Method and system for improving the signal-to-noise ratio of first arrival wave, and readable storage medium

Abstract

This application provides a method and system for improving the signal-to-noise ratio (SNR) of first-arrival waves, as well as a readable storage medium. The method includes: acquiring single-shot seismic data; mapping the original energy values ​​to a preset weighted interval using a nonlinear mapping function; performing PCA analysis on the weighted block set, fitting the data using incremental PCA, and mapping the high-dimensional block vector to a low-dimensional space when the cumulative explained variance ratio reaches a preset threshold; determining initial dictionary atoms based on the low-dimensional block vector, performing sparse coding using the OMP algorithm, and iteratively updating the dictionary using K-SVD; applying a two-dimensional Hanning window to each reconstructed sub-block; calculating the residual between the preliminary first-arrival denoised data and the original energy values, performing bandpass filtering on the residual when the residual energy quantile exceeds the first-arrival signal energy quantile, and proportionally superimposing the filtered residual onto the preliminary first-arrival denoised data. This technical solution can accurately suppress noise and effectively improve the quality of low SNR first-arrival data.

Description

Technical Field

[0001] This application relates to the field of geophysical exploration technology, and more specifically, to a method and system for improving the signal-to-noise ratio of first arrival waves, and a readable storage medium. Background Technology

[0002] In petroleum geophysical exploration, the first arrival wave, as the earliest signal to reach the receiver in the seismic wavefield, contains crucial information such as the shallow structure and velocity distribution of the subsurface medium. It is the core foundational data for subsequent processing, including static correction calculations, near-surface velocity modeling, and seismic data imaging. Its signal-to-noise ratio (SNR) directly determines the accuracy of exploration data processing and has a significant impact on key aspects such as oil and gas reservoir prediction and structural interpretation. Therefore, obtaining high SNR first arrival wave data is a prerequisite for ensuring the reliability of exploration results.

[0003] However, in actual exploration environments, first-arrival signals often face severe interference from complex noise. On the one hand, strong energy noise such as surface interference and surface waves can easily mask weak first-arrival signals, leading to an imbalance in the energy distribution of the spatial domain, with first-arrival signals in weak energy areas being submerged by strong interference. On the other hand, the widespread presence of random noise and instrument noise makes the characteristics of first-arrival signals in low signal-to-noise ratio areas blurred and difficult to identify. These problems directly lead to a decrease in the signal-to-noise ratio of the first-arrival, posing a significant challenge to subsequent data processing.

[0004] Existing first-arrival denoising methods have several limitations in practice: First, energy equalization processing is insufficient; traditional methods struggle to effectively suppress strong energy interference and tend to excessively weaken the first-arrival signal, failing to achieve precise energy balance in the spatial domain. Second, high-dimensional seismic data contains a lot of redundant information; existing dimensionality reduction methods struggle to retain the core features of the first-arrival while compressing data dimensions, resulting in limited accuracy in subsequent signal representation. Third, dictionary learning techniques often fail to accurately characterize the waveform features of the first-arrival due to insufficient stability of sparse coding and interference from weakly correlated signals during signal modeling. Fourth, during sub-block processing, boundary artifacts are prominent, and signal abrupt changes at sub-block splicing points can disrupt the spatial continuity of the first arrival wave. Fifth, the residual processing mechanism is lacking, and some effective first arrival signals cannot be recovered after being filtered out by noise, further resulting in signal energy loss. Sixth, parameter settings are mostly fixed values, making it difficult to adapt to the processing needs of data with different signal-to-noise ratios. It is difficult to achieve a dynamic balance between noise suppression and signal preservation, especially in low signal-to-noise ratio scenarios, where "past noise" can easily lead to signal loss or "insufficient denoising" can lead to noise residue. Summary of the Invention

[0005] This application aims to solve or improve the aforementioned technical problems.

[0006] Therefore, the primary objective of this application is to provide a method for improving the signal-to-noise ratio of the first arrival wave.

[0007] The second objective of this application is to provide a system for improving the signal-to-noise ratio of the first arrival wave.

[0008] The third objective of this application is to provide a system for improving the signal-to-noise ratio of the first arrival wave.

[0009] The fourth objective of this application is to provide a readable storage medium.

[0010] To achieve the first objective of this application, the technical solution of the first aspect of this application provides a method for improving the signal-to-noise ratio of first arrival waves, comprising: exciting and acquiring single-shot seismic data, wherein the single-shot seismic data includes receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines; acquiring single-trace energy through the root mean square algorithm, mapping the original energy value of the single-trace energy to a preset weight interval through a nonlinear mapping function, dividing the single-shot seismic data into blocks and weighting them element-wise according to the energy weight of the center trace to obtain a block set; performing PCA analysis on the weighted block set, calculating the mean of the block set and decentering it, fitting the data through incremental PCA, determining the number of principal components of the first arrival wave data when the cumulative explained variance ratio reaches a preset threshold, and mapping the high-dimensional block vector to a low-dimensional space according to the number of principal components of the first arrival wave data to obtain a low-dimensional block vector; determining the initial dictionary atoms according to the low-dimensional block vector, performing sparse coding through a regularized OMP algorithm, and introducing dynamic stopping... The threshold is set and weakly correlated signal encoding is skipped. The dictionary is updated iteratively using regularized K-SVD. Atoms that fail to decompose singular values ​​are reinitialized to obtain reconstructed sub-blocks. A two-dimensional Hanning window is applied to each reconstructed sub-block to obtain windowed sub-blocks. The windowed sub-blocks are superimposed according to their original positions and the window weights are accumulated. After weight normalization, preliminary first-arrival denoised data is generated. The residual between the preliminary first-arrival denoised data and the original energy value is calculated, and the residual energy quantile and the first-arrival signal energy quantile are extracted. When the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, the residual is bandpass filtered. The fusion factor is determined according to the energy ratio of the residual to the first-arrival signal. The filtered residual is superimposed proportionally to the preliminary first-arrival denoised data. The basic parameters are calculated based on the data dimension. The sparsity and regularization intensity are dynamically adjusted through the first-arrival data protection factor and regulation factor to obtain the parameter combination. The parameter combination is iteratively tested according to the preset first-arrival signal-to-noise ratio to obtain the optimal parameter combination.

[0011] According to the first-arrival signal-to-noise ratio (SNR) enhancement method provided in this application, firstly, single-shot seismic data is excited and collected, including receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines. Then, the energy characteristic values ​​of each trace are calculated, and the root mean square (RMS) algorithm is used to obtain the energy of each trace. The original energy values ​​are mapped to a weighted interval using a nonlinear mapping function. The seismic data is divided into blocks according to a preset size, and element-wise weighting is performed based on the energy weight of the central trace in each block to achieve spatial domain energy equalization. Then, PCA (principal component analysis) is performed on the weighted block set to calculate the mean of the block set and decenter it. Incremental PCA is used to fit the data, and when the cumulative explained variance ratio reaches a preset threshold, the number of principal components in the first-arrival data is determined. Based on the selected number of principal components in the first-arrival data, the high-dimensional block vector is mapped to a low-dimensional space, preserving the characteristics of the first-arrival signal while reducing the data dimensionality. Then, initial dictionary atoms are determined based on the low-dimensional block vectors. Sparse coding is performed using a regularized OMP algorithm, introducing a dynamic stopping threshold and skipping weakly correlated signal coding. The dictionary is updated iteratively using regularized K-SVD, and atoms that fail to decompose singular values ​​are reinitialized to enhance the dictionary's ability to represent seismic signals. A two-dimensional Hanning window matching the size of the data sub-blocks is generated, constructed using the outer product of cosine functions. The Hanning window is applied to each reconstructed sub-block to suppress abrupt signal changes at the sub-block edges. The windowed sub-blocks are stacked in their original positions, and the window weights are accumulated synchronously. The weight-normalized stacking result is divided by the accumulated weights to generate preliminary first-arrival denoised data. This preliminary first-arrival denoised data, generated through weight normalization, can reduce boundary artifacts caused by sub-block stitching and improve the spatial continuity of data reconstruction. Then, the residual between the preliminary first-arrival denoising result and the original data is calculated. The energy quantiles of the residual and the first-arrival signal are extracted. When the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, bandpass filtering is applied to the residual. A fusion factor is determined based on the energy ratio of the residual to the first-arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first-arrival denoising data to recover the effective first-arrival signal contained within the residual. Finally, basic parameters are calculated based on the data dimension, including patch size, step size, and dictionary capacity, determined according to the number of channels and sampling points. A signal protection factor is set, and the sparsity is dynamically reduced by adjusting the factor, while the regularization intensity is adjusted proportionally. Multiple parameter combinations are iteratively tested, and the optimal configuration is selected using the signal-to-noise ratio of the first-arrival data as the evaluation metric to achieve adaptive denoising and balance noise suppression and signal preservation.By suppressing strong energy interference through spatial domain energy equalization, preserving core features through PCA dimensionality reduction, accurately representing signals through regularized dictionary learning, suppressing boundary artifacts through Hanning window weighting, recovering effective signals through residual fusion, and optimizing adaptive parameters, a precise balance between efficient noise suppression and first-arrival signal fidelity is achieved, thereby improving the first-arrival signal-to-noise ratio. This provides high-quality basic data support for subsequent exploration data processing and meets the needs of exploration production for high-precision first-arrival analysis.

