Earthquake imaging profile noise suppression method and device

Abstract

The embodiment of the invention discloses a seismic imaging profile noise suppression method and device. The method comprises the following steps: acquiring to-be-processed seismic imaging profile data; based on a preset spatial sampling mode, performing decomposition processing on the to-be-processed seismic imaging profile data to obtain a first subset and a second subset; adding noise to the first subset and the second subset to obtain a third subset containing noise and a fourth subset containing noise; constructing a training sample pair; constructing a self-supervised noise suppression deep learning neural network, and performing model training on the self-supervised noise suppression deep learning neural network based on the training sample pair to obtain a seismic imaging profile noise suppression model; and carrying out noise suppression processing on to-be-processed seismic imaging profile data by using the seismic imaging profile noise suppression model to obtain denoised seismic imaging profile data. Therefore, the signal-to-noise ratio of the seismic data can be effectively improved without depending on a clean label, and meanwhile, good signal fidelity is kept.

Description

Technical Field

[0001] This application relates to the field of marine oil geophysical exploration, specifically to a method and apparatus for suppressing noise in seismic imaging profiles. Background Technology

[0002] Seismic profiles are inevitably affected by migration noise and random noise, impacting their signal-to-noise ratio (SNR) and consequently affecting the accuracy of structural interpretation, attribute analysis, and reservoir prediction. A high SNR is a crucial prerequisite for seismic data interpretation and inversion. Therefore, improving the SNR of seismic data is a vital step in the data processing workflow. However, traditional denoising methods are often ineffective in handling complex noise and maintaining signal fidelity. Therefore, a solution that can effectively suppress noise in seismic imaging profiles is urgently needed. Summary of the Invention

[0003] In view of the above problems, this application is made in order to provide a method, apparatus, computing device, computer storage medium and computer program product for suppressing seismic imaging profile noise to overcome or at least partially solve the above problems.

[0004] According to one aspect of the embodiments of this application, a method for suppressing noise in seismic imaging profiles is provided, comprising: Acquire seismic imaging profile data to be processed; Based on a preset spatial sampling method, the seismic imaging profile data to be processed is decomposed to obtain a first subset and a second subset, wherein the first subset and the second subset are spatially interleaved and complementary. Adding noise to the first subset yields a noisy third subset, and adding noise to the second subset yields a noisy fourth subset. Construct training sample pairs, wherein the training sample pairs include: using the fourth subset as input samples and the first subset as target samples; and using the third subset as input samples and the second subset as target samples; A self-supervised noise suppression deep learning neural network was constructed, and the model was trained on the self-supervised noise suppression deep learning neural network based on the training samples to obtain the seismic imaging profile noise suppression model. The seismic imaging profile noise suppression model is used to suppress noise in the seismic imaging profile data to be processed, resulting in denoised seismic imaging profile data.

[0005] Furthermore, based on a preset spatial sampling method, the seismic imaging profile data to be processed is divided into a first subset and a second subset, which further includes: A first subset is determined based on sampling points at a pair of diagonal locations in each 2×2 neighborhood of the seismic imaging profile data to be processed; a second subset is determined based on sampling points at another pair of diagonal locations in each 2×2 neighborhood of the seismic imaging profile data to be processed.

[0006] Furthermore, the sample value of each sampling point in the first subset is composed of the average of a pair of sampling points at diagonal positions, and the sample value of each sampling point in the second subset is composed of the average of another pair of sampling points at diagonal positions.

[0007] Furthermore, based on the training samples, a self-supervised noise suppression deep learning neural network is trained to obtain a seismic imaging profile noise suppression model, which further includes: All input samples from the training samples are fed into a self-supervised noise suppression deep learning neural network. The self-supervised noise suppression deep learning neural network undergoes multiple levels of iterative noise reduction processing until the model training ends, thus obtaining a seismic imaging profile noise suppression model. Each iteration of the denoising process performs the following steps: convolutional residual learning is performed based on the input samples; the convolutional residual learning result is then concatenated with the input samples; sparse regularization is performed based on the residual concatenation result, the previous-level auxiliary variable or the initial auxiliary variable with convergence, to obtain the model's denoised output; the auxiliary variables are updated based on the residual concatenation result and the model's denoised output; the loss function is calculated based on the model's denoised output and all target samples in the training sample pair; and the neural network parameters are optimized based on the loss function.

