CN110658557A - Seismic data surface wave suppression method based on generation of countermeasure network - Google Patents
Seismic data surface wave suppression method based on generation of countermeasure network Download PDFInfo
- Publication number
- CN110658557A CN110658557A CN201910829017.0A CN201910829017A CN110658557A CN 110658557 A CN110658557 A CN 110658557A CN 201910829017 A CN201910829017 A CN 201910829017A CN 110658557 A CN110658557 A CN 110658557A
- Authority
- CN
- China
- Prior art keywords
- data
- training
- generator
- seismic data
- discriminator
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 84
- 230000001629 suppression Effects 0.000 title claims abstract description 22
- 238000012549 training Methods 0.000 claims abstract description 51
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000013528 artificial neural network Methods 0.000 claims abstract description 6
- 238000004088 simulation Methods 0.000 claims abstract description 5
- 238000009826 distribution Methods 0.000 claims description 16
- 230000005484 gravity Effects 0.000 claims description 5
- 238000004364 calculation method Methods 0.000 claims description 3
- 230000036961 partial effect Effects 0.000 claims description 3
- 230000006870 function Effects 0.000 description 15
- 230000000694 effects Effects 0.000 description 11
- 238000001914 filtration Methods 0.000 description 8
- 230000009466 transformation Effects 0.000 description 7
- 238000013527 convolutional neural network Methods 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 230000001427 coherent effect Effects 0.000 description 4
- 238000003825 pressing Methods 0.000 description 4
- 230000006378 damage Effects 0.000 description 2
- 238000000354 decomposition reaction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 238000003672 processing method Methods 0.000 description 2
- 230000002829 reductive effect Effects 0.000 description 2
- 238000013519 translation Methods 0.000 description 2
- 230000002159 abnormal effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001010 compromised effect Effects 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 230000006735 deficit Effects 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000003208 petroleum Substances 0.000 description 1
- 238000012805 post-processing Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000003786 synthesis reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/36—Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/30—Noise handling
- G01V2210/32—Noise reduction
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention provides a seismic data surface wave suppression method based on a generation countermeasure network. The method mainly comprises the following steps: creating a set of training data obtained by data processing; providing training data to a neural network, training a generator and a discriminator, wherein the data generated by the generator can deceive the discriminator through network training, and the discriminator can identify the data and the training data generated by the generator; in the network training process, the generator can obtain good data by adopting a mutual confrontation training method of the generator and the discriminator, and the discriminator can carry out more accurate identification; after network training is completed, the raw seismic data containing the surface waves is provided to a generator, which produces seismic data without surface waves. The simulation data and the actual seismic data processing example show that the method provided by the invention can effectively suppress the surface waves in the seismic data, protect effective signals and improve the signal-to-noise ratio of the seismic data.
Description
Technical Field
The invention relates to a seismic data processing method, in particular to a method for suppressing surface waves in seismic data based on a generated countermeasure network.
Background
In the field seismic data acquisition process, due to the influence of construction environment and underground seismic geological conditions, various types of noise exist in the original seismic data acquired in the field, such as: surface waves, multiples, random noise, mechanical interference, alternating current interference and a small amount of abnormal amplitude and other interference waves. These interfering waves can be generally classified into two broad categories, coherent noise and random noise. Random noise refers to the random occurrence in seismic data, with no fixed frequency, nor fixed direction and velocity of propagation. Coherent noise refers to the fact that the coherent noise has certain regularity in time, has corresponding kinematic characteristics, and has corresponding regularity difference with effective reflected waves. The surface wave is a common coherent noise, and has the characteristics of low frequency, low speed and high amplitude. Surface waves, a special type of rayleigh wave, originate at the surface and are oriented almost perpendicular to the significant wave, often appearing in "broom-like" form in seismic data, and strong energy noise typically covers all seismic traces in a single shot record, overwhelming the significant reflected signal. Therefore, the signal-to-noise ratio of the seismic data is greatly reduced. If no effective measures are taken, the amplitude energy problem caused by the effective measures can seriously cover the reflection in-phase axis in the post-processing process, thereby reducing the processing quality of the seismic data.
