A method for determining inherent feature(s) of a geological formation and related computer program product

Abstract

The present invention concerns a method (100) for determining inherent feature(s) of a geological formation based on a structured database and on a seismic image, the method being implemented by a computer and comprising: - obtaining (110) the structured database comprising : - receiving (120) the seismic image, and an information corresponding to said seismic image. - identifiying (130), based on the received information, corresponding geological data in the structured database, - selecting (140) the model(s) aiming at being applied to the seismic image, based on the identified geological data, - parametrizing (150) said model(s), based on the identified geological data, - applying (160) the parameterized model(s) to the received seismic image to obtain output(s) characterizing the seismic image, and - determining (170) the inherent feature(s) of the geological formation corresponding to the seismic image by comparing the obtained output(s) to the geological data of the structured database.

Description

[0001] A method for determining inherent feature(s) of a geological formation and related computer program product

[0002] The present invention concerns a method for determining inherent feature(s) of a geological formation based on a structured database and on a seismic image.

[0003] The present invention in addition concerns a computer program product, comprising software instructions which, when being executed by a computer, implement such a method.

[0004] The invention concerns the domain of automatic analysis of seismic image.

[0005] Seismic images are images obtained by emitting soundwaves in a ground and measuring the reflection of these soundwaves on the different parts of the substructure.

[0006] It is known to apply machine learning based computer vision models to seismic images for detecting a predetermined geological event. The result of such methods will be series of geological events (objects) highlighted in the seismic image without any link to the real geological context behind the seismic image.

[0007] There is an important difference between this result and the conventional seismic interpretation performed by geoscientists:

[0008] Actually, the geoscientist starts by detecting the objects based on the geological context, while the machine learning model only tries to identify a predetermined event in an image no matter if it makes sense to find such event in such image or not.

[0009] The current object detection methods may be quick and effective for finding the possible geological events, but since they are unaware of the geological context, all results need to be checked by a geoscience expert. Considering that automated methods produce many results, this verification can take a lot of time for the experts.

[0010] Thus, it exist a need for a method able to simplify the determination of inherent feature(s) relative to a geological formation taking into account the geoscientists knowledge.

[0011] To tackle this issue, the present invention concerns a method for determining inherent feature(s) of a geological formation based on a structured database and on a seismic image, the method being implemented by a computer and comprising the following steps:

[0012] - obtaining (1 10) the structured database comprising : o a plurality of geological data, the geological data being related to a geological formation’s types and / or to the location in the world of said geological formation’s types, and o at least one parametrizable model, the or each model been associated with at least a geological datum in the structured database, said geological datum(a) defining a list of parameter to be selected for the model, the plurality of geological data comprising parameter(s) to parametrize each model,

[0013] - receiving (120): o the seismic image, and o an information corresponding to said seismic image, the information being related to a location in the world of said geological formation and / or to the geolocation formation’s type of said geological formation,

[0014] - identifiying, based on the received information, corresponding geological data in the structured database,

[0015] - selecting the model(s) aiming at being applied to the seismic image, based on the identified geological data,

[0016] - parametrizing said model(s), based on the identified geological data,

[0017] - applying the parameterized model(s) to the received seismic image to obtain output(s) characterizing the seismic image, and

[0018] - determining the inherent feature(s) of the geological formation corresponding to the seismic image by comparing the obtained output(s) to the geological data of the structured database.

[0019] The method makes it possible to interpret the seismic image and to identify inherent feature(s) relative to the geological formation without requiring the analysis of a geological expert. Consequently, the invention makes it possible to simplify said determination of inherent feature(s).

