A method and system for improving resolution of middle-deep seismic data

Abstract

The application discloses a method and system for improving resolution of middle-deep seismic data, and the method comprises the following steps: preprocessing three-dimensional seismic data of the same work area, combining with an automatic gain control function to construct seismic data; introducing a data augmentation method and a data splicing method; using a 3D Cycle-GAN network architecture to extract the characteristics of shallow high-resolution seismic data; introducing a loss function composed of a generator loss and a discriminator loss; inputting shallow, deep seismic data and corresponding shallow low-resolution data obtained by low-pass filtering of shallow high-resolution data into the network for training; inputting deep low-resolution data into the network, and performing AGC inverse transformation on the output, so as to finally improve the resolution of deep seismic data. The application uses the characteristics that the deep layer and the shallow layer in the same area have very similar geological structure characteristics, and uses unpaired deep layer and shallow layer data for weak supervision learning, so that the resolution of the deep layer seismic data can be effectively improved.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of seismic signal processing, and particularly relates to a method and system for improving resolution of medium-deep seismic data. BACKGROUND

[0002] Oil and gas resources are the basis of modern industry and play an irreplaceable role in the national economy and people's livelihood. Seismic exploration is a technology for revealing the characteristics of underground structures and inferring the properties and forms of strata. Through seismic exploration, seismic data can be obtained, which can be used to infer the depth and fluctuation of strata and calculate the changes in wave impedance in the strata to reveal geological structures and strata composition, which has important reference value for oil and gas exploration. In particular, high-frequency information often contains key stratum details, such as layering and fault information. Therefore, obtaining high-resolution seismic data is crucial for accurately interpreting underground structures. However, the acquisition process of seismic data is affected by factors such as stratum attenuation, resulting in a decrease in the bandwidth of the obtained original data and a low data resolution. Although modern seismic data acquisition technology has made significant progress, the quality and resolution of the acquired data have been greatly improved, but many acquired seismic data still face the problem of low resolution. With the development of exploration technology over the years, the exploration target has shifted to some deep strata with more complex geological structures, increasing the difficulty of data acquisition and the accuracy requirements. High-resolution seismic data are essential for accurate interpretation of geological structures, so it is necessary to study high-resolution methods for seismic data and improve the resolution of seismic data, which has attracted widespread attention from researchers

[0003] In recent years, with the rapid development of deep learning, deep learning-based techniques have been widely applied in many fields and have achieved remarkable results. Deep learning can automatically extract relevant features from the learning target based on selected input data and learning goals, which is very suitable for processing complex seismic data. Currently, deep learning-based techniques have been applied in many directions of seismic data processing, such as noise suppression, interpolation, fault detection, and resolution enhancement.

[0004] Deep learning-based seismic data processing methods are in a stage of rapid development in recent years, and scholars have conducted a lot of work in the fields of noise suppression, interpolation, and high resolution. Studies have proven the feasibility of deep learning in seismic data processing. However, the current research on applying deep learning to actual seismic data processing is still limited by sample quality and is in its early stages. To avoid the limitations of training samples, self-supervised learning has received widespread attention from scholars in the fields of image restoration, image completion, and image generation. Research on how to construct a non-paired data set suitable for network weak supervision learning can apply deep learning techniques to a wider range of actual seismic data high-resolution processing, which has valuable research value.

[0005] The existing deep learning method usually needs complete corresponding paired data and labels for training, but since the strong supervision learning has very strict requirements for the paired samples, each low resolution data needs to be completely paired with the high resolution label matched therewith, and thus the quality of the high resolution label directly determines the effect of the high resolution processing result. Therefore, in order to apply the strong supervision learning method to the field of high resolution processing of seismic data, it is a big difficulty to obtain a label with good quality. SUMMARY

[0006] The technical problem to be solved by the present application is to provide a method and system for improving the resolution of medium-deep seismic data, which solves the technical problem of the existing method that needs paired data for strong supervision learning.

[0007] The present application adopts the following technical solutions:

[0008] A method for improving the resolution of medium-deep seismic data, comprising the following steps:

[0009] S1, preprocessing three-dimensional seismic data of the same work area; combining an automatic gain control function to construct seismic data, and introducing a data augmentation method and a data splicing method to supplement the seismic data;

[0010] S2, constructing a 3D Cycle-GAN network architecture for extracting shallow high-resolution seismic data features;

[0011] S3, introducing a pair of loss functions composed of generator loss and discriminator loss into the 3D Cycle-GAN network obtained in step S2;

[0012] S4, inputting the shallow and deep seismic data obtained in step S1 and the corresponding shallow low-resolution data obtained by low-pass filtering the shallow high-resolution data into the 3D Cycle-GAN network obtained in step S3 for training, inputting the deep low-resolution data into the trained 3D Cycle-GAN network, and performing AGC inverse transformation on the output to improve the resolution of the deep seismic data.

[0013] Preferably, in step S1, for each trace d i of the input seismic data D in , the result d′ i of the signal after preprocessing using the automatic gain control algorithm is as follows:

[0014]

[0015] Wherein, W triangle (L) is a triangular window with length L.

[0016] Preferably, the data augmentation method is specifically:

[0017] Flip the two spatial dimensions of Inline and Crossline, exchange the two spatial dimensions of Inline and Crossline, take the complement of the data as a whole, and obtain the augmented data Each pair of braces represents randomly selecting one item with equal probability.

[0018] Preferably, the data splicing method is specifically:

[0019] When the input data is divided into blocks, 50% overlap is set for the data;

[0020] After high-resolution processing is completed, the overlapping part between high-resolution data blocks is multiplied by a set weight through windowing method, and spliced back to the original shape, so that the sum of the weights of the data at each position after splicing is 1;

[0021] The triangular window is extended to three-dimensional space, and the expression of the window function is as follows:

[0022]

[0023] Wherein, the length of the window function is 2N, i, j, k are the indexes of three dimensions, 0≤i,j,k≤2N.

[0024] Preferably, in step S2, the 3D Cycle-GAN network architecture includes a forward cycle and a reverse cycle, the forward cycle and the reverse cycle adopt two sets of generators and two sets of discriminators, the forward and reverse networks of the two sets of generators and discriminators maintain the same architecture, share the same architecture and are jointly trained, and the parameters are updated independently;

[0025] The generator includes a forward generator G LR2HR And a reverse generator G HR2LR The forward generator G HR2LR maps the LR data to the HR data, and the reverse generator G LR2HR maps the HR data to the LR data.

[0026] The discriminator includes an LR discriminator D LR And an HR discriminator D HR The LR discriminator D LR and the HR discriminator D HR are used to determine whether the input data is real LR or HR data, respectively.

[0027] Preferably, the forward generator G LR2HR and the reverse generator G HR2LR are three-dimensional networks, which simultaneously process two-dimensional data and three-dimensional data and adopt the same network architecture.

[0028] Data is first input by the first convolutional layer, and then passes through 16 concatenated residual blocks in the main part, each of which contains two convolutional layers and a residual connection, and there is a cross-layer connection outside the residual block group;

[0029] Finally, the data passes through multiple convolutional layers and is output through a tanh activation function.

[0030] Preferably, no pooling layer is included in the generator network; residual blocks are used, and no batch normalization layer is used.

[0031] Preferably, the LR discriminator D LR and the HR discriminator D HR are two-dimensional networks with the same network architecture.

[0032] Data is first input by the first convolutional layer, and then passes through 5 convolutional blocks;

[0033] Each convolutional block contains a convolutional layer, a batch normalization layer, and a Leaky ReLU activation function.

[0034] Then the data passes through a Sigmoid activation function to output a probability between 0 and 1.

