Seismic data frequency expanding method and device based on deep learning

Abstract

The invention discloses a seismic data frequency broadening method and device based on deep learning, and the method comprises the steps: carrying out the spectral bluing processing based on the obtained seismic data of a research region, and obtaining a spectral bluing data body; and inputting the spectral bluing data volume into a seismic data frequency-broadening model to obtain seismic data after frequency broadening. According to the seismic data after frequency expansion, the signal-to-noise ratio, the amplitude and the time-frequency characteristic are maintained, the frequency band of a seismic signal is effectively expanded, the resolution capability of the seismic signal is improved, the event reflection characteristic of a seismic section is clearer, interlayer information is richer, thin layer recognition is finer, the structure and reservoir characteristics in a research area can be better implemented, and the research efficiency is improved. The resolution of seismic data can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of geological exploration, and in particular to a method and device for seismic data frequency expansion based on deep learning. Background Art

[0002] In seismic exploration, the resolution of seismic data is crucial for geological interpretation and reservoir prediction. However, because seismic wave propagation underground is affected by various factors, such as stratum absorption, scattering, and multipath propagation, the resolution of seismic data is often limited. Therefore, improving the resolution of seismic data has long been a key research topic in seismic data processing. Existing techniques often use frequency spreading to expand the frequency range of seismic data, thereby obtaining richer geological information and improving the resolution of seismic data. Summary of the Invention

[0003] The present invention provides a method and device for seismic data frequency extension based on deep learning, aiming to apply deep learning to the frequency extension processing of seismic data to improve the resolution of seismic data.

[0004] In a first aspect, the present invention provides a seismic data frequency extension method based on deep learning, comprising:

[0005] Based on the seismic data of the study area, spectrum blueing processing is performed to obtain spectrum blueing data volume;

[0006] The spectrum blued data volume is input into a seismic data extension model to obtain extended seismic data; wherein the seismic data extension model is trained by the following method:

[0007] Performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples;

[0008] Using the spectral blueing data volume sample, training a pre-built deep learning algorithm model to obtain a trained model;

[0009] Evaluating the trained model based on a plurality of preset evaluation parameters, and optimizing the trained model according to the obtained evaluation results to obtain an optimized model;

[0010] The process of training, evaluating and optimizing the optimization model is repeated until the optimization model meets the preset evaluation conditions, thereby obtaining the seismic data spectrum extension model.

[0011] In one or some optional implementations of the embodiment of the present application, performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples includes:

[0012] Based on the training data samples, extracting the training data samples of different frequency bands by using a frequency division technology to obtain the frequency division data volume;

[0013] performing a spectrum blueing process on the frequency-divided data volume to obtain a spectrum blueing operator for the frequency-divided data volume, and calculating a weighting factor based on energy signals of each frequency band corresponding to the frequency-divided data volume;

[0014] Obtaining the optimization operator according to the spectrum blueing operator and the weighting factor;

[0015] Resampling the frequency-divided data volume to obtain reflection coefficients corresponding to each frequency band;

[0016] The spectral blueing data volume is obtained based on the reflection coefficient corresponding to each frequency band and the optimization operator.

[0017] In one or some optional implementations of the embodiment of the present application, before performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples, the method further includes:

[0018] The training data samples are preprocessed and denoised by constructing a guided filtering method until the training data samples meet the set resolution requirements.

[0019] In one or some optional implementations of the embodiment of the present application, after obtaining the seismic data after frequency spreading, the method further includes:

[0020] The spatial visualization technology is used to display the seismic data after frequency spreading, and a comparative analysis is performed with the seismic data of the study area before processing.

[0021] In one or some optional implementations of the embodiments of the present application, the evaluation parameters include accuracy, precision and recall, F1 score, AUC and confusion matrix;

[0022] The trained model is evaluated based on a plurality of preset evaluation parameters in the following manner:

[0023] The trained model is evaluated using accuracy, precision and recall, F1 score, AUC and confusion matrix to obtain corresponding evaluation results.

[0024] In a second aspect, the present invention provides a method for training a seismic data spectrum extension model, comprising:

[0025] Performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples;

[0026] Using the spectral blueing data volume sample, training a pre-built deep learning algorithm model to obtain a trained model;

[0027] Evaluating the trained model based on a plurality of preset evaluation parameters, and optimizing the trained model according to the obtained evaluation results to obtain an optimized model;

[0028] The process of training, evaluating and optimizing the optimization model is repeated until the optimization model meets the preset evaluation conditions, thereby obtaining the seismic data spectrum extension model.

