Automated interpretation of geological fault surfaces imaged in 3D seismic data

Abstract

A novel approach to the automatic extraction of geological faults from 3D seismic data is described and comparisons of manually and automatically picked fault geometries interpreted from a high-quality 3D seismic dataset are presented. Based on the premise that faults introduce discontinuities to otherwise coherent layers (horizons) in seismic data volumes, prototype software has been developed that employs a coherency measure to detect points of significant discontinuity. An iterative highest confidence first merging strategy then groups clusters of ‘faulty’ points to form many planar segments, or patches, of larger fault surfaces. Because a planar model cannot reflect the curvature and localized irregularities that characterize real fault surfaces, the 3D fault surface geometry is generated using a combined parametric and residual field model to link the fault patches. The parametric model, in our case a planar model, represents the basic structure of the fault surface whilst the residual field models the surface deviation from the planar model. This is one of the few projects of this kind that combines expertise from computer vision and structural geology. We conclude with a discussion on how empirical structural geological criteria describing the geometry of natural fault systems will be integrated with the autopicking software to optimize the geological validity of the interpreted fault geometry.