[0035] The discriminator uses a PatchGAN architecture, replacing the fully connected layer in the last layer with a convolutional layer.

[0036] Preferably, in step S3, the loss function is decomposed into a generator loss and a discriminator loss.

[0037] The generator loss is:

[0038] L Gen = w1L GAN + w2L cyc + w3L idt + w4L TV

[0039] where w1, w2, w3, and w4 represent the weights assigned to each loss function, L GAN is the adversarial loss, L cyc is the cycle consistency loss, L idt is the identity loss, and L TV is the total variation loss.

[0040] The discriminator loss is:

[0041]

[0042] where N is the number of training samples in a round of training, D HR is the HR discriminator, D LR is the LR discriminator, and GLR2HR (x i ) is low-resolution data x, and the high-resolution data obtained by inputting the forward generator, G HR2LR (y i ) is low-resolution data obtained by inputting the reverse generator to the high-resolution data y.

[0043] In a second aspect, an embodiment of the present application provides a system for improving resolution of middle-deep seismic data, comprising:

[0044] A data module is configured to preprocess three-dimensional seismic data of the same work area, construct seismic data in combination with an automatic gain control function, and introduce a data augmentation method and a data splicing method to supplement the seismic data.

[0045] A network module is configured to construct a 3D Cycle-GAN network architecture for extracting features of shallow high-resolution seismic data.

[0046] A function module is configured to introduce a pair of loss functions combined by a generator loss and a discriminator loss into the 3D Cycle-GAN network.

[0047] An improvement module is configured to input shallow, deep seismic data and corresponding shallow low-resolution data obtained by low-pass filtering of shallow high-resolution data into the 3D Cycle-GAN network obtained in step S3 for training, input deep low-resolution data into the trained 3D Cycle-GAN network, and perform AGC inverse transformation on the output to improve resolution of deep seismic data.

[0048] In a third aspect, a computer device comprises a memory, a processor, and a computer program stored in the memory and executable on the processor, and the processor implements steps of the method for improving resolution of middle-deep seismic data when executing the computer program.

[0049] In a fourth aspect, an embodiment of the present application provides a computer readable storage medium comprising a computer program, and the computer program implements steps of the method for improving resolution of middle-deep seismic data when executed by a processor.

[0050] Compared with the prior art, the present application has at least the following beneficial effects:

[0051] A method for improving the resolution of middle-deep seismic data is provided. Based on the Cycle-GAN network structure, a two-dimensional-three-dimensional Cycle-GAN architecture is proposed. Due to the long-term erosion of water flow in the deep ditch area of the loess plateau, the surface loess is lost, and thus the influence of the seismic wave is greatly reduced. Therefore, the resolution of the curved line seismic data collected along the deep ditch area is higher than that of the three-dimensional loess plateau data. Since the loess plateau data and the curved line seismic data are collected at the same geographical location, they actually reflect the same stratum characteristics, and thus the data of the two have strong correlation. Based on this feature, a deep learning method is used to establish the feature distribution mapping of the three-dimensional loess plateau data and the two-dimensional curved line seismic data by using a weakly supervised learning method. Meanwhile, data augmentation, automatic gain control and unique data splicing processing are added to the three-dimensional seismic data, so as to improve the resolution of the data of the entire loess plateau area.

[0052] Further, due to the unpaired characteristics of the three-dimensional work area data and the two-dimensional curved line seismic data, and the use of a weakly supervised learning method to learn the distribution and mapping relationship of the two, due to the different dimensions of the two, a convolutional neural network is designed to simultaneously process two-dimensional and three-dimensional data, thereby better preserving the spatial information of the seismic data.

[0053] Further, since the loess plateau data and the curved line seismic data have the same geographical location, but the dimensions of the data are different, and there is a complex corresponding relationship between the two, in order to solve the problem of neural network learning difficulty, bidirectional recurrent consistency is introduced, and the forward and reverse networks maintain the same architecture, share the same architecture and are jointly trained. Based on the above conditions, a unified learning process is ensured. Due to the introduction of bidirectional recurrence, the stability in training is significantly improved, the artifacts of the output data are reduced, and the overall quality of the generated HR seismic data is improved.

[0054] Further, since in actual seismic data, the reflection coefficients of different strata may differ greatly, and the seismic signals received by the geophone also vary within a large value range. If the selected part of the reflection signal is too weak, the network will have difficulty learning the characteristics of the weak reflection signal. In order to adjust the amplitude of the signal to a suitable range, an automatic gain control algorithm (AGC) is introduced to preprocess the signal, and the amplitude of the signal is adjusted to a range suitable for network training. After the network processing is completed, the amplitude of the signal is restored by using AGC inverse transformation. The AGC method is conducive to the network capturing the characteristics of the weak reflection signal.

[0055] Further, in order to solve the problem of improving the resolution of the seismic data, a pair of loss functions is designed according to the characteristics of the seismic data. The loss function is composed of two parts: generator loss and discriminator loss. Gen The loss function is composed of four parts: adversarial loss LGAN , cycle consistency loss L cyc , identity loss L idt , and total variation loss L TV The discriminator loss is also defined by the negative log-likelihood. This reduces the prediction error and prevents overfitting.

[0056] Further, in order to prevent overfitting due to insufficient training data, a large amount of training data is required, but due to the limitations of exploration technology, the old data used for high-resolution processing is often small in scale, exacerbating the problem of insufficient training data, so we introduce a series of data augmentation methods. Through this data augmentation method, the number of training samples is effectively expanded. Input data and labels can use different data augmentation methods, so that the network can access the entire data set from multiple angles, providing more flexibility compared to paired training data augmentation methods.

[0057] Further, due to the limitation of GPU memory, the network cannot input the entire three-dimensional seismic data body at once, so when processing high-resolution data, the data needs to be subdivided into fixed-size blocks, and the traditional direct splicing strategy may introduce discontinuity at the boundary of the processed block. In order to cope with this challenge and enhance the overall integrity of the seismic data, we propose an improved boundary processing strategy. This improved boundary processing strategy effectively reduces the artifacts that may be generated by the direct splicing method. By adopting our design of overlapping sampling and data windowing technology, our method ensures more accurate preservation of high-resolution seismic features, ensuring the effectiveness of the final output of high-resolution data in three-dimensional seismic data processing.

[0058] It can be understood that the beneficial effects of the above-mentioned second aspect can be referred to the related description in the above-mentioned first aspect, which will not be repeated here.

[0059] In summary, based on the Cycle-GAN network structure, the present application proposes a two-dimensional-three-dimensional Cycle-GAN architecture; the data of the loess tableland work area and the curved line seismic data have the same geographical acquisition location, and both actually reflect the same stratum characteristics, have strong correlation characteristics, using a deep learning method, by establishing a feature distribution mapping between three-dimensional work area loess tableland data and two-dimensional curved line seismic data, while adding data augmentation, automatic gain control and unique data splicing processing to the seismic data, thereby improving the resolution of the entire loess tableland data.

[0060] The technical solutions of the present application will be further described in detail below with the help of the drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0061] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the accompanying drawings used in the relative embodiment description are briefly introduced as follows. Obviously, the accompanying drawings in the following description only constitute some embodiments of the present application, and other accompanying drawings can be obtained by those skilled in the art without any creative effort on the basis of these accompanying drawings.