[0020] According to some embodiments, the invention comprises one or more of the following features, taken solely, or according to every possible combination:

[0021] - the inherent feature(s) of the geological formation comprises a first inherent feature locating, in the seismic image, a geological event;

[0022] - the geological event is a specific facies type of the geological formation at a specific location of the seismic image;

[0023] - the inherent feature(s) is / are chosen in a group comprising: o location(s) of dipping events in the geological formation,- direction(s) of seaward advancements in the geological formation, o presence and / or location(s) of downlaps in the geological formation, and o presence and / or location(s) of growth accumulation in the geological formation;

[0024] - the structured database comprises a plurality of graphs, each graph comprising: a central node, a plurality of peripheral nodes and a plurality of bounds respectively linking a peripheral node with the central node, each central and peripheral node being formed by a geological datum, the central node corresponding to a geological formation’s type and / or a location, each peripheral node corresponding to a specificity of the geological formation’s type and / or a location, that is quantifiable or qualifyable, each bound corresponding to a relationship between the specificity and the geological formation’s type and / or a location;

[0025] - during the step of parametrizing the model(s), parameter(s) of the model(s) is / are determined based on the specificity(ies) of the peripheral node(s) linked to the central node corresponding to the received information;

[0026] - the method further comprises the step of displaying a labeled image comprising the seismic image with at least one of the inherent feature(s) represented on the seismic image;

[0027] - the method further comprises a step of interacting with the labeled image comprising: o receiving, from a user, an instruction related to at least one inherent feature, and o displaying, on the labeled image, a response to the instruction resulting from the treatment of the instruction, based on the inherent feature(s);

[0028] - the or each model is based on : o neural network(s), o decision tree(s), o Gaussian process(es), and / or o Markov decision process(es);

[0029] - the step of obtaining the structured data base comprises the following sub-steps: o receiving, for a plurality of geological formation’s types and / or a plurality of location, a description of structural specificities of said formation’s type or location, o for each geological formation’s type and / or a location, generating a graph comprising a central node, a plurality of peripheral nodes and a plurality of bounds respectively linking a peripheral node with the central node, from the received description,

[0030] ■ the central node corresponding to the geological formation’s type and / or a location,

[0031] ■ each peripheral node corresponding to a respective specificity of the geological formation’s type and / or a location that is evaluable by a respective inherent feature, ■ each bound corresponding to a semantic relationship, in the description, between the specificity and the geological formation’s type and / or a location, and o for at least one specificity, obtaining a respective parametrizable model able to obtain value(s) evaluating the corresponding feature, o forming the structured data base by aggregating the graphs (30) and the model(s);

[0032] - the parametrizable model(s) comprise a model configured to detect geometrical shape in the seismic image; and

[0033] - the parametrizable model(s) comprise a model configured to evaluate a value of a seismic attribute.

[0034] The present invention also concerns a computer program product comprising software instructions which, when being executed by a computer, implement such a method

[0035] The invention will be better understood upon reading the following description, illustrating examples of embodiments of the invention, and in reference to the attached figures:

[0036] - [Fig. 1] Figure 1 illustrate a computer system implementing a method according to the invention, for determining inherent feature(s) of a geological formation;

[0037] - [Fig. 2] Figure 2 illustrate a flowchart of an embodiment of the method according to the invention,

[0038] - [Fig. 3] Figure 3 illustrates a graph that is obtained during a step of the method according to an embodiment of the invention,

[0039] - [Fig. 4] Figure 4 illustrates an example of an image received during a receiving step of an embodiment of the method according to the invention, and

[0040] - [Fig.5] Figure 5 illustrates a labeled image obtained by the method according to an embodiment of the invention.

[0041] A calculator 10 and a computer program product 12 are illustrated on figure 1 .

[0042] The calculator 10 is preferably a computer.

[0043] More generally, the controller 10 is a computer or computing system, or similar electronic computing device adapted to manipulate and / or transform data represented as physical, such as electronic, quantities within the computing system's registers and / or memories into other data similarly represented as physical quantities within the computing system's memories, registers or other such information storage, transmission or display devices.

[0044] The calculator 10 interacts with the computer program product 12. As illustrated on figure 1 , the calculator 10 comprises a processor 14 comprising a data processing unit 16, memories 18 and a reader 20 for information media. In the example illustrated on figure 1 , the calculator 10 comprises a human machine interface 22, such as a keyboard, and a display 24.