[0008] Furthermore, adding noise to the first subset to obtain a noisy third subset, and adding noise to the second subset to obtain a noisy fourth subset, further includes: Adding normally distributed random noise to the first subset yields a noisy third subset, and adding normally distributed random noise to the second subset yields a noisy fourth subset.

[0009] Furthermore, adding normally distributed random noise to the first subset yields a noisy third subset, and adding normally distributed random noise to the second subset yields a noisy fourth subset, which further includes: The following formulas are used to add normally distributed random noise to the first subset and the second subset respectively:

[0010]

[0011]

[0012]

[0013]

[0014]

[0015] Where a is any sampling point in the noisy third subset, A is any sampling point in the first subset, b is any sampling point in the noisy fourth subset, B is any sampling point in the second subset, N is the number of longitudinal sampling points in the seismic imaging profile data, and f is the noise addition intensity control factor.

[0016] According to another aspect of the embodiments of this application, a seismic imaging profile noise suppression device is provided, comprising: The acquisition module is suitable for acquiring seismic imaging profile data to be processed. The decomposition module is suitable for decomposing the seismic imaging profile data to be processed based on a preset spatial sampling method to obtain a first subset and a second subset, wherein the first subset and the second subset are spatially interleaved and complementary. The noise addition module is suitable for adding noise to the first subset to obtain a noisy third subset, and adding noise to the second subset to obtain a noisy fourth subset; The training sample construction module is suitable for constructing training sample pairs, wherein the training sample pairs include: using a fourth subset as input samples and a first subset as target samples; and using a third subset as input samples and a second subset as target samples. The model training module is suitable for constructing a self-supervised noise suppression deep learning neural network. Based on the training samples, the self-supervised noise suppression deep learning neural network is trained to obtain a seismic imaging profile noise suppression model. The noise suppression processing module is suitable for using a seismic imaging profile noise suppression model to suppress noise in the seismic imaging profile data to be processed, and to obtain denoised seismic imaging profile data.

[0017] According to another aspect of the embodiments of this application, a computing device is provided, including: a processor, a memory, a communication interface and a communication bus, wherein the processor, the memory and the communication interface communicate with each other through the communication bus; The memory is used to store at least one executable instruction, which causes the processor to perform the operation corresponding to the above-described seismic imaging profile noise suppression method.

[0018] According to another aspect of the embodiments of this application, a computer storage medium is provided, wherein at least one executable instruction is stored in the storage medium, the executable instruction causing a processor to perform an operation corresponding to the above-described seismic imaging profile noise suppression method.

[0019] According to another aspect of the embodiments of this application, a computer program product is provided, including at least one executable instruction that causes a processor to perform operations corresponding to the above-described seismic imaging profile noise suppression method.

[0020] The seismic imaging profile noise suppression method and apparatus provided in this application can effectively improve the signal-to-noise ratio of seismic data without relying on clean labels, while maintaining good signal fidelity. This method is computationally efficient, easy to integrate into existing processing workflows, and readily applicable in actual production.

[0021] The above description is merely an overview of the technical solutions of the embodiments of this application. In order to better understand the technical means of the embodiments of this application and to implement them in accordance with the contents of the specification, and to make the above and other objects, features and advantages of the embodiments of this application more obvious and understandable, specific implementation methods of the embodiments of this application are described below. Attached Figure Description

[0022] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the embodiments of this application. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings: Figure 1 A schematic flowchart of a seismic imaging profile noise suppression method according to an embodiment of this application is shown; Figure 2 A diagram showing the structure of a self-supervised noise suppression deep learning neural network; Figure 3A This is a schematic diagram of seismic imaging profile data from forward migration. Figure 3B A schematic diagram showing the addition of random noise to forward-modeled migration seismic imaging profile data; Figure 3C This is a schematic diagram of the seismic imaging profile data after noise suppression. Figure 4A A schematic diagram of seismic imaging profile data before denoising; Figure 4B This is a schematic diagram of the denoised seismic imaging profile data. Figure 4C A schematic diagram of noise in suppressed seismic imaging profile data; Figure 5 A structural block diagram of a seismic imaging profile noise suppression device according to an embodiment of this application is shown; Figure 6 A schematic diagram of the structure of a computing device according to an embodiment of this application is shown. Detailed Implementation

[0023] Exemplary embodiments of the present disclosure will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the present disclosure are shown in the drawings, it should be understood that the present disclosure may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the disclosure to those skilled in the art.

