A seismic horizon tracking method, device and storage medium

Abstract

The application discloses a seismic horizon tracking method, which comprises the following steps: constructing a deep learning training set label containing multiple seismic attributes and seed point information, training an improved neural network by using the training set label, obtaining a preliminary tracking result, and correcting the preliminary tracking result to obtain a final automatic horizon tracking result. The application fully considers the lateral continuity of strata, the vertical stratification, and the actual needs of referring to seed points and marker layer information in the interpretation process, and uses existing geophysical technology and computer technology to combine the seed points, reference layers, position information and multiple seismic attributes into the learning label, and combine the improved DenseNet network and the interpolation correction technology, so that the accuracy of the automatic horizon tracking is effectively improved. The application also discloses a computer device and a readable storage medium.

Description

Technical Field

[0001] This invention relates to the field of seismic data interpretation, and more specifically to a seismic horizon tracking method, device, and storage medium. Background Technology

[0002] Seismic exploration comprises three main stages: acquisition, processing, and interpretation. Stratigraphic interpretation is one of the most time-consuming and unavoidable fundamental stages in seismic data interpretation. High-precision automatic stratigraphic tracking is an urgent need for interpreters and a long-term research focus for methodology researchers. Traditional automatic stratigraphic tracking methods primarily rely on waveform similarity principles, judging the similarity between prediction points and seed points from the perspective of waveform correlation to track prediction points. Due to the limited information utilized, tracking errors are prone to occur when the continuity of seismic phase axes is poor.

[0003] Deep learning is one of the core methods of modern artificial intelligence technology. It uses operations such as multi-layer convolution, activation, pooling, normalization, feedback, and fully connected layers to build complex neural network systems. These systems are used to establish complex nonlinear relationships between inputs and outputs in labeled data (model training process), and then apply these relationships to a test set to obtain prediction results (model application process).

[0004] Deep learning methods can accurately extract various features of the target layer and represent an important development direction for automated stratigraphic interpretation technology, with tracking performance superior to traditional methods to some extent. However, most published literature currently utilizes existing intelligent image processing techniques to directly treat seismic profiles as images, neglecting the unique characteristics of the strata represented by seismic profiles, such as lateral continuity and vertical stratification, and failing to consider the role of seed points and reference layers. Therefore, it is prone to inaccurate tracking when the continuity of the phase axis is poor. Summary of the Invention

[0005] In view of this, in order to overcome at least one aspect of the above problems, embodiments of the present invention propose a seismic horizon tracking method, comprising the following steps:

[0006] S1. Obtain the seismic profile and perform layer interpretation on several first-direction survey lines in the seismic profile to obtain the target layer ordinate corresponding to each seismic trace in each interpreted first-direction survey line.

[0007] S2, each seismic trace in each interpreted first direction survey line is used as a seed seismic trace and extended to the uninterpreted seismic traces on the left and right sides by a preset width to obtain two profile segments corresponding to each seed seismic trace, wherein each profile segment includes one seed seismic trace and multiple other uninterpreted seismic traces.

[0008] S3, obtain multiple seismic attributes corresponding to each profile segment and parameters of seed seismic traces in each profile segment, wherein the parameters include the ordinate of the target layer.

[0009] S4, construct labels based on the parameters of the multiple seismic attributes and seed points, and input the labels into the neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment;

[0010] S5, take one seismic trace in the profile segment as a seed seismic trace, and extend the uninterpreted seismic trace to the left or right by a preset width to obtain a profile segment again, wherein the newly obtained profile segment includes a seed seismic trace, several interpreted seismic traces and several other uninterpreted seismic traces, and return to step S3 until all uninterpreted seismic traces are interpreted to obtain the first full-area ordinate corresponding to the target layer.

[0011] S6, calculate the average amplitude of the target layer based on the first full-area vertical coordinate, correct the full-area vertical coordinate based on the average amplitude of the full area, and smooth the correction result to obtain the final tracking result.

[0012] In some embodiments, S6, the average amplitude of the target layer is calculated based on the first full-area ordinate, and the full-area ordinate is corrected based on the average amplitude and the correction result is smoothed to obtain the final tracking result. The method further includes:

[0013] Replace the first direction survey line in steps S1-S5 with the second direction survey line to obtain the second full-area ordinate corresponding to the second direction survey line.

[0014] The average of the first and second regional longitudinal coordinates in each seismic trace is calculated to obtain the regional longitudinal coordinates.

[0015] The average amplitude of the target layer is obtained based on the vertical coordinate of the entire region. The vertical coordinate of the entire region is then corrected based on the average amplitude of the entire region, and the correction result is smoothed to obtain the final tracking result.

[0016] In some embodiments, step S5 involves using one seismic trace in the profile segment as a seed seismic trace, and extending a predetermined width to the uninterpreted seismic traces on the left or right to obtain a new profile segment. The newly obtained profile segment includes one seed seismic trace, several interpreted seismic traces, and several other uninterpreted seismic traces. Further steps include:

[0017] In response to a width that can be expanded to the left or right that is greater than a preset width, the outermost seismic trace in the profile segment is used as a seed seismic trace to expand the preset width to the left or right unexplained seismic traces to obtain a profile segment again, wherein the re-obtained profile segment includes a seed seismic trace and multiple other unexplained seismic traces.

[0018] In response to a width that can be expanded to the left or right that is less than a preset width, the seismic traces in the profile segment that are at a distance from the boundary equal to the preset width are used as seed seismic traces to expand the uninterpreted seismic traces to the left or right with a preset width to obtain a profile segment again, wherein the re-obtained profile segment includes a seed seismic trace, several interpreted seismic traces and several other uninterpreted seismic traces.

[0019] In some embodiments, S6, correcting the vertical coordinate of the entire region based on the average amplitude of the entire region, further includes:

[0020] The threshold is obtained based on the correction threshold control coefficient and the average amplitude of the entire region;

[0021] In response to the target layer being a peak target layer, it is determined whether the amplitude value of the point corresponding to the ordinate of each target layer in the vertical coordinate of the whole region is greater than the threshold.

[0022] In response to the point corresponding to the vertical coordinate of the target layer having an amplitude value greater than a threshold, a preset number of sampling points are extracted upwards and downwards from the point corresponding to the vertical coordinate of the target layer, and the amplitude corresponding to each sampling point is interpolated to obtain multiple amplitude values. The vertical coordinate of the target layer corresponding to the point with the largest amplitude value is selected as the correction result from the multiple amplitude values.