[0045] The computer program product 12 comprises an information medium 26.

[0046] The information medium 26 is a medium readable by the calculator 10, usually by the data processing unit 16. The readable information medium 26 is a medium suitable for storing electronic instructions and capable of being coupled to a computer system bus.

[0047] By way of example, the information medium 26 is a USB key, a floppy disk or flexible disk (of the English name "Floppy disc"), an optical disk, a CD-ROM, a magneto-optical disk, a ROM memory, a memory RAM, EPROM memory, EEPROM memory, magnetic card or optical card.

[0048] On the information medium 26 is stored the computer program 12 comprising program instructions.

[0049] The computer program 12 is loadable on the data processing unit 16 and is adapted to entail the implementation of a method for determining inherent feature(s) of a geological formation, when the computer program 12 is loaded on the processing unit 16 of the calculator 10.

[0050] In the present description, each inherent feature is related to a corresponding geological event. Thus, each inherent feature characterizes an aspect of the geological event, such as : its position in an image, its angle, its surface, its type, among others.

[0051] Preferably, the inherent feature may be all the geological events of sufficient size to enter the resolution zone of a seismic image (i.e. several meters to kilometers in size), their context, the types of rocks in these contexts, the faults and their types of corresponding geometries, which vary depending on the geological context, and the seismic image quality.

[0052] Operation of the calculator 10 will now be described with reference to figure 2, which diagrammatically illustrates an example of implementation of a method for determining inherent feature(s) of a geological formation from a structured database and a seismic image.

[0053] The method comprises a step 1 10 of obtaining a structured database 30.

[0054] Preferably, the structured database is a knowledge-based database. Optionally, the structured database is an ontology. An ontology is a database wherein the data are organized in a specific way and comprising semantic bounds relating the data. The set of organized data and semantic bounds allow to contextualize the data, thus forming knowledge. The structured database comprises a plurality of geological data and at least one parametrizable model.

[0055] The geological data are related to a geological formation’s types and / or to the location in the world of said geological formation’s types.

[0056] Each parametrizable model is associated with at least a geological datum in the structured database 30. Said geological datum(a) defines a list of parameter(s) to be selected for said model.

[0057] Preferably, the step 1 10 of obtaining the structured database 30 comprises a sub-step 112 of receiving, for a plurality of geological formation’s types and / or a plurality of location, a description of structural specificities of said formation’s type or location.

[0058] For example, the description of the structural specificities is a text, written by an expert of said formation or location, detailing the structure of the considered formation and / or the composition of different layers that may be found on said formation or location. Preferably, the description also explains how each specificity may be determined based on a seismic image.

[0059] Optionally, the description follows a specific template.

[0060] As an example, the description is the following: “Progradation slope in carbonate environment of North Libya region. The slope is characterized by a down -lap of lower reflection terminations, sigmoidal clinoforms moderate to high amplitude".

[0061] Preferably, the step 1 10 of obtaining the structured database in addition comprises, for each geological formation’s type and / or a location, a sub-step 1 14 of generating a graph 30 comprising a central node 32, a plurality of peripheral nodes 34 and a plurality of bounds 36 respectively linking a peripheral node 34 with the central node 32, from the received description.

[0062] An example of such a graph 30 is represented in figure 3.

[0063] The central node 32 corresponds to the geological formation’s type and / or a location. As an example, in figure 3, the central node corresponds to the type of “carbonate progradation”.

[0064] Each peripheral node 34 corresponds to a respective specificity of the geological formation’s type and / or a location that is evaluable by a respective feature.

[0065] In figure 3, the peripheral nodes related of the carbonate progradation are the following.

[0066] Firstly, “Seaward advancement” with the respective bound 36 representing the relation “displays”: The “Seaward advancement” node corresponds to the specificity of the carbonate progradation that indicates its direction of growth towards an open sea. Secondly, “Downlaps” with the respective bound 36 representing the relation “can have”. The “Downlaps” node corresponds to the specificity of the carbonate progradation that indicates the termination of its growth against a steeper slope or a basin margin.