[0062] Figure 1

[0063] Figure 2

[0064] Figure 3

[0065]

[0066] Figure 5

[0067]

[0068] Figure 7

[0069]

[0070] Figure 9

[0071] Figure 10

[0072] Figure 11

[0073] ​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​​Figure 12 (a) is a Crossline profile of the actual seismic data, the red box highlights the shallow part which has relatively higher resolution and is taken as the network input. In contrast, the blue box marks the deep part which has lower resolution due to seismic wave attenuation and is chosen as the training target for weakly supervised learning. (b) is the normalized amplitude spectrum comparison between the shallow high-resolution (HR) data (red) and the deep low-resolution (LR) data (blue).

[0074] Fig. 13 is a normalized amplitude spectrum, wherein (a) is the normalized amplitude spectrum of the Inline profile in Fig. 14, (b) is the normalized amplitude spectrum of the Crossline profile in Fig. 15;

[0075] Fig. 14 is an Inline profile of the actual seismic data, wherein (a) is the deep LR seismic data, (b) is the HR result obtained by the traditional time-varying spectral whitening method, (c) is the result obtained by using the method of the present application, the green box shows that the method of the present application significantly improves the resolution, showing its robustness in revealing geological structures;

[0076] Fig. 15 is a Crossline profile of the actual seismic data, wherein (a) is the deep LR seismic data, (b) is the HR result obtained by the traditional time-varying spectral whitening method, (c) is the result obtained by using the method of the present application, the yellow box shows that the method of the present application significantly improves the resolution under noise conditions, indicating its strong anti-interference ability to noise;

[0077] Fig. 16 is a time profile of the actual seismic data at t = 2284 ms, wherein (a) is the original LR seismic data, (b) is the HR result obtained by the traditional time-varying spectral whitening method, (c) is the result obtained by using the method of the present application, as shown by the yellow rectangle, the method of the present application depicts clearer geological structures than the spectral whitening method;

[0078] Fig. 17 is a time profile of the actual seismic data at t = 2504 ms, wherein (a) is the original LR seismic data, (b) is the HR result obtained by the traditional time-varying spectral whitening method, (c) is the result obtained by using the method of the present application, as shown by the yellow rectangle, the method of the present application reveals more subtle geological structures;

[0079] Figure 18 Fig. 18 is a flowchart of the method of the present application;

[0080] Figure 19 Fig. 19 is a schematic diagram of a computer device provided by an embodiment of the present application;

[0081] Figure 20 Fig. 20 is a block diagram of an electronic device provided by the present application according to an embodiment. DETAILED DESCRIPTION

[0082] The technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments of the present application. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort belong to the scope of the present application.

[0083] In the description of the present application, it should be understood that the terms "comprising" and "including" indicate the presence of described features, integers, steps, operations, elements, and / or components, but do not exclude one or more other features, integers, steps, operations, elements, components, and / or groups thereof.

[0084] It should also be understood that the terms used in the specification of the present application are only for the purpose of describing specific embodiments and are not intended to limit the present application. As used in the specification and the appended claims of the present application, the singular forms "a", "an" and "the" are intended to include the plural forms unless the context clearly indicates otherwise.

[0085] It should be further understood that the term "and / or" used in the specification and the appended claims of the present application means any combination of one or more of the associated listed items and all possible combinations, and includes these combinations, for example, A and / or B can represent three cases of A alone, A and B together, and B alone. In addition, the character " / " in the present application generally represents an "or" relationship between the front and rear associated objects.

[0086] It should be understood that although the terms first, second, third, etc. may be used in the embodiments of the present application to describe the preset ranges, etc., these preset ranges should not be limited to these terms. These terms are only used to distinguish the preset ranges from each other. For example, the first preset range can also be referred to as the second preset range, and similarly, the second preset range can also be referred to as the first preset range without departing from the scope of the embodiments of the present application.

[0087] Depending on the context, the word "if" as used herein can be interpreted to mean "when" or "while" or "in response to determining" or "in response to detecting". Similarly, the phrase "if it is determined" or "if (a stated condition or event) is detected" can be interpreted to mean "when it is determined" or "in response to determining" or "when (a stated condition or event) is detected" or "in response to detecting (a stated condition or event)".

[0088] Various structural schematic diagrams according to disclosed embodiments of the present application are shown in the drawings. These diagrams are not drawn to scale, in which certain details are exaggerated for clarity and others omitted. The shapes and relative sizes of the various regions, layers, and their relative positions illustrated in the drawings are exemplary only and can deviate in actual implementation due to manufacturing tolerances or technical limitations, and regions / layers with different shapes, sizes, and relative positions can be additionally designed according to actual needs by those skilled in the art.

[0089] The present application provides a method for improving the resolution of middle-deep seismic data. Based on the Cycle-GAN network structure, a 3D Cycle-GAN architecture is proposed. The deep and shallow layers of the same work area have very similar geological structure characteristics, and due to the unpaired characteristics of deep and shallow data, a deep learning method is used to learn the distribution and mapping relationship of deep and shallow data using a weakly supervised learning method. At the same time, data augmentation, automatic gain control and unique data splicing processing are added to the three-dimensional seismic data, so as to effectively improve the resolution of deep seismic data.

[0090] Referring to Figure 18 The present application provides a method for improving the resolution of middle-deep seismic data. Based on the Cycle-GAN network structure, a 3D Cycle-GAN architecture is proposed. The deep and shallow layers of the same work area have very similar geological structure characteristics, and due to the unpaired characteristics of deep and shallow data, a deep learning method is used to learn the distribution and mapping relationship of deep and shallow data using a weakly supervised learning method. At the same time, data augmentation, automatic gain control and unique data splicing processing are added to the three-dimensional seismic data, so as to effectively improve the resolution of deep seismic data.

[0091] S1, pre-processing the three-dimensional seismic data of the same work area, including deep and shallow layer division based on time window; combining with the automatic gain control function, constructing energy distribution more uniform, facilitating network learning input seismic data; introducing data augmentation method, constructing larger data set; introducing data splicing method, enhancing the overall integrity of three-dimensional seismic data;

[0092] Referring to Figure 4, in actual seismic data, the reflection coefficients of different strata can vary significantly, resulting in seismic signals received by the detectors varying over a large range. During training, only a small sample set is selected each time and fed into the network for training. If the selected reflection signals are too weak, the error obtained using loss functions such as the L2 norm is very small, making it difficult for the network to learn the characteristics of weak reflection signals. To adjust the signal amplitude to a suitable range, attenuating large signals and amplifying small signals, an automatic gain control (AGC) algorithm can be used for signal preprocessing. First, the signal amplitude is adjusted to a range suitable for network training. After network processing, the AGC inverse transform is used to recover the signal amplitude. The AGC method helps the network capture the characteristics of weak reflection signals. For the input seismic data D... in Every dish i The AGC result is calculated using the following formula:

[0093]

[0094] Where, d′ i For d i The AGC results, W triangle (L) represents a triangular window of length L. The results of AGC processing for actual seismic data are shown in Figure 4. The energy of the original data is concentrated in the time window range of t = 1.2 to 1.6 s, while the energy in the time window range of t = 0.5 to 1.2 s is much lower than that of the original data. However, the energy distribution of the data after AGC processing is more uniform, which facilitates the network learning the characteristics of the input seismic data.

[0095] Since 3D networks have more parameters than 2D networks for the same convolutional neural network depth, a large amount of training data is needed to prevent overfitting due to insufficient training data. However, due to limitations in exploration techniques, older datasets used for high-resolution processing are often small in size, further exacerbating the problem of insufficient training data. For training data D with a 3D structure, this invention uses the following three-point data augmentation method:

[0096] 1) Flip the Inline and Crossline spatial dimensions.

[0097]

[0098] in N i and N x The length of the data in the Inline and Crossline.

[0099] 2) Swap the Inline and Crossline spatial dimensions.