[0029] In a third aspect, the present invention provides a device for seismic data frequency spreading, comprising:

[0030] A data analysis module is used to perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples;

[0031] A first training module is configured to train a pre-built deep learning algorithm model using the spectral blued data volume sample to obtain a trained model;

[0032] An evaluation and optimization module is used to evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model;

[0033] A second training module is used to repeat the process of training, evaluating and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model;

[0034] A data acquisition module is used to perform spectrum blueing processing based on the acquired seismic data of the study area to obtain a spectrum blueing data volume;

[0035] The data processing module is used to input the spectrum blued data volume into the seismic data frequency extension model to obtain the seismic data after frequency extension.

[0036] In a fourth aspect, the present invention provides a training device for a seismic data spectrum extension model, comprising:

[0037] A data analysis module is used to perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples;

[0038] A first training module is configured to train a pre-built deep learning algorithm model using the spectral blued data volume sample to obtain a trained model;

[0039] An evaluation and optimization module is used to evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model;

[0040] The second training module is used to repeat the process of training, evaluating and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model.

[0041] In a fifth aspect, the present invention provides a computer-readable storage medium having a computer program stored thereon, characterized in that when the program is executed by a processor, it implements the deep learning-based seismic data frequency extension method as described in the first aspect of claim 1, and / or the training method of the seismic data frequency extension model as described in the second aspect.

[0042] In a sixth aspect, the present invention provides an electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory communicate with each other via the communication bus;

[0043] Memory for storing computer programs;

[0044] The processor is used to implement the deep learning-based seismic data frequency extension method as described in the first aspect of claim 1, and / or the seismic data frequency extension model training method as described in the second aspect when executing the program stored in the memory.

[0045] The present invention provides a deep learning-based seismic data topology method. By acquiring seismic data from a study area and performing spectrum blueing processing to obtain a spectrum blued data volume, the data volume is input into a seismic data topology model. This method can achieve spectrum topology of the seismic data and obtain the topology-derived seismic data. Since spectrum blueing is used to perform spectrum topology processing on training data samples during the construction of the seismic data topology model, and balanced calculations are performed on all frequency bands of the training data samples, compared with the existing topology processing methods, the topology resolution is improved, the problem of low-frequency jitter is eliminated, and the spectrum distribution anomalies are avoided. The topology results are superior and can retain some low-frequency information related to reservoir prediction in the original training data samples. This is an important technical solution for establishing a seismic data topology model that can effectively implement spectrum topology. Therefore, while maintaining the signal-to-noise ratio, amplitude, and time-frequency characteristics, the topology data can effectively broaden the seismic signal frequency band and improve the resolution of the seismic signal. This makes the seismic profile event reflection characteristics clearer, the interlayer information richer, and the thin layer identification more precise. This can better identify the structural and reservoir characteristics within the study area and effectively improve the resolution of the seismic data. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] In order to more clearly illustrate the technical solutions of the embodiments of the present invention or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the prior art description. The drawings described herein are used to provide a further understanding of the present invention, constitute part of this application, and do not constitute a limitation of the present invention. In the drawings:

[0047] Figure 1 A flowchart of a deep learning-based seismic data frequency extension method provided by one embodiment of the present invention;

[0048] Figure 2 A flowchart of a method for training a seismic data spectrum extension model provided by one embodiment of the present invention;

[0049] Figure 3 A schematic diagram illustrating the principle of spectrum blueing provided by an embodiment of the present invention;

[0050] Figure 4 A schematic diagram of a frequency division technology provided by an embodiment of the present invention;

[0051] Figure 5 A schematic diagram of a spectrum after conventional spectrum blueing and spectrum spreading provided by an embodiment of the present invention;

[0052] Figure 6 A schematic diagram of a spectrum after spectrum blueing and spectrum spreading provided by an embodiment of the present invention;

[0053] Figure 7 A schematic diagram of five evaluation parameters provided in one embodiment of the present invention;

[0054] Figure 8 A schematic diagram of a device for performing frequency spreading of seismic data provided by one embodiment of the present invention;

[0055] Figure 9 A schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following describes the embodiments of the present invention in detail with reference to the accompanying drawings and examples, so that the present invention can fully understand how to apply technical means to solve technical problems and achieve technical effects, and thus implement the invention accordingly. It should be noted that, as long as no conflict exists, the various embodiments of the present invention and the various features therein can be combined with each other, and the resulting technical solutions are all within the scope of protection of the present invention.