[0067] Thirdly, “Growth accumulation” with the respective bound 36 representing the relation “involves”. The “Growth accumulation” node corresponds to the specificity of the carbonate progradation that indicates its vertical and lateral stacking of sediments over time.

[0068] Fourthly, “Dipping events" with the respective bound 36 representing the relation “have”. The “Dipping event” node corresponds to the specificity of the carbonate progradation that indicates a change in its angle of growth relative to the horizontal plane.

[0069] Each bound 36 corresponds to a semantic relationship, in the description, between the specificity and the geological formation’s type and / or a location.

[0070] As an example, to implement the generating sub-step 114, the calculator 10 applies a Natural Language Processing (NLP) technique known per se. The NLP technique is for example able to extract key information of the description and to represent them as a graph, as represented in figure 3.

[0071] For instance, the NLP technique uses the library spaCy, which is an open-source library for NLP in the Python coding language.

[0072] With the description previously given as an example, the NLP technique is for example able to extract the following information:

[0073] Location: North Libya region,

[0074] Kind of formation: Progradation,

[0075] Environment: Carbonate,

[0076] Inherent specificities: sigmoidal clinoforms of moderate to high amplitude, lower reflection terminations.

[0077] The step 110 of obtaining the structured database preferably in addition comprises the sub-step 116 of obtaining, for at least some specif icity(ies) , a respective parametrizable model able to obtain value(s) evaluating the corresponding feature.

[0078] For example, each parametrizable model has been designed with the help of an expert. As an example, the model training has been done on a corporate account of Azure. Azure is a platform providing an environment, composed of virtual machine(s) with GPUs for training a deep learning model.

[0079] Each model is for example a machine learning-based model able to be parametrized based on the geological formation’s type and location, to evaluate the corresponding specif icity(ies). In that case, each model comprises some tunable weights (distinct of the parameters) that have been tuned, based on labelled data, during a training phase implemented previously than the method. Such training of the model is known per se. In other words, the parametrizable model(s) obtained during the obtaining sub-step 116 are already trained.

[0080] As an example, each model is based on neural network(s), decision tree(s), Gaussian process(es), and / or Markov decision process(es).

[0081] A neural network comprises an ordered succession of layers of neurons, each of which takes its inputs from the outputs of the previous layer.

[0082] More precisely, each layer comprises neurons taking their inputs from the outputs of the neurons in the previous layer, or from the input variables for the first layer. Each neuron is also associated with an operation, i.e. a type of processing, to be performed by said neuron within the corresponding processing layer.

[0083] Each layer is linked to the other layers by a plurality of synapses. A synaptic weight is associated with each synapse, and each synapse forms a link between two neurons. Each synaptic weight is preferably a real number, taking on both positive and negative values. In some cases, each synaptic weight is a complex number.

[0084] Each neuron is able to perform a weighted sum of the value(s) received from the neurons of the preceding layer, each value then being multiplied by the respective synaptic weight of each synapse, or link, between said neuron and the neurons of the preceding layer, then to apply an activation function, typically a non-linear function, to said weighted sum, and to deliver at the output of said neuron, in particular to the neurons of the following layer connected to it, the value resulting from the application of the activation function. The activation function introduces non-linearity into the processing performed by each neuron.

[0085] Examples of activation functions include the sigmoid function, the hyperbolic tangent function, the Heaviside function, the Rectified Linear Unit (ReLU) function or the softmax function.

[0086] As an optional extra, each neuron is also able to apply a multiplicative factor, also called bias, to the output of the activation function, and the value output by said neuron is then the product of the bias value and the value output by the activation function.

[0087] For instance, the parametrizable model(s) comprises two kinds of models: model(s) configured to detect geometric shape on an image (also denoted Object Detection Model(s) for geometrical shapes), and model configured to evaluate a value of a seismic attribute (also denoted Machine Learning model for quantitative analysis of seismic attributes).

