Seismic trace gather data denoising method, storage medium and program product
By using unsupervised deep neural networks and forward simulation techniques in seismic exploration, the noise interference problem is solved and the accuracy of seismic data is improved.
Patent Information
- Application Number
- CN202510251197.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-03-04
- Publication Date
- 2025-06-27
AI Technical Summary
In seismic exploration, seismic exogenous coherence noise interferes with the acquisition of seismic effective signals and affects the accuracy of the data.
Using an unsupervised deep neural network trained to convergence, seismic exogenous coherence noise in seismic track set data is predicted and removed through forward simulation and hyperbolic correction operators.
Effectively removes seismic external coherence noise, improves the accuracy and quality of effective seismic data, and makes the obtained noise closer to the real noise in amplitude, frequency and phase.
Smart Images
Figure CN120214924A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the technical field of underground medium structure exploration, and particularly to a method for denoising seismic trace gather data, a storage medium, and a program product. Background Art
[0002] At the data acquisition site of seismic exploration, in addition to the normal excitation source, there may be additional sources near the normal excitation source, such as pumping units, motor vehicles, and wind turbines. These additional sources will cause seismic external source coherent noise, affecting the seismic effective signal normally excited by the source, and thus affecting the collected seismic effective data.
[0003] Therefore, how to accurately obtain seismic effective data is an urgent problem to be solved today. Summary of the Invention
[0004] Embodiments of this application provide a method for denoising seismic trace gather data, a storage medium, and a program product, which are used to remove seismic external source coherent noise and obtain seismic effective data.
[0005] In a first aspect, embodiments of this application provide a method for denoising seismic trace gather data, including:
[0006] Obtain first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise;
[0007] Predict the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence, so as to obtain the first seismic external source coherent noise;
[0008] Remove the first seismic external source coherent noise in the first seismic trace gather data to obtain seismic effective data.
[0009] In a possible implementation manner, predicting the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence to obtain the first seismic external source coherent noise includes:
[0010] Perform forward modeling on the seismic external source coherent noise in the first seismic trace gather data to obtain the forward modeled seismic external source coherent noise;
[0011] Based on the forward modeled seismic external source coherent noise and the unsupervised deep neural network that has been trained to convergence, obtain the first seismic external source coherent noise, where the first seismic external source coherent noise is the predicted seismic external source coherent noise in the first seismic trace gather data.
[0012] In a possible implementation manner, based on the forward modeled seismic external source coherent noise and the unsupervised deep neural network that has been trained to convergence, obtaining the first seismic external source coherent noise includes:
[0013] Based on the hyperbola-based dynamic correction operator, the seismic external source coherent noise from the forward modeling of the hyperbola is corrected into horizontal event axes to obtain the second seismic external source coherent noise;
[0014] The second seismic external source coherent noise is input into the unsupervised deep neural network that has been trained to convergence, and the output obtained is the third seismic external source coherent noise;
[0015] Based on the inverse hyperbola-based dynamic correction operator, the third seismic external source coherent noise of the horizontal event axes is restored to a hyperbola to obtain the first seismic external source coherent noise.
[0016] In a possible implementation, before predicting the seismic external source coherent noise in the first seismic trace gather data through the unsupervised deep neural network that has been trained to convergence to obtain the first seismic external source coherent noise, it further includes:
[0017] Based on the first seismic trace gather data and the second seismic external source coherent noise, training the unsupervised deep learning network to obtain the unsupervised deep neural network that has been trained to convergence, where the second seismic external source coherent noise is obtained after correcting the seismic external source coherent noise from the forward modeling of the hyperbola into horizontal event axes based on the hyperbola-based dynamic correction operator.
[0018] In a possible implementation, training the unsupervised deep learning network based on the first seismic trace gather data and the second seismic external source coherent noise to obtain the unsupervised deep neural network that has been trained to convergence includes:
[0019] Based on the hyperbola-based dynamic correction operator, correcting the first seismic trace gather data of the hyperbola into horizontal event axes to obtain the second seismic trace gather data;
[0020] Based on the second seismic trace gather data and the second seismic external source coherent noise, constructing a deep neural network loss function;
[0021] Inputting the second seismic external source coherent noise into the unsupervised deep neural network, and training the unsupervised deep neural network until the deep neural network loss function converges to the minimum value to obtain the optimal deep neural network parameters;
[0022] Determining the unsupervised deep neural network corresponding to the optimal deep neural network parameters as the unsupervised deep neural network that has been trained to convergence.
