Multi-scale Visualisation and Automated Interpretations of Digital Outcrop

Clément Begotto and Guillaume Caumon. ( 2016 )
in: 2016 RING Meeting, ASGA

Abstract

Digital Outcrop Models (DOM) have a potential to quantify field observations and to propose virtual field trips. However the interpretation of DOMs remains a practical challenge because no dedicated software exists to manage, analyze and come up with 3D models consistent with such large data sets. This paper reports on our work to efficiently visualize DOMs at the appropriate level of detail in a geomodeling package, and describes our first steps towards the supervised inter- pretation of DOMs. For this, we integrated two open-source libraries within the SKUA-GOCAD TM geomodeling environment: OpenScenegraph (OSG) loads and manages level of details of complex triangulated outcrop models, and OpenCV is used to interpret these data with machine learning. We describe how the graphics environment of SKUA-GOCAD and OSG, both relying on OpenGL, were combined. We also show a preliminary application of support vector machines to classify the nodes of a DOM based on simple categories identified on a training set. The obtained results suggest that further investigations are needed in the selection of the support vector machine parameters and in the selection of DOM data.

Download / Links

BibTeX Reference

@INPROCEEDINGS{,
    author = { Begotto, Clément and Caumon, Guillaume },
     title = { Multi-scale Visualisation and Automated Interpretations of Digital Outcrop },
 booktitle = { 2016 RING Meeting },
      year = { 2016 },
 publisher = { ASGA },
  abstract = { Digital Outcrop Models (DOM) have a potential to quantify field observations and to propose
virtual field trips. However the interpretation of DOMs remains a practical challenge because no
dedicated software exists to manage, analyze and come up with 3D models consistent with such
large data sets. This paper reports on our work to efficiently visualize DOMs at the appropriate
level of detail in a geomodeling package, and describes our first steps towards the supervised inter-
pretation of DOMs. For this, we integrated two open-source libraries within the SKUA-GOCAD TM
geomodeling environment: OpenScenegraph (OSG) loads and manages level of details of complex
triangulated outcrop models, and OpenCV is used to interpret these data with machine learning.
We describe how the graphics environment of SKUA-GOCAD and OSG, both relying on OpenGL,
were combined. We also show a preliminary application of support vector machines to classify the
nodes of a DOM based on simple categories identified on a training set. The obtained results suggest
that further investigations are needed in the selection of the support vector machine parameters
and in the selection of DOM data. }
}