A method for removing shielded reflection wave based on generalized wavelet transform

Abstract

The application discloses a method for removing strong shielding reflection wave based on generalized wavelet transform, which comprises the following steps: S1, determining a wavelet matched with the physical form of an actual seismic wavelet as a mother wavelet; S2, constructing a wavelet dictionary according to the mother wavelet; S3, obtaining a wavelet coefficient vector according to the wavelet dictionary; S4, reconstructing a seismic signal containing a strong reflection background according to the wavelet coefficient vector and the wavelet dictionary; and S5, obtaining a weak reflection signal after removing shielding according to the reconstructed seismic signal containing the strong reflection background. The application adopts a main frequency focusing strategy, and only needs a single scale dense time shift to construct the wavelet dictionary, the dictionary scale is equal to the number of signal sampling points, and is far less than a multi-scale over-complete dictionary of a matching pursuit algorithm, so that the calculation efficiency is obviously improved.

Description

Technical Field

[0001] This invention belongs to the technical field, specifically relating to a method for removing strongly shielded reflected waves based on generalized wavelet transform. Background Technology

[0002] In seismic exploration, the seismic response characteristics of reservoirs are the core basis for hydrocarbon identification and reservoir prediction. However, in actual exploration, when strong impedance interfaces such as coal seams, gypsum-salt rocks, igneous rocks, and tight carbonate rocks exist underground, the huge impedance difference between these interfaces and the surrounding rocks generates extremely high-energy reflected waves. These strong reflected waves act as a "barrier," strongly shielding the seismic response of the underlying thin reservoir or weak reflective layer, resulting in the masking of effective signals, making reservoir identification difficult, significantly reducing the accuracy of seismic prediction, and directly affecting the exploration and development effectiveness of complex hydrocarbon reservoirs.

[0003] The core solution to this strong reflection shielding problem lies in "extraction and stripping of seismic wavelets". That is, through specific signal processing techniques, the original seismic signal is first decomposed into different components, then the waveform features representing strong reflections are extracted, and then stripped from the original signal to recover the underlying weak reflection signal.

[0004] Currently, seismic wavelet decomposition and reconstruction technology is the mainstream method in this field, the core of which lies in decomposing seismic signals into a linear combination of a series of wavelets. However, existing methods still have limitations in terms of basis function matching, coefficient solution accuracy, and fidelity in weak signals. For example, the matching pursuit algorithm is prone to getting trapped in local optima and has low computational efficiency; the basis functions of wavelet transform are fixed and differ greatly from the morphology of seismic wavelets, making it difficult to achieve high-fidelity separation when strong and weak signals are intertwined.

[0005] Therefore, there is an urgent need for a new wavelet decomposition and reconstruction method that can accurately match the physical form of seismic wavelets at the basis function level and obtain globally stable sparse solutions at the coefficient solution level, thereby achieving high-precision separation of strong and weak signals and high-fidelity recovery of weak reflection signals.

[0006] Wavelet transform is a classic method for time-frequency analysis and multi-wavelet decomposition of seismic signals, based on the theory of multi-resolution signal analysis. This method uses a set of orthogonal or bioorthogonal wavelet basis functions. Basis functions are generated through scaling and translation of the mother wavelet, representing the seismic signal as a linear combination of these basis functions. The wavelet coefficients are directly calculated through the inner product of the signal and the wavelet basis. In de-masking applications, the multi-scale decomposition capability of wavelet transform is mainly utilized to concentrate strong reflection energy into wavelet coefficients at a specific scale (frequency band). By suppressing these wavelet coefficients and then reconstructing the remaining signal, strong reflections are effectively removed.

[0007] The specific implementation process is as follows:

[0008] (1) Selecting wavelet basis functions

[0009] Choose an orthogonal or bioorthogonal wavelet basis according to the processing target. Commonly used wavelet families include Daubechies wavelets, Symlets wavelets, etc., and these basis functions have compact support.

[0010] (2) Multiscale wavelet decomposition

[0011] The original seismic signal is subjected to multi-scale discrete wavelet transform (DWT). Taking binary wavelet transform as an example, the signal is decomposed layer by layer through a set of high-pass and low-pass filters:

[0012] Level 1 decomposition: Obtains high-frequency detail coefficients CD1 and low-frequency approximation coefficients CA1.

[0013] Second-level decomposition: CA1 is further decomposed to obtain CD2 and CA2.

