Methods and apparatus for automatic identification of faults on noisy seismic data

Abstract

Methods and apparatus are provided for automatically identifying possible faults in large seismic datasets. An exemplary method comprises obtaining the seismic data; calculating a coherence cube of the seismic data; performing the following steps for a plurality of two-dimensional seismic sections of the coherence cube: (i) applying a threshold to the coherence cube to obtain a binary image representation comprising continuities; (ii) identifying edges of continuity areas in the binary image representation to identify changes in the continuities as fault point candidates; (iii) identifying fault points in the obtained seismic data based on a fault confidence value indicating a likelihood that a given point is part of a fault; (iv) creating one or more fault segments from the identified fault points; and (v) joining fault segments into fault lines using geological and/or geometrical constraints; and generating three-dimensional fault surfaces from the fault lines in the plurality of two-dimensional seismic sections. The exemplary automatic fault detection method can be parallelized.