ON THE USE OF TERRESTRIAL LIDAR SCANNING FOR SURFACE ROUGHNESS ESTIMATION IN THE GEOSCIENCES

Accurate estimation of surface parameters of planar features of rock outcrops is developing into an important task in the geosciences. The roughness and spatial orientation of fracture planes influence fluid flow properties and failure modes in rock units and record past tectonic movements. Geospatially referenced point cloud datasets representing rock surfaces can be processed with a variety of computer vision techniques to identify planar features and associated roughness information. For a set of points corresponding to a planar feature within a point cloud, roughness can be portrayed as the standard deviation of the residual distances of the points to their assigned least squares fit plane, projected along the plane’s normal direction. Terrestrial Laser Scanning (TLS) has been used to capture point cloud data for surface morphology studies in the geosciences in recent years, and is well suited to the application as it is capable of rapidly and accurately capturing 3D points at ranges up to hundreds of meters. Field procedures for TLS data collection require careful planning and adaptation to the circumstances of each project undertaken, meaning projects must be planned around practical constraints including available scan time and site accessibility. In this research we investigate whether user-influenced aspects of TLS data collection contribute to underestimation and increased uncertainty in measurements of planar surface roughness. A physical target was constructed and scanned under varying conditions of range, incident angle and sample density, and calculated roughness values compared to an a priori value derived from a theoretical model made from the measured target dimensions. More complete knowledge of the relationship between scan parameters and the quality of the final data product will allow researchers to design field procedures which more efficiently achieve the level of accuracy and precision their research requires.