# US 11,079,509 B1 · Building accurate training images for automatic seismic interpretation

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 11,079,509 B1
- Document ID: US11079509B1
- Status: Granted
- Kind: B1
- Application number: 17/152,329
- Filed: 2021-01-19
- Granted: 2021-08-03
- Earliest claimed: 2021-01-19
- Assignees: Emerson Paradigm Holding LLC
- Applicants: Emerson Paradigm Holding LLC
- Inventors: Anne-Laure Tertois; Jean-Laurent Mallet; Zvi Koren
- Web page: https://patentgrep.com/patent/US11079509B1

## CPC

- G01V 1/301: {for determining seismic cross-sections or geostructures}
- G01V 1/325: {Transforming one representation into another}
- G01V 1/345: {Visualisation of seismic data or attributes, e.g. in 3D cubes}
- G01V 20/00: Geomodelling in general
- G06F 18/214: Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06N 3/0464: Convolutional networks [CNN, ConvNet]
- G06N 3/08: Learning methods
- G06N 3/09: Supervised learning
- G06N 20/00: Machine learning
- G06T 3/06: Topological mapping of higher dimensional structures onto lower dimensional surfaces
- G06V 10/255: {Detecting or recognising potential candidate objects based on visual cues, e.g. shapes}
- G06V 10/7715: {Feature extraction, e.g. by transforming the feature space, e.g. multi-dimensional scaling [MDS]; Mappings, e.g. subspace methods}
- G06V 10/774: Generating sets of training patterns; Bootstrap methods, e.g. bagging or boosting
- G06V 20/64: Three-dimensional [3D] objects
- G01V 2210/64: Geostructures, e.g. in 3D data cubes
- G06N 20/10: using kernel methods, e.g. support vector machines [SVM]
- G01V: GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS (means for indicating the location of accidentally buried, e.g. snow-buried, persons A63B29/02)
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)
- G06N: COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06T: IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06V: IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING

## Abstract

Automatic propagation of real-world parent seismic images to efficiently generate a collection of realistic synthetic child training images to train a model for accurate automatic seismic interpretation. A 3D structural model in a present-day geological space (e.g., GB) depicting subsurface locations of particles (e.g., in region B) may be transformed by a 3D coordinate space transformation (e.g., uvtB) to a depositional space (e.g., G*B) depicting past depositional locations of those particles (e.g., corresponding depositional region B). A real-world parent image depicting subsurface locations of particles (e.g., in region A) may be transformed, via a forward transformation (e.g., uvtA), to a depositional seismic image in the depositional space of the three-dimensional structural model (e.g., G*A=G*B). A reverse transformation (e.g., uutB−1) may transform the depositional seismic image from the depositional space into synthetic child training images in the present-day geological space (e.g., GB) for training the model.

## Claims

### Claim 1

A system for automatically generating seismic training images and associated seismic interpretation training labels to train a model for automatic seismic interpretation, the system comprising:
one or more processors configured to:
receive a three-dimensional structural model in a present-day geological space depicting locations of particles of the subsurface geology of the Earth as observed at a current time period,
compute a three-dimensional coordinate space transformation to transform the three-dimensional structural model from the present-day geological space to a depositional space depicting past locations of the particles at times when the particles were determined to be originally deposited within the subsurface geology of the Earth,
receive a real-world parent image of the subsurface geology in a present-day geological space generated based on seismic data captured by sensors recording seismic waves reflected from the subsurface geology,
apply a forward transformation to transform the real-world parent image in the present-day geological space to a depositional seismic image in the depositional space of the three-dimensional structural model,
apply a reverse transformation to transform the depositional seismic image from the depositional space into a plurality of synthetic child training images in the present-day geological space,
associating the plurality of synthetic child training images with seismic interpretation training labels; and
one or more memories configured to store the plurality of synthetic child training images and associated seismic interpretation training labels.

### Claim 2 (depends on 1)

The system of claim 1, wherein the real-world parent image and each of the plurality of synthetic child training images differ by the addition, deletion, rotation, translation, scaling, location or intersection of one or more geological objects.

### Claim 3 (depends on 1)

The system of claim 1, wherein the one or more processors are configured to generate the plurality of synthetic child training images by applying a plurality of different respective reverse transformations.

