OVT domain self-supervision seismic data reconstruction method and device, electronic equipment and medium
By transforming seismic data to the OVT domain and using convolutional neural networks for self-supervised learning, the shortcomings of existing seismic data reconstruction methods are addressed, achieving high-precision seismic data reconstruction and improving the reliability of geological data interpretation.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-29
- Publication Date
- 2026-03-31
AI Technical Summary
Existing seismic data reconstruction methods are difficult to apply under complex geological conditions, are sensitive to noise, have limited sampling rates, and lack reconstruction accuracy. Furthermore, deep learning methods suffer from data volume dependence, training complexity, and overfitting issues, which affect the reliability of geological data interpretation.
Seismic data is transformed into the OVT domain, and self-supervised learning is performed using convolutional neural networks. By constructing training, validation, and test datasets, a weighted sum of root mean square error and structural similarity error functions is used as the loss function. Feature extraction and reconstruction are performed by combining Unet network, residual module, and convolutional attention module.
It improves the accuracy and completeness of earthquake data reconstruction, enhances the ability to reconstruct local detailed features, and improves the precision and reliability of earthquake data processing.
Smart Images
Figure CN121763365A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of seismic data reconstruction technology, and more specifically, to an OVT domain self-supervised seismic data reconstruction method, apparatus, electronic device, and medium. Background Technology
[0002] During seismic exploration and acquisition, data is often incomplete or irregular due to limitations in ground infrastructure, budget, time, and data processing and transmission. These issues severely impact the fidelity, resolution, and signal-to-noise ratio of subsequent data processing, thereby weakening the reliability of the final geological data interpretation. To improve the completeness of seismic data, it is necessary to reconstruct such data. High-quality reconstruction of missing seismic data is crucial for improving the accuracy of subsequent seismic data processing and interpretation.
[0003] Currently, the main methods for seismic data reconstruction encompass a variety of techniques, including wave equation-based methods, predictive filtering, mathematical transformations, and compressed sensing. However, these methods all have certain limitations. For example, wave equation-based methods may be difficult to apply under complex geological conditions; predictive filtering methods may be sensitive to noise; mathematical transformation-based methods may produce artifacts; and compressed sensing-based methods may be limited by sampling rate and reconstruction accuracy.
[0004] In recent years, deep learning methods have demonstrated significant advantages in various fields such as image recognition, classification, and speech recognition. This approach possesses powerful feature learning capabilities, able to fit complex functional relationships with a large number of training parameters, and exhibits good versatility. In the field of image reconstruction, deep learning techniques have successfully reconstructed high-resolution images and restored rich texture details. In particular, the introduction of techniques such as residual networks, attention mechanisms, and generative adversarial networks has made the reconstructed images more realistic. Nevertheless, deep learning methods also have some limitations, such as a high dependence on the amount of data, the complexity of the training process, and the potential for overfitting. Furthermore, the interpretability of deep learning models is relatively weak, which may be a critical issue in certain application scenarios.
[0005] Therefore, it is necessary to develop a method, device, electronic equipment, and medium for self-supervised seismic data reconstruction in the OVT domain.
[0006] The information disclosed in the background section of this invention is intended only to enhance the understanding of the general background of this invention, and should not be construed as an admission or in any way implying that such information constitutes prior art known to those skilled in the art. Summary of the Invention
[0007] This invention proposes a self-supervised seismic data reconstruction method, device, electronic device, and medium in the OVT domain. It transforms seismic data into the OVT domain as network input, extracts seismic image features in the OVT domain, enhances the network's ability to learn different information from seismic data, and can also reconstruct local details well while retaining the main information, thereby improving the reconstruction quality of seismic data. This is of great significance for high-precision seismic data reconstruction processing.
[0008] In a first aspect, embodiments of this disclosure provide a self-supervised seismic data reconstruction method in the OVT domain, including:
[0009] Using earthquake data with missing data as the original dataset, we established training, validation, and test datasets.
[0010] Convolutional neural networks are constructed, trained using the training dataset, and validated and tested using the validation and test datasets to obtain the final convolutional neural network.
[0011] The seismic data to be reconstructed is input into the final convolutional neural network to obtain the reconstructed OVT domain data.
[0012] As a specific implementation of this disclosure, establishing the training dataset, validation dataset, and test dataset includes:
[0013] Preprocess the original dataset;
[0014] The preprocessed data is transformed into the OVT domain to obtain the OVT gather;
[0015] Fill the missing data channels with zero values and create a sampling matrix R. i Sample the missing data y to obtain new missing data y. i ;
[0016] For y and y i The seismic data is sliced into training dataset, validation dataset, and test dataset.
[0017] Wherein, the sampling matrix R i This is a random sampling matrix or a sampling matrix created based on the missing data characteristics of the OVT domain.
[0018] As a specific implementation of this disclosure, the preprocessing includes amplitude compensation, noise suppression, static correction, and dynamic correction.
[0019] As one specific implementation of this disclosure, the missing data is:
[0020] y = R0x0
[0021] Where R0 is a sampling matrix containing 0 and 1, and x0 represents data without missing values.
[0022] As a specific implementation of this disclosure, the weighted sum of the root mean square error function and the structural similarity error function is used as the loss function of the convolutional neural network.
[0023] As a specific implementation of this disclosure, the seismic data to be reconstructed is input into the final convolutional neural network to obtain the reconstructed OVT domain data, including:
[0024] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into the final convolutional neural network to obtain the reconstructed data and stitch it together to the original size, resulting in the reconstructed OVT domain data.
[0025] Secondly, embodiments of this disclosure also provide an OVT domain self-supervised seismic data reconstruction apparatus, comprising:
[0026] The dataset creation module uses earthquake data with missing data as the original dataset to create training datasets, validation datasets, and test datasets;
[0027] The training module constructs a convolutional neural network, trains the convolutional neural network using the training dataset, and verifies and tests it using a validation dataset and a test dataset to obtain the final convolutional neural network.
[0028] The reconstruction module inputs the seismic data to be reconstructed into the final convolutional neural network to obtain the reconstructed OVT domain data.
[0029] As a specific implementation of this disclosure, establishing the training dataset, validation dataset, and test dataset includes:
[0030] Preprocess the original dataset;
[0031] The preprocessed data is transformed into the OVT domain to obtain the OVT gather;
[0032] Fill the missing data channels with zero values and create a sampling matrix R. i Sample the missing data y to obtain new missing data y. i ;
[0033] For y and y i The seismic data is sliced into training dataset, validation dataset, and test dataset.
[0034] Wherein, the sampling matrix R i This is a random sampling matrix or a sampling matrix created based on the missing data characteristics of the OVT domain.
[0035] As a specific implementation of this disclosure, the preprocessing includes amplitude compensation, noise suppression, static correction, and dynamic correction.
[0036] As one specific implementation of this disclosure, the missing data is:
[0037] y = R0x0
[0038] Where R0 is a sampling matrix containing 0 and 1, and x0 represents data without missing values.
[0039] As a specific implementation of this disclosure, the weighted sum of the root mean square error function and the structural similarity error function is used as the loss function of the convolutional neural network.
[0040] As a specific implementation of this disclosure, the seismic data to be reconstructed is input into the final convolutional neural network to obtain the reconstructed OVT domain data, including:
[0041] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into the final convolutional neural network to obtain the reconstructed data and stitch it together to the original size, resulting in the reconstructed OVT domain data.
[0042] Thirdly, embodiments of this disclosure also provide an electronic device, the electronic device comprising:
[0043] Memory, which stores executable instructions;
[0044] A processor that executes the executable instructions in the memory to implement the OVT domain self-supervised seismic data reconstruction method.
[0045] Fourthly, embodiments of this disclosure also provide a computer-readable storage medium storing a computer program that, when executed by a processor, implements the OVT domain self-supervised seismic data reconstruction method.
[0046] Its beneficial effects are as follows:
[0047] This invention can significantly improve the accuracy and completeness of reconstruction of continuously missing seismic data, providing more reliable data support for seismic exploration and oil and gas reservoir identification.
[0048] The methods and apparatus of the present invention have other features and advantages that will be apparent from or will be set forth in detail in the accompanying drawings and following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description
[0049] The above and other objects, features and advantages of the present invention will become more apparent from the more detailed description of exemplary embodiments of the invention in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same parts.
[0050] Figure 1 A flowchart illustrating the steps of an OVT domain self-supervised seismic data reconstruction method according to an embodiment of the present invention is shown.
[0051] Figure 2 A schematic diagram of OVT domain vector slice data according to an embodiment of the present invention is shown.
[0052] Figure 3 A flowchart illustrating network training and testing according to an embodiment of the present invention is shown.
[0053] Figure 4 A block diagram of an OVT domain self-supervised seismic data reconstruction apparatus according to an embodiment of the present invention is shown.
[0054] Explanation of reference numerals in the attached figures:
[0055] 201. Dataset creation module; 202. Training module; 203. Reconstruction module. Detailed Implementation
[0056] Preferred embodiments of the invention will now be described in more detail. While preferred embodiments of the invention are described below, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein.
[0057] To facilitate understanding of the solutions and effects of the embodiments of the present invention, six specific application examples are given below. Those skilled in the art should understand that these examples are merely for the purpose of understanding the present invention, and any specific details therein are not intended to limit the present invention in any way.
[0058] Example 1
[0059] Figure 1 A flowchart illustrating the steps of an OVT domain self-supervised seismic data reconstruction method according to an embodiment of the present invention is shown.
[0060] like Figure 1 As shown, the OVT domain self-supervised seismic data reconstruction method includes:
[0061] Step 101: Using the earthquake data with missing data as the original dataset, establish the training dataset, validation dataset, and test dataset;
[0062] Step 102: Construct a convolutional neural network. Train the convolutional neural network using the training dataset, and validate and test it using the validation dataset and the test dataset to obtain the final convolutional neural network.
[0063] Step 103: Input the seismic data to be reconstructed into the final convolutional neural network to obtain the reconstructed OVT domain data.
[0064] In one example, creating the training dataset, validation dataset, and test dataset includes:
[0065] Preprocess the original dataset;
[0066] The preprocessed data is transformed into the OVT domain to obtain the OVT gather;
[0067] Fill the missing data channels with zero values and create a sampling matrix R. i Sample the missing data y to obtain new missing data y. i ;
[0068] For y and y i The seismic data is sliced into training, validation, and test datasets.
[0069] Wherein, the sampling matrix R i This is a random sampling matrix or a sampling matrix created based on the missing data characteristics of the OVT domain.
[0070] In one example, preprocessing includes amplitude compensation, noise suppression, static correction, and dynamic correction.
[0071] In one example, the missing data is:
[0072] y = R0x0
[0073] Where R0 is a sampling matrix containing 0 and 1, and x0 represents data without missing values.
[0074] In one example, a weighted sum of the root mean square error function and the structural similarity error function is used as the loss function of the convolutional neural network.
[0075] In one example, the seismic data to be reconstructed is input into the final convolutional neural network to obtain the reconstructed OVT domain data, including:
[0076] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into the final convolutional neural network to obtain the reconstructed data and stitch it together to the original size, resulting in the reconstructed OVT domain data.
[0077] Figure 2 A schematic diagram of OVT domain vector slice data according to an embodiment of the present invention is shown.
[0078] Specifically, training, validation, and test datasets are established: missing seismic data x0 is obtained; the data is preprocessed, including amplitude compensation, noise suppression, static correction, and dynamic correction; the data is transformed to the OVT domain, OVT gathers are obtained, and vector slices are partitioned, such as... Figure 2 As shown; by filling the missing data with zero values and obtaining the sampling matrix of the missing data, the missing data can be represented as:
[0079] y = R0x0
[0080] Where R0 is a sampling matrix containing 0 and 1, x0 represents data without missing data, and y represents data with missing data.
[0081] Create a new random sampling matrix R i Sample y to obtain new missing data y i ; for y and y i The seismic data was sliced into uniformly sized slices, with 80% used as the training set, 10% as the validation set, and 10% as the test set.
[0082] Convolutional neural networks (CNNs) are constructed, including but not limited to UNet, Transformer, Generative Adversarial Networks (GANs), residual modules, and convolutional attention modules. CNNs are widely used in seismic data feature extraction, denoising, and reconstruction. Through symmetrical encoding and decoding structures and multiple upsampling and downsampling, they achieve feature fusion at different scales, preserving more texture details. Residual modules can reduce the vanishing gradient problem that occurs during training in multi-layer CNNs, while also facilitating gradient backpropagation, accelerating training efficiency, and improving network robustness. Convolutional attention modules are mainly used to enhance feature extraction capabilities and focus on local details.
[0083] The loss function is defined as a weighted sum of the root mean square error function and the structural similarity error function. The Adam adaptive learning rate method is used as the optimization algorithm, which can dynamically adjust the learning rate, resulting in fast convergence and good reconstruction performance.
[0084] After preparing the training set and network structure, the network is trained. The network parameters are continuously optimized through loss function and optimization algorithm until the set number of iterations is reached, and finally the trained network is obtained.
[0085] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into a trained network. The reconstructed data is output and then stitched together to the original size to obtain the reconstructed OVT domain data.
[0086] Example 2
[0087] The present invention also provides an OVT domain self-supervised seismic data reconstruction device, comprising:
[0088] The dataset creation module uses earthquake data with missing data as the original dataset to create training datasets, validation datasets, and test datasets;
[0089] The training module constructs a convolutional neural network by training the convolutional neural network using a training dataset and validating and testing it using a validation dataset and a test dataset to obtain the final convolutional neural network.
[0090] The reconstruction module inputs the seismic data to be reconstructed into the final convolutional neural network to obtain the reconstructed OVT domain data.
[0091] In one example, creating the training dataset, validation dataset, and test dataset includes:
[0092] Preprocess the original dataset;
[0093] The preprocessed data is transformed into the OVT domain to obtain the OVT gather;
[0094] Fill the missing data channels with zero values and create a sampling matrix R. i Sample the missing data y to obtain new missing data yi;
[0095] For y and y i The seismic data is sliced into training, validation, and test datasets.
[0096] Wherein, the sampling matrix R i This is a random sampling matrix or a sampling matrix created based on the missing data characteristics of the OVT domain.
[0097] In one example, preprocessing includes amplitude compensation, noise suppression, static correction, and dynamic correction.
[0098] In one example, the missing data is:
[0099] y = R0x0
[0100] Where R0 is a sampling matrix containing 0 and 1, and x0 represents data without missing values.
[0101] In one example, a weighted sum of the root mean square error function and the structural similarity error function is used as the loss function of the convolutional neural network.
[0102] In one example, the seismic data to be reconstructed is input into the final convolutional neural network to obtain the reconstructed OVT domain data, including:
[0103] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into the final convolutional neural network to obtain the reconstructed data and stitch it together to the original size, resulting in the reconstructed OVT domain data.
[0104] Figure 2 A schematic diagram of OVT domain vector slice data according to an embodiment of the present invention is shown.
[0105] Specifically, training, validation, and test datasets are established: missing seismic data x0 is obtained; the data is preprocessed, including amplitude compensation, noise suppression, static correction, and dynamic correction; the data is transformed to the OVT domain, OVT gathers are obtained, and vector slices are partitioned, such as... Figure 2 As shown; by filling the missing data with zero values and obtaining the sampling matrix of the missing data, the missing data can be represented as:
[0106] y = R0x0
[0107] Where R0 is a sampling matrix containing 0 and 1, x0 represents data without missing data, and y represents data with missing data.
[0108] Create a new random sampling matrix R i Sample y to obtain new missing data y i ; for y and y i The seismic data was sliced into uniformly sized slices, with 80% used as the training set, 10% as the validation set, and 10% as the test set.
[0109] Convolutional neural networks (CNNs) are constructed, including but not limited to UNet, Transformer, Generative Adversarial Networks (GANs), residual modules, and convolutional attention modules. CNNs are widely used in seismic data feature extraction, denoising, and reconstruction. Through symmetrical encoding and decoding structures and multiple upsampling and downsampling, they achieve feature fusion at different scales, preserving more texture details. Residual modules can reduce the vanishing gradient problem that occurs during training in multi-layer CNNs, while also facilitating gradient backpropagation, accelerating training efficiency, and improving network robustness. Convolutional attention modules are mainly used to enhance feature extraction capabilities and focus on local details.
[0110] The loss function is defined as a weighted sum of the root mean square error function and the structural similarity error function. The Adam adaptive learning rate method is used as the optimization algorithm, which can dynamically adjust the learning rate, resulting in fast convergence and good reconstruction performance.
[0111] After preparing the training set and network structure, the network is trained. The network parameters are continuously optimized through loss function and optimization algorithm until the set number of iterations is reached, and finally the trained network is obtained.
[0112] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into a trained network. The reconstructed data is output and then stitched together to the original size to obtain the reconstructed OVT domain data.
[0113] Example 3
[0114] Seismic data with missing data from actual seismic exploration projects were selected as the original dataset. Necessary preprocessing was performed on the data, including but not limited to amplitude compensation, noise suppression, static correction, and dynamic correction, to eliminate or reduce interference from non-geological factors and improve data quality.
[0115] Transforming the data from the time-space domain to the OVT domain creates OVT gathers. This step helps classify the data according to specific geometric and dynamic properties, facilitating subsequent deep learning processing.
[0116] Missing data traces are filled with zeros, and a sampling matrix for the missing data is recorded. This sampling matrix will be used in the subsequent self-supervised learning process. To increase data diversity, a new random sampling matrix is created to further sample the original missing data, generating more missing data samples. The processed data and the corresponding sampling matrix are then cut into seismic data slices of uniform size, divided into training, validation, and test sets in an 8:1:1 ratio.
[0117] Neural Network Model Construction. A deep learning model is constructed, whose feature extraction part mainly consists of a Unet network, a residual module, and a convolutional attention module. The Unet network is used for multi-scale feature fusion, the residual module is used to solve the gradient vanishing problem and accelerate training, and the convolutional attention module is used to improve the model's attention to key features. A composite loss function is defined, which is a weighted sum of the root mean square error function and the structural similarity error function. This loss function design aims to simultaneously optimize the numerical accuracy and structural similarity of the reconstructed data. The Adam optimization algorithm is used to train the model. The Adam algorithm can adaptively adjust the learning rate, thereby improving the convergence speed of training and the reconstruction performance of the model.
[0118] Figure 3 A flowchart illustrating network training and testing according to an embodiment of the present invention is shown.
[0119] like Figure 3As shown, the neural network model is trained using the prepared training and validation sets. The model parameters are continuously adjusted using the backpropagation algorithm and optimizer. During training, the model is periodically evaluated using the validation set to prevent overfitting and ensure the model's generalization ability. Training is stopped and the optimal model is saved when the model's performance on the validation set reaches a preset standard or the maximum number of iterations is reached.
[0120] The seismic data to be reconstructed is processed to generate input matrices of a specified size. These input matrices are then fed into a trained model to obtain the reconstructed data output. The output data blocks are then concatenated back to the original data size to form complete reconstructed OVT domain data. Necessary post-processing operations, such as inverse transformation back to the time-space domain, are performed on the reconstructed data to facilitate subsequent seismic interpretation.
[0121] Example 4
[0122] Figure 4 A block diagram of an OVT domain self-supervised seismic data reconstruction apparatus according to an embodiment of the present invention is shown.
[0123] like Figure 4 As shown, the OVT domain self-supervised seismic data reconstruction device includes:
[0124] The dataset creation module 201 uses the earthquake data with missing data as the original dataset to create the training dataset, validation dataset, and test dataset.
[0125] Training module 202 constructs a convolutional neural network by training the convolutional neural network with a training dataset and validating and testing it with a validation dataset and a test dataset to obtain the final convolutional neural network.
[0126] The reconstruction module 203 inputs the seismic data to be reconstructed into the final convolutional neural network to obtain the reconstructed OVT domain data.
[0127] In one example, creating the training dataset, validation dataset, and test dataset includes:
[0128] Preprocess the original dataset;
[0129] The preprocessed data is transformed into the OVT domain to obtain the OVT gather;
[0130] Fill the missing data channels with zero values and create a sampling matrix R. i Sample the missing data y to obtain new missing data y. i ;
[0131] For y and y iThe seismic data is sliced into training, validation, and test datasets.
[0132] Wherein, the sampling matrix R i This is a random sampling matrix or a sampling matrix created based on the missing data characteristics of the OVT domain.
[0133] In one example, preprocessing includes amplitude compensation, noise suppression, static correction, and dynamic correction.
[0134] In one example, the missing data is:
[0135] y = R0x0
[0136] Where R0 is a sampling matrix containing 0 and 1, and x0 represents data without missing values.
[0137] In one example, a weighted sum of the root mean square error function and the structural similarity error function is used as the loss function of the convolutional neural network.
[0138] In one example, the seismic data to be reconstructed is input into the final convolutional neural network to obtain the reconstructed OVT domain data, including:
[0139] The seismic data to be reconstructed is used to generate a matrix of a specified size, which is then input into the final convolutional neural network to obtain the reconstructed data and stitch it together to the original size, resulting in the reconstructed OVT domain data.
[0140] Example 5
[0141] This disclosure provides an electronic device comprising: a memory storing executable instructions; and a processor executing the executable instructions in the memory to implement the aforementioned OVT domain self-supervised seismic data reconstruction method.
[0142] An electronic device according to an embodiment of the present disclosure includes a memory and a processor.
[0143] This memory is used to store non-transitory computer-readable instructions. Specifically, the memory may include one or more computer program products, which may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may, for example, include random access memory (RAM) and / or cache memory. The non-volatile memory may, for example, include read-only memory (ROM), hard disk, flash memory, etc.
[0144] The processor may be a central processing unit (CPU) or other form of processing unit with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In one embodiment of this disclosure, the processor is used to execute computer-readable instructions stored in the memory.
[0145] Those skilled in the art will understand that, in order to solve the technical problem of how to achieve a good user experience, this embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of this disclosure.
[0146] For a detailed description of this embodiment, please refer to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.
[0147] Example 6
[0148] This disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the OVT domain self-supervised seismic data reconstruction method.
[0149] A computer-readable storage medium according to embodiments of the present disclosure stores non-transitory computer-readable instructions. When these non-transitory computer-readable instructions are executed by a processor, all or part of the steps of the methods described in the foregoing embodiments of the present disclosure are performed.
[0150] The aforementioned computer-readable storage media include, but are not limited to: optical storage media (e.g., CD-ROM and DVD), magneto-optical storage media (e.g., MO), magnetic storage media (e.g., magnetic tape or portable hard drive), media with built-in rewritable non-volatile memory (e.g., memory card), and media with built-in ROM (e.g., ROM cartridge).
[0151] Those skilled in the art should understand that the above description of the embodiments of the present invention is only intended to illustrate the beneficial effects of the embodiments of the present invention, and is not intended to limit the embodiments of the present invention to any of the examples given.
[0152] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.
Claims
1. An OVT domain self-supervised seismic data reconstruction method, characterized in that, The method comprises the following steps: Seismic data with missing data is taken as an original data set to establish a training data set, a verification data set and a test data set; A convolutional neural network is constructed, the convolutional neural network is trained by using the training data set, and the convolutional neural network is verified and tested by using the verification data set and the test data set to obtain a final convolutional neural network; Seismic data to be reconstructed is input into the final convolutional neural network to obtain reconstructed OVT domain data.
2. The OVT domain self-supervised seismic data reconstruction method of claim 1, wherein, The method comprises the following steps: The original data set is preprocessed; The preprocessed data is transformed into an OVT domain to obtain an OVT gather; Fill the missing data channel with zero value, create a sampling matrix R i Sample the missing data y to obtain new missing data yi; for y and y i slicing, establishing seismic data slices, dividing into the training dataset, the validation dataset and the test dataset; wherein the sampling matrix R i is a random sampling matrix or a sampling matrix made according to the missing feature of the OVT domain data.
3. The OVT domain self-supervised seismic data reconstruction method of claim 2, wherein, The preprocessing comprises amplitude compensation, noise suppression, static correction and dynamic correction.
4. The OVT domain self-supervised seismic data reconstruction method of claim 2, wherein, The missing data is y=R0x0, wherein R0 is a sampling matrix containing 0 and 1, and x0 represents non-missing data. A weighted sum of a root mean square error function and a structural similarity error function is taken as a loss function of the convolutional neural network. The method comprises the following steps:
5. The OVT domain self-supervised seismic data reconstruction method of claim 1, wherein, The seismic data to be reconstructed is generated into a matrix with a specified size, input into the final convolutional neural network, and spliced into the original size to obtain the reconstructed OVT domain data.
6. The OVT domain self-supervised seismic data reconstruction method of claim 1, wherein, The method comprises the following steps: A data set establishment module takes seismic data with missing data as an original data set to establish a training data set, a verification data set and a test data set; 7. An apparatus for OVT domain self-supervised seismic data reconstruction, characterized in that, A training module constructs a convolutional neural network, trains the convolutional neural network by using the training data set, and verifies and tests the convolutional neural network by using the verification data set and the test data set to obtain a final convolutional neural network; A reconstruction module inputs seismic data to be reconstructed into the final convolutional neural network to obtain reconstructed OVT domain data. The method comprises the following steps: The seismic data to be reconstructed is generated into a matrix with a specified size, input into the final convolutional neural network, and spliced into the original size to obtain the reconstructed OVT domain data.
8. The OVT domain self-supervised seismic data reconstruction apparatus of claim 7, wherein, The electronic device comprises: A memory storing executable instructions; 9. An electronic device, comprising: A processor running the executable instructions in the memory to implement the OVT domain self-supervised seismic data reconstruction method in any one of claims 1-6. The computer readable storage medium stores a computer program, and the computer program is executed by a processor to implement the OVT domain self-supervised seismic data reconstruction method in any one of claims 1-6. 10. A computer-readable storage medium, characterized in that,