is developed to obtain the contour of the trench surface. The extracted contours are utilized in a proposed conformity inspection method for trenches. The distribution of contours in ultradeep diaphragm walls was analyzed. The new CV-based method is applied to Metro Line 7 in Tianjin. Results show the new CV-based method outperforms the widely used Canny method. Statistical analysis indicates that the sequence of trenching and spatial overlap does not have a significant impact on the correlation of verticality. The distribution of trench contours differs in soft and hard soil layers, with a more dispersed pattern in layers with greater stiffness.