[0012] In some technical solutions, optionally, the single-channel energy is obtained through the root mean square algorithm, and the original energy value of the single-channel energy is mapped to a preset weight range through a nonlinear mapping function. This includes: obtaining the single-channel energy through the root mean square algorithm, calculating global energy parameters, which include a maximum energy value and a minimum energy value; converting the original channel energy value into weight coefficients through a nonlinear mapping function based on the single-channel energy, the maximum energy value, and the minimum energy value; wherein the maximum energy value is the maximum value among all channel energies, and the minimum energy value is the energy value of the first preset percentile among all channel energies, where the first preset percentile is 5%-15%.

[0013] In this technical solution, the single-channel energy is obtained through the root mean square (RMS) algorithm. A nonlinear mapping function then maps the original energy value of each channel to a preset weighted range. Specifically, the single-channel energy is first obtained using the RMS algorithm, and global energy parameters are calculated, including a maximum energy value and a minimum energy value. Then, based on the single-channel energy, the maximum energy value, and the minimum energy value, the original channel energy value is converted into weighting coefficients using the nonlinear mapping function. The maximum energy value is the highest value among all channel energies, and the minimum energy value is the energy value at the first preset percentile among all channel energies. The first preset percentile is 5%-15%, which avoids noise interference on the low signal-to-noise ratio first arrival wave, discarding the lowest extreme noise channel to ensure that the minimum energy represents a true weak signal rather than noise.

[0014] In some technical solutions, optionally, the single-channel energy is obtained through the root mean square algorithm, and the original energy value of the single-channel energy is mapped to a preset weight range through a nonlinear mapping function. It also includes: a preset weight range obtained by threshold clipping, wherein the weight range of the preset weight range is greater than or equal to 0 and less than or equal to 1.

[0015] In this technical solution, the original energy value of a single energy channel is mapped to a preset weight range through a nonlinear mapping function. This also includes a preset weight range obtained through threshold pruning, forming a differentiated weight system that enhances weak energy channels and moderately suppresses strong energy channels. The weight range of the preset weight range is greater than or equal to 0 and less than or equal to 1. Specifically, the weight range of the preset weight range is greater than or equal to 0.3 and less than or equal to 1.

[0016] In some technical solutions, optionally, the range of the preset threshold is greater than or equal to 0.7 and less than or equal to 0.9.

[0017] In this technical solution, the preset threshold ranges from 0.7 to 0.9. Specifically, the preset threshold is 0.8, which effectively preserves the main features of the first arrival signal and reduces the interference of redundant information on subsequent dictionary learning.

[0018] In some technical solutions, optionally, initial dictionary atoms are determined based on low-dimensional block vectors, and sparse coding is performed using a regularized OMP algorithm. A dynamic stopping threshold is introduced, and weakly correlated signal coding is skipped. This includes: randomly selecting samples from the low-dimensional block vectors as initial dictionary atoms, performing sparse coding using a regularized OMP algorithm, stopping iteration when the L2 norm of the difference between the original signal and the current approximate signal is less than a second preset percentage of the norm of the original signal; skipping the initial dictionary atoms when the correlation between the original signal and the initial dictionary atoms is less than a preset correlation value, and not including the initial dictionary atoms in the candidate atom set for subsequent coding; and adding regularization terms to the diagonal of the Gram matrix during the solution process.

[0019] In this technical solution, initial dictionary atoms are determined based on the low-dimensional block vector. Sparse coding is performed using a regularized OMP algorithm. A dynamic stopping threshold is introduced, and weakly correlated signal coding is skipped. Specifically, samples are randomly selected from the low-dimensional block vector as initial dictionary atoms, and sparse coding is performed using a regularized OMP algorithm. Iteration stops when the L2 norm of the difference between the original signal and the current approximate signal is less than a second preset percentage of the norm of the original signal. The second preset percentage can be... Initial correlation refers to the degree of matching between the original signal and a certain atom in the dictionary. If the correlation between an atom and the signal is less than a preset correlation value, i.e., almost no correlation, the atom is skipped and not included in the candidate atom set for subsequent encoding. The preset correlation value can be 0.05. Finally, during the solution process, a regularization term is added to the diagonal of the Gram matrix to improve the solution stability. The default value for the regularization term is 1e-6.

[0020] In some technical solutions, the residual energy quantile value and the first arrival signal energy quantile value are optionally the same; the bandpass filter value range is 5Hz-100Hz.

[0021] In this technical solution, the residual energy quantile and the first-arrival signal energy quantile have the same quantile, which can be 95. The lower the first-arrival signal-to-noise ratio (SNR), the larger the value, enhancing noise sensitivity; conversely, the higher the first-arrival SNR, the smaller the value, improving sensitivity to weak first-arrival signals. When the residual energy quantile exceeds a third preset percentage of the first-arrival signal energy quantile, the third preset percentage can be... The higher the first-arrival signal-to-noise ratio (SNR), the smaller the threshold adjustment formula (10% × (1 - SNR / 50)). Conversely, the lower the first-arrival SNR, the larger the threshold adjustment formula (10% × (1 + SNR / 20)). SNR is the SNR of the first-arrival data. The bandpass filter value range is determined based on the frequency characteristics of the first-arrival data, with a default range of 5Hz-100Hz. The fusion factor is negatively correlated with the SNR of the first-arrival data; for every 10dB decrease in SNR, the fusion weight increases by 50%, with an upper limit of 0.3 set to prevent noise amplification.

[0022] In some technical solutions, optionally, the value range of the first arrival wave data protection factor is 0.5-0.9; and the value range of the regulation factor is 0.6-0.9.

[0023] In this technical solution, the first-arrival data protection factor is related to the signal-to-noise ratio (SNR) of the first-arrival data. The lower the SNR of the first-arrival data, the larger the value. It preserves weak first-arrival characteristics in a strong noise environment, and its value ranges from 0.5 to 0.9. The adjustment factor is a scaling factor for the basic sparsity, which is also related to the SNR of the first-arrival data. The lower the SNR, the larger the value, and its value ranges from 0.6 to 0.9.

[0024] To achieve the second objective of this application, the technical solution of the second aspect of this application provides a first-arrival signal-to-noise ratio (SNR) enhancement system, comprising: an acquisition module for exciting and acquiring single-shot seismic data, the single-shot seismic data including receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines; a mapping module for acquiring single-trace energy through a root mean square (RMS) algorithm, mapping the original energy value of the single-trace energy to a preset weight interval through a nonlinear mapping function, dividing the single-shot seismic data into blocks and weighting them element-wise according to the energy weight of the central trace to obtain a block set; an analysis module for performing PCA analysis on the weighted block set, calculating the mean of the block set and decentering it, fitting the data through incremental PCA, determining the number of principal components of the first-arrival data when the cumulative explained variance ratio reaches a preset threshold, and mapping the high-dimensional block vector to a low-dimensional space based on the number of principal components of the first-arrival data to obtain a low-dimensional block vector; and a reconstruction module for determining initial dictionary atoms based on the low-dimensional block vector, performing sparse coding through a regularized OMP algorithm, and introducing... The system dynamically stops the threshold and skips weakly correlated signal encoding. It iteratively updates the dictionary using regularized K-SVD, reinitializing atoms that failed in singular value decomposition to obtain reconstructed sub-blocks. A generation module applies a two-dimensional Hanning window to each reconstructed sub-block, resulting in windowed sub-blocks. These windowed sub-blocks are then superimposed at their original positions, and window weights are accumulated. Weight normalization is performed to generate preliminary first-arrival denoised data. A calculation module calculates the residual between the preliminary first-arrival denoised data and the original energy value, extracting the residual energy quantile and the first-arrival signal energy quantile. When the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, bandpass filtering is applied to the residual. A fusion factor is determined based on the energy ratio of the residual to the first-arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first-arrival denoised data. An iteration module calculates basic parameters based on the data dimension, dynamically adjusting sparsity and regularization strength using first-arrival data protection and control factors to obtain parameter combinations. The optimal parameter combination is obtained by iteratively testing the parameter combinations based on a preset first-arrival signal-to-noise ratio.

[0025] The first-arrival signal-to-noise ratio (SNR) enhancement system provided in this application includes an acquisition module, a mapping module, an analysis module, a reconstruction module, a generation module, a calculation module, and an iteration module. The acquisition module is used to excite and acquire single-shot seismic data, which includes receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines. The mapping module is used to acquire the energy of each trace using a root mean square (RMS) algorithm, and to map the raw energy value of each trace to a preset weighted interval using a nonlinear mapping function. The single-shot seismic data is then divided into blocks and weighted element-wise according to the energy weight of the central trace to obtain a block set. The analysis module performs PCA analysis on the weighted block set, calculates the mean of the block set and decenters it, fits the data using incremental PCA, determines the number of principal components of the first-arrival data when the cumulative explained variance ratio reaches a preset threshold, and maps the high-dimensional block vector to a low-dimensional space based on the number of principal components of the first-arrival data to obtain a low-dimensional block vector. The reconstruction module determines initial dictionary atoms based on the low-dimensional block vector, performs sparse coding using a regularized OMP algorithm, introduces a dynamic stopping threshold and skips weakly correlated signal coding, iterates and updates the dictionary using regularized K-SVD, and reinitializes atoms that fail to decompose singular values, resulting in reconstructed sub-blocks. The generation module applies a two-dimensional Hanning window to each reconstructed sub-block, resulting in windowed sub-blocks. These windowed sub-blocks are then superimposed at their original positions, and window weights are accumulated. After weight normalization, preliminary first-arrival denoised data is generated. The calculation module calculates the residual between the preliminary first-arrival denoised data and the original energy value, extracts the residual energy quantile and the first-arrival signal energy quantile, performs bandpass filtering on the residual when the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, determines the fusion factor based on the energy ratio of the residual to the first-arrival signal, and proportionally superimposes the filtered residual onto the preliminary first-arrival denoised data. The iterative module calculates fundamental parameters based on data dimensions. It dynamically adjusts sparsity and regularization intensity using first-arrival data protection and control factors to obtain parameter combinations. Based on a preset first-arrival signal-to-noise ratio (SNR), iterative testing of these parameter combinations yields the optimal combination. Through spatial domain energy equalization to suppress strong energy interference, PCA dimensionality reduction to preserve core features, regularized dictionary learning to accurately represent the signal, Hanning window weighting to suppress boundary artifacts, residual fusion to recover effective signals, and adaptive parameter optimization, a precise balance between efficient noise suppression and first-arrival signal fidelity is achieved. This improves the first-arrival SNR, providing high-quality fundamental data support for subsequent exploration data processing and meeting the high-precision first-arrival analysis requirements of exploration production.

[0026] To achieve the third objective of this application, the technical solution of the third aspect of this application provides a first-arrival signal-to-noise ratio (SNR) enhancement system, comprising: a memory and a processor, wherein the memory stores a program or instructions that can be executed on the processor, and when the processor executes the program or instructions, it implements the first-arrival SNR enhancement method of any one of the technical solutions of the first aspect, thus having the technical effects of any one of the technical solutions of the first aspect, which will not be elaborated here.

[0027] To achieve the fourth objective of this application, the technical solution of the fourth aspect of this application provides a readable storage medium storing a program or instructions thereon. When the program or instructions are executed by a processor, they implement the steps of the first arrival signal-to-noise ratio enhancement method of any one of the technical solutions of the first aspect, and thus have the technical effects of any one of the technical solutions of the first aspect, which will not be repeated here.

[0028] Additional aspects and advantages of this application will become apparent in the following description or may be learned by practice of this application. Attached Figure Description

[0029] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the description of the embodiments taken in conjunction with the following drawings, in which: Figure 1 This is one of the step flow diagrams of a first-arrival signal-to-noise ratio enhancement method according to an embodiment of this application; Figure 2 This is a second schematic flowchart illustrating the steps of a first-arrival signal-to-noise ratio enhancement method according to an embodiment of this application; Figure 3 This is the third step of a flowchart illustrating the method for improving the signal-to-noise ratio of the first arrival wave according to an embodiment of this application; Figure 4 This is a fourth schematic flowchart illustrating the steps of a first-arrival signal-to-noise ratio enhancement method according to an embodiment of this application; Figure 5 This is a schematic block diagram of the first-arrival signal-to-noise ratio enhancement system according to one embodiment of this application; Figure 6 This is a schematic block diagram of the first-arrival signal-to-noise ratio enhancement system according to another embodiment of this application; Figure 7 This is a schematic diagram of an observation system for a first-arrival signal-to-noise ratio enhancement method according to an embodiment of this application; Figure 8 This is a schematic diagram of single-shot data collected by the observation system for an embodiment of the first-arrival signal-to-noise ratio enhancement method of this application; Figure 9 This is a schematic diagram of the result after denoising single-shot data in a first-arrival signal-to-noise ratio enhancement method according to an embodiment of this application.

[0030] 10: First arrival signal-to-noise ratio enhancement system; 110: Acquisition module; 120: Mapping module; 130: Analysis module; 140: Reconstruction module; 150: Generation module; 160: Calculation module; 170: Iteration module; 20: First arrival signal-to-noise ratio enhancement system; 300: Memory; 400: Processor. Detailed Implementation

[0031] To better understand the above-mentioned objectives, features, and advantages of this application, the application will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in these embodiments can be combined with each other.

[0032] Many specific details are set forth in the following description in order to provide a full understanding of this application. However, this application may also be implemented in other ways different from those described herein. Therefore, the scope of protection of this application is not limited to the specific embodiments disclosed below.

[0033] The following reference Figures 1 to 9 This application describes a method and system for improving the signal-to-noise ratio of first arrival waves, as well as a readable storage medium, according to some embodiments of this application.

[0034] like Figure 1 As shown, an embodiment of the first aspect of this application provides a method for improving the signal-to-noise ratio of the first arrival wave, comprising the following steps: Step S102: Excite and acquire single-shot seismic data, which includes receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines; Step S104: Obtain the energy of a single trace using the root mean square algorithm, map the original energy value of the single trace to a preset weight range using a nonlinear mapping function, divide the single-shot seismic data into blocks and perform element-wise weighting based on the energy weight of the central trace to obtain a block set; Step S106: Perform PCA analysis on the weighted block set, calculate the mean of the block set and decenter it, fit the data through incremental PCA, determine the number of principal components of the first arrival data when the cumulative explained variance ratio reaches the preset threshold, and map the high-dimensional block vector to the low-dimensional space according to the number of principal components of the first arrival data to obtain the low-dimensional block vector. Step S108: Determine the initial dictionary atoms based on the low-dimensional block vector, perform sparse coding through the regularized OMP algorithm, introduce a dynamic stopping threshold and skip the coding of weakly correlated signals, update the dictionary iteratively through regularized K-SVD, reinitialize the atoms that failed the singular value decomposition, and obtain the reconstructed sub-blocks. Step S110: Apply a two-dimensional Hanning window to each reconstructed sub-block to obtain windowed sub-blocks. Stack the windowed sub-blocks according to their original positions and accumulate the window weights. Generate preliminary first arrival denoised data by weight normalization. Step S112: Calculate the residual between the preliminary first arrival denoised data and the original energy value, and extract the residual energy quantile and the first arrival signal energy quantile. When the residual energy quantile exceeds the third preset percentage of the first arrival signal energy quantile, bandpass filter is applied to the residual. The fusion factor is determined according to the energy ratio of the residual to the first arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first arrival denoised data. Step S114: Calculate basic parameters based on data dimensions, dynamically adjust sparsity and regularization intensity through first-arrival data protection factor and regulation factor to obtain parameter combination, and iteratively test parameter combination according to preset first-arrival signal-to-noise ratio to obtain optimal parameter combination.

[0035] According to the first-arrival signal-to-noise ratio (SNR) enhancement method provided in this embodiment, firstly, single-shot seismic data is excited and collected, including receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines. Then, the energy characteristic values ​​of each trace are calculated, and the root mean square (RMS) algorithm is used to obtain the energy of each trace. The original energy values ​​are mapped to a weighted interval using a nonlinear mapping function. The seismic data is divided into blocks according to a preset size, and element-wise weighting is performed based on the energy weight of the central trace in each block to achieve spatial domain energy equalization. Then, PCA analysis is performed on the weighted block set to calculate the mean of the block set and decenter it. Incremental PCA is used to fit the data, and when the cumulative explained variance ratio reaches a preset threshold, the number of principal components of the first-arrival data is determined. Based on the selected number of principal components of the first-arrival data, the high-dimensional block vector is mapped to a low-dimensional space, preserving the characteristics of the first-arrival signal while reducing the data dimensionality. Then, initial dictionary atoms are determined based on the low-dimensional block vectors. Sparse coding is performed using a regularized OMP algorithm, introducing a dynamic stopping threshold and skipping weakly correlated signal coding. The dictionary is updated iteratively using regularized K-SVD, and atoms that fail to decompose singular values ​​are reinitialized to enhance the dictionary's ability to represent seismic signals. A two-dimensional Hanning window matching the size of the data sub-blocks is generated, constructed using the outer product of cosine functions. The Hanning window is applied to each reconstructed sub-block to suppress abrupt signal changes at the sub-block edges. The windowed sub-blocks are stacked in their original positions, and the window weights are accumulated synchronously. The weight-normalized stacking result is divided by the accumulated weights to generate preliminary first-arrival denoised data. This preliminary first-arrival denoised data, generated through weight normalization, can reduce boundary artifacts caused by sub-block stitching and improve the spatial continuity of data reconstruction. Then, the residual between the preliminary first-arrival denoising result and the original data is calculated. The energy quantiles of the residual and the first-arrival signal are extracted. When the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, bandpass filtering is applied to the residual. A fusion factor is determined based on the energy ratio of the residual to the first-arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first-arrival denoising data to recover the effective first-arrival signal contained within the residual. Finally, basic parameters are calculated based on the data dimension, including patch size, step size, and dictionary capacity, determined according to the number of channels and sampling points. A signal protection factor is set, and the sparsity is dynamically reduced by adjusting the factor, while the regularization intensity is adjusted proportionally. Multiple parameter combinations are iteratively tested, and the optimal configuration is selected using the signal-to-noise ratio of the first-arrival data as the evaluation metric to achieve adaptive denoising and balance noise suppression and signal preservation. By suppressing strong energy interference through spatial domain energy equalization, preserving core features through PCA dimensionality reduction, accurately representing signals through regularized dictionary learning, suppressing boundary artifacts through Hanning window weighting, recovering effective signals through residual fusion, and optimizing adaptive parameters, a precise balance between efficient noise suppression and first-arrival signal fidelity is achieved, thereby improving the first-arrival signal-to-noise ratio. This provides high-quality basic data support for subsequent exploration data processing and meets the needs of exploration production for high-precision first-arrival analysis.

[0036] like Figure 2 As shown, according to an embodiment of the first-arrival signal-to-noise ratio enhancement method proposed in this application, the single-channel energy is obtained through the root mean square algorithm, and the original energy value of the single-channel energy is mapped to a preset weight interval through a nonlinear mapping function, including the following steps: Step S202: Obtain the single-channel energy using the root mean square algorithm and calculate the global energy parameters, which include the maximum energy value and the minimum energy value; Step S204: Based on the single-channel energy, maximum energy value, and minimum energy value, convert the original channel energy value into weighting coefficients through a nonlinear mapping function; Among them, the maximum energy value is the maximum value among all Dao energies, and the minimum energy value is the energy value of the first preset percentile among all Dao energies, which is 5%-15%.

[0037] In this embodiment, the single-channel energy is obtained using the root mean square (RMS) algorithm. A nonlinear mapping function then maps the original energy value of each channel to a preset weighted range. Specifically, the single-channel energy is first obtained using the RMS algorithm, and global energy parameters are calculated, including a maximum energy value and a minimum energy value. Then, based on the single-channel energy, the maximum energy value, and the minimum energy value, the original channel energy value is converted into weighting coefficients using the nonlinear mapping function. The maximum energy value is the highest value among all channel energies, and the minimum energy value is the energy value at the first preset percentile among all channel energies. The first preset percentile is 5%-15%, which avoids noise interference on the low signal-to-noise ratio first arrival wave, discarding the lowest extreme noise channel to ensure that the minimum energy represents a real weak signal rather than noise.

[0038] like Figure 3 As shown, the first-arrival signal-to-noise ratio enhancement method according to an embodiment of this application obtains the single-channel energy through the root mean square algorithm, maps the original energy value of the single-channel energy to a preset weight interval through a nonlinear mapping function, and further includes the following steps: Step S302: The preset weight range obtained by threshold clipping has a weight range greater than or equal to 0 and less than or equal to 1.

[0039] In this embodiment, the original energy value of a single energy channel is mapped to a preset weight range using a nonlinear mapping function. The preset weight range is further obtained through threshold pruning, forming a differentiated weight system that enhances weak energy channels and moderately suppresses strong energy channels. The weight range of the preset weight range is greater than or equal to 0 and less than or equal to 1. Specifically, the weight range of the preset weight range is greater than or equal to 0.3 and less than or equal to 1.

[0040] In some embodiments, optionally, the preset threshold ranges from greater than or equal to 0.7 to less than or equal to 0.9. Specifically, a preset threshold of 0.8 effectively preserves the main features of the first arrival signal and reduces the interference of redundant information on subsequent dictionary learning.

[0041] like Figure 4 As shown, the first-arrival signal-to-noise ratio (SNR) enhancement method according to an embodiment of this application determines the initial dictionary atoms based on the low-dimensional block vector, performs sparse coding using the regularized OMP algorithm, introduces a dynamic stopping threshold, and skips the coding of weakly correlated signals. Specifically, it includes the following steps: Step S402: Randomly select samples from the low-dimensional block vector as initial dictionary atoms, and use the regularized OMP algorithm for sparse coding. Stop the iteration when the L2 norm of the difference between the original signal and the current approximate signal is less than the second preset percentage of the norm of the original signal. Step S404: When the correlation between the original signal and the initial dictionary atom is less than the preset correlation value, skip the initial dictionary atom and do not include the initial dictionary atom in the candidate atom set for subsequent encoding; Step S406: Add regularization terms to the diagonal of the Gram matrix during the solution process.

