Three-component seismic data random noise suppression method and device based on quaternion wavelet transform, medium and equipment

Abstract

The invention relates to a three-component seismic data random noise suppression method and device based on quaternion wavelet transformation, a medium and equipment. The method comprises the following steps: representing seismic data of X, Y and Z components as quaternion signals in a quaternion array form in a time-space domain; performing quaternion wavelet transform on the quaternion array, and performing multilayer decomposition on the quaternion signal in the form of the quaternion array by using the quaternion wavelet transform to obtain a quaternion wavelet coefficient corresponding to each layer; calculating a threshold value by using a new threshold value estimation expression, performing threshold value operation on quaternion wavelet coefficients of each layer by adopting a semi-soft threshold value function, reserving the wavelet coefficients greater than the threshold value, and setting the wavelet coefficients smaller than the threshold value to be zero; and performing quaternion wavelet inverse transformation on the quaternion wavelet coefficient after threshold processing, reconstructing a denoised quaternion signal, and extracting an X component, a Y component and a Z component from three imaginary part positions of the denoised quaternion signal to complete vector denoising.

Description

Technical Field

[0001] The present invention relates to the technical field of seismology or seismic exploration, and particularly to a method, device, medium and equipment for suppressing random noise in three-component seismic data based on quaternion wavelet transform. Background Art

[0002] In reflection seismic exploration, suppressing noise and improving the signal-to-noise ratio are one of the important tasks in seismic data processing. Seismic data with a high signal-to-noise ratio is crucial for subsequent processing (such as migration imaging, inversion interpretation). As the most common type of noise, random noise is mainly caused by factors such as environmental disturbances, instrument noise, and human operations. Its frequency band is wide and approximately satisfies the Gaussian distribution, and it is widely distributed throughout the seismic record. Based on the significant differences between random noise and effective signals in statistical characteristics, predictability, sparsity, and low-rankness, the effective separation of noise and signals can be achieved through corresponding methods.

[0003] With the shift of oil and gas exploration targets from conventional oil and gas reservoirs to unconventional oil and gas reservoirs such as hidden and tight reservoirs, the exploration difficulty has increased day by day. Traditional single-component longitudinal wave exploration has been difficult to meet the requirements of fine characterization of complex reservoirs. The multi-wave and multi-component seismic exploration technology can effectively reduce the non-uniqueness of reservoir prediction and improve the accuracy of fluid identification and prediction by receiving the full wavefield information, and thus has received extensive attention in the petroleum industry and is applied more and more. Multi-component seismic data acquisition usually uses a longitudinal wave source to generate a wavefield containing P-P wave and P-S wave components, which is received by multi-component geophones. In land exploration, three-component geophones (horizontal X, Y components and vertical Z component) are mainly used, while in marine exploration, four-component geophones (X, Y, Z components and pressure component P) are used. However, the signal-to-noise ratio of multi-component data is generally low, especially for the horizontal components (X, Y). Improving the signal-to-noise ratio of multi-component data is a key link in processing. Since traditional single-component denoising methods cannot effectively utilize the physical coupling relationship between components, it is difficult to meet the dual requirements of signal-to-noise ratio improvement and vector wavefield structure protection at the same time. Therefore, multi-component data needs to be regarded as a whole for joint denoising to fully explore the correlation information between components, protect the polarization characteristics of the vector wavefield while improving the signal-to-noise ratio, and provide a high-quality data basis for subsequent processing and interpretation.

[0004] However, traditional denoising methods cannot solve the problem of vector joint denoising of multi-component seismic data, and most of the existing vector denoising methods are the series combinations of scalar sparse transforms and do not use true vector global transforms. Summary of the Invention

[0005] Aiming at the above problems, the purpose of the present invention is to provide a method, device, medium and equipment for suppressing random noise in three-component seismic data based on quaternion wavelet transform, which can perform joint random noise suppression processing on three-component data.

