CN115097524A

CN115097524A - Mixed wave interference suppression method and device, storage medium and electronic equipment

Info

Publication number: CN115097524A
Application number: CN202210714894.5A
Authority: CN
Inventors: 路交通; 张庆淮; 于富文; 王国来; 王明星; 曾宁; 孟祥顺; 吴安楚; 巩长春; 张丽
Original assignee: Sinopec Oilfield Service Corp; Sinopec Petroleum Engineering Geophysics Co Ltd
Current assignee: Sinopec Oilfield Service Corp; Sinopec Petroleum Engineering Geophysics Co Ltd
Priority date: 2022-06-22
Filing date: 2022-06-22
Publication date: 2022-09-23

Abstract

The present application discloses a method for suppressing mixed-wave interference. The method includes: acquiring target mixed-wave interference seismic data; performing data domain conversion on the target mixed-wave interference seismic data to obtain converted seismic data; training a neural network model according to the converted seismic data to obtain a trained neural network model. A network model; performing mixing suppression processing on the converted seismic data through the trained neural network model, and obtaining the seismic data after the mixing of the target mixed-wave interference seismic data is removed. There is no need to synthesize a large number of mixed-wave sample seismic data, directly start from the mixed-wave seismic data itself, conduct unsupervised training of the network model according to the mixed-wave seismic data to be processed, and then use the trained network model directly for the mixed-wave seismic data. Wave interference processing has wider applicability, can more effectively protect seismic data, avoid the loss of effective seismic data, will not cause structural artifacts, and also improve the signal-to-noise ratio.

Description

Mixed wave interference suppression method and device, storage medium and electronic equipment

Technical Field

The present application relates to the field of seismic data processing technologies, and in particular, to a method and an apparatus for suppressing mixed wave interference, a computer-readable storage medium, and an electronic device.

Background

The high-efficiency seismic acquisition technology is more and more widely applied to land or offshore seismic exploration, and because the technology enables adjacent excitation shots to be excited simultaneously or the excitation time interval is very short, the acquisition efficiency can be greatly improved, and the density of the seismic acquisition shots-demodulator probes can also be greatly improved.

The high-efficiency seismic acquisition technology greatly reduces the high-density exploration cost, so that the method is the preferred acquisition technology for large-scale seismic exploration projects. However, the high-efficiency seismic acquisition technology such as the dynamic sliding scanning technology, the independent synchronous scanning technology, the ultra-high-efficiency acquisition technology and the like can introduce mixed wave interference to different degrees, so that the quality of seismic data of adjacent cannons is greatly influenced, and the mixed wave noise interference needs to be removed through mixed wave suppression.

The conventional mixed wave interference suppression method is based on the inversion reconstruction technology of sparse transformation to realize the separation of mixed wave and effective reflection seismic information, the construction false image is easy to appear, and the signal-to-noise ratio is relatively low to improve.

In recent two years, with the popularization of deep learning artificial intelligence algorithms, some researchers begin to utilize a deep learning method to attenuate and suppress the mixed wave noise of seismic data, and a certain processing effect is achieved.

However, the existing deep neural network-based mixed wave suppression technology is mainly characterized in that a network model needs to be supervised and trained, namely a large amount of clean seismic data and mixed wave interference-containing seismic data need to be synthesized to serve as labels to train the model, the mixed wave suppression effect in the synthesized seismic data is probably better, but the mixed wave suppression effect is not good due to the large difference of seismic data samples in the actual seismic data processing, and the adaptability of the method is restrained to a certain degree.

Disclosure of Invention

In order to solve the problems, the application provides a method and a device for suppressing mixed wave interference, a storage medium and electronic equipment, which realize the autonomous suppression of the mixed wave noise interference in high-efficiency seismic acquisition by using an unsupervised deep learning technology, and realize intelligent mixed wave suppression by driving a network model to update hyper-parameters from seismic data.

In a first aspect of the present application, a method for suppressing mixed wave interference is provided, the method including:

acquiring target mixed wave interference seismic data;

performing data domain conversion on the target mixed wave interference seismic data to obtain converted seismic data;

training a neural network model according to the converted seismic data to obtain a trained neural network model;

and performing mixed wave suppression processing on the converted seismic data through the trained neural network model to obtain the seismic data of the target mixed wave interference seismic data after the mixed wave is removed.

In some embodiments, the training a neural network model according to the converted seismic data to obtain a trained neural network model includes:

cutting and dividing the converted seismic data into blocks to obtain a block data set;

carrying out augmentation and expansion processing on the tile data set to obtain an expanded tile data set;

acquiring training data according to the expanded patch data set;

and training the neural network model according to the training data to obtain the trained neural network model.