[0023] In some embodiments, S6, correcting the vertical coordinate of the entire region based on the average amplitude of the entire region, further includes:

[0024] The threshold is obtained based on the correction threshold control coefficient and the average amplitude of the entire region;

[0025] In response to the target layer being a trough target layer, it is determined whether the amplitude value of the point corresponding to the ordinate of each target layer in the vertical coordinate of the whole region is less than the threshold.

[0026] In response to the point corresponding to the vertical coordinate of the target layer having an amplitude value less than a threshold, a preset number of sampling points are extracted upwards and downwards from the point corresponding to the vertical coordinate of the target layer, and the amplitude corresponding to each sampling point is interpolated to obtain multiple amplitude values. The vertical coordinate of the target layer corresponding to the point with the smallest amplitude value is selected as the correction result.

[0027] In some embodiments, it also includes:

[0028] Adjust the threshold control coefficient based on the calibration results and adjust the number of sampling points according to the width of the same axis.

[0029] In some embodiments, S6, the correction result is smoothed to obtain the final tracking result, and further includes:

[0030] The correction results are smoothed using the mean method or median filtering method.

[0031] In some embodiments, S4, constructing labels based on the parameters of the plurality of seismic attributes and seed points, and inputting the labels into a neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment, further includes:

[0032] A neural network is constructed, wherein the neural network includes a first convolutional layer with a kernel of 1 and an expansion coefficient of 1, a second convolutional layer with a kernel of 3 and an expansion coefficient of 1, a third convolutional layer with a kernel of 3 and an expansion coefficient of 2, a fourth convolutional layer with a kernel of 3 and an expansion coefficient of 5, a cat layer that concatenates the outputs of the four convolutional layers laterally, a reshape layer connected to the cat layer, a feature extraction layer connected to the reshape layer, and a regression layer connected to the feature extraction layer, wherein the regression layer uses a CNN+GRU+FC cascaded approach to achieve regression prediction;

[0033] Construct a training set and train the neural network.

[0034] In some embodiments, constructing a training set and training the neural network further includes:

[0035] Obtain the interpreted seismic backbone profile;

[0036] Multiple profile segments of preset widths are extracted along each survey line in the earthquake backbone profile;

[0037] Obtain N seismic attributes corresponding to each profile segment and parameters of each seismic trace in each profile segment, wherein the parameters include the target layer ordinate, the reference layer ordinate, the first direction survey line number, and the second direction survey line number of the seismic trace.

[0038] The N seismic attributes of each profile segment are used as the data of the first N channels of the label, and the parameters of each seismic trace are used as the data of the last channel to construct the input of the label. The target layer ordinate of each seismic trace in the profile segment is combined into a vector to construct the output of the label. In the input of the label, only the target layer ordinate of the first seismic trace is retained, and the target layer ordinate of the remaining seismic traces is set to zero or set to the ordinate of the reference layer.

[0039] The neural network is trained using the input and output of the labels corresponding to each profile segment.

[0040] In some embodiments, the parameters further include the starting ordinate, the number of sampling points, and the target layer unidirectional axis attribute number, wherein when it is a peak, the target layer unidirectional axis attribute number is 0, and when it is a trough, the target layer unidirectional axis attribute number is 1.

[0041] In some embodiments, it also includes:

[0042] The number of seismic attributes used to construct the label is incremented by 1 to obtain the number of channels for the label;

[0043] The height of the label is obtained based on the selected time window range and sampling rate;

[0044] The width of the label is determined based on the continuity of the same axis.

[0045] In some embodiments, it also includes:

[0046] The time window range is determined based on the maximum and minimum ordinates of the target layer.

[0047] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide a computer device, comprising:

[0048] At least one processor; and

[0049] A memory storing a computer program executable on the processor, characterized in that the processor executes the program to perform the steps of any of the seismic horizon tracking methods described above.

[0050] Based on the same inventive concept, according to another aspect of the present invention, embodiments of the present invention also provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of any of the seismic horizon tracking methods described above.

[0051] The present invention has one of the following beneficial technical effects: The automatic seismic horizon tracking method proposed in this invention fully considers the lateral continuity and vertical stratification of strata, as well as the practical needs such as the need to refer to seed points and marker layer information during the interpretation process. It utilizes existing geophysical and computer technologies to incorporate seed points, reference layers, location information, and various seismic attributes into the learning labels, and combines them with the improved DenseNet network and interpolation correction technology, which effectively improves the accuracy of automatic horizon tracking. Attached Figure Description

[0052] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other embodiments can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a schematic diagram of a research area;

[0054] Figure 2 right Figure 1 The three-dimensional data volume obtained from the sampling points in the study area shown;

[0055] Figure 3 for Figure 2 The diagram shows the internal structure of the three-dimensional data volume.

[0056] Figure 4 To Figure 2 A schematic diagram illustrating the process of obtaining a cross-section by slicing a three-dimensional data volume.

[0057] Figure 5 This is a cross-sectional schematic diagram;

[0058] Figure 6 A schematic diagram of a cross-section with the ordinate of the target layer marked;

[0059] Figure 7 A schematic flowchart of the seismic horizon tracking method provided in an embodiment of the present invention;

[0060] Figure 8 This is a top view of the survey line in the work area;

[0061] Figure 9 This is a diagram illustrating the effect of correcting the tracking result at a certain point in an embodiment of the present invention.

[0062] Figure 10 This is a comparison chart of tracking results for in-phase axes with good continuity in an embodiment of the present invention, wherein... Figure 10 (a) is the tracking result directly using the open-source DenseNet, (b) is the tracking result using the deep learning network of this invention, and (c) is the result after peak correction based on (b).

[0063] Figure 11 This is a comparison chart of tracking results for in-phase axes with poor continuity in an embodiment of the present invention, wherein... Figure 11 In the diagram, (a) is the result of manual interpretation, (b) is the result of tracking using the traditional automatic tracking method, (c) is the result of tracking using the deep learning network of this invention and performing peak correction, and (d) is the result of smoothing using the 3-point mean method based on (c).

