A method and system for suppressing seismic ambient noise
By constructing a deep learning-based background noise suppression network, the inefficiency and insufficient accuracy of conventional algorithms in handling complex seismic background noise are solved, achieving efficient and accurate noise suppression and improving the quality of seismic data.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- PETROCHINA CO LTD
- Filing Date
- 2024-03-20
- Publication Date
- 2026-07-21
AI Technical Summary
When dealing with complex and variable seismic background noise, conventional algorithms require mathematical assumptions and have low processing efficiency, which cannot accurately meet the actual data, resulting in poor denoising effects.
A deep learning-based background noise suppression network was constructed. Using training and validation datasets, deep learning techniques were employed to extract data features of earthquake background noise, and a multi-scale neural network structure was designed for noise suppression.
It improves the signal-to-noise ratio of seismic data, significantly enhances the accuracy of background noise suppression and processing efficiency, and provides high-quality data for subsequent seismic imaging and reservoir prediction.
Smart Images
Figure CN120703843B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of seismic data processing technology, and specifically relates to a method and system for suppressing seismic background noise. Background Technology
[0002] Noise suppression is a crucial foundational task in seismic data processing, especially in recent years. With increasingly complex exploration environments, the noise contained in acquired seismic data has become increasingly complex and variable. Influenced by multiple factors such as acquisition methods and surface environment, seismic data receiving instruments can pick up complex noise signals even before the first arrival wave arrives, continuing until signal reception ends. This type of noise varies in energy intensity, fails to reflect subsurface structural information, and severely interferes with effective reflected signals; we define this type of noise as background noise. Due to the irregularity of background noise in its morphology, frequency, and other characteristics, conventional processing techniques lack specific solutions. It is typically treated as random noise, anomalous amplitude, or linear noise, using techniques including bandpass filtering, fx-domain predictive filtering, wavelet transform, curvelet transform, FK filtering, FKK domain filtering, anomalous amplitude attenuation, KL transform, and tilt filtering.
[0003] While conventional noise suppression algorithms have a certain degree of noise reduction effect, they have the following two drawbacks when dealing with complex and diverse background noise:
[0004] 1. Conventional algorithms require abstracting information such as the propagation laws, physical characteristics, and morphological features of noise into mathematical language to separate signals from noise in a certain data space (such as the frequency domain, curve domain, etc.). These noise suppression methods based on mathematical models typically require theoretical assumptions during the mathematical abstraction process. While these assumptions can characterize certain features of seismic data to some extent, they cannot be precisely met in actual data processing (especially for low signal-to-noise ratio data), thus limiting the limitations of denoising algorithms.
[0005] 2. Conventional denoising algorithms are usually designed for a specific type of noise. In actual production, multiple denoising algorithms are needed to complete the task. Each denoising algorithm requires parameter tuning. The noise suppression process is highly subjective and inefficient when processing massive amounts of data. Summary of the Invention
[0006] To address the above problems, the present invention provides a method for suppressing seismic background noise, the method comprising:
[0007] Obtain a background noise dataset and divide the background noise dataset into a training dataset and a validation dataset;
[0008] An initial background noise suppression network architecture is constructed, and the training dataset is input into the initial background noise suppression network architecture for training to obtain the target background noise suppression network architecture.
[0009] The background noise suppression network architecture is used to process the background noise data to be tested, and the noise suppression data results are obtained.
[0010] Preferably, the background noise dataset consists of sample patches;
[0011] The sample sheets include background noise sample sheets and seismic signal sample sheets.
[0012] Preferably, the sample pieces are obtained through a combination of traversal acquisition and random acquisition.
[0013] Preferably, the training dataset includes a background noise-only training set, a seismic signal-only training set, and a noise and signal training set;
[0014] Among them, only the background noise training set consists of background noise sample patches;
[0015] The seismic signal training set consists only of seismic signal sample pieces;
[0016] The noise and signal training set consists of background noise sample pieces and seismic signal sample pieces.
[0017] Preferably, constructing the initial background noise network architecture includes:
[0018] Construct an initial background noise network architecture, and define the loss function and parameters to be optimized for the initial background noise network architecture.
[0019] Preferably, the initial background noise suppression network architecture consists of an input layer, a feature extraction layer, a data reconstruction layer, a downsampling layer, an upsampling layer, and an output layer.
[0020] Preferably, the training dataset is input into the initial background noise suppression network architecture for training, including:
[0021] Input the training dataset into the initial background noise suppression network architecture;
[0022] With the goal of minimizing the loss function, the parameters to be optimized are adjusted and determined.
[0023] Using the final determined parameters to be optimized as target parameters, the target background noise suppression network architecture is obtained.
[0024] Preferably, after obtaining the target background noise suppression network architecture, the process includes:
[0025] The validation dataset is input into the target background noise suppression network architecture to verify the accuracy of the background noise suppression network architecture.
[0026] The present invention also proposes a system for suppressing earthquake background noise, the system comprising an acquisition module, a training module, and a detection module;
[0027] The acquisition module is used to acquire a background noise dataset and divide the background noise dataset into a training dataset and a validation dataset.
[0028] The training module is used to construct an initial background noise suppression network architecture and input the training dataset into the initial background noise suppression network architecture for training to obtain the target background noise suppression network architecture.
[0029] The detection module is used to process the background noise data to be tested using a target background noise suppression network architecture to obtain noise suppression data results.
[0030] Preferably, the training module is used to construct an initial background noise suppression network architecture, including:
[0031] The training module is used to build the initial background noise network architecture, set the loss function of the initial background noise network architecture, and the parameters to be optimized.
[0032] Preferably, the training module is used to input the training dataset into the initial background noise suppression network architecture for training, including:
[0033] The training module is used to input the training dataset into the initial background noise suppression network architecture;
[0034] With the goal of minimizing the loss function, the parameters to be optimized are adjusted and determined.
[0035] Using the final determined parameters to be optimized as target parameters, the target background noise suppression network architecture is obtained.
[0036] Preferably, the system further includes a verification module;
[0037] The verification module is used to input the verification dataset into the target background noise suppression network architecture to verify the accuracy of the background noise suppression network architecture.
[0038] The present invention also proposes an electronic device, comprising:
[0039] Processor and memory;
[0040] The processor invokes the computer program stored in the memory to execute any of the methods described above for suppressing seismic background noise.
[0041] The present invention also proposes a computer-readable storage medium.
[0042] The computer-readable storage medium stores a computer program that, when executed by a processor, enables the processor to perform any of the methods described above for suppressing seismic background noise.
[0043] The present invention has the following beneficial effects:
[0044] This invention utilizes deep learning technology to extract data features of seismic background noise, thereby achieving targeted suppression of background noise, improving the basic signal-to-noise ratio of seismic data, providing high-quality basic data for subsequent seismic imaging and reservoir prediction, and helping to improve the quality and efficiency of oil and gas exploration and processing.
[0045] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description and the drawings. Attached Figure Description
[0046] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0047] Figure 1 This invention illustrates a method for suppressing seismic background noise in an embodiment of the invention;
[0048] Figure 2 This diagram illustrates a background noise segmentation line in an embodiment of the present invention.
[0049] Figure 3 This diagram illustrates the selection of seismic signals in an embodiment of the present invention.
[0050] Figure 4 This diagram illustrates the background noise suppression network architecture in an embodiment of the present invention.
[0051] Figure 5 This diagram illustrates the original single-shot record A in an embodiment of the present invention.
[0052] Figure 6 This diagram illustrates the denoising results for a single-shot record A in an embodiment of the present invention.
[0053] Figure 7 This diagram illustrates noise identification for a single-shot record A in an embodiment of the present invention.
[0054] Figure 8 This diagram illustrates the original single-shot record B in an embodiment of the present invention.
[0055] Figure 9 This diagram illustrates the denoising results for a single-shot record B in an embodiment of the present invention.
[0056] Figure 10 This diagram illustrates noise identification for single-shot record B in an embodiment of the present invention.
[0057] Figure 11 This invention illustrates a system for suppressing seismic background noise in an embodiment of the invention;
[0058] Figure 12 A diagram of an electronic device according to an embodiment of the present invention is shown. Detailed Implementation
[0059] Example embodiments will now be described more fully with reference to the accompanying drawings. However, example embodiments can be implemented in many forms and should not be construed as limited to the examples set forth herein; rather, these embodiments are provided to make this disclosure more comprehensive and complete, and to fully convey the concept of the example embodiments to those skilled in the art. The described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to give a full understanding of embodiments of this disclosure. However, those skilled in the art will recognize that the technical solutions of this disclosure can be practiced with one or more of the specific details omitted, or other methods, components, apparatus, steps, etc., can be employed. In other instances, well-known technical solutions are not shown or described in detail to avoid obscuring various aspects of this disclosure.
[0060] Furthermore, the accompanying drawings are merely illustrative of this disclosure and are not necessarily drawn to scale. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted. Some block diagrams shown in the drawings are functional entities and do not necessarily correspond to physically or logically independent entities. These functional entities may be implemented in software, in one or more hardware units or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.
[0061] The flowchart shown in the attached diagram is merely an illustrative example and does not necessarily include all steps. For example, some steps may be broken down, while others may be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.
[0062] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented, for example, in orders other than those illustrated or described herein.
[0063] Furthermore, the terms “comprising” and “having”, and any variations thereof, are intended to cover non-exclusive inclusion, such that a process, method, system, product, or device that includes a series of steps or sub-modules is not necessarily limited to those steps or sub-modules that are explicitly listed, but may include other steps or sub-modules that are not explicitly listed or that are inherent to such process, method, product, or device.
[0064] To address the issues of low noise recognition accuracy and poor adaptability in existing noise suppression technologies, this invention proposes a dataset construction method and a multi-scale neural network structure for background noise. Through interactive noise picking and data synthesis, a large set of denoised samples based on real data is formed. Compared with conventional noise suppression technologies, the trained denoised network model of this invention not only effectively improves the suppression accuracy of background noise but also greatly enhances the processing efficiency.
[0065] like Figure 1 As shown, the present invention proposes a method for suppressing seismic background noise, the method comprising:
[0066] S1 acquires a background noise dataset and divides the background noise dataset into a training dataset and a validation dataset;
[0067] S2 constructs an initial background noise suppression network architecture and inputs the training dataset into the initial background noise suppression network architecture for training to obtain the target background noise suppression network architecture.
[0068] S3 uses a target background noise suppression network architecture to process the background noise data to be tested and obtain the noise suppression data results.
[0069] Specifically, the technical solution of this invention is as follows:
[0070] (1) The method for constructing the background noise dataset, and the specific implementation steps are as follows:
[0071] (1.1) Obtaining the background noise data region: Calculate the background noise dividing line using the given zero-shot-receiver distance time t0 and formation velocity v. The calculation formula is as follows:
[0072]
[0073] Where x is the shot-receiver distance, which can be calculated from the coordinates of the shot point and the receiver point in the seismic trace head:
[0074]
[0075] Among them, (x src y src ) and (x rec y rec The coordinates of the shot point and receiver point are respectively, and this information can be read from the observation system. Figure 2 This is a schematic diagram of background noise segmentation. The area above the top tangent line in the diagram represents the region containing only background noise.
[0076] (1.2) Obtaining the data region for seismic signals: If there is no background noise or the background noise energy is extremely low in the data region preceding the first arrival wave, it can be considered that the seismic data in this region has no background noise. In this case, the data region for the seismic signals is determined by manually selecting an interactive polygon. Figure 3 As shown.
[0077] (1.3) Obtain seismic data samples (patches), including the following two methods (you can choose one or select all):
[0078] ① Traversal method: Based on the given sample piece size (px, pt) and step size (step_x, step_t), and the given region range {top left corner (nx0, nt0), bottom left corner (nx0, nt1), top right corner (nx2, nt2), bottom right corner (nx2, nt1)}, the process of obtaining the sample piece is as follows:
[0079] Step a): Initialize indexes i = 0, j = 0, given seismic data d;
[0080] Step b): Obtain the sample piece p = d(i:i+px,j:j+pt);
[0081] Step c): Update index i = i + step_x, j = j + step_t;
[0082] Step d): If i+px or j+pt exceeds the given range, the current loop will end; otherwise, repeat steps b) to d).
[0083] ② Randomized: Based on the given sample piece size (px, pt) and the expected number n, and the given region range {top left (nx0, nt0), bottom left (nx0, nt1), top right (nx2, nt2), bottom right (nx2, nt1)}, the process for obtaining the sample piece is as follows:
[0084] Step a): Initialize the counter variable num = 0, given the earthquake data d.
[0085] Step b): Randomly obtain indices i = rand(nx0, nx2-px), j = rand(k*i+b, nt1);
[0086] Step c): Obtain the sample piece p = d(i:i+px,j:j+pt);
[0087] Step d): Update the counter variable num = num + 1;
[0088] Step e): If num exceeds the given quantity n, then end; otherwise, repeat steps b) to e).
[0089] (1.4) Constructing a denoising training dataset: Obtain sample patches of the background noise described in (1.1) and the seismic signal described in (1.2) according to the method described in (1.3), and complete the process in the following three ways:
[0090] ① Background noise training set only: Randomly select a background noise sample as input data x, construct a zero matrix with the same shape as it as seismic signal label data y, and background noise label data n = x. The three together are used as a set of data {x, y, n} in the training set. Repeat the above operation until the given quantity requirement is met.
[0091] ② Seismic signal training set only: Randomly select a seismic signal sample piece as input data x, copy the seismic signal sample piece as seismic signal label data y, construct a zero matrix with the same shape as it as background noise label data n, and use the three together as a set of data {x,y,n} in the training set. Repeat the above operation until the given quantity requirement is met.
[0092] ③ Noise and signal training set: Randomly select a seismic signal sample s and a background noise sample n0, and weight them together to obtain the input data x = s + a * n0 (where a is a random number within a given range, such as a random number between (0.5, 1.5), used to characterize the strength of the noise). The seismic signal sample s is used as the seismic signal label y, and the background noise label data n = a * n0. The three together are used as a set of data {x, y, n} in the training set. Repeat the above operation until the given quantity requirement is met.
[0093] Combine the three training sets above in a given ratio (e.g., 3:3:4) to form a denoised training dataset; and construct a denoised validation dataset in the same way, the size of which is a certain proportion of the number of training datasets (e.g., 20%).
[0094] (1.5) Conventional seismic noise suppression techniques (combined techniques such as anomalous amplitude suppression, FKK domain filtering, and tilt filtering) are used to denoise the actual seismic data. After conventional noise suppression, the "two-width and one-height" seismic data are arranged in the order of "shot line-shot point-receiver line-receiver point" to form a five-dimensional data volume. The seismic data before denoising is used as the input data, and the data after denoising is used as (pseudo)label data.
[0095] (2) Design a background noise suppression network architecture
[0096] (2.1) Constructing a background noise suppression network:
[0097] The network structure used in the background noise suppression network, such as Figure 4 As shown, the network consists of an FK domain channel network, a time domain channel network, a shared network, a signal branch network, and a noise branch network. After the training dataset is input into the network, it passes through the FK domain channel network and the time domain channel network respectively. The output features of the two are connected and combined before entering the shared network. Then, it passes through the signal branch network to output the denoised signal and through the noise branch network to output the identified noise.
[0098] The FK domain channel network consists of one FK transform layer and five feature mapping layers.
[0099] The time-domain channel network consists of 6 sets of feature mapping layers;
[0100] The shared network consists of four feature extraction layers and four feature reconstruction layers. To avoid feature loss, skip connections are added between feature maps of the same shape.
[0101] Both the signal branch network and the noise branch network consist of 4 sets of feature mapping layers and 1 set of output layers;
[0102] The FK transform layer uses existing FK transform technology and has no learning parameters.
[0103] The feature extraction layer consists of five sets of convolutional layers and downsampling layers with 128 channels and a filter size of 3. Each set of convolutional layers contains the ReLU activation function.
[0104] The feature reconstruction layer consists of five sets of convolutional layers with 128 channels and a filter size of 3, and upsampling layers. Each set of convolutional layers contains the ReLU activation function.
[0105] The feature mapping layer consists of five convolutional layers with 128 channels and a filter size of 3, and each convolutional layer contains the ReLU activation function;
[0106] The downsampling layer consists of a set of convolutional layers with 128 channels, a filter size of 3, and a stride of 2.
[0107] The upsampling layer consists of a set of deconvolutional layers with 128 channels, a filter size of 3, and a stride of 2.
[0108] The output layer consists of a set of convolutional layers with 128 input channels, 2 output channels, and a filter size of 3.
[0109] (2.2) Constructing the loss function:
[0110]
[0111] Where K is the number of samples in the dataset, {x k ,y k ,n k} represents a set of training samples, x k For input data, y k For seismic signal label data, n k The background noise is the labeled data, and Θ represents the network parameters that the network needs to learn during training; ||·|| F It is the Frobenius norm.
[0112] (3) Train the network parameters and test the model using the test set.
[0113] (3.1) Randomly initialize the parameters of the background noise suppression network to obtain the initial model of the background noise suppression network;
[0114] (3.2) Input the training set of the earthquake data into the initial model of the background noise suppression network, obtain the data of the output layer, calculate the loss function as shown in formula (3), and determine whether the current iteration number meets the maximum training number (default value is 50). If not, adjust the network parameters of the current network through the backpropagation algorithm until the maximum training number is reached, and determine the network model that has reached the training number as the background noise suppression network model. The backpropagation algorithm of the background noise suppression network is the process of obtaining the optimal network parameters by minimizing the loss function as shown in formula (3). The minimization process of the objective function (formula (3)) can be achieved by the Adam optimization algorithm.
[0115] (4) Use network models to suppress noise in real data.
[0116] (4.1) Input the seismic data d into the background noise suppression network model to obtain the predicted seismic signal data. and background noise data Obtain the noise-suppressed data:
[0117]
[0118] Where μ∈[0,1], the default value of μ is 0.5.
[0119] Figure 5 This is the original single-shot record. Figure 6 As shown in the processing results of this invention, the background noise is significantly suppressed, proving the effectiveness of the noise suppression method of this invention. Figure 7 The noise identified by this invention contains almost no effective signal, demonstrating that this invention effectively suppresses noise while maintaining high amplitude preservation.
[0120] Example 1
[0121] This example uses actual seismic data collected from a basin (e.g., ...). Figure 5 , Figure 8 Using controlled-source excitation, severe background noise interference results in a low signal-to-noise ratio for seismic data. The first arrival of the seismic signal is submerged in background noise, making accurate identification impossible and hindering further processing and interpretation of the seismic data. This invention cleverly constructs a dataset and utilizes a deep neural network to learn the data characteristics of background noise. The trained denoising model effectively suppresses background noise (such as…). Figure 6 , Figure 9 The boundaries of the first arrival wave are clearer. Actual field tests show that after using this invention to suppress background noise, the automated first arrival wave pickup rate increased from 54% to 92%, demonstrating significant effectiveness. This invention can accurately identify background noise (such as…). Figure 7 , Figure 10 This indicates that the present invention has high amplitude preservation. After suppressing background noise using the present invention, the signal-to-noise ratio of seismic data is higher, which can provide high-quality basic data for subsequent processing and interpretation.
[0122] like Figure 11 As shown, the present invention also proposes a system for suppressing earthquake background noise, the system comprising an acquisition module 10, a training module 20 and a detection module 30;
[0123] The acquisition module 10 is used to acquire a background noise dataset and divide the background noise dataset into a training dataset and a validation dataset.
[0124] The training module 20 is used to construct an initial background noise suppression network architecture and input the training dataset into the initial background noise suppression network architecture for training to obtain the target background noise suppression network architecture.
[0125] The detection module 30 is used to process the background noise data to be tested using a target background noise suppression network architecture to obtain noise suppression data results.
[0126] like Figure 12As shown, corresponding to the method for suppressing seismic background noise provided above, the present invention also provides an electronic device. Since the embodiment of this device is similar to the method embodiment described above, the description is relatively simple; relevant details can be found in the description of the method embodiment section above. The device described below is merely illustrative. This device may include: a processor 1, a memory 2, a communication bus (i.e., the aforementioned device bus), and a lookup engine. The processor 1 and the memory 2 communicate with each other via the communication bus and communicate with external systems via a communication interface. The processor 1 can invoke logical instructions in the memory 2 to execute the method for suppressing seismic background noise.
[0127] Furthermore, the logical instructions in the aforementioned memory 2 can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as memory chips, USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0128] On the other hand, embodiments of the present invention also provide a processor-readable storage medium storing a computer program 3, which, when executed by a processor 1, is implemented to perform the methods for suppressing seismic background noise provided in the above embodiments.
[0129] The processor-readable storage medium can be any available medium or data storage device that the processor 1 can access, including but not limited to magnetic memory (e.g., floppy disk, hard disk, magnetic tape, magneto-optical disk (MO)), optical memory (e.g., CD, DVD, BD, HVD), and semiconductor memory (e.g., ROM, EPROM, EEPROM, non-volatile memory (NAND FLASH), solid-state drive (SSD)).
[0130] Those skilled in the art should understand that, despite the detailed description of the present invention with reference to the foregoing embodiments, modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.
Claims
1. A method for suppressing seismic background noise, characterized in that, The method includes: Obtain a background noise dataset and divide it into a training dataset and a validation dataset; wherein, the training dataset includes a background noise-only training set, a seismic signal-only training set, and a noise and signal training set; The background noise-only training set consists of background noise sample patches, seismic signal label data, and background noise label data, wherein the seismic signal label data is a zero matrix. The seismic signal training set consists of seismic signal sample pieces, seismic signal label data, and background noise label data, wherein the background noise label data is a zero matrix. The noise and signal training set consists of input data, seismic signal label data, and background noise label data. The input data is a weighted superposition of seismic signal sample pieces and background noise sample pieces. The seismic signal label data is a seismic signal sample piece, and the background noise label data is a weighted sum of background noise sample pieces. An initial background noise suppression network architecture is constructed, and the training dataset is input into the initial background noise suppression network architecture for training to obtain the target background noise suppression network architecture. The initial background noise suppression network architecture consists of an FK domain channel network, a time domain channel network, a shared network, a signal branch network, and a noise branch network. After the training dataset is input into the network, it passes through the FK domain channel network and the time domain channel network respectively. The output features of these two networks are then concatenated and combined before entering the shared network. Finally, the signal branch network outputs the denoised signal, and the noise branch network outputs the identified noise. The FK domain channel network consists of one set of FK transform layers and multiple sets of feature mapping layers; the time domain channel network consists of multiple sets of feature mapping layers; the shared network consists of multiple sets of feature extraction layers and multiple sets of feature reconstruction layers; the signal branch network and the noise branch network each consist of multiple sets of feature mapping layers and one set of output layers. The feature extraction layer consists of five convolutional layers and a downsampling layer, each containing a ReLU activation function; the feature reconstruction layer consists of five convolutional layers and an upsampling layer, each containing a ReLU activation function; the feature mapping layer consists of five convolutional layers, each containing a ReLU activation function; the downsampling layer consists of convolutional layers; and the upsampling layer consists of deconvolutional layers. The background noise suppression network architecture is used to process the background noise data to be tested, and the noise suppression data results are obtained.
2. The method for suppressing seismic background noise according to claim 1, characterized in that, The methods for acquiring the seismic signal sample pieces and background noise sample pieces include traversal acquisition and random acquisition.
3. The method for suppressing seismic background noise according to claim 1, characterized in that, The construction of the initial background noise network architecture includes: Construct an initial background noise network architecture, and define the loss function and parameters to be optimized for the initial background noise network architecture.