[0057] Additionally, the steps shown in the flowcharts of the accompanying drawings may be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowcharts, in some cases the steps shown or described may be performed in an order different from that shown.

[0058] The inventors discovered that existing methods for processing seismic data using spectrum spreading (such as deconvolution and spectral whitening) often suffer from data compensation distortion, loss of geological background significance, and low resolution due to oversimplification of low frequencies. This affects the accuracy of seismic reservoir prediction and makes it difficult to meet actual needs. Based on this, the inventors developed a spectrum blueing spectrum spreading method. The formation reflection coefficient after spectrum spreading is more realistic and the resolution is higher, which has better resolution for some thin reservoirs. The spectrum blueing spectrum spreading method only calculates the high-frequency data part and does not balance the calculation of all frequency bands of seismic data. This often leads to reduced spectrum spreading resolution, jitter in the low-frequency part, abnormal spectrum distribution, and other phenomena, which in turn causes some low-frequency information related to reservoir prediction in the original seismic data to be masked or missing.

[0059] Based on this, the inventors provide a seismic data frequency extension method and device based on deep learning, applying deep learning technology to seismic data processing to solve the problems existing in the above-mentioned existing technologies, realize effective frequency extension of seismic data, and improve the resolution of seismic data.

[0060] Example 1

[0061] Reference Figure 1 The present invention provides a deep learning-based seismic data frequency extension method, which specifically includes:

[0062] Step S1: performing spectrum blueing processing based on the acquired seismic data of the study area to obtain a spectrum blueing data volume.

[0063] Step S2: input the spectrum blued data volume into the seismic data frequency extension model to obtain the seismic data after frequency extension.

[0064] See Figure 2 , the earthquake data frequency extension model in step S2 can be obtained by the following training method:

[0065] Step S201: Perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples.

[0066] Due to the limitations of field seismic acquisition conditions and the requirements for surface wave suppression, low-frequency information is often lacking in the original seismic gather data. At the same time, as seismic waves propagate underground, their high-frequency components decay rapidly with increasing propagation distance, so the energy of the high-frequency part is also suppressed. These factors often lead to problems such as low main frequency, narrow frequency band, and low resolution in the original training data samples. Traditional spectrum spreading processing technology, due to the lack of objective constraints and prior means, has a certain degree of blindness in protecting low-frequency signals and recovering high-frequency signal energy.

[0067] Seismic frequency division is a method for improving seismic imaging resolution based on spectral decomposition. Because seismic signals of different frequencies exhibit a certain tuning effect corresponding to reservoirs of varying thicknesses, this frequency characteristic can be exploited to divide training data samples into distinct frequency bands. The energy of these training data samples can then be recovered within these bands, enabling detailed characterization of geological structures at varying levels.

[0068] Spectral blueing is a processing technique for training data samples that aims to improve the resolution and imaging quality of seismic data by enhancing the energy of high-frequency components. Spectral blueing enhances the high-frequency components in training data samples, making images clearer and sharper. However, this spectrum spreading method only processes the high-frequency band of the training data samples, rather than calculating the full frequency band. This can cause jitter and abnormal spectral distribution in the low-frequency portion of the training data samples, resulting in the masking or loss of low-frequency information. This can seriously affect the accuracy of structural imaging and reservoir prediction.

[0069] Due to the above-mentioned problems with conventional spectrum blueing technology, the inventors proposed a method for spectrum blueing processing, which aims to combine traditional spectrum blueing technology with frequency division technology and optimize the entire processing flow. In this way, in the process of improving resolution processing, the interference of mid- and high-frequency components on low-frequency signals can be avoided.

[0070] In order to obtain high-resolution training data samples, the obtained training data samples are subjected to spectrum blueing processing in step S201 to obtain spectrum blueing data volume samples. For details, see Figure 3 The specific implementation process of step S201 includes:

[0071] (1) The training data samples are converted into time-frequency and frequency-divided by frequency division technology to extract training data samples of different frequency bands. The training data samples of each frequency band correspond to a frequency-divided data, and all the frequency-divided data form a frequency-divided data body.