[0064] Figure 12 This refers to the data from the first three channels of the 999th tag in this embodiment of the invention. Figure 12 In the diagram, (a) corresponds to the earthquake amplitude attribute, (b) corresponds to the instantaneous phase attribute, and (c) corresponds to the integral trace attribute.

[0065] Figure 13 A schematic diagram of the structure of a computer device provided for an embodiment of the present invention;

[0066] Figure 14 A schematic diagram of the structure of a computer-readable storage medium provided for an embodiment of the present invention. Detailed Implementation

[0067] To make the objectives, technical solutions, and advantages of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to specific examples and the accompanying drawings.

[0068] It should be noted that all uses of "first" and "second" in the embodiments of the present invention are for the purpose of distinguishing two entities or parameters with the same name but different names. It is clear that "first" and "second" are only for the convenience of expression and should not be construed as limiting the embodiments of the present invention. Subsequent embodiments will not explain this in detail.

[0069] In embodiments of the present invention, such as Figure 1 The study area shown can be obtained by deploying points on the ground to obtain data about the study area. Figure 2 The data shown is a three-dimensional data volume. For example... Figure 3 As shown, each point within this 3D data body has x, y, and z coordinates and an amplitude value. For example... Figure 4 and Figure 5 As shown, if you cut perpendicularly along a certain line, you get a profile (a survey line). Each seismic trace on the profile ( Figure 5 The profile shown has 6 seismic traces, each composed of numerous "sampling points," and each sampling point has xyz coordinates and an amplitude value. For example... Figure 6 As shown, Figure 6 The dots in the diagram represent the points corresponding to the ordinates of the target layer. Marking the ordinates of the target layer for each seismic trace constitutes interpretation; the key to interpretation is determining the ordinates of the target layer. According to one aspect of the present invention, embodiments of the present invention propose a seismic horizon tracking method, such as... Figure 7 As shown, it may include the following steps:

[0070] S1. Obtain the seismic profile and perform layer interpretation on several first-direction survey lines in the seismic profile to obtain the target layer ordinate corresponding to each seismic trace in each interpreted first-direction survey line.

[0071] S2, each seismic trace in each interpreted first direction survey line is used as a seed seismic trace and extended to the uninterpreted seismic traces on the left and right sides by a preset width to obtain two profile segments corresponding to each seed seismic trace, wherein each profile segment includes one seed seismic trace and multiple other uninterpreted seismic traces.

[0072] S3, obtain multiple seismic attributes corresponding to each profile segment and parameters of seed seismic traces in each profile segment, wherein the parameters include the ordinate of the target layer.

[0073] S4, construct labels based on the parameters of the multiple seismic attributes and seed points, and input the labels into the neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment;

[0074] S5, take one seismic trace in the profile segment as a seed seismic trace, and extend the uninterpreted seismic trace to the left or right by a preset width to obtain a profile segment again, wherein the newly obtained profile segment includes a seed seismic trace, several interpreted seismic traces and several other uninterpreted seismic traces, and return to step S3 until all uninterpreted seismic traces are interpreted to obtain the first full-area ordinate corresponding to the target layer.

[0075] S6, calculate the average amplitude of the target layer based on the first full-area vertical coordinate, correct the full-area vertical coordinate based on the average amplitude of the full area, and smooth the correction result to obtain the final tracking result.

[0076] The solution proposed in this invention can improve the accuracy of automatic layer tracking for weakly continuous phase axes.

[0077] In some embodiments, in step S1, a seismic profile is acquired, and layer interpretation is performed on several first-direction seismic lines in the seismic profile to obtain the target layer ordinate corresponding to each seismic trace in each interpreted first-direction seismic line. Specifically, the target layer ordinate can refer to the target layer T0 value, which is the ordinate of the interpreted layer on the time-domain seismic profile. Alternatively, it can refer to the target layer D value, which is the ordinate of the interpreted layer on the depth-domain seismic profile. The first-direction seismic line can be an inline seismic line (main seismic line), or an X-line or trace seismic line (connecting line).

[0078] It should be noted that, as Figure 8 The diagram shows a top view of the seismic line. In this schematic diagram, each point represents a seismic trace, and all the points form a seismic line. A perpendicular cut along this seismic line yields the following results: Figure 4 and Figure 5The seismic profile shown, i.e., each seismic line profile, includes multiple seismic traces. The points along the longitudinal and transverse axes form two directions of seismic lines, one called the main seismic line and the other called the connecting line. That is, each seismic trace belongs to both the main seismic line profile segment and the connecting line profile segment.

[0079] In some embodiments, in step S2, each seismic trace in each interpreted first-direction survey line is used as a seed seismic trace and extended by a predetermined width to the uninterpreted seismic traces on the left and right sides to obtain two profile segments corresponding to each seed seismic trace. Each profile segment includes one seed seismic trace and multiple other uninterpreted seismic traces. Specifically, if the first-direction survey line is an Inline survey line, a profile segment of width W is extended along each interpreted Inline line, one seismic trace along the XLine direction and to both sides, that is, a profile segment of width W is extended to each of the two uninterpreted seismic traces. Figure 5 If the seismic trace in the middle of the profile is considered as an interpreted seismic trace in the Inline direction and the remaining seismic traces are considered as uninterpreted seismic traces in the Xline direction, then the profile segment can be obtained by extending along the XLine direction and to both sides.

[0080] For example, in a work area with 356 inline survey lines and 836 xline survey lines, data is missing in both the eastern and western parts of the area, therefore the work area is not a regular 356×836 survey line grid. The 20th, 120th, 220th, and 320th inline survey lines and the 200th, 400th, and 600th xline survey lines can be selected as the backbone profiles, and the interpretation can be completed using the functions provided by commercial interpretation software.

[0081] Taking the expansion of the first InLine as an example, let the seismic traces with InLine number i and XLine number j be denoted as Ti,j. The first InLine is numbered InLine20, and the trace numbers are XLine1 to XLine836. The seismic traces on it are denoted as T20,1, T20,2, ..., T20,836. The first seismic trace T20,1 on InLine20 expands into a profile segment with a width of 16 in both the direction of increasing and decreasing XLine. Thus, each profile segment includes seismic traces (T20,1, T21,1, ..., T35,1) and (T20,1, T19,1, ..., T5,1), respectively. The profile segment expanded from the second seismic trace T20,2 includes seismic traces (T20,2, T21,2, ..., T35,2) and (T20,2, T19,2, ..., T5,2), and so on. Then, the points of the other interpreted lines (InLine120, 220, 320) are extended in the same way. In this way, the first seismic trace of each profile segment is the seed seismic trace on the interpreted first-direction survey line profile, which has the known ordinate of the target layer.

