Line based reconstruction from terrestrial laser scanning data

Abstract

Abstract The methods for reconstructing objects from dense point clouds, typically acquired by terrestrial laser scanning (TLS), are currently unsatisfying, especially from a geodetic perspective, for a number of reasons. Digitization in the point cloud lacks redundancy and no stochastic measures are provided generally. Reconstruction from primitives, e.g., planes, spheres, and cylinders, enforces strong requirements on (large) object parts, which are strictly valid only for very special cases. In this article we propose a new reconstruction technique for terrestrial laser scanning point clouds addressing those issues. In line with photogrammetric restitution, lines, not necessarily straight, are reconstructed in a process providing precision, i.e. internal quality measures. Methods for reconstructing different types of lines will be presented, in particular break lines, form lines, step edges, and boundary lines. The process features a high degree of automation, reducing the manual interaction to supplying one approximate point or one short segment per relevant line. The method will be presented with two examples.