A sparse domain seismic diffraction wave separation method

Abstract

The present invention relates to the technical field of multiple wave suppression in seismic exploration, and more specifically, to a sparse domain seismic diffraction wave separation method, comprising constructing diffraction wave signals of different curvatures using the theory that the time-distance curve of the diffraction wave is a hyperbola and the amplitude energy is inversely proportional to the propagation distance, and forming a complete diffraction wave signal atom library. Then, the diffraction wave signal is iteratively decomposed using a matching pursuit algorithm to obtain sparse diffraction wave signal atoms, and finally the least squares technique is used to solve the diffraction wave signal atom coefficient obtained by the decomposition, and the separated diffraction wave data is obtained by weighted superposition of the diffraction wave signal atoms and their coefficients. The method of the present invention is simple, has high calculation efficiency, and good noise resistance. It has a good application effect on low signal-to-noise ratio data, and can also better solve the problem of separation integrity between the diffraction wave vertex and the two wings, and the diffraction wave signal is more thoroughly decomposed. As one of the conventional processes for seismic data processing, it has a wide range of applications and high promotion value.

Description

Technical Field

[0001] The present invention relates to the technical field of multiple wave suppression in seismic exploration, and more specifically, to a sparse domain seismic diffraction wave separation method. Background Art

[0002] With the deepening of exploration and development, the significance of small-scale, high-resolution seismic imaging in high-precision seismic exploration has become increasingly important. The focus of seismic exploration is to study the information related to lateral heterogeneous bodies such as faults, river channels, fractures, pinch points, and dissolved pore-type reservoirs. These target geological bodies have the characteristics of small scale, irregular shape, and strong spatial heterogeneity. Unlike reflection waves, seismic waves encounter these small-scale geological bodies during propagation, forming secondary sources to produce a large number of diffraction waves. Compared with reflection waves, these diffraction waves have obvious differences in both dynamics and kinematics. Therefore, making good use of diffraction wave imaging is of great significance for the exploration of lateral heterogeneous bodies. At the same time, it can also better explore the structural development zone and find favorable closures. The amplitude of diffraction waves is much weaker than that of reflection waves, and they are usually submerged in the reflection wave information and difficult to identify. Conventional seismic processing mainly targets primary reflection waves, and usually causes certain damage to diffraction waves other than continuous reflections during noise suppression. Therefore, a separate imaging technology for diffraction waves has been developed.

[0003] Diffraction wave separation imaging can be divided into "direct method" and "indirect method" diffraction wave imaging. The former is to directly image the diffraction wave according to the difference in characteristics between the reflection wave and the diffraction wave during or after the migration process, and the latter is to separate the diffraction wave from the wave field of the track gather before migration, and then migrate the separated diffraction wave for imaging. The "direct method" has the characteristics of high computational efficiency, while the diffraction wave separation technology of the track gather before migration mostly adopts the iterative inversion method and the amount of data for wave field separation is very large. The diffraction wave separation takes a very long time and has low computational efficiency. In addition, the diffraction wave energy of the diffraction wave separation of the track gather before migration is relatively weak. In areas with low signal-to-noise ratio, the diffraction wave signal is often difficult to identify. How to separate reliable weak diffraction wave signals from the strong reflection data has always been a technical difficulty of diffraction wave separation imaging.

[0004] The method for separating diffraction waves from pre-migration gathers mainly utilizes the differences in kinematic and dynamic characteristics between reflected waves and diffraction waves. Commonly used techniques include singular value filtering that utilizes the difference in amplitude energy between reflected waves and diffraction waves, dip filtering that utilizes the difference in travel time of seismic waves, or Radon transform filtering. These techniques are difficult to solve the problem of separating the apex and two wings of the diffraction wave at the same time. In addition, for low signal-to-noise ratio data, the diffraction wave separation effect is often not ideal due to the influence of noise.