[0014] And so on, breaking it down to the preset scale.

[0015] (3) Identification and suppression of strong reflection coefficient

[0016] The main frequency band range corresponding to the strong reflective layer is determined through spectral analysis or forward modeling. In the multi-scale decomposition results, the strong reflection energy is usually concentrated in the detail coefficients or approximate coefficients at a few specific scales. These coefficients are then suppressed.

[0017] (4) Wavelet inverse transform reconstructed signal

[0018] The suppressed coefficients are subjected to inverse wavelet transform to reconstruct the seismic signal after strong reflection.

[0019] The disadvantage of this existing technology is:

[0020] (1) Mismatch between basis functions and the physical form of seismic wavelet: Wavelet transform uses mathematically constructed orthogonal wavelet bases, and the waveforms of these basis functions differ significantly from those of actual seismic wavelets. The mismatch between the basis functions and the physical form of the signal results in the decomposition results failing to accurately characterize the seismic reflection features, inaccurate extraction of strong reflection components, and distorted recovery of weak reflections.

[0021] (2) The wavelet coefficient inner product projection solution method is not suitable for non-orthogonal dictionaries: Under the condition of orthogonal basis, the inner product projection can obtain unique coefficients; however, in actual seismic signal processing, in order to better match the signal characteristics, overcomplete representation is often required. At this time, the inner product projection can no longer obtain sparse and stable coefficients, which affects the decomposition accuracy.

[0022] (3) Poor weak signal recovery capability: Due to the mismatch of basis functions and lack of sparsity constraints, when wavelet transform is used to deal with the interweaving of strong and weak signals, the weak reflection signal is easily submerged by the side lobes of the strong reflection. The recovered weak signal waveform is severely distorted, has low fidelity, and is difficult to use for fine reservoir prediction.

[0023] (4) Incomplete suppression of strong reflections and easy damage to weak signals: In the frequency band with extremely strong reflection energy, direct suppression of wavelet coefficients often makes it difficult to achieve a balance between "preserving weak signals" and "removing strong reflections", which can easily lead to strong reflection residues or loss of weak signals.

[0024] Matching pursuit algorithm is a sparse signal decomposition algorithm. Its core idea is to iteratively select the atom with the largest inner product with the current residual signal from an overcomplete atom library, and then approximate the original signal using a linear combination of these atoms, thereby extracting and removing strong reflection components. The specific implementation steps are as follows:

[0025] (1) Constructing a complete atom library: Select Ricker wavelet or Morlet wavelet as the mother function, and generate an atom library by densely sampling parameters such as time shift, frequency scaling and phase rotation.

[0026] (2) Greedy iterative decomposition: Initialize the residual as the original seismic signal. In each iteration, search for the atom with the largest inner product with the residual, record the coefficient, update the residual, until the residual energy is lower than a certain set threshold or the maximum number of iterations is reached.

[0027] (3) Identification of strong reflective atoms: Through spectrum analysis or layer information, determine the dominant frequency range and time window of the atoms corresponding to the strong reflective layer. From all the atoms obtained by decomposition, identify the set of atoms that belong to strong reflection, and superimpose these atoms to obtain the strong reflected wave.

[0028] (4) Stripping strong reflection waves: Stripping strong reflection waves from the signal to obtain the seismic signal after strong reflection removal.

[0029] The disadvantage of this prior art is:

[0030] (1) Greedy iteration is prone to getting trapped in local optima: The MP algorithm only selects the atom with the largest inner product of the current residual each time, which is a local optimal strategy. When the strong and weak signal spectra are seriously overlapping, it is easy to mismatch the weak reflection signal as the strong reflection sidelobe, resulting in the weak signal being incorrectly stripped or distorted.

[0031] (2) Low computational efficiency: Each iteration requires traversing the entire incomplete atomic library to calculate its inner product. The atomic library is large and the number of iterations is large, resulting in extremely high computational costs, which makes it difficult to meet the processing needs of large-area three-dimensional seismic data.

[0032] (3) Limited physical matching of dictionary atoms: Although the MP algorithm can use physical wavelets such as Ricker wavelets to construct the dictionary, in order to cover a sufficient parameter space, the dictionary usually contains a large number of atoms with different scales, time shifts and phases. The large dictionary not only increases the computational burden, but the high correlation between atoms also makes the decomposition results sensitive to the search order, affecting the stability and repeatability of the final decomposition. Summary of the Invention

[0033] To address the above problems, this invention proposes a method for removing strongly shielded reflected waves based on generalized wavelet transform.