[0006] To achieve the above object, the present invention adopts the following technical solutions:

[0007] The method for suppressing random noise in three-component seismic data based on quaternion wavelet transform according to the present invention includes the following steps:

[0008] In the time-space domain (t-x domain), represent the seismic data of the X, Y, and Z components as quaternion signals in the form of quaternion arrays;

[0009] Perform quaternion wavelet transform on the quaternion array, use quaternion wavelet transform to perform multi-layer decomposition on the quaternion signal in the form of a quaternion array, and obtain the corresponding quaternion wavelet coefficients of each layer;

[0010] Calculate the threshold using a new threshold estimation expression, perform threshold operation on the quaternion wavelet coefficients of each layer using a semi-soft threshold function, the wavelet coefficients greater than the threshold are retained, and the wavelet coefficients less than the threshold are set to zero;

[0011] Perform inverse quaternion wavelet transform on the threshold-processed quaternion wavelet coefficients, reconstruct the denoised quaternion signal, and extract the X component, Y component, and Z component from the three imaginary part positions of the denoised quaternion signal to complete vector denoising.

[0012] In the method for suppressing random noise in three-component seismic data, preferably, the step of representing the seismic data of the X, Y, and Z components as quaternion signals in the form of quaternion arrays in the time-space domain (t-x domain) is as follows:

[0013] Select a set of 3C-2D seismic data D X (t,x), D Y (t,x) and D Z (t,x);

[0014] Using quaternion tools, combine the 3C-2D noisy data D X (t,x), D Y (t,x) and D Z (t,x) into a quaternion array: D Q (t,x) = D X (t,x)i + D Y (t,x)j + D Z (t,x)k;

[0015] D Q (t,x) can be expressed as D Q (t,x) = S Q (t,x) + N Q (t,x);

[0016] Among them, S Q (t, x) = S X (t, x)i + S Y (t, x)j + S Z (t, x)k represents a quaternion effective signal; N Q (t, x) = N X (t, x)i + N Y (t, x)j + N Z (t, x)k represents quaternion noise.

[0017] For the above-mentioned three-component seismic data random noise suppression method, preferably, perform quaternion wavelet transform on the quaternion array D Q (t, x). The specific steps are as follows:

[0018] In the quaternion wavelet domain, there are differences in the distribution characteristics of the wavelet energies of the effective signal and the noise. Utilize this difference and adopt the least squares inversion problem method to recover the effective signal S Q from the noisy signal D Q . This least squares inversion problem can be expressed as:

[0019]

[0020] Among them, D Q is the noisy data; is the quaternion discrete orthogonal wavelet inverse transform; W Q is the quaternion wavelet coefficient; λ > 0 is the regularization factor; represents the wavelet coefficient of the denoising result to be solved; p is the p-norm. When p = 0, the optimal solution can be obtained by performing a hard threshold operation on the wavelet coefficient W Q of the noisy data D Q . At this time, the threshold is denoted as λ; when p = 1, the optimal solution is obtained by performing a soft threshold operation on the wavelet coefficient W Q of D Q . The threshold is denoted as λ 2 / 2.

[0021] For the above-mentioned three-component seismic data random noise suppression method, preferably, the specific steps of the new threshold estimation are as follows:

[0022] When the random noise in the three-component noisy data satisfies the Gaussian distribution and the uncorrelated hypothesis, the probability density function f(r) and the probability distribution function F(r) corresponding to the amplitude value of the noise can be simplified into expressions with only the amplitude r as the variable, as shown in formulas (2) and (3) respectively.

[0023]

[0024]

[0025] Among them, σ is the noise standard deviation; u is the integration variable; r is the noise amplitude value;

[0026] When the probability value of the noise is 50%, the ratio of the noise amplitude value r to the noise standard deviation σ is 1.5382. Therefore, the formula (4) is used to estimate the noise standard deviation σ:

[0027]

[0028] Among them, median() represents the median function; |D1| represents the modulus of the first-layer wavelet decomposition coefficient sequence of the noisy data D Q (t,x);

[0029] The empirical threshold calculation formula (5) is used to calculate the new threshold λ expression:

[0030]

[0031] Among them, N represents the signal length; α > 0 represents the adjustment factor; σ is the noise standard deviation.