[0023] In a second aspect, an embodiment of the present application provides a denoising device for seismic trace gather data, including:
[0024] An acquisition module, configured to acquire the first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise;
[0025] A processing module, configured to predict seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence, so as to obtain the first seismic external source coherent noise;
[0026] A denoising module, configured to remove the first seismic external source coherent noise in the first seismic trace gather data, so as to obtain seismic effective data.
[0027] In a third aspect, an embodiment of the present application provides an electronic device, including: a memory and a processor;
[0028] The memory stores computer-executable instructions;
[0029] The processor executes the computer-executable instructions stored in the memory, so that the processor executes the above first aspect and / or various possible implementation manners of the first aspect.
[0030] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, in which computer-executable instructions are stored, and when the computer-executable instructions are executed by a processor, they are used to implement the above first aspect and / or various possible implementation manners of the first aspect.
[0031] In a fifth aspect, an embodiment of the present application provides a computer program product, including a computer program, and when the computer program is executed by a processor, it implements the above first aspect and / or various possible implementation manners of the first aspect.
[0032] A method, a storage medium, and a program product for denoising seismic trace gather data provided by an embodiment of the present application. Obtain first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise; predict the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence, so as to obtain the first seismic external source coherent noise; remove the first seismic external source coherent noise in the first seismic trace gather data, so as to obtain seismic effective data. Since the deep neural network has good non-linear mapping ability, and the unsupervised deep neural network does not need to rely on actual seismic external source coherent noise, for scenarios where actual seismic external source coherent noise cannot be extracted, it can well predict the seismic external source coherent noise in the first seismic trace gather data, making the obtained first seismic external source coherent noise closer to the actual seismic external source coherent noise in the first trace gather data in terms of amplitude, frequency, and phase, so as to obtain seismic effective data more accurately. Description of the Drawings
[0033] The drawings here are incorporated into the description and form a part of this description, showing embodiments that conform to the present application, and are used together with the description to explain the principles of the present application.
[0034] Figure 1aSchematic diagram 1 of the denoising method for seismic gather data provided by this application;
[0035] Figure 1b Schematic diagram 1 of the principle of seismic gather data provided by this application;
[0036] Figure 1c Principle of seismic gather data provided by this application Figure 2 ;
[0037] Figure 2 Schematic diagram of the process of the denoising method for seismic gather data provided by this application Figure 2 ;
[0038] Figure 3 Schematic diagram of the principle of the unsupervised deep neural network provided by this application;
[0039] Figure 4 Schematic diagram of the structure of the denoising device for seismic gather data provided by this application;
[0040] Figure 5 Schematic diagram of the structure of an electronic device provided by this application.
[0041] Through the above-mentioned drawings, specific embodiments of this application have been shown, and there will be more detailed descriptions hereinafter. These drawings and written descriptions are not intended to limit the scope of the concept of this application in any way, but to illustrate the concept of this application to those skilled in the art by referring to specific embodiments. Detailed implementation manners
[0042] Here, the exemplary embodiments will be described in detail, and the examples are shown in the drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementation manners described in the following exemplary embodiments do not represent all implementation manners consistent with this application. On the contrary, they are only examples of the devices and methods consistent with some aspects of this application as detailed in the appended claims.
[0043] At the data acquisition site in seismic exploration, in addition to the normal excitation source, there may be additional sources near the normal excitation source, such as pumping units, motor vehicles, and windmills. These additional sources will cause seismic external source coherent noise, which affects the seismic effective signal normally excited by the source, thereby affecting the collected seismic effective data. In the common shot domain, the amplitude of the external source coherent noise is very strong, the coherence is very strong, and the bandwidth is narrow. In the time-space domain, the external source coherent noise and the effective signal overlap a lot. The seismic external source coherent noise can seriously affect the seismic effective signal and needs to be suppressed or removed in the early stage of seismic data processing.
[0044] In some embodiments, a convolutional neural network is used to process data containing surface waves, and a supervised method is adopted to separate surface waves and seismic effective data. The full-wavefield data and the predicted multiple data are used as the input data of the supervised deep neural network, and the primary wave, multiple wave, and background noise obtained by traditional methods are used as label data. The finally trained supervised deep neural network can automatically separate the primary wave and multiple wave in the full-wavefield data. The above deep neural network methods all belong to supervised learning.
[0045] However, in the above supervised methods, real data is required to produce label data or the results processed by traditional methods are used as label data. Therefore, the effect of the supervised deep neural network method depends on the label data. In the actual processing of seismic trace gather data, real seismic external source coherent noise cannot be obtained, and the accuracy of the supervised deep neural network method in the actual seismic trace gather data processing scenario will be reduced.