[0072] Here, the training data samples can also be preprocessed before the time-frequency conversion is performed on the training data samples to reduce noise and interference and improve the signal-to-noise ratio of the training data samples. Denoising is performed by constructing a guided filtering method until the training data samples meet the set resolution requirements in order to better interpret the underground geological structure and stratum property information. Of course, the preprocessing method can also adopt the static correction processing, deconvolution and other methods in conventional technology, as long as the effect of improving the resolution of the training data samples can be achieved. Here, only one preprocessing embodiment is provided, and other preprocessing methods are also within the protection scope of this embodiment. Those skilled in the art can refer to the detailed description of the prior art for the specific preprocessing method, which will not be repeated here.

[0073] See Figure 4 Frequency division techniques can be used to divide training data samples into different frequency bands. These bands can then be processed separately to more clearly display the subtle features of the seismic signal, thereby obtaining more refined training data samples and improving the resolution of the training data samples. Frequency division techniques mainly include discrete Fourier transform, continuous wavelet transform, S transform, and other methods. The specific method of frequency division is not limited here; as long as it can achieve high frequency division efficiency and fine analysis results, it will be sufficient.

[0074] (2) Performing spectral blueing on the frequency-divided data volume to obtain a spectral blueing operator for the frequency-divided data volume, and calculating weighting factors based on the energy signals of each frequency band corresponding to the frequency-divided data volume. That is, the weighting factors of the spectral blueing operator within each frequency band are calculated based on the energy distribution characteristics within each frequency band. The calculation method of the spectral blueing operator can be used in conventional techniques and will not be described in detail here.

[0075] (3) According to the spectrum blueing operator and the weighting factor, an optimized operator is obtained, that is, the optimized spectrum blueing operator.

[0076] Because seismic signals in different frequency bands vary significantly in their ability to reflect formation reflections, a dominant frequency band exists for the target formation that better reflects seismic reflection signals, yielding more accurate reflector information. The dominant frequency band often has higher signal energy. By leveraging the energy distribution characteristics of seismic signals, we calculate the energy contribution of different frequency bands and use this as a weighting factor to optimize the spectral blueing operator. The resulting optimized operator is more accurate and reflects richer formation information.

[0077] (4) Resample the frequency-divided data volume to obtain the reflection coefficient corresponding to each frequency band.

[0078] (5) Based on the reflection coefficients and optimization operators corresponding to each frequency band, deconvolution is performed to obtain a spectral blued data volume.

[0079] See Figure 5 and Figure 6 Comparing the spectrum obtained after traditional spectrum blueing and the spectrum blueing process provided by this solution, we can see that the spectrum of the training data samples after spectrum blueing has stronger high-frequency energy overall. The dominant frequency obtained after traditional spectrum blueing is 30Hz, while the dominant frequency after spectrum blueing is increased to 50Hz, significantly enhancing weak high-frequency signals. Furthermore, the spectrum of the training data samples after spectrum blueing is smooth in the low-frequency portion, without sudden energy enhancement or distortion. Furthermore, the frequency band is also broadened to a certain extent, and the spectrum distribution of the training data samples after spectrum blueing is more reasonable, which enhances the ability to identify thin reservoirs.

[0080] Step S202: Use the spectral blued data volume samples to train the pre-built deep learning algorithm model to obtain a trained model.

[0081] When training the pre-built deep learning algorithm model in the above step S202, deep learning training algorithms such as back propagation and gradient descent in conventional methods can be used for training. Those skilled in the art can refer to the detailed description of the prior art for the specific training method, which will not be described in detail here.

[0082] Step S203: Evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model.

[0083] See Figure 7 In step S203, when evaluating the trained model, the accuracy, precision and recall, F1 score, AUC, and confusion matrix can be used to evaluate and analyze the accuracy and precision of the training model calculation results to evaluate the reliability and accuracy of the processed training data samples.

[0084] Accuracy: In single-label classification tasks, each sample has a single, definite category. Predicting that category is considered a correct classification, while not predicting it is considered an incorrect classification. Therefore, accuracy is the most intuitive metric for evaluating the classification performance of a trained model, representing the probability that all samples are correctly classified. In multi-class classification tasks, Top-1 Accuracy (the proportion of correctly predicted samples to the total number of samples) and Top-5 Accuracy (the proportion of correctly predicted samples to the top five predicted samples) are commonly used to evaluate the accuracy of a trained model.