[0034] The technical solution of this invention is: a method for removing strongly shielded reflected waves based on generalized wavelet transform, comprising the following steps:

[0035] S1. Determine the wavelet that matches the physical form of the actual seismic wavelet as the mother wavelet;

[0036] S2. Construct a wavelet dictionary based on the mother wavelet;

[0037] S3. Obtain the wavelet coefficient vector based on the wavelet dictionary;

[0038] S4. Based on the wavelet coefficient vector and wavelet dictionary, reconstruct the seismic signal containing the strong reflection background;

[0039] S5. Based on the reconstructed seismic signal containing a strong reflection background, the demasked weak reflection signal is obtained.

[0040] Furthermore, in S1, the time domain of the mother wavelet The expression is:

[0041] ;

[0042] in, For time, Main frequency, denoted as the time center of the wavelet, and e as the exponent.

[0043] Furthermore, wavelet dictionary The expression is:

[0044] ;

[0045] Where w is a column vector and N is the dimension.

[0046] Furthermore, in S3, an optimization problem is constructed and solved based on the wavelet dictionary to obtain the wavelet coefficient vector.

[0047] Furthermore, the expression for the optimization problem is:

[0048] ;

[0049] in, To obtain the solution, Let r be the wavelet dictionary, and r be the wavelet coefficients to be determined. For regularization parameters, It is an earthquake path.

[0050] Furthermore, the optimization problem includes L1 norm sparsity constraints.

[0051] Furthermore, in S4, a generalized inverse wavelet transform is performed based on the wavelet coefficient vector and the wavelet dictionary to reconstruct the seismic signal containing a strong reflection background.

[0052] Furthermore, seismic signals containing strong reflection backgrounds The expression is:

[0053] ;

[0054] in, These are the wavelet coefficients in the wavelet coefficient vector that correspond to this time window. This is a wavelet dictionary.

[0055] Furthermore, in S5, the seismic signal containing a strong reflection background is subtracted from the original seismic signal to remove the strong reflection background and output the deshielded weak reflection signal.

[0056] Furthermore, the weak reflection signal after deshielding is used to complete reservoir identification.

[0057] The beneficial effects of this invention are:

[0058] (1) Strong physical matching of basis functions: This invention breaks through the limitations of traditional orthogonal wavelet bases and selects wavelets (such as Ricker wavelets) that are consistent with the physical form of actual seismic wavelets to construct the wavelet dictionary, so that the decomposition results have clear geophysical meaning. Compared with the traditional wavelet transform using db, sym and other methods to construct wavelet bases, the basis functions of this invention are highly consistent with the physical form of seismic signals, which improves the decomposition accuracy from the source.

[0059] (2) Globally optimal and stable wavelet coefficient solution: This invention uses L1 norm sparse constraint optimization to globally solve the wavelet coefficients, avoiding the problem of greedy iteration in the matching pursuit algorithm easily getting trapped in local optima. Even when the dictionary is non-orthogonal and strong and weak signals are severely intertwined, L1 sparse optimization can still obtain stable, sparse and high-precision coefficient solutions.

[0060] (3) Precise strong reflection stripping and high fidelity of weak signal recovery: By using precisely solved sparse wavelet coefficients for coefficient separation and inverse transformation reconstruction within the strong reflection time window, a "surgical" precise stripping of strong reflections is achieved. Verification by examples shows that the weak reflection signal recovered after removing strong shielding is highly consistent with the theoretical real signal, with high waveform fidelity, avoiding the waveform distortion and energy leakage problems common in traditional methods.

[0061] (4) The theoretical framework has inheritance and extensibility: The generalized wavelet transform framework constructed in this invention is mathematically consistent with the traditional wavelet transform. At the same time, it achieves essential extension by introducing generalized basis and sparse constraints, and has a solid theoretical foundation and broad applicability potential.

[0062] (5) Simple and efficient dictionary construction: The present invention adopts the main frequency focusing strategy, which only requires a single-scale dense time shift to construct the wavelet dictionary. The size of the dictionary is equal to the number of signal sampling points, which is much smaller than the multi-scale overcomplete dictionary of the matching pursuit algorithm, and the computational efficiency is significantly improved. Attached Figure Description

[0063] Figure 1 The flowchart shows a method for removing strongly shielded reflected waves based on generalized wavelet transform.