Many geophysical researchers have developed many methods to attenuate surface waves in seismic data using the characteristics of surface waves. For example, Stockwell et al propose S-transforms to separate the surface wave and the reflected signal. And decomposing the seismic record into time-frequency signals or time scale coefficients in wavelet transformation by utilizing S transformation, and carrying out signal-noise separation in a time-frequency domain. Yarham et al propose the use of a curvelet transform to suppress surface waves in seismic data. Subsequently, Oliveira et al proposed a method based on the multi-scale and directional characteristics of curvelets, which utilized the L2 norm of the curvelet coefficients to remove the surface waves from the difference in the energy of the surface waves and reflected waves. In addition, Naghizadeh and Sacchi also adopt a proportion and angle guiding method to improve curvelet transformation and achieve good effect. The frequency-wavenumber (f-k) filter is a simple and effective method for removing surface waves, and utilizes the difference between the surface waves and signals in the frequency-wavenumber domain to perform surface wave suppression. On the basis, Askari and Siahkoohi also propose an x-f-k filtering surface wave suppression method. However, since the surface wave and the reflected wave often overlap in the frequency domain, removing the surface wave can harm the effective signal. For the problem of signal and noise frequency overlapping, Cai and Ma propose a method for solving the problem of signal and noise frequency overlapping by using a matched filter, and Jia et al propose a non-stationary matched filtering method for the non-stationarity of seismic data. In addition, there are other surface wave attenuation methods, such as: wavelet transformation, radial tracking transformation filtering, empirical mode decomposition, feature image filtering, singular value decomposition, gradient flow regularization and the like. Among the above methods, f-k filtering is one of the most widely used surface wave attenuation methods. However, f-k filters have certain limitations, and when the surface wave overlaps with the reflected wave in the f-k domain, the filter cannot suppress the surface wave well. Also, when the amplitude of the surface wave is much stronger than the reflected signal, the f-k filter causes significant distortion of the reflected waveform, while creating some aliasing in the reflected signal. Ideally, the above methods all give good results. However, at present, the domestic seismic exploration center has shifted from the eastern area with a simpler geological structure to the western area with a worse environment, and with the deterioration of the environment, the surface wave interference in the seismic data acquired in the field is more serious, which always causes the rapid decrease of the seismic record quality. The method has poor suppression effect on the surface waves in the seismic data acquired from the complex terrain. Therefore, a new method for effectively suppressing the surface waves in the seismic data of the complex area is urgently needed.
Neural network algorithm is a signal processing method widely used in the field of image processing, such as: alexnet, Googlenet, Vggnet, and Resnet, among which Convolutional Neural Networks (CNN) have become the most widely used method in image classification. With the development of deep learning techniques, convolutional neural networks have also recently begun to be used for noise suppression in seismic data. Goodfellow proposed a generate confrontation network (GAN) algorithm in 2014. GAN can sample from high-dimensional complex probability distributions without data assumptions and can obtain actual data faster than traditional neural network methods. After CNN and GAN were combined by Radford and Ronneberger et al, DCGAN and U-NET were obtained, respectively. As with convolutional neural networks, GAN was originally used primarily for various tasks in image processing, such as image-to-image translation, super-resolution, and even unpaired image-to-image translation. In recent years, with the development of artificial intelligence, GAN is also beginning to be applied in the field of seismic data processing, and has better effects in random noise attenuation and seismic missing data reconstruction.
In summary, the method in the prior art is only suitable for suppressing the situation that the original underground structure is simple, and is not suitable for suppressing the surface wave noise in the seismic data of the complex underground structure. The production of confrontational network algorithms has proven itself in the field of seismic data processing as one of several algorithms that have developed rapidly in recent years. There is therefore a need to develop a surface wave attenuation technique based on generation of a counterpoise network in a targeted manner to attenuate surface waves in seismic data.
Disclosure of Invention
Aiming at the defects of the prior art method, the invention provides a surface wave attenuation technology based on a generation countermeasure network based on a deep learning technology, so as to suppress the surface waves in the earthquake single cannon, improve the signal-to-noise ratio of the pre-stack earthquake data and lay a foundation for the subsequent earthquake data processing.
The purpose of the invention is realized by the following technical scheme:
a seismic data surface wave suppression method based on a generation countermeasure network comprises the following steps:
s1, creating a group of training data obtained through data processing;
s2, providing training data for the neural network, training a generator and a discriminator, wherein the data generated by the generator can deceive the discriminator through network training, and the discriminator can recognize the data and the training data generated by the generator;
s3, in the network training process, the generator can obtain good data and the discriminator can perform more accurate identification by adopting a mutual confrontation training method of the generator and the discriminator;
and S4, after the network training is completed, providing the original seismic data containing the surface waves to a generator, and generating the seismic data without the surface waves by the generator.
