Background noise suppression method for non-stationary signal decomposition of acquired data of vibroseis

Abstract

The invention discloses a background noise suppression method for non-stationary signal decomposition of vibroseis acquisition data, and the method comprises the steps: constructing an effective signal sparse representation model of time-frequency prior constraint of a downward propagation scanning signal according to the difference between an effective signal and background noise in uncorrelated seismic data in the efficient acquisition of a vibroseis in a time-frequency domain; according to the method, non-stationary signal decomposition is carried out on vibroseis acquisition data, adaptive separation of effective signals and background noise in the vibroseis is realized, effective signals in unrelated vibroseis seismic data are extracted, and background noise is abandoned, so that background noise suppression is realized. According to the method, background noise can be effectively suppressed, effective signals in the vibroseis are prevented from being suppressed, and a more reliable and efficient technical scheme is provided for processing vibroseis seismic data.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of land oil and gas seismic exploration data processing, and particularly relates to a controllable source high-efficiency acquisition data background noise suppression technology based on non-stationary signal decomposition, which is particularly suitable for suppressing background noise such as harmonic interference, mechanical vibration and aliasing noise in uncorrelated controllable source data in a high-efficiency acquisition mode, so as to improve the signal-to-noise ratio of land controllable source seismic data. BACKGROUND

[0002] In recent years, according to the literature [1], [2], controllable source high-efficiency acquisition technology has made significant progress, forming a series of advanced technologies such as alternating scanning, sliding scanning, multi-point synchronous scanning and high-fidelity acquisition. The application of these technologies has greatly improved the daily acquisition efficiency from several hundred shots to tens of thousands of shots, and has effectively promoted the land controllable source exploration into a new stage of high efficiency and high density. However, the high-efficiency acquisition mode inevitably introduces various background noises, which seriously affect the data quality. Specifically, these background noises include: (1) harmonic interference caused by nonlinear vibration of the source hydraulic servo system. (2) In the sliding scanning mode, the aliasing of adjacent shot signals in time and space. (3) Ground vibration caused by the controllable source mechanical system during operation, producing a kind of low-frequency mechanical noise propagating along the near-surface. This poses a fundamental challenge to traditional noise suppression methods based on the assumption of stationarity. Therefore, how to effectively eliminate the background noise generated by the controllable source in the high-efficiency acquisition process is a key link to obtain high-quality seismic data.

[0003] For the problem of suppressing background noise of the controllable source, it is mainly divided into two directions of outdoor acquisition technology optimization and indoor data processing. In the aspect of outdoor acquisition technology, researchers reduce harmonic interference from the source by improving the source system and optimizing the sweep signal design. In the literature [3], by improving the sweep signal design, the controllable source can effectively excite low-frequency signals and reduce distortion; in the literature [4], by adjusting the nonlinear sweep rate, the scan time is prolonged in the high-frequency band to compensate for the attenuation, so that the exploration resolution of Kazakhstan S area is significantly improved; in the literature [5], the response speed of the hydraulic servo valve and the vibrator mechanical structure are improved, the upper limit of the controllable source high-frequency output is increased, and the quality of the controllable source high-frequency output signal is improved to a certain extent; in the literature [6], a damping Ricker wavelet model is proposed, which expands the low-frequency component while maintaining the narrow side lobe characteristic. In the aspect of indoor data processing technology, suppressing background noise becomes the core of the research. In the literature [7], the traditional filtering method is extended, and a phase shift filter suitable for multi-shot data is designed by statistically analyzing the ground force signal, but this method can only be used to suppress high-order harmonics in the correlation data; in the literature [8], a frequency domain filtering method based on ground force signal decomposition is proposed, which designs the filter by accurately calculating the frequency spectrum ratio of the fundamental wave and the harmonic, but the processing needs to start from the last shot data, and cannot realize single-shot operation; in the literature [9], an optimization method based on adaptive dictionary learning is proposed, which dynamically adjusts the dictionary through data-driven, effectively overcomes the limitations of fixed dictionary in processing complex seismic data, and significantly improves the data fidelity and noise suppression effect. However, the above existing technologies cannot completely suppress the background noise in the high-efficiency acquisition of the controllable source.

[0004] Reference:

[0005] [1] TONG X Q, LIN J, JIANG T, et al. Development of land vibroseis. Progress in Geophysics, 2012, 27(5): 1912-1921.

[0006] [2] NI Y D, WANG J F, MA T, et al. Progress of controllable source seismic acquisition technology. Oil Geophysical Prospecting, 2011, 46(3): 349-356.

[0007] [3] Bagaini C. Low-frequency vibroseis data with maximum displacement sweeps. The Leading Edge, 2008, 27(5): 582-591.

[0008] [4] LAN J D. Application of controllable source nonlinear sweep in high resolution seismic acquisition. Geophysical Prospecting for Petroleum, 2008, 47(2): 208-211.

[0009] [5] Tao ZF. Discussion on improving the output quality of high-frequency signals of controlled source. Geophysical Prospecting Equipment, 2008, 18(2): 71-77.

[0010] [6] Xu LL, Zhang J, Zhao GY. Nonlinear sweep signal design method for controlled source based on damped Ricker wavelet. Oil Geophysical Prospecting, 2022, 57(3): 540-549.

[0011] [7] Huang JP, Zhou XF, Guo J, et al. Method of harmonic interference suppression in slip sweep recording. Journal of China University of Petroleum (Edition of Natural Science), 2012, 36(2): 81-85.

[0012] [8] Sicking C, Fleure T, Nelan S, et al. Slip sweep harmonic noise rejection on correlated shot data[J]. Expanded Abstracts of 79th Annual International SEG Meeting, 2009, 36-40.

[0013] [9] Mao HB, Ma JY, Wang XT, et al. Harmonic noise suppression method for controlled source data based on adaptive dictionary learning. Geophysical Prospecting, 2020, 59(5): 725-735. SUMMARY

[0014] In order to overcome the shortcomings of the prior art, the application provides a kind of based on non-stationary signal decomposition's controlled source high efficiency acquisition data background noise suppression technology, based on non-stationary signal decomposition technology (Non-stationary Signal Decomposition and Reconstruction, NSSDR), for the difference of effective signal and background noise in time-frequency domain in uncorrelated seismic data in controlled source high efficiency acquisition, the effective signal sparse representation model of the time-frequency priori constraint of downward propagation sweep signal is constructed, the adaptive separation of effective signal and background noise in controlled source is realized after the non-stationary signal decomposition of controlled source acquisition data using orthogonal matching pursuit algorithm, the effective signal and background noise in controlled source acquisition data are separated, finally, the background noise is discarded, and the purpose of suppressing background noise is achieved.

[0015] The core of the present application is a non-stationary signal decomposition technology based on the characteristic difference of the effective signal and background noise in the time-frequency domain in the uncorrelated controllable source seismic data. In the process of controllable source efficient acquisition, the uncorrelated controllable source seismic data not only contains effective signals, but also carries a variety of background noises. The characteristics of the effective signal in the uncorrelated controllable source seismic data in the time-frequency domain are consistent with the slope of the downward propagating scanning signal, and only the difference of the zero position exists in the time axis. In contrast, the background noise in the uncorrelated controllable source seismic data is complex and diverse, in which the harmonic noise signal has an integer multiple of the slope of the effective signal in the time-frequency domain, and the mechanical noise is randomly distributed in the whole time-frequency domain. Therefore, the effective signal in the uncorrelated controllable source seismic data in the time-frequency domain is sparse, while the background noise is dense as a whole. Therefore, by establishing the effective signal sparse representation model of the time-frequency prior constraint of the downward propagating scanning signal, the effective signal can be accurately extracted from the uncorrelated controllable source seismic data, so as to realize the suppression effect of the background noise. Compared with the traditional background noise suppression technology in the controllable source, the method of the present application performs better in suppressing the background noise, and can effectively prevent the effective signal in the controllable source from being suppressed, thereby providing a more reliable and efficient method for the processing of controllable source seismic data.

[0016] The technical scheme of the present application is as follows:

[0017] A controllable source efficient acquisition data non-stationary signal decomposition background noise suppression method based on non-stationary signal decomposition technology, specifically comprising the following steps:

[0018] A. Data acquisition: acquiring uncorrelated controllable source seismic two-dimensional data d(x,t) and scanning signal s(t), wherein x represents offset or trace number; t represents time;

[0019] B. Establishing an ideal model (effective signal sparse representation model in uncorrelated controllable source seismic data) to extract the effective signal in the uncorrelated controllable source seismic data:

[0020] Under ideal conditions, the uncorrelated controllable source seismic signal data d(x,t) received by the seismic detector is the sum of the signals generated by the convolution of multiple underground reflection interfaces and the scanning signal s(t). The signal characteristics generated by the convolution of a single underground reflection interface and the scanning signal s(t) only differ in amplitude and starting time from the scanning signal. Therefore, the effective signal sparse representation model of the time-frequency prior constraint of the downward propagating scanning signal of the i-th column of uncorrelated controllable source seismic data is established:

[0021]

[0022] where i is a constant, is an optional single column offset or trace number; K is the total number of reflection coefficients of the subsurface medium; t k is the time delay of the scanning signal after convolution with the kth subsurface medium reflection interface; a k is the unknown variable amplitude of the signal after the kth subsurface medium reflection interface is convolved with the scanning signal; f(t) is the known variable frequency of the signal after the kth subsurface medium reflection interface is convolved with the scanning signal; φ k is the initial phase of the signal after the kth subsurface medium reflection interface is convolved with the scanning signal, that is, the difference time delay with the scanning signal.

[0023] C. Adjust the sparse representation model of the uncorrelated vibroseis seismic data considering the actual noise:

[0024] In the actual acquisition process, the vibroseis seismic data received by the geophone usually contains various background noises n(t). Therefore, the sparse representation model of the ith column of the uncorrelated vibroseis seismic data is further represented as:

[0025]

[0026] where n(t) represents the background noise;

[0027] D. Matrix form transformation: representing the sparse representation model of the uncorrelated vibroseis seismic data in the form of a matrix;

[0028] The sparse representation model of the ith column of the uncorrelated vibroseis seismic data is represented in the form of a matrix S:

[0029] d i = Sa+n,

[0030] where S is the scanning signal matrix, a is the amplitude coefficient matrix, represented as a = [a1(t), a2(t), …, a K (t)], d i is the ith column signal in the vibroseis seismic data, and n represents the background noise n(t).

[0031] E. Constructing an objective function and solving it to obtain the final initial amplitude coefficient matrix:

[0032] Since the scanning signal matrix is known, the problem of extracting the effective signal in the uncorrelated vibroseis is transformed into solving the amplitude coefficient matrix a from the ith column record d i of the vibroseis seismic data. Therefore, the following objective function can be constructed:

[0033]

[0034] In the above formula: λ is a regularization parameter greater than 0, is a square of two norm, ‖‖1 is a one norm. Input initial amplitude coefficient matrix a, calculate current residual When the residual is less than a manually set noise threshold ε, stop iteration, and obtain final initial amplitude coefficient matrix a. Therefore, according to the final initial amplitude coefficient matrix, effective signal Sa in uncorrelated vibroseis seismic data and background noise d can be obtained i -Sa.

[0035] Through the above steps, the effective signal in uncorrelated vibroseis seismic data can be obtained, that is, the background noise suppression effect in uncorrelated vibroseis is realized.

[0036] Compared with the prior art, the beneficial effects of the present application are:

[0037] The present application provides a non-stationary signal decomposition background noise suppression method for controllable seismic source acquisition data, based on non-stationary signal decomposition reconstruction technology, a sparse representation model of time-frequency prior effective signal constraint of downward propagation scanning signal is constructed, and an orthogonal matching pursuit algorithm is used to realize the separation of effective signal and background noise in the controllable seismic source. Compared with the prior art, the present application has the following technical advantages:

[0038] (1) The traditional technology is to suppress specific background noise according to the characteristics of each type of background noise, but in the process of efficient acquisition, the types of background noise are very diverse and the characteristics are different. The present application anchors the characteristics of the effective signal in the time-frequency domain to be consistent with the slope of the scanning signal, and only the difference characteristics of the zero position exist on the time axis. The effective signal presents sparsity in the time-frequency domain. Therefore, by establishing a sparse representation model of time-frequency prior constraint of scanning signal, the effective signal can be accurately extracted, thereby realizing adaptive separation of background noise. Compared with the traditional multi-round noise reduction process, this method significantly reduces the processing steps and improves the efficiency.

[0039] (2) The existing method directly processes the correlated controllable seismic source seismic data, which is easy to suppress the effective signal while suppressing the background noise. The present application processes uncorrelated controllable seismic source seismic data, separates the effective signal and the background noise from the source, and then performs cross-correlation operation on the processed uncorrelated controllable seismic source seismic data and the scanning signal, which can effectively protect the effective signal. BRIEF DESCRIPTION OF DRAWINGS

[0040] Figure 1 is a flowchart of the method of the present application.

[0041] Figure 2 is the waveform and time-frequency diagram of each order harmonic in the test data.

[0042] Wherein, (a) is a first-order harmonic waveform diagram, (b) is a second-order harmonic waveform diagram, (c) is a third-order harmonic waveform diagram, (d) is a first-order harmonic time-frequency diagram, (e) is a second-order harmonic time-frequency diagram, and (f) is a third-order harmonic time-frequency diagram.

[0043] Figure 3 is a comparison chart of test data scanning signals before and after harmonic interference;

[0044] Wherein, (a) is a waveform diagram before harmonic interference, (b) is a waveform diagram after harmonic interference, (c) is a time-frequency diagram before harmonic interference, and (d) is a time-frequency diagram after harmonic interference.

[0045] Figure 4 is uncorrelated controllable seismic source test data before noise interference;

[0046] Wherein, (a) is an uncorrelated waveform diagram before noise interference, and (b) is an uncorrelated time-frequency diagram before noise interference.

[0047] Figure 5 is uncorrelated controllable seismic source test data after noise interference;

[0048] Wherein, (a) is an uncorrelated waveform diagram after noise interference, and (b) is an uncorrelated time-frequency diagram after noise interference.

[0049] Figure 6 is uncorrelated controllable seismic source test data after noise interference and after non-stationary signal decomposition processing;

[0050] Wherein, (a) is a waveform diagram after non-stationary signal decomposition processing, and (b) is a time-frequency diagram after non-stationary signal decomposition processing.

[0051] Figure 7 is uncorrelated controllable seismic source data;

[0052] Figure 8 is a comparison of controllable seismic source data processing results after correlation;

[0053] Wherein, (a) is a cross-correlation result, (b) is a result after F-K domain predictive deconvolution, and (c) is a result after non-stationary signal decomposition processing.

[0054] Figure 9 is a local amplification (3-5s) result of controllable seismic source data after correlation.

[0055] Wherein, (a) is a cross-correlation result, (b) is a result after F-K domain predictive deconvolution, (c) is a result after non-stationary signal decomposition processing, (d) is (b) minus (a), and (e) is (c) minus (a). DETAILED DESCRIPTION

[0056] The application will be further described by examples in connection with the drawings, but the application is not limited in any way by the scope of the application.

[0057] The application provides a background noise suppression method for efficient acquisition of data non-stationary signal decomposition of a controllable source, and the technical process comprises the following steps:

[0058] A. Obtain uncorrelated controllable source seismic two-dimensional data d(x,t) and scanning signal s(t), wherein x represents offset or trace number direction; t represents time direction;

[0059] B. In an ideal state, the seismic signal received by the detector is the sum of the convolution of multiple underground reflection interfaces and the scanning signal. The signal after the convolution of a single underground reflection interface and the scanning signal only differs from the scanning signal in amplitude and starting time. Therefore, before being correlated with the scanning signal, the single column record of the controllable source seismic can be represented as: wherein i is a constant, d is an optional single column; a k (t) is the unknown variable amplitude of the signal after the convolution of the kth underground reflection interface and the scanning signal; f(t) is the known variable frequency of the signal after the convolution of the kth underground reflection interface and the scanning signal; φ k is the initial phase of the signal after the convolution of the kth underground reflection interface and the scanning signal, that is, the time delay difference from the scanning signal; n(t) is background noise;

[0060] C. The single column record of the controllable source seismic before being correlated with the scanning signal can be represented in the form of a matrix: d i = Sa+n, wherein the elements in the S matrix are S k = sin[f(t)+φ k ], the a matrix is a=[a1(t),a2(t),…,a K (t)], and d i is the i-th column signal in the seismic data.

[0061] D. Since S is known, the problem is transformed into solving the amplitude coefficient a from d i . Therefore, the following objective function can be constructed: In the above formula: λ is a regularization parameter greater than 0, and when the two-norm of the residual is less than the noise threshold set artificially , the iteration is stopped.

[0062] Through the above steps, adaptive non-stationary signal decomposition reconstruction and signal enhancement of land deep seismic are realized, and the method flow chart is shown in Figure 1 .

[0063] In order to verify the effectiveness and advancement of the background noise suppression method proposed in this paper for the non-stationary signal decomposition of vibroseis data with high efficiency, the non-stationary signal decomposition technique was applied to test data and actual data to evaluate the effect of extracting effective signals from uncorrelated vibroseis earthquakes.

[0064] In the test data, the first, second and third order harmonic components are first superimposed on the basis of the scanning signal, such as Figure 2 The frequencies of these harmonics are 1, 2, and 3 times the frequency of the scanning signal, respectively. Figure 2 As shown in (d), (e) and (f), the energy intensity decreases to 80%, 50% and 40% of the scanning signal respectively.

[0065] from Figure 2 It can be clearly observed that as the harmonic frequency slope gradually increases, its energy intensity gradually decreases. When harmonic interference is added to the scanning signal, the time domain waveform is obviously distorted, and high-frequency interference components can be seen in the time-frequency domain, such as Figure 3 Under ideal conditions, that is, ignoring factors such as the absorption of energy by the formation medium, uncorrelated vibroseis data can be obtained by convolving the scanning signal with the synthetic wave impedance interface.

[0066] Figure 4 (a) shows the waveform of the data after the convolution of the scanning signal and the synthetic wave impedance interface without considering harmonic interference. This is the uncorrelated controllable source data. From this waveform, it is difficult to clearly distinguish the five wave impedance interfaces due to the long-term characteristics of the scanning signal. However, Figure 4 The time-frequency domain plot of the uncorrelated vibroseis data shown in (b) clearly reveals that the frequency slopes of the five effective signals are not only consistent with those of the sweep signal, but also independent of each other, with significant energy differences. Further observation reveals that the starting points of the five tilted signals in the time-frequency domain perfectly coincide with the temporal positions of the five coefficients of the synthetic wave impedance interface, and the energy differences also correspond to the absolute values ​​of these coefficients. Secondly, in addition to the existing harmonic interference, random noise interference is added to the sweep signal. Without considering complex conditions such as formation medium absorption, the sweep signal with background noise is convolved with the synthetic wave impedance interface to obtain uncorrelated vibroseis data with background noise.

[0067] Figure 5 (a) shows the total response after the convolution of the scanning signal with background noise and the synthetic wave impedance interface, that is, the uncorrelated vibroseis data with background noise. It is difficult to clearly distinguish the five wave impedance interfaces from the waveform diagram. Figure 4(a), the interference of background noise can be observed obviously. This shows that the background noise increases the complexity of the signal and the effective information is covered. Figure 5 (b) is the time-frequency domain graph of the uncorrelated vibroseis data with background noise. It can be clearly seen from the graph that there are not only harmonic noises with greater slope around the effective signal, but also random noise distribution in the global range. These noises undoubtedly seriously affect the identification of the effective signal.

[0068] The uncorrelated vibroseis data with background noise is subjected to non-stationary signal decomposition processing, and the result is shown in Figure 6 By observing the waveform graph and the time-frequency graph, it can be seen that after the non-stationary signal decomposition processing, the effective signal is completely preserved, and the processed result clearly shows the consistency with the result shown in Figure 4 This proves the excellent ability of the non-stationary signal decomposition technology in extracting effective signals from uncorrelated vibroseis data with complex background noise.

[0069] In order to further verify the superiority of the non-stationary signal decomposition technology in extracting effective signals and suppressing background noise in uncorrelated vibroseis data with complex background noise, compared with the F-K domain predictive deconvolution technology, the high-efficiency acquisition actual data of a controlled source in the southwest Gobi work area of a basin is selected as the test sample. The surface of this work area is mainly composed of gravel, and the geological structure of the shallow layer is complex, resulting in low signal-to-noise ratio of the original single-shot data, and the background noise is various and very active. The scanning acquisition technology is used in the acquisition process.

[0070] Figure 7 The uncorrelated vibroseis data selected in this paper as test data is shown, each shot contains 230 channels, and each channel records for 30 seconds with a sampling interval of 1 millisecond. From Figure 7 It can be clearly seen that due to the long time of the long scanning signal of the excitation source, the seismic data information similar to that generated by the explosive source cannot be directly extracted. Therefore, the cross-correlation with the scanning signal is needed to obtain the seismic data similar to the interpretability of the explosive source.

[0071] The selected uncorrelated vibroseis data is processed by the F-K domain predictive deconvolution technology and the non-stationary signal decomposition technology, respectively. Figure 8 The whole shot effect graph of the uncorrelated vibroseis data after the above-mentioned technology processing and then cross-correlation with the scanning signal is shown, where (a) is the cross-correlation processing result, (b) is the result after the F-K domain predictive deconvolution technology, and (c) is the result after the non-stationary signal decomposition processing.

[0072] From Figure 8(a) It can be seen that the field data after only cross-correlation processing is disturbed by various complex factors, containing a large amount of background noise. This makes the field data less interpretable. As shown in Figure 8 (b), the overall background noise suppression effect of the F-K domain predictive deconvolution technique is not satisfactory, and there is still background noise disturbance in the near-mid offset shallow deep seismic data, and the events are not continuous enough. While Figure 8 (c) after non-stationary signal decomposition processing, the global seismic data events are clear and continuous, proving that the background noise suppression is more complete.

[0073] In order to further show the details of the technical effect, Figure 9 then the local (3-5s) of the vibroseis processing result is enlarged and displayed. Among them, (a) is the cross-correlation processing result, (b) is the F-K domain predictive deconvolution technique processing result, (c) is the non-stationary signal decomposition processing result, (d) is the residual error of the F-K domain predictive deconvolution technique processing result and the cross-correlation processing result, and (e) is the residual error of the non-stationary signal decomposition technique processing result and the cross-correlation processing result. From the area marked by the red box in the local enlarged view, Figure 9 (a) there is a large amount of strong background noise disturbance, especially around trace number 140, and the events are almost indistinguishable. Figure 9 (b) and Figure 9 (c) can be seen that compared with the F-K domain predictive deconvolution technique, the non-stationary signal decomposition technique can obtain more coherent and clear seismic data events. While using the F-K domain predictive deconvolution technique, there is still some background noise disturbance, and overall, the non-stationary signal decomposition technique processed seismic data performs better in terms of signal-to-noise ratio.

[0074] It should be noted that the purpose of publishing the embodiments is to help further understand the present application, but those skilled in the art can understand that various substitutions and modifications are possible without departing from the scope of the present application and the appended claims. Therefore, the present application should not be limited to the disclosed content of the embodiments, and the scope of protection claimed by the present application is subject to the scope defined by the claims.

Claims

1. A method for suppressing background noise in the decomposition of non-stationary signals from vibroseis acquisition data, characterized by: Aiming at the difference between the effective signal and background noise in the time-frequency domain in the uncorrelated seismic data during the efficient acquisition of controllable source, a sparse representation model of effective signal with time-frequency prior constraints on the downward propagating scanning signal is constructed. The controllable source acquisition data is decomposed into non-stationary signals to achieve adaptive separation of effective signal and background noise in the controllable source, extract the effective signal from the uncorrelated controllable source seismic data, and discard the background noise, thereby suppressing the background noise.

2. The background noise suppression method for decomposing non-stationary signals of vibroseis acquisition data according to claim 1, characterized in that: Construct an effective signal sparse representation model with time-frequency prior constraints on the downward-propagating scan signal, based on the actual noise adjustment model; The effective signal sparse representation model is expressed as: Where i is a constant, which is an optional single column offset or track number; t represents time; t k is the time delay after the scanning signal is convolved with the kth underground medium reflection interface; d is the uncorrelated vibroseis seismic signal data received by the seismic detector; K is the total number of underground medium reflection coefficients; s is the scanning signal; a k (t) is the unknown amplitude of the signal after the k-th underground medium reflection interface is convolved with the scanning signal; f(t) is the known frequency of the signal after the k-th underground medium reflection interface is convolved with the scanning signal; φ k It is the initial phase of the signal after the convolution of the k-th underground medium reflection interface and the scanning signal, that is, the difference time delay from the scanning signal; n(t) represents the background noise.

3. The background noise suppression method for decomposing non-stationary signals of vibroseis acquisition data according to claim 2, characterized in that: The sparse representation model of uncorrelated vibroseis seismic data is expressed in matrix form. The objective function is constructed and solved to obtain the final initial amplitude coefficient matrix. The objective function is expressed as: Among them, S is the scanning signal matrix; a is the amplitude coefficient matrix; λ is the regularization parameter greater than 0, is the square of the two-norm, ‖‖1 is the one-norm; Input the initial amplitude coefficient matrix and calculate the current residual. When the residual is less than the noise threshold, stop the iteration and obtain the final initial amplitude coefficient matrix. According to the final initial amplitude coefficient matrix, the effective signal in the uncorrelated vibroseis seismic data is Sa and the background noise is d. i -Sa; This achieves background noise suppression in the decomposition of non-stationary signals of vibroseis acquisition data.

4. The background noise suppression method for decomposing non-stationary signals of vibroseis acquisition data according to claim 3, characterized in that: The single column record of vibroseis earthquake before being correlated with the scanning signal is expressed in matrix form: d i =Sa+n, Among them, the elements in the S matrix are S k = sin[f(t)+φ k ], the matrix a is a=[a1(t),a2(t)…,a K (t)],d i is the signal in column i of the seismic data.

5. The background noise suppression method for decomposing non-stationary signals of vibroseis acquisition data according to claim 1, characterized in that: Specifically, the orthogonal matching pursuit algorithm is used to decompose the non-stationary signal of the controllable source acquisition data, so as to realize the adaptive separation of the effective signal and background noise in the controllable source.

Images