[0024] Figure 1 A schematic flowchart of a seismic imaging profile noise suppression method according to an embodiment of this application is shown, as follows: Figure 1 As shown, the method includes the following steps: Step S101: Obtain the seismic imaging profile data to be processed.

[0025] The seismic imaging profile data to be processed is noisy data that needs to be suppressed. It can be uploaded by users or imported from the data.

[0026] Step S102: Based on the preset spatial sampling method, the seismic imaging profile data to be processed is decomposed to obtain a first subset and a second subset, wherein the first subset and the second subset are spatially interleaved and complementary.

[0027] Based on the preset spatial sampling method, spatial decomposition processing is performed on the seismic imaging profile data to be processed. The preset spatial sampling method can be a checkerboard downsampling method or other sampling methods, such as sampling rules set based on the longitudinal seismic trace sequence and spatial sampling point location of the seismic imaging profile. These can be flexibly set according to actual needs. Through spatial decomposition processing, the seismic imaging profile data to be processed can be decomposed into two subsets, namely, the first subset and the second subset, which are spatially interleaved and complementary.

[0028] In one optional embodiment of this application, a first subset can be determined based on sampling points at a pair of diagonal positions in each 2×2 neighborhood of the seismic imaging profile data, wherein the sample value of each sampling point in the first subset is composed of the average value of the sampling points at the pair of diagonal positions; and a second subset can be determined based on sampling points at another pair of diagonal positions in each 2×2 neighborhood of the seismic imaging profile data, wherein the sample value of each sampling point in the second subset is composed of the average value of the sampling points at the other pair of diagonal positions.

[0029] For example, suppose the seismic imaging profile data to be processed is D. It can be decomposed into two subsets using checkerboard downsampling. Here, the first subset is denoted as A, and the second subset as B: Subset A consists of the average values ​​of the samples at positions (2i, 2j) and (2i+1, 2j+1).

[0030] Subset B consists of the average values ​​of the sample points at positions (2i+1, 2j) and (2i, 2j+1), where i is the longitudinal sampling point of the seismic imaging profile data, j is the lateral sampling point of the seismic imaging profile data, i is an integer greater than or equal to 0, and j is an integer greater than or equal to 0.

[0031] Right now:

[0032]

[0033] These two subsets are spatially complementary and contain most of the information from the original data.

[0034] Step S103: Add noise to the first subset to obtain a noisy third subset, and add noise to the second subset to obtain a noisy fourth subset.

[0035] Noise is added to the first subset and the second subset respectively. For example, an appropriate amount of random noise following a normal distribution is added to the first subset and the second subset respectively, thereby obtaining a noisy third subset and a noisy fourth subset.

[0036] Here, the third subset is denoted as 'a', and the fourth subset as 'b'. The sampling points in the third subset and the sampling points in the fourth subset can be determined using the following formulas:

[0037]

[0038]

[0039]

[0040]

[0041]

[0042] Where a is any sampling point in the noisy third subset, A is any sampling point in the first subset, b is any sampling point in the noisy fourth subset, B is any sampling point in the second subset, N is the number of longitudinal sampling points in the seismic imaging profile data, and f is the noise addition intensity control factor, used to adjust the noise intensity, which is usually 0.3.

[0043] Step S104: Construct training sample pairs, wherein the training sample pairs include: using the fourth subset as input samples and the first subset as target samples; and using the third subset as input samples and the second subset as target samples.

[0044] Following step S103, a training sample pair construction process needs to be performed to form input-target pairs for training the model. Specifically, any sampling point in the fourth subset is used as an input sample, and any sampling point in the first subset is used as a target sample to form a training sample pair; any sampling point in the third subset is used as an input sample, and any sampling point in the second subset is used as a target sample to form another training sample pair. All of the above-mentioned training sample pairs are used for subsequent model training.

[0045] Through the processing in steps S102-S104, a training dataset is created from a single seismic data volume without any external labels, solving the unlabeled problem and expanding the scale of the training samples.

[0046] Step S105: Construct a self-supervised noise suppression deep learning neural network, and train the self-supervised noise suppression deep learning neural network based on training samples to obtain a seismic imaging profile noise suppression model.

