Deformation detection of mine tunnel based on VSLAM 3D dense point cloud slice

Abstract

At present, the non-contact measurement of mine tunnel deformation is mainly realized by laser scanning technology. To address the issues of poor flexibility, slow detection speed, and low automation in multi-site measurement, a new method for tunnel deformation detection is proposed that combines visual SLAM (simultaneous localization and mapping) and 3D point cloud slicing technology. The mobile robot carries a depth camera to continuously capture RGB and depth images of underground tunnels and uses the visual SLAM algorithm to reconstruct the entire 3D dense map of underground tunnels. A composite filtering method is designed for point cloud denoising. Comparative tests were conducted on the reconstruction effect of point cloud structures under different lighting conditions. The dense point clouds in the whole area at different times are sliced, and the deformation volume and position of the roadway are accurately identified through pairing comparison with skeleton line algorithm and point cloud slicing algorithm. The experimental results show that this method can establish a high-precision 3D dense model of underground tunnels under lighting conditions above 50lx. The composite filtering method can remove a large amount of point cloud noise, and the measurement error of tunnel deformation point cloud is less than 5 mm. It can achieve precise detection of tunnel deformation and determination of deformation position, with fast detection speed, and can meet the needs of under-ground tunnel deformation detection in coal mines.