4. The method for suppressing seismic background noise according to claim 3, characterized in that, The training dataset is input into the initial background noise suppression network architecture for training, including: Input the training dataset into the initial background noise suppression network architecture; With the goal of minimizing the loss function, the parameters to be optimized are adjusted and determined. Using the final determined parameters to be optimized as target parameters, the target background noise suppression network architecture is obtained.
5. The method for suppressing seismic background noise according to claim 1, characterized in that, After obtaining the target background noise suppression network architecture, the following is included: The validation dataset is input into the target background noise suppression network architecture to verify the accuracy of the background noise suppression network architecture.
6. A system for suppressing seismic background noise, characterized in that, The system includes an acquisition module, a training module, and a detection module; The acquisition module is used to acquire a background noise dataset and divide the background noise dataset into a training dataset and a validation dataset; wherein, the training dataset includes a background noise-only training set, a seismic signal-only training set, and a noise and signal training set; The background noise-only training set consists of background noise sample patches, seismic signal label data, and background noise label data, wherein the seismic signal label data is a zero matrix. The seismic signal training set consists of seismic signal sample pieces, seismic signal label data, and background noise label data, wherein the background noise label data is a zero matrix. The noise and signal training set consists of input data, seismic signal label data, and background noise label data. The input data is a weighted superposition of seismic signal sample pieces and background noise sample pieces. The seismic signal label data is a seismic signal sample piece, and the background noise label data is a weighted sum of background noise sample pieces. The training module is used to construct an initial background noise suppression network architecture and input the training dataset into the initial background noise suppression network architecture for training to obtain the target background noise suppression network architecture. The initial background noise suppression network architecture consists of an FK domain channel network, a time domain channel network, a shared network, a signal branch network, and a noise branch network. After the training dataset is input into the network, it passes through the FK domain channel network and the time domain channel network respectively. The output features of the two are connected and combined and then enter the shared network. After passing through the signal branch network, the denoised signal is output, and after passing through the noise branch network, the identified noise is output. The FK domain channel network consists of one set of FK transform layers and multiple sets of feature mapping layers; the time domain channel network consists of multiple sets of feature mapping layers; the shared network consists of multiple sets of feature extraction layers and multiple sets of feature reconstruction layers; the signal branch network and the noise branch network each consist of multiple sets of feature mapping layers and one set of output layers. The feature extraction layer consists of five convolutional layers and a downsampling layer, each containing a ReLU activation function; the feature reconstruction layer consists of five convolutional layers and an upsampling layer, each containing a ReLU activation function; the feature mapping layer consists of five convolutional layers, each containing a ReLU activation function; the downsampling layer consists of convolutional layers; and the upsampling layer consists of deconvolutional layers. The detection module is used to process the background noise data to be tested using a target background noise suppression network architecture to obtain noise suppression data results.
7. The system for suppressing seismic background noise according to claim 6, characterized in that, The training module is used to construct the initial background noise suppression network architecture, including: The training module is used to build the initial background noise network architecture, set the loss function of the initial background noise network architecture, and the parameters to be optimized.
8. The system for suppressing seismic background noise according to claim 7, characterized in that, The training module is used to input the training dataset into the initial background noise suppression network architecture for training, including: The training module is used to input the training dataset into the initial background noise suppression network architecture; With the goal of minimizing the loss function, the parameters to be optimized are adjusted and determined. Using the final determined parameters to be optimized as target parameters, the target background noise suppression network architecture is obtained.
9. The system for suppressing seismic background noise according to claim 6, characterized in that, The system also includes a verification module; The verification module is used to input the verification dataset into the target background noise suppression network architecture to verify the accuracy of the background noise suppression network architecture.
10. An electronic device, characterized in that, include: Processor and memory; The processor invokes the computer program stored in the memory to execute the method for suppressing seismic background noise as described in any one of claims 1 to 5.
11. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, enables the processor to perform the method for suppressing seismic background noise as described in any one of claims 1 to 5.