### Claim 4 (depends on 3)

The system of claim 3, wherein the one or more processors are configured to apply the plurality of different respective reverse transformations to transform the depositional seismic image to a plurality of different respective regions of the subsurface geology in the present-day geological space.

### Claim 5 (depends on 1)

The system of claim 1, wherein the real-world parent image and the three-dimensional structural model depict different regions of the subsurface geology, such that the forward and reverse transforms map the real-world parent image depicting a first region of the subsurface geology to the plurality of the synthetic child training images representing one or more different regions of the subsurface geology.

### Claim 6 (depends on 1)

The system of claim 1, wherein the one or more processors are configured to generate the plurality of synthetic child training images by using a plurality of different respective configurations of the three-dimensional structural model in the present-day geological space.

### Claim 7 (depends on 1)

The system of claim 1, wherein the one or more processors are configured to generate the plurality of synthetic child training images by using a plurality of different respective configurations of the three-dimensional structural model in the depositional space.

### Claim 8 (depends on 1)

The system of claim 1, wherein the real-world parent image and the three-dimensional structural model depict the same region of the subsurface geology, wherein the plurality of the synthetic child training images represent a plurality of different perturbations induced in the present-day or depositional space.

### Claim 9 (depends on 1)

The system of claim 1, wherein the one or more processors are configured to alter one or more of the synthetic child training images in a post-processing stage.

### Claim 10 (depends on 1)

The system of claim 1, wherein the one or more processors are configured to convert:
a full-size real-world parent image occupying the entire three-dimensional structural model to a plurality of sub-volume synthetic child training images occupying a plurality of respective sub-volumes of the three-dimensional structural model,
a full-size real-world parent image to a plurality of full-size synthetic child training images each occupying the entire three-dimensional structural model, or
a sub-volume real-world parent image occupying a sub-volume of the three-dimensional structural model to the plurality of sub-volume synthetic child training images.

### Claim 11 (depends on 10)

The system of claim 10, wherein the plurality of sub-volume synthetic child training images are distributed at locations in the three-dimensional structural model that are regularly spaced, randomly positioned, or preferentially distributed along one or more geologic objects.

### Claim 12 (depends on 1)

The system of claim 1, wherein the one or more processors are configured to automatically generate the seismic interpretation training labels associated with one or more geologic objects from the three-dimensional structural model.

### Claim 13 (depends on 1)

The system of claim 1, wherein one or more of the seismic interpretation training labels encode information selected from the group consisting of: the presence or absence of a geological object of interest, spatial characteristics of a geological object of interest, and geological-time.

### Claim 14 (depends on 1)

The system of claim 1, wherein, during a training phase, the one or more processors are configured to train the model to perform automatic seismic interpretation using the plurality of the synthetic child training images and associated seismic interpretation training labels.

### Claim 15 (depends on 14)

The system of claim 14, wherein, during a run-time phase, the one or more processors are configured to input a new seismic image into the trained model and output one or more new seismic interpretation training labels associated with the new seismic image.

### Claim 16 (depends on 1)

The system of claim 1, wherein the real-world parent image is derived from seismic data captured from shallow and substantially unfolded and unfaulted parts of a seismic acquisition survey.

### Claim 17 (depends on 1)

The system of claim 1, wherein the real-world parent image is derived from seismic data captured from folded and faulted parts of a seismic acquisition survey.

### Claim 18

A method for automatically generating seismic training images and associated seismic interpretation training labels to train a model for automatic seismic interpretation, the method comprising:
receiving a three-dimensional structural model in a present-day geological space depicting locations of particles of the subsurface geology of the Earth as observed at a current time period;
computing a three-dimensional coordinate space transformation to transform the three-dimensional structural model from the present-day geological space to a depositional space depicting past locations of the particles at times when the particles were determined to be originally deposited within the subsurface geology of the Earth;
receiving a real-world parent image of the subsurface geology in a present-day geological space generated based on seismic data captured by sensors recording seismic waves reflected from the subsurface geology;
applying a forward transformation to transform the real-world parent image in the present-day geological space to a depositional seismic image in the depositional space of the three-dimensional structural model;
applying a reverse transformation to transform the depositional seismic image from the depositional space into a plurality of synthetic child training images in the present-day geological space;
associating the plurality of synthetic child training images with seismic interpretation training labels; and
storing the plurality of synthetic child training images and associated seismic interpretation training labels.

