Abstract: To improve the deformation detection and analysis accuracy of a tunnel, studies are conducted on the axis extraction of a 3D tunnel point cloud, generation and segmentation of tunnel intensity image, pointcloud segmentation of tunnel, and deformation detection of tunnel crosssection. Bassed on the original pointcloud data of the tunnel obtained using a 3D laser scanner, an improved RANSAC algorithm is proposed to fit the center of the equaldistance tunnel crosssection based on the designed axis, and a fitting method of cubic Bspline curve is applied to obtain the spatial axis of the tunnel based on the least square method. Furthermore, the tunnel intensity image is generated using algorithm of data transformation, and the boundary detection and geometric weights calculation algorithms are designed to extract all boundary lines in the image. By mapping the boundary line pixels to the threedimensional space, the segmentation of the tunnel point cloud is achieved. Finally, accurate detection metrics for tunnel crosssection deformation is proposed. The results show that the algorithm obtained can segment the tunnel point cloud and improve the accuracy and feasibility of tunnel crosssection deformation detection.

Key words: deformation detection, point cloud data, axis fitting; tunnel point cloud segmentation