Three-dimensional multi-seismic-source mixed collection seismic data rapid separation method based on prior information constraint

Abstract

The invention discloses a three-dimensional multi-seismic-source mixed collection seismic data rapid separation method based on prior information constraint, and the method comprises the steps: dividing three-dimensional seismic data into a series of slices along an x axis or a y axis, carrying out the multi-scale and multi-direction two-dimensional curvelet transformation of the three-dimensional seismic data to obtain a curvelet domain coefficient, and introducing a conventional iterative shrinkage threshold algorithm for separation; then, according to the characteristic that curvelet coefficient support sets of two adjacent slices have a large intersection, a priori information constraint condition is established in a curvelet domain by using a priori support set provided by a previous slice of which aliasing noise is removed, the constraint condition is fused into an iterative shrinkage threshold algorithm, and an estimation precision factor formula is proposed; and a novel index threshold parameter formula is adopted to perform multi-seismic source mixed collection seismic data separation. According to the method, the signal-to-noise ratio of the separated seismic data is increased, the calculation speed is increased, the requirement for a computer memory is lowered, weak effective wave signals are recovered, and the single-shot reflected wave event is more continuous and clearer.

Description

Technical Field

[0001] The present invention relates to the field of three-dimensional multi-source mixed sampling seismic data rapid separation method, and specifically discloses a three-dimensional multi-source mixed sampling seismic data rapid separation method based on prior information constraints. Background Art

[0002] Multi-source acquisition, also known as hybrid acquisition, can effectively reduce acquisition costs and improve data quality, but it also generates significant aliasing noise, which can negatively impact the effectiveness of subsequent processing methods (Zu et al., 2017; Zhang et al., 2021). Therefore, eliminating aliasing noise has become a challenge in multi-source seismic exploration. Although many traditional seismic processing methods, such as multiple attenuation, wavefield inversion, and migration imaging, can be directly applied to hybrid data (Ma et al., 2016; Chen et al., 2017; Bai et al., 2018), these methods struggle to suppress the strong aliasing noise when the aliasing factor is high, resulting in poor processing quality and failing to meet the requirements of high-precision seismic interpretation. It is well known that during seismic data acquisition, due to the use of dithering strategies, in hybrid seismic data, in non-common-shot regions, the signal is correlated while the aliasing noise is random. Therefore, the current mainstream processing method for hybrid seismic data is to eliminate aliasing noise and separate the hybrid data into single-shot data generated by conventional single sources for subsequent processing and interpretation. Therefore, to improve the effectiveness of traditional processing methods, removing mixed noise becomes a key preprocessing step (Huang et al., 2018). Generally speaking, multi-source data processing methods are mainly divided into two types: filtering-based denoising and sparse inversion-based separation methods. These two types of separation methods play different roles at different stages. However, numerous studies have shown that inversion-based separation methods are more effective than filtering-based denoising methods (Chen et al., 2015; Gan et al., 2016). Among all inversion-based separation methods, sparse transforms, including Fourier transform, wavelet transform, Radon transform, and curvelet transform, are often used as constraints for multi-source separation, achieving good separation results (Abma et al., 2010; Ibrahim and Sacchi, 2013; Zhang et al., 2019; Wang et al., 2021). To further improve the separation effect, many methods have been extended to the separation of 3D mixed-sample seismic data. However, for the separation of 3D mixed-sample seismic data, a common processing method is to divide the 3D data into a series of 2D non-common shot gathers and separate them separately to improve computing time and save memory space (Wang et al., 2020).However, studies by Friedlander et al. and Mansour et al. have shown that the support sets of the curvelet domain coefficients of adjacent seismic data gathers are highly correlated (Friedlander et al., 2011; Mansour et al., 2013). The traditional method of independently separating non-common-shot gathers one by one prevents the effective utilization of the hidden prior information between gathers, reducing the separation accuracy of mixed-sample data. Furthermore, traditional algorithms are slow when separating three-dimensional mixed-sample seismic data. In particular, when using the highly redundant curvelet transform as a sparse basis for multi-source data separation, although separation accuracy can be guaranteed, the computational speed is far from meeting the requirements of processing large-scale mixed-sample data (Zhang et al., 2022).