[0047] Specifically, this deep learning neural network is designed for signal reconstruction tasks involving noisy data, and the network as a whole adopts a multi-round iterative processing paradigm. In each iteration, the input data is first extracted using a convolutional module, which consists of two convolutional layers: the first layer maps the single-channel input to 32-channel features using a 3×3 convolutional kernel; the second layer maps the 32-channel features back to the single-channel output, also using a 3×3 convolutional kernel. To alleviate the gradient vanishing problem in deep network training, a residual connection is introduced after the output of the convolutional module, adding the original input data and the output of the convolutional module element-wise to retain the basic features of the original data. Then, sparsity regularization is performed on the data: a learnable threshold parameter is used to perform a soft thresholding operation on the data to suppress noise components in the signal and enhance the sparsity of the effective signal; at the same time, the network dynamically adjusts the balance between signal reconstruction and sparsity regularization in each iteration, as well as the update magnitude of consistency constraints during the iteration process, through a learnable balance parameter, and finally outputs a sparsely reconstructed signal optimized through multiple iterations. All threshold and balancing parameters in the network are set as learnable parameters and are adaptively adjusted during the training process. The overall network is gradually optimized through multiple rounds of iteration.

[0048] Define the following main variables: y represents the input noisy seismic data, specifically the noisy seismic data in the third and fourth subsets, namely a and b.

[0049] x is the estimate of the clean seismic data in each iteration (fitted from the original data by convolutional layers).

[0050] z represents the denoised output of the model after sparse regularization.

[0051] u is an auxiliary variable used to ensure the convergence of the iterative process and to balance the consistency of x and z.

[0052] Perform network initialization: Set the learnable neural network parameters during the iteration process, including: Penalty parameters The initial value is set to 0.1 to control the convergence speed of the neural network iteration; Balance parameters The initial value is set to 0.1 to balance the data fidelity item and the regularization item; Threshold parameter: Initial value is set to 0.005, used to control the strength of soft thresholding operation; Threshold factor_f: Initial value is set to 0.005, used for fine adjustment of the threshold; The learnable neural network parameters mentioned above are all automatically learned through backpropagation during the iteration process, achieving adaptive optimization of the parameters.

[0053] Initial value: z 0 =0, u 0 =0 Here, all input samples from the training sample pair are fed into the self-supervised noise suppression deep learning neural network for model training. The self-supervised noise suppression deep learning neural network undergoes multiple levels of iterative noise reduction processing until the model training ends, resulting in a seismic imaging profile noise suppression model. Each iteration of the denoising process performs the following steps: convolutional residual learning is performed based on the input samples; the convolutional residual learning result is concatenated with the input samples to obtain a clean seismic data estimate; sparse regularization is performed based on the residual concatenation result (clean seismic data estimate) and the previous-level auxiliary variable or initial auxiliary variable with convergence to obtain the model denoising output; the auxiliary variables are updated based on the residual concatenation result and the model denoising output; the loss function is calculated based on the model denoising output and all target samples in the training sample pair; and the neural network parameters are optimized based on the loss function.

[0054] Figure 2 The structure diagram of a self-supervised noise-suppressed deep learning neural network is shown below. Figure 2As shown, this example uses noisy seismic data from the third subset as the input sample and seismic data from the second subset as the target sample. The same principle applies to noisy seismic data from the fourth subset as the input sample and seismic data from the first subset as the target sample. In the following introduction, we will combine noisy seismic data from the third subset as the input sample and seismic data from the second subset as the target sample for a detailed explanation. After multiple iterations, the signal is denoised and filtered once in each iteration. After k iterations, the noise is gradually suppressed, resulting in the seismic imaging profile noise suppression model. Finally, z is the clean data after removing the noise.

[0055] Among them, x-update convolutional residual learning: The x-update step uses a convolutional neural network to implement a non-linear mapping and update the estimates of clean seismic data. First, based on the previous iteration's z, auxiliary variable u, and noisy seismic data, intermediate feature data is generated, for example... , in, For intermediate feature data, For penalty parameters, To balance the parameters, k is an integer greater than or equal to 0. This intermediate feature data is then input into a convolutional block (ConvBlock) for processing. The convolutional block consists of two layers: the first layer: 1→32 channels, 3×3 convolution; the second layer: 32→1 channel, 3×3 convolution. After this iteration, the estimated values ​​of the clean seismic data are updated, i.e., .

[0056] Add residual join: To effectively mitigate the gradient vanishing problem in deep networks while preserving the effective signals from the original seismic data, this stage involves adding residual connections between the updated clean seismic data estimates and the input samples. This summation and fusion of the two results in an enhanced clean seismic data estimate. This operation avoids over-filtering of the effective signals, ensuring that useful information from the original seismic data is retained while removing noise. The addition of residual connections is a key step in this process. .

