Rank-based NMO-free three-dimensional near-surface correction method and device

Abstract

The invention provides a rank-based NMO-free three-dimensional near-surface correction method and device. According to the rank-based NMO-free three-dimensional near-surface correction scheme provided by the invention, residual static correction calculation can be carried out, more accurate static correction data can be obtained, dynamic correction is not needed, and relatively low processing efficiency caused by the fact that existing static correction still needs to iteratively calculate an NMO speed value and a residual static correction value for multiple times is avoided.

Description

Technical Field

[0001] The present invention relates to the field of seismic exploration data processing methods, and more specifically, to a rank-based non-NMO three-dimensional near-surface correction method and apparatus. Background Art

[0002] The complex near-surface weathered layer poses challenges to the imaging and inversion of land seismic data. Acquisition and modeling limitations may prevent the creation of an accurate near-surface model that cannot capture its heterogeneity and strong lateral variation characteristics. To overcome these limitations, static correction is used as an effective alternative solution. However, due to the surface-consistent assumption, static correction estimation relies on the simplification of the near-surface model. Therefore, surface-consistent short-wave static correction estimation and correction are usually performed after a non-surface-consistent step to obtain most of the static correction amount.

[0003] The near-surface weathered layer is characterized by complexity and inhomogeneity, which poses great challenges to the imaging and inversion of land seismic data. The strata near the surface contain loose minerals and rocks, resulting in severe and complex acoustic wave attenuation. If this complex wavefield perturbation is not processed, it will seriously affect the analysis results of seismic data. Due to limitations in aspects such as spatial resolution and modeling capabilities, current seismic exploration technologies are difficult to accurately depict the complex and variable detailed geological structure of the near-surface layer and cannot effectively parameterize model its high heterogeneity and strong lateral variation characteristics, which further exacerbates the negative impact of the near-surface on seismic data. To try to overcome this difficulty, a static correction method is introduced in seismic processing to compensate for surface noise. The basic idea of static correction is to eliminate the interference of the near-surface layer on the seismic record as much as possible without changing the deep reflection coefficient, so that the reflected wave packet is as close as possible to the source wave packet. Currently, a large number of studies have shown that static correction can indeed greatly reduce the damage of near-surface noise to seismic data, improve the signal-to-noise ratio, and lay a foundation for subsequent geological interpretation and analysis.

[0004] However, traditional static correction techniques themselves also have certain problems and limitations, mainly reflected in that the calculation of static correction relies on the assumption of surface consistency, which often does not conform to the actual geological situation. The surface consistency assumption simplifies the near-surface model and adopts a single approximation of delay time. However, in fact, due to the complexity of the near-surface layer heterogeneity, the input wave fields at different positions will indeed experience different attenuation, scattering, and deflection, resulting in different static correction residuals. Therefore, strict surface consistency does not hold. The static correction amounts obtained under the traditional surface consistency method cannot completely eliminate the influence of the near surface, and the processing effect has limitations. In addition, in the specific processing flow, the short-wave static correction processing of surface consistency is also based on a non-surface-consistent processing. Usually, it needs to go through two steps: the first step is to use a non-surface-consistent method to obtain the main static correction amount; the second step is to use the surface consistency method for more detailed residual static correction processing to further improve the accuracy of static correction. This two-step separate strategy has low processing efficiency, and there may be multiple iterations in the middle, resulting in a large overall workload. The specific analysis is as follows.

[0005] First, the existing technologies generally need to perform multiple iterative calculations for NMO velocity values and residual static correction amounts because there is an interactive relationship between the two, restricting each other. Without determining one of the quantities accurately, it is impossible to calculate the other quantity precisely. However, the time cost required for such repeated iterative calculations is extremely large, and the efficiency is severely low. In the traditional processing flow, accurate dynamic correction must be obtained before static correction. However, NMO correction relies on the time difference mean of the trace gather, which requires correct estimation of the initial velocity of each trace in advance, so as to implement appropriate NMO correction to achieve the purpose of normal time correction. At the same time, accurate estimation of the residual static correction amount can only be carried out after performing sufficiently accurate NMO correction. Because estimating the residual static correction amount requires analyzing and distinguishing the causes of the residual noise, and incorrect time difference correction will cause the mixture of static correction residuals and NMO correction residuals. Therefore, these two steps restrict each other and are interdependent. Sufficient number of circular iterations must be performed to converge to sufficiently accurate respective correction results. And each iteration requires repeated solution of NMO velocity and static correction amount, with extremely high computational complexity and time cost.

[0006] Second, the workload of this iterative calculation is huge, and it is very difficult to automatically terminate the judgment. Various algorithms can be used to calculate the NMO velocity, such as the Gaussian attenuation algorithm, semblance peak method, etc. For the static correction amount, the optimal solution can also be searched through various methods. These operations themselves have a heavy computational load and also require a large amount of manual experience to participate in parameter adjustment and effect interpretation, and it is impossible to simply preset the termination condition. The processing workload is large and the efficiency is low.

[0007] Thirdly, repeated iteration forms a closed loop, and the cumulative error affects the result. Each calibration calculation itself has an approximation error of the algorithm itself. After cyclic accumulation, the error will also accumulate and amplify, making it difficult for the true solution to converge and forming a vicious cycle of "error - iteration". This is also an important reason for low efficiency.

[0008] In summary, the existing three - dimensional seismic data calibration schemes are difficult to be satisfactory. Summary of the Invention

[0009] In view of this, the present invention aims to propose a three - dimensional near - surface calibration scheme without the NMO calibration link to obtain more accurate static correction data and with a smaller amount of calculation.

