Rock mass characterization is very important in rock engineering. The characterization encompasses the description of the apparent joint network as well as the spatial distribution of ground types or lithologies. Currently, the rock mass is documented and mapped either on-sight or with digital mapping-software like ShapeMetriX3D on the desktop. In the latter, digital images are used to generate 3D point clouds and manually map the joint planes or traces. With an increasing trend to automate the mapping procedure, the analysis of the digital images experiences a digital rebirth. In this paper, digital image processing is applied to detect edges in the images of two selected tunnel faces, clusters the signals and a structure map is generated which can be used for determining joint orientations, spacings and trace lengths. The applied method resulted in both cases in a very good detection joint traces, which could be distinguished into four distinct joint sets in the first step and reduced into three joint set in the second clustering process. In total, 743 (case study I) and 1233 (case study II) joint traces were detected automatically. However, in the current application, the segment linking process is incomplete and leads to scattering of the orientation values due to short line segments. Additionally, no effort was made yet to exclude artificial color changes, like caused by chiselling, from the analysis. However, with the large number of measurements, their influence is considered negligible. The applied method shows strengths especially in detecting geological features, which do not per se occur as joint planes and which will be missed in an automated vector analysis of the point clouds.
The extraction of joint traces from photographs of rock masses is nothing new in rock mass characterization (e.g. Franklin et al., 1988, Reid & Harrison, 2000, Lemy & Hadjigeorgiou, 2003, Deb et al., 2008). However, the analyses were restricted in dimension (2D) and heavily affected by different light and rock mass conditions. With the up come of photogrammetry in rock mass characterization and an increased trend for automatic rock mass characterization (e.g. Kemeny & Post, 2003, van Knapen & Slob, 2006, Riquelme et al., 2014), digital image processing experiences a digital rebirth for rock engineering purposes (Vasuki et al., 2014, 2017), still facing the old problems (e. g. exposition and light condition, rock mass conditions), but with the possibility to include spatial analyses like the determination of the joint trace and plane orientations in point clouds.