[0057] z-update soft threshold operation: This stage employs a soft thresholding function to perform sparsity regularization on the estimates of clean seismic data, aiming to suppress noise while preserving effective signals. First, the threshold τ for the current iteration is calculated based on preset threshold parameters and threshold factors. k , τ k The threshold parameter is the product of the threshold parameter and the threshold factor; subsequently, a soft thresholding process is applied to the sum of the enhanced clean seismic data estimate and the auxiliary variables: values ​​with absolute values ​​less than the threshold τ are considered thresholded. k The value is set to 0 to suppress noise; values ​​with absolute values ​​greater than the threshold τ are... kThe numerical values ​​are retained to preserve the sparse characteristics of the effective signal, ultimately yielding the denoising result z for this iteration. k+1 .

[0058] z-update uses a soft threshold function to achieve sparsity regularization:

[0059] Where the threshold Sparse noise suppression is achieved through soft thresholding, preserving the sparsity characteristics of the effective signal. The function of soft thresholding is to set values ​​with absolute values ​​less than the threshold to 0 (suppressing noise) and retain values ​​greater than the threshold (preserving the effective signal).

[0060] u-update auxiliary variable update: To ensure the convergence of the iterative process and balance the consistency between the clean seismic data estimates and the denoised results, this stage uses the clean seismic data estimates xxx obtained from this round of iterations. k With the denoising result z k+1 Update the auxiliary variable to obtain the updated auxiliary variable u. k+1 For the next iteration, specifically, .

[0061] Loss function design: This embodiment employs a combined loss function to balance denoising effectiveness and detail preservation. This loss function is a weighted average of mean squared error (MSE) loss and mean absolute error (L1) loss, with weights of 0.8 and 0.2, respectively. Wherein: MSE loss (mean squared error): used to ensure the overall accuracy of the denoised data; L1 loss (mean absolute error): used to reduce the loss of detail during the denoising process.

[0062] During the iteration process, the network parameters are automatically learned through backpropagation based on the combined loss function, thereby achieving adaptive optimization of the learnable parameters and continuously improving the denoising performance.

[0063] Specifically,

[0064] Where: MSE loss (mean squared error): This ensures the overall accuracy of the data after noise reduction.

[0065] L1 loss (mean absolute error): This reduces the loss of details during the noise reduction process.

[0066] In this embodiment of the application, a forward model with a complex structure can be selected for testing. Figure 3AFor forward-migrated seismic imaging profile data, such as Figure 3A As shown, give Figure 3A The forward-migrated seismic imaging profile data shown has random noise added, such as... Figure 3B As shown, from Figure 3B As can be seen, the main offset arc noise is masked.

[0067] The seismic imaging profile noise suppression method provided in this application is used to... Figure 3B The seismic imaging profile data from forward migration with added random noise shown has undergone noise suppression processing. The noise-suppressed seismic imaging profile data is as follows: Figure 3C As shown, using Figure 2 A self-supervised noise suppression deep learning neural network architecture. Key parameters: The data is divided into small blocks of size 208x208. The neural network uses a lightweight two-layer convolutional layer for x updates, and mean squared error is used as the loss function. Figure 3C As can be seen, the noise from the offset arc is well attenuated, thus confirming that the seismic imaging profile noise suppression model trained using the above method can effectively suppress noise.

[0068] Step S106: Use the seismic imaging profile noise suppression model to suppress noise in the seismic imaging profile data to be processed, and obtain the denoised seismic imaging profile data.

[0069] After training the seismic imaging profile noise suppression model according to step S105, the seismic imaging profile data to be processed can be input into the seismic imaging profile noise suppression model. The seismic imaging profile noise suppression model performs noise suppression processing on the seismic imaging profile data to be processed, thereby filtering out the arc noise in the seismic imaging profile data to be processed, and obtaining the denoised seismic imaging profile data.

[0070] Taking the actual data processing of a certain sea area as an example, the entire work area is a shallow water environment. Figure 4A This is a schematic diagram of the seismic imaging profile data before denoising, as shown below. Figure 4A As shown, its data is characterized by a relatively complex structure, well-developed faults, and a low signal-to-noise ratio.

[0071] The seismic imaging profile noise suppression method provided in the embodiments of this application is used to suppress noise in seismic imaging profiles. Figure 4A The seismic imaging profile data shown is subjected to noise suppression processing to obtain denoised seismic imaging profile data, as follows. Figure 4B As shown, from Figure 4BAs can be seen, the effective reflection axis is clearly visible. After noise suppression processing, the signal-to-noise ratio of random noise in the complex structural regions of the middle and deep layers is significantly improved. Simultaneously, it can be seen that the clarity of the fault is almost unaffected by the noise reduction. To intuitively and conveniently demonstrate the noise suppression effect of the seismic imaging profile noise suppression method provided in this application embodiment, the suppressed noise is presented here in graphical form. Figure 4C A schematic diagram of noise in suppressed seismic imaging profile data.

[0072] In this embodiment, the seismic imaging profile data D itself contains arc-drawing noise. Random noise is added to the first subset A and the second subset B to mask the arc-drawing noise, effectively transforming it into random noise before model training. Since random noise is a non-sparse signal and cannot be learned, self-supervised learning using a neural network is employed. The data with added random noise learns from the data without added noise. Therefore, the arc-drawing noise cannot be learned during model training (because it is masked by random noise). Consequently, the seismic imaging profile noise suppression model obtained after subsequent training does not learn the arc-drawing noise. Thus, the seismic imaging profile noise suppression model can suppress the arc-drawing noise in the seismic imaging profile data to achieve denoising.

[0073] The results of practical application in the work area show that this method can effectively suppress the noise generated by the offset, improve the continuity of the in-phase axis, and is superior to traditional noise reduction methods.

[0074] The seismic imaging profile noise suppression method provided in this application can effectively improve the signal-to-noise ratio of seismic data without relying on clean labels, while maintaining good signal fidelity. This method is computationally efficient, easy to integrate into existing processing workflows, and readily applicable in actual production.

[0075] Figure 5 A structural block diagram of a seismic imaging profile noise suppression device according to an embodiment of this application is shown, as follows: Figure 5 As shown, the device includes: Acquisition module 501 is adapted to acquire seismic imaging profile data to be processed; The decomposition module 502 is adapted to decompose the seismic imaging profile data to be processed based on a preset spatial sampling method to obtain a first subset and a second subset, wherein the first subset and the second subset are spatially interleaved and complementary. The noise addition module 503 is adapted to add noise to the first subset to obtain a noisy third subset, and to add noise to the second subset to obtain a noisy fourth subset; The training sample construction module 504 is adapted to construct training sample pairs, wherein the training sample pairs include: using a fourth subset as input samples and a first subset as target samples; and using a third subset as input samples and a second subset as target samples. The model training module 505 is suitable for constructing a self-supervised noise suppression deep learning neural network. Based on the training samples, the self-supervised noise suppression deep learning neural network is trained to obtain a seismic imaging profile noise suppression model. The noise suppression processing module 506 is suitable for using a seismic imaging profile noise suppression model to suppress noise in the seismic imaging profile data to be processed, so as to obtain denoised seismic imaging profile data.

[0076] Optionally, the decomposition module is further adapted to: determine a first subset based on sampling points at a pair of diagonal locations in each 2×2 neighborhood of the seismic imaging profile data to be processed; and determine a second subset based on sampling points at another pair of diagonal locations in each 2×2 neighborhood of the seismic imaging profile data to be processed.

[0077] Optionally, the sample value of each sample point in the first subset is composed of the average of a pair of sample points at diagonal positions, and the sample value of each sample point in the second subset is composed of the average of another pair of sample points at diagonal positions.

[0078] Optionally, the model training module is further adapted to: input all input samples in the training samples into a self-supervised noise suppression deep learning neural network, and the self-supervised noise suppression deep learning neural network undergoes multi-level iterative noise reduction processing until the model training termination condition is met, thereby obtaining a seismic imaging profile noise suppression model. Each iteration of the denoising process performs the following steps: convolutional residual learning is performed based on the input samples; the convolutional residual learning result is then concatenated with the input samples; sparse regularization is performed based on the residual concatenation result, the previous-level auxiliary variable or the initial auxiliary variable with convergence, to obtain the model's denoised output; the auxiliary variables are updated based on the residual concatenation result and the model's denoised output; the loss function is calculated based on the model's denoised output and all target samples in the training sample pair; and the neural network parameters are optimized based on the loss function.

[0079] Optionally, the noise addition module is further adapted to: add random noise following a normal distribution to the first subset to obtain a noisy third subset, and add random noise following a normal distribution to the second subset to obtain a noisy fourth subset.