[0003] To this end, the present invention uses a multi-scale, multi-directional 2D curvelet transform to transform each slice of the 3D mixed data along the y-axis into the curvelet domain. An iterative threshold shrinkage algorithm is then introduced to suppress aliasing noise on each slice individually. During this process, based on the large intersection between the support sets of the curvelet domain coefficients of two adjacent slices, a priori information constraint is established using the prior information provided by adjacent slices that have already been aliased. This constraint is then incorporated into the iterative shrinkage threshold algorithm, which then rapidly separates adjacent slices to be aliased. Summary of the Invention

[0004] The purpose of the present invention is to be able to separate field multi-source mixed seismic data with high precision, greatly improve the separation accuracy and calculation speed, save memory space, separate weak effective wave signals, and make the main source reflection wave phase axis more continuous and clear. A three-dimensional multi-source mixed seismic data rapid separation method based on prior information constraints is proposed.

[0005] In order to achieve the above object, the technical solution adopted by the present invention is as follows:

[0006] A method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints includes the following steps:

[0007] S1. Divide the 3D multi-source seismic data into a series of slices along the x-axis or y-axis, and perform a multi-scale and multi-directional 2D curvelet transform on them to obtain the curvelet domain coefficients. Since the effective signal is continuous and the aliasing noise is random and uncorrelated, a traditional iterative shrinkage threshold algorithm is introduced for separation.

[0008] S2. Based on the large intersection between the support sets of the curvelet coefficients of two adjacent slices, a priori information constraint condition is established in the curvelet domain using the prior support set provided by the previous slice with aliasing removed.

[0009] S3. Incorporate the constraint conditions into the iterative shrinkage threshold algorithm, propose an estimation precision factor formula, and use a new exponential threshold parameter formula to separate multi-source mixed seismic data, thus realizing a rapid separation method for three-dimensional multi-source mixed seismic data based on prior information constraints.

[0010] Furthermore, in the 3D common receiver gather domain, the multi-source mixing problem is expressed as follows:

[0011] d=d1+Γd2

[0012] Where d represents the mixed data of multiple sources, d1 and d2 represent the data excited by the source ship 1 and the source ship 2, respectively, and Γ is the random time delay code, which is expressed as follows:

[0013] Γ=F -1 PF,

[0014] Here F and F -1 Represents the forward and reverse Fourier transform respectively, and P is the following diagonal matrix

[0015]

[0016] where t k Indicates the random delay time of the kth shot in source 2.

[0017] In the three-dimensional common detection point domain, for each mixed seismic record, the main source is correlated, while the mixed noise is uncorrelated. Therefore, in the sparse transform domain, the main source coefficient is sparse, and the amplitude energy is larger than the aliasing noise. The aliasing noise can be suppressed by the iterative shrinkage threshold algorithm to restore the main source signal. Apply the inverse transform operator Γ to the above equation -1 , we can get the following equation:

[0018] Γ -1 d = Γ -1 d1+d2.

[0019] Γ -1 d represents pseudo-separation data. Combining the above equations, we can get an augmented estimation equation

[0020]

[0021] in

[0022]

[0023] Assuming that the coefficient x is the sparse representation of the data m in the curvelet domain C, the equation can be expressed as:

[0024]

[0025] Here, H represents the conjugate rank, and the sparse solution can be obtained by solving the following optimization problem:

[0026]

[0027] Furthermore, the traditional iterative shrinkage threshold algorithm is as follows:

[0028] In order to solve the above equation, we use the iterative shrinkage threshold algorithm to solve the underdetermined ill-conditioned equation. The iterative format of the iterative method is as follows:

[0029]

[0030] Where n = 1, 2, ..., N. N represents the maximum number of iterations, T λ The soft threshold operator representing the threshold λ is as follows:

[0031] T λ (x) = sign(x) max((x) - λ, 0)

[0032] Furthermore, based on the large intersection between the support sets of the curvelet coefficients of two adjacent slices, we use the prior support set provided by the previous slice that has been de-aliased to establish a priori information constraint in the curvelet domain. The constraint is expressed as follows:

[0033]

[0034] Among them, w i represents the weight value (i=1,2,3,…,J), and its value is as follows:

[0035]

[0036] It represents the a priori support set provided by the curvelet domain coefficients of the l-1th time slice, which means directly taking the position corresponding to the maximum curvelet coefficient of a certain part. The weight value γ outside the a priori support set is generally between 0 and 1. If it is selected too small, it may damage the effective signal. If it is selected too large, the filtering process will not be performed. In particular, when γ = 1, the method of the present invention will degenerate into a traditional iterative shrinkage threshold algorithm.