[0085] Precision and Recall: Precision indicates the proportion of samples predicted as positive among all samples predicted as positive. Recall indicates the proportion of samples predicted as positive among all true positive samples. Precision and recall are conflicting metrics, and a trade-off is often necessary. The performance of a trained model is evaluated by plotting a curve with recall as the horizontal axis and precision as the vertical axis. The area enclosed by the curve and the coordinate axes is used for quantitative evaluation.

[0086] F1 score: The F1 score is the harmonic mean of precision and recall, comprehensively evaluating both precision and recall performance. The F1 score is calculated as: F1 = 2 * (precision * recall) / (precision + recall). Its maximum value is 1, and its minimum value is 0. A higher F1 score indicates that the trained model performs well in both precision and recall.

[0087] AUC (Area Under Curve): AUC represents the area under the receiver operating characteristic (ROC) curve and is used to evaluate the performance of a trained model at all thresholds. It comprehensively considers the impact of both false positives and false negatives, providing a more comprehensive assessment of the performance of a trained model.

[0088] Confusion Matrix: A confusion matrix is ​​a table that displays the prediction results of a trained model, detailing the misclassification between categories. Using the confusion matrix, we can calculate the true and predicted class labels for each category, and then calculate various evaluation metrics such as precision, recall, and F1 score.

[0089] Based on the evaluation results corresponding to the above evaluation parameters, the trained model is optimized through back propagation and optimization algorithms, such as gradient descent, to obtain an optimized model. Other model optimization methods can also be used, which are not described in detail here.

[0090] Step S204: Repeat the process of training, evaluating, and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model.

[0091] Users can set preset thresholds for accuracy, precision and recall, F1 score, AUC, and confusion matrix. When the evaluation results of the above evaluation parameters reach the preset thresholds, that is, when the preset evaluation conditions are met, the seismic data spectrum extension model is obtained to improve the precision and accuracy of the seismic data resolution.

[0092] Furthermore, after obtaining the seismic data extension model, the spectral blued data volume in step S1 is processed based on the model to obtain extended seismic data. This extended seismic data can then be displayed using spatial visualization techniques to achieve data visualization. Comparing this data with the pre-processed seismic data from the study area can more intuitively demonstrate the difference between the two, significantly improving the resolution of the extended seismic data. Furthermore, the extended seismic data can be exported in a standard format for subsequent seismic interpretation and reservoir inversion.

[0093] In the embodiment of the present invention, the process of performing spectral blueing processing on the seismic data of the study area to obtain a spectral blueing data volume in the above-mentioned step S1 can refer to the implementation method of performing spectral blueing processing on the training data samples to obtain a spectral blueing data volume sample in the above-mentioned step S201, and will not be repeated here.

[0094] The deep learning-based seismic data extension method provided in an embodiment of the present invention obtains seismic data from a study area and performs spectrum blueing processing to obtain a spectrum blueing data volume. The blueing data volume is then input into a seismic data extension model to implement spectrum extension of the seismic data and obtain extended seismic data. Since spectrum blueing is used to perform spectrum extension processing on training data samples during the construction of the seismic data extension model, all frequency bands of the training data samples are balanced. Compared with the existing spectrum extension processing methods, the spectrum extension resolution is improved, the problem of low-frequency jitter is eliminated, and the spectrum distribution anomaly is avoided. The spectrum extension result is more superior and can retain some low-frequency information related to reservoir prediction in the original training data samples. It is an important technical solution for establishing a seismic data extension model that can effectively implement spectrum extension. Therefore, the extended seismic data can effectively widen the seismic signal band while maintaining the signal-to-noise ratio, amplitude and time-frequency characteristics, improve the resolution of the seismic signal, make the seismic profile phase axis reflection characteristics clearer, enrich the interlayer information, and more precisely identify thin layers. It can better identify the structural and reservoir characteristics in the study area and effectively improve the resolution of seismic data.

[0095] Example 2

[0096] Based on the same inventive concept, the present invention also provides a method for training a seismic data spectrum extension model. Figure 2 ,include:

[0097] Step S201: Perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples.

[0098] Step S202: Using the spectral blued data volume samples, a pre-built deep learning algorithm model is trained to obtain a trained model;

[0099] Step S203: Evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model;

[0100] Step S204: Repeat the process of training, evaluating, and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model.

[0101] In the embodiment of the present invention, the specific implementation process of the training method of the seismic data frequency extension model can refer to the detailed description of the seismic data frequency extension method based on deep learning in the above embodiment 1, and will not be repeated here.

