A method, apparatus, and medium for noise suppression

Abstract

The application discloses a noise suppression method, device and medium, and relates to the field of digital signal processing. The method comprises the following steps: obtaining original observation data obtained by a detector observing a seismic signal; introducing a damping operator to perform a rank reduction operation on the original observation data and obtaining a matrix after the rank reduction; obtaining low-rank filtering data according to the matrix after the rank reduction; and filtering a signal of a preset frequency in the low-rank filtering data by using a filter to obtain first denoising data. In the method, the damping operator is introduced in the process of performing the rank reduction operation on the original data. Since the damping operator can weaken the artificial interference introduced by the low-rank projection of strong disturbance noise and random noise, the suppression of incoherent noise in the seismic data is realized. In addition, the application further provides a noise suppression device and a computer readable storage medium, which have the same or corresponding technical features as the above noise suppression method, and the effects are the same.

Description

Technical Field

[0001] This application relates to the field of digital signal processing, and in particular to a method, apparatus and medium for noise suppression. Background Technology

[0002] During the recording of seismic data, random noise (which is a type of incoherent interference wave) may be mixed in. Some of the noise with a high intensity can cause great interference to the recording of seismic data, resulting in inaccurate seismic data. Therefore, it is necessary to suppress the incoherent noise in the seismic data.

[0003] Currently, Singular Spectrum Analysis (SSA) is commonly used to suppress incoherent noise in seismic data. SSA utilizes the low-rank characteristics of seismic signals, treating noise suppression as a low-rank reconstruction problem. However, in the process of using SSA, significant artificial interference is introduced due to the low-rank projection in order to fit high-amplitude instability interference in the least-squares sense. Therefore, SSA cannot effectively suppress incoherent noise in seismic data.

[0004] Therefore, how to suppress incoherent noise in seismic data is a technical problem that urgently needs to be solved by those in the field. Summary of the Invention

[0005] The purpose of this application is to provide a method, apparatus, and medium for noise suppression, used to suppress incoherent noise in seismic data.

[0006] To solve the above-mentioned technical problems, this application provides a method for noise suppression, comprising:

[0007] Acquire the raw observation data obtained by the detector observing the seismic signal;

[0008] A damping operator is introduced to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix.

[0009] Obtain low-rank filtered data based on the reduced-rank matrix;

[0010] A filter is used to filter the signal of a preset frequency in the low-rank filtered data in order to obtain the first denoised data.

[0011] Preferably, after filtering the signal of a preset frequency in the low-rank filtered data using a filter to obtain the first denoised data, the method further includes:

[0012] The weighting matrix is ​​obtained based on the low-rank filtered data and the first denoised data;

[0013] The filter is used to filter the low-rank filtered data and the signal of the preset frequency in the weighting matrix in order to obtain the second denoised data.

[0014] Preferably, the damping operator is determined based on the damping factor, and obtaining the damping factor includes:

[0015] Obtain the upper limit, lower limit, maximum number of iterations, and current number of iterations of the preset damping factor;

[0016] The damping factor is obtained based on the upper limit value, the lower limit value, the maximum number of iterations, and the current number of iterations.

[0017] Preferably, the step of introducing a damping operator to perform a rank reduction operation on the original observation data and obtaining the rank-reduced matrix includes:

[0018] The original observation data is transformed in the frequency domain to obtain the transformed original observation data;

[0019] The transformed original observation data are rearranged to obtain the Hankelized matrix;

[0020] The rank of the Hankelized matrix is ​​reduced based on the damping factor and the truncated rank, and the reduced-rank matrix is ​​obtained.

[0021] Preferably, obtaining low-rank filtered data from the reduced-rank matrix includes:

[0022] The reduced-rank matrix is ​​averaged using an anti-diagonal method to obtain the low-rank filtered data.

[0023] Preferably, after filtering the low-rank filtered data and the signal of the preset frequency in the weighting matrix using the filter to obtain the second denoised data, the method further includes:

[0024] Determine whether the current iteration number is greater than the maximum iteration number;

[0025] If so, then obtain the second denoised data;

[0026] If not, return to the step of introducing a damping operator to perform a rank reduction operation on the original observation data and obtaining the rank-reduced matrix.

