Method and system to analyze geologic formation properties
Systems and methods for determining properties of subterranean formations surrounding a wellbore are provided. An example method can include receiving an image of a formation sample; partitioning the image into a plurality of patches; detecting, via a semantic extraction processor, textures captured in the plurality of patches; associating the textures to a location of the image of the formation sample; reducing a dimension of representation of the textures to obtain one or more vectors, the one or more vectors being based on the textures; and providing a plurality of curves based on the one or more vectors.
1. A computer-implemented method for analyzing geological formation properties, comprising:
receiving, via a processor, an image of a training sample;
partitioning, via the processor, the image of the training sample into a first plurality of patches;
receiving, via the processor, input including an identification of a texture for each patch of the first plurality of patches;
training, via the processor, a neural network with the plurality of patches and identified textures;
receiving, via the processor, an image of a formation sample to be evaluated;
partitioning, via the processor, the image of the formation sample into a second plurality of patches; and
for each patch of the second plurality of patches:
providing, via the processor, the patch as an input to the neural network;
obtaining, via the processor, values from an output layer of the neural network for the patch, the values of the output layer representing semantic features of the patch in a first number of dimensions;
associating, via the processor, the patch to a location of the image of the formation sample; and
reducing, via the processor, the first number of dimensions representing the semantic features of the patch to obtain a vector corresponding to the patch having a second number of dimensions; and
generating, via the processor, a plurality of curves, one curve being displayed for each of the second number of dimensions, each curve being a two-dimensional graph with one axis corresponding to a position along the formation sample at which the patch is located and the second axis corresponding to a respective dimension of the vector for the corresponding patch;
displaying, via the processor, the plurality of curves on a display device.
2. The method of claim 1 , further comprising imaging the formation sample.
3. The method of claim 1 , further comprising extracting the formation sample from a wellbore.
4. The method of claim 1 , wherein the formation sample is extracted by a core sample collection tool coupled to a conveyance, a sampling-while-drilling tool, or by forming cuttings by drilling and circulating cuttings to a surface.
5. The method of claim 1 , wherein the neural network comprises a convolutional neural network.
6. The method of claim 1 , further comprising displaying the plurality of curves adjacent the image of the formation sample.
7. The method of claim 1 , further comprising dividing the image of the formation sample into texture segments.
8. The method of claim 1 , wherein each patch from the plurality of patches comprises a plurality of pixels associated with the image.
9. The method of claim 1 , wherein each texture represents a classification by the expert of a formation rock type.
10. The method of claim 1 , further comprising determining a statistical representation of the plurality of patches in a horizontal direction based on the textures.
11. The method of claim 1 , wherein each of the vectors comprises at least two components.
12. The method of claim 11 , wherein the at least two components comprise a semantic feature representation and one or more associated location parameters, the semantic feature representation comprising a texture classification.
13. A system comprising:
one or more processors; and
at least one non-transitory computer-readable medium having stored therein instructions which, when executed by the one or more processors, cause the system to:
receive an image of a training sample;
partition the image of the training sample into a first plurality of patches;
receive input including an identification of a texture for each patch of the first plurality of patches;
train a neural network with the plurality of patches and identified textures;
receive an image of a formation sample to be evaluated;
partition the image of the formation sample into a second plurality of patches; and
for each patch of the second plurality of patches:
provide the patch as an input to the neural network;
obtain values from an output layer of the neural network for the patch, the values of the output layer representing semantic features of the patch in a first number of dimensions;
associate the patch to a location of the image of the formation sample; and
reduce the first number of dimensions representing the semantic features of the patch to obtain a vector corresponding to the patch having a second number of dimensions; and
display a plurality of curves, one curve being displayed for each of the second number of dimensions, each curve being displayed as a two-dimensional graph with one axis corresponding to a position along the formation sample at which the patch is located and the second axis corresponding to a respective dimension of the vector for the corresponding patch.
14. The system of claim 13 , wherein neural network comprises a convolutional neural network.
15. The system of claim 13 , the at least one non-transitory computer-readable medium having stored therein instructions which, when executed by the one or more processors, cause the system to render the plurality of curves and the image of the formation sample.
16. The system of claim 13 , wherein each patch from the plurality of patches comprises a plurality of pixels associated with the image, and wherein each texture represents a classification by the expert of a formation rock type.
17. The system of claim 13 , the at least one non-transitory computer-readable medium having stored therein instructions which, when executed by the one or more processors, cause the system to determine a statistical representation of the plurality of patches in a horizontal direction based on the textures.
18. The system of claim 13 , wherein each of the one or more vectors comprises at least two components, the at least two components comprising a semantic feature representation and one or more associated location parameters, the semantic feature representation comprising a texture classification associated with a respective texture.
19. A non-transitory computer-readable medium having stored therein instructions which, when executed by one or more processors, cause the one or more processors to:
receive an image of a training sample;
partition the image of the training sample into a first plurality of patches;
receive input including an identification of a texture for each patch of the first plurality of patches;
train a neural network with the plurality of patches and identified textures;
receive an image of a formation sample to be evaluated;
partition the image of the formation sample into a second plurality of patches; and
for each patch of the second plurality of patches:
provide the patch as an input to the neural network;
obtain values from an output layer of the neural network for the patch, the values of the output layer representing semantic features of the patch in a first number of dimensions;
associate the patch to a location of the image of the formation sample; and
reduce the first number of dimensions representing the semantic features of the patch to obtain a vector corresponding to the patch having a second number of dimensions; and
display a plurality of curves, one curve being displayed for each of the second number of dimensions, each curve being displayed as a two-dimensional graph with one axis corresponding to a position along the formation sample at which the patch is located and the second axis corresponding to a respective dimension of the vector for the corresponding patch.
20. The non-transitory computer-readable medium of claim 19 , wherein the textures are detected via a convolutional neural network, wherein each patch from the plurality of patches comprises a plurality of pixels associated with the image, and wherein each texture represents a classification by the expert of a formation rock type.
21. The non-transitory computer-readable medium of claim 19 , storing instructions which, when executed by the one or more processors, cause the one or more processors to:
determine a statistical representation of the plurality of patches in a horizontal direction based on the textures; and
present the plurality of curves and the image of the formation sample.
22. The non-transitory computer-readable medium of claim 19 , wherein each of the one or more vectors comprises at least two components, the at least two components comprising a semantic feature representation and one or more associated location parameters, the semantic feature representation comprising a texture classification associated with a respective texture.