[0100]

[0101] 3) Data overall negation

[0102]

[0103] wherein, is the augmented data, and each brace in the formula indicates that one of them is randomly selected according to equal probability. Through this data augmentation method, the number of training samples is effectively expanded. Input data and labels can use different data augmentation methods, so that the network can access the entire data set from multiple angles, providing greater flexibility compared to paired training data augmentation methods.

[0104] Please refer to Figure 5 Due to the GPU memory limit, the network cannot input the entire three-dimensional seismic data body at once, so when processing at high resolution, the data needs to be subdivided into fixed-size blocks. The network processes each block of data and then splices it back to the original size, so the selected boundary processing method is crucial to the quality of the output data. The traditional direct splicing strategy may introduce discontinuity at the boundary of the processed block. To address this challenge and enhance the overall integrity of the seismic data, the present application proposes an improved boundary processing strategy.

[0105] When the input data is blocked, 50% overlap is set for the data. After high-resolution processing is completed, the overlapping part between the high-resolution data blocks is multiplied by the appropriate weight value through the windowing method, and it is spliced back to the original shape, and the sum of the weight values of the data at each position after splicing is ensured to be 1. The present application extends the triangular window to three-dimensional space, making it suitable for the splicing of three-dimensional data. Please refer to Fig. 6 for a three-dimensional window function diagram, and the expression of the window function is as follows:

[0106]

[0107] wherein, the length of the window function is 2N, i, j, k are the indexes of the three dimensions respectively, 0≤i,j,k≤2N,

[0108]

[0109] By adopting the design of overlapping sampling and data windowing technology, the effect of the final output of high-resolution data in three-dimensional seismic data processing is ensured.

[0110] S2, a 3D Cycle-GAN network architecture is proposed to extract the features of shallow high-resolution seismic data;

[0111] Please refer to Figure 1The application provides a three-dimensional seismic data network training flowchart. Due to the unpaired characteristics of deep and shallow data, it is difficult to extract paired labels for training using a strong supervision method. The application uses a weak supervision learning method to learn the distribution and mapping relationship of deep and shallow data. In order to improve the resolution of deep data based on the resolution of shallow data, Cycle-GAN is used as a backbone network in this chapter, and unpaired deep and shallow data are used for training. Since there is a complex corresponding relationship between shallow and deep data in the same data acquisition, in order to solve the problem of neural network learning difficulty, the application introduces bidirectional cycle consistency. Since only the forward cycle is used, the network performance is unstable, especially when facing some complex data sets, the HR output often generates misleading features. After introducing bidirectional cycle, the network training flowchart is as shown in Figure 1 , and the specific training process includes forward cycle (LR→HR→LR) and reverse cycle (HR→LR→HR).

[0112] Such improvement enables the network in this chapter to more comprehensively understand the subtle correlation in seismic data. Bidirectional cycle uses two sets of generators (forward generator G LR2HR and reverse generator G HR2LR ) and two sets of discriminators (LR discriminator D LR and HR discriminator D HR ). The forward and reverse networks of the two are of the same architecture, share the same architecture and are jointly trained, but their parameters are updated independently.

[0113] Wherein, the generator G LR2HR maps the LR data to the HR data, and the generator G HR2LR maps the HR data to the LR data; the discriminators D LR and HR are used to determine whether the input data is real LR or HR data. The consistent architecture of the forward and reverse components ensures a unified learning process. The bidirectional cycle significantly improves the stability in training, reduces the artifacts of the output data, and improves the overall quality of the generated HR seismic data due to the introduction of additional constraints.

[0114] Specifically, for the forward cycle, the forward generator G LR2HR learns the mapping from deep LR seismic data to shallow HR data. The input is deep LR seismic data, and the label is the corresponding shallow HR seismic data. It should be noted that the LR data and the HR data used here are three-dimensional data bodies. The output of G LR2HR is the high-resolution result χ HR of the deep data. The reverse generator G HR2LR simulates G LR2HRThe reverse process is essentially mapping HR seismic data back to the original LR data. G LR2HR Output χ HR Input reverse generator G HR2LR To obtain the raw LR data Approximate results By adding to the loss function and By imposing consistency constraints, it is possible to ensure G LR2HR High-resolution reversibility. HR discriminator D HR Used to evaluate and improve G LR2HR The generation performance. It distinguishes between G and other components. LR2HR The generated HR data χ HR and real HR data with tags The probability that the generated HR data and the target HR data are not true labeled data is estimated separately to guide the training of the generator. This adversarial training process forces G... LR2HR By synthesizing more realistic HR data, and ultimately enabling the generation of HR data closely related to the target HR data, a meaningful resolution improvement was achieved in LR seismic data. The reverse loop is the opposite of the forward loop, swapping all shallow and deep data from the forward loop process. The forward generator G... LR2HR and the reverse generator G HR2LR Interchange, LR discriminator D LR and HR discriminator D HR Interchange is a reverse loop process.

[0115] Please see Figure 2 This invention presents the generator part of the proposed 3D Cycle-GAN network architecture. For the generator, data is first input through the first convolutional layer, then passes through 16 contiguous residual blocks in the main body. Each residual block contains two convolutional layers and residual connections, with a larger cross-layer connection outside the residual block group. Finally, the data passes through several convolutional layers and is output through the tanh activation function. To prevent the loss of useful information, pooling layers are not included in the generator network; residual blocks are used to enhance the performance of deep networks; batch normalization (BN) layers are not used because as the network deepens, especially when training GAN networks, BN layers may introduce artifacts, reducing the network's generalization ability. Removing BN layers helps improve the stability and consistency of training.

[0116] Please see Figure 3For the discriminator part in the 3D Cycle-GAN network architecture proposed by the present application, for the discriminator, the data first enters by the input three-dimensional convolutional layer, passes through 5 convolutional blocks, of which the first 2 are three-dimensional convolution, and the last 3 are two-dimensional convolution. Each convolutional block contains a three-dimensional convolutional layer, a batch normalization layer and a Leaky ReLU activation function (α=0.2);

[0117] Then the data passes through the Sigmoid activation function, and outputs a probability between 0 and 1. Unlike traditional discriminators, the discriminator of the present application adopts a PatchGAN architecture, replacing the fully connected layer in the last layer with a convolutional layer. This allows the discriminator to focus on the local area of the data, which is crucial for high-resolution processing of seismic data. In addition, this architecture reduces the number of model parameters and enables the discriminator to converge quickly, resulting in data with higher resolution and higher accuracy.

[0118] S3, introduce a pair of loss functions combined by generator loss and discriminator loss, reduce prediction error and prevent overfitting;

[0119] Loss functions are crucial in deep learning as they determine how to punish network outputs during weight optimization. Choosing the right loss function is very important to reduce prediction error and prevent overfitting. To solve the problem of improving seismic resolution, the present application designs a pair of loss functions according to the characteristics of seismic data. The loss function can be decomposed into two components: generator loss and discriminator loss.

[0120] 1) Generator loss

[0121] The forward and backward generators have the same network architecture and use the same loss function. The generator loss is:

[0122] L Gen generator GAN + L cyc discriminator idt + L TV cycle

[0123] where w1, w2, w3 and w4 represent the weights assigned to each loss function, L GAN is the adversarial loss, L cyc is the cycle consistency loss, L idt is the identity loss, and L TV is the total variation loss. In an ideal case, the individual contributions of the four loss functions to the final loss function are synergistic, working together to improve the resolution of seismic data.