[0027] Preferably, during the iteration process, the method further includes:

[0028] Determine whether the current iteration number is the first iteration;

[0029] If so, then a quadratic fitting function is used to iteratively approximate the original observation data;

[0030] If not, then a non-quadratic fitting function is used to iteratively approximate the original observation data.

[0031] To address the aforementioned technical problems, this application also provides a noise suppression device, comprising:

[0032] The first acquisition module is used to acquire the raw observation data obtained by the detector from observing the seismic signal;

[0033] The rank reduction and acquisition module is used to introduce a damping operator to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix.

[0034] The second acquisition module is used to acquire low-rank filtered data based on the reduced-rank matrix.

[0035] The filtering module is used to filter the signal of a preset frequency in the low-rank filtered data using a filter in order to obtain the first denoised data.

[0036] To address the aforementioned technical problems, this application also provides a noise suppression device, comprising:

[0037] Memory, used to store computer programs;

[0038] A processor, used to implement the above-described noise suppression method when executing the computer program.

[0039] To address the aforementioned technical problems, this application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, implements the steps of the noise suppression method described above.

[0040] This application provides a noise suppression method, comprising: acquiring raw observation data obtained by a detector observing seismic signals; introducing a damping operator to perform a rank reduction operation on the raw observation data and obtaining a rank-reduced matrix; obtaining low-rank filtered data based on the rank-reduced matrix; and using a filter to filter signals of a preset frequency in the low-rank filtered data to obtain first denoised data. In this method, a damping operator is introduced during the rank reduction operation on the raw data. Since the damping operator can weaken the artificial interference introduced by the low-rank projection of strong disturbance noise and random noise, it achieves the suppression of incoherent noise in the seismic data.

[0041] In addition, this application also provides a noise suppression apparatus and a computer-readable storage medium, which have the same or corresponding technical features as the noise suppression method mentioned above, and have the same effect. Attached Figure Description

[0042] To more clearly illustrate the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0043] Figure 1 A flowchart of a noise suppression method provided in an embodiment of this application;

[0044] Figure 2 A structural diagram of a noise suppression apparatus provided in an embodiment of this application;

[0045] Figure 3 This is a structural diagram of a noise suppression device provided in another embodiment of this application. Detailed Implementation

[0046] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the protection scope of this application.

[0047] The core of this application is to provide a method, apparatus, and medium for noise suppression, used to suppress incoherent noise in seismic data.

[0048] Adding and removing seismic noise can be represented as a pair of forward and inverse models. Currently, the Seismic Subtraction Annoying Sensor (SSA) method is used for seismic data denoising. The process of using SSA for seismic denoising is as follows:

[0049] D = S + N (1.1)

[0050]

[0051] Where D is the observed data, S is the noise-free signal, and N is the incoherent noise. Let ||·||p represent the regularization operator, ||·||p represent a certain norm, and S' be a low-rank matrix. Generally, the first term in equation (1.2) is called the fidelity term, ensuring that the estimated signal S is consistent with the observed value. The second term is the penalty term, which imposes some prior constraints on the estimated signal S. Different combinations of fidelity and penalty terms lead to different denoising methods. In most cases, it is assumed that the incoherent noise N follows a Gaussian distribution, and the norm ||·||p is chosen as the Frobenius norm.

[0052] The inversion problem in Equation (1.2) can be solved using the SSA method. In this case, Equation (1.2) is typically rewritten as an optimization problem with equality constraints:

[0053]

[0054] In formula (1.3), This represents the Frobenius norm. k is an integer, typically chosen as the number of tilt components. The SSA provides a filter D for the ω-frequency component. ω To solve formula (1.3), which involves three main steps.

[0055]

[0056] In formula (1.4), S' w This represents the data after filtering with a filter of frequency ω and then performing singular spectrum analysis for noise reduction. This is a rearrangement operator used to rearrange data into a Hankel matrix structure, as shown below:

[0057]

[0058] In formula (1.5), D(1), D(2), ..., D(N) are D ω The elements of the matrix. x = N - y + 1 is a predefined integer, and when this integer is chosen, the Hankel matrix... Approximately squared. It is a rank-reducing operator that uses the truncated singular value decomposition (SVD) method to compute H. ω The low-rank approximation matrix.