[0005] For example, the prior art discloses a diffraction wave separation method based on near-path edge sparse Radon transform, which uses efficient sparse Radon transform to perform wave field separation on a pre-stack track gather containing diffraction waves to obtain separated reflection waves and diffraction waves. Although this separation method can effectively reduce the near-path edge effect of the Radon transform and improve the separability of the reflection wave and the diffraction wave in the Radon domain, this separation method still has the problems of low noise resistance and poor separation effect. Summary of the invention

[0006] In order to overcome the technical problem that the diffraction wave separation technology described in the above-mentioned prior art is affected by noise and the diffraction wave separation effect is not very ideal, the present invention provides a sparse domain seismic diffraction wave separation method with high noise resistance and more thorough diffraction wave signal separation.

[0007] In order to solve the above technical problems, the technical solution adopted by the present invention is: a sparse domain seismic diffraction wave separation method, the method comprising:

[0008] Step 1: Using the theory that the time-distance curve of the diffraction wave is a hyperbola and the amplitude energy is inversely proportional to the propagation distance, construct diffraction wave signals with different curvatures and generate a diffraction wave signal atomic library;

[0009] Step 2: Using the matching pursuit algorithm, iteratively decompose and search for signal atoms in the diffraction wave signal atom library obtained in step 1 to obtain sparse diffraction wave signal atoms; and using the least square method to solve the coefficients of the sparse signal atoms in each iteration of the matching pursuit decomposition process;

[0010] Step 3: Obtain separated diffraction wave data by weighted superposition of the diffraction wave signal atoms and their coefficients obtained in step 2.

[0011] The present invention uses a complete diffraction wave signal atom library to perform diffraction wave decomposition in a sparse domain, which has good noise resistance and is effective for low signal-to-noise ratio data. It can also better solve the problem of separation integrity between the diffraction wave vertex and the two wings. Compared with traditional matching pursuit technology, it uses iterative matching pursuit to solve the coefficients of the diffraction wave signal atoms, and the diffraction wave signal decomposition is more thorough.

[0012] Furthermore, in the step 1, before generating the diffraction wave signal atomic library, data needs to be input and rearranged, specifically: input shot gathers or common center point gathers, and rearrange the input shot gathers or common center point gathers according to the offset distance to obtain the common offset gather d; at the same time, input the root mean square velocity field v rms .

[0013] Furthermore, in the step 1, after inputting the data and rearranging the data, the parameters in the diffraction wave separation process are set, including the curvature range [σ1-σ2] of the diffraction wave signal atom, the curvature interval Δσ, the damping coefficient ε obtained by the least square method, and the minimum acceptable correlation coefficient threshold α between the diffraction wave signal atom and the actual data. acp , the maximum acceptable residual percentage ζ acp , the maximum number of matching pursuit iterations N.

[0014] Furthermore, in the step 1, a diffraction wave signal atom library Atom(σ,t0,x) is generated, wherein:

[0015] The time-distance curve calculation formula of the diffraction wave signal adopts the hyperbola formula, Where t0 is the two-way reflection time of the diffraction wave vertex, x is the horizontal distance from the hyperbola vertex, and i is the curvature index, i = 1, 2, ..., N.

[0016] Furthermore, in the step 1, a diffraction wave signal atom library Atom(σ,t0,x) is generated, wherein:

[0017] The amplitude of the diffracted wave is calculated using the formula Where A0 and v are the seismic amplitude and RMS velocity at the vertex of the hyperbola, respectively; x is the horizontal distance from the vertex of the hyperbola; and a and b are unknown coefficients related to the amplitude.

[0018] Furthermore, generating the diffraction wave signal atom library Atom(σ, t0, x) in step 1 further includes the following operations:

[0019] The least squares fitting is performed based on the searched diffraction wave signal, that is, ||A(x)-A real (x)||→min,A real (x) is the amplitude value of the actual diffraction wave data.