Preferably, the training set that can be used for network training is obtained by numerical simulation software and commercial seismic data processing software.
Preferably, the objective function is represented as:
wherein x is input data, i.e., data with noise, y is corresponding clean data, and z is a random noise vector; pdata(x,y),Pdata(x) And Pz(z) is the distribution of data pairs (x, y), x, and random noise z, respectively. E is the mathematical expectation, and x-P represent the samples x from the distribution P.
Preferably, the distance between the two distributions is calculated using the following formula:
wherein | | | purple hair1Is the L1 distance, which represents the difference between the result generated by the generator and the true result y.
Preferably, the final objective function is expressed as:
where λ is a coefficient controlling LL1The specific gravity occupied in the loss function, and then the objective function is iteratively calculated using the partial derivatives, the iteration being performed until G is the global minimum and the corresponding global maximum, D, such that the resulting distribution P isG(x)(x) Will resemble the distribution Pdata(x,y)。
Preferably, in the training process, when a set of data with surface waves is input, the arbiter gives different results to the objective function for the following cases:
when the input is the real data pair (x, y), the previous term L of the objective functioncGanA larger specific gravity can be obtained, and the data is judged to be true by the discriminator as much as possible;
when the input is a generated data pair (x, G (x, z)), the latter term L of the objective functionL1There will be a large proportion and the discriminator will try to discriminate the data as false.
Preferably, the discriminator section is trained by equation (1); judging the difference between the data generated by the generator and the real data through the calculation of the formula (2), thereby training the generator part; and combining the formula (1) and the formula (2) together to form a formula (3) as a final objective function of the whole network, and finally obtaining the minimum value of G and the corresponding maximum value of D aiming at the objective function.
Preferably, the data simulation is implemented by MATLAB software.
Preferably, the commercial seismic data processing software is any one of CGG software, geoeast software, focus software, omega software or sessmic unix software.
Compared with the prior art, the embodiment of the invention has at least the following advantages:
because the method directly processes data in a time-space domain, the method does not need to carry out complex domain transformation. Meanwhile, the method starts based on data and automatically learns useful signals, so that the useful signals are basically not damaged. Compared with the pressing result of the prior art, the method provided by the invention can greatly improve the signal-to-noise ratio of the seismic data.
Drawings
FIG. 1 is a flow chart of a seismic data surface wave suppression method provided by the present invention;
FIGS. 2(a) -2 (f) are graphs of the effect of attenuating surface waves in simulated data using prior art methods;
3(a) -3 (c) are graphs of the effect of suppressing the same simulated data surface wave as in FIG. 2 using the method provided by the present invention;
4(a) -4 (d) are graphs of the effect of suppressing surface waves in actual seismic data for a work area using prior art methods;
fig. 5(a) -5 (c) are graphs showing the effect of suppressing the same actual seismic mid-surface wave as in fig. 5 by the method provided by the present invention.
Detailed Description
The present invention will be further described with reference to the following examples and fig. 1-5 thereof, which are illustrative and not limiting, and the scope of the present invention is not limited thereby.
As shown in fig. 1, fig. 1 is a flow chart of a method for suppressing surface waves in seismic data according to the present invention.
A seismic data surface wave suppression method based on a generation countermeasure network comprises the following steps:
s1, creating a group of training data obtained through data processing; different training sets need to be used for the model data and the actual data. The training data obtained by data processing is as follows:
for the model data, 100 pairs of synthetic seismic records are collectively made in this embodiment, and these records include both reflection signals synthesized by Rake (Ricker) wavelets with a main frequency of 20Hz, surface waves with a frequency of 6Hz to 10Hz, and white Gaussian noise. The synthetic record has 128 seismic traces and a record length of 2000 milliseconds. The training data is composed of pure signal data and noisy data which are paired, and the pure signal data and the noisy data are in a one-to-one corresponding relation.
For the actual data, in this embodiment, the surface wave of the original seismic data is removed by using a surface wave suppression module in commercial software (e.g., geoeast developed by geophysics corporation of petroleum eastern china, CGG software developed by geophysics (CGG) corporation of france located in paris, france, focus software, omega software, or seismic unix software, and other commercially available seismic data processing software), and the obtained denoised data and the original seismic data constitute a pair of training data. Specifically, a total of 50 pairs of training data are created using actual three-dimensional seismic data, and a training set of the network is formed together with training data for synthetic data.