[0124] The adversarial loss is used to ensure that G LR2HR is indistinguishable from G HR2LRThe output can deceive D as much as possible. HR and D LR When χ HR D HR The more likely the data is to be classified as genuine HR data, the smaller the value of the adversarial loss. Negative log-likelihood is used as the adversarial loss L. GAN Its definition is as follows:

[0125]

[0126] Where N is the number of training samples in one round of training.

[0127] The cycle consistency loss aims to ensure that the processes of the high-resolution forward generator and backward generator are reversible, thereby ensuring the consistency of data before and after passing through the generator; the cycle consistency loss is defined as follows:

[0128]

[0129] L1 loss and multi-scale structural similarity (MS-SSIM) loss are used. MS-SSIM The comprehensive method defines the identity loss L idt As shown below:

[0130]

[0131] in, The HR data y is low-pass filtered to obtain the corresponding degraded data. L1 loss measures the difference between two seismic images, given two 3D seismic data sets. and The formula for calculating L1 loss is as follows:

[0132]

[0133] For two-dimensional seismic profiles U and V extracted from three-dimensional data, different multi-scale structural self-similarity is combined to calculate MS-SSIM, as shown below:

[0134]

[0135] Where M represents the maximum scale, and l(U,V), c(U,V), and s(U,V) represent the multi-scale self-similarity of U and V in terms of brightness, contrast, and structure; here, α, β, and γ are the weights of each term. The definitions of l(U,V), c(U,V), and s(U,V) are as follows:

[0136]

[0137] Where, μ and σ represent the mean and standard deviation, respectively; C1, C2 and C3 are used to stabilize the division to prevent the denominator from being zero. When the Inline length of the input data is less than the Crossline length, the MS-SSIM loss is defined as follows:

[0138]

[0139] Where, and represent the Inline and Crossline two-dimensional sections randomly selected from the three-dimensional data and r i is the Inline index randomly selected. When the Inline length of the input data is greater than the Crossline length, similarly, the Crossline two-dimensional section is randomly selected and are also calculated according to the above formula.

[0140] For the output data D of the forward generator or the backward generator, L TV is defined as follows:

[0141]

[0142] Where, and are the gradients of D in three dimensions, respectively.

[0143] Following the core idea of GAN, the present application uses the discriminator D HR and D LR to conduct adversarial training on the generator G LR2HR and G HR2LR .

[0144] Specifically, the loss function of the discriminator is used to ensure that it can correctly determine the difference between true and false data; when it judges that the input data is real data, it outputs 1, and when it judges that it is the output of the generator, it outputs 0.

[0145] The discriminator loss is also defined by the negative log-likelihood as follows:

[0146]

[0147] S4, the corresponding shallow low-resolution data obtained by low-pass filtering the shallow and deep seismic data and the shallow high-resolution data is input into the network for training, and after the training is completed, the deep low-resolution data is input into the network, and the output thereof is subjected to AGC inverse transformation, and finally the resolution of the deep seismic data is improved.

[0148] After the data preprocessing and model training are completed, the deep low-resolution 3D seismic data is input into the network, and the AGC inverse transform is performed on the output of the network to realize the resolution improvement of the deep seismic data.

[0149] Those skilled in the art can understand that various aspects of the present application can be implemented as a system, a method or a program product. Therefore, various aspects of the present application can be embodied as a complete hardware implementation, a complete software implementation (including firmware, microcode, etc.), or a combination of hardware and software aspects, which can be collectively referred to as "circuitry", "module" or "platform" here.

[0150] In another embodiment of the present application, a system for improving the resolution of middle-deep seismic data is provided, which can be used to implement the method for improving the resolution of middle-deep seismic data. Specifically, the system for improving the resolution of middle-deep seismic data comprises a data module, a network module, a function module and an improvement module.

[0151] The data module pre-processes 3D seismic data of the same work area, constructs seismic data in combination with an automatic gain control function, and supplements the seismic data by introducing a data augmentation method and a data splicing method.

[0152] The network module constructs a 3D Cycle-GAN network architecture for extracting features of shallow high-resolution seismic data.

[0153] The function module introduces a pair of loss functions composed of generator loss and discriminator loss into the 3D Cycle-GAN network.

[0154] The improvement module inputs the shallow, deep seismic data and the corresponding shallow low-resolution data obtained by low-pass filtering the shallow high-resolution data into the 3D Cycle-GAN network for training, inputs the deep low-resolution data into the trained 3D Cycle-GAN network, and performs AGC inverse transform on the output to realize the resolution improvement of the deep seismic data.

[0155] In still another embodiment of the present application, a terminal device is provided, which comprises a processor and a memory, the memory is configured to store a computer program, the computer program comprises program instructions, and the processor is configured to execute the program instructions stored in the computer storage medium. The processor can be a central processing unit (CPU), and can also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computing core and control core of the terminal, and are suitable for implementing one or more instructions, and are specifically suitable for loading and executing one or more instructions to implement a corresponding method flow or a corresponding function; the processor in the embodiments of the present application can be used for the operation of the method for improving the resolution of medium-deep seismic data, comprising:

[0156] Pretreat the three-dimensional seismic data of the same work area; construct the seismic data in combination with an automatic gain control function, and introduce a data augmentation method and a data splicing method to supplement the seismic data; construct a 3D Cycle-GAN network architecture for extracting shallow high-resolution seismic data features; introduce a pair of loss functions composed of generator loss and discriminator loss into the 3D Cycle-GAN network; input the shallow, deep seismic data and the corresponding shallow low-resolution data obtained by low-pass filtering of the shallow high-resolution data into the 3D Cycle-GAN network for training, input the low-resolution data of the deep layer into the trained 3D Cycle-GAN network, and perform AGC inverse transformation on the output to improve the resolution of the deep seismic data.

[0157] In still another embodiment of the present application, a computer readable storage medium, specifically a computer readable storage medium (Memory) is also provided. The computer readable storage medium is a memory device in the terminal equipment, for storing programs and data. It can be understood that the computer readable storage medium herein can include the built-in storage medium in the terminal equipment, and of course can also include the expansion storage medium supported by the terminal equipment, and can be any tangible medium containing or storing programs, which can be used by or in combination with the instruction execution system, device or apparatus. The computer readable storage medium provides a storage space, which stores the operating system of the terminal. Moreover, one or more instructions suitable for being loaded and executed by the processor are also stored in the storage space, and the instructions can be one or more computer programs (including program codes). It should be noted that more specific examples (non-exhaustive list) of the computer readable storage medium include: an electrical connection with one or more conductive wires, a portable disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above.

[0158] The computer readable storage medium also includes a data signal carried in the baseband or as a part of a carrier wave, in which the readable program code is borne. Such a propagated data signal can take any of a variety of forms, including but not limited to electro-magnetic, optical, or any suitable combination thereof. The computer readable storage medium can also be any medium that can be read by the instruction execution system, device or apparatus, and can be used or in combination with the same. The program code contained in the computer readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0159] The program code for carrying out operations of the present application can be written in any combination of one or more programming languages, including an object oriented programming language such as Java, C++, etc., and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The program code can execute entirely on the user's computing device, partly on the user's device, as a stand-alone software package, partly on the user's computing device and partly on a remote computing device or entirely on the remote computing device or server. In the latter scenario, the remote computing device can be connected to the user's computing device through any kind of network, including a local area network (LAN) or a wide area network (WAN), or the connection can be made to an external computing device, such as through the Internet using an Internet service provider.

[0160] The one or more instructions stored in the computer-readable storage medium can be loaded and executed by the processor to implement the corresponding steps of the method for improving resolution of intermediate-deep seismic data in the above embodiments; the one or more instructions stored in the computer-readable storage medium are loaded and executed by the processor to implement the following steps:

[0161] The three-dimensional seismic data of the same work area is preprocessed; the seismic data is constructed in combination with an automatic gain control function, and a data augmentation method and a data splicing method are introduced to supplement the seismic data; a 3D Cycle-GAN network architecture is constructed to extract shallow high-resolution seismic data features; a pair of loss functions combined by generator loss and discriminator loss are introduced into the 3D Cycle-GAN network; the shallow, deep seismic data and the corresponding shallow low-resolution data obtained by low-pass filtering of the shallow high-resolution data are input into the 3D Cycle-GAN network for training, the deep low-resolution data is input into the trained 3D Cycle-GAN network, and the output is subjected to AGC inverse transformation, thereby improving the resolution of deep seismic data.

[0162] Please refer to Figure 19 , the terminal device is a computer device, the computer device 60 of the embodiment includes a processor 61, a memory 62, and a computer program 63 stored in the memory 62 and executable on the processor 61, and the computer program 63 is executed by the processor 61 to implement the method for improving resolution of intermediate-deep seismic data in the embodiment, to avoid repetition, which will not be described here. Alternatively, the computer program 63 is executed by the processor 61 to implement the functions of each model / unit in the system for improving resolution of deep seismic data in the embodiment, to avoid repetition, which will not be described here.

[0163] The computer device 60 can be a desktop computer, a notebook computer, a palm computer, and a cloud server, etc. The computer device 60 can include, but is not limited to, a processor 61 and a memory 62. Those skilled in the art can understand that Figure 19 The computer device 60 is only an example and does not constitute a limitation on the computer device 60, and can include more or fewer components than the illustration, or combine certain components, or different components, for example, the computer device can also include an input / output device, a network access device, a bus, etc.

[0164] The processor 61 can be a central processing unit (CPU), and can also be other general-purpose processors, central processing units, graphics processing units, digital signal processors (DSPs), application specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs) or other programmable logic devices, discrete gates or transistor logic components, quantum computing-based data processing logic components, discrete hardware components, etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor.

[0165] The memory 62 can be an internal storage unit of the computer device 60, such as a hard disk or a memory of the computer device 60. The memory 62 can also be an external storage device of the computer device 60, such as a plug-in hard disk, a smart media card (SMC), a secure digital (SD) card, a flash card, etc. equipped on the computer device 60.

[0166] Further, the memory 62 can include both an internal storage unit and an external storage device of the computer device 60. The memory 62 is used to store computer programs and other programs and data required by the computer device. The memory 62 can also be used to temporarily store data that has been output or will be output.

[0167] Any reference to storage, databases or other media used to store data in the embodiments provided herein is intended to include at least one of volatile and non-volatile storage. Non-volatile storage can include, for example, optical, floppy disks, hard disks, or solid state drives. Volatile storage can include, for example, random access memory (RAM). A basic input / output system (BIOS), containing the basic routines that help to transfer information between elements within the electronic device, such as during startup, can typically be stored in non-volatile memory. By way of illustration, and not limitation, a basic input / output system based on the BIOS, can include a BIOS, a unified extensible firmware interface (UEFI), or the like, including without limitation basic input / output system software stored in nonvolatile memory that

[0168] The databases involved in the embodiments provided herein can include at least one of a relational database and a non-relational database. The non-relational database can include a distributed database based on a blockchain, and the like, without being limited thereto. The processor involved in the embodiments provided herein can be a general processor, a central processing unit, a graphics processing unit, a digital signal processor, a programmable logic device, a data processing logic device based on quantum computing, and the like, without being limited thereto.

[0169] Referring to Figure 20 , the terminal device 600 is an electronic device, which is manifested in the form of a general computing device. The components of the electronic device can include, but are not limited to, at least one processing unit 610, at least one storage unit 620, a bus 630 connecting different platform components including the storage unit 620 and the processing unit 610, a display unit 640, and the like.

[0170] The storage unit stores program codes, which can be executed by the processing unit 610, so that the processing unit 610 performs the steps according to various exemplary embodiments of the present application described in the method part of the present specification. For example, the processing unit 610 can perform the steps as shown in Figure 18 .

[0171] The storage unit 620 can include a readable medium in the form of volatile storage such as random access memory (RAM) 6201 and / or cache memory 6202, and also can include a non-volatile storage such as read only memory (ROM) 6203.

[0172] The storage unit 620 also can include a program / utility 6204 having a set of programs / modules 6205, including an operating system, one or more application programs, other program modules, and program data, each of which can implement aspects of a network environment, as each of these example or some combination thereof.

[0173] The bus 630 can represent one or more of several types of bus structures, including a storage bus or bus controller, a peripheral bus, a graphics bus, a processor or local bus using any of a variety of bus architectures.

[0174] The electronic device 600 also can communicate with one or more external devices 700 such as a keyboard or pointing device, a Bluetooth device, etc.; other devices that enable a user to interact with the electronic device 600; and / or one or more devices that enable the electronic device 600 to communicate with one or more other computing devices. Such communication can occur via an input / output (I / O) interface 650. Still yet, the electronic device 600 can communicate with one or more networks such as a local area network (LAN), a general wide area network (WAN), and / or a public network such as the Internet, via a network adapter 660. The network adapter 660 can be communicatively coupled to the other components of the electronic device 600 via the bus 630. It should be appreciated that the electronic device 600 can be a part of one or more networks, such as virtual networks, which further can include more than one network.

[0175] In order to make the purposes, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described clearly and completely below with reference to the drawings in the embodiments of the present application. Obviously, the described embodiments are some embodiments of the present application but not all embodiments of the present application. The components of the embodiments of the present application described and shown in the drawings herein can be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of the present application provided in the drawings is not intended to limit the scope of the claimed present application but only represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present application belong to the scope of protection of the present application.

[0176] Referring to Figure 7 , in order to evaluate the effectiveness of the method proposed in the present application on the simulated seismic data in the present experiment, the present application evaluates the results from two aspects: profile results and multi-channel average amplitude spectrum, to verify the effectiveness and superiority of the network method.

[0177] Referring to Fig. 8, the results of the simulated seismic data in the Inline profile (Inline = 200) are shown, wherein Fig. 8(a) and Fig. 8(b) are the original data of the deep low-resolution and the corresponding high-resolution region respectively, Fig. 8(c) is the result of the deep low-resolution data after spectral whitening processing, and Fig. 8(d) is the result of the deep low-resolution data after network high-resolution processing. From the comparison of the experimental results of the two groups, it can be seen that the resolution of the deep part data after network processing is effectively improved, and the fault information in the original data is effectively preserved, which is closer to the corresponding high-resolution data compared with the spectral whitening method, fully illustrating the effectiveness of the method of the present application for correctly improving the resolution of seismic data.

[0178] Referring to Figure 9 , the results of the simulated seismic data in the Crossline profile (Crossline = 200) are shown, wherein Figure 9 (a) and Figure 9 (b) are the original data of the deep low-resolution and the corresponding high-resolution region respectively, Figure 9 (c) is the result of the deep low-resolution data after spectral whitening processing, Figure 9 (d) is the result of the deep low-resolution data after network high-resolution processing. From the comparison of the experimental results of the two groups, it can be seen that the spectral whitening method cannot clearly restore the event at the fault position, and almost no event structure can be seen at this position; compared with the spectral whitening method, the network output result can more effectively restore the event of the original data, which illustrates that the method of the present application can more effectively improve the resolution of the seismic signal compared with the traditional method.

