A computer vision system based on digital panoramic images is used to overcome current problems in geotechnical data acquisition. High resolution images enable spatial measurement of geometric entities of a rock mass, like discontinuity lengths, the size of a region, discontinuity orientations, and roughness parameters. Besides, the images represent an objective documentation of the visual rock mass conditions. A system based a digital panoramic camera was tested on-tunnel construction sites and quarries in Austria.
The collection and evaluation of geotechnical data is vital for any rock mechanical analysis. Present field work often relies on measurements using compass-clinometer devices and measuring tapes combined with manual sketches and eventually conventional photographs. This requires tactile access to the exposed rock mass which can be hazardous and time-consuming (Fasching 2000). Besides, this data acquisition process is strongly subjective and, what is even more problematic, it is not a comprehensive documentation of the rock mass. If for any reason the rock mass is altered, e.g. by erosion or excavation work, the original conditions are lost. However, the actual rock mass conditions are preserved, if visual data at a sufficient quality are acquired (Gaich 2000). Generally, the results from current geotechnical data acquisition are incomplete, mostly Incorrect (due they do not have metrics) and often inconsistent. Anyhow, further rock mass analyses rely on these data.
Computer vision can be understood as connecting a computer with (digital) cameras and processing the images on it in order to get information on the recorded scene or objects. Three- dimensional computer vision (3d vision) is inspired by human perception and tries either to reconstruct objects from images or to understand the contents of a scene (Sonka et al. 1999). Figure 1 outlines the approach. An imaging system is used to generate high quality digital images. The images are processed with software components that allow an automatic reconstruction of the three-dimensional shape (surface) and to interactively annotate rock mass structures by means of a stereoscopic inspection tool. Various approaches for 3d vision systems were introduced in the past and many algorithms were established (Faugeras 1993) and it showed to be no universal solution for the manifold applications which themselves are strongly liable for the configuration of a 3d vision system and the used principles. As one can imagine the image quality is an essential design criterion for a computer vision system.
The imaging source is a digital panoramic line-scan camera. It has a single CCD-line sensor for each of the three colour channels and has to be rotated in order to get an image. This scanning principle implies that only static scenes or objects can be recorded. The camera is mounted on a tripod and controlled by standard notebook computer that is eventually adapted for the application in rough environments. The whole system weighs less that 10 kilogramme, thus is can be transported and set up by a single person easily.