[0032] For the three-component seismic data random noise suppression method described above, preferably, the semi-soft threshold function has the expression shown in formula (6):

[0033]

[0034] Among them, m > 0 is the adjustment factor; W is the wavelet transform coefficient; λ is the new threshold; when m → 0, the threshold function approaches the hard threshold function, and when m → ∞, the threshold function approaches the soft threshold function.

[0035] For the three-component seismic data random noise suppression method described above, preferably, the denoising result is:

[0036] The present invention also provides a three-component seismic data random noise suppression device based on quaternion wavelet transform, including:

[0037] The first processing unit is used to represent the seismic data of the X, Y, and Z three components as a quaternion signal in the form of a quaternion array in the time-space domain;

[0038] The second processing unit is used to perform quaternion wavelet transform on the quaternion array, and perform multi-layer decomposition on the quaternion signal in the form of a quaternion array using quaternion wavelet transform to obtain the corresponding quaternion wavelet coefficients of each layer;

[0039] A third processing unit, configured to calculate a threshold using a new threshold estimation expression, perform a threshold operation on the quaternion wavelet coefficients of each layer by adopting a semi-soft threshold function, keep the wavelet coefficients greater than the threshold, and set to zero the wavelet coefficients less than the threshold;

[0040] A fourth processing unit, configured to perform an inverse quaternion wavelet transform on the threshold-processed quaternion wavelet coefficients, reconstruct the denoised quaternion signal, and take out the X component, Y component, and Z component from the positions of the three imaginary parts of the denoised quaternion signal to complete vector denoising.

[0041] The present invention also provides a computer storage medium, on which a computer program is stored, and when the computer program is executed by a processor, the steps of the three-component seismic data random noise suppression method based on quaternion wavelet transform are implemented.

[0042] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor, and when the processor executes the computer program, the steps of the three-component seismic data random noise suppression method based on quaternion wavelet transform are implemented.

[0043] Due to the adoption of the above technical solutions, the present invention has the following advantages:

[0044] (1) The present invention uses quaternions to jointly represent three-component seismic data in the time domain, and proposes a three-component seismic data vector denoising technique based on quaternion wavelet transform. When denoising, the inherent orthogonality of quaternions enables it to have the advantages of protecting the non-linear mutual relationship between the data of each component and maintaining the vector structure characteristics of the seismic wave field;

[0045] (2) The present invention uses a tower algorithm to implement multi-scale decomposition and reconstruction of quaternion wavelet transform. Compared with the traditional scalar wavelet transform, this vector overall transform can make full use of the correlation between the components of the seismic signal to enhance the wavelet coefficient energy of the effective signal in the wavelet domain, making it easier to separate the signal and noise when applying the threshold operation, and thus suppressing random noise more thoroughly;

[0046] (3) The threshold estimation expression used in the traditional wavelet threshold denoising method is developed for single-component data and is not applicable to the joint denoising situation of three-component data. The present invention gives a threshold estimation expression applicable to the joint denoising situation of three-component data under the assumption that the three-component noise satisfies an independent Gaussian distribution, and is applicable to the joint denoising situation of three-component data. Description of the Drawings

[0047] By reading the following detailed description of the preferred embodiments, various other advantages and benefits will become clear to those of ordinary skill in the art. The drawings are only for the purpose of showing the preferred embodiments and are not considered to be a limitation of the present invention. Throughout the drawings, the same reference numerals are used to represent the same components. In the drawings:

[0048] Figure 1 are the profiles of the X, Y, and Z three-component original noise-free data and the data with random noise synthesized in the first embodiment of the present invention, where Figure 1 (a)-1(c) represent the X, Y, and Z components of the original noise-free data, Figure 1 (d)-1(f) represent the X, Y, and Z component data with random noise;

[0049] Figure 2 are the denoising results using scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform of the embodiment of the present invention, where Figure 2 (a)-2(c) are the denoising results of the X component using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method, respectively; Figure 2 (d)-2(f) are the denoising results of the Y component using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method, respectively; Figure 2 (g)-2(i) are the denoising results of the Z component using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method, respectively;