[0088] The Object Detection Model(s) are configured to detect geometrical shapes on seismic profiles, such as a delimitation between two layers in the seismic image, or dipping shapes. The Object Detection Model(s) are optionally based on Deep Learning and preferably based on Detectron2 framework. Such model are for example trained on historical images that have been labeled by an expert. The Machine Learning models for quantitative analysis are preferably designed by an expert to obtain quantitative outputs that evaluate some specificities of a geological formation’s type or location. For example, one of said model comprises a classifier configured to output a vector of numbers, each number being representative of a probability that said seismic image, or a part of said seismic image, belongs to the class corresponding to said number.

[0089] The obtaining step 1 10 preferably in addition comprises a sub-step 118 of forming the structured database by aggregating the graphs and the model(s).

[0090] In particular, if a specificity is common to different geological formation’s types and / or locations, said specificity is only present once in the structured database and is linked to each of the considered different geological formation’s types and / or locations by a respective bound 36.

[0091] Alternatively, during the obtaining step 1 10, the structured databased is received by the calculator 10 from another system.

[0092] The method then comprises a receiving step 120. During the receiving step 120, the calculator 10 receives the seismic image, such as the seismic image represented on figure 4.

[0093] A seismic image, also denoted seismic section, is an image visualization of seismic data record, which is an acquisition of acoustic wave propagation in the subsurface.

[0094] In other words, the seismic image represents a vertical section of a geological formation, obtained from the analysis of the acoustic waves reflected or refracted by the different layers of rocks. The seismic image allows to visualize the structure and composition of the geological formation, as well as the possible fluid reservoirs that it contains.

[0095] This acoustic waves recording can be processed in a variety of ways to obtain different kinds of seismic images, in order, among other, to:

[0096] - identify rock types,

[0097] - highlight discontinuities in the subsurface (such as faults),

[0098] - to get a better view of reservoir fluid types.

[0099] Such processing may include filtering or applying different algorithms to the recording to obtain images.

[0100] In a seismic image, the x-axis represents the distance, and the y-axis represents the depth. The depth may be expressed in, meter, or in seconds (corresponding to the “double time” also denoted “back and forth”), depending on the processing applied to the seismic data, with a velocity model established by geophysics experts using complex technics. Such relation between a length and time, is made based on the equation v=2d / t, where v is the velocity of the acoustic wave, d the distance reached and t the time back and forth, and where the velocity v may change from a material to another.

[0101] The seismic image is based on the variation of acoustic impedance between the different types of rocks and fluids. The acoustic impedance is the product of the density and the velocity of the acoustic waves in a given medium. When an acoustic wave encounters a boundary between two media with different acoustic impedances, part of the wave is reflected and part of the wave is transmitted. The amplitude of the reflected wave depends on the contrast of acoustic impedance between the two media. The higher the contrast, the brighter the reflection on the seismic image is. The lower the contrast, the darker the reflection on the seismic image is.

[0102] As represented in Figure 4, the "black and white" seismic section highlights acoustic impedance contrasts, in other words, geological level transitions where acoustic waves propagate more or less rapidly, with more or less intensity. If the contrasts are strong, the seismic image would make appear deep black and white contrast zone. If the contrasts are less strong, the image would make appear more homogeneous grey areas or areas tending towards white.

[0103] This provide an indication of the type of rock and / or fluids they may contain (stronger contrasts may indicate geological successions containing more competent ("hard") rocks, or the presence of contacts between several fluids, for example).

[0104] In addition, during the receiving step 120, the calculator 10 receives an information corresponding to said image, the information being related to a location in the world of said geological formation or to the geolocation formation’s type of said geological formation.

[0105] Preferentially, the information corresponding to said image are extracted from the integrated geoscience platform from which the seismic image comes from.

[0106] As an example, the integrated geoscience platform is Sismage-CIG. From this platform, a file in SEG-Y format is obtained. In that case, the information are for example extracted using the Segyio library in Python, to transform it in a format compatible for Python environment, e.g. similar to the structured database.

[0107] In the precedent example, the information preferably comprises the following elements: carbonate environment, progradation and a location located in North Libya region, for example the longitude 22.992 and the latitude 20.895.