[0064] Figure 2 Geological model diagram;

[0065] Figure 3 This is a sequence diagram of reflection coefficients;

[0066] Figure 4 To synthesize the seismic record map;

[0067] Figure 5 This is a diagram showing a theoretically weak reflection signal.

[0068] Figure 6 Ricker wavelet plot for 30Hz main frequency;

[0069] Figure 7 The wavelet dictionary matrix diagram;

[0070] Figure 8 Wavelet coefficient diagram;

[0071] Figure 9 A diagram showing the separation and reconstruction of wavelet coefficients with strong reflection;

[0072] Figure 10 Comparison of effects on reducing strong reflections. Detailed Implementation

[0073] The embodiments of the present invention will be further described below with reference to the accompanying drawings.

[0074] MP: Matching Tracking.

[0075] An overcomplete atomic library, also known as an overcomplete dictionary, is a set of atoms with a unit norm, where each atom is a wavelet with a different time shift, frequency, and scale.

[0076] DWT: Discrete Wavelet Transform.

[0077] LASSO: Minimal absolute shrinkage and selection operator.

[0078] FISTA: Fast Iterative Shrink Threshold Algorithm.

[0079] L1 norm: the sum of the absolute values ​​of the elements of a vector.

[0080] Ricker wavelet: A zero-phase seismic wavelet.

[0081] Morlet wavelet: A complex exponential oscillation wave modulated by Gaussian envelope, also called a Gabor wavelet.

[0082] db wavelet (Daubechies wavelet): Daubechies compactly supported orthogonal wavelet.

[0083] Symlet wavelet: An improved Daubechies-style approximately symmetric orthogonal compactly supported wavelet.

[0084] ADMM: Alternating Direction Multiplier Method, a convex relaxation algorithm for finding the global optimal sparse solution.

[0085] like Figure 1 As shown, this invention provides a method for removing strongly shielded reflected waves based on generalized wavelet transform, comprising the following steps:

[0086] S1. Determine the wavelet that matches the physical form of the actual seismic wavelet as the mother wavelet;

[0087] S2. Construct a wavelet dictionary based on the mother wavelet;

[0088] S3. Obtain the wavelet coefficient vector based on the wavelet dictionary;

[0089] S4. Based on the wavelet coefficient vector and wavelet dictionary, reconstruct the seismic signal containing the strong reflection background;

[0090] S5. Based on the reconstructed seismic signal containing a strong reflection background, the demasked weak reflection signal is obtained.

[0091] This invention is based on wavelet transform theory, which extends the wavelet basis functions from fixed orthogonal basis to generalized basis that matches the physical morphology of seismic wavelets. It uses L1 norm sparse constraint optimization to replace the traditional greedy iteration or inner product projection to solve the wavelet coefficients. In terms of application strategy, it introduces the idea of ​​dominant frequency focusing and constructs a wavelet dictionary based on the dominant frequency of the seismic signal, thereby achieving high-precision strong reflection stripping and high-fidelity recovery of weak signals.

[0092] Before introducing the specific technical solution, the "generalized wavelet transform" constructed in this invention will be defined and explained first:

[0093] The core mathematical form of wavelet transform is to represent a signal as a linear combination of wavelet basis functions:

[0094] ;

[0095] in, For wavelet basis functions, These correspond to the wavelet coefficients. Traditional wavelet transform directly obtains wavelet coefficients through inner product projection:

[0096] ;

[0097] This invention, while inheriting this core mathematical form, expands upon it theoretically in two dimensions:

[0098] 1) Extension of basis function systems: from orthogonal basis to generalized basis.

[0099] Traditional wavelet transform typically employs orthogonal or bioorthogonal wavelet bases, which possess excellent mathematical properties but differ from the physical morphology of actual seismic wavelets. This invention extends wavelet base functions to generalized bases, selecting wavelets whose morphology closely matches that of actual seismic wavelets as the wavelet bases, thus making the base function system more closely resemble the physical essence of seismic signals. This extension does not require orthogonality between base functions; instead, stable signal representation is achieved through subsequent sparse constraint optimization.

[0100] 3) Extension of wavelet coefficient solution methods: from inner product projection to sparse constraint optimization.

