Synthetic Borehole Images From Outcrop Photographs to Improve Subsurface Geological Interpretations

The interpretation of geological features on borehole image logs (BHIs) is not trivial, and it is susceptible to misinterpretation. Since the BHIs are projected on a two-dimensional (2D) cylindrical plane view, planar features will be seen as sinusoids, which is very unusual to those who work on 2D views such as core slabs, road cuts, or photographs. To interpret BHIs with higher confidence, geologists and petrophysicists require field analogs, cores, or at least sidewall cores to calibrate their eyes for rock textures and structures that may be present in this kind of data. However, until now, only the cylindrical X-ray computed tomography log (CT scans) is adequate for being truly compared with a borehole image view. To expand the possibilities of the correlation between outcrop and BHIs, this paper brings a new methodology for the creation of pseudo-borehole images from outcrop photographs. The method consists of Python algorithms that generate a pseudo volume from an outcrop image and extract a pseudo image that respects the borehole dimensions. This method was validated by comparing the pseudo images created from the slab of a core CT scan with the corresponding 2D plane view of the tomography log from the same core surface. This procedure was finally attested in outcrop photographs of different kinds of rocks and improved the interpretation of real borehole images. Alongside, the utilization of this method has shown good results for reproducing planar features and three-dimensional (3D) structures equal to or bigger than the borehole size. In this work, we discuss the scenarios where this method improves the confidence of image analysis, its potential to reduce misinterpretations, and the pitfalls related to the procedure. In addition, we demonstrate how synthetic images can be useful in providing accurate geometric analogs for subsurface reservoir sections by creating suitable representations based on the geometry and textures of outcropping geological structures and their corresponding features in BHIs. Finally, the application of this method also has significant potential in developing input images for machine learning or other artificial computational intelligence systems.