In some embodiments, the training data comprises model input data, the obtaining training data from the augmented patch data set comprises:

respectively performing mask processing on the tile data in the expanded tile data set based on a preset mask strategy to obtain a masked tile data set;

and taking the masked tile data set as the model input data.

In some embodiments, the training data further includes model tag data, the obtaining training data from the augmented patch data set further includes:

and taking the extended patch data set as the model tag data.

In some embodiments, the preset masking policy includes: at least one of a dotted mask policy, a blocky mask policy, and a striped mask policy, where the masking processing is performed on the tile data in the expanded tile data set based on a preset mask policy, respectively, and includes:

acquiring the signal-to-noise ratio of the block data;

under the condition that the signal-to-noise ratio of the tile data is greater than a first preset threshold value, performing mask processing on tiles in the expanded tile data set based on a dot-shaped mask strategy;

under the condition that the signal-to-noise ratio of the tile data is smaller than a second preset threshold value, performing mask processing on tiles in the expanded tile data set based on a striped mask strategy;

and under the condition that the signal-to-noise ratio of the tile data is not greater than a first preset threshold and not less than a second preset threshold, performing mask processing on the tiles in the expanded tile data set based on a block mask strategy.

In some embodiments, the data domain conversion comprises:

any one of co-detection point domain conversion, co-offset domain conversion, and cross domain conversion.

In some embodiments, the training the neural network model according to the training data to obtain the trained neural network model includes:

setting initial network parameters of the neural network model;

training the neural network model according to model input data and model label data of the training data to obtain optimized network parameters;

and taking the optimized network parameters as the network parameters of the neural network model to obtain the trained neural network model.

In a second aspect of the present application, there is provided a mixed wave interference suppression apparatus, the apparatus including:

the acquisition module is used for acquiring target mixed wave interference seismic data;

the conversion module is used for carrying out data domain conversion on the target mixed wave interference seismic data to obtain converted seismic data;

the training module is used for training the neural network model according to the converted seismic data to obtain a trained neural network model;

and the interference suppression module is used for performing mixed wave suppression processing on the converted seismic data through the trained neural network model to obtain the seismic data of the target mixed wave interference seismic data after the mixed waves are removed.

In a third aspect of the present application, a computer-readable storage medium is provided, where the computer-readable storage medium stores a computer program, which is executable by one or more processors, to implement the method for suppressing mixed wave interference as described above.

In a fourth aspect of the present application, an electronic device is provided, which includes a memory and one or more processors, the memory stores a computer program, the memory and the one or more processors are communicatively connected to each other, and the computer program, when executed by the one or more processors, implements the method for suppressing the mixed wave interference as described above.

Compared with the prior art, the technical scheme of the application has the following advantages or beneficial effects:

(1) the method has the advantages that a large amount of seismic data containing mixed wave samples are not required to be synthesized, an unsupervised training network model is directly carried out according to the mixed wave seismic data to be processed, and then the trained network model is directly used for mixed wave interference processing of the mixed wave seismic data, so that the method has the characteristic of wider applicability compared with a traditional sparse transform mixed wave pressing method;

(2) compared with the traditional sparse transform mixed wave suppression method, the method and the device have the advantages that the seismic data can be more effectively protected, the loss of effective seismic data information is avoided, the construction false image cannot be caused, and meanwhile, the signal to noise ratio is also improved.

Drawings

In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the following description is briefly made with respect to the drawings required to be used in the embodiments or the prior art, and it is obvious that the drawings in the following description are only embodiments of the present application, and for those skilled in the art, other drawings can be obtained according to the provided drawings without creative efforts.

Fig. 1 is a flowchart of a method for suppressing mixed wave interference according to an embodiment of the present disclosure;

fig. 2 is a diagram of a network structure model of the Unet according to an embodiment of the present application;

fig. 3 is a schematic diagram illustrating a suppression effect of mixed wave interference according to an embodiment of the present disclosure;

fig. 4 is a schematic structural diagram of a mixed wave interference suppression device according to an embodiment of the present application;

fig. 5 is a connection block diagram of an electronic device according to an embodiment of the present application.

Detailed Description

The following detailed description will be provided with reference to the accompanying drawings and embodiments, so that how to apply the technical means to solve the technical problems and achieve the corresponding technical effects can be fully understood and implemented. The embodiments and various features in the embodiments of the present application can be combined with each other on the premise of no conflict, and the formed technical solutions are all within the protection scope of the present application.

As can be known from the background art, the high-efficiency seismic acquisition technology is more and more widely applied to onshore or offshore seismic exploration, and because the technology enables adjacent excitation shots to be excited simultaneously or the excitation time interval to be very short, the acquisition efficiency can be greatly improved, and the density of the seismic acquisition shots and the detection shots can be greatly improved.