[0046] In order to accurately obtain seismic effective data, during the denoising process of actual seismic trace gather data, the actually collected seismic trace gather data includes seismic effective data and seismic external source coherent noise. Since the deep neural network has good non-linear mapping ability, a deep neural network can be used to predict the seismic external source coherent noise. At this time, only the predicted seismic external source coherent noise in the seismic trace gather data needs to be removed to obtain seismic effective data; further, in order to improve accuracy, the deep neural network adopts an unsupervised deep neural network that does not rely on real seismic external source coherent noise, uses the available actual seismic trace gather data as input data and label data, has a wider range of use, and further makes the predicted seismic external source coherent noise closer to the real seismic external source coherent noise.
[0047] Therefore, an embodiment of the present application provides a method, a storage medium, and a program product for denoising seismic trace gather data. Obtain first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise; predict the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence to obtain the first seismic external source coherent noise; remove the first seismic external source coherent noise in the first seismic trace gather data to obtain seismic effective data. Since the deep neural network has good non-linear mapping ability and the unsupervised deep neural network does not need to rely on actual seismic external source coherent noise, for scenarios where actual seismic external source coherent noise cannot be extracted, it can well predict the seismic external source coherent noise in the first seismic trace gather data, making the obtained first seismic external source coherent noise closer to the real seismic external source coherent noise in terms of amplitude, frequency, and phase, so as to more accurately obtain seismic effective data.
[0048] The technical solution of the present application and how the technical solution of the present application solves the above technical problems will be described in detail below with specific embodiments. The following several specific embodiments can be combined with each other, and the same or similar concepts or processes may not be repeated in some embodiments. The embodiments of the present application will be described below with reference to the accompanying drawings.
[0049] Figure 1a FIG. 1 is a schematic flow chart of a method for denoising seismic gather data provided by the present application. As Figure 1a shown, the method includes:
[0050] S101. Obtain first seismic gather data.
[0051] Among them, the first seismic gather data includes seismic external source coherent noise.
[0052] Seismic gather data refers to a data set composed of multiple seismic trace records in the process of seismic exploration. Each seismic trace records the vibration information of seismic waves at a specific position and time, and this information is stored in digital form, including parameters such as the amplitude, frequency, and phase of seismic waves.
[0053] In some embodiments, in land data acquisition, there are additional seismic sources such as pumping units and windmills that continuously operate in the work area, which will cause strong-amplitude seismic external source coherent noise in the collected seismic gather data. The number of additional seismic sources generating seismic external source coherent noise is limited and can be easily identified in the common shot gather. A windowing processing function can be used to select the first seismic gather data centered on the seismic external source coherent noise, and the first seismic gather data includes all the limited seismic external source coherent noise.
[0054] In a possible implementation manner, seismic activities can be monitored through seismic monitoring instruments (such as geophones, data collectors, clock synchronization devices, etc.) to obtain the original seismic gather data, and then the original seismic gather data is screened to filter out the seismic gather data including seismic external source coherent noise, and further the first seismic gather data is obtained. Therefore, the first seismic gather data is the seismic gather data obtained by screening and filtering the original seismic gather data.
[0055] Exemplarily, the obtained first seismic gather data can be as Figure 1b shown, Figure 1b FIG. 2 is a schematic diagram of the principle of a seismic gather data provided by the present application. The vertical axis corresponds to the time of propagation of seismic effective signals and seismic external source coherent noise, with the unit of seconds; the horizontal axis corresponds to the number of seismic wave channels; among them, the black arrow indicates the seismic external source coherent noise, and the in-phase axis of the noise is more curved, and the white arrow indicates the seismic effective signal corresponding to the seismic effective data, and the in-phase axis of the seismic effective signal is flatter.
[0056] S102. Predict the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence, so as to obtain the first seismic external source coherent noise.
[0057] Specifically, the first seismic trace gather data is composed of seismic external source coherent noise and seismic effective data. Therefore, if the seismic external source coherent noise in the first trace gather data can be predicted, the seismic effective data can be obtained based on the known first seismic trace gather data and the predicted seismic external source coherent noise. Further, the unsupervised deep neural network has good non-linear mapping ability and can well predict the seismic external source coherent noise in the first seismic trace gather data, making the obtained first seismic external source coherent noise closer to the real seismic external source coherent noise in the first trace gather data in terms of amplitude, frequency and phase.
[0058] In a possible implementation manner, input the first seismic trace gather data into an unsupervised deep neural network that has been trained to convergence, so that the unsupervised deep neural network that has been trained to convergence predicts the first seismic external source coherent noise from the first seismic trace gather data.
[0059] S103. Remove the first seismic external source coherent noise in the first seismic trace gather data to obtain seismic effective data.