### Claim 19 (depends on 18)

The method of claim 18, wherein the real-world parent image and each of the plurality of synthetic child training images differ by the addition, deletion, rotation, translation, scaling, location or intersection of one or more geological objects.

### Claim 20 (depends on 18)

The method of claim 18 comprising generating the plurality of synthetic child training images by applying a plurality of different respective reverse transformations.

### Claim 21 (depends on 20)

The method of claim 20, wherein applying the plurality of different respective reverse transformations comprises transforming the depositional seismic image to a plurality of different respective regions in the present-day geological space depicting a plurality of different respective parts of the subsurface geology.

### Claim 22 (depends on 18)

The method of claim 18, wherein the real-world parent image and the three-dimensional structural model depict different regions of the subsurface geology, such that the forward and reverse transforms map the real-world parent image depicting a first region of the subsurface geology to the plurality of the synthetic child training images representing one or more different regions of the subsurface geology.

### Claim 23 (depends on 18)

The method of claim 18 comprising generating the plurality of synthetic child training images by using a plurality of different respective configurations of the three-dimensional structural model in the present-day geological space.

### Claim 24 (depends on 18)

The method of claim 18 comprising generating the plurality of synthetic child training images by using a plurality of different respective configurations of the three-dimensional structural model in the depositional space.

### Claim 25 (depends on 18)

The method of claim 18, wherein the real-world parent image and the three-dimensional structural model depict the same region of the subsurface geology, comprising inducing a plurality of different perturbations in the present-day or depositional space to generate respective variations in the plurality of the synthetic child training images.

### Claim 26 (depends on 18)

The method of claim 18 comprising converting:
a full-size real-world parent image occupying the entire three-dimensional structural model to a plurality of sub-volume synthetic child training images occupying a plurality of respective sub-volumes of the three-dimensional structural model,
a full-size real-world parent image to a plurality of full-size synthetic child training images each occupying the entire three-dimensional structural model, or
a sub-volume real-world parent image occupying a sub-volume of the three-dimensional structural model to the plurality of sub-volume synthetic child training images.

### Claim 27 (depends on 18)

The method of claim 18 comprising, during a training phase, training the model to perform automatic seismic interpretation using the plurality of the synthetic child training images and associated seismic interpretation training labels.

### Claim 28 (depends on 18)

The method of claim 18 comprising, during a run-time phase:
inputting a new seismic image into the trained model; and
outputting one or more new seismic interpretation training labels associated with the new seismic image.

### Claim 29

A non-transitory computer readable storage medium having instructions stored thereon which, when executed by one or more processors, causes the processors to:
receive a three-dimensional structural model in a present-day geological space depicting locations of particles of the subsurface geology of the Earth as observed at a current time period;
compute a three-dimensional coordinate space transformation to transform the three-dimensional structural model from the present-day geological space to a depositional space depicting past locations of the particles at times when the particles were determined to be originally deposited within the subsurface geology of the Earth;
receive a real-world parent image of the subsurface geology in a present-day geological space generated based on seismic data captured by sensors recording seismic waves reflected from the subsurface geology;
apply a forward transformation to transform the real-world parent image in the present-day geological space to a depositional seismic image in the depositional space of the three-dimensional structural model;
apply a reverse transformation to transform the depositional seismic image from the depositional space into a plurality of synthetic child training images in the present-day geological space;
associate the plurality of synthetic child training images with seismic interpretation training labels; and
store the plurality of synthetic child training images and associated seismic interpretation training labels.

### Claim 30 (depends on 29)

The non-transitory computer readable storage medium of claim 29, further storing instructions which, when executed by the one or more processors, causes the processors to:
execute a training phase comprising training the model to perform automatic seismic interpretation using the plurality of the synthetic child training images and associated seismic interpretation training labels; and
execute a run-time phase comprising inputting a new seismic image into the trained model, and outputting one or more new seismic interpretation training labels associated with the new seismic image.