阅读说明：本技术 地震数据的面波噪声去除方法及装置 (Method and device for removing surface wave noise of seismic data ) 是由 张才 于 2020-05-18 设计创作，主要内容包括：本发明公开了一种地震数据的面波噪声去除方法及装置,其中,该方法包括：获取包含面波噪声的原始地震数据；基于预定规则将原始地震数据分割为低频数据和高频数据,其中,低频数据包括：第一低频子数据和第二低频子数据,高频数据包括：第一高频子数据和第二高频子数据,第二低频子数据包括面波噪声；根据已训练的神经网络和预先设置的滤波器对第二高频子数据进行校正处理,以生成第二高频校正数据,校正处理包括：高低频转换处理；根据第二高频校正数据和第二低频子数据对原始地震数据进行去除面波噪声操作,以生成去除面波噪声后的地震数据。通过本发明,可以更有效地去除原始地震数据中的面波噪声。(The invention discloses a method and a device for removing surface wave noise of seismic data, wherein the method comprises the following steps: acquiring original seismic data containing surface wave noise; segmenting the raw seismic data into low frequency data and high frequency data based on a predetermined rule, wherein the low frequency data comprises: the first low frequency subdata and the second low frequency subdata, the high frequency data comprising: the first high-frequency subdata and the second high-frequency subdata comprise surface wave noise; and performing correction processing on the second high-frequency sub-data according to the trained neural network and a preset filter to generate second high-frequency correction data, wherein the correction processing comprises the following steps: high-low frequency conversion processing; and performing surface wave noise removal operation on the original seismic data according to the second high-frequency correction data and the second low-frequency subdata to generate seismic data with surface wave noise removed. By the method and the device, the surface wave noise in the original seismic data can be removed more effectively.)

1. A method of surface wave noise removal for seismic data, the method comprising:

acquiring original seismic data containing surface wave noise;

segmenting the raw seismic data into low frequency data and high frequency data based on a predetermined rule, wherein the low frequency data comprises: first low frequency sub-data and second low frequency sub-data, the high frequency data comprising: the first high-frequency subdata and the second high-frequency subdata comprise the surface wave noise;

performing correction processing on the second high-frequency sub-data according to the trained neural network and a preset filter to generate second high-frequency correction data, wherein the correction processing comprises: high-low frequency conversion processing;

and performing surface wave noise removal operation on the original seismic data according to the second high-frequency correction data and the second low-frequency subdata to generate seismic data with surface wave noise removed.

2. The method of claim 1, wherein performing a surface wave noise removal operation on the raw seismic data based on the second high frequency correction data and the second low frequency sub-data to generate surface wave noise removed seismic data comprises:

determining the surface wave noise according to the second high-frequency correction data and the second low-frequency subdata;

and performing subtraction operation on the original seismic data and the surface wave noise to generate the seismic data with the surface wave noise removed.

3. The method of claim 1, wherein segmenting the raw seismic data into low frequency data and high frequency data based on predetermined rules comprises:

segmenting the original seismic data into low-frequency data and high-frequency data according to a frequency domain segmentation rule, wherein the low-frequency data comprises the surface wave noise;

dividing the low-frequency data into the first low-frequency subdata and the second low-frequency subdata based on a time-space domain division rule;

and dividing the high-frequency data into the first high-frequency subdata and the second high-frequency subdata based on the time-space domain division rule.

4. The method of claim 1, wherein performing correction processing on the second high frequency sub-data according to the trained neural network and the filter to generate second high frequency correction data comprises:

inputting the second high-frequency subdata into a trained neural network for high-frequency and low-frequency conversion processing so as to generate prediction data;

inputting the prediction data to the filter to output the second high frequency correction data.

5. The method of claim 1, wherein the neural network is trained by:

acquiring historical seismic data, the historical seismic data comprising: historical first low-frequency subdata and historical first high-frequency subdata;

training the neural network by using the historical first low-frequency subdata and the historical first high-frequency subdata as training data.

6. The method of claim 5, wherein the historical seismic data further comprises: historical second high frequency correction data and historical second low frequency sub-data, the filter being set by:

setting the filter according to a remaining error value after performing a subtraction operation on the historical second high-frequency correction data and the historical second low-frequency sub-data.

7. An apparatus for surface wave noise removal of seismic data, the apparatus comprising:

the data acquisition unit is used for acquiring original seismic data containing surface wave noise;

a segmentation unit for segmenting the original seismic data into low frequency data and high frequency data based on a predetermined rule, wherein the low frequency data includes: first low frequency sub-data and second low frequency sub-data, the high frequency data comprising: the first high-frequency subdata and the second high-frequency subdata comprise the surface wave noise;

a correction unit, configured to perform correction processing on the second high-frequency sub-data according to a trained neural network and a preset filter to generate second high-frequency correction data, where the correction processing includes: high-low frequency conversion processing;

and the noise removing unit is used for removing surface wave noise from the original seismic data according to the second high-frequency correction data and the second low-frequency subdata to generate seismic data with surface wave noise removed.