[0082] In some embodiments, step S3 involves obtaining multiple seismic attributes corresponding to each profile segment and parameters of the seed seismic traces in each profile segment. These parameters include the target layer's longitudinal coordinates. Specifically, there are hundreds of seismic attributes, such as amplitude, phase, and frequency, which can be calculated and generated using traditional, well-known methods or existing commercial interpretation software. Preferably, three attributes—seismic amplitude, instantaneous phase, and integral trace—can be selected for label construction. Other attributes can be selected based on research experience when applying this invention.

[0083] In some embodiments, in step S4, labels are constructed based on the parameters of the multiple seismic attributes and seed points, and the labels are input into a neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment. Specifically, after the labels of each profile segment are input into the neural network, the target layer ordinates of other seismic traces in that profile segment can be obtained, and the first round tracking result R1 is obtained.

[0084] For example, in the profile segment (T20,1, T21,1, ..., T35,1), before inputting into the neural network, only the target layer ordinate of seismic trace T20,1 is known. After inputting into the neural network, the target layer ordinates of seismic traces T21,1, ..., T35,1 are all obtained through neural network inference. Thus, after inputting all the profile segments into the neural network, the InLine lines numbered 5-19, 21-35, 105-119, 121-135, 205-219, 221-235, 305-319, and 321-335 are interpreted, meaning that the target layer ordinates of all seismic traces on them are interpreted.

[0085] It should be noted that the size of the labels during inference must be the same as the size of the labels in the training set when training the neural network. For example, if the size of the labels in the training set is C×H×W (C for the number of channels, H for the height, and W for the width), then the size of the labels during inference must also be C×H×W. For details on the label construction process, please refer to the training set construction process below.

[0086] In some embodiments, a marker layer near the target layer can be selected and interpreted across the entire region as a reference layer for automatic tracking of the target layer. The ordinate of the reference layer is then also used as an element in constructing the label.

[0087] In some embodiments, in step S5, one seismic trace in the profile segment is used as a seed seismic trace, and the uninterpreted seismic traces on the left or right are extended by a predetermined width to obtain a new profile segment. The newly obtained profile segment includes one seed seismic trace, several interpreted seismic traces, and several other uninterpreted seismic traces. The process then returns to step S3 until all uninterpreted seismic traces have been interpreted to obtain the first full-area ordinate corresponding to the target layer. This further includes:

[0088] In response to a width that can be expanded to the left or right that is greater than a preset width, the outermost seismic trace in the profile segment is used as a seed seismic trace to expand the preset width to the left or right unexplained seismic traces to obtain a profile segment again, wherein the re-obtained profile segment includes a seed seismic trace and multiple other unexplained seismic traces.

[0089] In response to a width that can be expanded to the left or right that is less than a preset width, the seismic traces in the profile segment that are at a distance from the boundary equal to the preset width are used as seed seismic traces to expand the uninterpreted seismic traces to the left or right with a preset width to obtain a profile segment again, wherein the re-obtained profile segment includes a seed seismic trace, several interpreted seismic traces and several other uninterpreted seismic traces.

[0090] Specifically, based on the profile segments from the first round, new profile segments can be extended in both directions of XLine into the unexplained regions using a method similar to the previous step. New profile segments and corresponding labels can be constructed and input into the network to obtain the second round tracking result R2.

[0091] For example, InLines 16, 105, 205, and 305 can be expanded to the left, and InLines 35, 135, 235, and 335 can be expanded to the right to construct new profile segments and corresponding labels. These are then input into the network to obtain the second round of tracking results R2. In the first round of inference, all seismic traces in InLines 5-19 have been interpreted. However, since there are 356 InLines in the work area, directly using the seismic traces in InLine 5 as seed traces would not allow for the construction of profile segments of the same width. Therefore, the profile segment is expanded to the left based on InLine 16. This resulting profile segment includes one seed trace, several interpreted traces, and several uninterpreted traces. That is, if the width of the expandable data in a certain direction is less than W, the expandable width can be made equal to W by adjusting the starting line position in the opposite direction, ensuring that all data in the area is covered. Repeated expansion does not affect the final tracking effect.

[0092] It should be noted that when different survey lines expand their profile segments in two directions and then overlap, further expansion will cease. For example, if InLine35 expands to the right while InLine105 expands to the left, after several expansions, if InLine35-InLine105 has been fully interpreted, then that interval will no longer be expanded.

[0093] Repeat the above process to obtain several intermediate tracing results R3, R4, ... until the stratum tracing of the entire region is completed. Then merge R1, R2, ... to obtain the preliminary tracing result of the entire region, denoted as R.

[0094] In some embodiments, step S6, correcting the ordinate of the entire region based on the average amplitude of the entire region, further includes:

[0095] The threshold is obtained based on the correction threshold control coefficient and the average amplitude of the entire region;

[0096] In response to the target layer being a peak target layer, it is determined whether the amplitude value of the point corresponding to the ordinate of each target layer in the vertical coordinate of the whole region is greater than the threshold.

[0097] In response to the amplitude value of the point corresponding to the ordinate of the target layer being greater than the threshold, a preset number of sampling points are extracted up and down centered on the point corresponding to the ordinate of the target layer, and interpolation processing is performed on the amplitude corresponding to each sampling point to obtain multiple amplitude values, and the ordinate of the target layer corresponding to the point with the largest amplitude value is selected from the multiple amplitude values as the correction result.

[0098] In some embodiments, S6, correcting the全区 ordinate according to the全区 amplitude average value further includes:

[0099] Obtain a threshold according to the correction threshold control coefficient and the全区 amplitude average value;

[0100] In response to the target layer being a trough target layer, determine whether the amplitude value of the point corresponding to each ordinate of the全区 ordinate is less than the threshold;

[0101] In response to the amplitude value of the point corresponding to the ordinate of the target layer being less than the threshold, a preset number of sampling points are extracted up and down centered on the point corresponding to the ordinate of the target layer, and interpolation processing is performed on the amplitude corresponding to each sampling point to obtain multiple amplitude values, and the ordinate of the target layer corresponding to the point with the smallest amplitude value is selected from the multiple amplitude values as the correction result.