[0059]

[0060] In formula (1.6), H represents ω The low-rank approximation. and Representation matrix H ω The first k largest singular values ​​and the associated k singular vectors. () H This represents the Hermite transpose of a matrix. It is an averaging operator that applies to matrices The average of the anti-diagonal lines is used to recover the filtered data, as shown in formula (1.7) below:

[0061]

[0062] In formula (1.7), This data is filtered using the SSA method. The central idea of ​​SSA is the predictability of seismic plane waves. Considering a plane wave uniformly sampled along the spatial direction, it can be expressed in the frequency domain as shown in the following formula (1.8):

[0063] d n (ω)=w(ω)z n-1 (ω) (1.8)

[0064] In formula (1.8), d n (w) represents the signal in the frequency domain, and w(ω) is the Fourier transform of the plane wavelet. n-1 (ω)z n-1 (ω) is the phase shift at frequency ω. For two adjacent recording tracks, the following recursive expression can be used:

[0065] d n (ω)=P(ω)d n-1 (ω). (1.9)

[0066] In formula (1.9), P(ω) represents the recursion coefficient.

[0067] Due to the plane wave assumption, the SSA method is highly effective for reconstructing linear events. For curved events, this method is typically implemented within a local spatial window to approximate the ideal signal model. The SSA method and its various similar algorithms handle unstable disturbances by iteratively estimating the absolute difference between the original and filtered data. However, the SSA method may fail for strongly unstable disturbances because the low-rank projection in the SSA method forces the introduction of significant artificial disturbances to fit high-amplitude unstable disturbances in the least-squares sense.

[0068] This application extends the SSA method based on the theory of SSA (Seismic Subtraction Anomaly) for suppressing seismic noise, thereby enhancing the stability of low-rank reconstruction of seismic data. A damping operator is used to mitigate artificial interference introduced by the low-rank projection of strong disturbance noise and random noise. By adjusting the damping factor, the projection of irregular strong disturbance noise and random noise can be reduced to an acceptable level, avoiding errors accumulated during iteration. When seismic data is disturbed by irregular strong disturbance noise and random noise, the method of this application can achieve stable suppression of these disturbances and random noise.

[0069] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to the accompanying drawings and specific embodiments. Figure 1 A flowchart of a noise suppression method provided in this application embodiment is shown below. Figure 1 As shown, the method includes:

[0070] S10: Acquire the raw observation data obtained by the detector from observing the seismic signal;

[0071] S11: Introduce a damping operator to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix;

[0072] S12: Obtain low-rank filtered data based on the reduced-rank matrix;

[0073] S13: Use a filter to filter the signal of a preset frequency in the low-rank filtered data in order to obtain the first denoised data.

[0074] To acquire seismic data, this embodiment uses a geophone to obtain seismic signals and raw observation data. Raw observation data typically contains other interference signals in addition to the seismic signal. It should be noted that after acquiring the raw observation data, a frequency domain transformation is required. In this embodiment, a damping operator is introduced during the rank reduction operation of the raw observation data to reduce artificial interference introduced during low-rank projection. After obtaining the rank reduction matrix, the anti-diagonal of the rank reduction matrix is ​​averaged using an averaging operator to recover the filtered data, i.e., low-rank filtered data. Filtering the signal at a preset frequency using a filter yields the first denoised data. Since the raw data undergoes a frequency domain transformation after acquisition, a corresponding inverse frequency domain transformation is required after filtering the low-rank filtered data to finally obtain the first denoised data. The value of the preset frequency is not limited and is determined based on the actual situation. Typically, the selected preset frequency value is between the upper and lower limits of the filtering frequency.

[0075] This embodiment provides a noise suppression method, comprising: acquiring raw observation data obtained by a detector observing seismic signals; introducing a damping operator to perform a rank reduction operation on the raw observation data and obtaining a rank-reduced matrix; obtaining low-rank filtered data based on the rank-reduced matrix; and using a filter to filter signals of a preset frequency in the low-rank filtered data to obtain first denoised data. In this method, a damping operator is introduced during the rank reduction operation on the raw data. Since the damping operator can weaken the artificial interference introduced by the low-rank projection of strong disturbance noise and random noise, it achieves the suppression of incoherent noise in the seismic data.

[0076] The above embodiments achieve denoising of the original observation data by introducing a damping operator. This embodiment, based on the above embodiments, further denoises the original observation matrix using a weighted matrix. Specifically, after filtering the signal of a preset frequency in the low-rank filtered data to obtain the first denoised data, the noise suppression method further includes:

[0077] The weighting matrix is ​​obtained based on the low-rank filtered data and the first denoised data.