[0179] Referring to Figure 10 , the results of the simulated seismic data in the isochronous slice (t = 450 ms) are shown, wherein Figure 10 (a) and Figure 10 (b) are the original data of the deep low-resolution and the corresponding high-resolution region respectively, Figure 10 (c) is the result of the deep low-resolution data after spectral whitening processing, Figure 10(d) is the result of deep low-resolution data processed by the network. From the comparison of the experimental results of the two groups, it can be seen that the network output result is more similar to the corresponding high-resolution data part under the premise of completely retaining the fault information, which shows that compared with the traditional method, the method of the application can more effectively improve the resolution of the seismic signal.

[0180] Please refer to Figure 11 , which shows the comparison of multi-channel amplitude spectrum of simulated seismic data using spectral whitening method and network method processing result, wherein the solid line is the original low-resolution data, the dotted line is the corresponding high-resolution data, the dotted line is the result of using the spectral whitening method, and the dot-dash line is the high-resolution result of using the network processing. From the spectrum, it can be seen that both the network method and the spectral whitening method can effectively improve the high-frequency component of the data, and the bandwidth expansion of the two is basically the same. But compared with the spectral whitening method, using the network method to process the data, the spectrum obtained is closer to the original data at low frequency, especially at low frequency below 20Hz, the network method better preserves the low-frequency information of the deep data, while the spectral whitening method has the problem of over-high low frequency.

[0181] Compared with synthetic data, actual data presents greater complexity and is significantly affected by noise. In order to further evaluate the effectiveness of the high-resolution method proposed in the application, the spectral whitening method is selected as the benchmark method, and a three-dimensional post-stack seismic data set obtained in a certain work area is used for comparison. The Inline and Crossline sizes of the data are 931 and 2471 respectively, and the time sampling rate of the data is 2ms. Select the 600x1800 part near the center with relatively high signal-to-noise ratio as the training data, and test the entire data after training.

[0182] Please refer to Figure 12 , the shallow part is selected from the time window of 0.6-1.6s, and the deep part is selected from the relatively high signal-to-noise ratio area of 1.9-2.9s. After training, the network is applied to the deep data to enhance the resolution. Figure 12 (a) shows the Crossline profile of actual seismic, Figure 12 (b) shows the normalized multi-channel amplitude spectrum of deep and shallow parts.

[0183] Referring to FIG. 14, the results of actual seismic data in Inline profile (Inline = 8 km) are shown. FIG. 14(a) shows the original deep LR data, which is distorted by some noise, resulting in discontinuity. FIG. 14(b) is the enhanced result after spectral whitening, and FIG. 14(c) is the result after network processing. The results show that the resolution of the original seismic data has been significantly improved, the reflection is more refined, and the geological structure is more refined. In addition, it is observed that the two methods have great similarity in structure, which confirms the reliability of our method.

[0184] Referring to FIG. 15, the results of actual seismic data in Crossline profile (Crossline = 6 km) are shown. FIG. 15(a) is the original Crossline profile, which shows relatively wide features due to the lack of high-frequency energy. Compared with the original profile, both the spectral whitening technique in FIG. 15(b) and our method in FIG. 15(c) have significantly improved resolution and effectively restored valuable high-frequency information. However, spectral whitening has difficulty in maintaining spatial continuity and energy consistency, resulting in increased noise. In contrast, our method significantly enhances lateral continuity and improves spatial energy consistency. Even under noisy conditions, the network still maintains excellent enhancement performance, demonstrating its strong anti-interference ability to noise, as shown. This enhancement is crucial for reliable stratum tracking and detailed stratum interpretation.

[0185] Referring to FIG. 13, the normalized amplitude spectrum is shown, where FIG. 13(a) is the normalized amplitude spectrum of the Inline profile in FIG. 14, and FIG. 13(b) is the normalized amplitude spectrum of the Crossline profile in FIG. 15. The original LR data, the HR result after spectral whitening, and the method of the present application are represented respectively. Both our method and spectral whitening effectively enhance the high-frequency components, achieving comparable bandwidth expansion. They are closely related to the original LR data, preserving the inherent characteristics while improving resolution. However, compared with the spectral whitening method, the data processed using the present method corresponds better to the original data. Spectral whitening shows a significant deviation from the original spectrum beyond 20 Hz and cannot follow the original frequency conversion around 35 Hz, losing signal integrity in the mid-frequency range, highlighting the advantages of the present method in smoothly expanding bandwidth, accurately enhancing high-frequency components, and maintaining overall data integrity, making it more effective in improving seismic data resolution.

[0186] Referring to FIGS. 16 and 17, the results of actual seismic data at isochronous slices (t = 2284 ms) and (t = 2504 ms) are shown. The resolution enhancement capability of the method of the present application is further emphasized. The improvement of deep time slices after network processing is more obvious than spectral whitening. As shown in the boxes, the method of the present application not only achieves higher clarity in delineating geological structures, but also reveals clearer details that spectral whitening cannot capture. The spatial relationship between different strata can be more accurately depicted, and the geological interface is more clearly delineated. This detailed representation not only helps to more accurately interpret the subsurface geology, but also greatly enhances our understanding of complex stratum relationships, thus providing an important tool for advancing seismic exploration and analysis.

[0187] In summary, the method and system for improving the resolution of middle-deep seismic data of the present application designs a three-dimensional cyclic generative adversarial network structure, and corresponding loss function, data augmentation and data splicing method, and achieves the purpose of improving the resolution of deep seismic data. The feasibility and effectiveness of the method of the present application are also proved in the experiments of simulated seismic data and actual seismic data. Compared with the traditional spectral whitening method, the continuity of the event and the high-frequency details are improved.

[0188] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above functional units and modules is exemplified, and in actual application, the above functions can be completed by different functional units and modules according to needs, that is, the internal structure of the device is divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated in one processing unit, or each unit can exist physically, or two or more units can be integrated in one unit. The above integrated unit can be realized in the form of hardware or in the form of software functional unit. In addition, the specific names of each functional unit and module are only for easy distinction, and do not limit the protection scope of the present application. The specific working process of the units and modules in the above system can refer to the corresponding process in the foregoing method embodiments, which will not be repeated here.

[0189] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in detail in a certain embodiment can be referred to the relevant description of other embodiments.

[0190] Those skilled in the art can understand that the units and algorithm steps of each example described in combination with the embodiments disclosed in the present application can be realized in electronic hardware or in combination of computer software and electronic hardware. Whether the functions are realized in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to realize the described functions for each specific application, but such implementation should not be considered beyond the scope of the present application.

[0191] In the embodiments provided by the present application, it should be understood that the disclosed apparatus / terminal and method can be implemented in other ways. For example, the apparatus / terminal embodiments described above are merely schematic. The division of the modules or units is merely a logical function division, and there can be another division manner in actual implementation. For example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the displayed or discussed coupling or direct coupling or communication connection between the units can be indirect coupling or communication connection through some interface, device or unit, and can be electrical, mechanical or in other forms.

[0192] The units described as separate components can or can not be physically separate, and the components shown as units can or can not be physical units, i.e. they can be located in one place, or distributed on a plurality of network units. Part or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0193] In addition, each functional unit in each embodiment of the present application can be integrated into a processing unit, or each unit can exist physically independently, or two or more units can be integrated into one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0194] The integrated module / unit, if realized in the form of a software function unit and sold or used as an independent product, can be stored in a computer-readable storage medium. Based on such understanding, all or part of the processes in the above-mentioned embodiment methods can also be completed by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and the computer program can implement the steps of each method embodiment when executed by a processor. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or some intermediate forms, etc. The computer-readable medium can include any entity or device capable of carrying the computer program code, recording medium, U disk, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electrical carrier signal, telecommunication signal, and software distribution medium, etc. It should be noted that the computer-readable medium can include or exclude contents according to the requirements of legislation and patent practice in the jurisdiction, for example, in some jurisdictions, according to legislation and patent practice, the computer-readable medium does not include electrical carrier signals and telecommunication signals.