[0102] Specifically, taking the peak target layer as an example, the全区 seismic amplitude average value A corresponding to the preliminary tracking result R can be obtained first, and a correction threshold control coefficient c (0 < c < 0.2) is set.

[0103] For each point r of the horizon R, correction is performed in the following two cases: If the amplitude value a corresponding to r is < c * A, its ordinate value remains unchanged. Otherwise, centered on its ordinate value, 10 sampling points are extracted up and down, and cubic spline interpolation is performed on the amplitude values corresponding to these 21 points to encrypt them into 201 points, and then the ordinate value corresponding to the point with the largest amplitude among these 201 points is obtained as the correction result.

[0104] For the trough target layer, a similar method can be used. Retain the T0 value of the points greater than the threshold amplitude value c * A, and correct the T0 value of the points within the threshold range to the position corresponding to the minimum amplitude to achieve correction.

[0105] For example, as Figure 9 shown, the original tracking point is at the maximum amplitude on the dotted line, and its (time, amplitude) coordinates are (9.9, 89). The corrected tracking point is at the maximum amplitude on the solid line, and its coordinates are (10.8, 90.1). The horizon is corrected to the peak maximum.

[0106] In some embodiments, it further includes:

[0107] Adjust the threshold control coefficient based on the calibration results and adjust the number of sampling points according to the width of the same axis.

[0108] Specifically, the threshold control coefficient c can be adjusted according to the correction effect; and the number of sampling points used for interpolation can be adjusted according to the width of the phase axis.

[0109] In some embodiments, S6, the correction result is smoothed to obtain the final tracking result, and further includes:

[0110] The correction results are smoothed using the mean method or median filtering method.

[0111] Specifically, well-known curve smoothing methods, such as the mean method and median filtering method, can be used to smooth the correction results. For example, the correction results can be smoothed using a three-point mean.

[0112] In some embodiments, Figure 10 The comparison of tracking results for in-phase axes with good continuity is shown. Among them, (a) is the tracking result directly using an open-source neural network (such as DenseNet), which has large layer fluctuations; (b) is the tracking result using the deep learning network proposed in this invention, and the final result is basically reasonable but does not fall completely on the peak; (c) is the result after peak correction based on (b). It can be seen that the result of (c) has the highest accuracy and can meet production requirements. Figure 11 This is a comparison of tracking results for in-phase axes with poor continuity. Figure (a) shows the manually interpreted result for comparison; (b) shows the tracking result of the traditional automatic tracking method, which shows obvious errors; (c) shows the result after tracking and correction by the deep learning network of this invention, demonstrating its effective handling of the "cross-phase" problem (ellipse markings); (d) is the result after smoothing (c) using the traditional mean method, showing a near-identical relationship with (a), with errors only at the boxed areas. The overall accuracy far exceeds that of traditional methods.

[0113] In some embodiments, S6, the average amplitude of the target layer is calculated based on the first full-area ordinate, and the full-area ordinate is corrected based on the average amplitude and the correction result is smoothed to obtain the final tracking result. The method further includes:

[0114] Replace the first direction survey line in steps S1-S5 with the second direction survey line to obtain the second full-area ordinate corresponding to the second direction survey line.

[0115] The average of the first and second regional longitudinal coordinates in each seismic trace is calculated to obtain the regional longitudinal coordinates.

[0116] The average amplitude of the target layer is obtained based on the vertical coordinate of the entire region. The vertical coordinate of the entire region is then corrected based on the average amplitude of the entire region, and the correction result is smoothed to obtain the final tracking result.

[0117] Specifically, the tracking results can be obtained through two types of survey lines. After obtaining the full-area tracking result R based on the InLine survey line, the above process is repeated in the XLine direction to obtain another set of full-area tracking results XR. The average value of the tracking results in the two directions is taken by the formula (R+XR) / 2 as the initial full-area tracking result R.

[0118] In some embodiments, S4, constructing labels based on the parameters of the plurality of seismic attributes and seed points, and inputting the labels into a neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment, further includes:

[0119] A neural network is constructed, wherein the neural network includes a first convolutional layer with a kernel of 1 and an expansion coefficient of 1, a second convolutional layer with a kernel of 3 and an expansion coefficient of 1, a third convolutional layer with a kernel of 3 and an expansion coefficient of 2, a fourth convolutional layer with a kernel of 3 and an expansion coefficient of 5, a cat layer that concatenates the outputs of the four convolutional layers laterally, a reshape layer connected to the cat layer, a feature extraction layer connected to the reshape layer, and a regression layer connected to the feature extraction layer, wherein the regression layer uses a CNN+GRU+FC cascaded approach to achieve regression prediction;

[0120] Construct a training set and train the neural network.

[0121] Specifically, you can choose the feature layer of the DenseNet network provided by the PyTorch platform as the base network. Then, add convolutional layers with different receptive fields before the feature layer, concatenate their outputs along the W direction, and reshape them into a tensor with 3 channels, which serves as the input to the feature layer.

[0122] For example, we can first perform four convolutions on the input data: one with a kernel size of 1 and dilation of 1, one with a kernel size of 3 and dilation of 1, one with a kernel size of 3 and dilation of 2, and one with a kernel size of 3 and dilation of 5. These four convolutional layers all have C input channels and 6 output channels. Then, the outputs of these convolutional layers are concatenated along the W direction using a cat layer, resulting in a concatenated output with 6 channels. Finally, a reshape layer is used to transform this into a tensor with 3 channels, which is then used as the input to the feature layer.

[0123] Next, the classification layer at the back end of DenseNet is replaced with a regression layer. Preferably, the classification layer at the back end of DenseNet can be replaced with a regression layer. As a preferred approach, regression prediction can be achieved by using a CNN+GRU+FC concatenated layer. Here, CNN is a convolutional layer, GRU is a recurrent neural network layer, and FC is a fully connected layer. The network design ensures that the output of the concatenated three layers is a one-dimensional vector of length 16, consistent with the shape of the label output.

