Fault detection based on seismic data interpretation

1. A method for determining a position of a geological feature in a formation comprising:

receiving, with one or more seismic sensors, signals from within the formation, wherein the signals are based on seismic waves generated by a seismic source;

determining a seismic dataset, wherein the seismic dataset is based on the signals;

determining, with an edge detector, a set of indicators of candidate discontinuities in the formation based on the seismic dataset;

labeling, with a neural network processor, a subset of the set of indicators of candidate discontinuities using a neural network with a label based on the set of indicators of candidate discontinuities, wherein the label distinguishes an indicator of a candidate discontinuity between being an indicator of a target discontinuity or being an indicator of a non-target discontinuity; and

determining the position of the geological feature in the formation, wherein the geological feature in the formation is associated with at least one target discontinuity based on the subset of the set of indicators of candidate discontinuities.

2. The method of claim 1 , wherein the determining the set of indicators of candidate discontinuities based on the seismic dataset comprises applying a phase congruency operation on the seismic dataset.

3. The method of claim 1 , wherein labeling the subset of the set of indicators of candidate discontinuities comprises:

generating a convoluted dataset based on one or more convolution layers of the neural network, wherein the one or more convolution layers are applied to at least one of the seismic dataset and the set of indicators of candidate discontinuities; and

labeling the subset of the set of indicators of candidate discontinuities using one or more rectified linear units layers of the neural network based on the convoluted dataset.

4. The method of claim 1 , wherein determining the set of indicators of candidate discontinuities comprises partitioning the seismic dataset into seismic subsamples.

5. The method of claim 1 , wherein labeling the subset of the set of indicators of candidate discontinuities comprises performing a binary classification operation using the neural network.

6. The method of claim 1 , wherein the geological feature is at least one of a fracture and a geological fault.

7. The method of claim 6 , wherein labeling the subset of the set of indicators of candidate discontinuities comprises distinguishing an indicator of a candidate discontinuity between being an indicator of a target discontinuity or an indicator of a signal reflector from the formation.

8. One or more non-transitory machine-readable media comprising program code for determining a position of a geological feature in a formation, the program code to:

receive, with one or more seismic sensors, signals from within the formation, wherein the signals are based on seismic waves generated by a seismic source;

determine a seismic dataset, wherein the seismic dataset is based on the signals;

determine, with an edge detector, a set of indicators of candidate discontinuities in the formation based on the seismic dataset;

label, with a neural network processor, a subset of the set of indicators of candidate discontinuities using a neural network with a label based on the set of indicators of candidate discontinuities, wherein the label distinguishes an indicator of a candidate discontinuity between being an indicator of a target discontinuity or being an indicator of a non-target discontinuity; and

determine the position of the geological feature in the formation, wherein the geological feature in the formation is associated with at least one target discontinuity based on the subset of the set of indicators of candidate discontinuities.

9. The machine-readable media of claim 8 , wherein the program code to determine the set of indicators of candidate discontinuities based on the seismic dataset comprises program code to apply a phase congruency operation on the seismic dataset.

10. The machine-readable media of claim 8 , wherein the program code to label the subset of the set of indicators of candidate discontinuities comprises program code to:

generate a convoluted dataset based on one or more convolution layers of the neural network, wherein the one or more convolution layers are applied to at least one of the seismic dataset and the set of indicators of candidate discontinuities; and

label the subset of the set of indicators of candidate discontinuities using one or more rectified linear units layers of the neural network based on the convoluted dataset.

11. The machine-readable media of claim 8 , wherein the program code to determine the set of indicators of candidate discontinuities comprises program code to partition the seismic dataset into seismic subsamples.

12. The machine-readable media of claim 8 , wherein the program code to label the subset of the set of indicators of candidate discontinuities comprises program code to perform a binary classification operation using the neural network.

13. The machine-readable media of claim 8 , wherein the geological feature is at least one of a fracture and a geological fault.

14. The machine-readable media of claim 8 , wherein the program code to label the subset of the set of indicators of candidate discontinuities comprises program code to distinguish an indicator of a candidate discontinuity between being an indicator of a target discontinuity or an indicator of a signal reflector from the formation.

15. An apparatus comprising:

one or more seismic sensors;

a processor; and

a machine-readable medium having program code executable by the processor to cause the apparatus to,

receive, with the one or more seismic sensors, signals from within a formation, wherein the signals are based on seismic waves generated by a seismic source

determine a seismic dataset, wherein the seismic dataset is based on the signals,

determine, with an edge detector, a set of indicators of candidate discontinuities in the formation based on the seismic dataset,

label, with a neural network processor, a subset of the set of indicators of candidate discontinuities using a neural network with a label based on the set of indicators of candidate discontinuities, wherein the label distinguishes an indicator of a candidate discontinuity between being an indicator of a target discontinuity or being an indicator of a non-target discontinuity, and

determine a position of a geological feature in the formation, wherein the geological feature in the formation is associated with at least one target discontinuity based on the subset of the set of indicators of candidate discontinuities.

16. The apparatus of claim 15 , wherein the program code to determine the set of indicators of candidate discontinuities based on the seismic dataset comprises program code to apply a phase congruency operation on the seismic dataset.

17. The apparatus of claim 15 , wherein the program code to label the subset of the set of indicators of candidate discontinuities comprises program code to:

generate a convoluted dataset based on one or more convolution layers of the neural network, wherein the one or more convolution layers are applied to at least one of the seismic dataset and the set of indicators of candidate discontinuities; and

label the subset of the set of indicators of candidate discontinuities using one or more rectified linear units layers of the neural network based on the convoluted dataset.

18. The apparatus of claim 15 , wherein the program code to determine the set of indicators of candidate discontinuities comprises program code to partition the seismic dataset into seismic subsamples.

19. The apparatus of claim 15 , wherein the program code to label the subset of the set of indicators of candidate discontinuities comprises program code to perform a binary classification operation using the neural network.

20. The apparatus of claim 15 , wherein the geological feature is at least one of a fracture and a geological fault.