[0078] A filter is used to filter the low-rank filtered data and the signals of a preset frequency in the weighting matrix in order to obtain the second denoised data.

[0079] In this embodiment, a non-quadratic fitting function is used instead of a quadratic fitting function to enhance the stability of the low-rank approximation. The denoising method of this application is called Reweighted Damped Singular Spectrum Analysis (RD-SSA). The RD-SSA method uses Tukey's bisquare function, the expression of which is shown in equation (2.0).

[0080]

[0081] In formula (2.0), ε is a constant. Replacing the Frobenius norm with Tukey's bisquare norm, the inversion problem in formula (1.3) is rewritten as formula (2.1):

[0082]

[0083] Because Tukey's bisquare norm is a non-quadratic norm, the minimization in Equation (2.1) is a non-convex optimization problem. If a non-quadratic norm is chosen for the fidelity term, then this type of inversion problem has no closed-form solution. Therefore, a reweighting strategy is used to minimize the non-convex optimization problem. W=[ω i，j ] is a weighted matrix, and its specific expression is shown in formula (2.2):

[0084]

[0085] In formula (2.2), γ i，j ε is the element in the i-th row and j-th column of the absolute difference |DS′|. ε is related to the standardized median absolute deviation, and the expression for ε is shown in formula (2.3):

[0086] ε=6.9681·(median(|DS′-median(DS′)|)) (2.3)

[0087] In formula (2.3), median() represents the median of all elements. Formula (2.2) can be simply expressed as W = g T (|DS′|), where g T This represents a function of |DS′|.

[0088] The embodiment provides a method that uses a non-quadratic fitting function instead of a quadratic fitting function to enhance the stability of the low-rank approximation. After performing the non-quadratic fitting, a reweighting strategy is used to achieve a closed-form solution for denoising seismic noise.

[0089] The damping operator is determined based on the damping factor. In practice, to obtain the damping factor, a preferred embodiment includes obtaining the damping factor by:

[0090] Obtain the upper limit, lower limit, maximum number of iterations, and current number of iterations of the preset damping factor;

[0091] The damping factor is obtained based on the upper limit, lower limit, maximum number of iterations, and current number of iterations.

[0092] As the number of iterations increases, the iterations approximate the original observation data more closely. The truncation rank k in the RD-SSA method is equal to the number of plane waves contained in the seismic profile. The damping factor N increases with the number of iterations, and the formula for calculating the damping factor N is shown in formula (2.4).

[0093] N = N L +(N U -N L )·i / I (2.4)

[0094] In formula (2.4), N L N U These are the lower and upper limits of the damping factor, respectively, given by the user. I is the maximum number of iterations, and i is the current iteration number.

[0095] In this embodiment, the damping operator is used to weaken the artificial interference introduced by the low-rank projection of strong disturbance noise and random noise. By adjusting the damping factor, the irregular strong disturbance noise and random noise can be projected to an acceptable level, and the error accumulated by iteration can be avoided.

[0096] Based on the damping factor obtained in the above embodiments, this embodiment introduces a damping operator during the rank reduction process. Specifically, introducing a damping operator to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix includes:

[0097] Perform frequency domain transformation on the original observation data and obtain the transformed original observation data;

[0098] The transformed original observation data are rearranged to obtain the Hankelized matrix;

[0099] The rank of the Hankelized matrix is ​​reduced based on the damping factor and the truncated rank, and the reduced-rank matrix is ​​obtained.

[0100] The weighting matrix W is closely related to the low-rank matrix S'. Since it is necessary to fit irregular, strongly perturbed noise in the least-squares sense, traditional truncated singular value decomposition easily introduces additional noise artifacts. According to linear algebra, the Hankel matrix H... ω The singular value decomposition of can be expressed by formula (2.5):

[0101]

[0102] In formula (2.5), subscripts 1 and 2 represent the first k maximum values, the remaining singular values, and their corresponding singular vectors. The rank-k approximation of truncated singular value decomposition is achieved by discarding... It is implemented using singular vectors and their related features, but singular value decomposition can only decompose data into a noise subspace and a signal-plus-noise subspace. Because unstable perturbations have strong energy and good focusing, projections onto the low-rank space are prone to severe distortion. Therefore, a damping factor is introduced in the low-rank approximation. Equation (2.6) is the low-rank approximation matrix in the RD-SSA method. The expression for the damping operator is given by formula (2.7).