[0060] Specifically, when the first seismic external source coherent noise predicted according to the seismic external source coherent noise in the first seismic trace gather data is obtained, the first trace gather data can be subtracted by the first seismic external source coherent noise to obtain seismic effective data.
[0061] Exemplarily, the obtained seismic effective data is as Figure 1c shown, Figure 1c which is the principle of a seismic trace gather data provided by this application. Figure 2 . The vertical axis corresponds to the time when the seismic wave and the seismic external source coherent noise propagate, with the unit of second; the horizontal axis corresponds to the number of seismic wave channels; among them, the white arrow indicates the seismic effective signal corresponding to the seismic effective data restored after suppressing the seismic external source coherent noise.
[0062] A denoising method for seismic trace gather data provided by an embodiment of the present application. Obtain first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise; predict the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence to obtain the first seismic external source coherent noise; remove the first seismic external source coherent noise in the first seismic trace gather data to obtain seismic effective data. Since the deep neural network has good non-linear mapping ability, and the unsupervised deep neural network does not need to rely on actual seismic external source coherent noise, for scenarios where actual seismic external source coherent noise cannot be extracted, it can well predict the seismic external source coherent noise in the first seismic trace gather data, making the obtained first seismic external source coherent noise closer to the seismic external source coherent noise in the real first trace gather data in terms of amplitude, frequency, and phase, so as to obtain seismic effective data more accurately.
[0063] Figure 2 Flow schematic of a denoising method for seismic trace gather data provided by the present application Figure 2 As Figure 2 shown, the method includes:
[0064] S201. Obtain first seismic trace gather data.
[0065] It should be noted that for the specific implementation process of step S201, reference can be made to the specific implementation process of step S101, and details will not be elaborated here.
[0066] S202. Based on the first seismic trace gather data and the second seismic external source coherent noise, train an unsupervised deep learning network to obtain an unsupervised deep neural network that has been trained to convergence.
[0067] Among them, the second seismic external source coherent noise is obtained by correcting the seismic external source coherent noise of the forward modeling of the hyperbola to a horizontal event after using a hyperbola-based moveout operator.
[0068] Specifically, for the unsupervised deep neural network, it does not need to rely on actual seismic external source coherent noise for training. Therefore, only the second seismic external source coherent noise and the first seismic trace gather data need to be input into the unsupervised deep learning network, and after multiple rounds of training until the unsupervised deep neural network is trained to convergence, the seismic external source coherent noise in the first seismic trace gather data can be accurately predicted; and the seismic external source coherent noise of the horizontal event can increase self-similarity. By correcting the seismic external source coherent noise of the forward modeling of the hyperbola to a horizontal event, the recognition effect of the deep neural network can be improved, and further the training effect of the unsupervised deep neural network can be improved.
[0069] Exemplarily, the unsupervised deep neural network can be a U-net network. The U-Net cleverly integrates multi-layer convolution operations, and the sampling information of each layer can extract rich characteristic information of seismic trace gather data after fusion. The U-net network mainly includes an encoding process and a decoding process, as Figure 3 shown Figure 3 FIG. 163 is a schematic diagram of an unsupervised deep neural network provided by the present application. The U-net network mainly includes a convolutional layer, a max-pooling layer, and an upsampling layer.
[0070] Among them, the size of the convolution kernel in the convolutional layer is 3*3. In the first 21 convolutional layers, the activation function is the Rectified Linear Unit (ReLU). When the data value input to the ReLU activation function is greater than 0, the ReLU function outputs identically. When the data value input to the ReLU activation function is less than 0, the output value of the ReLU function is 0. The original data and the amplitude of the input data input to the deep neural network are both normalized to between -1 and 1. Therefore, in the last convolutional layer of the deep neural network, the activation function is the hyperbolic tangent function (Tanh).
[0071] In some embodiments, based on the first seismic trace gather data and the second seismic external coherent noise, the unsupervised deep learning network is trained to obtain an unsupervised deep neural network that has been trained to convergence, including the following steps S11-S14:
[0072] S11. Based on the hyperbolic normal moveout operator, correct the first seismic trace gather data of the hyperbola into horizontal event axes to obtain the second seismic trace gather data.
[0073] Based on the hyperbolic normal moveout operator, multiply the hyperbolic normal moveout operator by the first seismic trace gather data, and the obtained output is the second seismic trace gather data. In this embodiment, the first seismic trace gather data of the hyperbola is corrected into the second seismic trace gather data of horizontal event axes. Since the seismic trace gather data of horizontal event axes can increase self-similarity, when the deep neural network processes the second seismic trace gather data of horizontal event axes, it can more easily capture the characteristics of similarity in waveform, amplitude, etc. of the second seismic trace gather data of horizontal event axes, thereby improving the recognition effect of the deep neural network and further improving the training effect of the unsupervised deep neural network.