S2, after obtaining the training set, providing training data for the neural network, training a generator and a discriminator, wherein the data generated by the generator can deceive the discriminator through network training, and the discriminator can recognize the data generated by the generator and the training data; as shown in fig. 2(a) - (e), the reaction of the arbiter to the real sample and the data generated by the generator is shown, respectively.
S3, in the network training process, the generator can obtain good data and the discriminator can perform more accurate identification by adopting a mutual confrontation training method of the generator and the discriminator; wherein the objective function can be expressed as:
where x is the input data, i.e., the noisy data, y is the corresponding clean data, and z is the random noise vector. Pdata(x,y),Pdata(x) And Pz(z) is the distribution of data pairs (x, y), x, and random noise z, respectively. E is the mathematical expectation, and x-P represent the samples x from the distribution P.
The distance between the two distributions is calculated using the following formula:
wherein | | | purple hair1Is the L1 distance, which represents the difference between the result generated by the generator and the true result y.
Preferably, the final objective function can be obtained by the above two formulas, which can be expressed as:
where λ is a coefficient. Iteratively computing the objective function using the partial derivatives until G is the global minimum and the corresponding global maximum, such that the resulting distribution PG(x)(x) Similar to distribution Pdata(x,y)。
S4, after the training of the confrontation network is completed, the raw seismic data containing the surface waves is provided to the generator, and the generator will generate seismic data without surface waves.
According to the constraint condition in the formula (1), in the training process, when a group of data with surface waves is input, the arbiter can make the objective function obtain different results for different situations:
when the input data is the real data pair (x, y), the previous term of the objective function gets a larger proportion, and the discriminator determines the data as true as possible.
When the input is a generated data pair (x, G (x, z)), the latter term of the objective function is a large specific gravity, and the discriminator discriminates the data as false as possible.
In this case, the discriminator is mainly trained by equation (1). The calculation of equation (2) may be used to judge the difference between the data generated by the generator and the real data, thereby training the generator. And finally combining the formula (1) and the formula (2) together as an objective function of the whole countermeasure network. The final target is the minimum value of G for this function and the corresponding maximum value of D.
Since the method processes data directly in the time-space domain, the method in this embodiment does not require complex domain transformation. Meanwhile, the method in the embodiment starts based on data and automatically learns useful signals, so that the useful signals are basically not damaged. It can be seen that the method in the present embodiment theoretically solves the disadvantages of the existing methods mentioned last in the background art.
The difference between the surface wave pressing method in the prior art and the technical scheme of the invention is described below with reference to the accompanying drawings:
FIGS. 2(a) -2 (e) are graphs of the effect of attenuating surface waves in simulated data using a prior art method; fig. 2(a) is a shot record containing a surface wave obtained by simulation synthesis, and fig. 2(b) is a corresponding shot record containing no surface wave. FIG. 2(c) is the result of a surface wave being compressed using a prior art f-k filtering technique, and FIG. 2(d) is the compressed surface wave. It can be seen that some significant residual noise is still present in fig. 2(c), while some filtered-out reflected signal is also noted in fig. 2 (d). It can be seen that the f-k filtering method has the disadvantages of incomplete suppression of surface waves and signal impairment. Fig. 2(e) shows the result of compressing the surface wave using the conventional S-domain filter technique, and fig. 2(f) shows the compressed surface wave. The S-domain filter method, while removing more of the surface waves than the f-k filtering method, attenuates some of the reflected signal energy. Thus, a large number of valid signals are compromised while the surface waves are suppressed using prior art methods.
3(a) -3 (c) are graphs of the effect of suppressing the surface wave in the same simulation data as in FIG. 2 using the method provided by the present invention; wherein, fig. 3(a) is the same data as fig. 2(a), fig. 3(b) is the pressing result by the method of the present invention, and fig. 3(c) is the pressed surface wave. It can be seen from the figure that after the surface wave is suppressed by the method provided by the present invention, the surface wave almost disappears in fig. 3(b), the reflected signal originally covered by the surface wave is better recovered, and in fig. 3(c), only the surface wave is suppressed, and the effective signal is not damaged. Therefore, the application effect of the method provided by the invention on the simulated seismic data is obviously better than the suppression effect of the method in the prior art.