[0010] According to one aspect of the present invention, a rank - based NMO - free three - dimensional near - surface calibration method is proposed. The method includes:

[0011] Step 1: Arrange the three - dimensional seismic data into a two - dimensional matrix to form a two - dimensional matrix D.

[0012] Step 2: Perform a Fourier transform on the matrix D to obtain a matrix E.

[0013] Step 3: Loop through all frequencies and ranks of the matrix E, calculate the singular value decomposition of each frequency slice, and then traverse each rank based on the singular value decomposition result to obtain the low - rank approximation F of each rank of the frequency slice.

[0014] Step 4: Aggregate the low - rank approximations F of all ranks of all frequency slices, and perform an inverse Fourier transform on the aggregated matrix to obtain a matrix G.

[0015] Step 5: Calculate the cross - correlation between the matrix G and the matrix D, and select the maximum delay of the cross - correlation as the static correction amount.

[0016] Step 6: Use the static correction amount to perform static correction on the three - dimensional seismic data.

[0017] In some embodiments, step 1 specifically includes:

[0018] Place the x - coordinates in the source point coordinates and receiver point coordinates of the three - dimensional seismic data on one dimension of the matrix, and place the y - coordinates in the source point coordinates and receiver point coordinates on the other dimension of the matrix.

[0019] In some embodiments, calculating the singular value decomposition of each frequency slice specifically includes:

[0020] For each frequency, take out the frequency slice of this frequency from the matrix E, denoted as sub - matrix A.

[0021] Establish the singular value decomposition equation of the sub - matrix A.

[0022] Based on the singular value decomposition equation, the singular values of A and the corresponding eigenvectors are obtained by solving the eigenvalue problem of the submatrix A.

[0023] In some embodiments, at each frequency slice, when traversing each rank, it is in ascending order.

[0024] In some embodiments, based on the singular value decomposition result, each rank is traversed to obtain the low-rank approximation F of each rank of this frequency slice, including:

[0025] For rank n, based on the first n singular values and the corresponding eigenvectors obtained from the singular value decomposition, the low-rank approximation F of rank n is obtained.

[0026] According to another aspect of the present invention, a rank-based non-NMO three-dimensional near-surface correction device is proposed. The device includes:

[0027] A matrix unit for arranging three-dimensional seismic data into a two-dimensional matrix to form a two-dimensional matrix D;

[0028] A Fourier transform unit for performing a Fourier transform on the matrix D to obtain a matrix E;

[0029] A loop calculation unit for looping through all frequencies and ranks of the matrix E, calculating the singular value decomposition of each frequency slice, and then traversing each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice;

[0030] An inverse Fourier transform unit for summarizing the low-rank approximations F of all ranks of all frequency slices and performing an inverse Fourier transform on the summarized matrix to obtain a matrix G;

[0031] A static correction amount calculation unit for performing cross-correlation between the matrix G and the matrix D and selecting the maximum delay of the cross-correlation as the static correction amount;

[0032] A static correction unit for performing static correction on the three-dimensional seismic data using the static correction amount.

[0033] In some embodiments, the matrix unit is specifically used for:

[0034] Placing the x coordinate in the shot point coordinates and receiver point coordinates of the three-dimensional seismic data on one dimension of the matrix, and placing the y coordinate in the shot point coordinates and receiver point coordinates on the other dimension of the matrix.

[0035] In some embodiments, calculating the singular value decomposition of each frequency slice specifically includes:

[0036] For each frequency, the frequency slice of this frequency is taken out from the matrix E and denoted as the submatrix A.

[0037] Establish the singular value decomposition equation of the sub - matrix A;

[0038] Based on the singular value decomposition equation, obtain the singular values of A and the corresponding eigenvectors by solving the eigenvalue problem of the sub - matrix A.

[0039] In some embodiments, based on the singular value decomposition result, traverse each rank to obtain the low - rank approximation F of each rank of this frequency slice, including:

[0040] For rank n, based on the first n singular values and the corresponding eigenvectors obtained from the singular value decomposition, obtain the low - rank approximation F of rank n.

[0041] In some embodiments, based on the singular value decomposition result, traverse each rank to obtain the low - rank approximation F of each rank of this frequency slice, including:

[0042] For rank n, based on the first n singular values and the corresponding eigenvectors obtained from the singular value decomposition, obtain the low - rank approximation F of rank n.

[0043] According to another aspect of the present invention, an electronic device is also proposed. The electronic device includes:

[0044] A memory storing executable instructions;

[0045] A processor that runs the executable instructions in the memory to implement the rank - based NMO - free 3D near - surface correction method described above.

[0046] According to another aspect of the present invention, a computer - readable storage medium is also proposed. The computer - readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the rank - based NMO - free 3D near - surface correction method described above.

[0047] The beneficial effects of the present invention at least include:

[0048] 1. Improve the calculation accuracy of static correction

[0049] The present invention abandons the surface - consistency assumption, directly estimates the static correction amount for the original record, avoids the static correction residuals caused by surface inhomogeneity, and the calculation result is closer to the true optimal solution, improving the effectiveness of static correction;

[0050] 2. No need for NMO correction

[0051] The NMO velocity error does not affect the residual static correction estimation, reducing the multiple iterations of NMO velocity and the estimation of residual static correction amount. Therefore, in order to avoid the NMO stretching effect, it is also not necessary to cut the data after NMO correction;

[0052] 3. Simplify the calculation process and improve efficiency

[0053] Since the steps of iteratively calculating the NMO velocity and static correction amount are skipped, the redundant computational amount is greatly reduced, and the entire calculation process is more concise and efficient.

[0054] 4. Reduced the influence of manual experience

[0055] In the traditional process, many parameters need to be repeatedly adjusted and interpreted manually. This method realizes end-to-end fully automatic calculation, avoiding accidental errors caused by manual experience;

[0056] 5. Improved the processing effect of low signal-to-noise ratio data

[0057] In low signal-to-noise ratio records, repeated iteration is often difficult to converge. This method achieves the goal in one step and has better adaptability to low signal-to-noise ratio data;

[0058] 6. Reduced the algorithm and implementation difficulty

[0059] The calculation process is simple and straightforward, the method is easy to understand and implement, and the algorithm difficulty is also greatly reduced.

[0060] 7. Expanded the application scope and is more flexible in processing

[0061] This method is not limited by the frequency band. In theory, it can process signals of any bandwidth, expanding the application scope, and can be flexibly applied to 2D or 3D data.

[0062] 8. Laid a better foundation for subsequent menu analysis

[0063] Through more accurate and effective static correction, the subsequent imaging quality can be effectively improved, laying a better foundation for fine description analysis such as feature extraction and reservoir identification.

[0064] In summary, the present invention proposes a rank-based NMO-free 3D near-surface correction scheme, which can perform residual static correction calculation, obtain more accurate static correction data, and does not require dynamic correction, avoiding the low processing efficiency caused by the need for multiple iterative calculations of the NMO velocity value and residual static correction amount in the existing static correction. The present invention has significant improvements, good application prospects, and important scientific research value and application value.

[0065] The method and apparatus of the present invention have other characteristics and advantages, which will be obvious in the accompanying drawings and subsequent specific embodiments incorporated herein, or will be described in detail in the accompanying drawings and subsequent specific embodiments incorporated herein. These drawings and specific embodiments are used together to explain the specific principles of the present invention. Brief Description of the Drawings

[0066] The above and other objects, features, and advantages of the present invention will become more apparent by describing the exemplary embodiments of the present invention in more detail with reference to the accompanying drawings, in which, in the exemplary embodiments of the present invention, the same reference numerals generally represent the same components.

[0067] Figure 1 The flowchart of a rank-based NMO-free 3D near-surface correction method according to an embodiment of the present invention is shown.

[0068] Figure 2 A 30 Hz frequency slice of a certain exploration area without static correction problems is shown.

[0069] Figure 3 A 30 Hz frequency slice of a certain exploration area with static correction problems is shown.

[0070] Figure 4 A 30 Hz frequency slice of a certain exploration area processed according to an embodiment of the present invention is shown. Detailed implementation manners

[0071] The preferred embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided to make the present invention more thorough and complete, and to fully convey the scope of the present invention to those skilled in the art.

[0072] The principle of the present invention is briefly introduced here. The frequency slices affected by the weathered layer often exhibit slowly decaying singular values, while the frequency slices without static correction problems have low rank. Therefore, the calculation of the residual static correction amount can use the low-rank approximation of the frequency slices affected by the weathered layer to enhance the low-rank structure and obtain frequency slices with fewer near-surface acquisition footprints, and these frequency slices may also contain incorrect amplitudes. In order to maintain the relationship between the amplitude and the offset (AVO) response, the frequency slices of the low-rank approximation are cross-correlated with the original frequency slices to estimate the residual static correction amount. This process is iteratively applied at different rank scales and different frequency bands, that is, low-rank approximations of different ranks are obtained, and the low-rank approximation of high frequencies is improved.

[0073] The basic steps of two-dimensional residual static correction amount estimation and correction are to transform the data from the shot-receiver domain to the midpoint-offset domain to meet the singular value decay requirement, so as to achieve accurate low-rank approximation. Therefore, in order to extend this method to 3D data, it is necessary to find a suitable organization to obtain a two-dimensional matrix, so as to calculate the singular value decomposition (SVD). In addition, another important requirement is that these matrices need to meet the singular value decay requirement.

[0074] The following further introduces various embodiments of the present invention.

[0075] Example 1

[0076] Figure 1 The flowchart of a rank-based NMO-free 3D near-surface correction method according to an embodiment of the present invention is shown. As shown, this example includes steps 1 to 6.

[0077] Step 1: Arrange the 3D seismic data into a 2D matrix to form a 2D matrix D.

[0078] In some embodiments, the x coordinates in the shot point coordinates and receiver point coordinates of the 3D seismic data are placed on one dimension of the matrix, and the y coordinates in the shot point coordinates and receiver point coordinates are placed on the other dimension of the matrix.

[0079] The basic steps of 2D residual static correction amount estimation and correction are to transform the data from the shot-receiver domain to the midpoint-offset domain to meet the singular value decay requirement, so as to achieve an accurate low-rank approximation. Therefore, in order to extend this method to 3D data, it is necessary to find a suitable organization to obtain a 2D matrix, so as to calculate the singular value decomposition (SVD). In addition, another important requirement is that these matrices need to meet the singular value decay requirement. According to this embodiment, the 3D data is arranged into a 2D matrix, that is, matrixization is performed. The shot point coordinates and receiver point coordinates are placed on one dimension of the matrix along one direction (xsrc, xrec), and the shot point coordinates and receiver point coordinates in the other direction (ysrc, yrec) are placed on the other dimension to organize the data. It has the following advantages:

[0080] (1) Ensure that the matrix meets the calculation requirements of subsequent singular value decomposition. The 3D data is organized into a 2D matrix with distinct rows and columns, and SVD decomposition can be directly performed;

[0081] (2) Use the spatial distribution characteristics of the source point and receiver point coordinates for matrix mapping, convert the physical coordinate association into matrix index coordinates, and ensure the corresponding relationship of geographical space information;

[0082] (3) Take the source coordinates and receiver coordinates in one direction as one dimension of the matrix, and the source coordinates and receiver coordinates in the other direction as the other dimension of the matrix, retaining the spatial distribution information of the 3D data to the greatest extent;

[0083] (4) The rows and columns of the matrix have clear physical meanings, which is convenient for interpreting the results after matrix operations and mapping them back to the actual 3D space;

[0084] (5) The matrix structure is simple, the operation process is intuitive, and it is convenient for matrix operations such as singular value decomposition;

[0085] (6) The matrix organization form is ingenious, effectively converting three-dimensional sequence data into a layout suitable for two-dimensional matrix operations, creating conditions for the subsequent introduction of low-rank methods;

[0086] (7) The matrix conversion process can be automated, with little repetitive workload and good practicality;

[0087] (8) The matrix result is convenient for storage and also suitable for network transmission, and can be easily used as the input for subsequent operations such as low-rank decomposition.

[0088] Step 2: Perform a Fourier transform on matrix D to obtain matrix E.

[0089] By performing a Fourier transform on matrix D, the seismic data in the time domain can be converted to the frequency domain.

[0090] In some examples, the process of obtaining matrix E by performing a Fourier transform on matrix D is as follows:

[0091] One-dimensional Fourier transforms can be performed on each column of matrix D, and the specific calculation method can use the FFT (Fast Fourier Transform algorithm);

[0092] Through the FFT algorithm, the frequency domain information corresponding to each column of time domain signals in matrix D can be obtained;

[0093] Then, the frequency domain results of all columns are rearranged and combined in column order to form a new complex matrix E.

[0094] Matrix E is the result matrix after the Fourier transform of matrix D to the frequency domain, which contains the frequency component information of the three-dimensional seismic data in matrix D.

[0095] The number of rows and columns of matrix E can be the same as those of matrix D, but the elements change from real numbers to complex numbers, reflecting the frequency characteristics of the signal.

[0096] Thus, the frequency domain conversion of the time domain seismic data can be achieved, facilitating subsequent operations such as low-rank analysis of frequencies.

[0097] Step 3: Loop through all frequencies and ranks of matrix E, calculate the singular value decomposition of each frequency slice, and then traverse each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of the frequency slice.

[0098] Matrix E includes the information of each frequency in the expected frequency band for static correction amount estimation. Each frequency corresponds to a sub-matrix block in matrix E, serving as the frequency slice of that frequency. The rank reflects the number of linearly independent column vectors or row vectors of the matrix.

[0099] The frequency slices of all frequencies in the expected frequency band of static correction amount estimation can be subjected to singular value decomposition, and then each rank can be traversed based on the singular value decomposition result, and finally the low-rank approximation matrix F corresponding to each rank at each frequency can be obtained.

[0100] In some embodiments, calculating the singular value decomposition of each frequency slice specifically includes:

[0101] For each frequency, the frequency slice of this frequency is taken out from the matrix E and denoted as the sub-matrix A.

[0102] Establish the singular value decomposition equation of the sub-matrix A;

[0103] Based on the singular value decomposition equation, the singular values of A and the corresponding eigenvectors are obtained by solving the eigenvalue problem of the sub-matrix A.

[0104] In some embodiments, when traversing each rank under each frequency slice, it is in ascending order.

[0105] In some embodiments, traversing each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice includes:

[0106] For rank n, based on the first n singular values and the corresponding eigenvectors obtained by singular value decomposition, the low-rank approximation F of rank n is obtained.

[0107] The singular values on the diagonal of the singular value matrix obtained in the singular value decomposition (SVD) are by default sorted from large to small, that is, the singular values with earlier indices on the diagonal are larger, and the singular values with later indices will become smaller and smaller. Correspondingly, the singular vectors in the singular vector matrices U and V are also stored corresponding to the singular values in this order.

[0108] Therefore, when obtaining the low-rank approximation of rank n, the first n singular values can be taken, that is, the largest first n singular values on the diagonal, and the corresponding first n singular vectors in the singular value vector matrix can be taken, that is, the first n columns in the vector matrix. In this way, the low-rank approximation of the matrix can be reconstructed through the largest singular values and singular vectors. Thus, by retaining the first n singular values and vectors and truncating the other parts, the low-rank approximation matrix at this frequency and this rank can be obtained.

[0109] Performing cyclic calculations for all ranks and frequencies in this way takes into account both the influence of frequency and the influence of different rank values, and can comprehensively obtain all the low-rank approximation results.

[0110] A simple example is given for illustration. Suppose the expected frequency band of static correction amount estimation is 10 - 50 Hz, and the low-rank value range is 1 - 5.

[0111] First, the frequency slice corresponding to 10 Hz can be taken out, and SVD decomposition is performed on it to obtain all the singular values and eigenvectors at this frequency. Then, the low-rank approximation matrices are calculated by successively taking ranks from 1 to 5. First, when taking rank 1, the first singular value and the corresponding eigenvector in the singular value decomposition result at 10 Hz are retained, and a low-rank approximation matrix is reconstructed through their product; when taking rank 2, the first two singular values and the corresponding eigenvectors in the singular value decomposition result at 10 Hz are retained, and a low-rank approximation matrix is reconstructed through their product, and so on. In this way, all 5 low-rank approximation matrices with ranks from 1 to 5 at 10 Hz are calculated.

[0112] Then, the frequency slice corresponding to 11 Hz is taken out, and SVD decomposition is performed on it to obtain all the singular values and eigenvectors at this frequency. Then, the low-rank approximation matrices are calculated by successively taking ranks from 1 to 5,... until 50 Hz is reached, and all 5 low-rank approximation matrices with ranks from 1 to 5 at 50 Hz are calculated.

[0113] In other words, each "frequency-rank" combination can correspond to a low-rank approximation matrix F.

[0114] Step 4: Aggregate the low-rank approximations F of all ranks of all frequency slices, and perform an inverse Fourier transform on the aggregated matrix to obtain matrix G.

[0115] The matrix obtained after aggregation comprehensively reflects the low-rank information of all frequencies and all ranks. The low-rank approximation matrices F are distributed at different frequencies and different ranks. Combining them together can comprehensively reflect the low-rank characteristics of the data matrix. Then, an inverse transform is performed on it to align its result with D on the time axis for subsequent cross-correlation.

[0116] Performing an inverse transform on the aggregated matrix can ensure that low-frequency and high-frequency signals can participate in subsequent processing simultaneously, which is of great significance for subsequent processing.

[0117] Step 5: Perform cross-correlation on matrix G and matrix D, and select the maximum delay of the cross-correlation as the static correction amount.

[0118] Matrix G is a time-domain matrix obtained by performing an inverse Fourier transform on the aggregation of the low-rank approximation matrix F, and matrix D is the original seismic data matrix.

[0119] According to the present invention, the cross-correlation function between matrix G and matrix D can be calculated to reflect their similarity. Through the parameterized time shift τ, the relationship between the cross-correlation function value and the time shift can be obtained. In other words, cross-correlation can reflect the similarity between matrix G and D. By finding the best coincidence point between the two through parameterized time shift, the time shift corresponding to the maximum coincidence degree is the maximum similarity time shift between the two, that is, the above-mentioned maximum delay.

[0120] On the cross-correlation function graph, the time shift τ corresponding to the maximum cross-correlation peak can be found. This τ is the maximum delay of G relative to D. Since the matrix G reflects the best-fitting formation model after low-rank processing, its maximum delay τ can be used as the static correction amount required for static correction.

[0121] Step 6, perform static correction on the 3D seismic data using the static correction amount.

[0122] In seismic data, due to the influence of the near-surface low-velocity layer, etc., time shift errors often occur in deep reflection signals. Performing static correction on the 3D seismic data using the static correction amount can eliminate the influence of near-surface layer time shift errors, restore the true phase of deep reflections, make the deep reflection signals return to the original correct occurrence time, and can also improve the phase information. After the phase information is corrected, the formation interface will be clearer, the system noise will be reduced, the signal-to-noise ratio will be increased, the overall imaging quality will be improved, and it will provide a better basis for subsequent quantitative interpretation and analysis, improving the accuracy of data analysis and model establishment.

[0123] Therefore, the static correction of the present invention plays an important role in maintaining the efficiency of the seismic data processing chain and improving the quality of the final analysis results.

[0124] This embodiment proposes a rank-based non-NMO 3D near-surface correction scheme, which can perform residual static correction calculation, can obtain more accurate static correction data, and does not require dynamic correction, avoiding the low processing efficiency caused by the existing static correction still requiring multiple iterative calculations of NMO velocity values and residual static correction amounts.

[0125] Example 2

[0126] The present invention also proposes a rank-based non-NMO 3D near-surface correction device, and the device includes:

[0127] A matrix unit for arranging 3D seismic data into a 2D matrix to form a 2D matrix D;

[0128] A Fourier transform unit for performing Fourier transform on the matrix D to obtain a matrix E;

[0129] A loop calculation unit for looping through all frequencies and ranks of the matrix E, calculating the singular value decomposition of each frequency slice, and then traversing each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of the frequency slice;

[0130] An inverse Fourier transform unit for summarizing the low-rank approximations F of all ranks of all frequency slices and performing inverse Fourier transform on the summarized matrix to obtain a matrix G;

[0131] The static correction quantity calculation unit is used to perform cross-correlation on matrix G and matrix D, and select the maximum delay of the cross-correlation as the static correction quantity;

[0132] The static correction unit is used to perform static correction on the three-dimensional seismic data by using the static correction quantity.

[0133] In some embodiments, the matrix unit is specifically configured to:

[0134] Place the x coordinates in the shot point coordinates and receiver point coordinates of the three-dimensional seismic data on one dimension of the matrix, and place the y coordinates in the shot point coordinates and receiver point coordinates on another dimension of the matrix.

[0135] In some embodiments, calculating the singular value decomposition of each frequency slice specifically includes:

[0136] For each frequency, take out the frequency slice of this frequency from matrix E, denoted as sub-matrix A.

[0137] Establish the singular value decomposition equation of sub-matrix A;

[0138] Based on the singular value decomposition equation, obtain the singular values and corresponding eigenvectors of A by solving the eigenvalue problem of sub-matrix A.

[0139] In some embodiments, traversing each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice includes:

[0140] For rank n, based on the first n singular values and corresponding eigenvectors obtained from the singular value decomposition, obtain the low-rank approximation F of rank n.

[0141] In some embodiments, traversing each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice includes:

[0142] For rank n, based on the first n singular values and corresponding eigenvectors obtained from the singular value decomposition, obtain the low-rank approximation F of rank n.

[0143] The beneficial effects of this embodiment at least include:

[0144] 1. The calculation accuracy of static correction is improved

[0145] The present invention abandons the surface consistency assumption, directly estimates the static correction quantity for the original record, avoids the static correction residuals caused by surface inhomogeneity, the calculation result is closer to the true optimal solution, and the effectiveness of static correction is improved;

[0146] 2. NMO correction is not required

[0147] The NMO velocity error does not affect the residual static correction estimation, reducing the multiple iterations of NMO velocity and the estimation of residual static correction. Therefore, in order to avoid the NMO stretching effect, it is not necessary to cut the data after NMO correction either.

[0148] 3. Simplify the calculation process and improve efficiency

[0149] Since the steps of calculating NMO velocity and static correction amount through multiple iterations are skipped, the redundant calculation amount is greatly reduced, and the entire calculation process is more concise and efficient.

[0150] 4. Reduce the influence of manual experience

[0151] In the traditional process, many parameters need to be repeatedly adjusted and judged manually. This method realizes end-to-end fully automatic calculation, avoiding accidental errors caused by manual experience.

[0152] 5. Improve the processing effect of low signal-to-noise ratio data

[0153] In low signal-to-noise ratio records, repeated iteration is often difficult to converge. This method achieves the goal in one step and has better adaptability to low signal-to-noise ratio data.

[0154] 6. Reduce the algorithm and implementation difficulty

[0155] The calculation process is simple and straightforward, the method is easy to understand and implement, and the algorithm difficulty is also greatly reduced.

[0156] 7. Expand the application scope and be more flexible in processing

[0157] This method is not restricted by the frequency band and can theoretically process signals of any bandwidth, expanding the application scope and can be flexibly applied to 2D or 3D data.

[0158] 8. Lay a better foundation for subsequent menu analysis

[0159] Through more accurate and effective static correction, the subsequent imaging quality can be effectively improved, laying a better foundation for fine description analysis such as feature extraction and reservoir identification.

[0160] In summary, this embodiment proposes a rank-based NMO-free 3D near-surface correction scheme, which can perform residual static correction calculation, obtain more accurate static correction data, and does not require dynamic correction, avoiding the low processing efficiency caused by the need for multiple iterations to calculate NMO velocity values and residual static correction amounts in the existing static correction.

[0161] For other detailed descriptions and advantages of this embodiment, reference can be made to the corresponding descriptions in the foregoing embodiments, which will not be elaborated here.

[0162] Example 3

[0163] According to another aspect of the present invention, an electronic device is also provided. The electronic device includes:

[0164] A memory storing executable instructions;

[0165] A processor that runs the executable instructions in the memory to implement the rank-based NMO-free 3D near-surface correction method according to the present invention.

[0166] The method includes the following steps 1 to 6.

[0167] The method includes:

[0168] Step 1: Arrange the 3D seismic data into a 2D matrix to form a 2D matrix D;

[0169] Step 2: Perform a Fourier transform on the matrix D to obtain a matrix E;

[0170] Step 3: Loop through all frequencies and ranks of the matrix E, calculate the singular value decomposition of each frequency slice, and then traverse each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of the frequency slice;

[0171] Step 4: Aggregate the low-rank approximations F of all ranks of all frequency slices, and perform an inverse Fourier transform on the aggregated matrix to obtain a matrix G;

[0172] Step 5: Calculate the cross-correlation between the matrix G and the matrix D, and select the maximum delay of the cross-correlation as the static correction amount;

[0173] Step 6: Use the static correction amount to perform static correction on the 3D seismic data.

[0174] In some embodiments, step 1 specifically includes:

[0175] Place the x coordinates in the source point coordinates and receiver point coordinates of the 3D seismic data on one dimension of the matrix, and place the y coordinates in the source point coordinates and receiver point coordinates on the other dimension of the matrix.

[0176] In some embodiments, calculating the singular value decomposition of each frequency slice specifically includes:

[0177] For each frequency, take out the frequency slice of this frequency from the matrix E, denoted as sub-matrix A.

[0178] Establish the singular value decomposition equation of the sub-matrix A;

[0179] Based on the singular value decomposition equation, obtain the singular values and corresponding eigenvectors of A by solving the eigenvalue problem of the sub-matrix A.

[0180] In some embodiments, at each frequency slice, when traversing each rank, it is in ascending order.

[0181] In some embodiments, based on the singular value decomposition result, traverse each rank to obtain the low-rank approximation F of each rank at this frequency slice, including:

[0182] For rank n, based on the first n singular values and the corresponding eigenvectors obtained by singular value decomposition, obtain the low-rank approximation F of rank n.

[0183] Specifically, the memory may include one or more computer program products, and the computer program products may include various forms of computer-readable storage media, such as volatile memory and / or non-volatile memory. The volatile memory may include, for example, random access memory (RAM) and / or cache memory, etc. The non-volatile memory may include, for example, read-only memory (ROM), hard disk, flash memory, etc.

[0184] The processor may be a central processing unit (CPU) or other forms of processing units with data processing capabilities and / or instruction execution capabilities, and may control other components in the electronic device to perform desired functions. In an embodiment of the present invention, the processor is used to run the computer-readable instructions stored in the memory.

[0185] The beneficial effects of this embodiment at least include:

[0186] 1. Improve the calculation accuracy of static correction

[0187] The present invention abandons the surface consistency assumption, directly estimates the static correction amount for the original record, avoids the static correction residuals caused by surface inhomogeneity, and the calculation result is closer to the true optimal solution, improving the effectiveness of static correction;

[0188] 2. No need for NMO correction

[0189] The NMO velocity error does not affect the remaining static correction estimation, reducing the multiple iterations of NMO velocity and the remaining static correction amount estimation. Therefore, in order to avoid the NMO stretching effect, it is also not necessary to cut the data after NMO correction;

[0190] 3. Simplify the calculation process and improve efficiency

[0191] Since the steps of calculating the NMO velocity and static correction amount through multiple iterations are skipped, the redundant operation amount is greatly reduced, and the entire calculation process is more concise and efficient.