[0042] In this embodiment, initial dictionary atoms are determined based on the low-dimensional block vector. Sparse coding is performed using a regularized OMP algorithm. A dynamic stopping threshold is introduced, and weakly correlated signal coding is skipped. Specifically, samples are randomly selected from the low-dimensional block vector as initial dictionary atoms, and sparse coding is performed using a regularized OMP algorithm. Iteration stops when the L2 norm of the difference between the original signal and the current approximate signal is less than a second preset percentage of the norm of the original signal. The second preset percentage can be... Initial correlation refers to the degree of matching between the original signal and a certain atom in the dictionary. If the correlation between an atom and the signal is less than a preset correlation value, i.e., almost no correlation, the atom is skipped and not included in the candidate atom set for subsequent encoding. The preset correlation value can be 0.05. Finally, a regularization term is added to the diagonal of the Gram matrix during the solution process to improve the solution stability. The default value for the regularization term is 1e-6.

[0043] In some embodiments, optionally, the residual energy quantile value and the first-arrival signal energy quantile value have the same quantile, which can be 95. A lower first-arrival signal-to-noise ratio (SNR) results in a larger value, enhancing noise sensitivity; conversely, a higher first-arrival SNR results in a smaller value, improving sensitivity to weak first-arrival signals. When the residual energy quantile value exceeds a third preset percentage of the first-arrival signal energy quantile value, the third preset percentage can be... The higher the first-arrival signal-to-noise ratio (SNR), the smaller the threshold adjustment formula (10% × (1 - SNR / 50)). Conversely, the lower the first-arrival SNR, the larger the threshold adjustment formula (10% × (1 + SNR / 20)). SNR is the SNR of the first-arrival data. The bandpass filter value range is determined based on the frequency characteristics of the first-arrival data, with a default range of 5Hz-100Hz. The fusion factor is negatively correlated with the SNR of the first-arrival data; for every 10dB decrease in SNR, the fusion weight increases by 50%, with an upper limit of 0.3 set to prevent noise amplification.

[0044] In some embodiments, optionally, the first-arrival data protection factor is related to the signal-to-noise ratio (SNR) of the first-arrival data; the lower the SNR of the first-arrival data, the larger the value, preserving weak first-arrival characteristics in a strong noise environment; the value is 0.5-0.9. The adjustment factor is a preset factor used to scale the basic sparsity; it is related to the SNR of the first-arrival data; the lower the SNR, the larger the value; the value is 0.6-0.9.

[0045] like Figure 5As shown, an embodiment of the second aspect of this application provides a first-arrival signal-to-noise ratio (SNR) enhancement system 10, comprising: an acquisition module 110 for exciting and acquiring single-shot seismic data, the single-shot seismic data including receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines; a mapping module 120 for acquiring single-trace energy through a root mean square (RMS) algorithm, mapping the original energy value of the single-trace energy to a preset weight interval through a nonlinear mapping function, dividing the single-shot seismic data into blocks and weighting them element-wise according to the center trace energy weight to obtain a block set; an analysis module 130 for performing PCA analysis on the weighted block set, calculating the mean of the block set and decentering it, fitting the data through incremental PCA, determining the number of principal components of the first-arrival data when the cumulative explained variance ratio reaches a preset threshold, and mapping the high-dimensional block vector to a low-dimensional space according to the number of principal components of the first-arrival data to obtain a low-dimensional block vector; and a reconstruction module 140 for determining initial dictionary atoms based on the low-dimensional block vector, performing sparse coding through a regularized OMP algorithm, and introducing dynamic... The process involves stopping the threshold and skipping weakly correlated signal encoding, updating the dictionary through regularized K-SVD iterations, reinitializing atoms that failed in singular value decomposition, and obtaining reconstructed sub-blocks. A generation module 150 applies a two-dimensional Hanning window to each reconstructed sub-block to obtain windowed sub-blocks. These windowed sub-blocks are then superimposed at their original positions, and window weights are accumulated. Weight normalization is then performed to generate preliminary first-arrival denoised data. A calculation module 160 calculates the residual between the preliminary first-arrival denoised data and the original energy value, and extracts the residual energy quantile and the first-arrival signal energy. The residual energy quantile value is used to perform bandpass filtering on the residual when it exceeds the third preset percentage of the energy quantile value of the first arrival signal. The fusion factor is determined based on the energy ratio of the residual to the first arrival signal. The filtered residual is then proportionally superimposed on the initial first arrival denoised data. The iteration module 170 is used to calculate basic parameters based on the data dimension. The sparsity and regularization intensity are dynamically adjusted through the first arrival data protection factor and the control factor to obtain the parameter combination. The optimal parameter combination is obtained by iteratively testing the parameter combination based on the preset first arrival signal-to-noise ratio.

[0046] The first-arrival signal-to-noise ratio (SNR) enhancement system 10 provided in this embodiment includes an acquisition module 110, a mapping module 120, an analysis module 130, a reconstruction module 140, a generation module 150, a calculation module 160, and an iteration module 170. The acquisition module 110 is used to excite and acquire single-shot seismic data, which includes receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines. The mapping module 120 is used to acquire single-trace energy using a root mean square (RMS) algorithm, map the raw energy values ​​of single-trace data to a preset weighted interval using a nonlinear mapping function, divide the single-shot seismic data into blocks, and perform element-wise weighting based on the center trace energy weight to obtain a block set. Analysis module 130 performs PCA analysis on the weighted block set, calculates the mean of the block set and decenters it, fits the data through incremental PCA, determines the number of principal components of the first arrival data when the cumulative explained variance ratio reaches a preset threshold, and maps the high-dimensional block vector to a low-dimensional space based on the number of principal components of the first arrival data to obtain the low-dimensional block vector. Reconstruction module 140 determines the initial dictionary atoms based on the low-dimensional block vector, performs sparse coding using a regularized OMP algorithm, introduces a dynamic stopping threshold and skips the coding of weakly correlated signals, iterates and updates the dictionary using regularized K-SVD, and reinitializes atoms that failed singular value decomposition to obtain reconstructed sub-blocks. Generation module 150 applies a two-dimensional Hanning window to each reconstructed sub-block to obtain windowed sub-blocks, superimposes the windowed blocks at their original positions and accumulates the window weights, and generates preliminary first arrival denoised data after weight normalization. The calculation module 160 calculates the residual between the preliminary first-arrival denoised data and the original energy value, and extracts the residual energy quantile and the first-arrival signal energy quantile. When the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, bandpass filtering is applied to the residual. A fusion factor is determined based on the energy ratio of the residual to the first-arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first-arrival denoised data. The iteration module 170 calculates basic parameters based on the data dimension. It dynamically adjusts the sparsity and regularization intensity through the first-arrival data protection factor and regulation factor to obtain parameter combinations. The optimal parameter combination is obtained by iteratively testing the parameter combinations based on the preset first-arrival signal-to-noise ratio. By suppressing strong energy interference through spatial domain energy equalization, preserving core features through PCA dimensionality reduction, accurately representing signals through regularized dictionary learning, suppressing boundary artifacts through Hanning window weighting, recovering effective signals through residual fusion, and optimizing adaptive parameters, a precise balance between efficient noise suppression and first-arrival signal fidelity is achieved, thereby improving the first-arrival signal-to-noise ratio. This provides high-quality basic data support for subsequent exploration data processing and meets the needs of exploration production for high-precision first-arrival analysis.

[0047] like Figure 6As shown, an embodiment of the third aspect of this application provides a first-arrival signal-to-noise ratio (SNR) enhancement system 20, including a memory 300 and a processor 400. The memory 300 stores a program or instructions that can be executed on the processor 400. When the processor 400 executes the program or instructions, it implements the steps of the first-arrival SNR enhancement method of any one of the embodiments of the first aspect, and thus has the technical effects of any embodiment of the first aspect, which will not be repeated here.

[0048] An embodiment of the fourth aspect of this application provides a readable storage medium storing a program or instructions that, when executed by a processor, implement the steps of the first arrival signal-to-noise ratio enhancement method of any one of the embodiments of the first aspect, and thus has the technical effects of any embodiment of the first aspect described above, which will not be repeated here.

[0049] like Figure 7 , Figure 8 and Figure 9 As shown, the first-arrival signal-to-noise ratio (SNR) enhancement method according to a specific embodiment of this application achieves a precise balance between efficient noise suppression and first-arrival signal fidelity by suppressing strong energy interference through spatial domain energy equalization, preserving core features through PCA dimensionality reduction, accurately representing the signal through regularized dictionary learning, suppressing boundary artifacts through Hanning window weighting, recovering effective signals through residual fusion, and optimizing adaptive parameters. This improves the first-arrival SNR, provides high-quality basic data support for subsequent exploration data processing, and meets the needs of exploration production for high-precision first-arrival analysis.

[0050] This embodiment is achieved through the following technical solution: 1) Excite and collect single-shot seismic data. The receiver line spacing is L meters, the channel spacing is R meters, each receiver line has M channels, each channel has N sampling points, and the number of receiver lines is S.

[0051] 2) Calculate the energy characteristic values ​​of each channel. Use the root mean square algorithm to obtain the energy of a single channel and calculate the global energy parameters, including the maximum and minimum energy values. The maximum energy is the highest value among all channel energies, and the minimum energy is the lowest value among all channel energies. Percentile value. The original energy value is mapped to a value using a nonlinear mapping function. Weighting intervals. Seismic data is divided into blocks according to a preset size, and element-wise weighting is performed based on the energy weight of the central trace of each block to achieve spatial domain energy equalization.

[0052] In step 2), the original Dao energy value is converted into weight coefficients through a nonlinear mapping function. The mapping relationship is: weight = 0.6 + 0.4 × (Dao energy - minimum energy) / (maximum energy - minimum energy).

[0053] The minimum energy in step 2) is the th energy among all the energy levels. Percentile value, to avoid noise interference with the low signal-to-noise ratio first arrival, is set by default. It might be set in the 5%-15% range (e.g., 10%), discarding the lowest extreme noise channels to ensure that the minimum energy represents a "true weak signal" rather than noise. The lower the signal-to-noise ratio of the first-arrival data, the better. The higher the value, the better the noise immunity.

[0054] Weight range in step 2) It is obtained through threshold pruning, forming a differentiated weighting system that enhances weak energy channels and moderately suppresses strong energy channels. The default value is 0.3. The default value is 1.0. Greater than or equal to 0, Less than or equal to 1.0.

[0055] 3) Perform PCA analysis on the weighted block set, calculate the mean of the block set, and then decenter it. Fit the data using incremental PCA. When the cumulative explained variance ratio reaches a preset threshold, determine the number of principal components of the first arrival data. Based on the selected number of principal components of the first arrival data, map the high-dimensional block vectors to a low-dimensional space, preserving the characteristics of the first arrival signal while reducing the data dimensionality.

[0056] Threshold in step 3) The default value is 0.8, which effectively preserves the main features of the first arrival signal and reduces the interference of redundant information on subsequent dictionary learning.

[0057] 4) Randomly select samples from the low-dimensional block vector as initial dictionary atoms. Use the regularized OMP algorithm for sparse coding, introducing a dynamic stopping threshold (signal norm). For weakly correlated signals (initial correlation less than 10%) The encoding is skipped, and a regularization term is added to the diagonal of the Gram matrix during the solution process. The dictionary is updated through regularized K-SVD iteration, and atoms that fail to be decomposed by singular value decomposition are reinitialized to enhance the dictionary's ability to represent seismic signals.

[0058] The dynamic stopping threshold in step 4) refers to the threshold value when the L2 norm of the difference between the original signal and the current approximate signal is less than the norm of the original signal. Stop iteration here. The default value is 10.

[0059] The initial correlation in step 4) refers to the degree of matching between the original signal and a certain atom in the dictionary. If the correlation between a certain atom and the signal is less than... If an atom is almost unrelated to another atom, it is skipped and not included in the set of candidate atoms for subsequent encoding. The default value is 0.05.

[0060] In step 4), add regularization terms to the diagonal of the Gram matrix. Improve solution stability The default value is 1e-6.

[0061] 5) Generate a two-dimensional Hanning window (constructed by the outer product of cosine functions) that matches the size of the data sub-blocks. Apply the Hanning window to each reconstructed sub-block to suppress abrupt changes in the sub-block edge signal. Stack the windowed sub-blocks in their original positions, accumulate the window weights synchronously, and generate preliminary first-arrival denoised data by weight normalization (dividing the stacking result by the accumulated weights).

[0062] Step 5) generates preliminary first-arrival denoised data by weight normalization (overlay result divided by cumulative weight), which mainly weakens the boundary artifacts generated by sub-block splicing and improves the spatial continuity of data reconstruction.

[0063] 6) Calculate the residual between the preliminary first-arrival denoising result and the original data, and extract the k% quantile of the residual energy and the k% quantile of the signal energy. When the k% quantile of the residual energy exceeds the k% quantile of the signal energy... At that time, bandpass filtering is applied to the residual. The fusion factor is determined based on the energy ratio of the residual to the signal. The filtered residuals are proportionally superimposed onto the preliminary results to recover the effective first arrival wave signals contained in the residuals.

[0064] In step 6), the k% quantile value of the residual energy and the k% quantile value of the signal energy are generally taken to be the same value, with k defaulting to 95. The lower the first-arrival signal-to-noise ratio, the larger the value should be, which enhances noise sensitivity; conversely, the higher the first-arrival signal-to-noise ratio, the smaller the value should be, which improves sensitivity to weak first-arrival signals.

[0065] In step 6), when the residual energy (k% quantile) exceeds the signal energy (k% quantile)... hour, The default value is 10. The higher the first-arrival signal-to-noise ratio (SNR), the smaller the threshold adjustment formula is: 10% × (1-SNR / 50); the lower the first-arrival SNR, the larger the threshold adjustment formula is: 10% × (1+SNR / 20). Here, SNR is the signal-to-noise ratio of the first-arrival data.

[0066] In step 6), the residual is bandpass filtered. The range of values ​​is determined based on the frequency characteristics of the first arrival wave data, with a default value of 5Hz-100Hz.

[0067] The fusion factor in step 6) is negatively correlated with the signal-to-noise ratio of the first arrival wave data. For every 10dB decrease in SNR, the fusion weight increases by 50%, and an upper limit of 0.3 is set to prevent noise amplification.

[0068] 7) Calculate basic parameters based on the data dimension. These parameters include patch size, step size, dictionary capacity, etc., determined according to the number of channels and sampling points. Set a signal protection factor, dynamically adjust the sparsity by regulating the factor, and adjust the regularization intensity proportionally. Iteratively test multiple parameter combinations, using the signal-to-noise ratio of the first arrival data as the evaluation index to select the optimal configuration, achieving adaptive denoising and balancing noise suppression and signal preservation.

[0069] The signal protection factor set in step 7) is related to the signal-to-noise ratio of the first arrival data. The lower the signal-to-noise ratio of the first arrival data, the larger the value should be. In a strong noise environment, the weak first arrival characteristics should be preserved. The value should be 0.5-0.9.

[0070] The adjustment factor in step 7) is to scale the basic sparsity by a preset factor. It is related to the signal-to-noise ratio of the first arrival wave data. The lower the signal-to-noise ratio, the larger the value. The value is 0.6-0.9.

[0071] This embodiment employs spatial domain energy equalization, calculating the energy of a single channel using the root mean square (RMS) and obtaining weights through nonlinear mapping, then weighting the data in blocks to achieve equalization. PCA dimensionality reduction preserves the main features of the first arrival wave during weighted data processing. Regularized dictionary learning uses regularized OMP encoding and K-SVD iterative updates to the dictionary. Hanning window processing and normalization suppress abrupt changes at sub-block edges to generate preliminary denoised data. Residual fusion conditionally superimposes filtered residuals to recover effective signals. Adaptive parameter optimization dynamically adjusts parameters iteratively to select the optimal configuration, significantly improving the signal-to-noise ratio of the first arrival wave data and providing foundational data for subsequent data processing.

[0072] The core of this embodiment is a multi-stage cascaded processing flow: First, the energy of a single channel is calculated using the root mean square algorithm, and weights are obtained through nonlinear mapping. Spatial domain energy equalization is achieved through block weighting. Second, PCA analysis is performed on the weighted data to reduce dimensionality and retain the main features of the first-arrival data. Then, regularized OMP algorithm sparse coding is used, combined with regularized K-SVD iterative dictionary updates to enhance signal representation capabilities. Next, a two-dimensional Hanning window is used to process and reconstruct sub-blocks, suppressing edge abrupt changes and normalizing to generate preliminary denoised data. Subsequently, residuals are calculated, and energy quantile values ​​are used for judgment, filtering, and fusion to recover effective first-arrival signals. Finally, parameters are calculated based on data dimensions, and protection and control factors are dynamically optimized. The optimal solution is selected using the first-arrival signal-to-noise ratio (SNR) as an indicator. This method balances noise suppression and first-arrival signal preservation, improves the first-arrival SNR, and demonstrates good reliability and practicality, meeting exploration requirements.

[0073] The main technologies in this embodiment include: First, spatial domain energy equalization, which calculates the energy of a single channel using the root mean square algorithm, obtains weights through nonlinear mapping, and then weights the data according to the energy weight of the central channel after segmentation; Second, PCA dimensionality reduction, which decenters the weighted data and fits it with incremental PCA, determines the number of principal components based on the cumulative explained variance ratio, achieves dimensionality reduction of high-dimensional data and retains the main features of the first arrival wave; Third, regularized dictionary learning, which randomly selects samples as initial dictionary atoms, uses regularized OMP algorithm for sparse coding (including dynamic stopping threshold, skipping weak correlation signal coding and adding regularization terms to the Gram matrix), and iteratively updates the dictionary using regularized K-SVD; Fourth, Hanning window processing and normalization, which uses a two-dimensional Hanning window to suppress abrupt changes at the edges of reconstructed sub-blocks, and generates preliminary denoised first arrival data after weight normalization after superposition; Fifth, residual fusion, which calculates residuals and extracts energy quantile values, and when the threshold is exceeded, filters the residuals and superimposes them according to the fusion factor to recover the effective first arrival wave signal; Sixth, adaptive parameter optimization, which calculates basic parameters based on the data dimension, sets protection and control factors to dynamically adjust parameters, and iteratively selects the optimal configuration.

[0074] 1) Excite and record single-shot seismic data. The receiver line spacing is 200 meters, the excitation line spacing (shot row spacing) is 200 meters, the trace spacing is 40 meters, and the shot point spacing is 40 meters. Each receiver line has 200 traces, each trace has 1100 sampling points, and the sampling interval is 2 ms. The designed observation system is as follows: Figure 7 As shown.

[0075] Based on the shot collection records collected by the aforementioned observation system, as follows: Figure 8 As shown, due to the urbanization and industrialization of the exploration area, there are many sources of interference, the signal-to-noise ratio of a single shot is low, the first arrival wave is submerged by noise, the signal-to-noise ratio of the first arrival wave data is low, and the first arrival time is difficult to determine.

[0076] 2) Calculate the energy characteristic values ​​for each channel. Use the root mean square algorithm to obtain the energy of each channel and calculate the global energy parameters, including the maximum and minimum energy values. The maximum energy is the highest value among all channel energies, and the minimum energy is the 10th percentile value among all channel energies. Map the original energy values ​​to a nonlinear mapping function. Weighting intervals. Seismic data is divided into blocks according to a preset size, and element-wise weighting is performed based on the energy weight of the central trace of each block to achieve spatial domain energy equalization.

[0077] 3) Perform PCA analysis on the weighted block set, calculate the mean of the block set, and then center it. Fit the data using incremental PCA, and when the cumulative explained variance ratio reaches a preset threshold... The number of principal components in the first-arrival data is determined. Based on the selected number of principal components, the high-dimensional block vectors are mapped to a low-dimensional space, preserving the characteristics of the first-arrival signal while reducing the data dimensionality.

[0078] 4) Randomly select samples from the low-dimensional block vector as initial dictionary atoms. Use a regularized OMP algorithm for sparse coding, introducing a dynamic stopping threshold (10% of the signal's L2 norm) for weakly correlated signals (initial correlation less than...). The encoding is skipped, and a regularization term is added to the diagonal of the Gram matrix during the solution process. The dictionary is updated iteratively through regularized K-SVD, and atoms that fail to be decomposed by singular value decomposition are reinitialized to enhance the dictionary's ability to represent seismic signals.

[0079] 5) Generate a two-dimensional Hanning window (constructed by the outer product of cosine functions) that matches the size of the data sub-blocks. Apply the Hanning window to each reconstructed sub-block to suppress abrupt changes in the sub-block edge signal. Stack the windowed sub-blocks in their original positions, accumulate the window weights synchronously, and generate preliminary first-arrival denoised data by weight normalization (dividing the stacking result by the accumulated weights).

[0080] 6) Calculate the residual between the preliminary first-arrival denoising result and the original data. Extract the 95th percentile of the residual energy and the 95th percentile of the signal energy. When the 95th percentile of the residual energy exceeds 10% of the 95th percentile of the signal energy, perform bandpass filtering on the residual. Determine the fusion factor (0~0.3) based on the energy ratio of the residual to the signal. Proportionally superimpose the filtered residual onto the preliminary result to recover the effective first-arrival signal contained in the residual.

[0081] 7) Calculate basic parameters based on data dimensions. These parameters include patch size, step size, dictionary capacity, etc., determined according to the number of channels and sampling points. Set a signal protection factor (0.6) and dynamically adjust the sparsity by regulating the factor to adjust the regularization intensity proportionally. Iteratively test multiple parameter combinations, using the signal-to-noise ratio of the first arrival data as the evaluation index to select the optimal configuration, achieving adaptive denoising and balancing noise suppression and signal preservation.

[0082] According to steps 2)-7), Figure 8 The single-shot data shown is denoised, such as Figure 9 As shown, it can be seen that, with Figure 8 In comparison, after processing, a large amount of random noise is suppressed, the first arrival characteristics are enhanced, the first arrivals of adjacent channels are smoother and more continuous, and the number of first arrival channels that can be picked up increases.

[0083] In summary, the beneficial effects of the embodiments of this application are as follows: single-channel energy is calculated by root mean square (RMS) and weights are obtained through nonlinear mapping, and block weighting is used to achieve equilibrium; PCA dimensionality reduction preserves the main features of the first arrival wave in the weighted data processing; regularized dictionary learning uses regularized OMP encoding and K-SVD iterative updating of the dictionary; Hanning window processing and normalization suppress abrupt changes at the edges of sub-blocks to generate preliminary denoised data; residual fusion uses conditional superposition of filtered residuals to recover effective signals; adaptive parameter optimization dynamically adjusts parameters iteratively to select the optimal configuration, which significantly improves the signal-to-noise ratio of the first arrival wave data and provides basic data for subsequent data processing.