In view of the above, the application discloses a mixed wave interference suppression method, a mixed wave interference suppression device, a storage medium and electronic equipment, mixed wave noise interference is removed through an unsupervised autonomous network model, a large amount of mixed wave sample seismic data does not need to be synthesized, the mixed wave seismic data is directly started from the mixed wave seismic data, the trained model is directly used for mixed wave seismic data processing through an unsupervised network autonomous seismic data characteristic structure, and the applicability is wider.

Compared with the traditional sparse transform mixed wave suppression method, the method has the advantages that the seismic data can be more effectively protected, the loss of effective seismic information is avoided, the construction false image cannot be caused, and the signal-to-noise ratio is relatively improved.

Example one

Fig. 1 is a flowchart of a mixed wave interference suppression method according to an embodiment of the present application, and as shown in fig. 1, the method according to the present embodiment includes:

and S110, acquiring target mixed wave interference seismic data.

Optionally, target mixed wave interference seismic data are obtained from seismic data of the target mining area.

And S120, performing data domain conversion on the target mixed wave interference seismic data to obtain converted seismic data.

In some embodiments, the data domain conversion comprises:

Optionally, the target mixed wave interference seismic data is converted from the common shot point domain to other data domains such as a common demodulation point domain, a common offset domain, a cross domain and the like, so as to obtain converted seismic data, and mixed wave coherent noise interference on effective reflection seismic information is represented as strong energy random noise interference in other data domains such as the common demodulation point domain, and characteristic difference occurs between the interference and an effective reflection phase axis, so that the interference can be used in subsequent processing.

S130, training the neural network model according to the converted seismic data to obtain the trained neural network model.

A U-shaped network structure such as Unet or Unet + + built based on a coding layer Encoder and a decoding layer Decoder can be adopted, fig. 2 is a Unet network structure model diagram provided by the embodiment of the application, for example, a Unet network structure model shown in fig. 2, a slice feature structure is extracted through a multilayer deep network coding layer Encoder in the Unet network structure model, and a compressed low-dimensional seismic data feature structure is obtained along with the increase of the number of layers and the continuous depth of a network; and then, continuously reconstructing and recovering the high-dimensional seismic data through a Decoder layer Decoder to obtain the reconstructed and recovered seismic data.

In order to improve the suppression effect of mixed wave noise and simultaneously keep effective seismic signals, the coding layer and the decoding layer are connected by jump connection, and effective seismic information extracted from the coding layer at the same height is directly spliced into the decoding layer at the same layer, so that the high-order characteristics of a depth network are not lost in the processes of down-sampling and up-sampling, and the effective information loss caused in the process of down-sampling can be recovered.

In order to improve the generalization capability of a network architecture and increase the diversity of a network model, the neurons in a multilayer network are randomly inactivated by random abandonment, so that the network model can effectively keep seismic data characteristics and abandon noise characteristics.

acquiring training data according to the expanded patch data set;

Optionally, the tile cutting and dividing process includes the following steps:

setting the size of the tile as M × N, then sliding the tile on the seismic data from top to bottom and from left to right, and setting a sliding step length comprising a step length x _ step along the space direction and a step length t _ step along the time direction, wherein the size of the tile and the sliding step lengths in different directions are actually adjusted according to the size of the seismic data section. If the seismic data to be processed are large, the size of the slice and the sliding step length can be properly increased, and when the size of the seismic data is not large enough for slicing, zero filling processing needs to be carried out on the seismic data, so that the seismic data can be completely sliced and divided. Through the steps, a mixed wave interference seismic data section is changed into thousands of pieces of data with the same size, and a piece data set B is formed.

Optionally, the augmentation and expansion process includes the following steps:

the unsupervised mixed wave interference suppression technology is adopted, a large amount of seismic data and tag data do not need to be synthesized, and the block seismic data set obtained in the steps needs to be subjected to amplification and expansion processing for expanding the seismic data set. The seismic data augmentation method includes polarity inversion of the seismic data, rotation at different angles (for example, rotation by 90 degrees, 180 degrees, and 270 degrees counterclockwise), up-down or left-right inversion, and the like. Meanwhile, in order to further enhance the data diversity, the size of the seismic data can be enlarged by using a mirror image inversion mode, for example, the size of 32 × 32 seismic data is mirror-expanded into 56 × 56, and sliding cutting is performed on the basis of the seismic data to obtain different types of 32 × 32 seismic data. By the aid of the amplification and expansion processing method, seismic data can be expanded by several times or even more than ten times to obtain a new block data set B', due to the fact that diversity of sample data is increased, a deep neural network cannot be trapped in a local extreme value trap at a later stage, and the problem of overfitting of a network model is avoided.

