Abstract: The existing over- and under-excavation detection methods cannot effectively adapt to variable tunnel sections. To address this issue, an over- and under-excavation detection method based on three-dimensional image reconstruction technology is proposed for variable large cross-sectional tunnels. The tunnel axis serves as the streamline, and the designed excavation cross-section acts as the reference line. The variable-section characteristics of the tunnel are represented by a section enlargement factor k, which is used to construct a surface calculation equation suitable for such tunnels and to form a calculation algorithm for the designed excavation contour. The proposed method was applied to the Liuzhi-side tunnel anchorage project of the Huajiang Canyon bridge. The control points are established, and field photographs are collected. The actual excavation contours of the variable-section tunnel anchorage are reconstructed as point clouds and three-dimensional (3D) surfaces using 3D image reconstruction technology. The grid-ray method is then employed to compare the actual excavation with the designed variable section profiles, enabling the calculation of the over- and under-excavation. The accuracy of the proposed method is validated by comparing its results with actual sprayed concrete volumes and total station measurements. The results demonstrate that the proposed method enables rapid and accurate detection of over- and under-excavation for each excavation cycle in the variable-section tunnels. The average relative deviation rate of the detection results is only 1.50%, which is significantly lower than that of traditional methods (2.83%). The proposed method expands the applicability of over- and under-excavation detection techniques, demonstrates excellent performance in field applications, and provides a reliable basis for quality control during the excavation of variable-section tunnels.

Key words: tunnel anchorages, variable-section tunnel, three-dimensional image reconstruction, over- , and under-excavation detection