[0108] The method 100 then comprises a step 130 of identifying, based on the received information, corresponding geological data in the structured database.

[0109] To this end, the identifying step 130 preferably comprises comparing the receiving information to the structured database and identifying the appropriate graph in the structured database. In the example of the image of figure 4, the corresponding graph 30 is the graph of figure 3 corresponding to carbonate progradation.

[0110] The method 100 in addition comprises a step 140 of selecting the model(s) aiming at being applied to the seismic image, based on the identified geological data.

[0111] In the previous example, the graph 30 corresponding to the identified geological data, i.e. carbonate progradation, indicates, among others dipping events may happen with this geological formation’s type. Consequently, during the selection step 140, a shape detection model is selected.

[0112] The method 100 then comprises a step 150 of parametrizing the selected model(s).

[0113] In the previous example, the shape detection model is parametrized to detect dipping events. In particular, the graph specifies that, for carbonate progradation, said shapes display seaward advancement. Therefore, the selected model is additionally parameterized to recognize shapes associated with seaward advancement.

[0114] As another example if one parametrizable model comprises a classifier, during the parametrizing step 150, a respective sense is given to each class of said classifier based on the graph peripheral nodes 34.

[0115] For example, if a model includes a classifier designed to identify geological formations, the step of parametrizing 150 involves assigning specific weights to features for each class. This assignment is determined by the information contained in the peripheral nodes 34, ensuring that the model is finely tuned to distinguish and categorize various geological features with a good accuracy.

[0116] The step of parameterizing 150 may comprise selecting a classifier among a plurality of stored models, based on the peripheral nodes 34.

[0117] Then, the method 100 comprises a step 160 of applying the parameterized model(s) to the received seismic image to obtain output(s) characterizing the seismic image.

[0118] Preferentially, for the Object Detection Model(s) for geometrical shapes, the outputs correspond to a straight, or curved, line delimiting two parts of the seismic image, or to a region of the seismic image.

[0119] Preferably, for the Machine Learning model for quantitative analysis, the outputs is a number or a vector of numbers respective to said model.

[0120] The method then comprises a step 170 of determining the inherent feature(s) corresponding to the seismic image.

[0121] In the preceding example, the inherent features are preferably chosen in the group comprising :

[0122] - location(s) of dipping events in the geological formation,- direction(s) of seaward advancements in the geological formation, - presence and / or location(s) of downlaps in the geological formation, and

[0123] - presence and / or location(s) of growth accumulation in the geological formation.The determining step 170 corresponds to the interpretation of the output(s) of the parametrized model(s).

[0124] To this end, each output is compared to the geological data of the structured database to contextualize the output.

[0125] Optionally, the inherent feature(s) of the geological formation comprises a first inherent feature locating, in the seismic image, a geological event. Preferably, the location is defined by an area and a frontier of such area, or by a punctual position.

[0126] For example, the geological event is a specific facies type of the geological formation at a specific location of the seismic image.

[0127] A geological facies is a body of rock that has distinct physical, chemical, and biological characteristics from other rocks in the same formation. Facies can be identified by their appearance, composition, fossil content, or other features that reflect the conditions of their formation.

[0128] Alternatively or in a complement, the inherent feature(s) comprises the composition of a layer of the geological formation corresponding to the received seismic image. To this end, optionally, a vector that is output from a classifier trained to determine composition of a layer, is compared to geological data according providing a corresponding composition for each class.

[0129] In an optional complement, the inherent feature(s) in addition comprises the delimitation of said layer in the received seismic image. Optionally, the lines on the image output from a model is compared to the geological data to interpret it as a delimitation between two layers of the geological formation.

[0130] In the progradation carbonate example, one specificity is the angle of said progradation. The output of the corresponding model is a number such as 5. During the determining step 170, the output of said model is for example compared the graph related to the carbonate progradation and, since the corresponding model is related to the specificity: “angle”, it is determined that the angle of the progradation is equal to 5 degrees.