4(a) -4 (d) are graphs of the effect of suppressing surface waves in actual seismic data for a work area using prior art methods; where fig. 4(a) is a shot gather record extracted from actual three-dimensional seismic data, the seismic record length of which is 5s, of which only 2.4s is shown. Wherein the surface wave with strong energy in the single shot record almost covers the reflected signal of the shot concentration near offset track, and the signal-to-noise ratio of the seismic data is seriously reduced. Fig. 4(b) shows the result of removing the surface wave by the S-domain filter. Note that some significant residual noise is still present in the data of fig. 4(b), and it can be seen that the S-domain filter does not completely remove the surface wave. FIG. 4(c) is a shot set for the training set, which was obtained by compressing the surface wave of FIG. 4(a) using commercial software. In comparison to fig. 4(b), commercial software, while suppressing most of the surface waves, attenuates some of the energy of the effective signal. Fig. 4(d) shows a commercial software suppressed surface wave. It can be seen that although some surface waves are suppressed in fig. 4(d), the method also has a serious influence on the effective signal, and the energy of the effective wave is greatly attenuated. It can be seen that the prior art methods have not been able to thoroughly suppress surface waves in seismic data.
FIGS. 5(a) -5 (c) are graphs of the effect of suppressing the same actual seismic data surface wave as in FIG. 4 using the method provided by the present invention; fig. 5(a) is the same as the seismic data in fig. 4(a), fig. 5(b) is the result of compressing a surface wave by the method of the present invention, and fig. 5(c) is the compressed surface wave. From the analysis on the graph, the method provided by the invention can better suppress the surface wave, obviously enhances the continuity of the reflected signal and has less damage to the effective signal. Compared with the pressing result of the prior art, the method provided by the invention greatly improves the signal-to-noise ratio of the seismic data, thereby laying a foundation for subsequent data processing.
The above description is only for the preferred embodiment of the present invention, but the scope of the present invention is not limited thereto, and any changes or substitutions that can be easily conceived by those skilled in the art within the technical scope of the present invention are included in the scope of the present invention. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (9)
1. A seismic data surface wave suppression method based on a generated countermeasure network is characterized by comprising the following steps:
s1, creating a group of training data obtained through data processing;
s2, providing training data for the neural network, training a generator and a discriminator, wherein the data generated by the generator can deceive the discriminator through network training, and the discriminator can recognize the data and the training data generated by the generator;
s3, in the network training process, the generator can obtain good data and the discriminator can perform more accurate identification by adopting a mutual confrontation training method of the generator and the discriminator;
and S4, after the network training is completed, providing the original seismic data containing the surface waves to a generator, and generating the seismic data without the surface waves by the generator.
2. The seismic data surface wave suppression method based on generation of countermeasure networks according to claim 1, characterized in that a training set that can be used for network training is obtained by numerical simulation software and commercial seismic data processing software.
3. The seismic data surface wave suppression method based on generation of a countermeasure network of claim 2, wherein the objective function is expressed as:
wherein x is input data, i.e., data with noise, y is corresponding clean data, and z is a random noise vector; pdata(x,y),Pdata(x) And Pz(z) is the distribution of data pairs (x, y), x, and random noise z, respectively. E is the mathematical expectation, and x-P represent the samples x from the distribution P.
4. The seismic data surface wave suppression method based on generation of a countermeasure network of claim 3, wherein the distance between two distributions is calculated using the following formula:
wherein | | | purple hair1Is the L1 distance, which represents the difference between the result generated by the generator and the true result y.
5. The seismic data surface wave suppression method based on generation of a countermeasure network of claim 4, wherein a final objective function is expressed as:
where λ is a coefficient controlling LL1The specific gravity occupied in the loss function, and then the objective function is iteratively calculated using the partial derivatives, the iteration being performed until G is the global minimum and the corresponding global maximum, D, such that the resulting distribution P isG(x)(x) Will resemble the distribution Pdata(x,y)。
6. The seismic data surface wave suppression method based on generation of a countermeasure network of claim 5, wherein during training, when a set of data with surface waves is input, the arbiter gives different results to the objective function for the following cases:
when the input is the real data pair (x, y), the previous term L of the objective functioncGanA larger specific gravity can be obtained, and the data is judged to be true by the discriminator as much as possible;
when the input is a generated data pair (x, G (x, z)), the latter term L of the objective functionL1There will be a large proportion and the discriminator will try to discriminate the data as false.
