Training method for strong wave impedance elimination model and method and device for eliminating strong wave impedance.

Abstract

This invention provides a training method for a strong wave impedance elimination model, comprising the following steps: using seismic sample data and seismic data after strong wave impedance elimination, training a preset neural network model to obtain the strong wave impedance elimination model; the preset neural network model is a convolutional neural network model, and during training, the convolutional neural network model uses a total variation regularization term to calculate the loss function. Calculating the loss function of the convolutional neural network using a total variation regularization term, the total variation algorithm can remove noise from the seismic data, improve the continuity of the phase axis of the seismic attributes, and efficiently establish a strongly nonlinear mapping relationship between the original seismic data and the seismic data after strong wave impedance elimination through the convolutional neural network, significantly improving the accuracy of seismic data identification.

Description

Technical Field

[0001] This invention relates to the field of geophysical exploration, and in particular to a training method for a strong wave impedance elimination model and a method and apparatus for eliminating strong wave impedance. Background Technology

[0002] In oil and gas geophysical exploration, when seismic waves propagate from one medium to another, some of the energy is reflected, while the rest continues to penetrate into the next layer of medium. If the reflection coefficients of the subsurface interfaces differ significantly, the amplitude of the reflected wave will be much greater than that of the transmitted wave, resulting in strong reflected signals in the seismic record. These reflected signals are interference caused by strong wave impedance, leading to the inability to identify reflected signals from nearby reservoirs.

[0003] The formation of strong wave impedance is relatively complex, including differences in reflection coefficients at stratigraphic interfaces, medium inhomogeneity, and multipath propagation of seismic waves. Traditional methods for eliminating strong wave impedance either have multiple solutions or lead to a decrease in the accuracy of seismic data during the process. Summary of the Invention

[0004] In view of the above problems, the present invention is proposed to provide a training method for a strong wave impedance elimination model and a method and apparatus for eliminating strong wave impedance to overcome or at least partially solve the above problems.

[0005] In a first aspect, embodiments of the present invention provide a training method for a strong wave impedance cancellation model, comprising:

[0006] The pre-acquired seismic sample data with strong wave impedance eliminated is used as the label data corresponding to the seismic sample data.

[0007] Based on the earthquake sample data and the corresponding label data, a preset neural network model is trained to obtain a strong wave impedance cancellation model; the preset neural network model is a convolutional neural network model, and during training, the convolutional neural network model uses a total variation regularization term to calculate the loss function.

[0008] In one embodiment, the loss function of the convolutional neural network is defined as follows:

[0009]

[0010] Where: Y true For label data; Y pred For prediction data; smooth is a preset constant; lambda is a preset coefficient; TV(Y) pred ) represents the total variation of the predicted data.

[0011] In one embodiment, the calculation of the total variation of the predicted data during the training of the neural network includes:

[0012] For each dimension of the predicted data, the absolute value of the difference between the predicted data and the gradient is calculated to obtain the sum of the absolute values ​​of the dimensions; the gradient data is the maximum value of the first derivative of the function of the predicted data obtained in advance;

[0013] The total variation of the predicted data is obtained by summing the absolute values ​​of each dimension of the predicted data.

[0014] In one embodiment, the pre-acquired seismic data with high impedance eliminated is obtained in the following manner:

[0015] By employing a matching pursuit algorithm or a multi-wavelet decomposition and reconstruction method to remove strong reflections, the strong wave impedance of the original seismic data is eliminated, resulting in seismic data with strong wave impedance removed.

[0016] In one embodiment, after eliminating the strong wave impedance of the seismic data and before the step of training the neural network, the method further includes:

[0017] The total variation algorithm is used to denoise seismic data after eliminating strong wave impedance.

[0018] In one embodiment, the convolutional neural network is a UNet network.

[0019] Secondly, embodiments of the present invention provide a method for eliminating strong wave impedance in seismic data, comprising:

[0020] The earthquake data to be predicted is input into the strong wave impedance elimination model to obtain earthquake data with strong wave impedance eliminated; the strong wave impedance elimination model is trained by the training method of the strong wave impedance elimination model.