[0124] Preferably, the network uses the mean squared loss (MSELoss) function as the loss function and the Adam function as the optimization function. The initial values ​​for the network hyperparameters can be: batch size 50, learning rate 0.002, and number of iterations 250. Then, following well-known methods in the field of deep learning, other necessary code, especially quality control code, is added to construct a complete deep learning network.

[0125] Finally, the improved network is trained using the training set labels. This includes: following well-known methods in the field of deep learning, inputting the training set into the improved deep learning network, training the network, and saving the network parameters; using quality control methods, adjusting the network structure and hyperparameters according to accuracy requirements until the expected prediction accuracy is achieved. If necessary, the number of training set labels can be appropriately increased.

[0126] In some embodiments, constructing a training set and training the neural network further includes:

[0127] Obtain the interpreted seismic backbone profile;

[0128] Multiple profile segments of preset widths are extracted along each survey line in the earthquake backbone profile;

[0129] Obtain N seismic attributes corresponding to each profile segment and parameters of each seismic trace in each profile segment, wherein the parameters include the target layer ordinate, the reference layer ordinate, the first direction survey line number, and the second direction survey line number of the seismic trace.

[0130] The N seismic attributes of each profile segment are used as the data of the first N channels of the label, and the parameters of each seismic trace are used as the data of the last channel to construct the input of the label. The target layer ordinate of each seismic trace in the profile segment is combined into a vector to construct the output of the label. In the input of the label, only the target layer ordinate of the first seismic trace is retained, and the target layer ordinate of the remaining seismic traces is set to zero or set to the ordinate of the reference layer.

[0131] The neural network is trained using the input and output of the labels corresponding to each profile segment.

[0132] In some embodiments, the parameters further include the starting ordinate, the number of sampling points, and the target layer unidirectional axis attribute number, wherein when it is a peak, the target layer unidirectional axis attribute number is 0, and when it is a trough, the target layer unidirectional axis attribute number is 1.

[0133] In some embodiments, it also includes:

[0134] The number of seismic attributes used to construct the label is incremented by 1 to obtain the number of channels for the label;

[0135] The height of the label is obtained based on the selected time window range and sampling rate;

[0136] The width of the label is determined based on the continuity of the same axis.

[0137] In some embodiments, it also includes:

[0138] The time window range is determined based on the maximum and minimum ordinates of the target layer.

[0139] Specifically, backbone profiles can be selected and interpreted. Ideally, the backbone profile should consist of at least three evenly distributed profiles in the InLine direction and three or more in the XLine direction, plus one arbitrary line profile connecting the wells. The target layer should be interpreted using traditional, well-known methods or commercial software. Optionally, marker layers near the target layer can be selected and interpreted throughout the entire area as reference layers for automatic tracking of the target layer. Marker layers should be stably developed throughout the area, with distinct characteristics, and capable of accurate automatic interpretation using traditional, well-known methods. For example, for a work area with 356 InLine survey lines and 836 XLine survey lines, the 20th, 120th, 220th, and 320th InLine survey line profiles and the 200th, 400th, and 600th XLine survey line profiles can be selected as backbone profiles, and the interpretation can be completed using the functions provided by commercial interpretation software.

[0140] Then, determine the dimensions of each label, including three parameters: number of channels C, height H, and width W. C equals the number of seismic attributes to be used to construct the label plus 1. H is the number of sampling points for each selected trace, calculated from the selected time window range t and the seismic data sampling rate s: H = t / s + 1. The time window range t can be uniform across the entire region, preferably obtained by extrapolating the maximum and minimum T0 values ​​of the target layer by at least 200 ms; or it can be variable across the entire region, preferably obtained by extrapolating the rough layer obtained from the interpolation of the target layer upwards and downwards by at least 500 ms each. The T0 value refers to the ordinate of the interpreted layer on the time-domain seismic profile, with units in milliseconds (ms). W is the number of seismic data traces for each selected label, preferably 8 to 16 traces based on the continuity of the phase axis. For depth-domain profiles, simply change the time-domain T0 value in the above steps to the depth-domain D value, and change the corresponding unit from milliseconds (ms) to meters (m); other operations remain unchanged. For example, if three seismic attributes can be used, then C = 4. Based on the target layer T0 value range, a uniform time window range of 1400ms to 1700ms was determined for the entire region, with t = 300 and sampling rate s = 1, therefore H = 301. Based on the continuity of the in-phase axis, W = 16 was set.

[0141] Next, the seismic attributes to be used for label construction are determined. There are hundreds of seismic attributes, including amplitude, phase, and frequency, which can be calculated and generated using traditional, well-known methods or existing commercial interpretation software. Preferably, this invention selects three attributes—seismic amplitude, instantaneous phase, and integral trace—for label construction. Other attributes can be selected based on research experience when applying this invention.

[0142] Next, n profile segments with width W and height H are extracted along each interpreted seismic line profile, and the three attributes corresponding to each segment are obtained: seismic amplitude, instantaneous phase, and integral trace. The extraction process is performed segment by segment. Assuming the extracted roll is k (1≤k≤W), the seismic traces corresponding to the first profile segment are (1, 2, ..., W), the seismic traces corresponding to the second profile segment are (k+1, k+2, ..., k+W), and so on. If the total number of traces for a profile is m, then n = (mW) / / k+1. The operator " / / " represents floor division, and the result is the largest integer smaller than the quotient. The three attributes of the same profile segment are used as the first, second, and third channels of a training set label, respectively. Figure 5 The profile shown can be divided into three sections: the first two seismic traces, the middle two seismic traces, and the last two seismic traces.

[0143] For example, on the interpreted backbone profile and its corresponding instantaneous phase and integral trace profiles, with k=1, 7 survey lines (the 20th, 120th, 220th, and 320th Inline survey lines and the 200th, 400th, and 600th XLine survey lines) were used, extracting a total of 4307 profile segments. The seismic amplitude, instantaneous phase, and integral trace attributes of the same profile segment were used as the first, second, and third channels of the training set labels, respectively. The first three channels of the 999th label are as follows: Figure 12 As shown in the figure, the horizontal line represents the location of the stratigraphic layer. It is evident that these three attributes represent the data characteristics of the stratigraphic layer's location from different perspectives, which will help improve the accuracy of stratigraphic tracking.

