Semi-automatic interpretation of 3D-Sedimentological Structures modeling from Geological Images.

in: Proc. 34th Gocad Meeting, Nancy

Abstract

The characterization of sedimentary structures from quantitative analysis of outcrops is important to understand their genetic controls and to provide subsurface analogs. However, quantitative interpretation of these structures generally consist in line pickings, which can be tedious due to the possible abundance of objects. Meanwhile, a vast literature deals with automatic image interpretations, particularly in the medical domain. This type of approaches are usually adapted for high quality images with little noise, hence are difficult to apply to seismic images. In this context we propose to enable the user to initialize the shape and position of a template form, such that it is a first rough interpretation of the object to be interpreted. These templates are constructed using Non-Uniform Rational B-Splines (NURBS). NURBS are a compact parametrizations based on a restricted number of points and can be either globally and locally deformed. Information about the sedimentary structure shapes are extracted from the image using a Canny Edge detection approach. The edges are postprocessed in order to strengthen the most relevant ones according to their orientations and distances to the user defined initialization. An automatic optimization is then performed using these image parameters to fit the template form to the actual object traces on the image. We apply this approach on satellite pictures showing alluvial channels and use a retro-deformation method to automatically reconstruct the point bars. This method shows good potential to help interpreters rapidly come up with 3D analog models of sedimentary structures.

Download / Links

BibTeX Reference

@inproceedings{Ruiu2GM2014,
 abstract = { The characterization of sedimentary structures from quantitative analysis of outcrops is important to understand their genetic controls and to provide subsurface analogs. However, quantitative interpretation of these structures generally consist in line pickings, which can be tedious due to the possible abundance of objects. Meanwhile, a vast literature deals with automatic image interpretations, particularly in the medical domain. This type of approaches are usually adapted for high quality images with little noise, hence are difficult to apply to seismic images. In this context we propose to enable the user to initialize the shape and position of a template form, such that it is a first rough interpretation of the object to be interpreted. These templates are constructed using Non-Uniform Rational B-Splines (NURBS). NURBS are a compact parametrizations based on a restricted number of points and can be either globally and locally deformed. Information about the sedimentary structure shapes are extracted from the image using a Canny Edge detection approach. The edges are postprocessed in order to strengthen the most relevant ones according to their orientations and distances to the user defined initialization. An automatic optimization is then performed using these image parameters to fit the template form to the actual object traces on the image. We apply this approach on satellite pictures showing alluvial channels and use a retro-deformation method to automatically reconstruct the point bars. This method shows good potential to help interpreters rapidly come up with 3D analog models of sedimentary structures. },
 author = { Ruiu, Jeremy AND Caumon, Guillaume AND Viseur, Sophie AND Antoine, Christophe },
 booktitle = { Proc. 34th Gocad Meeting, Nancy },
 title = { Semi-automatic interpretation of 3D-Sedimentological Structures modeling from Geological Images. },
 year = { 2014 }
}