Vectorization, Edge Preserving Smoothing and Dimensioning of Profiles in Laser Scanner Point Clouds

The 3D geometry of an object can be captured very efficiently using a terrestrial laser scanner. By modelling and visualisation of these 3D data, it is possible to obtain vectorized geometric information of the object. Many users prefer to work on profiles, motivated by the handling of paper prints in the field by their own familiarization. Profiles extracted from laser scanner point clouds will inherit the noise characteristics of the original points. The effect of noise can be reduced by applying filtering operations. Straight profile sections can be smoothed by using a straight line filter kernel, while incurved profiles can be smoothed by arcshaped structure elements. Because of the scanner resolution and the beam divergence, edges are usually not measured exactly. The paper presents an edge preserving algorithm to extract and smooth profiles and an approach for automatic vectorization. Smoothed laser scanner data profiles are represented as key points, straight line and arc segments. In addition to the profile vectorization, a CAD-format oriented dimensioning is derived from the data and added to the output. Results from practical applications in cultural heritage documentation and as-built documentation are shown. A profile length comparison between automatically extracted profiles and manually vectorized lines shows a standard deviation σ of 7 mm and a maximum deviation of 1.2 cm.