7. The seismic data surface wave suppression method based on generation of a countermeasure network of claim 6, wherein the discriminator section is trained by equation (1); judging the difference between the data generated by the generator and the real data through the calculation of the formula (2), thereby training the generator part; and combining the formula (1) and the formula (2) together to form a formula (3) as a final objective function of the whole network, and finally obtaining the minimum value of G and the corresponding maximum value of D aiming at the objective function.
8. The seismic data surface wave suppression method based on generation of a countermeasure network according to any one of claims 2-6, wherein numerical simulation is implemented by MATLAB software.
9. The seismic data surface wave suppression method based on generation of the countermeasure network according to any one of claims 2-6, wherein the commercial seismic data processing software is any one of CGG software, geoeast software, focus software, omega software or sessmic unix software.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910829017.0A CN110658557B (en) | 2019-09-03 | 2019-09-03 | Seismic data surface wave suppression method based on generation of countermeasure network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910829017.0A CN110658557B (en) | 2019-09-03 | 2019-09-03 | Seismic data surface wave suppression method based on generation of countermeasure network |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110658557A true CN110658557A (en) | 2020-01-07 |
CN110658557B CN110658557B (en) | 2021-11-02 |
Family
ID=69036672
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910829017.0A Active CN110658557B (en) | 2019-09-03 | 2019-09-03 | Seismic data surface wave suppression method based on generation of countermeasure network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110658557B (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111190227A (en) * | 2020-01-09 | 2020-05-22 | 吉林大学 | Low signal-to-noise ratio seismic data denoising method based on residual convolution generation countermeasure model |
CN113589383A (en) * | 2020-04-30 | 2021-11-02 | 中国石油化工股份有限公司 | Seismic data linear interference noise elimination method based on deep learning |
CN113592000A (en) * | 2021-08-03 | 2021-11-02 | 成都理工大学 | Convolution-based crack identification technology for antagonistic neural network |
CN113687420A (en) * | 2020-05-18 | 2021-11-23 | 中国石油天然气股份有限公司 | Method and device for removing surface wave noise of seismic data |
CN113900146A (en) * | 2020-07-06 | 2022-01-07 | 中国石油天然气股份有限公司 | Surface wave pressing method and system |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107437077A (en) * | 2017-08-04 | 2017-12-05 | 深圳市唯特视科技有限公司 | A kind of method that rotation face based on generation confrontation network represents study |
CN109143353A (en) * | 2018-08-20 | 2019-01-04 | 电子科技大学 | A kind of pre-stack seismic waveform classification generating confrontation network based on depth convolution |
CN109523463A (en) * | 2018-11-20 | 2019-03-26 | 中山大学 | A kind of face aging method generating confrontation network based on condition |
CN110069656A (en) * | 2019-03-28 | 2019-07-30 | 天津大学 | A method of threedimensional model is retrieved based on the two-dimension picture for generating confrontation network |
CN110082821A (en) * | 2019-03-26 | 2019-08-02 | 长江大学 | A kind of no label frame microseism signal detecting method and device |
WO2019162364A1 (en) * | 2018-02-23 | 2019-08-29 | Robert Bosch Gmbh | Adversarial probabilistic regularization |
KR102016977B1 (en) * | 2019-04-01 | 2019-09-02 | (주)지오시스템리서치 | Sea surface temperature images regenerating system and method usign gan deep learning |
-
2019
- 2019-09-03 CN CN201910829017.0A patent/CN110658557B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107437077A (en) * | 2017-08-04 | 2017-12-05 | 深圳市唯特视科技有限公司 | A kind of method that rotation face based on generation confrontation network represents study |
WO2019162364A1 (en) * | 2018-02-23 | 2019-08-29 | Robert Bosch Gmbh | Adversarial probabilistic regularization |
CN109143353A (en) * | 2018-08-20 | 2019-01-04 | 电子科技大学 | A kind of pre-stack seismic waveform classification generating confrontation network based on depth convolution |
CN109523463A (en) * | 2018-11-20 | 2019-03-26 | 中山大学 | A kind of face aging method generating confrontation network based on condition |