[0074] Exemplarily, the hyperbolic normal moveout operator can be the hyperbolic normal moveout correction (HNMC) operator T designed for the hyperbola event axis dip angle in the forward formula.
[0075] S12. Construct a loss function of the deep neural network based on the second seismic trace gather data and the second seismic external source coherent noise.
[0076] Among them, the constructed loss function of the deep neural network is used to ensure that the deep neural network has the correct optimization direction, so as to obtain accurate predicted seismic external source coherent noise.
[0077] Exemplarily, the loss function of the deep neural network constructed based on the second seismic trace gather data and the second seismic external source coherent noise is shown in the following formula (1):
[0078] (1)
[0079] Among them, is the parameter of the deep neural network in the i-th round of training, is the move correction operator of the hyperbola, is the second seismic trace gather data in the i-th round of training, is the data output by the unsupervised deep learning network in the i-th round of training, is the second seismic external source coherent noise in the i-th round of training, is the regularization factor (represents the regularization factor), is the first seismic trace gather data in the i-th round of training, is the forward modeled seismic external source coherent noise in the i-th round of training.
[0080] S13. Input the second seismic external source coherent noise into the unsupervised deep neural network, and train the unsupervised deep neural network until the loss function of the deep neural network converges to the minimum value to obtain the optimal deep neural network parameters.
[0081] Specifically, input the second seismic external source coherent noise into the unsupervised deep neural network, and perform multiple rounds of training on the unsupervised deep neural network. During this process, by adjusting the parameters of the deep neural network, the loss function of the deep neural network is made to converge to the minimum value. When the loss function of the deep neural network converges to the minimum value, the corresponding deep neural network parameters are the optimal deep neural network parameters.
[0082] S14. Determine the unsupervised deep neural network corresponding to the optimal deep neural network parameters as the unsupervised deep neural network that has been trained to convergence.
[0083] Among them, the seismic external source coherent noise predicted by the unsupervised deep neural network corresponding to the optimal deep neural network parameters is closer to the actual seismic external source coherent noise, so as to obtain more accurate seismic effective data.
[0084] S203. Predict the seismic source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence to obtain the first seismic source coherent noise.
[0085] In some embodiments, forward model the seismic source coherent noise in the first seismic trace gather data to obtain the forward modeled seismic source coherent noise; based on the forward modeled seismic source coherent noise and the unsupervised deep neural network that has been trained to convergence, obtain the first seismic source coherent noise, where the first seismic source coherent noise is the predicted seismic source coherent noise in the first seismic trace gather data.
[0086] Specifically, since the seismic source coherent noise has good periodicity in the time domain and there is a constant frequency band in the Frequency - Wave number Domain (FK), the forward modeling method can well simulate the seismic source coherent noise in the actual seismic trace gather data to obtain the forward modeled seismic source coherent noise.
[0087] The forward modeling method refers to a method that predicts the distribution and variation of the geophysical field through mathematical calculations and simulations based on known geological models and physical laws. Exemplarily, based on the elastic wave theory, assuming that the underground medium is an elastic body, when a seismic source generates seismic waves, the seismic waves propagate in the underground medium and will undergo reflection, refraction, and transmission when encountering the interfaces of media with different properties. By collecting data related to the seismic waves, the seismic trace gather data can be obtained, and based on the seismic trace gather data, the elastic wave equation can be solved to calculate the propagation path and wave field characteristics of the seismic waves in the underground medium, thereby obtaining the forward modeled seismic source coherent noise.
[0088] Furthermore, since there are differences in amplitude, frequency, and phase between the forward modeled seismic source coherent noise and the actual seismic source coherent noise, a matching filter is needed to correct the forward modeled seismic source coherent noise. Since the deep neural network has good non - linear mapping ability and can be used as a matching filter, the forward modeled seismic source coherent noise is input into the unsupervised deep neural network to filter the forward modeled seismic source coherent noise, and the output obtained is the first seismic source coherent noise to accurately predict the seismic source coherent noise in the first seismic trace gather data.
[0089] In some embodiments, based on the forward modeled seismic source coherent noise and the unsupervised deep neural network that has been trained to convergence, obtaining the first seismic source coherent noise includes the following steps S21 - S23:
[0090] S21. Based on the hyperbolic dynamic correction operator, the seismic external source coherent noise from the forward modeling of the hyperbola is corrected into horizontal event axes to obtain the second seismic external source coherent noise.