[0050] Figure 3 are the random noises removed using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method of the embodiment of the present invention, where Figure 3 (a)-3(c) are the random noises removed from the X component using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method, respectively; Figure 3 (d)-3(f) are the random noises removed from the Y component using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method, respectively; Figure 3 (g)-3(i) are the random noises removed from the Z component using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method, respectively;

[0051] Figure 4 are the comparison between the single-channel signal after denoising using the scalar wavelet transform, cascaded combined wavelet transform, and the quaternion wavelet transform threshold method of the embodiment of the present invention and the original noise-free single-channel signal, where Figure 4 (a) is the comparison between the denoising result of the X component using the scalar wavelet transform threshold method and the original noise-free single-channel signal; Figure 4 (b) is the comparison between the denoising result of the Y component using the cascaded combined wavelet transform threshold method and the original noise-free single-channel signal;Figure 4 (c) is the comparison between the denoising result of the Z - component quaternion wavelet transform threshold method and the original noise - free single - channel signal;

[0052] Figure 5 is the comparison of the vector - end diagrams of the denoising results of the X, Y, and Z components and the original noise - free signal; among them, Figure 5 (a) is the comparison of the vector - end diagram of the denoising result using the scalar wavelet transform and the original noise - free signal; Figure 5 (b) is the comparison of the vector - end diagram of the denoising result using the cascaded wavelet transform and the original noise - free signal; Figure 5 (c) is the comparison of the vector - end diagram of the denoising result using the quaternion wavelet transform of the embodiment of the present invention and the original noise - free signal;

[0053] Figure 6 is the original data of the actual OBS data in a certain area, among which, Figure 6 (a), Figure 6 (b), and Figure 6 (c) respectively represent the X, Y, and Z components of the actual OBS data.

[0054] Figure 7 is the denoising result of the actual OBS data using the scalar wavelet transform, the cascaded combined wavelet transform, and the quaternion wavelet transform of the embodiment of the present invention. Among them, Figure 7 (a), Figure 7 (b), and Figure 7 (c) respectively represent the denoising results of the X - component using the scalar wavelet transform, the cascaded combined wavelet transform, and the quaternion wavelet transform threshold method of the embodiment of the present invention; Figure 7 (d), Figure 7 (e), and Figure 7 (f) respectively represent the denoising results of the Y - component using the scalar wavelet transform, the cascaded combined wavelet transform, and the quaternion wavelet transform threshold method of the embodiment of the present invention; Figure 7 (g), Figure 7 (h), and Figure 7 (i) respectively represent the denoising results of the Z - component using the scalar wavelet transform, the cascaded combined wavelet transform, and the quaternion wavelet transform threshold method of the embodiment of the present invention. Detailed implementation manners

[0055] The exemplary embodiments of the present invention will be described in more detail with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided so that the present invention can be more thoroughly understood and the scope of the present invention can be fully conveyed to those skilled in the art.

[0056] The present invention provides a method for suppressing random noise in three-component seismic data based on quaternion wavelet transform. First, the multi-scale decomposition and reconstruction of quaternion wavelet transform are realized by using the tower algorithm. Secondly, a threshold estimation method in the three-component case is derived, and a vector threshold denoising method is proposed based on this. Compared with the traditional wavelet scalar threshold denoising method, the present invention can effectively improve the signal-to-noise ratio of the denoising result and better protect the weak signals and the vector polarization characteristics of multi-component data.

[0057] The method for suppressing random noise in three-component seismic data based on quaternion wavelet transform provided by the present invention includes the following steps:

[0058] S1. Represent the seismic data of X, Y, and Z components in the time-space domain (t-x domain) as a quaternion signal in the form of a quaternion array;

[0059] S2. Perform quaternion wavelet transform on the quaternion array, and use quaternion wavelet transform to perform multi-layer decomposition on the quaternion signal in the form of a quaternion array to obtain the corresponding quaternion wavelet coefficients of each layer;

[0060] S3. Calculate the threshold using a new threshold estimation expression, and perform threshold operation on the quaternion wavelet coefficients of each layer using a semi-soft threshold function. The wavelet coefficients greater than the threshold are retained, and the wavelet coefficients less than the threshold are set to zero;

[0061] S4. Perform inverse quaternion wavelet transform on the threshold-processed quaternion wavelet coefficients to reconstruct the denoised quaternion signal, and extract the X component, Y component, and Z component from the three imaginary part positions of the denoised quaternion signal to complete vector denoising.