[0101] Under a generalized basis framework, wavelet dictionaries typically exhibit inter-atomic correlations, making inner product projection no longer sufficient to obtain sparse and stable coefficient solutions. Therefore, this invention extends the solution of wavelet coefficients from a deterministic inner product projection to a constrained optimization problem. Considering the sparse representation of seismic signals under a wavelet dictionary, L1 norm sparsity constraints are introduced to solve for the wavelet coefficients of the wavelet forward transform.

[0102] ;

[0103] in, For wavelet bases of different scales The constructed wavelet dictionary, Let S be the wavelet coefficients to be determined, and S be the actual seismic signal. This is the regularization parameter.

[0104] Within the constructed framework, the generalized wavelet transform is defined as follows:

[0105] The generalized wavelet transform refers to a class of analytical and synthetic transform frameworks that use wavelet bases with time-frequency localization characteristics as basic units to represent signals in the form of wavelet coefficients, where wavelet coefficients are obtained through sparse constraint optimization.

[0106] The generalized forward wavelet transform refers to analyzing a signal using a wavelet dictionary and obtaining its wavelet coefficients to acquire a wavelet domain representation of the signal. Its expression is:

[0107] ;

[0108] The generalized inverse wavelet transform refers to the reconstruction of a signal based on wavelet coefficients, thereby recovering its time-domain representation. Its expression is:

[0109] ;

[0110] This framework has the following implications:

[0111] Inheritance: When the wavelet dictionary satisfies the orthogonality condition and the regularization parameter When =0, The solution will degenerate into This is the inner product projection form of the traditional wavelet transform. Therefore, under certain conditions, the generalized wavelet transform and the traditional wavelet transform are consistent in form.

[0112] Extensibility: By introducing generalized bases and sparse constraints, the generalized wavelet transform extends wavelet analysis from "inner product projection transformation based on fixed bases" to "sparse optimization adaptive representation based on generalized bases", realizing the organic integration of the advantages of time-frequency localization and sparse representation capabilities.

[0113] Multi-scale inclusiveness: The generalized wavelet transform framework itself possesses multi-scale analysis capabilities, inheriting from the traditional wavelet transform, and theoretically still falls within the scope of multi-wavelet decomposition and reconstruction methods. This method addresses the practical need for strong shielding removal by constructing a wavelet dictionary based on the dominant frequency of the seismic signal. This does not negate multi-scale representation capabilities, but rather represents a scale optimization based on the band-limited characteristics and spectral dominance of the seismic signal. Strong reflection responses are mainly concentrated within the dominant frequency band, and the difference between strong and weak reflections is mainly manifested in the changes in the amplitude and energy distribution of the decomposition coefficients within that frequency band. Focusing on the dominant frequency band helps enhance the local characterization capability of the signal and improve the distinguishability of strong and weak reflections. Therefore, the selection of the dominant frequency can be regarded as a scale optimization based on information contribution and time-frequency characteristics within a multi-scale framework.

[0114] In this embodiment of the invention, in S1, the limitations of the traditional orthogonal wavelet basis are broken, and a wavelet that matches the physical morphology of the actual seismic wavelet is selected as the mother wavelet.

[0115] Time domain of mother wavelet The expression is:

[0116] ;

[0117] in, For time, Main frequency, denoted as the time center of the wavelet, and e as the exponent.

[0118] :Second; Peak frequency (Hz); π: Pi. Based on the obtained dominant frequency and with fixed scale parameters, a wavelet dictionary focusing the dominant frequency is constructed through intensive time shifting. This dictionary utilizes a multi-scale framework with a dominant frequency optimization strategy, ensuring that each atom is time-aligned with the sampling point and frequency-accurately matched to the target signal.

[0119] A sub-wavelet of the wavelet dictionary, It is a set of column vectors whose sample point time intervals are related to the seismic traces. It's the same, among which... This is the maximum value of the wavelet. Therefore, the wavelet... Wavelet dictionaries can be constructed .

[0120] In this embodiment of the invention, in step S3, an optimization problem is constructed and solved based on the wavelet dictionary to obtain the wavelet coefficient vector.

[0121] In this embodiment of the invention, the wavelet dictionary The expression is:

[0122] ;

[0123] Where w is a column vector and N is the dimension.

[0124] In this embodiment of the invention, considering the sparse representation characteristics of seismic signals under the generalized wavelet dictionary, the signal decomposition process is transformed into a constrained optimization problem. A LASSO objective function containing L1 norm sparse constraints is established.