8. The apparatus of claim 7, wherein the noise removing unit comprises:

the noise determining module is used for determining the surface wave noise according to the second high-frequency correction data and the second low-frequency subdata;

and the noise removing module is used for performing subtraction operation on the original seismic data and the surface wave noise to generate the seismic data after the surface wave noise is removed.

9. The apparatus of claim 7, wherein the segmentation unit comprises:

a frequency domain segmentation module, configured to segment the original seismic data into low-frequency data and high-frequency data according to a frequency domain segmentation rule, where the low-frequency data includes the surface wave noise;

the low-frequency subdata generating module is used for dividing the low-frequency data into the first low-frequency subdata and the second low-frequency subdata based on a time-space domain division rule;

and the high-frequency subdata generating module is used for dividing the high-frequency data into the first high-frequency subdata and the second high-frequency subdata based on the time-space domain division rule.

10. The apparatus of claim 7, wherein the correction unit comprises:

the prediction data generation module is used for inputting the second high-frequency subdata into a trained neural network to perform high-frequency and low-frequency conversion processing so as to generate prediction data;

a correction module for inputting the prediction data to the filter to output the second high frequency correction data.

11. The apparatus of claim 7, further comprising:

a neural network training unit for training the neural network,

the neural network training unit includes:

a historical data acquisition module for acquiring historical seismic data, the historical seismic data comprising: historical first low-frequency subdata and historical first high-frequency subdata;

and the training module is used for training the neural network by taking the historical first low-frequency subdata and the historical first high-frequency subdata as training data.

12. The apparatus of claim 11, wherein the historical seismic data further comprises: the apparatus further includes historical second high frequency correction data and historical second low frequency sub-data:

a filter setting unit for setting the filter,

the filter setting unit is specifically configured to: setting the filter according to a remaining error value after performing a subtraction operation on the historical second high-frequency correction data and the historical second low-frequency sub-data.

13. An electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the steps of the method of any of claims 1 to 6 are implemented when the processor executes the program.

14. A computer-readable storage medium, on which a computer program is stored which, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6.

Technical Field

The invention relates to the field of seismic data processing, in particular to a method and a device for removing surface wave noise of seismic data.

Background

Denoising is an important link in seismic data processing. The original seismic data contains a large amount of noise interference, the noise interference energy is often higher than the effective signal energy, and the original seismic data needs to be denoised to carry out subsequent seismic data processing and research.

The surface wave is a common noise type of land seismic data, has the characteristics of low speed, low frequency, strong energy and the like, and is completely submerged in noise in a surface wave area. If the terrain is severely fluctuated or the ground surface has speed abnormal bodies, the strong energy surface waves can also generate scattering interference, and the signal-to-noise ratio of the data is seriously reduced.

Conventional surface wave denoising techniques are mostly based on differences of surface wave noise and effective reflection signals in frequency and visual velocity, such as high-pass filtering, FK filtering, pyramid filtering, time-frequency domain denoising, curvelet domain denoising, and the like.

The method comprises the steps of firstly transforming an original signal into another domain, enabling an effective signal and a surface wave in the transformed domain to have certain separability, suppressing the surface wave signal by designing a filter, and then reversely transforming the surface wave signal into a time domain to obtain data of removing the surface wave. Since the effective signal and the surface wave noise are not strictly limited in the transform domain, the effective wave is inevitably damaged while the surface wave is removed.

The time-frequency domain denoising method generally adopts methods such as short-time Fourier transform and the like to transform original data into a time-frequency domain, prediction is carried out in a local space according to the frequency and the speed characteristics of a surface wave, strong energy in a surface wave region is replaced by a local space average value, and data after the surface wave energy is compressed are obtained by inverse transformation back to the time-space domain. The method cannot really remove the surface wave, but suppresses the energy of the surface wave to reduce the surface wave to an effective signal energy level. Since the surface wave noise is not really removed, surface wave interference still exists in the data.

With the rise of deep learning technology, some methods based on deep learning appear in the field of seismic data processing, how to obtain high-quality samples is the core of deep learning, conventional denoised data is mainly used as sample labels in the current deep learning method, and the previous discussion shows that the traditional surface wave denoising method has technical defects and cannot provide accurate samples, so that the deep learning technology is difficult to surpass the traditional denoising method in the field of surface wave denoising.

Disclosure of Invention

Accordingly, the present invention is directed to a method and apparatus for removing surface wave noise from seismic data to solve at least one of the above-mentioned problems.

According to a first aspect of the present invention, there is provided a method of surface wave noise removal for seismic data, the method comprising: acquiring original seismic data containing surface wave noise; segmenting the raw seismic data into low frequency data and high frequency data based on a predetermined rule, wherein the low frequency data comprises: first low frequency sub data (LP1) and second low frequency sub data (LP2), the high frequency data comprising: first high frequency sub-data (HP1) and second high frequency sub-data (HP2), the second low frequency sub-data including the surface wave noise; performing correction processing on the second high-frequency sub-data (HP2) according to the trained neural network and a preset filter to generate second high-frequency correction data, the correction processing including: high-low frequency conversion processing; and performing surface wave noise removal operation on the original seismic data according to the second high-frequency correction data and the second low-frequency subdata (LP2) to generate seismic data with surface wave noise removed.