[0091] Specifically, the kinematic positions of the seismic external source coherent noise from the forward modeling and the actual seismic external source coherent noise are similar. Therefore, the event axis position of the actual seismic external source coherent noise in the first seismic trace gather data can be marked by the event axis position of the seismic external source coherent noise from the forward modeling. Moreover, the seismic external source coherent noise with horizontal event axes can increase self-similarity. Based on the hyperbolic dynamic correction operator, multiplying the hyperbolic dynamic correction operator by the seismic external source coherent noise from the forward modeling, the output obtained is the second seismic external source coherent noise, thus correcting the seismic external source coherent noise from the forward modeling of the hyperbola into horizontal event axes, thereby improving the recognition effect of the deep neural network on the second seismic external source coherent noise to accurately predict the seismic external source coherent noise in the first seismic trace gather data.
[0092] S22. Input the second seismic external source coherent noise into the unsupervised deep neural network that has been trained to convergence, and the output obtained is the third seismic external source coherent noise.
[0093] Specifically, based on the optimal deep neural network parameters, input the second seismic external source coherent noise into the unsupervised deep neural network that has been trained to convergence to correct the differences in amplitude, frequency, and phase between the second seismic external source coherent noise and the actual seismic external source coherent noise. Then, the output obtained is the third seismic external source coherent noise. The third seismic external source coherent noise is closer to the actual seismic external source coherent noise than the second seismic external source coherent noise, so as to obtain seismic effective data more accurately.
[0094] S23. Based on the hyperbolic inverse dynamic correction operator, restore the third seismic external source coherent noise with horizontal event axes to a hyperbola to obtain the first seismic external source coherent noise.
[0095] Specifically, in order to obtain more realistic seismic external source coherent noise, based on the hyperbolic inverse dynamic correction operator, multiply the hyperbolic inverse dynamic correction operator by the third seismic external source coherent noise, and the output is the first seismic external source coherent noise, thus restoring the third seismic external source coherent noise with horizontal event axes to the first seismic external source coherent noise of a hyperbola, and then obtaining the first seismic external source coherent noise that is closer to the actual seismic external source coherent noise.
[0096] S204. Remove the first seismic external source coherent noise from the first seismic trace gather data to obtain seismic effective data.
[0097] In a possible implementation, after removing the first seismic external source coherent noise in the first seismic trace gather data, if the obtained seismic effective data still has residual seismic external source coherent noise, steps S201 - S204 can be repeated until all the seismic external source coherent noise is removed.
[0098] A denoising method for seismic trace gather data provided by an embodiment of the present application. Obtain first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise; based on the first seismic trace gather data and the second seismic external source coherent noise, train an unsupervised deep learning network to obtain an unsupervised deep neural network that has been trained to convergence; predict the seismic external source coherent noise in the first seismic trace gather data through the unsupervised deep neural network that has been trained to convergence to obtain the first seismic external source coherent noise; remove the first seismic external source coherent noise in the first seismic trace gather data to obtain seismic effective data. Since the deep neural network has good non - linear mapping ability, and for scenarios where actual seismic external source coherent noise cannot be extracted, combined with the loss function of the deep neural network, the unsupervised deep neural network does not need to rely on actual seismic external source coherent noise for training, and can well predict the seismic external source coherent noise in the first seismic trace gather data, making the obtained first seismic external source coherent noise closer to the seismic external source coherent noise in the real first trace gather data in terms of amplitude, frequency, and phase, so as to obtain seismic effective data more accurately.
[0099] Figure 4 It is a schematic structural diagram of a denoising device for seismic trace gather data provided by the present application. As Figure 4 shown, the denoising device 40 for seismic trace gather data provided in this embodiment includes:
[0100] An acquisition module 401, configured to acquire first seismic trace gather data, where the first seismic trace gather data includes seismic external source coherent noise.
[0101] A processing module 402, configured to predict the seismic external source coherent noise in the first seismic trace gather data through an unsupervised deep neural network that has been trained to convergence to obtain the first seismic external source coherent noise.
[0102] A denoising module 403, configured to remove the first seismic external source coherent noise in the first seismic trace gather data to obtain seismic effective data.
[0103] In a possible implementation, the processing module 402 is specifically configured to:
[0104] Perform forward modeling on the seismic external source coherent noise in the first seismic trace gather data to obtain the forward - modeled seismic external source coherent noise;
[0105] Based on the forward simulation of seismic external source coherent noise and an unsupervised deep neural network that has been trained to convergence, the first seismic external source coherent noise is obtained. The first seismic external source coherent noise is the seismic external source coherent noise in the predicted first seismic trace gather data.