[0020] Furthermore, the specific operation of searching for diffraction wave signal atoms in step 2 is:

[0021] (I) Set input data d;

[0022] Initialize residual data r0 = d, output diffraction wave seismic data d s = 0, the searched diffraction wave atom library Atom searched =Φ, number of iterations k = 0;

[0023] Time sample point loop processing;

[0024] Cyclic processing of seismic traces;

[0025] (II), matching pursuit loop: k≤N;

[0026] (III) Calculate the diffraction wave signal atomic library Atom (σ, t0, x) and the data residual r k Correlation coefficient

[0027] (IV) Calculate the maximum index value of the correlation coefficient C(σ) m = argmax(C(σ i )),i=1,2,...,N,if C(σ m )>α acp Then go to step (V), otherwise end the matching pursuit loop;

[0028] (V) Add the searched diffraction wave signal atom to the searched atom library. searched =Atom searched ∪Atom(σ m ,t0,x), and exclude it from the atom library of the next iteration. Indicates the removal of atomic operations.

[0029] Furthermore, solving the coefficients of the sparse signal atoms in the step 2 is solving the global optimization coefficients of the diffraction wave atoms, and the equation used is an overdetermined equation.

[0030] Furthermore, the specific operation of solving the coefficients of the sparse signal atoms in step 2 is:

[0031] Let L = Atom searched , the least square method is used to obtain the global optimization coefficient y of the searched diffraction wave atom searched =(L T L+εI) -1 L T d, T represents matrix transpose, I represents the unit diagonal matrix, ε represents the damping coefficient solved by least squares, and d represents the common offset gather.

[0032] Furthermore, the specific operations of step three are:

[0033] Reconstructing diffraction wave seismic data s =Ly searched ;

[0034] Update residual r k+1 =dd s , calculate the percentage coefficient of the residual If ζ>ζ acp , let k = k + 1, then go to step (III), otherwise end the matching pursuit loop;

[0035] After the matching pursuit cycle ends;

[0036] The cyclic processing of seismic trace numbers is completed;

[0037] The time sample point loop processing ends;

[0038] Output diffraction wave seismic data d s .

[0039] Compared with the prior art, the present invention has the following beneficial effects:

[0040] 1) In the present invention, the present invention adopts a complete diffraction wave signal atomic library to perform diffraction wave decomposition in a sparse domain, so its noise resistance is very good, and the application effect is significant for low signal-to-noise ratio data. At the same time, it can also better solve the problem of separation integrity of the diffraction wave vertex and the two wings;

[0041] 2) Compared with the traditional matching pursuit technology, the present invention adopts iterative matching pursuit to solve the coefficients of the diffraction wave signal atoms, and the diffraction wave signal is decomposed more thoroughly;

[0042] 3) The overall method of the present invention is simple and has high computational efficiency. As one of the conventional processes for seismic data processing, it has a wide range of applications and is worthy of popularization and application. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 It is a flow chart of the sparse domain seismic diffraction wave separation method of the present invention;

[0044] Figure 2 It is a specific flow chart of the sparse domain seismic diffraction wave separation method of the present invention;

[0045] Figure 3 This is a noise-free data separation effect diagram of the sparse domain seismic diffraction wave separation method of the present invention;

[0046] Figure 4 This is a diagram showing the separation effect when the diffraction wave signal and noise signal-to-noise ratio of the sparse domain seismic diffraction wave separation method of the present invention is -6db;

[0047] Figure 5 This is a diagram showing the separation effect when the signal-to-noise ratio of the diffraction wave signal to the noise in the sparse domain seismic diffraction wave separation method of the present invention is -20db. DETAILED DESCRIPTION

[0048] The drawings are only for illustrative purposes and cannot be construed as limiting the present invention. To better illustrate the present embodiment, some parts of the drawings may be omitted, enlarged, or reduced, and do not represent the size of the actual product. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the drawings. The positional relationships described in the drawings are only for illustrative purposes and cannot be construed as limiting the present invention.