[0103]

[0104]

[0105] In formula (2.7), T is the damping operator, N is the damping factor, I is the identity matrix, and δ is... The largest element. After obtaining the reduced-rank matrix, the anti-diagonal average of the reduced-rank matrix is ​​performed to restore the filtered data. The specific calculation method is similar to that shown in formula (1.7) above.

[0106] The damping operator introduced in this embodiment during the rank reduction process can weaken the artificial interference introduced by the low-rank projection of strong disturbance noise and random noise. Furthermore, by adjusting the damping factor, the irregular strong disturbance noise and random noise can be projected to an acceptable level, thus minimizing the error accumulated during iteration.

[0107] To make the obtained results closer to the original observation data, a preferred implementation method is that, after filtering the low-rank filtered data and the signals of preset frequencies in the weighting matrix using filters to obtain the second denoised data, the noise suppression method further includes:

[0108] Determine if the current iteration count is greater than the maximum iteration count;

[0109] If so, then obtain the second denoised data;

[0110] If not, return to the steps of introducing a damping operator to reduce the rank of the original observation data and obtaining the reduced-rank matrix.

[0111] The RD-SSA method achieves this by alternately solving the weighted W and the low-rank matrix S', and can be expressed as:

[0112]

[0113]

[0114]

[0115] In formula (3.0), ⊙ represents the Hadamard product, where 1 represents a matrix in which all elements are equal to 1. This represents the result after the next iteration of the current iteration. The final result is obtained by alternately solving formulas (2.8), (2.9), and (3.0).

[0116] During the iteration process, noise suppression methods also include:

[0117] Determine if the current iteration is the first iteration;

[0118] If so, then use a quadratic fitting function to iteratively approximate the original observation data;

[0119] If not, then use a non-quadratic fitting function to iteratively approximate the original observation data.

[0120] There are no restrictions on the quadratic fitting function or non-quadratic fitting function used. For example, in the embodiments of this application, the quadratic fitting function is the Frobenius norm, and the non-quadratic function is the Tukey bisquare norm.

[0121] The method provided in this embodiment involves setting a maximum number of iterations. If the current iteration number does not exceed the maximum number of iterations, the iteration continues, resulting in denoised data that more closely approximates the original observation data, thus achieving a better denoising effect.

[0122] In the above embodiments, the method for noise suppression has been described in detail. This application also provides embodiments corresponding to the noise suppression apparatus. It should be noted that this application describes the embodiments of the apparatus from two perspectives: one based on functional modules and the other based on hardware.

[0123] Figure 2A structural diagram of a noise suppression device provided according to an embodiment of this application. This embodiment, based on functional modules, includes:

[0124] The first acquisition module 10 is used to acquire the raw observation data obtained by the detector from observing the seismic signal;

[0125] The rank reduction and acquisition module 11 is used to introduce a damping operator to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix.

[0126] The second acquisition module 12 is used to acquire low-rank filtered data based on the reduced-rank matrix.

[0127] The filtering module 13 is used to filter the signal of a preset frequency in the low-rank filtered data to obtain the first denoised data.

[0128] Since the embodiments of the apparatus and the embodiments of the method correspond to each other, please refer to the description of the embodiments of the method for the embodiments of the apparatus, which will not be repeated here.

[0129] The noise suppression device provided in this embodiment acquires the original observation data obtained by the detector observing the seismic signal through a first acquisition module; it introduces a damping operator through a rank reduction and acquisition module to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix; a second acquisition module obtains low-rank filtered data based on the rank-reduced matrix; and a filtering module uses a filter to filter the signals of a preset frequency in the low-rank filtered data to obtain the first denoised data. In this device, a damping operator is introduced during the rank reduction operation of the original data. Since the damping operator can weaken the artificial interference introduced by the low-rank projection of strong disturbance noise and random noise, it achieves the suppression of incoherent noise in the seismic data.