[0131] It is clear that the determining step 170 makes it possible to interpret the outputs of the models to provide an analysis that goes way belong a simple frontier detection, as in the prior art.

[0132] The method 100 then optionally comprises a step 180 of displaying, on a display, the seismic image with at least one of the inherent feature(s) represented on the seismic image, called labeled image. An example of a labeled image is represented in figure 5. As can be seen on figure 5, the seismic image has been interpreted highlighting regions and delimitations, represented by bounding boxes.

[0133] In this specific instance, Figure 5 represents a 400*300-sized tile extracted from a seismic image, for example another seismic image than the one of Figure 4. The bounding boxes serve to delineate instances of dipping events, each accompanied by corresponding confidence scores. The confidence score not only signifies the model's level of certainty regarding the presence of an object within the bounding box but also provides an indication of the model's confidence in the accuracy of its predictions.

[0134] As an example, if the corresponding model is a classifier, the score corresponds to the value of the classifier output vector, corresponding to the class “dip”. The determined inherent feature thus may be the location of said dip zones in the geological formation

[0135] In the example of figure 5, the detected bounding boxes correspond to a facies of type “diping”. This facies can be related to the dipping events mentioned on figure 3. But as there is no detected “growth accumulation”, “seaward advancement”, or “Downlaps” in these zones, there will be no carbonate progradation as described on figure 3.

[0136] As an example, to confirm the occurrence of a geological event, such as carbonate progradation, each feature, represented in Figure 3 as a peripheral node 34, should be checked.

[0137] In a non-represented alternative, the labeled image further comprises some additional tags corresponding to the determined inherent features relative to specificities of the identified geological formation’s type or location.

[0138] In an embodiment, these tags have the form of a short text corresponding to a specificity of the geological formation’s type or location, and optionally an arrow pointing toward the position of said specificity on the labeled image.

[0139] For example, the tags comprises, in a non-exhaustive manner:

[0140] “Seismic artefacts” that are distortions or anomalies in the seismic image that do not reflect the true structure or properties of the subsurface. They may result from various sources, such as noise, multiple reflections, acquisition geometry, processing errors, or velocity variations. Seismic artefacts can affect the quality and interpretation of the seismic image, and therefore need to be identified and corrected or minimized,

[0141] “Location(s) of clastic remobilization” that are areas where sedimentary rocks have been deformed by gravity-driven processes, such as slumping, sliding, or spreading. They are characterized by chaotic seismic facies, disrupted bedding, and variable thickness, and “Location(s) of fault-progradation folds” that a type of fold that forms when a fault propagates through a sedimentary sequence, creating a ramp and a flat geometry.

[0142] Optionally, the method in addition comprises a step 190 of interacting with the labeled image, wherein the calculator 10 receives an instruction from a user. The instruction is preferably an instruction to display a specific content related to inherent feature(s).

[0143] For example, the instruction concerns one inherent feature that is not apparent on the labeled image or a combination of inherent features. As an example, the instructions is to display the detected objects with a specific confidence of the model.

[0144] The step of interacting 190 then comprises the treatment of the instruction by the calculators 10, in combination with the structured database, and the display of response to the received instruction.

[0145] With the method 100 according to the invention, it is possible for a non-expert user to obtain an analysis of a seismic image, contextualized by the geological background of said image. In addition, thanks to the optional interpreting step 190, it is possible for the nonexpert user to obtain specific information regarding the inherent feature(s).