[0144] It should be noted that the three attributes of seismic amplitude, instantaneous phase, and integral trace in the same profile segment refer to the profile segment extracted from the same location in the seismic amplitude, instantaneous phase, and integral trace profile segment, which can be used as the three attributes of the same profile segment.

[0145] Next, the data for the fourth channel of the label is constructed. At each corresponding position in each trace of each profile segment, seven data points are written in a uniform format: the starting T0 value of the seismic trace, the number of sampling points, the InLine number, the XLine number, the T0 value of the target layer for this trace, the T0 value of the reference layer, and the target layer phase axis attribute number. Each data point is written three times consecutively, occupying positions 1 through 21 in the H direction; the remaining positions are filled with 0 or nan. nan represents a null value. If there is no reference layer, the corresponding position is filled with 0 or nan. The phase axis attribute number is: 0 for peaks, 1 for peaks. Then, the target layer T0 values ​​for all traces except the first trace are rewritten to 0 or nan.

[0146] For example, one of the tags is located in work area InLine220, XLine400-416, and the T0 value of the first channel of the tag is 1557. As an example, assuming that the T0 values ​​of the reference layer are all 1510, then according to the method of the present invention, the values ​​of some channels of the fourth channel of the tag are shown in the following table:

[0147]

[0148]

[0149] The construction of the fourth channel data of the above label is mainly to provide the deep learning network with the spatial positional relationship between the tracked point and other points. Therefore, the order and format can be adjusted, and the content can be added or removed, but the four items of target layer T0 value, reference layer T0 value, InLine number and XLine number cannot be completely deleted.

[0150] Then, the channel data obtained above are concatenated along the C direction to form a three-dimensional tensor with C=4, height H, and width W, which is used as the input for the training set labels. For example, it can be concatenated along the C direction to form a three-dimensional tensor with C=4, height 301, and width 16, which is used as the input for the training set labels.

[0151] Finally, the W target layer T0 values ​​corresponding to each profile segment are combined into a one-dimensional vector of length W, which is used as the output of the training set labels. The label construction is now complete.

[0152] The automatic seismic horizon tracking method proposed in this invention fully considers the lateral continuity and vertical stratification of strata, as well as the practical needs of referencing seed points and marker layer information during the interpretation process. It utilizes existing geophysical and computer technologies to incorporate seed points, reference layers, location information, and / or multiple seismic attributes into the learning labels, and combines them with an improved DenseNet network and interpolation correction techniques to effectively improve the accuracy of automatic horizon tracking.

[0153] Based on the same inventive concept, according to another aspect of the present invention, such as Figure 13 As shown, an embodiment of the present invention also provides a computer device 501, comprising:

[0154] At least one processor 520; and

[0155] The memory 510 stores a computer program 511 that can run on a processor. When the processor 520 executes the program, it performs the steps of any of the seismic horizon tracking methods described above.

[0156] Based on the same inventive concept, according to another aspect of the present invention, such as Figure 14 As shown, embodiments of the present invention also provide a computer-readable storage medium 601, which stores computer program instructions 610. When the computer program instructions 610 are executed by a processor, they perform the steps of any of the seismic horizon tracking methods described above.

[0157] Finally, it should be noted that those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods.

[0158] Furthermore, it should be understood that the computer-readable storage medium (e.g., memory) described herein may be volatile memory or non-volatile memory, or may include both volatile memory and non-volatile memory.

[0159] Those skilled in the art will also understand that the various exemplary logic blocks, modules, circuits, and algorithm steps described in conjunction with the disclosure herein can be implemented as electronic hardware, computer software, or a combination of both. To clearly illustrate this interchangeability between hardware and software, the functionality of various illustrative components, blocks, modules, circuits, and steps has been generally described. Whether this functionality is implemented as software or as hardware depends on the specific application and the design constraints imposed on the system as a whole. Those skilled in the art can implement the functionality in various ways for each specific application, but such implementation decisions should not be construed as departing from the scope of the embodiments disclosed herein.

[0160] The above are exemplary embodiments disclosed in this invention. However, it should be noted that various changes and modifications can be made without departing from the scope of the embodiments of this invention as defined by the claims. The functions, steps, and / or actions of the methods according to the disclosed embodiments described herein do not need to be performed in any particular order. Furthermore, although the elements disclosed in the embodiments of this invention may be described or claimed individually, they may be understood as multiple unless explicitly limited to a singular number.

[0161] It should be understood that, as used herein, the singular form “a” is intended to include the plural form as well, unless the context clearly supports an exception. It should also be understood that, as used herein, “and / or” refers to any and all possible combinations of one or more of the associated listed items.

[0162] The embodiment numbers disclosed in the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0163] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0164] Those skilled in the art should understand that the discussion of any of the above embodiments is merely exemplary and is not intended to imply that the scope of the invention (including the claims) is limited to these examples. Within the framework of the invention, technical features of the above embodiments or different embodiments can be combined, and many other variations of different aspects of the invention exist, which are not provided in the details for the sake of brevity. Therefore, any omissions, modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the invention should be included within the protection scope of the invention.

Claims

1. A seismic horizon tracing method, characterized in that, Includes the following steps: S1. Obtain the seismic profile and perform layer interpretation on several first-direction survey lines in the seismic profile to obtain the target layer ordinate corresponding to each seismic trace in each interpreted first-direction survey line. S2, each seismic trace in each interpreted first direction survey line is used as a seed seismic trace and extended to the uninterpreted seismic traces on the left and right sides by a preset width to obtain two profile segments corresponding to each seed seismic trace, wherein each profile segment includes one seed seismic trace and multiple other uninterpreted seismic traces. S3, obtain multiple seismic attributes corresponding to each profile segment and parameters of seed seismic traces in each profile segment, wherein the parameters include the ordinate of the target layer. S4, construct labels based on the parameters of the multiple seismic attributes and seed points, and input the labels into the neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment; S5, take one seismic trace in the profile segment as a seed seismic trace, and extend the uninterpreted seismic trace to the left or right by a preset width to obtain a profile segment again, wherein the newly obtained profile segment includes a seed seismic trace, several interpreted seismic traces and several other uninterpreted seismic traces, and return to step S3 until all uninterpreted seismic traces are interpreted to obtain the first full-area ordinate corresponding to the target layer. S6, calculate the average amplitude of the target layer based on the first full-area vertical coordinate, correct the first full-area vertical coordinate based on the average amplitude of the target layer, and smooth the correction result to obtain the final tracking result; or, Replace the first direction survey line in steps S1-S5 with the second direction survey line to obtain the second full-area ordinate corresponding to the second direction survey line. The average of the first and second regional longitudinal coordinates in each seismic trace is calculated to obtain the regional longitudinal coordinates. The average amplitude of the target layer is obtained based on the vertical coordinate of the entire region. The vertical coordinate of the entire region is then corrected based on the average amplitude of the entire region, and the correction result is smoothed to obtain the final tracking result.