[0130] Figure 3 This is a structural diagram of a noise suppression device provided in another embodiment of this application. This embodiment is based on a hardware perspective, such as... Figure 3 As shown, the noise suppression device includes:

[0131] Memory 20 is used to store computer programs;

[0132] The processor 21 is configured to implement the steps of the noise suppression method as described in the above embodiments when executing a computer program.

[0133] The processor 21 may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor 21 may be implemented using at least one of the following hardware forms: Digital Signal Processor (DSP), Field-Programmable Gate Array (FPGA), or Programmable Logic Array (PLA). The processor 21 may also include a main processor and a coprocessor. The main processor, also known as the Central Processing Unit (CPU), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor 21 may integrate a Graphics Processing Unit (GPU), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, the processor 21 may also include an Artificial Intelligence (AI) processor, which is used to handle computational operations related to machine learning.

[0134] The memory 20 may include one or more computer-readable storage media, which may be non-transitory. The memory 20 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In this embodiment, the memory 20 is used to store at least the following computer program 201, which, after being loaded and executed by the processor 21, is capable of implementing the relevant steps of the noise suppression method disclosed in any of the foregoing embodiments. In addition, the resources stored in the memory 20 may also include an operating system 202 and data 203, and the storage method may be temporary or permanent storage. The operating system 202 may include Windows, Unix, Linux, etc. The data 203 may include, but is not limited to, the data involved in the noise suppression method mentioned above.

[0135] In some embodiments, the noise suppression device may further include a display screen 22, an input / output interface 23, a communication interface 24, a power supply 25, and a communication bus 26.

[0136] Those skilled in the art will understand that Figure 3 The structure shown does not constitute a limitation on the noise suppression device and may include more or fewer components than shown.

[0137] The noise suppression apparatus provided in this application includes a memory and a processor. When the processor executes the program stored in the memory, it can implement the following method: the noise suppression method has the same effect as above.

[0138] Finally, this application also provides an embodiment corresponding to a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps described in the above method embodiments.

[0139] It is understood that if the methods in the above embodiments are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and executes all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0140] The computer-readable storage medium provided in this application includes the noise suppression method mentioned above, and has the same effect.

[0141] To enable those skilled in the art to better understand the present application, the present application will be further described in detail below with reference to specific embodiments. The following is the overall process of denoising using the RD-SSA method.

[0142] Step 1: Input data, including raw observation data D, cutoff rank k, and upper limit of damping factor N. L Damping factor lower limit N U Maximum number of iterations I, upper limit of filter frequency ω L , lower limit of filter frequency ω U ;

[0143] Step 2: Perform frequency domain transformation on D in the input data;

[0144] Step 3: When the frequency ω is between the upper and lower limits of the filter frequency, and the current iteration number is less than or equal to the maximum iteration number, perform the following operations:

[0145] By using the upper limit N of the damping factor L Damping factor lower limit N U Given the current iteration number i, the maximum iteration number I, and the damping factor N, calculate using formula (2.4);

[0146] Use formula (1.5) for D ω Rearrange the matrix to obtain the Hankelized matrix H. ω ;

[0147] Using the damping factor N and the cutoff rank k, equations (2.6) and (2.7) are applied to H. ω Perform a rank reduction operation to obtain the rank-reduced data.

[0148] Data after rank reduction Using formula (1.7) to perform anti-diagonal averaging, the low-rank filtered data obtained by the RD-SSA method is obtained.

[0149] Using low-rank filtered data The data D is obtained by filtering using the frequency ω. ω The weighting matrix W is solved using formula (2.9). );

[0150] Using a weighted matrix W and low-rank filtered data Data D is obtained by filtering using frequency ω. ω Using formula (3.0) for iterative calculation, a new D is obtained. ω ;

[0151] Use the new D ω Repeat the above steps;

[0152] Step 4: Finally, D ω Perform the inverse frequency domain transformation to obtain the denoised data D.

[0153] In the method provided in this embodiment, the damping operator is used to weaken the artificial interference introduced by the low-rank projection of strong perturbation noise and random noise; by adjusting the damping factor, the random perturbation noise and random noise can be projected to an acceptable level and the error accumulated by iteration can be avoided; when the data is disturbed by random perturbation noise and random noise, the stability of random perturbation noise and random noise can be stably suppressed.