[0192] 4. Reduce the influence of manual experience

[0193] In the traditional process, many parameters need to be manually adjusted and interpreted repeatedly. This method realizes end-to-end fully automatic calculation, avoiding accidental errors caused by manual experience;

[0194] 5. Improved the processing effect of low signal-to-noise ratio data

[0195] In low signal-to-noise ratio records, repeated iteration often fails to converge. This method achieves the goal in one step and has better adaptability to low signal-to-noise ratio data;

[0196] 6. Reduced the algorithm and implementation difficulty

[0197] The calculation process is simple and straightforward, the method is easy to understand and implement, and the algorithm difficulty is also greatly reduced.

[0198] 7. Expanded the application scope and is more flexible in processing

[0199] This method is not limited by the frequency band. In theory, it can process signals of any bandwidth, expanding the application scope and can be flexibly applied to two-dimensional or three-dimensional data.

[0200] 8. Laid a better foundation for subsequent menu analysis

[0201] Through more accurate and effective static correction, the subsequent imaging quality can be effectively improved, laying a better foundation for fine description analysis such as feature extraction and reservoir identification.

[0202] In summary, this embodiment proposes a rank-based non-NMO three-dimensional near-surface correction scheme, which can perform residual static correction calculation, obtain more accurate static correction data, and does not require dynamic correction, avoiding the low processing efficiency caused by the need for multiple iterative calculations of NMO velocity values and residual static correction amounts in existing static correction.

[0203] For the detailed description of this embodiment, reference can be made to the corresponding descriptions in the foregoing embodiments, which will not be elaborated here.

[0204] Example 4

[0205] According to another aspect of the present invention, a computer-readable storage medium is also provided. The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, it implements the rank-based non-NMO three-dimensional near-surface correction method according to the present invention.

[0206] The method includes the following steps 1 to 6.

[0207] The method includes the following steps 1 to 6.

[0208] The method includes:

[0209] Step 1, arrange the three-dimensional seismic data into a two-dimensional matrix to form a two-dimensional matrix D;

[0210] Step 2: Perform a Fourier transform on matrix D to obtain matrix E;

[0211] Step 3: Loop through all frequencies and ranks of matrix E, calculate the singular value decomposition of each frequency slice, and then traverse each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice;

[0212] Step 4: Aggregate the low-rank approximations F of all ranks of all frequency slices, and perform an inverse Fourier transform on the aggregated matrix to obtain matrix G;

[0213] Step 5: Perform cross-correlation on matrix G and matrix D, and select the maximum delay of the cross-correlation as the static correction amount;

[0214] Step 6: Use the static correction amount to perform static correction on the three-dimensional seismic data.

[0215] In some embodiments, step 1 specifically includes:

[0216] Place the x coordinates in the source point coordinates and receiver point coordinates of the three-dimensional seismic data on one dimension of the matrix, and place the y coordinates in the source point coordinates and receiver point coordinates on another dimension of the matrix.

[0217] In some embodiments, calculating the singular value decomposition of each frequency slice specifically includes:

[0218] For each frequency, extract the frequency slice of this frequency from matrix E, denoted as sub-matrix A.

[0219] Establish a singular value decomposition equation for sub-matrix A;

[0220] Based on the singular value decomposition equation, obtain the singular values of A and the corresponding eigenvectors by solving the eigenvalue problem of sub-matrix A.

[0221] In some embodiments, under each frequency slice, when traversing each rank, it is in ascending order.

[0222] In some embodiments, traversing each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice includes:

[0223] For rank n, based on the first n singular values and the corresponding eigenvectors obtained from the singular value decomposition, obtain the low-rank approximation F of rank n.

[0224] According to the computer-readable storage medium of the embodiments of the present invention, non-temporary computer-readable instructions are stored thereon. When the non-temporary computer-readable instructions are run by a processor, all or part of the steps of the methods of the various embodiments of the present invention described above are executed.

[0225] The above computer-readable storage mediums include, but are not limited to: optical storage mediums (e.g., CD-ROMs and DVDs), magneto-optical storage mediums (e.g., MOs), magnetic storage mediums (e.g., magnetic tapes or external hard drives), media with built-in rewritable non-volatile memories (e.g., memory cards), and media with built-in ROMs (e.g., ROM cartridges).

[0226] Those skilled in the art should understand that, in order to solve the technical problem of how to obtain good user experience effects, the present embodiment may also include well-known structures such as communication buses and interfaces, and these well-known structures should also be included within the protection scope of the present invention.

[0227] The beneficial effects of the present embodiment at least include:

[0228] 1. The calculation accuracy of static correction is improved

[0229] The present invention abandons the surface-consistent assumption, directly estimates the static correction amount for the original records, avoids the static correction residuals caused by surface inhomogeneities, and the calculation results are closer to the true optimal solution, improving the effectiveness of static correction;

[0230] 2. No NMO correction is required

[0231] The NMO velocity error does not affect the residual static correction estimation, reducing the multiple iterations of NMO velocity and the estimation of residual static correction amount. Therefore, in order to avoid the NMO stretching effect, it is also not necessary to excise the data after NMO correction;

[0232] 3. The calculation process is simplified and the efficiency is improved

[0233] Since the steps of calculating the NMO velocity and static correction amount through multiple iterations are skipped, the redundant computation amount is greatly reduced, and the entire calculation process is more concise and efficient.