CN110082821A (en) * | 2019-03-26 | 2019-08-02 | 长江大学 | A kind of no label frame microseism signal detecting method and device |
CN110069656A (en) * | 2019-03-28 | 2019-07-30 | 天津大学 | A method of threedimensional model is retrieved based on the two-dimension picture for generating confrontation network |
KR102016977B1 (en) * | 2019-04-01 | 2019-09-02 | (주)지오시스템리서치 | Sea surface temperature images regenerating system and method usign gan deep learning |
Non-Patent Citations (3)
Title |
---|
代丽艳 等: "微地震数据去噪方法综述", 《吉林大学学报(地球科学版)》 * |
刘继伟 等: "使用卷积神经网络的面波识别技术", 《CPS/SEG北京2018国际地球物理会议暨展览电子论文集》 * |
王坤峰 等: "生成式对抗网络GAN的研究进展与展望", 《自动化学报》 * |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111190227A (en) * | 2020-01-09 | 2020-05-22 | 吉林大学 | Low signal-to-noise ratio seismic data denoising method based on residual convolution generation countermeasure model |
CN111190227B (en) * | 2020-01-09 | 2021-06-04 | 吉林大学 | Low signal-to-noise ratio seismic data denoising method based on residual convolution generation countermeasure model |
CN113589383A (en) * | 2020-04-30 | 2021-11-02 | 中国石油化工股份有限公司 | Seismic data linear interference noise elimination method based on deep learning |
CN113589383B (en) * | 2020-04-30 | 2024-03-19 | 中国石油化工股份有限公司 | Seismic data linear interference noise elimination method based on deep learning |
CN113687420A (en) * | 2020-05-18 | 2021-11-23 | 中国石油天然气股份有限公司 | Method and device for removing surface wave noise of seismic data |
CN113687420B (en) * | 2020-05-18 | 2023-08-22 | 中国石油天然气股份有限公司 | Method and device for removing surface wave noise of seismic data |
CN113900146A (en) * | 2020-07-06 | 2022-01-07 | 中国石油天然气股份有限公司 | Surface wave pressing method and system |
CN113592000A (en) * | 2021-08-03 | 2021-11-02 | 成都理工大学 | Convolution-based crack identification technology for antagonistic neural network |
Also Published As
Publication number | Publication date |
---|---|
CN110658557B (en) | 2021-11-02 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110658557B (en) | Seismic data surface wave suppression method based on generation of countermeasure network | |
CA2600657C (en) | Radon transformations for removal of noise from seismic data | |
CN110806602A (en) | Intelligent seismic data random noise suppression method based on deep learning | |
WO2006096348A1 (en) | Removal of noise from seismic data using radon transformations | |
CA2855734A1 (en) | Coherent noise attenuation | |
CN111190227B (en) | Low signal-to-noise ratio seismic data denoising method based on residual convolution generation countermeasure model | |
US8315124B2 (en) | System and method for suppression of seismic multiple reflection signals | |
WO2008112036A1 (en) | Imaging of multishot seismic data | |
Cambois et al. | Surface-consistent deconvolution in the log/Fourier domain | |
WO2020033465A1 (en) | Surface wave estimation and removal from seismic data | |
Li et al. | Distributed acoustic sensing vertical seismic profile data denoising based on multistage denoising network | |
CN115905805A (en) | DAS data multi-scale noise reduction method based on global information judgment GAN | |
Wang et al. | Attribute-guided target data separation network for DAS VSP data | |
CA2497296C (en) | Removal of noise from seismic data using improved radon transformations | |
Chen et al. | Robust adaptive polarization analysis method for eliminating ground roll in 3C land seismics | |
Wu et al. | Multi-scale progressive fusion attention network based on small sample training for DAS noise suppression | |
CN112926232A (en) | Seismic low-frequency component recovery method based on layered fusion | |
Jia et al. | Separating ground-roll from land seismic record via convolutional neural network | |
CN109212609A (en) | Near surface Noise Elimination method based on wave equation continuation | |
Ma et al. | A Global and Multi-Scale Denoising Method Based on Generative Adversarial Network for DAS VSP Data | |
CN113093282A (en) | Desert data denoising method based on geometric modal characteristic parallel network | |
Ma et al. | Unpaired training: Optimize the seismic data denoising model without paired training data | |
Sun et al. | Deep learning-based Vz-noise attenuation for OBS data | |
CN111722286B (en) | Random function median filtering technique | |
Wang et al. | Deep learning-based attenuation for shear-wave leakage from ocean-bottom node data |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |