Seismic data deconvolution method and device

Abstract

The invention provides a seismic data deconvolution method and device. The method comprises the following steps: acquiring to-be-measured seismic data and historical seismic data; establishing a seismic sample database based on the historical seismic data; acquiring a seismic data training sample from a seismic sample database; the seismic data sample comprises common shot gather data and a corresponding sample category label; based on the common shot set data and the corresponding sample category labels, constructing a target consistency deconvolution network; and inputting the seismic data to be measured into the consistency deconvolution network to obtain a corresponding prediction category. And establishing an intelligent earth surface consistency deconvolution deep neural network in which a self-attention mechanism is introduced, realizing extraction of earth surface-oriented deconvolution operators based on the deep neural network, and realizing pre-stack data seismic consistency deconvolution processing.

Description

Technical Field

[0001] The present invention relates to the field of geophysical exploration technology, and in particular to a seismic data deconvolution method and device. Background Art

[0002] Deconvolution is a particularly important link in the process of seismic data processing, and it plays a connecting role in the whole process. Deconvolution mainly improves the vertical resolution by compressing the basic seismic wavelets in the seismic records, suppressing reverberation and short-period multiple waves.

[0003] Although at present, deconvolution technology has achieved certain progress at home and abroad, and with the continuous development of artificial intelligence in recent years, seismic data processing technology based on deep learning has also achieved certain application results, but the method of achieving consistent processing of seismic data based on deep neural networks to better serve the consistent deconvolution process of seismic data preprocessing has not yet been proposed. Summary of the invention

[0004] The present invention provides a deconvolution method and device for seismic data, which realizes the extraction of surface-oriented deconvolution operators based on a deep neural network, improves network accuracy, and realizes the processing of seismic consistency deconvolution of pre-stack data.

[0005] In a first aspect, the present invention provides a deconvolution method for seismic data, comprising:

[0006] Obtain the earthquake data to be measured and historical earthquake data;

[0007] Establishing an earthquake sample database based on the historical earthquake data;

[0008] Acquire seismic data training samples from a seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0009] Based on the common shot gather data and the corresponding sample category labels, constructing a target consistency deconvolution network;

[0010] The earthquake data to be measured is input into the consistency deconvolution network to obtain the corresponding prediction category.

[0011] Optionally, the sample category labels include: common detection domain data and common offset domain data; establishing a seismic sample database based on the historical seismic data includes:

[0012] Sorting the historical seismic data to obtain two data types including common detection domain data and common offset domain data;

[0013] According to the data type of the historical earthquake data, marking corresponding data category labels;

[0014] The earthquake sample database is constructed based on the historical earthquake data and the corresponding data category labels.

[0015] Optionally, based on the common shot gather data and corresponding sample category labels, a target consistency deconvolution network is constructed, including:

[0016] Input all the common shot collections of the common shot collection data into a consistent deconvolution network to generate corresponding sample categories;

[0017] According to the sample category label and the sample category, a training error is determined, and based on the training error, the consistent deconvolution network is adjusted to obtain optimal network parameters, and the target consistent deconvolution network is generated using the optimal network parameters.

[0018] Optionally, before inputting all the common shot gathers of the common shot gather data into a consistent deconvolution network to generate corresponding sample categories, the method further includes:

[0019] The common shot gather data is cut or resampled in blocks according to a preset size.

[0020] In a second aspect, the present invention provides a deconvolution device for seismic data, comprising:

[0021] A data acquisition module is used to acquire the earthquake data to be measured and the historical earthquake data;

[0022] A database establishment module, used for establishing an earthquake sample database based on the historical earthquake data;

[0023] A sample acquisition module is used to acquire seismic data training samples from a seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0024] A construction module, used to construct a target consistency deconvolution network based on the common shot gather data and the corresponding sample category labels;

[0025] The prediction module is used to input the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

[0026] Optionally, the sample category label includes: common detection domain data and common offset domain data; the database establishment module includes:

[0027] A sorting submodule, used for sorting the historical seismic data into two data types including common detection domain data and common offset domain data;

[0028] A labeling submodule, used for labeling corresponding data category labels according to the data type of the historical earthquake data;

[0029] The construction submodule is used to construct the earthquake sample database based on the historical earthquake data and the corresponding data category labels.

[0030] Optionally, the building blocks include:

[0031] An input submodule, used for inputting all the common shot collections of the common shot collection data into a consistent deconvolution network to generate corresponding sample categories;

[0032] The network generation submodule is used to determine the training error according to the sample category label and the sample category, and based on the training error, adjust the consistency deconvolution network to obtain optimal network parameters, and use the optimal network parameters to generate a target consistency deconvolution network.

[0033] Optionally, the building block further includes:

[0034] The preprocessing submodule is used to cut or resample the common shot gather data into blocks according to a preset size.

[0035] In a third aspect, the present application provides an electronic device, comprising a processor and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the steps in the method provided in the first aspect are performed.

[0036] In a fourth aspect, the present application provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps in the method provided in the first aspect are performed.

[0037] It can be seen from the above technical solutions that the present invention has the following advantages:

[0038] The present invention provides a method and device for deconvolution of seismic data, the method comprising: obtaining seismic data to be tested and historical seismic data; establishing a seismic sample database based on the historical seismic data; obtaining seismic data training samples from the seismic sample database; the seismic data samples comprising: common shot data and corresponding sample category labels; constructing a target consistency deconvolution network based on the common shot data and the corresponding sample category labels; inputting the seismic data to be tested into the consistency deconvolution network to obtain the corresponding prediction category. An intelligent surface consistency deconvolution deep neural network with a self-attention mechanism is established, and the extraction of surface-oriented deconvolution operators is realized based on the deep neural network, so as to realize the processing of pre-stack data seismic consistency deconvolution. BRIEF DESCRIPTION OF THE DRAWINGS

[0039] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0040] Figure 1 A flowchart of a first embodiment of a method for deconvolution of seismic data according to the present invention;

[0041] Figure 2 A flowchart of a second embodiment of a method for deconvolution of seismic data according to the present invention;

[0042] Figure 3 A flowchart of a third embodiment of a method for deconvolution of seismic data according to the present invention;

[0043] Figure 4 A flowchart of a fourth embodiment of a method for deconvolution of seismic data according to the present invention;

[0044] Figure 5 A network structure diagram of a seismic data deconvolution method embodiment 4 of the present invention that introduces a self-attention mechanism;

[0045] Figure 6 The data before processing after introducing the self-attention mechanism in Embodiment 4 of the deconvolution method for seismic data of the present invention;

[0046] Figure 7 This is a processing effect diagram of commercial software introducing a self-attention mechanism in Embodiment 4 of a deconvolution method for seismic data of the present invention;

[0047] Figure 8 This is a diagram showing the denoising effect after introducing a self-attention mechanism according to Embodiment 4 of a deconvolution method for seismic data of the present invention;

[0048] Fig. 9 The present invention is a structural block diagram of a deconvolution method and device embodiment of seismic data. DETAILED DESCRIPTION

[0049] The embodiment of the present invention provides a deconvolution method and device for seismic data, which realizes the extraction of surface-oriented deconvolution operators based on a deep neural network, improves network accuracy, and realizes the processing of seismic consistency deconvolution of pre-stack data.

[0050] In order to make the purpose, features and advantages of the present invention more obvious and easy to understand, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the embodiments described below are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0051] For example, see Figure 1 , Figure 1 The flowchart of the first embodiment of the VTI medium reverse time migration imaging method for geological targets of the present invention includes:

[0052] Step S101, obtaining earthquake data to be measured and historical earthquake data;

[0053] Step S102, establishing an earthquake sample database based on the historical earthquake data;

[0054] Step S103, obtaining seismic data training samples from a seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0055] Step S104, constructing a target consistency deconvolution network based on the common shot gather data and the corresponding sample category labels;

[0056] Step S105, inputting the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

[0057] The embodiment of the present invention discloses a deconvolution method for seismic data, including: obtaining seismic data to be tested and historical seismic data; establishing a seismic sample database based on the historical seismic data; obtaining seismic data training samples from the seismic sample database; the seismic data samples include: common shot data and corresponding sample category labels; constructing a target consistency deconvolution network based on the common shot data and the corresponding sample category labels; inputting the seismic data to be tested into the consistency deconvolution network to obtain the corresponding prediction category. An intelligent surface consistency deconvolution deep neural network with a self-attention mechanism is established, and the extraction of surface-oriented deconvolution operators is realized based on the deep neural network, so as to realize the processing of pre-stack data seismic consistency deconvolution.

[0058] For example 2, please refer to Figure 2 , Figure 2 This is a flow chart of a second embodiment of a method for deconvolution of seismic data of the present invention, the method comprising:

[0059] Step S201, obtaining earthquake data to be measured and historical earthquake data;

[0060] Step S202, sorting the historical seismic data to obtain two data types including common detection domain data and common offset domain data;

[0061] Step S203, marking corresponding data category labels according to the data type of the historical earthquake data;

[0062] Step S204, constructing the earthquake sample database according to the historical earthquake data and the corresponding data category labels;

[0063] Step S205, obtaining seismic data training samples from a seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0064] Step S206, constructing a target consistency deconvolution network based on the common shot gather data and the corresponding sample category labels;

[0065] Step S207, inputting the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

[0066] The embodiment of the present invention discloses a deconvolution method for seismic data. The method sorts historical seismic data to obtain two data types including common detection domain data and common offset domain data for labeling. The historical seismic data is used as a training sample to ensure the authenticity and accuracy of the sample.

[0067] For example 3, please refer to Figure 3 , Figure 3 The present invention is a flowchart of a third embodiment of a method for deconvolution of seismic data, the method comprising:

[0068] Step S301, obtaining earthquake data to be measured and historical earthquake data;

[0069] Step S302, establishing an earthquake sample database based on the historical earthquake data;

[0070] Step S303, obtaining seismic data training samples from a seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0071] Step S304, inputting all the common shot gathers of the common shot gather data into a consistent deconvolution network to generate corresponding sample categories;

[0072] Step S305, determining a training error according to the sample category label and the sample category, and adjusting the consistent deconvolution network based on the training error to obtain optimal network parameters, and using the optimal network parameters to generate a target consistent deconvolution network;

[0073] Step S306: input the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

[0074] The embodiment of the present invention discloses a deconvolution method for seismic data, which generates corresponding sample categories by inputting all common shot gathers of common shot gather data into a consistent deconvolution network, thereby determining a training error, and then obtaining a target consistent deconvolution network based on the training error.

[0075] Example 4, please refer to Figure 4 , Figure 4 This is a flowchart of a fourth embodiment of a method for deconvolution of seismic data of the present invention, the method comprising:

[0076] Step S401, obtaining earthquake data to be measured and historical earthquake data;

[0077] Step S402, sorting the historical seismic data to obtain two data types including common detection domain data and common offset domain data;

[0078] Step S403, marking corresponding data category labels according to the data type of the historical earthquake data;

[0079] Step S404, constructing the earthquake sample database according to the historical earthquake data and the corresponding data category labels;

[0080] Step S405, obtaining seismic data training samples from a seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0081] In an embodiment of the present invention, consistent deconvolution processing of historical seismic data is carried out, the effect of consistent processing of common shot gather data with a greater demand for consistent deconvolution in an actual work area is evaluated, the historical seismic data is screened, and the data with good processing effect is selected, and then the selected common shot gather data is sorted to obtain corresponding common detection domain data and common offset domain data as sample data for constructing a seismic sample database.

[0082] Step S406, cutting or resampling the common shot gather data into blocks according to a preset size;

[0083] In an embodiment of the present invention, the selected seismic data samples are subjected to data block cropping, and each input common shot set data and the corresponding sample category label are cropped or resampled according to the size of 256*256 to perform data enhancement, thereby improving sample diversity and enhancing the generalization ability of the network.

[0084] Step S407, inputting all the common shot gathers of the common shot gather data into a consistent deconvolution network to generate corresponding sample categories;

[0085] The network structure adopted in the embodiment of the present invention is as follows: Figure 5 As shown in the network structure diagram of the fourth embodiment of the deconvolution method of seismic data of the present invention, the self-attention mechanism is introduced. In actual training, the network input data can be the common shot data of multiple work areas, but the common shot data of the same work area contains the corresponding shot domain, detection domain and offset domain. The output of the network is the result obtained after the corresponding gather consistency deconvolution processing. The network as a whole is a UNET network structure of a full convolutional neural network, which is divided into three parts: downsampling, upsampling and self-attention network blocks. The basic structural unit in the downsampling path is two convolutional layers connected to a maximum pooling layer, with a total of three groups. The number of channels of the convolutional layer is the same, the convolution kernel size is 3x3, the maximum pooling layer is 2x2, the step size is 2x2, the number of channels of the convolutional layer is increased from the original 64 to 512, and the number of channels in each group is twice that of the previous group. The image is downsampled under the action of the pooling layer, and the number of horizontal and vertical sampling points of the image becomes half of the previous group. In the upsampling path, the deconvolution algorithm is used for image upsampling, and the cropping and copying methods are further used to splice with the high-resolution image in the left downsampling to ensure the richness and resolution of the information in the upsampling process. The number of channels in the upsampling convolution process is continuously reduced, and finally a 1x1 convolution is performed to obtain the final output result. In addition, the self-attention mechanism network block introduced is mainly composed of three parallel Q, K, and V convolution kernels with 1x1 convolution layers, where the number of channels of Q and K is 128, and the number of channels of V is 1024. The tensor after the two convolution layers Q and K is obtained by inner product to obtain the correlation, and then the tensor is activated by the softmax function and the result after parallel convolution of V is inner product, and then the weighted sum is input into the initial input of the self-attention network block as the final output to be connected to the upsampling part of the entire network.

[0086] In the specific implementation, the designed consistent deconvolution network is a single-channel input and single-channel output structure, and the size of the input and output data is consistent. In order to ensure that the relative relationship between the energy and amplitude of the spatiotemporal data before and after processing remains unchanged, the loss function used is the L2 norm, and the Adam optimizer is used for iterative updates.

[0087] Step S408, determining a training error according to the sample category label and the sample category, and adjusting the consistent deconvolution network based on the training error to obtain optimal network parameters, and using the optimal network parameters to generate a target consistent deconvolution network;

[0088] In the embodiment of the present invention, the result after consistency processing during the training process is used as a label, and the data of different domains is used as input data to input the consistency deconvolution network for training, and finally the trained consistency deconvolution network is obtained, and the trained network is saved;

[0089] Then reload the saved network, perform surface consistency processing on the test data, and verify the effect of the trained consistent deconvolution network. Since the network of this method is a fully convolutional network, the input data does not need to be cropped during the inference test phase, and the original data size can be directly input for testing. The trained consistent deconvolution network that passes the test is defined as the target consistent deconvolution network.

[0090] In the specific implementation, the input data size of this training is 256x256, and a total of 90,000 samples are prepared. The data sample label data is prepared by TFRecord. The above network is established, and the learning rate is set to 0.0001, batch_size is set to 10, and the number of learning rounds is set to 50. In addition, the cluster GPU used in this test is RTX 2080Ti, the video memory size is 11G, and the number of GPU cards used for training is four cards. The use of data parallelism for training improves training efficiency.

[0091] In the embodiment of the present invention, network training is performed based on the above parameters, and the consistent deconvolution network is saved. During the test, the data size is 266*2001, and the input pre-processing data is as follows Figure 6 The data before processing by introducing the self-attention mechanism in the fourth embodiment of the deconvolution method for seismic data of the present invention is shown in the processing effect based on commercial software. Figure 7 The processing effect of commercial software introducing the self-attention mechanism in the fourth embodiment of the deconvolution method of seismic data of the present invention is shown in the figure. The data after denoising by the technical method is as follows Figure 8 The denoising effect after introducing the self-attention mechanism of the fourth embodiment of the deconvolution method of seismic data of the present invention is shown in the figure. It can be clearly seen from the figure that both the commercial software and the method of the present invention have good effects in consistency processing, and the sub-wave consistency of the processed results is significantly enhanced. The processing effect of the present invention is equivalent to that of the commercial software. After comparison, the comprehensive processing efficiency based on the present method is 3 times that of the commercial software. It can be concluded that the processing effect based on the present invention is equivalent to that of the commercial software, but the processing efficiency is greatly improved.

[0092] Step S409: input the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

[0093] The embodiment of the present invention discloses a deconvolution method for seismic data. By establishing a sample label set of multiple data domains and introducing a self-attention mechanism to fully explore the influencing factors of different data domains on surface consistent deconvolution processing, the extraction of surface-oriented deconvolution operators is realized based on a deep neural network, the network accuracy is improved, and the processing of pre-stack data seismic consistent deconvolution is realized.

[0094] For example 5, please refer to Fig. 9 , Fig. 9 The structure block diagram of an embodiment of a deconvolution device for seismic data of the present invention comprises:

[0095] The data acquisition module 501 is used to acquire the earthquake data to be measured and the historical earthquake data;

[0096] A database establishment module 502 is used to establish an earthquake sample database based on the historical earthquake data;

[0097] The sample acquisition module 503 is used to acquire seismic data training samples from the seismic sample database; the seismic data samples include: common shot gather data and corresponding sample category labels;

[0098] A construction module 504 is used to construct a target consistency deconvolution network based on the common shot gather data and the corresponding sample category labels;

[0099] The prediction module 505 is used to input the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

[0100] In an optional embodiment, the sample category label includes: common detection domain data and common offset domain data; the database establishment module 502 includes:

[0101] A sorting submodule, used for sorting the historical seismic data into two data types including common detection domain data and common offset domain data;

[0102] A labeling submodule, used for labeling corresponding data category labels according to the data type of the historical earthquake data;

[0103] The construction submodule is used to construct the earthquake sample database based on the historical earthquake data and the corresponding data category labels.

[0104] In an optional embodiment, the building module 504 includes:

[0105] An input submodule, used for inputting all the common shot collections of the common shot collection data into a consistent deconvolution network to generate corresponding sample categories;

[0106] The network generation submodule is used to determine the training error according to the sample category label and the sample category, and based on the training error, adjust the consistency deconvolution network to obtain optimal network parameters, and use the optimal network parameters to generate a target consistency deconvolution network.

[0107] In an optional embodiment, the construction module 504 further includes:

[0108] The preprocessing submodule is used to cut or resample the common shot gather data into blocks according to a preset size.

[0109] An embodiment of the present invention further provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor executes the steps of a deconvolution method for seismic data as described in any of the above embodiments.

[0110] An embodiment of the present invention further provides a computer storage medium on which a computer program is stored. When the computer program is executed by the processor, the steps of a seismic data deconvolution method as described in any of the above embodiments are implemented.

[0111] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0112] In the several embodiments provided in the present application, it should be understood that the methods, devices, electronic devices and storage media disclosed in the present invention can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as 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 mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0113] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0114] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0115] If the integrated unit 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 is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a readable storage medium, including several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the method described in each embodiment of the present invention. The aforementioned readable storage medium includes: U disk, mobile hard disk, read-only memory (ROM, Read-Only Memory), random access memory (RAM, Random Access Memory), disk or optical disk and other media that can store program codes.

[0116] As described above, the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit the same. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that the technical solutions described in the aforementioned embodiments may still be modified, or some of the technical features thereof may be replaced by equivalents. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A deconvolution method for seismic data, characterized in that: include: Obtain the earthquake data to be measured and historical earthquake data; Establishing an earthquake sample database based on the historical earthquake data; Obtain earthquake data training samples from an earthquake sample database; The seismic data samples include: total shot gather data and corresponding sample category labels; Based on the common shot gather data and the corresponding sample category labels, constructing a target consistency deconvolution network; The earthquake data to be measured is input into the consistency deconvolution network to obtain the corresponding prediction category.

2. The deconvolution method of seismic data according to claim 1, characterized in that: The sample category labels include: common detection domain data and common offset domain data; establishing a seismic sample database based on the historical seismic data includes: Sorting the historical seismic data to obtain two data types including common detection domain data and common offset domain data; According to the data type of the historical earthquake data, marking corresponding data category labels; The earthquake sample database is constructed based on the historical earthquake data and the corresponding data category labels.

3. The deconvolution method of seismic data according to claim 1, characterized in that: Based on the common shot data and the corresponding sample category labels, a target consistency deconvolution network is constructed, including: Input all the common shot collections of the common shot collection data into a consistent deconvolution network to generate corresponding sample categories; According to the sample category label and the sample category, a training error is determined, and based on the training error, the consistent deconvolution network is adjusted to obtain optimal network parameters, and the target consistent deconvolution network is generated using the optimal network parameters.

4. The deconvolution method of seismic data according to claim 3, characterized in that: Before inputting all the common shot collections of the common shot collection data into a consistent deconvolution network to generate corresponding sample categories, the method further includes: The common shot gather data is cut or resampled in blocks according to a preset size.

5. A deconvolution device for seismic data, characterized in that: include: A data acquisition module is used to acquire the earthquake data to be measured and the historical earthquake data; A database establishment module, used for establishing an earthquake sample database based on the historical earthquake data; A sample acquisition module is used to acquire seismic data training samples from a seismic sample database; The seismic data samples include: total shot gather data and corresponding sample category labels; A construction module, used to construct a target consistency deconvolution network based on the common shot gather data and the corresponding sample category labels; The prediction module is used to input the earthquake data to be measured into the consistency deconvolution network to obtain the corresponding prediction category.

6. The deconvolution device for seismic data according to claim 5, characterized in that: The sample category labels include: common detection domain data and common offset domain data; the database establishment module includes: A sorting submodule, used for sorting the historical seismic data into two data types including common detection domain data and common offset domain data; A labeling submodule, used for labeling corresponding data category labels according to the data type of the historical earthquake data; The construction submodule is used to construct the earthquake sample database based on the historical earthquake data and the corresponding data category labels.

7. The deconvolution device for seismic data according to claim 5, characterized in that: The building blocks include: An input submodule, used for inputting all the common shot collections of the common shot collection data into a consistent deconvolution network to generate corresponding sample categories; The network generation submodule is used to determine the training error according to the sample category label and the sample category, and based on the training error, adjust the consistency deconvolution network to obtain optimal network parameters, and use the optimal network parameters to generate a target consistency deconvolution network.

8. The deconvolution device for seismic data according to claim 7, characterized in that: The building blocks also include: The preprocessing submodule is used to cut or resample the common shot gather data into blocks according to a preset size.

9. An electronic device, characterized in that: The method comprises a processor and a memory, wherein the memory stores computer-readable instructions, and when the computer-readable instructions are executed by the processor, the method according to any one of claims 1 to 4 is executed.

10. A storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 4 is executed.

Images