[0102] Example 3

[0103] Based on the same inventive concept, an embodiment of the present invention further provides a device for seismic data frequency spreading, see Figure 8 ,include:

[0104] The data analysis module 101 is configured to perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples.

[0105] The first training module 102 is used to train a pre-built deep learning algorithm model using the spectral blued data volume samples to obtain a trained model.

[0106] The evaluation and optimization module 103 is used to evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model.

[0107] The second training module 104 is used to repeat the process of training, evaluating and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model.

[0108] The data acquisition module 105 is used to perform spectrum blueing processing based on the acquired seismic data of the study area to obtain a spectrum blueing data volume.

[0109] The data processing module 106 is used to input the spectrum blued data volume into the seismic data frequency extension model to obtain the seismic data after frequency extension.

[0110] The implementation principle and technical effects of the device for seismic data frequency spreading provided in the embodiment of the present invention are similar to those of the first embodiment and will not be described in detail here.

[0111] Example 4

[0112] Based on the same inventive concept, an embodiment of the present invention further provides a training device for a seismic data spectrum extension model, comprising:

[0113] The data analysis module is used to perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples.

[0114] The first training module is used to train a pre-built deep learning algorithm model using spectral blued data body samples to obtain a trained model.

[0115] The evaluation and optimization module is used to evaluate the trained model based on multiple preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model.

[0116] The second training module repeats the process of training, evaluating and optimizing the optimization model until the optimization model meets the preset evaluation conditions and obtains the seismic data topology model.

[0117] The implementation principle and technical effects of the training device for the seismic data frequency extension model provided in the embodiment of the present invention are similar to those of the second embodiment and will not be described in detail here.

[0118] Example 5

[0119] Based on the same inventive concept, a computer-readable storage medium is provided in an embodiment of the present invention. When the program is executed by a processor, it implements the deep learning-based seismic data frequency extension method in embodiment one, and / or the training method of the seismic data frequency extension model in embodiment two.

[0120] The computer-readable storage medium may be included in the device / apparatus described in the above embodiments, or may exist independently and not be incorporated into the device / apparatus. The computer-readable storage medium carries one or more programs, which, when executed, implement the method according to Embodiment 1 or Embodiment 2 of the present invention.

[0121] According to an embodiment of the present invention, a computer-readable storage medium may be a non-volatile computer-readable storage medium, such as, but not limited to, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device.

[0122] Example 6

[0123] Based on the same inventive concept, an embodiment of the present invention provides an electronic device, referring to Figure 9 , including a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112 and the memory 113 communicate with each other through the communication bus 114,

[0124] Memory 113, for storing computer programs;

[0125] The processor 111 is configured to implement the deep learning-based seismic data frequency extension method of the first embodiment and / or the seismic data frequency extension model training method of the second embodiment when executing the program stored in the memory 113 .

[0126] The implementation principle and technical effects of the electronic device provided by the embodiment of the present invention are similar to those of the aforementioned embodiment 1 or embodiment 2, and will not be repeated here.

[0127] The memory 113 can be an electronic memory such as a flash memory, an EEPROM (Electrically Erasable Programmable Read-Only Memory), an EPROM, a hard disk, or a ROM. The memory 113 has storage space for program code for executing any of the method steps described above. For example, the storage space for program code can include individual program codes for implementing each step of the method described above. These program codes can be read from or written to one or more computer program products. These computer program products include program code carriers such as hard disks, compact disks (CDs), memory cards, or floppy disks. Such computer program products are typically portable or fixed storage units. The storage unit can have storage segments or storage spaces arranged similarly to the memory 113 in the electronic device described above. The program code can be compressed, for example, in a suitable form. Typically, the storage unit includes a program for executing the method steps according to Embodiment 1 or Embodiment 2 of the present invention, i.e., code that can be read by, for example, the processor 111. When executed by an electronic device, these codes cause the electronic device to execute the various steps of the method described in Embodiment 1 or Embodiment 2.