[0234] 4. The influence of manual experience is reduced

[0235] In the traditional process, many parameters need to be repeatedly adjusted and judged manually. This method realizes end-to-end fully automatic calculation, avoiding the accidental errors caused by manual experience;

[0236] 5. The processing effect of low signal-to-noise ratio data is improved

[0237] In low signal-to-noise ratio records, repeated iterations are often difficult to converge. This method achieves the goal in one step and has better adaptability to low signal-to-noise ratio data;

[0238] 6. The algorithm and implementation difficulty are reduced

[0239] The calculation process is simple and straightforward, the method is easy to understand and implement, and the algorithm difficulty is also greatly reduced.

[0240] 7. Expand the application scope and be more flexible in processing

[0241] This method is not limited by the frequency band. In theory, it can process signals with any bandwidth, expanding the application scope and can be flexibly applied to 2D or 3D data.

[0242] 8. Lay a better foundation for subsequent menu analysis

[0243] Through more accurate and effective static correction, the subsequent imaging quality can be effectively improved, laying a better foundation for fine description analysis such as feature extraction and reservoir identification.

[0244] In summary, this embodiment proposes a rank-based NMO-free 3D near-surface correction scheme, which can perform residual static correction calculation, obtain more accurate static correction data, and does not require dynamic correction, avoiding the low processing efficiency caused by the need for multiple iterative calculations of NMO velocity values and residual static correction amounts in existing static correction.

[0245] For the detailed description of this embodiment, reference can be made to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0246] Example 5

[0247] To verify the effectiveness of the method in the present invention, the data of a certain exploration area was tested. Figure 2 is a 30Hz frequency slice without static correction problems; Figure 3 is a 30Hz frequency slice with static correction problems; Figure 4 is a 30Hz frequency slice processed according to the embodiment of the present invention. It can be seen that the signal-to-noise ratio of the frequency slice processed by the present invention is significantly improved, verifying the effectiveness of the solution of the present invention.

[0248] For other detailed descriptions of this embodiment, reference can be made to the corresponding descriptions in the foregoing embodiments, which will not be repeated here.

[0249] The embodiments of the present invention have been described above. The above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and variations are obvious to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The choice of terms used herein is intended to best explain the principles of the embodiments, practical applications, or technical improvements to the technologies in the market, or to enable other ordinary skill in the art in this technical field to understand the embodiments disclosed herein.

Claims

1. A rank-based 3D near-surface correction method without NMO, characterized in that, the method includes: Step 1, arrange the 3D seismic data into a 2D matrix to form a 2D matrix D; Step 2, perform a Fourier transform on matrix D to obtain matrix E; Step 3, loop through all frequencies and ranks of matrix E, calculate the singular value decomposition of each frequency slice, and then traverse each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice; Step 4, summarize the low-rank approximations F of all ranks of all frequency slices, and perform an inverse Fourier transform on the summarized matrix to obtain matrix G; Step 5, perform cross-correlation on matrix G and matrix D, and select the maximum delay of the cross-correlation as the static correction amount; Step 6, use the static correction amount to perform static correction on the 3D seismic data.

2. The method according to claim 1, characterized in that, Step 1 specifically includes: Place the x coordinates in the source point coordinates and receiver point coordinates of the 3D seismic data on one dimension of the matrix, and place the y coordinates in the source point coordinates and receiver point coordinates on the other dimension of the matrix.

3. The method according to claim 1, characterized in that, Calculating the singular value decomposition of each frequency slice specifically includes: For each frequency, extract the frequency slice of this frequency from matrix E, denoted as sub-matrix A. Establish the singular value decomposition equation of sub-matrix A; Based on the singular value decomposition equation, obtain the singular values and corresponding eigenvectors of A by solving the eigenvalue problem of sub-matrix A.

4. The method according to claim 1, characterized in that, At each frequency slice, when traversing each rank, it is in ascending order.

5. The method according to claim 1, characterized in that, Based on the singular value decomposition result, traversing each rank to obtain the low-rank approximation F of each rank of this frequency slice includes: For rank n, based on the first n singular values and corresponding eigenvectors obtained from the singular value decomposition, obtain the low-rank approximation F of rank n.

6. A rank-based 3D near-surface correction device without NMO, characterized in that, the device includes: A matrix unit, used to arrange the 3D seismic data into a 2D matrix to form a 2D matrix D; A Fourier transform unit, used to perform a Fourier transform on matrix D to obtain matrix E; A loop calculation unit, used to loop through all frequencies and ranks of matrix E, calculate the singular value decomposition of each frequency slice, and then traverse each rank based on the singular value decomposition result to obtain the low-rank approximation F of each rank of this frequency slice; An inverse Fourier transform unit, used to summarize the low-rank approximations F of all ranks of all frequency slices, and perform an inverse Fourier transform on the summarized matrix to obtain matrix G; A static correction amount calculation unit, used to perform cross-correlation on matrix G and matrix D, and select the maximum delay of the cross-correlation as the static correction amount; A static correction unit, used to perform static correction on the 3D seismic data using the static correction amount.

7. The device according to claim 6, characterized in that, The matrix unit is specifically used for: Place the x - coordinates of the shot points and receiver points in a three - dimensional seismic data on one dimension of the matrix, and place the y - coordinates of the shot points and receiver points on another dimension of the matrix.

8. The apparatus according to claim 6, wherein, Traversing each rank based on the singular value decomposition result to obtain the low - rank approximation F of each rank of this frequency slice includes: For rank n, based on the first n singular values and the corresponding eigenvectors obtained from the singular value decomposition, obtain the low - rank approximation F of rank n.

9. An electronic device, wherein, The electronic device includes: A memory storing executable instructions; A processor that runs the executable instructions in the memory to implement the method according to any one of claims 1 - 5.

10. A computer - readable storage medium storing a computer program, which when executed by a processor implements the method according to any one of claims 1 - 5.

Images