[0106] In a possible implementation manner, the processing module 402 is specifically configured to:
[0107] Based on the hyperbolic normal moveout operator, correct the forward-simulated seismic external source coherent noise of the hyperbola into horizontal event axes to obtain the second seismic external source coherent noise;
[0108] Input the second seismic external source coherent noise into the unsupervised deep neural network that has been trained to convergence, and the obtained output is the third seismic external source coherent noise;
[0109] Based on the inverse hyperbolic normal moveout operator, restore the third seismic external source coherent noise of the horizontal event axes to a hyperbola to obtain the first seismic external source coherent noise.
[0110] In a possible implementation manner, the processing module 402 is further configured to:
[0111] Based on the first seismic trace gather data and the second seismic external source coherent noise, train the unsupervised deep learning network to obtain an unsupervised deep neural network that has been trained to convergence. The second seismic external source coherent noise is obtained after correcting the forward-simulated seismic external source coherent noise of the hyperbola into horizontal event axes based on the hyperbolic normal moveout operator.
[0112] In a possible implementation manner, the processing module 402 is specifically configured to:
[0113] Based on the hyperbolic normal moveout operator, correct the first seismic trace gather data of the hyperbola into horizontal event axes to obtain the second seismic trace gather data;
[0114] Based on the second seismic trace gather data and the second seismic external source coherent noise, construct a deep neural network loss function;
[0115] Input the second seismic external source coherent noise into the unsupervised deep neural network, and train the unsupervised deep neural network until the deep neural network loss function converges to the minimum value to obtain the optimal deep neural network parameters;
[0116] Determine the unsupervised deep neural network corresponding to the optimal deep neural network parameters as the unsupervised deep neural network that has been trained to convergence.
[0117] The denoising device for seismic trace gather data provided in this embodiment can execute the method provided in the above method embodiment, and its implementation principle and technical effects are similar. Details are not described herein again.
[0118] Figure 5 A structural schematic diagram of an electronic device provided for this application. As Figure 5 shown, the electronic device 50 provided in this embodiment includes: at least one processor 501 and a memory 502. Optionally, the device 50 further includes a communication component 503. Among them, the processor 501, the memory 502, and the communication component 503 are connected through a bus 504.
[0119] In the specific implementation process, at least one processor 501 executes the computer-executable instructions stored in the memory 502, so that at least one processor 501 executes the above method.
[0120] For the specific implementation process of the processor 501, reference can be made to the above method embodiment, and its implementation principle and technical effect are similar, so they will not be elaborated here in this embodiment.
[0121] In the above embodiment, it should be understood that the processor may be a central processing unit (English: Central Processing Unit, abbreviated: CPU), or other general-purpose processors, digital signal processors (English: Digital Signal Processor, abbreviated: DSP), application specific integrated circuits (English: Application Specific Integrated Circuit, abbreviated: ASIC), etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc. The steps of the method disclosed in combination with the invention can be directly embodied as being executed by a hardware processor, or executed by a combination of hardware and software modules in the processor.
[0122] The memory may include a high-speed memory (Random Access Memory, RAM), and may also include a non-volatile memory (Non-volatile Memory, NVM), such as at least one disk memory.
[0123] The bus may be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. The bus can be divided into an address bus, a data bus, a control bus, etc. For the convenience of representation, the bus in the drawings of this application is not limited to only one bus or one type of bus.
[0124] The present application also provides a computer program product, including a computer program which, when executed by a processor, implements the above-mentioned method.
[0125] The present application also provides a computer-readable storage medium storing computer-executable instructions, which, when executed by a processor, implement the above-mentioned method.
[0126] The above-mentioned readable storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, a magnetic disk or an optical disk. The readable storage medium can be any available medium accessible by a general-purpose or special-purpose computer.
[0127] An exemplary readable storage medium is coupled to a processor, enabling the processor to read information from and write information to the readable storage medium. Of course, the readable storage medium can also be a component of the processor. The processor and the readable storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the readable storage medium can also exist as discrete components in a device.
[0128] The division of units is only a logical function division. In actual implementation, there may be other division methods. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the displayed or discussed couplings or direct couplings or communication connections to each other can be indirect couplings or communication connections through some interfaces, devices or units, and can be in electrical, mechanical or other forms.
[0129] The units described as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units, that is, they can be located in one place or distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.
[0130] In addition, in each embodiment of the present invention, the functional units can be integrated in one processing unit, or each unit can exist physically alone, or two or more units can be integrated in one unit.
[0131] If a function is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art or a part of this technical solution can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods of the various embodiments of the present invention. The aforementioned storage medium includes: USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs and other various media that can store program codes.