[0080] Optionally, the noise addition module is further adapted to add normally distributed random noise to the first subset and the second subset respectively using the following formula:

[0081]

[0082]

[0083]

[0084]

[0085]

[0086] Where a is any sampling point in the noisy third subset, A is any sampling point in the first subset, b is any sampling point in the noisy fourth subset, B is any sampling point in the second subset, N is the number of longitudinal sampling points in the seismic imaging profile data, and f is the noise addition intensity control factor.

[0087] The descriptions of the above modules refer to the corresponding descriptions in the method embodiments, and will not be repeated here.

[0088] The seismic imaging profile noise suppression device provided in this application embodiment can effectively improve the signal-to-noise ratio of seismic data without relying on clean labels, while maintaining good signal fidelity. This method is computationally efficient, easy to integrate into existing processing workflows, and readily applicable in actual production.

[0089] This application provides a non-volatile computer storage medium storing at least one executable instruction or computer program that enables a processor to perform the operation corresponding to the seismic imaging profile noise suppression method in any of the above method embodiments.

[0090] This application provides a computer program product, which includes at least one executable instruction or computer program that enables a processor to perform the operation corresponding to the seismic imaging profile noise suppression method in any of the above method embodiments.

[0091] Figure 6 The diagram shows a structural schematic of an embodiment of the computing device of this application. The specific embodiments of this application do not limit the specific implementation of the computing device.

[0092] like Figure 6 As shown, the computing device may include: a processor 602, a communications interface 604, a memory 606, and a communications bus 608.

[0093] The processor 602, communication interface 604, and memory 606 communicate with each other via communication bus 608. Communication interface 604 is used to communicate with other network elements such as clients or other servers. The processor 602 executes program 610, specifically performing the relevant steps in the above-described embodiment of the seismic imaging profile noise suppression method for computing devices.

[0094] Specifically, program 610 may include program code that includes computer operation instructions.

[0095] The processor 602 may be a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits configured to implement the embodiments of this application. The computing device includes one or more processors, which may be processors of the same type, such as one or more CPUs; or processors of different types, such as one or more CPUs and one or more ASICs.

[0096] Memory 606 is used to store program 610. Memory 606 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0097] Specifically, program 610 can be used to cause processor 602 to execute the seismic imaging profile noise suppression method in any of the above method embodiments. The specific implementation of each step in program 610 can be found in the corresponding descriptions of the steps and units in the above seismic imaging profile noise suppression embodiments, and will not be repeated here. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the devices and modules described above can be referred to the corresponding process descriptions in the foregoing method embodiments, and will not be repeated here.

[0098] The algorithms and displays provided herein are not inherently related to any particular computer, virtual system, or other device. Various general-purpose systems can also be used in conjunction with the teachings herein. The required structure for constructing such systems is apparent from the above description. Furthermore, the embodiments of this application are not directed to any particular programming language. It should be understood that the contents of the embodiments of this application described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best implementation of the embodiments of this application.

[0099] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0100] Similarly, it should be understood that, in order to simplify this disclosure and aid in understanding one or more of the various inventive aspects, in the foregoing description of exemplary embodiments of the present application, various features of the present application embodiments are sometimes grouped together into a single embodiment, figure, or description thereof. However, this approach to disclosure should not be construed as reflecting an intention that the claimed embodiments of the present application require more features than expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of the present application.

[0101] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0102] Furthermore, those skilled in the art will understand that although some embodiments described herein include certain features but not others included in other embodiments, combinations of features from different embodiments are meant to be within the scope of the embodiments of this application and form different embodiments. For example, in the following claims, any one of the claimed embodiments can be used in any combination.

[0103] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components according to the embodiments of this application. The embodiments of this application can also be implemented as device or apparatus programs (e.g., computer programs and computer program products) for performing part or all of the methods described herein. Such programs implementing the embodiments of this application can be stored on a computer-readable medium, or can be in the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0104] It should be noted that the above embodiments are illustrative of the embodiments of this application and not limiting of the embodiments of this application, and those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. Embodiments of this application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

Claims

1. A method for suppressing noise in seismic imaging profiles, comprising: Acquire seismic imaging profile data to be processed; Based on a preset spatial sampling method, the seismic imaging profile data to be processed is decomposed to obtain a first subset and a second subset, wherein the first subset and the second subset are spatially interleaved and complementary. Adding noise to the first subset yields a noisy third subset, and adding noise to the second subset yields a noisy fourth subset; Construct training sample pairs, wherein the training sample pairs include: using the fourth subset as input samples and the first subset as target samples; and using the third subset as input samples and the second subset as target samples; A self-supervised noise suppression deep learning neural network is constructed, and the model is trained on the self-supervised noise suppression deep learning neural network based on the training samples to obtain a seismic imaging profile noise suppression model. The seismic imaging profile noise suppression model is used to suppress noise in the seismic imaging profile data to be processed, resulting in denoised seismic imaging profile data.