Performing surface wave noise removal operation on the original seismic data according to the second high-frequency correction data and the second low-frequency subdata (LP2) to generate surface wave noise-removed seismic data, including: determining the surface wave noise according to the second high-frequency correction data and the second low-frequency sub-data (LP 2); and performing subtraction operation on the original seismic data and the surface wave noise to generate the seismic data with the surface wave noise removed.

Specifically, the segmenting the raw seismic data into low frequency data and high frequency data based on a predetermined rule includes: segmenting the original seismic data into low-frequency data and high-frequency data according to a frequency domain segmentation rule, wherein the low-frequency data comprises the surface wave noise; partitioning the low frequency data into the first low frequency sub data (LP1) and the second low frequency sub data (LP2) based on a time-space domain partitioning rule; segmenting the high frequency data into the first high frequency sub-data (HP1) and the second high frequency sub-data (HP2) based on the time-space domain segmentation rule.

The performing the correction processing on the second high-frequency sub-data (HP2) according to the trained neural network and the filter to generate second high-frequency correction data includes: inputting the second high-frequency subdata (HP2) into a trained neural network for high-low frequency conversion processing to generate prediction data (PLP 2); inputting the prediction data (PLP2) to the filter to output the second high frequency correction data.

Preferably, the neural network is trained by: acquiring historical seismic data, the historical seismic data comprising: a history first low frequency sub data (LP1) and a history first high frequency sub data (HP 1); training the neural network using the historical first low frequency sub-data (LP1) and the historical first high frequency sub-data (HP1) as training data.

The historical seismic data further comprises: historical second high frequency correction data and historical second low frequency sub-data (LP2), the filter being set by: setting the filter according to a remaining error value after performing a subtraction operation on the historical second high frequency correction data and the historical second low frequency sub-data (LP 2).

According to a second aspect of the present invention, there is provided a surface wave noise removing apparatus for seismic data, the apparatus comprising: the data acquisition unit is used for acquiring original seismic data containing surface wave noise; a segmentation unit for segmenting the original seismic data into low frequency data and high frequency data based on a predetermined rule, wherein the low frequency data includes: first low frequency sub data (LP1) and second low frequency sub data (LP2), the high frequency data comprising: first high frequency sub-data (HP1) and second high frequency sub-data (HP2), the second low frequency sub-data including the surface wave noise; a correction unit configured to perform correction processing on the second high-frequency sub-data (HP2) according to the trained neural network and a preset filter to generate second high-frequency correction data, the correction processing including: high-low frequency conversion processing; and the noise removing unit is used for carrying out surface wave noise removing operation on the original seismic data according to the second high-frequency correction data and the second low-frequency subdata (LP2) so as to generate seismic data with surface wave noise removed.

The noise removing unit includes: a noise determination module for determining the surface wave noise according to the second high frequency correction data and the second low frequency sub-data (LP 2); and the noise removing module is used for performing subtraction operation on the original seismic data and the surface wave noise to generate the seismic data after the surface wave noise is removed.

The dividing unit includes: a frequency domain segmentation module, configured to segment the original seismic data into low-frequency data and high-frequency data according to a frequency domain segmentation rule, where the low-frequency data includes the surface wave noise; a low-frequency sub-data generating module, configured to divide the low-frequency data into the first low-frequency sub-data (LP1) and the second low-frequency sub-data (LP2) based on a time-space domain division rule; a high-frequency sub-data generating module, configured to divide the high-frequency data into the first high-frequency sub-data (HP1) and the second high-frequency sub-data (HP2) based on the time-space domain division rule.

The correction unit includes: a prediction data generation module, which is used for inputting the second high-frequency sub-data (HP2) into a trained neural network to perform high-frequency and low-frequency conversion processing so as to generate prediction data (PLP 2); a correction module for inputting said prediction data (PLP2) to said filter to output said second high frequency correction data.

Further, the above apparatus further comprises: and the neural network training unit is used for training the neural network.

The neural network training unit includes: a historical data acquisition module for acquiring historical seismic data, the historical seismic data comprising: a history first low frequency sub data (LP1) and a history first high frequency sub data (HP 1); a training module for training the neural network using the historical first low frequency sub-data (LP1) and the historical first high frequency sub-data (HP1) as training data.

The historical seismic data further comprises: historical second high frequency correction data and historical second low frequency sub-data (LP2), the apparatus further comprising: a filter setting unit for setting the filter.

The filter setting unit is specifically configured to: setting the filter according to a remaining error value after performing a subtraction operation on the historical second high frequency correction data and the historical second low frequency sub-data (LP 2).

According to a third aspect of the present invention, there is provided an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of the method when executing the program.

According to a fourth aspect of the invention, the invention provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of the above-mentioned method.

