The Automated Interpretation of Photorealistic Outcrop Models
Erik Monsen, David William Hunt, Aicha Bounaim, Bérengère Savary-Sismondini, Trond Brenna, Michael Nickel, Lars Sonneland, John Bernard Thurmond, Paul Gillespie, 2011. "The Automated Interpretation of Photorealistic Outcrop Models", Outcrops Revitalized: Tools, Techniques and Applications, Ole J. Martinsen, Andrew J. Pulham, Peter D.W. Haughton, Morgan D. Sullivan
Download citation file:
During the last decade, the laser scanning of outcrop analogues for the construction of 3D geologic models has become commonplace. Many elements of both laser-scan data collection and photorealistic model creation have dedicated software and are semi-automated or fully automated. However, much of the subsequent geological interpretation process relies heavily on manual digitization. Consequently, only a small fraction of the inherent 3D and 2D information captured within photorealistic models is normally extracted and analyzed. This situation is perhaps most acute in the study of fractured reservoir analogues where hundreds of thousands or even millions of fractures can be captured in multi-kilometer-scale datasets, making full manual digitization of the dataset impractical.
In order to maximize and standardize the extraction of quantitative data from digital outcrop models, we sought to develop new technologies specifically for the automated extraction of lithologic information, bedding, and fracture discontinuities from photorealistic data. This is achieved through the adaption of algorithms originally developed for automated fault interpretation from 3D seismic data. Projection of the 2D bedding and fracture discontinuities extracted from digital outcrop images onto 3D outcrop surface, using transform matrices calculated during the construction of a photorealistic model, permits calculation of the strike, dip, and a confidence value for every extracted feature. The high-resolution image-derived discontinuity data is then integrated with larger-scale 3D discontinuities extracted from topologic analysis of laser-scan-derived data. This combined approach maximizes the potential information from both (1) high-resolution 2D digital image data and (2) the 3D laser scan topologic data in order to extract and combine different scales of discontinuities into a single 3D dataset.
Independent quality control of the automated bedding and fracture extraction methods is achieved through the third-party collection, analysis, and comparison of structural data collected in the test area. Comparison between automatically extracted and field-mapped and measured fractures demonstrates the accuracy of the methods developed and some limitations. The automated techniques developed are designed not to replace outcrop studies but to augment them. They limit the tedious and time-consuming manual interpretation tasks by providing unbiased, pre-generated geological primitives that can be edited and analyzed with a set of structural-analysis tools. These data and tools allow time to be better spent on the analysis of large numbers of quantitative data and their incorporation into 3D models. The main limitations on quality of the automatic mapping appear to be related to the coverage and quality of the original dataset.