Automatic creation of as-is building information model from single-track railway tunnel point clouds

Abstract

Abstract The growing use of Building Information Model (BIM) is propelling an increase in demand for creating or updating as-is models for existing tunnels. However, the geometric modeling can be time-consuming and expensive, especially as a tunnel does not have an as-designed BIM. To meet this demand, this article develops a general technique to automatically identify the different types of components (rails, cross-sections, pipes, catenary equipment and refuges) and create parametric as-is BIMs for single-track railway tunnels using the Terrestrial Laser Scanning (TLS) data. This technique includes two core operations: point cloud classification and model parameters estimation. Several classification algorithms are developed to find the point group belonging to each tunnel component based on the geometric features of points, such as relative distance, relative angle, convex hull and linearity. Based on these point groups, the model parameters of each component are estimated to create the final as-is BIM product. The experiment on two operating single-track railway tunnels was conducted and demonstrated that the proposed technique can automatically create as-is BIMs with identified components. The evaluation on the accuracies of models was conducted by analysis of point-to-model distances as well as comparisons with the existing methods and total station surveying, results of which show that the models of rails, linings and refuges achieve millimeter-level accuracy, and that the models of pipes and catenary equipment achieve centimeter-level accuracy.