[0128] It will be understood by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Thus, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0129] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowcharts and / or block diagrams, as well as combinations of processes and / or blocks in the flowcharts 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, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowcharts and / or block diagrams. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0130] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0131] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0132] The specific embodiments described above further illustrate the objectives, technical solutions and beneficial effects of the present invention in detail. It should be understood that the above description is only a specific embodiment of the present invention and is not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A seismic data frequency extension method based on deep learning, characterized in that: include: Based on the seismic data of the study area, spectrum blueing processing is performed to obtain spectrum blueing data volume; The spectrum blued data volume is input into a seismic data extension model to obtain extended seismic data; wherein the seismic data extension model is trained by the following method: Performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples; Using the spectral blueing data volume sample, training a pre-built deep learning algorithm model to obtain a trained model; Evaluating the trained model based on a plurality of preset evaluation parameters, and optimizing the trained model according to the obtained evaluation results to obtain an optimized model; The process of training, evaluating and optimizing the optimization model is repeated until the optimization model meets the preset evaluation conditions, thereby obtaining the seismic data spectrum extension model.

2. The deep learning-based seismic data frequency extension method according to claim 1, wherein: The performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples includes: Based on the training data samples, extracting the training data samples of different frequency bands by using a frequency division technology to obtain the frequency division data volume; performing a spectrum blueing process on the frequency-divided data volume to obtain a spectrum blueing operator for the frequency-divided data volume, and calculating a weighting factor based on energy signals of each frequency band corresponding to the frequency-divided data volume; Obtaining the optimization operator according to the spectrum blueing operator and the weighting factor; Resampling the frequency-divided data volume to obtain reflection coefficients corresponding to each frequency band; The spectral blueing data volume is obtained based on the reflection coefficient corresponding to each frequency band and the optimization operator.

3. The seismic data frequency extension method based on deep learning according to claim 1, characterized in that: Before performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples, the method further includes: The training data samples are preprocessed and denoised by constructing a guided filtering method until the training data samples meet the set resolution requirements.

4. The deep learning-based seismic data frequency extension method according to claim 1, wherein: After obtaining the seismic data after frequency spreading, the following is also included: The spatial visualization technology is used to display the seismic data after frequency spreading, and a comparative analysis is performed with the seismic data of the study area before processing.

5. The deep learning-based seismic data frequency extension method according to claim 1, wherein: The evaluation parameters include accuracy, precision and recall, F1 score, AUC and confusion matrix; The trained model is evaluated based on a plurality of preset evaluation parameters in the following manner: The trained model is evaluated using accuracy, precision and recall, F1 score, AUC and confusion matrix to obtain corresponding evaluation results.

6. A method for training a seismic data frequency extension model, characterized in that: include: Performing spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples; Using the spectral blueing data volume sample, training a pre-built deep learning algorithm model to obtain a trained model; Evaluating the trained model based on a plurality of preset evaluation parameters, and optimizing the trained model according to the obtained evaluation results to obtain an optimized model; The process of training, evaluating and optimizing the optimization model is repeated until the optimization model meets the preset evaluation conditions, thereby obtaining the seismic data spectrum extension model.

7. A device for frequency spreading of seismic data, characterized in that: include: A data analysis module is used to perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples; A first training module is configured to train a pre-built deep learning algorithm model using the spectral blued data volume sample to obtain a trained model; An evaluation and optimization module is used to evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model; A second training module is used to repeat the process of training, evaluating and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model; A data acquisition module is used to perform spectrum blueing processing based on the acquired seismic data of the study area to obtain a spectrum blueing data volume; The data processing module is used to input the spectrum blued data volume into the seismic data frequency extension model to obtain the seismic data after frequency extension.

8. A training device for a seismic data frequency extension model, characterized in that: include: A data analysis module is used to perform spectrum blueing processing on the acquired training data samples to obtain spectrum blueing data volume samples; A first training module is configured to train a pre-built deep learning algorithm model using the spectral blued data volume sample to obtain a trained model; An evaluation and optimization module is used to evaluate the trained model based on a plurality of preset evaluation parameters, and optimize the trained model according to the obtained evaluation results to obtain an optimized model; The second training module is used to repeat the process of training, evaluating and optimizing the optimization model until the optimization model meets the preset evaluation conditions, thereby obtaining a seismic data spectrum extension model.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it implements the deep learning-based seismic data frequency extension method as described in any one of claims 1 to 5, and / or the training method of the seismic data frequency extension model as described in claim 6.

10. An electronic device, characterized in that: The processor, the communication interface, the memory and the communication bus are connected to each other via the communication bus. Memory for storing computer programs; The processor is used to implement the deep learning-based seismic data frequency extension method as described in any one of claims 1 to 5, and / or the training method of the seismic data frequency extension model as described in claim 6 when executing the program stored in the memory.

Images