[0132] Those of ordinary skill in the art can understand that all or part of the steps for implementing the above method embodiments can be completed by hardware related to program instructions. The aforementioned program can be stored in a computer-readable storage medium. When this program is executed, it executes the steps including the above method embodiments; and the aforementioned storage medium includes: ROMs, RAMs, magnetic disks, or optical discs and other various media that can store program codes.
[0133] Finally, it should be noted that: After considering the specification and practicing the invention disclosed herein, those skilled in the art will easily think of other implementation manners of the present invention. The present invention is intended to cover any variations, uses, or adaptations of the present invention, which follow the general principles of the present invention and include the common general knowledge or conventional technical means in the technical field not disclosed in the present invention. It is not limited to the exact structures described above and shown in the drawings, and various modifications and changes can be made without departing from its scope. The scope of the present invention is only limited by the appended claims.
Claims
1. A method for denoising seismic gather data, characterized in that: include: Acquiring first seismic gather data, wherein the first seismic gather data includes seismic exogenous coherent noise; Predicting the seismic exogenous coherent noise in the first seismic trace gather data by using an unsupervised deep neural network that has been trained to convergence to obtain the first seismic exogenous coherent noise; The first seismic external source coherent noise in the first seismic track gather data is removed to obtain seismic valid data.
2. The method according to claim 1, characterized in that: The method of predicting the seismic exogenous coherent noise in the first seismic gather data by using an unsupervised deep neural network that has been trained to convergence to obtain the first seismic exogenous coherent noise comprises: By forward modeling the seismic exogenous coherent noise in the first seismic gather data, a forward modeled seismic exogenous coherent noise is obtained; Based on the forward simulated seismic exogenous coherent noise and the unsupervised deep neural network that has been trained to convergence, a first seismic exogenous coherent noise is obtained, and the first seismic exogenous coherent noise is the predicted seismic exogenous coherent noise in the first seismic track gather data.
3. The method according to claim 2, characterized in that The method of obtaining a first seismic exogenous coherent noise based on the forward modeling and the unsupervised deep neural network trained to convergence comprises: Based on the dynamic correction operator of the hyperbola, the seismic external source coherent noise simulated by the forward modeling of the hyperbola is corrected to the horizontal phase axis to obtain the second seismic external source coherent noise; Inputting the second seismic exogenous coherent noise into the unsupervised deep neural network that has been trained to convergence, and obtaining an output as a third seismic exogenous coherent noise; Based on the hyperbola's dynamic correction inverse transformation operator, the third seismic exogenous coherent noise of the horizontal event axis is restored to a hyperbola to obtain the first seismic exogenous coherent noise.
4. The method according to claim 1, characterized in that: Before predicting the seismic exogenous coherent noise in the first seismic trace gather data by using the unsupervised deep neural network that has been trained to convergence to obtain the first seismic exogenous coherent noise, the method further includes: Based on the first seismic track gather data and the second seismic exogenous coherent noise, an unsupervised deep learning network is trained to obtain an unsupervised deep neural network that has been trained to convergence. The second seismic exogenous coherent noise is obtained by correcting the seismic exogenous coherent noise of the forward simulation of the hyperbola into a horizontal phase axis based on a dynamic correction operator of the hyperbola.
5. The method according to claim 4, characterized in that The step of training an unsupervised deep learning network based on the first seismic gather data and the second seismic exogenous coherent noise to obtain an unsupervised deep neural network that has been trained to convergence includes: Based on the dynamic correction operator of the hyperbola, the first seismic gather data of the hyperbola is corrected to a horizontal event axis to obtain the second seismic gather data; Constructing a deep neural network loss function based on the second seismic gather data and the second seismic exogenous coherent noise; Inputting the second seismic exogenous coherent noise into the unsupervised deep neural network, training the unsupervised deep neural network until the deep neural network loss function converges to a minimum value, so as to obtain optimal deep neural network parameters; The unsupervised deep neural network corresponding to the optimal deep neural network parameters is determined as the unsupervised deep neural network that has been trained to convergence.
6. A denoising device for seismic gather data, characterized in that: include: An acquisition module, used for acquiring first seismic gather data, wherein the seismic gather data includes seismic exogenous coherent noise; A processing module, configured to predict the seismic exogenous coherent noise in the first seismic trace gather data by using an unsupervised deep neural network that has been trained to convergence, so as to obtain a first seismic exogenous coherent noise; A denoising module is used to remove the first seismic external source coherent noise in the seismic track gather data to obtain effective seismic data.
7. An electronic device, characterized in that: include: Memory, processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory, so that the processor performs the method according to any one of claims 1 to 5.
8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 5 when executed by a processor.
9. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 5 when being executed by a processor.