According to the technical scheme, based on the low-frequency band and low-speed characteristics of surface wave noise, the frequency domain segmentation processing and the time-space domain segmentation processing are carried out on original seismic data to obtain first low-frequency subdata (LP1) and second low-frequency subdata (LP2), first high-frequency subdata (HP1) and second high-frequency subdata (HP2), wherein only the LP2 contains the surface wave noise, and the rest three subdata do not contain the surface wave noise, then the HP2 is corrected based on a pre-trained neural network and a filter to obtain PLP2', then the surface wave noise of the original seismic data is removed according to the PLP2' and the second low-frequency subdata (LP2) to generate seismic data after the surface wave noise is removed, and as the training data of the neural network are historical first low-frequency subdata LP1 and historical first high-frequency subdata 1, compared with the prior art, the two subdata do not contain the surface wave noise, therefore, a more accurate neural network can be obtained through training, more accurate prediction data can be obtained, and surface wave noise in the original seismic data can be effectively removed.

Drawings

In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to these drawings without creative efforts.

FIG. 1 is a flow chart of a seismic data surface wave noise removal method according to an embodiment of the invention;

FIG. 2 is a schematic illustration of raw seismic data;

FIG. 3 is a schematic diagram of low frequency data (LP) according to an embodiment of the present invention;

FIG. 4 is a schematic diagram of high frequency data (HP) according to an embodiment of the present invention;

FIG. 5 is a schematic diagram of first low frequency sub-data (LP1) according to an embodiment of the present invention;

FIG. 6 is a schematic diagram of second low frequency sub-data (LP2) according to an embodiment of the present invention;

FIG. 7 is a schematic diagram of first high frequency sub-data (HP1) according to an embodiment of the present invention;

FIG. 8 is a schematic diagram of second high frequency sub-data (HP2) according to an embodiment of the present invention;

FIG. 9 is a detailed flow diagram of a seismic data surface wave noise removal method according to an embodiment of the invention;

FIG. 10 is a block diagram of the structure of a seismic-data surface-wave noise removing apparatus according to an embodiment of the present invention;

FIG. 11 is a block diagram showing the detailed structure of a seismic-data surface-wave noise removing apparatus according to an embodiment of the present invention;

fig. 12 is a block diagram of the structure of the neural network training unit 5 according to the embodiment of the present invention;

fig. 13 is a block diagram of the structure of the segmentation unit 2 according to an embodiment of the present invention;

fig. 14 is a block diagram of the structure of the correction unit 3 according to the embodiment of the present invention;

fig. 15 is a block diagram of the structure of the noise removing unit 4 according to the embodiment of the present invention;

fig. 16 is a schematic block diagram of a system configuration of an electronic apparatus 600 according to an embodiment of the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.

In the prior art, in the process of removing the surface wave noise in the seismic data based on the deep learning technology, because the conventional denoised data is adopted as a sample label in the current deep learning technology, and the conventional surface wave denoising method has technical defects, an accurate sample label cannot be provided, so that the deep learning technology is difficult to exceed the conventional denoising method in the surface wave denoising field, and the surface wave in the seismic data cannot be well removed. Based on the scheme, the method and the device for removing the surface wave noise of the seismic data can provide accurate samples, so that the surface waves in the seismic data can be removed more effectively. The following describes an embodiment of the present invention in detail based on the above-ground seismic data as an example, with reference to the drawings.

Fig. 1 is a flowchart of a seismic data surface wave noise removal method according to an embodiment of the present invention, as shown in fig. 1, the method including:

step 101, acquiring original seismic data containing surface wave noise, where the original seismic data is shown in fig. 2, and an abscissa is a channel Number (Trace Number) and an ordinate is Time (Time).

Step 102, segmenting the original seismic data into low-frequency data and high-frequency data based on a predetermined rule, wherein the low-frequency data comprises: first low frequency sub data (LP1) and second low frequency sub data (LP2), the high frequency data comprising: first high frequency sub-data (HP1) and second high frequency sub-data (HP 2). The predetermined rule here is a frequency domain division rule and a time-space domain division rule.

Since the surface wave noise exists in the low frequency band and the high frequency band does not contain the surface wave noise, the original seismic data may be divided into the low frequency data (LP) shown in fig. 3 and the high frequency data (HP) shown in fig. 4 according to the frequency domain division rule, so that the low frequency data includes the surface wave noise and the effective information, and the high frequency data does not contain the surface wave noise and only includes the effective information.

In one embodiment, because the surface wave exists only in the low frequency portion of the data, a low pass filter may be constructed with a high cutoff frequency greater than the maximum effective frequency of the surface wave noise. The filter is used to input the original seismic data and filter the data to obtain data containing surface wave noise (LP), and the original seismic data is subtracted from the data containing surface wave to obtain data not containing surface wave (HP). Similarly, a wavelet transform method may be used, where the large scale portion contains surface wave noise and the small scale portion does not contain surface wave noise.

Then, according to the characteristic that the surface wave noise has low speed, the low-frequency data is divided into first low-frequency sub-data (LP1) shown in fig. 5 and second low-frequency sub-data (LP2) shown in fig. 6 based on a time-space domain division rule, wherein the second low-frequency sub-data comprises the surface wave noise. Meanwhile, the high frequency data is divided into first high frequency sub-data (HP1) shown in fig. 7 and second high frequency sub-data (HP2) shown in fig. 8 based on the time-space domain division rule. The HP1 corresponds to the LP1, and the HP2 corresponds to the LP 2.

In one embodiment, given an apparent velocity greater than the maximum apparent velocity of the surface waves in the input seismic data, the input seismic data is decomposed into two parts at the given apparent velocity, based on the apparent velocity of the surface wave noise being low, where apparent velocity refers to the value of the ground observation shot distance divided by time, data with apparent velocity greater than the given apparent velocity (e.g., LP1) and data with apparent velocity less than the given apparent velocity (e.g., LP 2).

103, performing correction processing on the second high-frequency sub-data (HP2) according to the trained neural network and a preset filter to generate second high-frequency correction data (corrected PLP2), wherein the correction processing comprises: and (5) high-low frequency conversion processing.

In actual operation, the neural network may be a multi-layer neural network, such as a multi-layer fully-connected Convolutional Neural Network (CNN). The neural network model achieves the purpose of predicting low-frequency band signals from high-frequency band signals, namely, the neural network model is used for achieving high-frequency and low-frequency conversion processing.

In one embodiment, the neural network may be trained from historical seismic data. The training process specifically comprises: acquiring historical seismic data, the historical seismic data comprising: a history first low frequency sub data (LP1) and a history first high frequency sub data (HP 1); the historical first low-frequency sub-data LP1 and the historical first high-frequency sub-data HP1 are then used as training data to train the neural network.

Here, the historical first low-frequency sub-data (LP1) and the historical first high-frequency sub-data (HP1) are used as training samples, so that the prediction result obtained by inputting HP1 data is minimally different from LP1 data, and thus a deep learning neural network model is trained to realize conversion of high-frequency data into low-frequency data.

The historical first low-frequency subdata (LP1) and the historical first high-frequency subdata (HP1) serving as the training samples are obtained based on a time-space domain segmentation rule, and compared with a sample label (namely training data) in the prior art, the two subdata do not contain surface wave noise and only contain effective information, so that the two subdata are more accurate than the training data in the prior art, a more accurate neural network can be trained, and the surface wave noise in seismic data can be better removed.

Then, the second high-frequency sub-data (HP2) is input to the trained neural network, and prediction data (PLP2) is generated, where the PLP2 is low-frequency data and can be considered as data LP2 without a surface wave portion.

In the correction process, the second high-frequency subdata (HP2) is input into the trained neural network to generate prediction data (PLP 2); the predicted data (PLP2) is then input to the set filter, and second high frequency correction data (PLP2') can be output.

And 104, performing surface wave noise removing operation on the original seismic data according to the second high-frequency correction data (PLP2') and the second low-frequency subdata (LP2) to generate seismic data with surface wave noise removed.

Specifically, the surface wave noise is determined according to the second high-frequency correction data (PLP2') and the second low-frequency sub-data LP 2; and then, performing subtraction operation on the original seismic data and the surface wave noise to generate the seismic data with the surface wave noise removed.

In one embodiment, the difference between the second high frequency correction data (PLP2') and the second low frequency sub data LP2 is the final predicted surface wave noise (N) (including the surface wave and its scattering noise). And subtracting the obtained noise N from the original seismic data to obtain the seismic data with the final surface wave removed.

The method comprises the steps of carrying out frequency domain segmentation processing and time-space domain segmentation processing on original seismic data based on low-frequency band and low-speed characteristics of surface wave noise to obtain first low-frequency subdata (LP1) and second low-frequency subdata (LP2), first high-frequency subdata (HP1) and second high-frequency subdata (HP2), wherein only the LP2 contains the surface wave noise, and the rest three subdata do not contain the surface wave noise, then carrying out correction on HP2 based on a pre-trained neural network and a filter to obtain PLP2', and then carrying out surface wave noise removal operation on the original seismic data according to PLP2' and the second low-frequency subdata (LP2) to generate seismic data after the surface wave noise is removed, wherein the training data of the neural network are historical first low-frequency subdata LP1 and historical first high-frequency subdata HP 35 1, and compared with the prior art, the two subdata do not contain the surface wave noise, therefore, a more accurate neural network can be obtained through training, more accurate prediction data can be obtained, and surface wave noise in the original seismic data can be effectively removed.

In a specific implementation process, the filter corrects the low-frequency data PLP2 output by the neural network so that the residual error after subtraction between LP2 and the corrected PLP2 (i.e., PLP2') is the minimum, and the difference between LP2 and PLP2' is the final predicted surface wave noise (N).

The filter may be preset based on historical seismic data used to train the neural network, and the historical seismic data may further include: historical second high frequency correction data (which may be obtained through a trained neural network) and historical second low frequency sub-data (LP 2). Specifically, the filter may be set according to a residual error value after the subtraction operation is performed on the historical second high-frequency correction data and the historical second low-frequency sub-data (LP 2).

For example, the filter may be set by the following formula:

wherein w represents a convolution matrix composed of one-dimensional filter operators, | | | | | non-calculation₁Representing the L1 norm.

It should be noted that all references to seismic data herein may be considered as matrices.

Based on the above equation, a filter w is constructed or set by solving an optimization problem that minimizes the error under the L1 norm after LP2 is subtracted from corrected PLP 2.

In actual practice, the filter w may also be set using the norm of L2.

Fig. 9 is a detailed flowchart of a seismic data surface wave noise removing method according to an embodiment of the present invention, and as shown in fig. 9, the flowchart includes:

step 901, based on the fact that the surface wave exists in the low frequency band and the surface wave is not contained in the high frequency band, a frequency domain segmentation module is designed to decompose the original seismic data (D) into a surface wave containing part (LP) and a surface wave free part (HP), wherein the surface wave containing data not only contains effective signals but also contains surface wave noise.

Step 902, according to the low-speed characteristic of the surface wave, designing a time-space domain segmentation module to divide the data LP into two parts: a non-surface wave containing part (LP1) and a surface wave containing part (LP 2). The data HP is divided into two parts, namely HP1 and HP2, by using the same time-space domain division module, wherein HP1 corresponds to LP1, and HP2 corresponds to LP 2.

Step 903, deep learning neural network training is carried out, a multilayer neural network is built, and the purpose that low-frequency signals can be predicted from high-frequency signals can be achieved through the deep learning neural network model. The data LP1 and HP1 are used as training samples, and a deep learning neural network model is trained, so that the prediction result obtained by inputting HP1 data is the least different from the LP1 data.

And 904, applying the deep learning neural network model obtained in 903, inputting the data HP2 into the deep learning neural network to obtain a prediction result PLP2, wherein the predicted PLP2 can be regarded as a part without surface waves in the data LP 2.

Step 905, design a filter based on the following formula, correct PLP2, and minimize the residual error after subtraction between LP2 and corrected PLP2(PLP2'), where the difference between the two is the final predicted surface wave and its scattering noise (N).

The filter w is constructed by solving the optimization problem described above to minimize the error at the L1 norm after subtraction of LP2 from the corrected PLP 2.

Step 906, subtracting the noise N predicted in step 905 from the original seismic data D to obtain the final seismic data from which the surface waves are removed.

According to the method and the device, based on the characteristics of low surface wave frequency and low apparent velocity, the frequency domain and the time-space domain of the original seismic data are segmented, accurate label data are manufactured by utilizing the body wave data outside the surface wave region, the problems of insufficient deep learning samples and low accuracy are solved, the purpose of removing the surface waves and the scattering noise thereof in the seismic data in a fidelity manner is achieved, and the signal-to-noise ratio of the seismic data is improved.

Based on similar inventive concepts, the embodiment of the present invention further provides a device for removing surface wave noise of seismic data, and preferably, the device can be used for implementing the processes in the above method embodiments.

Fig. 10 is a block diagram showing the structure of a seismic-data surface wave noise removing apparatus according to an embodiment of the present invention, as shown in fig. 10, the apparatus including: a data acquisition unit 1, a segmentation unit 2, a correction unit 3, and a noise removal unit 4, wherein:

a data acquisition unit 1 for acquiring original seismic data containing surface wave noise;

a segmentation unit 2 configured to segment the original seismic data into low-frequency data and high-frequency data based on a predetermined rule, wherein the low-frequency data includes: first low-frequency sub data LP1 and second low-frequency sub data LP2, the high-frequency data including: the first high-frequency sub data HP1 and the second high-frequency sub data HP2 are provided, and the second low-frequency sub data comprise the surface wave noise;

a correcting unit 3, configured to perform correction processing on the second high-frequency sub-data HP2 according to the trained neural network and a preset filter to generate second high-frequency correction data (corrected PLP2), where the correction processing includes: high-low frequency conversion processing;

and the noise removing unit 4 is configured to perform surface wave noise removing operation on the original seismic data according to the second high-frequency correction data and the second low-frequency sub-data LP2, so as to generate seismic data with surface wave noise removed.

Based on the low-frequency band and low-speed characteristics of surface wave noise, the dividing unit 2 performs frequency domain division processing and time-space domain division processing on original seismic data acquired by the data acquisition unit 1 to obtain first low-frequency sub-data (LP1) and second low-frequency sub-data (LP2), first high-frequency sub-data (HP1) and second high-frequency sub-data (HP2), wherein only LP2 contains surface wave noise, and the rest three sub-data do not contain surface wave noise, then the correction unit 3 corrects HP2 based on a pre-trained neural network and a filter to obtain PLP2', then the noise removal unit 4 performs surface wave noise removal operation on the original seismic data according to PLP2' and the second low-frequency sub-data (LP2) to generate seismic data after surface wave noise removal, and as training data of the neural network are historical first low-frequency sub-data LP1 and historical first high-frequency sub-data HP1, compared with the prior art, the two subdata do not contain surface wave noise, so that a more accurate neural network can be obtained through training, more accurate prediction data can be obtained, and the surface wave noise in the original seismic data can be removed more effectively.

In practical operation, as shown in fig. 11, the above apparatus further comprises: and the neural network training unit 5 is used for training the neural network.

As shown in fig. 12, the neural network training unit 5 includes: a historical data acquisition module 51 and a training module 52, wherein:

a historical data acquisition module 51 configured to acquire historical seismic data, where the historical seismic data includes: the history first low-frequency sub data LP1 and the history first high-frequency sub data HP 1;

a training module 52, configured to train the neural network by using the historical first low-frequency sub-data LP1 and the historical first high-frequency sub-data HP1 as training data.

In actual practice, the historical seismic data further includes: the historical second high-frequency correction data and the historical second low-frequency sub-data LP 2.

With continued reference to fig. 11, the apparatus further comprises: a filter setting unit 6 for setting the filter. Specifically, the filter setting unit 6 is specifically configured to: the filter is set according to the remaining error value after the subtraction operation is performed on the historical second high-frequency correction data and the historical second low-frequency sub-data LP 2.

Specifically, as shown in fig. 13, the above-described dividing unit 2 includes: a frequency domain division module 21, a low frequency subdata generation module 22 and a high frequency subdata generation module 23, wherein:

a frequency domain segmentation module 21, configured to segment the original seismic data into low-frequency data and high-frequency data according to a frequency domain segmentation rule, where the low-frequency data includes the surface wave noise;

the low-frequency sub-data generating module 22 is configured to divide the low-frequency data into the first low-frequency sub-data LP1 and the second low-frequency sub-data LP2 based on a time-space domain division rule;

the high-frequency sub-data generating module 23 is configured to divide the high-frequency data into the first high-frequency sub-data HP1 and the second high-frequency sub-data HP2 based on the time-space domain division rule.

As shown in fig. 14, the correction unit 3 includes: a prediction data generation module 31 and a correction module 32, wherein:

the prediction data generation module 31 is used for inputting the second high-frequency sub-data HP2 to a trained neural network for high-frequency and low-frequency conversion processing so as to generate prediction data PLP 2;

a correction module 32 for inputting said prediction data PLP2 to said filter to output said second high frequency correction data.

Specifically, as shown in fig. 15, the noise removing unit 4 includes: a noise determination module 41 and a noise removal module 42, wherein:

a noise determining module 41, configured to determine the surface wave noise according to the second high-frequency correction data and the second low-frequency sub-data LP 2;

and a noise removing module 42, configured to perform a subtraction operation on the original seismic data and the surface wave noise to generate the seismic data after the surface wave noise is removed.

For specific execution processes of the units and the modules, reference may be made to the description in the foregoing method embodiments, and details are not described here again.

In practical operation, the units and the modules may be combined or may be singly arranged, and the present invention is not limited thereto.

The present embodiment also provides an electronic device, which may be a desktop computer, a tablet computer, a mobile terminal, and the like, but is not limited thereto. In this embodiment, the electronic device may be implemented by referring to the above method embodiment and the embodiment of the seismic data surface wave noise removing device, and the contents thereof are incorporated herein, and repeated details are not repeated.

Fig. 16 is a schematic block diagram of a system configuration of an electronic apparatus 600 according to an embodiment of the present invention. As shown in fig. 16, the electronic device 600 may include a central processor 100 and a memory 140; the memory 140 is coupled to the central processor 100. Notably, this diagram is exemplary; other types of structures may also be used in addition to or in place of the structure to implement telecommunications or other functions.

In one embodiment, the seismic data surface wave noise removal function may be integrated into the central processor 100. The central processor 100 may be configured to control as follows:

step 101, acquiring original seismic data containing surface wave noise;

step 102, segmenting the original seismic data into low-frequency data and high-frequency data based on a predetermined rule, wherein the low-frequency data comprises: first low-frequency sub data LP1 and second low-frequency sub data LP2, the high-frequency data including: the first high-frequency sub data HP1 and the second high-frequency sub data HP2 are provided, and the second low-frequency sub data comprise the surface wave noise;

103, performing correction processing on the second high-frequency sub-data HP2 according to the trained neural network and a preset filter to generate second high-frequency correction data (corrected PLP2), where the correction processing includes: high-low frequency conversion processing;

and 104, performing surface wave noise removing operation on the original seismic data according to the second high-frequency correction data and the second low-frequency sub-data LP2 to generate seismic data with surface wave noise removed.

As can be seen from the above description, the electronic device provided in the embodiment of the present application can perform frequency domain segmentation processing and time-space domain segmentation processing on the original seismic data based on the low-frequency band and low-speed characteristics of the surface wave noise to obtain first low-frequency sub-data (LP1) and second low-frequency sub-data (LP2), first high-frequency sub-data (HP1) and second high-frequency sub-data (HP2), where only LP2 contains the surface wave noise, and none of the remaining three sub-data contains the surface wave noise, and then correct HP2 based on a pre-trained neural network and a filter to obtain PLP2', and then perform surface wave noise removal operation on the original seismic data according to PLP2' and the second low-frequency sub-data (LP2) to generate seismic data after the surface wave noise is removed, and since the training data of the neural network are historical first low-frequency sub-data LP1 and historical first high-frequency sub-data HP1, compared with the prior art, the two subdata do not contain surface wave noise, so that a more accurate neural network can be obtained through training, more accurate prediction data can be obtained, and the surface wave noise in the original seismic data can be removed more effectively.

In another embodiment, the seismic data surface wave noise removing device may be configured separately from the central processor 100, for example, the seismic data surface wave noise removing device may be configured as a chip connected to the central processor 100, and the seismic data surface wave noise removing device is implemented by the control of the central processor.

As shown in fig. 16, the electronic device 600 may further include: communication module 110, input unit 120, audio processing unit 130, display 160, power supply 170. It is noted that the electronic device 600 does not necessarily include all of the components shown in FIG. 6; furthermore, the electronic device 600 may also include components not shown in fig. 16, which may be referred to in the prior art.

As shown in fig. 16, the central processor 100, sometimes referred to as a controller or operational control, may include a microprocessor or other processor device and/or logic device, the central processor 100 receiving input and controlling the operation of the various components of the electronic device 600.

The memory 140 may be, for example, one or more of a buffer, a flash memory, a hard drive, a removable media, a volatile memory, a non-volatile memory, or other suitable device. The information relating to the failure may be stored, and a program for executing the information may be stored. And the central processing unit 100 may execute the program stored in the memory 140 to realize information storage or processing, etc.

The input unit 120 provides input to the cpu 100. The input unit 120 is, for example, a key or a touch input device. The power supply 170 is used to provide power to the electronic device 600. The display 160 is used to display an object to be displayed, such as an image or a character. The display may be, for example, an LCD display, but is not limited thereto.

The memory 140 may be a solid state memory such as Read Only Memory (ROM), Random Access Memory (RAM), a SIM card, or the like. There may also be a memory that holds information even when power is off, can be selectively erased, and is provided with more data, an example of which is sometimes called an EPROM or the like. The memory 140 may also be some other type of device. Memory 140 includes buffer memory 141 (sometimes referred to as a buffer). The memory 140 may include an application/function storage section 142, and the application/function storage section 142 is used to store application programs and function programs or a flow for executing the operation of the electronic device 600 by the central processing unit 100.

The memory 140 may also include a data store 143, the data store 143 for storing data, such as contacts, digital data, pictures, sounds, and/or any other data used by the electronic device. The driver storage portion 144 of the memory 140 may include various drivers of the electronic device for communication functions and/or for performing other functions of the electronic device (e.g., messaging application, address book application, etc.).

The communication module 110 is a transmitter/receiver 110 that transmits and receives signals via an antenna 111. The communication module (transmitter/receiver) 110 is coupled to the central processor 100 to provide an input signal and receive an output signal, which may be the same as in the case of a conventional mobile communication terminal.

Based on different communication technologies, a plurality of communication modules 110, such as a cellular network module, a bluetooth module, and/or a wireless local area network module, may be provided in the same electronic device. The communication module (transmitter/receiver) 110 is also coupled to a speaker 131 and a microphone 132 via an audio processor 130 to provide audio output via the speaker 131 and receive audio input from the microphone 132 to implement general telecommunications functions. Audio processor 130 may include any suitable buffers, decoders, amplifiers and so forth. In addition, an audio processor 130 is also coupled to the central processor 100, so that recording on the local can be enabled through a microphone 132, and so that sound stored on the local can be played through a speaker 131.

Embodiments of the present invention also provide a computer-readable storage medium, on which a computer program is stored, where the computer program is executed by a processor to implement the steps of the above-mentioned seismic data surface wave noise removing method.

In summary, the embodiment of the invention divides the original seismic data into the frequency domain and the time-space domain based on the characteristics of low surface wave frequency and low apparent velocity, and utilizes the body wave data outside the surface wave region to make accurate tag data, thereby solving the problems of insufficient deep learning samples and low accuracy, achieving the purpose of removing the surface wave and the scattering noise thereof in the seismic data in a fidelity way, and improving the signal-to-noise ratio of the seismic data.

The preferred embodiments of the present invention have been described above with reference to the accompanying drawings. The many features and advantages of the embodiments are apparent from the detailed specification, and thus, it is intended by the appended claims to cover all such features and advantages of the embodiments which fall within the true spirit and scope thereof. Further, since numerous modifications and changes will readily occur to those skilled in the art, it is not desired to limit the embodiments of the invention to the exact construction and operation illustrated and described, and accordingly, all suitable modifications and equivalents may be resorted to, falling within the scope thereof.

As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The principle and the implementation mode of the invention are explained by applying specific embodiments in the invention, and the description of the embodiments is only used for helping to understand the method and the core idea of the invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, there may be variations in the specific embodiments and the application scope, and in summary, the content of the present specification should not be construed as a limitation to the present invention.