2. The method as described in claim 1, characterized in that, Step S5: One seismic trace in the profile segment is used as a seed seismic trace, and the uninterpreted seismic traces on the left or right are extended by a predetermined width to obtain another profile segment. The newly obtained profile segment includes one seed seismic trace, several interpreted seismic traces, and several other uninterpreted seismic traces. Further, it includes: In response to a width that can be expanded to the left or right that is greater than a preset width, the outermost seismic trace in the profile segment is used as a seed seismic trace to expand the preset width to the left or right unexplained seismic traces to obtain a profile segment again, wherein the re-obtained profile segment includes a seed seismic trace and multiple other unexplained seismic traces. In response to a width that can be expanded to the left or right that is less than a preset width, the seismic traces in the profile segment that are at a distance from the boundary equal to the preset width are used as seed seismic traces to expand the uninterpreted seismic traces to the left or right with a preset width to obtain a profile segment again, wherein the re-obtained profile segment includes a seed seismic trace, several interpreted seismic traces and several other uninterpreted seismic traces.

3. The method as described in claim 1, characterized in that, S6, correcting the vertical coordinate of the entire region based on the average amplitude of the entire region, further includes: The threshold is obtained based on the correction threshold control coefficient and the average amplitude of the entire region; In response to the target layer being a peak target layer, it is determined whether the amplitude value of the point corresponding to the ordinate of each target layer in the vertical coordinate of the whole region is greater than the threshold. In response to the point corresponding to the vertical coordinate of the target layer having an amplitude value greater than a threshold, a preset number of sampling points are extracted upwards and downwards from the point corresponding to the vertical coordinate of the target layer, and the amplitude corresponding to each sampling point is interpolated to obtain multiple amplitude values. The vertical coordinate of the target layer corresponding to the point with the largest amplitude value is selected as the correction result from the multiple amplitude values.

4. The method as described in claim 3, characterized in that, S6, correcting the vertical coordinate of the entire region based on the average amplitude of the entire region, further includes: The threshold is obtained based on the correction threshold control coefficient and the average amplitude of the entire region; In response to the target layer being a trough target layer, it is determined whether the amplitude value of the point corresponding to the ordinate of each target layer in the vertical coordinate of the whole region is less than the threshold. In response to the point corresponding to the vertical coordinate of the target layer having an amplitude value less than a threshold, a preset number of sampling points are extracted upwards and downwards from the point corresponding to the vertical coordinate of the target layer, and the amplitude corresponding to each sampling point is interpolated to obtain multiple amplitude values. The vertical coordinate of the target layer corresponding to the point with the smallest amplitude value is selected as the correction result.

5. The method as described in claim 4, characterized in that, Also includes: Adjust the threshold control coefficient based on the calibration results and adjust the number of sampling points according to the width of the same axis.

6. The method as described in claim 1, characterized in that, S6, the correction results are smoothed to obtain the final tracking result, and further includes: The correction results are smoothed using the mean method or median filtering method.

7. The method as described in claim 1, characterized in that, S4, constructing labels based on the parameters of the multiple seismic attributes and seed points, and inputting the labels into a neural network to obtain the target layer ordinates corresponding to the unexplained seismic traces in each profile segment, further including: A neural network is constructed, wherein the neural network includes a first convolutional layer with a kernel of 1 and an expansion coefficient of 1, a second convolutional layer with a kernel of 3 and an expansion coefficient of 1, a third convolutional layer with a kernel of 3 and an expansion coefficient of 2, a fourth convolutional layer with a kernel of 3 and an expansion coefficient of 5, a cat layer that concatenates the outputs of the four convolutional layers laterally, a reshape layer connected to the cat layer, a feature extraction layer connected to the reshape layer, and a regression layer connected to the feature extraction layer, wherein the regression layer uses a CNN+GRU+FC cascaded approach to achieve regression prediction; Construct a training set and train the neural network.

8. The method as described in claim 7, characterized in that, Constructing a training set and training the neural network further includes: Obtain the interpreted seismic backbone profile; Multiple profile segments of preset widths are extracted along each survey line in the earthquake backbone profile; Obtain N seismic attributes corresponding to each profile segment and parameters of each seismic trace in each profile segment, wherein the parameters include the target layer ordinate, the reference layer ordinate, the first direction survey line number, and the second direction survey line number of the seismic trace. The N seismic attributes of each profile segment are used as the data of the first N channels of the label, and the parameters of each seismic trace are used as the data of the last channel to construct the input of the label. The target layer ordinate of each seismic trace in the profile segment is combined into a vector to construct the output of the label. In the input of the label, only the target layer ordinate of the first seismic trace is retained, and the target layer ordinate of the remaining seismic traces is set to zero or set to the ordinate of the reference layer. The neural network is trained using the input and output of the labels corresponding to each profile segment.

9. The method as described in claim 8, characterized in that, The parameters also include the starting ordinate, the number of sampling points, and the target layer unidirectional axis attribute number, wherein when it is a peak, the target layer unidirectional axis attribute number is 0, and when it is a trough, the target layer unidirectional axis attribute number is 1.

10. The method as described in claim 8, characterized in that, Also includes: The number of seismic attributes used to construct the label is incremented by 1 to obtain the number of channels for the label; The height of the label is obtained based on the selected time window range and sampling rate; The width of the label is determined based on the continuity of the same axis.

11. The method as described in claim 10, characterized in that, Also includes: The time window range is determined based on the maximum and minimum ordinates of the target layer.

12. A computer device, comprising: At least one processor; as well as A memory storing a computer program executable on the processor, characterized in that the processor executes the program and performs the steps of the method as described in any one of claims 1-11.

13. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by a processor, it performs the steps of the method as described in any one of claims 1-11.

Images