| CPC G01V 1/302 (2013.01) [G01V 1/50 (2013.01); G06N 20/00 (2019.01); G01V 2210/1234 (2013.01); G01V 2210/64 (2013.01)] | 20 Claims |

|
1. A method, comprising:
training a first machine learning model using data from a crossline direction and an inline direction of a seismic volume;
determining, using the first machine learning model, a top of salt (TOS) surface in the seismic volume based on crossline direction slices of the seismic volume and inline direction slices of the seismic volume;
determining a binary mask based upon the determined TOS surface;
initializing training data of the seismic volume based on output from the trained first machine learning model;
determining a depth in the seismic volume above which a predicted salt surface has a predetermined level of clarity;
sampling seismic data in the seismic volume in a depth direction to obtain a training seismic slice, wherein the training seismic slice includes information from the crossline direction, the inline direction, and the depth direction;
sampling the binary mask in the depth direction to obtain a mask slice;
selecting a first coordinate in the training seismic slice to produce a first tile;
selecting a second coordinate in the mask slice to produce a second tile;
updating the training data of the seismic volume based upon the first tile and the second tile;
training a second machine learning model using the updated training data;
determining, using the second machine learning model, a presence of a salt body in an evaluation seismic slice;
generating, by the second machine learning model, a three-dimensional (3D) model of a subterranean formation with the salt body delineated in the seismic volume;
displaying the 3D model of the subterranean formation with the salt body in the seismic volume;
determining, based on the 3D model of the subterranean formation, fluid flow characteristics of a reservoir in a field of the subterranean formation; and
controlling, based on the 3D model of the subterranean formation, drilling in the reservoir.
|