[0195] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present application. It should be understood that each flow and / or block in the flowcharts and / or block diagrams, and the combination of flows 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 apparatus to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing apparatus generate a means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The means for implementing the functions specified in the flowcharts and / or block diagrams.

[0196] These computer program instructions can also be stored in a computer-readable memory capable of instructing the computer or other programmable data processing apparatus to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including instruction means, which implements the functions specified in the flowcharts and / or block diagrams. Figure 1 The means for implementing the functions specified in the flowcharts and / or block diagrams. Figure 1 The means for implementing the functions specified in the flowcharts and / or block diagrams.

[0197] These computer program instructions can also be loaded into a computer or other programmable data processing devices, so that a series of operational steps are performed on the computer or other programmable data processing devices to generate a computer implemented process, so that the instructions executed on the computer or other programmable data processing devices provide a process for implementing the flowchart Figure 1 one flow or multiple flows and / or blocks Figure 1 one block or multiple blocks.

[0198] The above is only to illustrate the technical idea of the present application, and cannot limit the protection scope of the present application. Any modification made according to the technical idea of the present application on the basis of the technical scheme falls within the protection scope of the claims of the present application.

Claims

1. A method for improving resolution of intermediate-depth seismic data, characterized in that, The method comprises the following steps: S1, preprocessing three-dimensional seismic data of the same work area; combining an automatic gain control function to construct seismic data, and introducing a data augmentation method and a data splicing method to supplement the seismic data; S2, constructing a 3D Cycle-GAN network architecture for extracting shallow high-resolution seismic data features, the 3D Cycle-GAN network architecture comprises a forward cycle and a reverse cycle, the forward cycle and the reverse cycle adopt two sets of generators and two sets of discriminators, the forward and reverse networks of the two sets of generators and discriminators maintain the same architecture, share the same architecture and are jointly trained, and the parameters are updated independently; The generator comprises a forward generator and a backward generator , the forward generator maps data to data, the backward generator maps data to data; The discriminator comprises The discriminator And The discriminator , The discriminator And The discriminator Respectively used for determining the data input therein as real Or Data; Forward generator And backward generator For three-dimensional networks, while processing two-dimensional data and three-dimensional data, and using the same network architecture; The data is first input by the first convolutional layer, passes through 16 residual blocks connected at the beginning and end in the main part, each residual block contains two convolutional layers and a residual connection, and there is a cross-layer connection outside the residual block group; Ultimately, the data is passed through multiple convolutional layers, through an activation function output; discriminator and discriminator are two-dimensional networks, employing the same network architecture; The data is first input by the convolutional layer, and then passes through five convolutional blocks; Each convolutional block contains a convolutional layer, a batch normalization layer and a Leaky ReLU activation function; Then the data passes through a Sigmoid activation function to output a probability between 0 and 1; The discriminator adopts a PatchGAN architecture, and the last full connection layer is replaced by a convolutional layer; S3, introducing a pair of loss functions composed of generator loss and discriminator loss into the 3D Cycle-GAN network obtained in step S2; S4, inputting the shallow, deep seismic data obtained in step S1 and the corresponding shallow low-resolution data obtained by low-pass filtering the shallow high-resolution data into the 3D Cycle-GAN network obtained in step S3 for training, inputting the deep low-resolution data into the trained 3D Cycle-GAN network, and performing AGC inverse transformation on the output to improve the resolution of the deep seismic data.

2. The method for improving resolution of middle-deep seismic data according to claim 1, characterized in that, In step S1, for each trace of the input seismic data , the result of pre-processing the signal using an automatic gain control algorithm is as follows: ​ wherein is a triangular window of length is a triangular window of length 3. The method for improving resolution of middle-deep seismic data according to claim 2, characterized in that, The data augmentation method specifically comprises: Flip both Inline and Crossline dimensions, swap both Inline and Crossline dimensions, negate the data as a whole, and obtain the augmented data Each curly brace indicates that one of the items is randomly selected with equal probability.

4. The method for improving resolution of middle-deep seismic data according to claim 2, characterized in that, The data splicing method specifically comprises: When the input data is blocked, 50% overlap is set for the data; After the high-resolution processing is completed, the overlapping parts between the high-resolution data blocks are multiplied by a set weight through windowing, and spliced back to the original shape, so that the sum of the weights of the data at each position after splicing is 1; The triangular window is expanded to three-dimensional space, and the expression of the window function is as follows: wherein the window function has a length of , , , are indices of the three dimensions, respectively, .

5. The method for improving resolution of middle-deep seismic data according to claim 1, characterized in that, The generator network does not include a pooling layer; a residual block is used, and a batch normalization layer is not used.

6. The method for improving resolution of middle-deep seismic data according to claim 1, characterized in that, In step S3, the loss function is decomposed into generator loss and discriminator loss; The generator loss is: wherein, , , and denote the weight assigned to each loss function, is the adversarial loss, is the cycle consistency loss, is the identity loss, is the total variation loss; The discriminator loss is: wherein, is the number of training samples for a network in a round of training, is is a discriminator, is is a discriminator, is low resolution data is high resolution data obtained by inputting the forward generator, is high resolution data obtained by inputting the forward generator, is low resolution data obtained by inputting the backward generator.

7. A system for improving resolution of intermediate-depth seismic data, the system comprising: It comprises: A data module for preprocessing three-dimensional seismic data of the same work area; Combining an automatic gain control function to construct seismic data, and introducing a data augmentation method and a data splicing method to supplement the seismic data; A network module for constructing a 3D Cycle-GAN network architecture for extracting shallow high-resolution seismic data features, the 3D Cycle-GAN network architecture comprises a forward cycle and a reverse cycle, the forward cycle and the reverse cycle adopt two sets of generators and two sets of discriminators, the forward and reverse networks of the two sets of generators and discriminators maintain the same architecture, share the same architecture and are jointly trained, and the parameters are updated independently; The generator comprises a forward generator and a backward generator The forward generator maps data to data, and the backward generator maps data to data; The discriminator comprises The discriminator and The discriminator , The discriminator and The discriminator is used for determining whether the data input therein is real or data; forward generator and a backward generator for three-dimensional networks, while handling both two-dimensional and three-dimensional data, and using the same network architecture; Data first input by the first convolutional layer, through the main part of 16 head-to-tail residual blocks, each residual block contains two convolutional layers and a residual connection, and there is a cross-layer connection outside the residual block group; Ultimately, the data is passed through multiple convolutional layers, through an activation function output; discriminator and discriminator are two-dimensional networks, employing the same network architecture; Data first input by the first convolutional layer, through the main part of 16 head-to-tail residual blocks, each residual block contains two convolutional layers and a residual connection, and there is a cross-layer connection outside the residual block group; Each convolutional block contains a convolutional layer, a batch normalization layer and a Leaky ReLU activation function. Then the data passes through the Sigmoid activation function, outputting a probability between 0 and 1. The discriminator adopts the PatchGAN architecture, replacing the fully connected layer in the last layer with a convolutional layer. The function module introduces a pair of loss functions composed of generator loss and discriminator loss into the 3D Cycle-GAN network. The improvement module inputs the corresponding shallow low-resolution data obtained by low-pass filtering of shallow, deep seismic data and shallow high-resolution data into the 3D Cycle-GAN network for training, inputs the deep low-resolution data into the trained 3D Cycle-GAN network, and performs AGC inverse transformation on the output to improve the resolution of deep seismic data.

Images