[0049] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if the terms "upper", "lower", "left", "right", "long", "short" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limitations on this patent. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0050] The technical solution of the present invention is further described in detail below through specific embodiments and in conjunction with the accompanying drawings:

[0051] Example 1

[0052] like Figure 1-Figure 2 A sparse domain seismic diffraction wave separation method is shown, comprising:

[0053] Step 1: Using the theory that the time-distance curve of the diffraction wave is a hyperbola and the amplitude energy is inversely proportional to the propagation distance, construct diffraction wave signals with different curvatures and generate a diffraction wave signal atomic library;

[0054] Step 2: Using the matching pursuit algorithm, iteratively decompose and search for signal atoms in the diffraction wave signal atom library obtained in step 1 to obtain sparse diffraction wave signal atoms; and using the least square method to solve the coefficients of the sparse signal atoms in each iteration of the matching pursuit decomposition process;

[0055] Step 3: Obtain separated diffraction wave data by weighted superposition of the diffraction wave signal atoms and their coefficients obtained in step 2.

[0056] The specific operation of the above step 1 is: input data and data rearrangement: rearrange the input shot gather or common center point gather according to the offset distance to obtain the common offset gather d, where shot gather and common center point gather are existing professional terms and will not be repeated here; at the same time, input the root mean square velocity field v rms The main purpose of data rearrangement is to make full use of the good hyperbolic characteristics of the diffraction wave signal in the common offset domain, so as to facilitate the matching pursuit decomposition using the hyperbolic diffraction wave signal atomic library, and the root mean square velocity field is used for the subsequent amplitude fitting of the diffraction wave signal atomic library;

[0057] Set the parameters in the diffraction wave separation process: the curvature range of the diffraction wave signal atom [σ1~σ2], the curvature interval Δσ, the damping coefficient ε of the least square solution, and the minimum acceptable correlation coefficient threshold α between the diffraction wave signal atom and the actual data acp , the maximum acceptable residual percentage ζacp , Due to the influence of the absolute size of the amplitude, the residual definition of the present invention adopts a more adaptable relative percentage to measure the size of the residual energy, and the maximum number of matching pursuit iterations N;

[0058] Generate a diffraction wave signal atom library Atom (σ, t0, x). Each atom in the diffraction wave signal atom library consists of a hyperbolic travel time and the corresponding amplitude, and only has a value at the position of the hyperbola. The time-distance curve calculation formula of the diffraction wave signal adopts the hyperbola formula. Where t0 is the two-way reflection time of the diffraction wave vertex, x is the distance from the hyperbola vertex in the horizontal direction, i is the curvature index, i=1,2,...,N; since the energy of the diffraction wave signal obviously decays with the increase of the lateral distance from the diffraction point, the traditional diffraction wave separation technology based on hyperbolic Radon transform generally does not consider the lateral change of the amplitude, and there is a certain degree of residual in the diffraction wave energy separation. The present invention adopts an approximate relationship between the lateral change of the diffraction wave amplitude, and the relationship between the amplitude of the diffraction wave and the absolute distance of the fixed point is expressed by the formula: Where A0 and v are the seismic amplitude and root mean square velocity at the vertex of the hyperbola, respectively, and a and b are unknown coefficients related to the amplitude. The least squares fitting is performed based on the searched diffraction wave signal, that is: ||A(x)-A real (x)||→min,A real (x) is the amplitude value of the actual diffraction wave data;

[0059] The specific operation of searching for diffraction wave signal atoms in the above step 2 is:

[0060] (I) Set initial parameters:

[0061] Initialize residual data r0 = d, output data d s = 0, the searched diffraction wave atom library Atom searched =Φ, number of iterations k = 0;

[0062] Time sample point loop processing;

[0063] Cyclic processing of seismic traces;

[0064] (II), matching pursuit loop: k≤N;

[0065] (III) Calculate the diffraction wave signal atomic library Atom (σ, t0, x) and the data residual r k Correlation coefficient

[0066] (IV) Calculate the index value with the largest correlation coefficient m = argmax(C(σ i )),i=1,2,...,N,if C(σm )>α acp Then go to step (V), otherwise end the matching pursuit loop;

[0067] (V) Add the searched diffraction wave signal atom to the searched atom library. searched =Atom searched ∪Atom(σ m ,t0,x), and exclude it from the atom library of the next iteration. Indicates that atomic operations are removed and each atom will not be used in repeated searches.

[0068] The specific operation of solving the coefficients of sparse signal atoms in the above step 2 is:

[0069] Let L = Atom searched , compared with the sparse diffraction wave atoms, the sampling points of the diffraction wave data are much larger than their number, so the equation for solving the global optimization coefficient of the diffraction wave atoms is an overdetermined equation. The least squares method is used to obtain the global optimization coefficient solution of the searched diffraction wave atoms and is expressed as: y searched =(L T L+εI) -1 L T d, T represents matrix transpose, I represents unit diagonal matrix, ε represents the damping coefficient solved by least squares, and d represents common offset gathers;

[0070] The specific operations of the above step three are:

[0071] Reconstructing diffraction wave seismic data s =Ly searched ;

[0072] Update residual r k+1 =dd s , calculate the percentage coefficient of the residual If ζ>ζ acp , let k = k + 1, then go to step (III), otherwise end the matching pursuit loop;

[0073] The matching pursuit loop ends;

[0074] The cyclic processing of seismic trace numbers is completed;

[0075] The time sample loop processing ends.

[0076] Output diffraction wave seismic data d s .

[0077] The present invention uses a complete diffraction wave signal atom library to perform diffraction wave decomposition in a sparse domain, and its noise resistance is relatively good, and the application effect is more significant for low signal-to-noise ratio data. At the same time, it can also better solve the problem of separation integrity between the diffraction wave vertex and the two wings, so that the diffraction wave signal separation is relatively complete. In addition, compared with the traditional matching pursuit technology, the present invention uses iterative matching pursuit to solve the coefficients of the diffraction wave signal atoms, and the diffraction wave signal decomposition is more thorough. The overall method of the present invention is simple, and the calculation efficiency is very high. As one of the conventional processes for seismic data processing, it has a wide range of applications and is worthy of application and promotion.

[0078] The advantages of the method of the present invention are illustrated below by a specific data test. The specific operation is as follows: the test data of the present invention uses synthetic seismic data, the data track spacing is 12.5m, there are 8 groups of adjacent diffraction waves with a certain spacing in the data, and the diffraction spacing changes are 12.5m, 25m, 37.5m, 50m, 62.5m, 75m, 87.5m, and 100m. The reflection coefficient at the vertex of the left branch diffraction wave is 0.0023, and the reflection coefficient at the vertex of the right branch diffraction wave is 0.0028. The data also contains two upper and lower horizontal layered interfaces, and their reflection coefficients are 0.01 and 0.012 respectively.

[0079] In order to test the diffraction wave separation effect of the present invention, this embodiment uses noise-free data for testing. The test results include four sub-graphs (A), (B), (C), and (D), and their data are: Figure (A) is the input data graph, Figure (B) is the separated diffraction wave data graph, Figure (C) is the other data graph other than the diffraction wave, and Figure (D) is the difference graph between the separated diffraction wave and the real diffraction wave.

[0080] like Figure 3 It can be seen that the reflected wave signal and the diffraction wave signal are completely separated. The separated diffraction wave signal is relatively complete and thorough, and the difference with the real diffraction wave signal is very small, making it difficult to identify with the naked eye.