[0037] Furthermore, the constraint condition is incorporated into the iterative shrinkage threshold algorithm, which is expressed as follows:

[0038]

[0039] Among them, y l Indicates the lth slice;

[0040] After a maximum of N iterations, the curvelet coefficients that remove the aliasing noise can be obtained, and then the separated main source data can be obtained by the following formula:

[0041]

[0042] Furthermore, the formula for estimating the precision factor is as follows:

[0043]

[0044] In the formula and The support sets of the lth and l+1th slice curvelet coefficients are represented. The larger a is, the greater the overlap ratio of the same support sets of adjacent slices is, indicating that the method of the present invention is more effective.

[0045] Furthermore, the new index threshold parameter formula is as follows:

[0046]

[0047] Where Max is the maximum value of |Cy|, that is, the maximum absolute value of the curvelet coefficient.

[0048] To address the problem that traditional 3D multi-source mixed-sample seismic data separation methods ignore the potential correlation between adjacent slices, this paper first divides the 3D multi-source mixed-sample seismic data into a series of slices along the y-axis. A multi-scale and multi-directional 2D curvelet transform is then performed to obtain the curvelet domain coefficients. An iterative shrinkage threshold algorithm is then introduced for reconstruction. In this process, based on the large intersection between the curvelet coefficient support sets of two adjacent slices, a priori information constraint condition is established in the curvelet domain using the prior support set provided by the previous slice that has been de-aliased. This constraint condition is then incorporated into the iterative shrinkage threshold algorithm, and a novel exponential threshold parameter formula is used for multi-source mixed-sample seismic data separation. This results in a rapid 3D multi-source mixed-sample seismic data separation method based on prior information constraints.

[0049] The beneficial effects of the present invention are as follows: 1. Compared with the filter-based multi-source separation method, the separation accuracy of the present invention is higher and it can process multi-source mixed seismic data with a high degree of mixing;

[0050] 2. Compared with the 3D multi-source mixed sampling seismic data separation based on the iterative shrinkage threshold algorithm, the technology of the present invention not only improves the single-shot separation accuracy and speeds up the calculation, but also reduces the requirement for computer memory, separates the weak effective wave signal, and makes the main source reflection wave phase axis more continuous and clear;

[0051] In summary, the present invention overcomes the shortcomings of traditional multi-source data separation methods that cannot quickly and accurately separate mixed seismic data, as well as the problems of low separation accuracy and limited application of traditional separation methods. Moreover, the invented method can not only separate mixed seismic data with a large degree of mixing with high precision, but also separate noisy multi-source seismic data, greatly improving the calculation speed and signal-to-noise ratio of the separation method, saving computing space, and restoring weak effective wave signals, thereby making the reflection wave phase axis of the main source seismic record after separation more continuous and clear. This has important practical value in guiding the improvement of field seismic data acquisition efficiency and saving field exploration costs. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to more clearly illustrate the technical solutions in the embodiments of the present application, a brief introduction will be given below to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0053] Figure 1 This is a flowchart of multi-source seismic data separation in an embodiment of the present invention;

[0054] Figure 2a The theoretical seismic data of the present invention Figure 1 ;

[0055] Figure 2b This is the theoretical seismic data Figure 2 of the present invention;

[0056] Figure 3 Curvelet spectra of adjacent slices;

[0057] Figure 4a This is the separation result of the present invention under different algorithms Figure 1 ;

[0058] Figure 4b Figure 2 is the separation results of the present invention under different algorithms;

[0059] Figure 5a is the separation error of different iterative algorithms of the present invention Figure 1 ;

[0060] Figure 5b Figure 2 is the separation error diagram of different iterative algorithms of the present invention;

[0061] Figure 6 is a graph showing the number of iterations and the signal-to-noise ratio after separation according to the present invention;

[0062] Figure 7 This is a signal-to-noise ratio curve diagram after separation at different weight values ​​of the present invention;

[0063] Figure 8 is a graph showing the signal-to-noise ratio after separation of different slices according to the present invention;

[0064] Figure 9 is a graph of the estimated precision factor in an embodiment of the present invention. DETAILED DESCRIPTION

[0065] The technical solution of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the embodiments described are only some embodiments of the present invention, not all embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0066] Reference Figure 1 A method for rapid separation of three-dimensional multi-source mixed-sample seismic data based on prior information constraints is proposed. The steps for implementing the method mainly include: establishing a seismic data reconstruction model, a traditional iterative shrinkage threshold algorithm, establishing prior information constraints, an iterative shrinkage threshold algorithm based on the constraints, an exponential threshold parameter formula, and an estimation precision factor formula.

[0067] The specific steps are as follows:

[0068] Step 1: Establish the data reconstruction model; in the common receiver gather domain, the 3D multi-source mixing problem is expressed as follows:

[0069] d=d1+Γd2

[0070] Where d represents the mixed data of multiple sources, d1 and d2 represent the data excited by source ship 1 and source ship 2 respectively, and Γ is the random time delay code, which is expressed as follows:

[0071] Γ=F -1 PF,

[0072] Here F and F -1 Represents the forward and reverse Fourier transform respectively, and P is the following diagonal matrix

[0073]

[0074] where t k Indicates the random delay time of the kth shot in source 2.

[0075] In the three-dimensional common detection point domain, for each mixed seismic record, the main source is correlated, while the mixed noise is uncorrelated. Therefore, in the sparse transform domain, the main source coefficient is sparse, and the amplitude energy is larger than the aliasing noise. The aliasing noise can be suppressed by the iterative shrinkage threshold algorithm to restore the main source signal. Apply the inverse transform operator Γ to the above equation -1 , we can get the following equation:

[0076] Γ -1 d = Γ -1 d1+d2.

[0077] Γ -1 d represents pseudo-separation data. Combining the above equations, we can get an augmented estimation equation

[0078]

[0079] in

[0080]

[0081] Assuming that the coefficient x is the sparse representation of the data m in the curvelet domain C, the equation can be expressed as:

[0082]

[0083] Here, H represents the conjugate rank, and the sparse solution can be obtained by solving the following optimization problem:

[0084]

[0085] Step 2: Traditional iterative shrinkage threshold algorithm; Solving the above equation is an underdetermined ill-conditioned problem, and it is not easy to get an exact solution. In order to solve the above equation, the traditional iterative shrinkage threshold algorithm can be used to solve the underdetermined ill-conditioned equation. The iterative format of this iterative method is as follows:

[0086]

[0087] Where n = 1, 2, ..., N. N represents the maximum number of iterations, T λ The soft threshold operator representing the threshold λ is as follows:

[0088] T λ (x) = sign(x) max((x) - λ, 0)

[0089] Step 3: Establish prior information constraints. Based on the large intersection between the support sets of the curvelet coefficients of two adjacent slices, the prior information constraints are established in the curvelet domain using the prior support set provided by the previous slice that has been de-aliased. The constraint is expressed as follows:

[0090]

[0091] Among them, w i represents the weight value (i=1,2,3,…,J), and its value is as follows:

[0092]

[0093] It represents the a priori support set provided by the curvelet domain coefficients of the l-1th time slice, which means directly taking the position corresponding to the maximum curvelet coefficient of a certain part. The weight value γ outside the a priori support set is generally between 0 and 1. If it is selected too small, it may damage the effective signal. If it is selected too large, the filtering process will not be performed. In particular, when γ = 1, the method of the present invention will degenerate into a traditional iterative shrinkage threshold algorithm.

[0094] Step 4: Iterative shrinkage threshold algorithm based on constraints. Incorporating constraints into the iterative shrinkage threshold algorithm, its expression is as follows:

[0095]

[0096] Among them, y l represents the lth slice. After a maximum of N iterations, the curvelet coefficients that remove the aliasing noise can be obtained. Then, the separated main source data can be obtained by the following formula:

[0097]

[0098] Step 5: Estimation of the precision factor formula. In order to understand the accuracy of the 3D multi-source mixed sampling seismic data reconstruction, the proposed estimation precision factor formula is as follows:

[0099]

[0100] In the formula and The support sets of the lth and l+1th slice curvelet coefficients are represented. The larger a is, the greater the overlap ratio of the same support sets of adjacent slices is, indicating that the method of the present invention is more effective.

[0101] Step 6: Exponential threshold parameter formula. In the process of solving the above nonlinear problem, the threshold parameter λ n Satisfied, ||Cy|| ∞ =λ1>λ2>…>ε, where ε is a small value close to zero. Commonly used threshold parameters do not meet the required accuracy and efficiency. To improve efficiency, the present invention introduces a new exponentially decaying threshold parameter, which allows the threshold parameter to gradually decrease by a larger amplitude. While ensuring solution accuracy, it also accelerates convergence to meet the requirements of massive seismic data processing. The formula for this new exponential threshold parameter is as follows:

[0102]

[0103] Where Max is the maximum value of |Cy|, that is, the maximum absolute value of the curvelet coefficient.

[0104] The specific operations to implement this method are:

[0105] In order to measure the separation effect of the method of the present invention, 20log 10 ||f0||2 / ||f-f0||2 is used as the signal-to-noise ratio formula, where f0 represents the unmixed seismic data and f represents the demixed seismic data. The larger the signal-to-noise ratio, the higher the accuracy of the method.

[0106] The sizes of the three-dimensional common detection point theoretical seismic data along the x-axis, y-axis and time direction are 750, 256 and 256 respectively. Then, an aliasing operator with a random delay time range of 300 to 300ms is used to obtain artificial three-dimensional aliased seismic data.

[0107] In order to display the 3D seismic data in detail, the 64th slice is selected in the y-axis direction for display. The unmixed seismic data and the aliased seismic data after pseudo separation are shown in FIG. Figure 2a-2b shown.

[0108] In order to show the correlation between different slices, two adjacent slices of theoretical data are extracted and their amplitudes in the curvelet domain are plotted, as shown in Figure 3 As shown in the figure, it can be seen that the support sets of the curvelet spectra of adjacent slices have considerable overlap. The prior support set provided by the previous slice with aliasing noise removed is used to establish the prior information constraint condition, which can provide the prior support set information for the next slice to be separated and aliasing noise removed. In order to demonstrate the advantages of the method of the present invention, the traditional threshold iteration algorithm is first used to separate the aliasing noise. The number of iterations is 18, and the separation results are shown in the figure. Figure 4a As shown in Figure 1, the signal-to-noise ratio after separation is 22.18dB. It can be seen that although most of the aliasing noise has been removed and the useful signal has been restored, some aliasing noise has not been removed completely. Then the separation method of the present invention is used, the weight value γ is 0.3, and the prior support set is the position corresponding to the first 2% maximum curvelet coefficients. The separation results are shown in Figure 1. Figure 4b As shown, the signal-to-noise ratio is 28.34 dB, which shows that the separation accuracy of the method of the present invention is high.

[0109] Figure 5a-5b This is the error diagram after separation of the traditional method and the method of the present invention. It can be seen that the three-dimensional multi-source mixed sampling data separation method proposed in the present invention produces a small error and almost no loss of effective wave signals. The reason is that new constraints have been added, which can effectively utilize the prior information of the previous slice for removing aliasing noise, provide support set information for the next adjacent slice to remove aliasing noise, delineate the main effective wave energy distribution range, and suppress low-energy random noise outside the range, thereby suppressing aliasing noise with high precision.

[0110] In order to further compare the separation effect of the method of the present invention with that of the traditional method under different iteration times, the weight value γ is set to 0.3, the prior support set is set to the number of the first 2% maximum curvelet coefficients, and then the maximum number of iterations is changed from 2 to 60, and the results are analyzed in turn. Figure 2b Separate and calculate the relationship between the maximum number of iterations each time and the signal-to-noise ratio after separation, such as Figure 6 As shown, it can be seen that no matter how many iterations there are, the separation effect of the method of the present invention is better than that of the traditional separation method. It also shows that the use of prior information of adjacent slices can effectively improve the separation accuracy. It can also be seen that in order to obtain a separation result with a signal-to-noise ratio of about 25.5dB, the traditional method requires 60 iterations, while the separation method based on prior information constraints proposed by the present invention only requires about 11 iterations, which can save nearly 5 times the computing time of the traditional threshold iterative separation method. Although the calculation of the weight value takes a little time, it is negligible compared to the entire calculation time. In general, the signal-to-noise ratio after reconstruction of these two iterative algorithms increases with the increase of the number of iterations, but too many iterations will waste computing time. For the traditional method, when the number of iterations is greater than 18, the increase in the signal-to-noise ratio is relatively small as the number of iterations increases. For this reason, the number of iterations selected for the theoretical simulation data application of the present invention is 18.

[0111] In fact, in the reconstruction process, the selection of the weight value γ is particularly important. If it is too small or too large, the effect of removing aliasing noise will not be good. Therefore, the weight value is kept changing from 0 to 1, and then the signal-to-noise ratio curve after separation by the method of the present invention is calculated. The result is as follows Figure 7 As shown in the figure, it can be seen that when γ is 0.3, the signal-to-noise ratio after removing aliasing noise is the highest. When the weight value is greater than or less than 0.3, the signal-to-noise ratio decreases. This is why the weight value of 0.3 is selected in the theoretical simulation experiment of the present invention. It should be noted that if the weight value is 1, the method of the present invention degenerates into a conventional iterative shrinkage threshold algorithm.

[0112] Since the theoretical data has 256 slices in the y-axis direction, in order to further compare the advantages of the method of the present invention, Figure 8 The figure shows the signal-to-noise ratio curves after the separation of the entire three-dimensional multi-source theoretical seismic data using the traditional method and the method of the present invention. Generally speaking, in each aliased seismic data separation, the separation effect of the method of the present invention is higher than the corresponding signal-to-noise ratio of the traditional method, which also shows that there is indeed a large intersection between the support sets of the curved wave coefficients of adjacent slices. The previous slice can provide prior support set information for the next adjacent slice to be aliased. In the past, the potential correlation between adjacent slices was ignored, so the aliasing removal effect was affected to a certain extent. Since the method of the present invention utilizes the prior information between different slices and adds constraints, the separation effect is better.

[0113] It should be pointed out that since the present invention only uses 18 iterations, if the number of iterations continues to increase, the difference between the signal-to-noise ratio after separation by the traditional method and the signal-to-noise ratio after separation by the present invention will gradually narrow. Figure 9 The curve of the estimated accuracy a of the theoretical data is shown in Figure 2. For the theoretical data, the estimated accuracy a of most slices along the y-axis is about 0.8, and the change trend is also similar to that of the theoretical data. Figure 8 This further demonstrates that theoretical data can use prior support information as a constraint to improve the separation accuracy of aliasing noise.

[0114] Therefore, compared with the filter-based multi-source separation method, the separation technology of the present invention has higher separation accuracy and can process multi-source mixed seismic data with a high degree of mixing;

[0115] Compared with the 3D multi-source mixed-sample seismic data separation based on the iterative shrinkage threshold algorithm, the technology of the present invention not only improves the single-shot separation accuracy and speeds up the calculation, but also reduces the requirement for computer memory. It also separates the weak effective wave signal and makes the main source reflection wave phase axis more continuous and clear.

