Fault Extraction Using Point Cloud Approach to a Seismic Enhanced Discontinuity Cube

Bounaim, Aicha; Boe, Trond Hellem; Athmer, Wiebke; D'Hamonville, Pierre Tardiff; Sønneland, Lars

In petroleum exploration, faults are very important geological structures as they are often associated with subsurface traps for oil and gas accumulation. Moreover, knowledge of the location of the faults is crucial for an effective planning of drilling sites, for reservoir model building and compartmentalization as well as geo-mechanical modelling.

In general, faults are rarely completely captured by individual attributes across a 3-D seismic dataset, and optimal results are often obtained by combining a number of attributes. For large fault identification, appropriate seismic attributes should allow for multi-scale feature detection. Moreover, extracting a large listric fault in one 3-D surface object as highlighted on the edge volume is still a challenge facing seismic interpreters.

The workflow adopted for fault extraction consists of three main steps: 1) Seismic conditioning or filtering to improve signal-to-noise ratio; 2) Edge detection based on one or combined seismic attributes; 3) Edge enhancement using the AntTracking algorithm (Pedersen, 2002); and 4) Fault patches/surfaces extraction.

Step 2 and Step 3 are combined and improved by inserting a 3-D Radon edge enhancement iteration (Boe, 2012) resulting in the edge cube used in Step 4.

The present work will focus on the step from discontinuity cube to fault patch creation in the special case of large listric faults. To our knowledge, different methods have shown some limitations for this process. In many cases, although well highlighted in the edge cube, large listric faults are not extracted as full complete objects, but rather as a set of small patches that require extra-editing.

The proposed methodology consists in geometrically processing a thresholded version of the edge cube in a point cloud approach. 3-D geometrical and orientation properties are calculated for each non-zero value voxels using neighborhood information. A set of rules is built based on these new attributes - namely azimuth and planarity leading to a sequence of constrained sparse cubes with value of 1 for the discontinuity, and 0 elsewhere. Then, from morphological image processing methods (Haralick et al, 1992), connected and labeled objects are extracted from the new edge cube and converted to three-dimensional point sets each representing a fault. Finally, these point-sets are geologically quality checked before the generation of the final fault objects as 3-D surfaces.

AAPG Search and Discovery Article #90163©2013AAPG 2013 Annual Convention and Exhibition, Pittsburgh, Pennsylvania, May 19-22, 2013