[0125] The expression for the optimization problem is:

[0126] ;

[0127] in, To obtain the solution, Let r be the wavelet dictionary, and r be the wavelet coefficients to be determined. For regularization parameters, It is an earthquake path.

[0128] The objective function is solved using a sparse optimization algorithm to obtain the wavelet coefficient vector of the original seismic signal under the wavelet dictionary. Due to the use of sparse optimization, high-precision wavelet coefficients can still be obtained even if the dictionary is non-orthogonal and the signal is aliased.

[0129] In this embodiment of the invention, the time window containing the strong reflection wave in the original seismic signal is determined. Wavelet coefficients corresponding to this time window are extracted from the obtained wavelet coefficient vector. Using these extracted wavelet coefficients and a wavelet dictionary, a generalized inverse wavelet transform is performed to reconstruct the seismic signal containing only the strong reflection background.

[0130] Seismic signals containing strong reflection background The expression is:

[0131] ;

[0132] in, These are the wavelet coefficients in the wavelet coefficient vector that correspond to this time window. This is a wavelet dictionary.

[0133] In this embodiment of the invention, in step S5, the seismic signal containing a strong reflection background is subtracted from the original seismic signal to remove the strong reflection background and output the deshielded weak reflection signal.

[0134] The reconstructed strong reflection wave is directly subtracted from the original seismic signal to remove the strong reflection background, and the output seismic data that finally restores the underlying weak reflection characteristics is used for subsequent reservoir identification.

[0135] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the following detailed explanation of the method and effects of this invention is provided in conjunction with a typical strong shielding theoretical model of coal-bearing strata and accompanying drawings.

[0136] A typical layered geological model designed includes strong reflections from coal seams, such as... Figure 2 As shown, the model, from top to bottom, consists of: overlying mudstone (thickness 100m, velocity 4200m / s, density 2456kg / m³); target coal seam (thickness 30m, velocity 2900m / s, density 2242kg / m³); mudstone interlayer (thickness 20m, velocity 4200m / s, density 2456kg / m³); underlying target sandstone reservoir (thickness 40m, velocity 4500m / s, density 2557kg / m³); and underlying basal mudstone (thickness 50m, velocity 4200m / s, density 2456kg / m³).

[0137] The reflection coefficient sequence calculated based on the above parameters is as follows: Figure 3 As shown, the top and bottom interfaces of the coal seam exhibit a strong reflection coefficient pair with an absolute value of 0.227 due to the huge difference in wave impedance; while the reflection coefficient of the top surface of the underlying target sandstone reservoir is only 0.055.

[0138] The reflection coefficient is convolved with a Ricker wavelet with a dominant frequency of 30Hz to generate a raw seismic record containing strong reflections (e.g., Figure 4 As shown in the figure, the huge coal seam reflection amplitude is extremely prominent, and its strong sidelobe oscillations extend downward, completely covering the weak reflection in-phase axis of the sandstone that should have appeared below, forming a typical geological strong shielding phenomenon.

[0139] As a benchmark for algorithm verification, Figure 5 The theoretical seismic record without shielding is presented after removing the strong reflection coefficient of the coal seam. The figure shows the true weak reflection morphology of the sandstone reservoir without coal seam interference.

[0140] Input signal ( Figure 4 The advantageous main frequency is 30Hz. Therefore, a Ricker wavelet with a clearly defined physical characteristic of 30Hz (such as...) is selected. Figure 6 (As shown) is used as the mother wavelet. The sampling time interval is set to 1ms, and the 30Hz wavelet is shifted point-by-point along the time axis to generate a series of column vectors. These are combined to form the wavelet dictionary matrix with the main frequency focused, as shown. Figure 7 As shown, this matrix is ​​a non-orthogonal Toeplitz matrix with local maxima distributed on the main diagonal. This dictionary ensures full-time domain coverage of all possible reflected waves within the 30Hz dominant frequency band.

[0141] Will Figure 4 The original earthquake records are used as the observation column vector S, and the wavelet dictionary matrix is ​​constructed. This is then incorporated into the optimization objective function with L1 norm sparsity constraints. The regularization constraint parameter is set to λ = 0.001, and the FISTA algorithm is used for global iterative optimization. The iteration termination condition is set to the relative change in the objective function value between two consecutive iterations being less than 10. −4 After the algorithm converges, the wavelet coefficient vector of the signal in the wavelet dictionary is obtained. Figure 8 As can be seen, the originally severely interfering seismic waveforms were transformed into a series of wavelet coefficients representing different energies and intensities in an extremely clean and sparse manner.