2. The method according to claim 1, wherein, The step of dividing the seismic imaging profile data to be processed into a first subset and a second subset based on a preset spatial sampling method further includes: The first subset is determined based on sampling points at a pair of diagonal positions in each 2×2 neighborhood of the seismic imaging profile data to be processed; the second subset is determined based on sampling points at another pair of diagonal positions in each 2×2 neighborhood of the seismic imaging profile data to be processed.

3. The method according to claim 2, wherein, The sample value of each sampling point in the first subset is composed of the average value of the sampling points at the pair of diagonal positions, and the sample value of each sampling point in the second subset is composed of the average value of the sampling points at the other pair of diagonal positions.

4. The method according to any one of claims 1-3, wherein, The step of training the self-supervised noise suppression deep learning neural network based on the training sample pairs to obtain the seismic imaging profile noise suppression model further includes: All input samples in the training samples are input into the self-supervised noise suppression deep learning neural network. The self-supervised noise suppression deep learning neural network undergoes multi-level iterative noise reduction processing until the model training ends, thus obtaining the seismic imaging profile noise suppression model. Each iteration of the denoising process performs the following steps: convolutional residual learning is performed based on the input samples; the convolutional residual learning result is then concatenated with the input samples; sparse regularization is performed based on the residual concatenation result, the previous-level auxiliary variable or the initial auxiliary variable that represents convergence, to obtain the model's denoised output result; the auxiliary variables are updated based on the residual concatenation result and the model's denoised output result; a loss function is calculated based on the model's denoised output result and all target samples in the training sample pair; and the neural network parameters are optimized based on the loss function.

5. The method according to any one of claims 1-3, wherein, The step of adding noise to the first subset to obtain a noisy third subset, and adding noise to the second subset to obtain a noisy fourth subset, further includes: Adding normally distributed random noise to the first subset yields a noisy third subset, and adding normally distributed random noise to the second subset yields a noisy fourth subset.

6. The method according to claim 5, wherein, The step of adding normally distributed random noise to the first subset to obtain a noisy third subset, and adding normally distributed random noise to the second subset to obtain a noisy fourth subset, further includes: The following formulas are used to add normally distributed random noise to the first subset and the second subset respectively: Where a is any sampling point in the noisy third subset, A is any sampling point in the first subset, b is any sampling point in the noisy fourth subset, B is any sampling point in the second subset, N is the number of longitudinal sampling points in the seismic imaging profile data, and f is the noise addition intensity control factor.

7. A seismic imaging profile noise suppression device, comprising: The acquisition module is suitable for acquiring seismic imaging profile data to be processed. The decomposition module is adapted to decompose the seismic imaging profile data to be processed based on a preset spatial sampling method to obtain a first subset and a second subset, wherein the first subset and the second subset are spatially interleaved and complementary. The noise addition module is adapted to add noise to the first subset to obtain a noisy third subset, and to add noise to the second subset to obtain a noisy fourth subset; A training sample construction module is adapted to construct training sample pairs, wherein the training sample pairs include: using the fourth subset as input samples and the first subset as target samples; and using the third subset as input samples and the second subset as target samples; The model training module is suitable for constructing a self-supervised noise suppression deep learning neural network. Based on the training samples, the self-supervised noise suppression deep learning neural network is trained to obtain a seismic imaging profile noise suppression model. The noise suppression processing module is adapted to use the seismic imaging profile noise suppression model to perform noise suppression processing on the seismic imaging profile data to be processed, so as to obtain denoised seismic imaging profile data.

8. A computing device, comprising: The processor, memory, communication interface, and communication bus are provided, wherein the processor, memory, and communication interface communicate with each other via the communication bus. The memory is used to store at least one executable instruction that causes the processor to perform the operation corresponding to the seismic imaging profile noise suppression method as described in any one of claims 1-6.

9. A computer storage medium storing at least one executable instruction that causes a processor to perform an operation corresponding to the seismic imaging profile noise suppression method as described in any one of claims 1-6.

10. A computer program product comprising at least one executable instruction that causes a processor to perform an operation corresponding to the seismic imaging profile noise suppression method as described in any one of claims 1-6.

Images