[0021] Thirdly, embodiments of the present invention provide a training apparatus for a strong wave impedance cancellation model, comprising:

[0022] The tag creation module is used to use pre-acquired earthquake sample data with strong wave impedance eliminated as the tag data corresponding to the earthquake sample data.

[0023] The training module is used to train a preset neural network model based on the earthquake sample data and the corresponding label data to obtain a strong wave impedance cancellation model; the preset neural network model is a convolutional neural network model, and the regularization term of the loss function of the convolutional neural network model is a total variation regularization term.

[0024] Fourthly, embodiments of the present invention provide an apparatus for eliminating strong wave impedance in seismic data, comprising:

[0025] The elimination module is used to input the earthquake data to be predicted into the strong wave impedance elimination model to obtain earthquake data with strong wave impedance eliminated; the strong wave impedance elimination model is trained by the training method of the strong wave impedance elimination model.

[0026] Fifthly, embodiments of the present invention provide a computing device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the program executed by the processor implements a training method for a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data.

[0027] In a sixth aspect, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements a training method for a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data.

[0028] In a seventh aspect, embodiments of the present invention provide a computer program product, the computer program product including a computer program, which, when executed by a processor, implements a training method for a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data.

[0029] The beneficial effects of the above-described technical solutions provided in the embodiments of the present invention include at least the following:

[0030] This invention provides a training method for a strong wave impedance elimination model, comprising the following steps: using seismic sample data and corresponding label data, training a preset neural network model to obtain the strong wave impedance elimination model; the preset neural network model is a convolutional neural network model, and during training, the convolutional neural network model uses a total variation regularization term to calculate the loss function. Calculating the loss function of the convolutional neural network using a total variation regularization term, the total variation algorithm can remove noise from the seismic data, improve the continuity of the phase axis of the seismic attributes, and efficiently establish a strongly nonlinear mapping relationship between the original seismic data and the seismic data with eliminated strong wave impedance through the convolutional neural network, significantly improving the accuracy of seismic data identification.

[0031] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description, claims, and drawings.

[0032] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0033] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings:

[0034] Figure 1 A flowchart illustrating the training method for the strong wave impedance elimination model provided in this embodiment of the invention;

[0035] Figure 2 Earthquake sample data diagram provided for embodiments of the present invention;

[0036] Figure 3 This is a seismic data map provided by an embodiment of the present invention after eliminating strong wave impedance using a matching pursuit algorithm;

[0037] Figure 4 This is a structural block diagram of the matching and tracking algorithm provided in an embodiment of the present invention;

[0038] Figure 5 This is a seismic data image provided by an embodiment of the present invention after eliminating strong wave impedance using a Unet network and the TV total variation algorithm;

[0039] Figure 6 Seismic profile of the original seismic data provided in the embodiments of the present invention;

[0040] Figure 7 This is a seismic profile of seismic data processed by the method for eliminating strong wave impedance in seismic data provided in this embodiment of the invention.

[0041] Figure 8 Waveform diagram of seismic profile of raw seismic data provided in the embodiments of the present invention;

[0042] Figure 9 The waveform of the seismic profile of the seismic data after processing by the method for eliminating strong wave impedance of seismic data provided in this embodiment of the invention;

[0043] Figure 10 The maximum amplitude attribute map of seismic data after eliminating strong wave impedance using conventional methods provided in embodiments of the present invention;

[0044] Figure 11 The maximum amplitude attribute map of seismic data after processing by the method for eliminating strong wave impedance of seismic data provided in this embodiment of the invention;

[0045] Figure 12 This is a structural block diagram of the training method for the strong wave impedance elimination model provided in an embodiment of the present invention. Detailed Implementation

[0046] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0047] The flowchart of the training method for the strong wave impedance cancellation model provided in this embodiment of the invention can be, for example, as shown below. Figure 1 As shown, it includes the following steps:

[0048] S11. Use the pre-acquired seismic sample data with strong wave impedance eliminated as the label data corresponding to the seismic sample data;

[0049] Earthquake sample data graphs can be, for example, as follows Figure 2 As shown, this is the raw seismic data used for training; the labeled data can be, for example, as shown in the image. Figure 3 As shown;

[0050] S12. Based on the earthquake sample data and the corresponding label data, train the preset neural network model to obtain the strong wave impedance cancellation model; the preset neural network model is a convolutional neural network model, and during training, the convolutional neural network model uses the total variation regularization term to calculate the loss function.

[0051] The aforementioned convolutional neural network can be exemplified by the UNet neural network, a type of convolutional neural network primarily used for medical image segmentation. The UNet network structure mainly consists of an encoder and a decoder, forming a U-shaped structure. The encoder is responsible for feature extraction, and the decoder is responsible for reconstructing the feature information.

[0052] Encoder: The encoder part consists of multiple convolutional and pooling layers for feature extraction and dimensionality reduction. Each convolutional layer is typically followed by an activation function, and the pooling layers are used for downsampling to further extract features and reduce the size of the feature map.

[0053] Decoder: The decoder consists of upsampling layers and convolutional layers for feature map reconstruction. The upsampling layers restore the feature map to the same size as the input image, and the convolutional layers refine the feature map and generate the final segmentation result.

[0054] The UNet neural network has a U-shaped architecture, with the encoder and decoder connected via skip connections. Skip connections concatenate the encoder's and decoder's feature maps, helping to recover detailed information and improve segmentation accuracy. The encoder uses convolution and pooling operations for feature extraction and dimensionality reduction, while the decoder uses upsampling and convolution operations for feature map restoration and refinement. Skip connections: By concatenating the encoder's and decoder's feature maps, skip connections help recover detailed information and solve the gradient vanishing problem.

[0055] The loss function of the convolutional neural network is calculated using the total variation regularization term. The total variation algorithm can remove noise from the seismic data and improve the continuity of the phase axis of the seismic attributes. Furthermore, the convolutional neural network can efficiently establish a strong nonlinear mapping relationship between the original seismic data and the seismic data with strong wave impedance eliminated, which significantly improves the accuracy of seismic data identification.

[0056] Before implementing methods for eliminating strong wave impedance in seismic data and training methods for strong wave impedance elimination models, for example, time-window processing can be performed on the strong wave impedance region of the seismic data to be eliminated, and the seismic data within the time window can be used as seismic sample data. This not only provides uniform seismic sample data, which is convenient for training the strong wave impedance elimination model, but also helps to improve the accuracy of the strong wave impedance elimination model in eliminating strong wave impedance, thereby improving the effectiveness of the strong wave impedance elimination model in eliminating strong wave impedance.

[0057] The seismic data with eliminated strong wave impedance, which was acquired in step S11 above, can be obtained, for example, in the following manner:

[0058] The matching pursuit algorithm is used to remove strong reflections, eliminating strong wave impedance in seismic sample data, resulting in seismic data with eliminated strong wave impedance, such as... Figure 3 As shown; earthquake sample data, for example, are as follows: Figure 2 As shown; the aforementioned elimination of strong wave impedance in seismic data can also be accomplished using existing techniques for eliminating strong wave impedance in seismic data, and the embodiments of the present invention do not limit this.

[0059] The aforementioned matching pursuit algorithm is a sparse representation method for signal processing. Its basic idea is to represent the signal as a linear combination of a set of basis functions and iteratively approximate the original signal. The process can be illustrated as follows: Figure 4 As shown. The aforementioned multi-wavelet decomposition and reconstruction strong reflection stripping method is a signal processing method. Its basic principle is: to select seismic wavelets of different frequency bands to reconstruct the original signal, and to remove the components of the strong wave impedance frequency band from the reconstructed components.

[0060] After eliminating the strong wave impedance of the seismic data and before the aforementioned step S11, the training method for the strong wave impedance elimination model may, for example, include:

[0061] The total variational algorithm is used to denoise seismic sample data after eliminating strong impedance. This reduces noise in the seismic sample data after eliminating strong impedance and also improves the continuity of the in-phase axis after eliminating strong impedance using conventional methods.

[0062] The aforementioned phase axis refers to the seismic reflection properties, namely the components of a seismic profile; the physical seismological properties of the phase axis include: amplitude, frequency, and continuity; amplitude refers to the displacement of a particle from its equilibrium position; frequency reflects the distance between adjacent reflecting interfaces; and continuity refers to the lateral extension of the amplitude and frequency of the phase axis.

[0063] In step S12, the loss function of the convolutional neural network is defined as follows:

[0064]

[0065] Where: Y true For label data; Y pred For prediction data; smooth is a preset constant; lambda is a preset coefficient; TV(Y) pred ) represents the total variation of the predicted data.

[0066] TV(Y pred The calculation of ) can be performed, for example, in the following way:

[0067] For each dimension of the predicted data, calculate the absolute value of the difference between the predicted data and the gradient to obtain the sum of the absolute values ​​of the dimensions; the gradient data is the maximum value of the first derivative of the function of the pre-acquired predicted data; add the sum of the absolute values ​​of each dimension of the predicted data to obtain the total variation of the predicted data.

[0068] Based on the same inventive concept, embodiments of the present invention also provide a method for eliminating strong wave impedance in seismic data using the aforementioned strong wave impedance elimination model, comprising:

[0069] The earthquake sample data and the earthquake data after the strong wave impedance was eliminated by the matching pursuit algorithm are used together as label data and input into the strong wave impedance elimination model to obtain the earthquake data with the strong wave impedance eliminated. The strong wave impedance elimination model can be trained, for example, by the training method of the strong wave impedance elimination model provided in the embodiment of the present invention. The label data is divided into training set and test set according to an 8:2 ratio.

[0070] Therefore, earthquake sample data, such as Figure 2The image shows the original seismic data used for prediction. Seismic data for which strong wave impedance is eliminated using the strong wave impedance elimination model provided in this embodiment of the invention can be, for example, as shown below. Figure 5 As shown; comparison Figure 3 and Figure 5 It can be seen that the strong wave impedance elimination model provided by the embodiments of the present invention predicts the elimination of strong wave impedance. Figure 5 The continuity of the in-phase axis of the seismic data in Figure 3 In seismic data, good phase axis continuity is desirable; better phase axis continuity indicates a lower impact of strong wave impedance on the seismic data. For example, according to... Figure 3 The lateral discontinuity between the red and blue regions in the 2160ms-2200ms range indicates that the continuity of the in-phase axis of the aforementioned common center point gather is poor.

[0071] exist Figure 2 , Figure 3 and Figure 5 In this context, CMP stands for Comment Middle Point, which refers to the collection of seismic traces in the observation system where the center point of the detector and the excitation point are the same. Red represents peak amplitude, blue represents trough amplitude, and white represents inflection point amplitude. Darker colored areas represent areas with strong amplitude, while lighter colored areas represent areas with weak amplitude.

[0072] To verify the effectiveness of the method for eliminating strong wave impedance in seismic data provided in this embodiment of the invention, the seismic profile of the original seismic data and the seismic profile of the seismic data after eliminating strong wave impedance using the method provided in this embodiment of the invention can be compared. Additionally, the waveform of the seismic profile of the original seismic data can be compared with the waveform of the seismic profile of the seismic data after eliminating strong wave impedance using the method provided in this embodiment of the invention. For example, the seismic profile of the original seismic data and the seismic profile of the seismic data after eliminating strong wave impedance can be imported into the GeoEaset software. The seismic profile of the original seismic data can be, for example, as shown in the example... Figure 6 As shown in the embodiment of the present invention, the seismic profile of the seismic data after eliminating the strong wave impedance can be, for example, as shown in the following example. Figure 7 As shown; the method for eliminating strong wave impedance in seismic data provided in this embodiment of the invention can clearly show the strong wave impedance interface in the seismic profile of the seismic data. Figure 6 and Figure 7 The geological information of the weakly reflective layer beneath the circled area (indicated by the arrow, the cyan interface) is even more evident in the waveform display of the profile. For example, the waveform of the seismic profile from the original seismic data can be seen as follows: Figure 8As shown, the waveform diagram of the seismic profile of the seismic data after eliminating strong wave impedance according to the present invention can be, for example, as shown in the figure. Figure 9 As shown, according to Figure 8 The circled part and Figure 9 By comparing the circled parts, it can be seen that the seismic data after eliminating strong wave impedance according to the embodiment of the present invention has better continuity of the in-phase axis. Figures 6-9 The three different colors in the image correspond to the peaks, troughs, and inflection points of the seismic data, respectively.

[0073] To further verify the effectiveness of the method for eliminating strong wave impedance in seismic data provided in this embodiment of the invention, taking seismic data in image form as an example, this embodiment of the invention also provides three evaluation metrics for image similarity, including: Mean Square Error (MSE), Peak Signal to Noise Ratio (PSNR), and Structural Similarity (SSIM). Calculations show that after eliminating strong wave impedance in seismic data using existing methods, the MSE of the seismic data with eliminated strong wave impedance compared to the original seismic data is 3.8e, the PSNR is 4.1, and the SSIM is 0.87. After eliminating seismic data using the method provided in this embodiment of the invention, the MSE of the seismic data with eliminated strong wave impedance compared to the original seismic data is 2.6e, the PSNR is 25.8, and the SSIM is 0.95. By comparing the similarity between seismic data with strong wave impedance eliminated by conventional methods and the original seismic data, and the similarity between seismic data with strong wave impedance eliminated by the method provided in this embodiment of the invention, it can be seen that the method provided in this embodiment of the invention has a lower MSE index, which describes noise, indicating that the seismic data processed by the method provided in this embodiment of the invention has less noise. At the same time, the PSNR and SSIM indexes, which describe image similarity, are higher in the method provided in this embodiment of the invention, indicating that the method provided in this embodiment of the invention is more similar to the original seismic data than conventional methods.

[0074] For example, a comparison can be made between the maximum amplitude attribute of strong wave impedance in seismic data eliminated by conventional methods and the maximum amplitude attribute of strong wave impedance in seismic data eliminated by the method provided in this embodiment of the invention. For example, the maximum amplitude attribute diagram of seismic data after eliminating strong wave impedance using conventional methods can be shown as follows: Figure 10 As shown in the embodiment of the present invention, the method for eliminating strong wave impedance in seismic data, after eliminating strong wave impedance, results in a maximum amplitude attribute map of the seismic data, which can be illustrated as follows: Figure 11 As shown; in Figure 10 and Figure 11 In the diagram, L and T refer to the corresponding line number and trace number, respectively, and the numbers represent the line number and trace number. It can be seen that compared to conventional methods, the method for eliminating strong wave impedance in seismic data provided in this embodiment of the invention eliminates strong wave impedance, resulting in seismic data in the weak reflection layer (…). Figure 11 The energy of the red part of L6436 between T6351 and T6551 is stronger, indicating that the method for eliminating strong wave impedance in seismic data provided in this embodiment of the invention can effectively eliminate strong wave impedance, significantly improve the accuracy of underground structure identification, and provide data support for subsequent high-precision reservoir prediction.

[0075] Based on the same inventive concept, this invention also provides a training device for a strong wave impedance elimination model, the structural block diagram of which is shown below. Figure 12 As shown, it includes:

[0076] The tag creation module 121 is used to use the pre-acquired seismic sample data with strong wave impedance eliminated as the tag data corresponding to the seismic sample data.

[0077] Training module 122 is used to train a preset neural network model based on earthquake sample data and the corresponding label data to obtain a strong wave impedance cancellation model; the preset neural network model is a convolutional neural network model, and the regularization term of the loss function of the convolutional neural network model is a total variation regularization term.

[0078] Based on the same inventive concept, embodiments of the present invention also provide an apparatus for eliminating strong wave impedance in seismic data, comprising:

[0079] The elimination module is used to input the earthquake data to be predicted into the strong wave impedance elimination model to obtain earthquake data with strong wave impedance eliminated; the strong wave impedance elimination model can be trained, for example, by the training method of the strong wave impedance elimination model provided in the embodiments of the present invention.

[0080] Based on the same inventive concept, embodiments of the present invention also provide a computing device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The program executed by the processor implements a training method for a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data.

[0081] Based on the same inventive concept, embodiments of the present invention also provide a computer-readable storage medium storing a computer program, which, when executed by a processor, implements a training method for a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data.

[0082] Based on the same inventive concept, embodiments of the present invention also provide a computer program product, which includes a computer program that, when executed by a processor, implements a training method for a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data.