[0062] In the above embodiment, preferably, the step of representing the seismic data of X, Y, and Z components in the time-space domain (t-x domain) as a quaternion signal in the form of a quaternion array is as follows:

[0063] Select a set of 3C-2D seismic data D X (t,x), D Y (t,x) and D Z (t,x);

[0064] Use quaternion tools to combine the 3C-2D noisy data D X (t,x), D Y (t,x) and D Z (t,x) into a quaternion array: D Q (t,x) = D X (t,x)i + D Y (t,x)j + D Z (t,x)k;

[0065] D Q (t, x) can be expressed as D Q (t, x) = S Q (t, x) + N Q (t, x);

[0066] wherein, S Q (t, x) = S X (t, x)i + S Y (t, x)j + S Z (t, x)k represents a quaternion valid signal; N Q (t, x) = N X (t, x)i + N Y (t, x)j + N Z (t, x)k represents quaternion noise.

[0067] In the above embodiment, preferably, a quaternion wavelet transform is performed on the quaternion array D Q (t, x), and the specific steps are as follows:

[0068] In the quaternion wavelet domain, there are differences in the distribution characteristics of the wavelet energy between the valid signal and the noise. Using this difference, the least squares inversion problem method is used to recover the valid signal S Q from the noisy signal D Q , and this least squares inversion problem can be expressed as:

[0069]

[0070] wherein, D Q is the noisy data; quaternion discrete orthogonal wavelet inverse transform; W Q is the quaternion wavelet coefficient; λ > 0 is the regularization factor; represents the wavelet coefficient of the denoising result to be solved; p is the p-norm. When p = 0, the optimal solution can be obtained by performing a hard threshold operation on the wavelet coefficient W Q of the noisy data D Q , and the threshold is denoted as λ at this time; when p = 1, the optimal solution is obtained by performing a soft threshold operation on the wavelet coefficient W Q of D Q , and the threshold is denoted as λ 2 / 2.

[0071] In the above embodiment, preferably, the specific steps of the new threshold estimation are as follows:

[0072] When the random noise in the three-component noisy data satisfies the Gaussian distribution and the uncorrelated assumption, the probability density function f(r) and the probability distribution function F(r) corresponding to the amplitude value of the noise can be simplified to expressions with only the amplitude r as the variable, as shown in formulas (2) and (3) respectively.

[0073]

[0074] Among them, σ is the noise standard deviation; u is the integration variable; r is the noise amplitude value.

[0075] When the probability value of the noise is 50%, the ratio of the noise amplitude value r to the noise standard deviation σ is 1.5382. Therefore, formula (4) is used to estimate the noise standard deviation σ:

[0076]

[0077] Among them, median() represents the median function; |D1| represents the modulus of the first-layer wavelet decomposition coefficient sequence of the noisy data D Q (t,x).

[0078] The empirical threshold calculation formula (5) is used to calculate the new threshold λ expression:

[0079]

[0080] Among them, N represents the signal length, α>0 represents the adjustment factor; σ is the noise standard deviation.

[0081] In the above embodiment, preferably, the semi-soft threshold function has the expression shown in formula (6):

[0082]

[0083] Among them, m>0 is the adjustment factor; W is the wavelet transform coefficient; λ is the new threshold; when m→0, the threshold function approaches the hard threshold function, and when m→∞, the threshold function approaches the soft threshold function.