Claims

CLAIMS1.- A method (100) for determining inherent feature(s) of a geological formation based on a structured database and on a seismic image, the method being implemented by a computer and comprising the following steps:- obtaining (110) the structured database comprising : o a plurality of geological data, the geological data being related to a geological formation’s types and / or to the location in the world of said geological formation’s types, and o at least one parametrizable model, the or each model been associated with at least a geological datum in the structured database, said geological datum(a) defining a list of parameter to be selected for the model, the plurality of geological data comprising parameter(s) to parametrize each model,- receiving (120): o the seismic image, and o an information corresponding to said seismic image, the information being related to a location in the world of said geological formation and / or to the geolocation formation’s type of said geological formation,- identifiying (130), based on the received information, corresponding geological data in the structured database,- selecting (140) the model(s) aiming at being applied to the seismic image, based on the identified geological data,- parametrizing (150) said model(s), based on the identified geological data,- applying (160) the parameterized model(s) to the received seismic image to obtain output(s) characterizing the seismic image, and- determining (170) the inherent feature(s) of the geological formation corresponding to the seismic image by comparing the obtained output(s) to the geological data of the structured database.2.- The method (100) according to claim 1 , wherein the inherent feature(s) of the geological formation comprises a first inherent feature locating, in the seismic image, a geological event.3.- The method (100) according to claim 2, wherein the geological event is a specific facies type of the geological formation at a specific location of the seismic image.4.- The method (100) according to any claims 1 to 3, wherein the inherent feature(s) is / are chosen in a group comprising:- location(s) of dipping events in the geological formation,- direction(s) of seaward advancements in the geological formation,- presence and / or location(s) of downlaps in the geological formation, and- presence and / or location(s) of growth accumulation in the geological formation.5.- The method (100) according to any of claims 1 to 4, wherein the structured database comprises a plurality of graphs (30), each graph (30) comprising: a central node (32), a plurality of peripheral nodes (34) and a plurality of bounds (36) respectively linking a peripheral node (34) with the central node (32), each central (32) and peripheral (34) node being formed by a geological datum, the central node (32) corresponding to a geological formation’s type and / or a location, each peripheral node (34) corresponding to a specificity of the geological formation’s type and / or a location, that is quantifiable or qualifyable, each bound (36) corresponding to a relationship between the specificity and the geological formation’s type and / or a location.6.- The method (100) according to claim 5, wherein, during the step of parametrizing the model(s), parameter(s) of the model(s) is / are determined based on the specif icity(ies) of the peripheral node(s) (34) linked to the central node (32) corresponding to the received information.7.-The method (100) according to any of claims 1 to 6, further comprising the step (180) of displaying a labeled image comprising the seismic image with at least one of the inherent feature(s) represented on the seismic image.8.- The method (100) according to claim 7, further comprising a step (190) of interacting with the labeled image comprising:- receiving, from a user, an instruction related to at least one inherent feature, and- displaying, on the labeled image, a response to the instruction resulting from the treatment of the instruction, based on the inherent feature(s).9.- The method (100) according to any of claims 1 to 8, wherein the or each model is based on :- neural network(s),- decision tree(s),- Gaussian process(es), and / or- Markov decision process(es).10.- The method (100) according to any of claims 1 to 9, wherein the step (110) of obtaining the structured data base comprises the following sub-steps: receiving (1 12), for a plurality of geological formation’s types and / or a plurality of location, a description of structural specificities of said formation’s type or location,- for each geological formation’s type and / or a location, generating (114) a graph (30) comprising a central node (32), a plurality of peripheral nodes (34) and a plurality of bounds (36) respectively linking a peripheral node (34) with the central node (32), from the received description, o the central node (32) corresponding to the geological formation’s type and / or a location, o each peripheral node (34) corresponding to a respective specificity of the geological formation’s type and / or a location that is evaluable by a respective inherent feature, o each bound (36) corresponding to a semantic relationship, in the description, between the specificity and the geological formation’s type and / or a location, and- for at least one specificity, obtaining (116) a respective parametrizable model able to obtain value(s) evaluating the corresponding feature,- forming (1 18) the structured data base by aggregating the graphs (30) and the model(s).1 1.- The method (100) according to any of the preceding claims, wherein the parametrizable model(s) comprise a model configured to detect geometrical shape in the seismic image.12.- The method (100) according to any of claims 1 to 1 1 , wherein the parametrizable model(s) comprise a model configured to evaluate a value of a seismic attribute.13.- A computer program product, comprising software instructions which, when being executed by a computer, implement a method according to any of claims 1 to 12.

Images