[0081] Example 2

[0082] This embodiment is an embodiment 2 of a sparse domain seismic diffraction wave separation method. The difference between this embodiment and embodiment 1 is that: in this embodiment, the test data uses synthetic seismic data, the data track spacing is 12.5m, there are 8 groups of adjacent diffraction waves with a certain spacing in the data, and the diffraction spacing changes are 12.5m, 25m, 37.5m, 50m, 62.5m, 75m, 87.5m, and 100m respectively. The reflection coefficient at the vertex of the left branch diffraction wave is 0.0023, and the reflection coefficient at the vertex of the right branch diffraction wave is 0.0028. The data also contains two upper and lower horizontal layered interfaces, whose reflection coefficients are 0.01 and 0.012 respectively.

[0083] In order to test the diffraction wave separation effect of the present invention, this embodiment uses data with a signal-to-noise ratio of 12db for testing. The test results include four sub-graphs (A), (B), (C), and (D), and their data are: Figure (A) is the input data graph, Figure (B) is the separated diffraction wave data graph, Figure (C) is the other data graph other than the diffraction wave, and Figure (D) is the difference graph between the separated diffraction wave and the real diffraction wave.

[0084] like Figure 4 It can be seen that as the amplitude of the two wings of the diffraction wave gradually decays with the increase of distance, the diffraction wave signal becomes difficult to identify relative to the vertex of the diffraction wave. If the separation is based on the slope, the diffraction wave signal is often difficult to separate well due to its low signal-to-noise ratio. However, the diffraction wave signal separation of the method of the present invention is still relatively complete and thorough. Figure 4 The difference between the separated diffraction wave and the true diffraction wave in (D) can also be seen, and there is only a slight difference between the two.

[0085] Example 3

[0086] This embodiment is an embodiment 3 of a sparse domain seismic diffraction wave separation method. The difference between this embodiment and embodiment 1 is that: in this embodiment, the test data uses synthetic seismic data, the data track spacing is 12.5m, there are 8 groups of adjacent diffraction waves with a certain spacing in the data, and the diffraction spacing changes are 12.5m, 25m, 37.5m, 50m, 62.5m, 75m, 87.5m, and 100m respectively. The reflection coefficient at the vertex of the left branch diffraction wave is 0.0023, and the reflection coefficient at the vertex of the right branch diffraction wave is 0.0028. The data also contains two upper and lower horizontal layered interfaces, and their reflection coefficients are 0.01 and 0.012 respectively.

[0087] In order to test the diffraction wave separation effect of the present invention, this embodiment uses data with a signal-to-noise ratio of 6db for testing. The test results include four sub-graphs (A), (B), (C), and (D), and their data are: Figure (A) is the input data graph, Figure (B) is the separated diffraction wave data graph, Figure (C) is the other data graph other than the diffraction wave, and Figure (D) is the difference graph between the separated diffraction wave and the real diffraction wave.

[0088] like Figure 5 It can be seen that as the noise energy becomes stronger, the signals on both wings of the diffraction wave are basically unrecognizable, and only a relatively clear diffraction wave signal can be seen near the apex of the diffraction wave. From the separation results, the diffraction wave separation effect is still relatively ideal. Although the difference between the real diffraction wave and the separated diffraction wave has a stronger energy, its energy value is still relatively weak compared to the real diffraction wave energy.

[0089] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, and are not intended to limit the embodiments of the present invention. For those skilled in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to list all the embodiments here. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the protection scope of the claims of the present invention.