[0084] In the above embodiment, preferably, the quaternion wavelet inverse transform is performed on the quaternion wavelet coefficients after the threshold operation to obtain the denoising result: Embodiment 1:

[0085] The present invention is applied to the denoising of synthetic 3C-2D model data, and the effects are compared with the scalar wavelet transform method and the tandem combined wavelet transform threshold denoising method. As Figure 1 shown, the original three-component noise-free data is as Figure 1(As shown in (a)-1(c), Gaussian random noise is added to each of the three component data, such as Figure 1 (d)-1(f), where the signal-to-noise ratio of the horizontal X and Y component data is 1.0 dB, and the signal-to-noise ratio of the vertical Z component is 2.0 dB.

[0086] The scalar wavelet transform method, the series combination wavelet transform method, and the quaternion wavelet transform method are respectively used to denoise the noisy data. The denoising results are as shown in Figure 2 shown, where Figure 2 (a)-2(c) is the denoising result of the X component; Figure 2 (d)-2(f) is the denoising result of the Y component; Figure 2 (g)-2(i) is the denoising result of the Z component. By comparison, it can be seen that the smoothness and continuity of the in-phase axis of the effective wave in the denoising result of the scalar wavelet method are poor, and there is obvious noise residue near the edge of the in-phase axis, as indicated by the black arrow; although the denoising effect of the series combination wavelet transform method has improved, the continuity and smoothness of the in-phase axis of the effective wave are still insufficient; while the denoising effect of the quaternion wavelet transform method is the most ideal, the in-phase axis of the effective wave is naturally continuous and has good smoothness. In addition, in the area shown by the black rectangular box, the weak reflection axis in the denoising result of the scalar wavelet transform method is almost suppressed as noise, while the weak reflection axis is also well retained in the series combination wavelet transform method and the quaternion wavelet transform method.

[0087] Figure 3 The random noise filtered by the three methods is shown. By comparison, it can be seen that there is obvious residual energy of the effective reflection wave in the noise filtered by the scalar wavelet transform method, which causes great damage to the energy of the in-phase axis, as shown by the black rectangular box and the black arrow; the residual energy of the effective wave in the noise filtered by the series combination wavelet transform method is weak; the residual energy of the effective wave in the noise filtered by the quaternion wavelet transform method is the weakest, and the energy damage of the in-phase axis of the reflection wave is the smallest. The signal-to-noise ratio is used to quantitatively analyze the denoising performance of the three methods. The calculation formula of the signal-to-noise ratio is as follows,

[0088]

[0089] where D cl ean represents the original noise-free data; D d enoised represents the denoising result.

[0090] The SNR values of the denoising results of the three methods are shown in Table 1. The average signal-to-noise ratio of the three components of the denoising result of the quaternion wavelet transform is 1.65 dB higher than that of the scalar wavelet transform.

[0091] In addition, a single trace data is extracted for further comparison of the denoising effect. Figure 4(a)-4(c) respectively show the single-trace comparison of the denoising results of scalar wavelet transform, cascaded combined wavelet transform method, and quaternion wavelet transform with the original noise-free data X, Y, and Z components. As shown by the black rectangular frames in the figure, the weak effective signals corresponding to the X component and Y component are suppressed as noise by the scalar wavelet transform method; there is still some residual noise energy in the denoising results of the X and Y components obtained by the cascaded combined wavelet transform; while the denoising result of the quaternion wavelet transform better protects the weak signals and matches well with the original noise-free real data. Select Figure 4 the waveform within the time window of [730, 1110] ms in the green dashed box in Figure 5 as shown. It can be seen that the vibration trajectory of the particles in the denoising result of the scalar wavelet transform deviates greatly from that of the particles in the original noise-free signal, followed by the cascaded combined wavelet transform method, while the denoising result of the quaternion wavelet transform method matches best with the hodograph of the original noise-free signal. This experiment demonstrates the superiority of the quaternion wavelet transform threshold denoising method in protecting the weak signal energy and maintaining the vector polarization structure characteristics of seismic signals.

[0092] Table 1. Comparison of signal-to-noise ratios of the denoising results of three methods

[0093]

[0094] Example 2:

[0095] Apply the present invention to the random noise suppression of 3C-2D OBS data in a certain work area. This data set is an OBS common receiver gather, including X, Y, and Z components. Each component data contains 283 traces, the trace interval is 20 m, the time sampling length of each trace is 8 s, and the sampling rate is 2 ms. Select the seismic data within the time window of [1.6, 1.9] s and trace numbers 235-283 for denoising tests. Figure 6 (a)-6(c) show the X, Y, and Z components of the original data. It can be seen from the figure that there is a large amount of random noise in this data. Figure 7 Show the denoising results of the scalar wavelet transform, cascaded combined wavelet transform, and quaternion wavelet transform threshold methods, where Figure 7 (a)-7(c) are the X components after denoising by the three methods respectively, Figure 7 (d)-7(f) are the Y components after denoising by the three methods respectively, Figure 7 (g)-7(i) are the Z components after denoising by the three methods respectively. By comparing the areas indicated by the black boxes, it can be seen that the quaternion wavelet transform threshold method has a stronger ability to suppress random noise. The reflection wave event structure in the denoising result is clear and continuous, there are fewer residual random noise energy points, and the section is cleaner.

[0096] The present invention also provides a three-component seismic data random noise suppression device based on quaternion wavelet transform, which includes:

[0097] A first processing unit, configured to represent the seismic data of X, Y, and Z components as a quaternion signal in the form of a quaternion array in the time-space domain;

[0098] A second processing unit, configured to perform quaternion wavelet transform on the quaternion array, and perform multi-layer decomposition on the quaternion signal in the form of a quaternion array using quaternion wavelet transform to obtain the corresponding quaternion wavelet coefficients of each layer;

[0099] A third processing unit, configured to calculate a threshold using a new threshold estimation expression, and perform a threshold operation on the quaternion wavelet coefficients of each layer using a semi-soft threshold function. The wavelet coefficients greater than the threshold are retained, and the wavelet coefficients less than the threshold are set to zero;

[0100] A fourth processing unit, configured to perform inverse quaternion wavelet transform on the threshold-processed quaternion wavelet coefficients to reconstruct the denoised quaternion signal, and extract the X component, Y component, and Z component from the three imaginary part positions of the denoised quaternion signal to complete vector denoising.

[0101] The present invention also provides a computer storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the three-component seismic data random noise suppression method based on quaternion wavelet transform are implemented.

[0102] The present invention also provides a computer device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the three-component seismic data random noise suppression method based on quaternion wavelet transform are implemented.

[0103] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit them. Although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for suppressing random noise in three-component seismic data based on quaternion wavelet transform, characterized in that It includes the following steps: Represent the seismic data of X, Y, and Z three components in the time - space domain as a quaternion signal in the form of a quaternion array; Perform a quaternion wavelet transform on the quaternion array, and use the quaternion wavelet transform to perform multi - layer decomposition on the quaternion signal in the form of a quaternion array to obtain the corresponding quaternion wavelet coefficients of each layer; Calculate the threshold using a new threshold estimation expression, and perform a threshold operation on the quaternion wavelet coefficients of each layer using a semi - soft threshold function. The wavelet coefficients greater than the threshold are retained, and the wavelet coefficients less than the threshold are set to zero; Perform an inverse quaternion wavelet transform on the threshold - processed quaternion wavelet coefficients to reconstruct the denoised quaternion signal, and extract the X - component, Y - component, and Z - component from the three imaginary part positions of the denoised quaternion signal to complete vector denoising.

2. The three-component seismic data random noise suppression method according to claim 1, wherein The specific steps of representing the seismic data of X, Y, and Z three components in the time - space domain as a quaternion signal in the form of a quaternion array are as follows: Select a set of 3C-2D seismic data D containing random noise X (t, x), D Y (t, x) and D Z (t, x); Using quaternion tools, the 3C-2D noisy data D X (t, x), D Y (t, x) and D Z (t, x) are combined into a quaternion array: D Q (t, x) = D X (t, x)i + D Y (t, x)j + D Z (t, x)k; D Q (t, x) is represented as D Q (t, x) = S Q (t, x) + N Q (t, x); where, S Q (t, x) = S X (t, x)i + S Y (t, x)j + S Z (t, x)k represents a quaternion valid signal; N Q (t, x) = N X (t, x)i + N Y (t, x)j + N Z (t, x)k represents quaternion noise.