[0154] The above provides a detailed description of a noise suppression method, apparatus, and medium provided in this application. The various embodiments in the specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to in the method section. It should be noted that those skilled in the art can make several improvements and modifications to this application without departing from the principles of this application, and these improvements and modifications also fall within the protection scope of the claims of this application.

[0155] It should also be noted that, in this specification, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.

Claims

1. A method for noise suppression, characterized in that, include: Acquire the raw observation data obtained by the detector observing the seismic signal; A damping operator is introduced to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix. Obtain low-rank filtered data based on the reduced-rank matrix; A filter is used to filter the signal of a preset frequency in the low-rank filtered data in order to obtain the first denoised data; After filtering the signal of a preset frequency in the low-rank filtered data using a filter to obtain the first denoised data, the method further includes: The weighting matrix is ​​obtained based on the low-rank filtered data and the first denoised data; The filter is used to filter the low-rank filtered data and the signal of the preset frequency in the weighting matrix in order to obtain the second denoised data; The damping operator is determined based on the damping factor. Obtaining the damping factor includes: Obtain the upper limit, lower limit, maximum number of iterations, and current number of iterations of the preset damping factor; The damping factor is obtained based on the upper limit value, the lower limit value, the maximum number of iterations, and the current number of iterations; After filtering the low-rank filtered data and the signal of the preset frequency in the weighting matrix using the filter to obtain the second denoised data, the method further includes: Determine whether the current iteration number is greater than the maximum iteration number; If so, then obtain the second denoised data; If not, return to the step of introducing the damping operator to perform a rank reduction operation on the original observation data and obtaining the rank-reduced matrix; During the iteration process, the method further includes: Determine whether the current iteration number is the first iteration; If so, then a quadratic fitting function is used to iteratively approximate the original observation data; If not, then a non-quadratic fitting function is used to iteratively approximate the original observation data.

2. The noise suppression method according to claim 1, characterized in that, The step of introducing a damping operator to reduce the rank of the original observation data and obtaining the reduced-rank matrix includes: The original observation data is transformed in the frequency domain to obtain the transformed original observation data; The transformed original observation data are rearranged to obtain the Hankelized matrix; The rank of the Hankelized matrix is ​​reduced based on the damping factor and the truncated rank, and the reduced-rank matrix is ​​obtained.

3. The noise suppression method according to claim 2, characterized in that, The step of obtaining low-rank filtered data based on the reduced-rank matrix includes: The reduced-rank matrix is ​​averaged using an anti-diagonal method to obtain the low-rank filtered data.

4. A noise suppression device, characterized in that, include: The first acquisition module is used to acquire the raw observation data obtained by the detector from observing the seismic signal; The rank reduction and acquisition module is used to introduce a damping operator to perform a rank reduction operation on the original observation data and obtain the rank-reduced matrix. The second acquisition module is used to acquire low-rank filtered data based on the reduced-rank matrix. The filtering module is used to filter the signal of a preset frequency in the low-rank filtered data using a filter in order to obtain the first denoised data; After filtering the signal of a preset frequency in the low-rank filtered data using a filter to obtain the first denoised data, the method further includes: The weighting matrix is ​​obtained based on the low-rank filtered data and the first denoised data; The filter is used to filter the low-rank filtered data and the signal of the preset frequency in the weighting matrix in order to obtain the second denoised data; The damping operator is determined based on the damping factor. Obtaining the damping factor includes: Obtain the upper limit, lower limit, maximum number of iterations, and current number of iterations of the preset damping factor; The damping factor is obtained based on the upper limit value, the lower limit value, the maximum number of iterations, and the current number of iterations; After filtering the low-rank filtered data and the signal of the preset frequency in the weighting matrix using the filter to obtain the second denoised data, the method further includes: Determine whether the current iteration number is greater than the maximum iteration number; If so, then obtain the second denoised data; If not, return to the step of introducing the damping operator to perform a rank reduction operation on the original observation data and obtaining the rank-reduced matrix; The iteration process also includes: Determine whether the current iteration number is the first iteration; If so, then a quadratic fitting function is used to iteratively approximate the original observation data; If not, then a non-quadratic fitting function is used to iteratively approximate the original observation data.

5. A noise suppression device, characterized in that, include: Memory, used to store computer programs; A processor for executing the computer program to implement the steps of the noise suppression method as described in any one of claims 1 to 3.

6. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the noise suppression method as described in any one of claims 1 to 3.

Images