Claims

1. A sparse domain seismic diffraction wave separation method, characterized in that: The following steps are involved: Step 1: Using the theory that the time-distance curve of the diffraction wave is a hyperbola and the amplitude energy is inversely proportional to the propagation distance, construct diffraction wave signals with different curvatures and generate a diffraction wave signal atomic library; Step 2: Using the matching pursuit algorithm, iteratively decompose and search for signal atoms in the diffraction wave signal atom library obtained in step 1 to obtain sparse diffraction wave signal atoms; and using the least square method to solve the coefficients of the sparse signal atoms in each iteration of the matching pursuit decomposition process; Step 3: Obtain separated diffraction wave data by weighted superposition of the diffraction wave signal atoms and their coefficients obtained in step 2; In the step 1, data needs to be input and rearranged before generating the diffraction wave signal atomic library, specifically: input shot gathers or common center point gathers, and rearrange the input shot gathers or common center point gathers according to the offset distance to obtain the common offset distance gather d; at the same time, input the root mean square velocity field v rms In the step 1, after inputting the data and rearranging the data, the parameters in the diffraction wave separation process are set, including the curvature range [σ1~σ2] of the diffraction wave signal atom, the curvature interval Δσ, the damping coefficient ε obtained by least squares, the minimum acceptable correlation coefficient threshold α between the diffraction wave signal atom and the actual data acp , the maximum acceptable residual percentage ζ acp , the maximum number of matching pursuit iterations N; in the step 1, the diffraction wave signal atom library Atom (σ, t0, x) is generated, wherein the time-distance curve calculation formula of the diffraction wave signal adopts the hyperbola formula, Where t0 is the two-way reflection time of the vertex of the diffraction wave, x is the distance from the vertex of the hyperbola in the horizontal direction, i is the curvature index, i=1,2,...,N; in the step 1, the diffraction wave signal atom library Atom(σ,t0,x) is generated, wherein the amplitude of the diffraction wave is calculated using the formula Where A0 and v are the seismic amplitude and root mean square velocity at the vertex of the hyperbola, respectively; x is the distance from the vertex of the hyperbola in the horizontal direction; a and b are unknown coefficients related to the amplitude; generating the diffraction wave signal atom library Atom(σ,t0,x) in step 1 also includes the following operations: performing least squares fitting according to the searched diffraction wave signal, that is, ||A(x)-A real (x)||→min,A real (x) is the amplitude value of the actual diffraction wave data.

2. A sparse domain seismic diffraction wave separation method according to claim 1, characterized in that: The specific operation of searching for diffraction wave signal atoms in step 2 is: (I) Set initial parameters: Initialize residual data r0 = d, output diffraction wave seismic data d s = 0, the searched diffraction wave atom library Atom searched =Φ, number of iterations k = 0; Time sample point loop processing; Cyclic processing of seismic traces; (II), matching pursuit loop: k≤N; (III) Calculate the diffraction wave signal atomic library Atom (σ, t0, x) and the data residual r k Correlation coefficient (IV) Calculate the maximum index value of the correlation coefficient C(σ) m = argmax(C(σ i )),i=1,2,...,N,if C(σ m )>α acp , then go to step (V), otherwise end the matching pursuit loop; (V) adding the searched diffraction wave signal atom to the searched atom library, Atom searched =Atom searched ∪Atom(σ m ,t0,x), and exclude it from the atom library of the next iteration. Indicates the removal of an atomic operation.

3. A sparse domain seismic diffraction wave separation method according to claim 2, characterized in that: Solving the coefficients of the sparse signal atoms in the step 2 is solving the global optimization coefficients of the diffraction wave atoms, and the equation used is an overdetermined equation.

4. A sparse domain seismic diffraction wave separation method according to claim 3, characterized in that: The specific operation of solving the coefficients of the sparse signal atoms in step 2 is: Let L = Atom searched , the least square method is used to obtain the global optimization coefficient y of the searched diffraction wave atom searched =(L T L+εI) -1 L T d, T represents matrix transpose, I represents the unit diagonal matrix, ε represents the damping coefficient solved by least squares, and d represents the common offset gather.

5. A sparse domain seismic diffraction wave separation method according to claim 4, characterized in that: The specific operations of step three are: Reconstructing diffraction wave seismic data s =Ly searched ; Update residual r k+1 =dd s , calculate the percentage coefficient of the residual If ζ>ζ acp , let k = k + 1, then go to step (III), otherwise end the matching pursuit loop; The matching pursuit loop ends; The cyclic processing of seismic trace numbers is completed; The time sample point loop processing ends; Output diffraction wave seismic data d s .

Images