3. The method for suppressing random noise in three-component seismic data according to claim 2, wherein Perform quaternion wavelet transform on the quaternion array D Q (t, x), and the specific steps are as follows: In the quaternion wavelet domain, there are differences in the distribution characteristics of the wavelet energy between the effective signal and the noise. Using this difference, the least squares inversion problem method is adopted to recover the effective signal S Q from the noisy signal D Q , and this least squares inversion problem can be expressed as: Among them, D Q is the noisy data; is the inverse discrete quaternion orthogonal wavelet transform; W Q is the quaternion wavelet coefficient; λ > 0 is the regularization factor; represents the wavelet coefficient of the denoising result to be obtained; p is the p-norm. When p = 0, the optimal solution can be obtained by performing a hard thresholding operation on the wavelet coefficient W Q of the noisy data D Q . At this time, the threshold is denoted as λ; when p = 1, the optimal solution is obtained by performing a soft thresholding operation on the wavelet coefficient W Q of D Q , and the threshold is denoted as λ 2 / 2.

4. The method for suppressing random noise in three-component seismic data according to claim 3, characterized in that The specific steps of the new threshold estimation are as follows: When the random noise in the three - component noisy data satisfies the Gaussian distribution and the uncorrelated assumption, the probability density function f(r) and the probability distribution function F(r) corresponding to the amplitude value of the noise are simplified to expressions that only take the amplitude r as a variable, as shown in formulas (2) and (3) respectively, where σ is the noise standard deviation; u is the integration variable; r is the noise amplitude value; When the probability value of the noise is 50%, the ratio of the noise amplitude value r to the noise standard deviation σ is 1.5382. Therefore, formula (4) is used to estimate the noise standard deviation σ: Among them, median() represents the median function; |D1| represents the noisy data D Q The modulus of the first-layer wavelet decomposition coefficient sequence of (t, x); Use the empirical threshold calculation formula (5) to calculate the new threshold λ expression: where N represents the signal length; α>0 represents the adjustment factor; σ is the noise standard deviation.

5. The three-component seismic data random noise suppression method according to claim 4, wherein The semi-soft threshold function is expressed as shown in formula (6): where m > 0 is a regulation factor; W is the wavelet transform coefficient; λ is the new threshold; when m → 0, the threshold function approaches the hard threshold function, and when m → ∞, the threshold function approaches the soft threshold function.

6. The three-component seismic data random noise suppression method according to claim 5, wherein The denoising result is:

7. A three-component seismic data random noise suppression device based on quaternion wavelet transform, characterized in that, It includes: The first processing unit is used to represent the seismic data of X, Y, and Z three components in the time - space domain as a quaternion signal in the form of a quaternion array; The second processing unit is used to perform a quaternion wavelet transform on the quaternion array, and use the quaternion wavelet transform to perform multi - layer decomposition on the quaternion signal in the form of a quaternion array to obtain the corresponding quaternion wavelet coefficients of each layer; The third processing unit is used to calculate the threshold using a new threshold estimation expression, and perform a threshold operation on the quaternion wavelet coefficients of each layer using a semi - soft threshold function. The wavelet coefficients greater than the threshold are retained, and the wavelet coefficients less than the threshold are set to zero; The fourth processing unit is used to perform an inverse quaternion wavelet transform on the threshold - processed quaternion wavelet coefficients to reconstruct the denoised quaternion signal, and extract the X - component, Y - component, and Z - component from the three imaginary part positions of the denoised quaternion signal to complete vector denoising.

8. A computer storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the three - component seismic data random noise suppression method based on quaternion wavelet transform described in any one of claims 1 - 6.

9. A computer device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the three - component seismic data random noise suppression method based on quaternion wavelet transform described in any one of claims 1 - 6.

Images