Comment
Author: Admin | 2025-04-28
Image, and the chord length, vault net height, and catenary guide height of the section were calculated. However, this method requires manual point selection. Owing to large manual interference, the repeatability comparison error of multiphase data is large, which is not conducive to deformation positioning and multiphase detection.To summarize, owing to its structural rules, the deformation detection of the shield tunnel has been extensively studied. The irregular section structure of the mine tunnel has caused several problems in the deformation detection process, such as inaccurate positioning of deformation points, difficulties in the unification of multiphase benchmarks, and difficulties in the calculation of section parameters. Therefore, there has been limited research on mining tunnels. In this paper, a deformation detection method based on automatic target recognition is proposed to solve the problems encountered in tunnel deformation detection by the mining method. Through automatic target recognition, the positioning and multiphase data datum are unified. Specifically, a preview image is first generated according to the tunnel point cloud obtained by the mobile tunnel laser detection system. Second, the You Only Look Once version 4 (YOLOv4) [35] target detection algorithm is used to automatically identify the tunnel section containing the target position, and the recognition accuracy is optimized by comprehensively considering the prediction confidence threshold, target spatial position, and target grayscale rule. Then, the chord length and vault net height are calculated by using the gross error elimination and curve-fitting methods. Finally, the applicability and accuracy of the method were verified using the measured
Add Comment