[0116] In summary, the present invention overcomes the shortcomings of traditional multi-source data separation methods that cannot quickly and accurately separate mixed seismic data, as well as the problems of low separation accuracy and limited application of traditional separation methods. Moreover, the invented method can not only separate mixed seismic data with a large degree of mixing with high precision, but also separate noisy multi-source seismic data, greatly improving the calculation speed and signal-to-noise ratio of the separation method, saving computing space, and restoring weak effective wave signals, thereby making the reflection wave phase axis of the main source seismic record after separation more continuous and clear. This has important practical value in guiding the improvement of field seismic data acquisition efficiency and saving field exploration costs.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints, characterized in that: The steps include: S1. Divide the three-dimensional multi-source seismic data into a series of slices along the x-axis or y-axis, and perform multi-scale and multi-directional two-dimensional curvelet transform on them to obtain curvelet domain coefficients. Since the effective signal is continuous and the aliasing noise is random and uncorrelated, the traditional iterative shrinkage threshold algorithm is introduced for separation; S2. Based on the large intersection between the support sets of the curvelet coefficients of two adjacent slices, the prior information constraint condition is established in the curvelet domain using the prior support set provided by the previous slice from which the aliasing noise has been removed; S3. Incorporate constraints into the iterative shrinkage threshold algorithm, propose an estimation precision factor formula, use a new exponential threshold parameter formula to separate multi-source mixed seismic data, and realize a three-dimensional multi-source mixed seismic data rapid separation method based on prior information constraints.

2. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 1 is characterized in that: In the 3D common receiver gather domain, the multi-source mixing problem is expressed as follows: d=d1+Γd2 Where: d represents multi-source mixed data, d1 and d2 represent the data excited by source ship 1 and source ship 2 respectively, Γ is the random time delay code, and its expression is: Γ=F -1 PF, Here F and F -1 They represent the forward and reverse Fourier transforms respectively, and P is the following diagonal matrix where t k Represents the random delay time of the kth shot in source 2.

3. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 2 is characterized in that: In the three-dimensional common detection point domain, for each mixed seismic record, the main source is correlated, while the mixed noise is uncorrelated. Therefore, in the sparse transform domain, the main source coefficient is sparse, and the amplitude energy is larger than the aliasing noise. The aliasing noise can be suppressed by the iterative shrinkage threshold algorithm to restore the main source signal. Applying the inverse transform operator Γ to the above equation -1 , we can get the following equation: C -1 d=C -1 d1+d2. Γ -1 d represents pseudo-separation data. Combining the above equations, we can get an augmented estimation equation in Assuming that the coefficient x is a sparse representation of the data m in the curvelet domain C, the equation can be expressed as: And A=FC H Here, H represents the conjugate transformation rank, and the sparse solution can be obtained by solving the following optimization problem:

4. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 1 is characterized in that: The traditional iterative shrinkage threshold algorithm is as follows: In order to solve the above equation, an iterative shrinkage threshold algorithm is used to solve the underdetermined ill-conditioned equation. The iterative format of the iterative method is as follows: Where n = 1, 2, ···, N. N represents the maximum number of iterations, T λ The soft threshold operator representing the threshold λ is given by the following formula: T λ (x)=sign(x)·max((x)-λ,0)。 5. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 1 is characterized in that: According to the large intersection between the support sets of the curvelet coefficients of two adjacent slices, the prior information constraint condition is established in the curvelet domain using the prior support set provided by the previous slice from which the aliasing noise has been removed. The constraint condition is expressed as follows: Among them, w i represents the weight value (i=1,2,3,···,J), and its value is as follows: It represents the a priori support set provided by the curvelet domain coefficients of the l-1th time slice, which means directly taking the position corresponding to a certain part of the maximum curvelet coefficient. The weight value γ outside the priori support set is generally between 0 and 1. If it is selected too small, it may damage the effective signal. If it is selected too large, the filtering process will not work. In particular, when γ=1, the method will degenerate into a traditional iterative shrinkage threshold algorithm.

6. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 5 is characterized in that: The constraint condition is incorporated into the iterative shrinkage threshold algorithm, and its expression is as follows: Among them, y l represents the lth slice; after a maximum of N iterations, the curve coefficients of the aliasing noise can be obtained, and then the separated main source data can be obtained by the following formula:

7. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 1, characterized in that: The formula for estimating the precision factor is as follows: In the formula and Represents the support set of the lth and l+1th slice curvelet coefficients. The larger a is, the greater the overlap ratio of the same support set of adjacent slices.

8. The method for rapid separation of three-dimensional multi-source mixed sampling seismic data based on prior information constraints according to claim 1 is characterized in that: The new index threshold parameter formula is as follows: Where Max is the maximum value of |Cy|, that is, the maximum absolute value of the curve coefficient.

Images