[0083] Since the principle by which these devices solve problems is similar to the training method of a strong wave impedance elimination model or a method for eliminating strong wave impedance in seismic data, the implementation of these devices can be referred to the implementation of the aforementioned methods, and the repetitions will not be repeated.

[0084] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage and optical storage) containing computer-usable program code.

[0085] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0086] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0087] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1The steps of the functions specified in one or more boxes. Obviously, those skilled in the art can make various modifications and variations to this invention without departing from the spirit and scope of the invention. Therefore, if these modifications and variations of the invention fall within the scope of the claims of the invention and their equivalents, the invention is also intended to include these modifications and variations.

Claims

1. A method for training a strong wave impedance cancellation model, characterized in that, The method comprises the following steps: obtaining, in advance, seismic sample data with strong wave impedance removed as label data corresponding to the seismic sample data; training a preset neural network model according to the seismic sample data and the label data corresponding to the seismic sample data to obtain a strong wave impedance removal model; the preset neural network model is a convolutional neural network model, and the convolutional neural network model uses a total variation regular term to calculate a loss function during training.

2. The method of claim 1, wherein, The loss function of the convolutional neural network is defined as follows: wherein: Y true is the label data; Y pred is the prediction data; smooth is a pre-set constant; lambda is a pre-set coefficient; TV(Y pred ) is the total variation of the prediction data.

3. The method of claim 2, wherein, During the training of the neural network, the calculation of the total variation of the prediction data comprises the following steps: for each dimension of the prediction data, calculating the absolute value of the difference between the prediction data and the gradient to obtain the sum of the absolute values of the dimension; the gradient data is the maximum value of the first derivative of the function of the prediction data obtained in advance; adding the sum of the absolute values of each dimension of the prediction data to obtain the total variation of the prediction data.

4. The method according to any one of claims 1 to 3, characterized in that, The seismic data with strong wave impedance removed obtained in advance is obtained by the following method: using a matching pursuit algorithm or a multi-wavelet decomposition and reconstruction strong reflection stripping method to remove the strong wave impedance of the original seismic data to obtain seismic data with strong wave impedance removed.

5. The method of claim 4, wherein, After removing the strong wave impedance of the seismic data, before the step of training the neural network, the method further comprises the following steps: using a total variation algorithm to denoise the seismic data with strong wave impedance removed.

6. The method according to any one of claims 1 to 5, wherein, The convolutional neural network is a UNet network.

7. A method of removing strong impedance from seismic data, characterized by, The method comprises the following steps: inputting the seismic data to be predicted into the strong wave impedance removal model to obtain seismic data with strong wave impedance removed; The strong wave impedance removal model is trained by the training method of the strong wave impedance removal model according to any one of claims 1-6.

8. A training device of a strong wave impedance elimination model, characterized in that, The method comprises the following steps: a label making module for obtaining, in advance, seismic sample data with strong wave impedance removed as label data corresponding to the seismic sample data; a training module for training a preset neural network model according to the seismic sample data and the label data corresponding to the seismic sample data to obtain a strong wave impedance removal model; the preset neural network model is a convolutional neural network model, and the regular term of the loss function of the convolutional neural network model is a total variation regular term.

9. An apparatus for removing strong impedance in seismic data, characterized by, The method comprises the following steps: a removal module for inputting the seismic data to be predicted into the strong wave impedance removal model to obtain seismic data with strong wave impedance removed; The strong wave impedance removal model is trained by the training method of the strong wave impedance removal model according to any one of claims 1-6.

10. A computing device, comprising: The method comprises the following steps: a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the processor executes the program to implement the training method of the strong wave impedance removal model according to any one of claims 1-6 or the method for removing strong wave impedance of seismic data according to claim 7.

11. A computer readable storage medium characterized by, The computer readable storage medium stores a computer program, and the computer program is executed by the processor to implement the training method of the strong wave impedance removal model according to any one of claims 1-6 or the method for removing strong wave impedance of seismic data according to claim 7.

12. A computer program product, characterised in that, The computer program product includes a computer program that, when executed by a processor, implements a training method for a strong wave impedance elimination model as described in any one of claims 1-6 or a method for eliminating strong wave impedance in seismic data as described in claim 7.

Images