[0142] Based on the known coal seam interface, determine the time window corresponding to strong reflection ( Figure 8 (Within the first and second dashed lines). Figure 8In the wavelet coefficient vector shown, the wavelet coefficients within the strong reflection time window are separated. The time window can be extended vertically as appropriate; in this example, it is extended vertically by 5ms to include the wavelet coefficients of strong reflection. Figure 9 a) Obtain the strong reflection wavelet coefficient vector and combine it with the dictionary matrix. Multiplying and performing a generalized wavelet inverse transform, a high-precision reconstruction of the "strong reflection background waveform" containing only the coal seam response (such as...) is obtained. Figure 9 (as shown in b).

[0143] The reconstructed strong reflection waveform ( Figure 9 b) From the original seismic record ( Figure 4 Subtract it directly from the output, i.e., remove the strong reflections from the coal seam. The output result of the de-shielding process is as follows: Figure 10 The red line indicates this signal. It is compared with the theoretical signal without strong reflection in the same coordinate system. The results show that the signal output after removing the strong shielding using the method of this invention not only eliminates the shielding effect of the upper coal seam but also recovers the weak reflection from the deep sand layer, which highly matches the theoretical true signal.

[0144] To quantitatively evaluate the processing effect of this invention, the Pearson correlation coefficient and mean square error between the de-shielding result and the theoretically no-strong-reflection signal within the time window of a weak-reflection target were calculated. The results show that the correlation coefficient is above 0.98 and the mean square error is less than 0.001, further verifying the high-precision separation capability and high-fidelity weak-signal recovery capability of the method of this invention under conditions of severe interleaving of strong and weak signals.

[0145] Those skilled in the art will recognize that the embodiments described herein are intended to help the reader understand the principles of the invention, and should be understood that the scope of protection of the invention is not limited to such specific statements and embodiments. Those skilled in the art can make various other specific modifications and combinations based on the technical teachings disclosed in this invention without departing from the spirit of the invention, and these modifications and combinations are still within the scope of protection of this invention.

Claims

1. A method for removing strongly shielded reflected waves based on generalized wavelet transform, characterized in that, Includes the following steps: S1. Determine the wavelet that matches the physical form of the actual seismic wavelet as the mother wavelet; S2. Construct a wavelet dictionary based on the mother wavelet; S3. Obtain the wavelet coefficient vector based on the wavelet dictionary; S4. Based on the wavelet coefficient vector and wavelet dictionary, reconstruct the seismic signal containing the strong reflection background; S5. Based on the reconstructed seismic signal containing a strong reflection background, obtain the deshielded weak reflection signal; The wavelet dictionary The expression is: ； Where w is a column vector and N is the dimension; In step S3, an optimization problem is constructed and solved based on the wavelet dictionary to obtain the wavelet coefficient vector; The expression for the optimization problem is: ； in, To obtain the solution, Let r be the wavelet dictionary, and r be the wavelet coefficients to be determined. For regularization parameters, For earthquake tunnels; In step S4, a generalized inverse wavelet transform is performed based on the wavelet coefficient vector and the wavelet dictionary to reconstruct the seismic signal containing a strong reflection background.

2. The method for removing strongly shielded reflected waves based on generalized wavelet transform according to claim 1, characterized in that, In S1, the time domain of the mother wavelet The expression is: ； in, For time, Main frequency, denoted as the time center of the wavelet, and e as the exponent.

3. The method for removing strongly shielded reflected waves based on generalized wavelet transform according to claim 1, characterized in that, The optimization problem includes L1 norm sparsity constraints.

4. The method for removing strongly shielded reflected waves based on generalized wavelet transform according to claim 1, characterized in that, The seismic signal containing a strong reflection background The expression is: ； in, These are the wavelet coefficients in the wavelet coefficient vector that correspond to the time window. This is a wavelet dictionary.

5. The method for removing strongly shielded reflected waves based on generalized wavelet transform according to claim 1, characterized in that, In step S5, the seismic signal containing a strong reflection background is subtracted from the original seismic signal to remove the strong reflection background and output the deshielded weak reflection signal.

6. The method for removing strongly shielded reflected waves based on generalized wavelet transform according to claim 5, characterized in that, The weak reflection signal after deshielding is used to complete reservoir identification.

Images