Accurate positioning of boundary points for complex shaped tunnel contour changes and extraction of segmented contours can provide strong support for tunnel deformation monitoring, point cloud orthophoto generation, and reconstruction of existing tunnel models. Based on high-precision tunnel point cloud data collected by mobile laser scanning, the original section point cloud is first extracted and preprocessed. Then, contour changes are identified by using contour feature point neighborhood density features. Finally, tunnel segmentation and contour extraction are completed by combining mileage positioning method and boundary extraction algorithm. The feasibility of this method was verified through experiments using measured data from a subway tunnel and a highway tunnel. Compared with existing research, the method proposed in this paper eliminates the high dependence on the tunnel axis and can autonomously and accurately identify contour changes, effectively improving processing efficiency. It provides an efficient and innovative solution for engineering applications related to complex shape tunnel segmentation and contour extraction, and has certain reference value for optimizing the application of high-performance laser scanning technology.