[0084] In this application, the terms "first," "second," and "third" are used for descriptive purposes only and should not be construed as indicating or implying relative importance; the term "multiple" refers to two or more unless otherwise expressly defined. The terms "install," "connect," "link," and "fix" should be interpreted broadly. For example, "connect" can mean a fixed connection, a detachable connection, or an integral connection; "link" can mean a direct connection or an indirect connection through an intermediate medium. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0085] In the description of this application, it should be understood that the terms "upper", "lower", "front", "rear", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this application and simplifying the description, and do not indicate or imply that the device or module referred to must have a specific orientation or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this application.

[0086] In the description of this specification, the terms "one embodiment," "some embodiments," "specific embodiment," etc., refer to a specific feature, structure, material, or characteristic described in connection with that embodiment or example, which is included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0087] The above are merely preferred embodiments of this application and are not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A method for improving the signal-to-noise ratio of the first arrival wave, characterized in that, include: Excite and acquire single-shot seismic data, which includes receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines; The energy of a single trace is obtained by means of the root mean square algorithm. The original energy value of the single trace is mapped to a preset weight range by a nonlinear mapping function. The single-shot seismic data is divided into blocks and weighted element by element according to the energy weight of the central trace to obtain a block set. PCA analysis is performed on the weighted block set to calculate the mean of the block set and decenter it. The data is fitted by incremental PCA. When the cumulative explained variance ratio reaches a preset threshold, the number of principal components of the first arrival data is determined. Based on the number of principal components of the first arrival data, the high-dimensional block vector is mapped to the low-dimensional space to obtain the low-dimensional block vector. The initial dictionary atoms are determined based on the low-dimensional block vector. Sparse coding is performed using the regularized OMP algorithm. A dynamic stopping threshold is introduced and weakly correlated signal coding is skipped. The dictionary is updated iteratively using regularized K-SVD. Atoms that fail to decompose singular values ​​are reinitialized to obtain reconstructed sub-blocks. A two-dimensional Hanning window is applied to each of the reconstructed sub-blocks to obtain windowed sub-blocks. The windowed sub-blocks are then superimposed according to their original positions and the window weights are accumulated. After weight normalization, preliminary first arrival denoising data is generated. Calculate the residual between the preliminary first-arrival denoised data and the original energy value, and extract the residual energy quantile and the first-arrival signal energy quantile. When the residual energy quantile exceeds the first-arrival signal energy quantile by a third preset percentage, bandpass filter is applied to the residual. A fusion factor is determined based on the energy ratio of the residual to the first-arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first-arrival denoised data. Based on the data dimension, the basic parameters are calculated, and the sparsity and regularization intensity are dynamically adjusted by the first arrival wave data protection factor and the regulation factor to obtain the parameter combination. The parameter combination is iteratively tested according to the preset first arrival wave signal-to-noise ratio to obtain the optimal parameter combination.

2. The method for improving the signal-to-noise ratio of the first arrival wave according to claim 1, characterized in that, The step of obtaining single-channel energy through the root mean square algorithm and mapping the original energy value of the single-channel energy to a preset weight range through a nonlinear mapping function includes: The energy of a single channel is obtained by the root mean square algorithm, and the global energy parameters are calculated. The global energy parameters include the maximum energy value and the minimum energy value. Based on the single-channel energy, the maximum energy value, and the minimum energy value, the original channel energy value is converted into weighting coefficients through a nonlinear mapping function; The maximum energy value is the maximum value among all channel energies, and the minimum energy value is the energy value of the first preset percentile among all channel energies, where the first preset percentile is 5%-15%.

3. The method for improving the signal-to-noise ratio of the first arrival wave according to claim 2, characterized in that, The step of obtaining single-channel energy through the root mean square algorithm and mapping the original energy value of the single-channel energy to a preset weight range through a nonlinear mapping function further includes: The preset weight range obtained by threshold clipping has a weight range greater than or equal to 0 and less than or equal to 1.

4. The method for improving the signal-to-noise ratio of the first arrival wave according to any one of claims 1 to 3, characterized in that, The preset threshold ranges from 0.7 to 0.

9.

5. The method for improving the signal-to-noise ratio of the first arrival wave according to any one of claims 1 to 3, characterized in that, The step of determining the initial dictionary atoms based on the low-dimensional block vector, performing sparse coding using the regularized OMP algorithm, introducing a dynamic stopping threshold, and skipping the coding of weakly correlated signals includes: Samples are randomly selected from the low-dimensional block vector as initial dictionary atoms, and sparse coding is performed using the regularized OMP algorithm. The iteration stops when the L2 norm of the difference between the original signal and the current approximate signal is less than the second preset percentage of the norm of the original signal. When the correlation between the original signal and the initial dictionary atom is less than a preset correlation value, the initial dictionary atom is skipped and not included in the candidate atom set for subsequent encoding. During the solution process, regularization terms are added to the diagonal of the Gram matrix.

6. The method for improving the signal-to-noise ratio of the first arrival wave according to any one of claims 1 to 3, characterized in that, The residual energy quantile value and the first arrival signal energy quantile value have the same quantile value; The bandpass filter has a value range of 5Hz-100Hz.

7. The method for improving the signal-to-noise ratio of the first arrival wave according to any one of claims 1 to 3, characterized in that, The value range of the first arrival wave data protection factor is 0.5-0.9; The value range of the regulatory factor is 0.6-0.

9.

8. A first-arrival signal-to-noise ratio enhancement system, characterized in that, include: The acquisition module (110) is used to excite and acquire single-shot seismic data, which includes receiver line spacing, trace spacing, number of traces per receiver line, number of sampling points per trace, and number of receiver lines; The mapping module (120) is used to obtain the energy of a single trace through the root mean square algorithm, map the original energy value of the single trace to a preset weight range through a nonlinear mapping function, divide the single-shot seismic data into blocks and perform element-wise weighting based on the energy weight of the central trace to obtain a block set. The analysis module (130) is used to perform PCA analysis on the weighted block set, calculate the mean of the block set and decenter it, fit the data through incremental PCA, determine the number of principal components of the first arrival data when the cumulative explained variance ratio reaches a preset threshold, and map the high-dimensional block vector to the low-dimensional space according to the number of principal components of the first arrival data to obtain the low-dimensional block vector. The reconstruction module (140) is used to determine the initial dictionary atoms according to the low-dimensional block vector, perform sparse coding through the regularized OMP algorithm, introduce a dynamic stopping threshold and skip the coding of weakly correlated signals, update the dictionary through regularized K-SVD iteration, reinitialize the atoms that failed the singular value decomposition, and obtain the reconstructed sub-blocks. The generation module (150) is used to apply a two-dimensional Hanning window to each of the reconstructed sub-blocks to obtain windowed sub-blocks, and to superimpose the windowed sub-blocks according to their original positions and accumulate window weights, and generate preliminary first arrival denoising data after weight normalization. The calculation module (160) is used to calculate the residual between the preliminary first arrival denoised data and the original energy value, and extract the residual energy quantile and the first arrival signal energy quantile. When the residual energy quantile exceeds the third preset percentage of the first arrival signal energy quantile, the residual is bandpass filtered. The fusion factor is determined according to the energy ratio of the residual to the first arrival signal. The filtered residual is then proportionally superimposed onto the preliminary first arrival denoised data. The iterative module (170) is used to calculate basic parameters based on data dimensions, dynamically adjust sparsity and regularization intensity through first-arrival data protection factor and regulation factor to obtain parameter combinations, and iteratively test the parameter combinations according to the preset first-arrival signal-to-noise ratio to obtain the optimal parameter combination.

9. A first-arrival wave signal-to-noise ratio enhancement system, characterized in that, include: A memory (300) and a processor (400), wherein the memory (300) stores a program or instructions executable on the processor (400), and the processor (400) implements the steps of the first arrival signal-to-noise ratio enhancement method as described in any one of claims 1 to 7 when executing the program or instructions.

10. A readable storage medium having a program or instructions stored thereon, characterized in that, When the program or the instructions are executed by the processor, they implement the steps of the first arrival signal-to-noise ratio enhancement method as described in any one of claims 1 to 7.

Images