In some embodiments, the training data comprises model input data, the obtaining training data from the augmented patch data set comprising:

and taking the masked tile data set as the model input data.

and taking the extended patch data set as the model tag data.

In some embodiments, the preset masking policy includes: at least one of a dotted mask policy, a block mask policy, and a striped mask policy, wherein the masking processing of the tile data in the expanded tile data set based on a preset mask policy includes:

acquiring the signal-to-noise ratio of the block data;

Optionally, after completing the seismic sample data set expansion, performing mask processing, where the mask processing includes the following steps:

randomly extracting one of the tiles b from the augmented tile data set _i Wherein i ═ 1, 2.. and N, N are the number of patch sample sets, and the patch is subjected to masking policy processing to obtain masked patch b' _i As input to the deep network, original tile b _i And as expected output of the deep network, training network model parameters.

Optionally, the signal-to-noise ratio of the block data may be obtained by a mean variance analysis method, specifically, the block data b containing the noise of the mixed wave seismic data _i Pre-evaluating the noise strength, firstly obtaining the block data b _i Then obtaining the block data b by the following formula _i Signal to noise ratio of (c):

mse＝10*(b _i _mean/b _i _sd)

wherein, b _i Mean is chunk data b _i Mean value of b _i Sd is chunk data b _i Standard deviation of (1), mse is tile data b _i The signal-to-noise ratio result value of (c).

Optionally, the mask policy includes three ways shown below, and may be flexibly selected according to factors such as actual demand or intensity degree and distribution characteristics of noise:

(1) dotted mask policy

For a certain tile b _i And performing dot mask processing, randomly selecting a data point p _ point of a certain proportion of p _ percent in the block, then setting a mask radius m _ r within a certain range, and randomly selecting other data point values to replace the point value within the set radius range around the p _ point.

The randomly drawn data point p _ point may also be zeroed out or filled in with random noise.

It should be noted that the random point extraction ratio p _ percentage is selected from the range of a few tenths of a percent to tens of percents according to the characteristics of the seismic data.

(2) Block mask policy

For a certain tile b _i Performing block mask processing, and randomSelecting a data block b _ block with a certain proportion b _ percent in the blocks, then setting a mask radius m _ r in a certain range, and randomly selecting other data point values around the b _ block in the set radius range to replace the values in the data block.

The seismic data in the randomly extracted data block b _ block may also be zeroed out or filled in with random noise.

It should be noted that the random block extraction ratio b _ percentage can be selected from a range of a few tenths of a percent to a few tens of percent according to the characteristics of the actual seismic data.

(3) Striped mask policy

For a certain tile b _i Masking processing is carried out, strip-shaped seismic channels s _ trace with a certain proportion of s _ percent in a patch are randomly selected in the space direction, then mask radius m _ r in a certain range is set, and other strip-shaped seismic channels are randomly selected in the set radius range around s _ trace to replace the seismic channels.

The randomly extracted strip-shaped seismic traces s _ trace may also be zeroed out or padded with random noise.

It should be noted that the random seismic trace extraction ratio s _ percentage is selected from the range of a few tenths of a percent to a few tens of percent according to the characteristics of the seismic data.

It should be further noted that when the masking is performed on a certain patch, the point, the patch, and the strip seismic data therein are randomly extracted, and exhibit a random distribution rule.

Optionally, when the mixed wave noise interference energy of the target mixed wave interference seismic data is weak and is close to the random noise, for example, when the signal-to-noise ratio is greater than a first preset threshold, a dot mask strategy is adopted; when the mixed wave noise interference energy of the target mixed wave interference seismic data is strong, for example, the signal-to-noise ratio is smaller than a second preset threshold value, a strip mask strategy is adopted; when the mixed wave noise interference energy of the target mixed wave interference seismic data is between weak and strong, for example, the signal-to-noise ratio is not greater than a first preset threshold and not less than a second preset threshold, a block mask strategy is adopted.

For example, when the interference energy of the mixed noise is weak and close to the random noise, the signal-to-noise ratio of the seismic data is high, for example, when mse >20db, a dot mask strategy is adopted; when the interference energy of the mixed wave noise is strong, the signal-to-noise ratio of the seismic data is low, for example, when mse is less than 10db, a strip mask strategy is adopted; when the interference energy of the mixed wave noise is in a medium intensity level and the signal-to-noise ratio of the seismic data is moderate, for example, when 10db is less than or equal to mse less than or equal to 20db, a block mask strategy is adopted.

It should be noted that, on the premise that the first preset threshold is greater than the second preset threshold, both the first preset threshold and the second preset threshold may be set according to the actual requirement of the user, and specifically, no special limitation is made here.

The conventional method for removing the mixed waves through deep learning needs to synthesize a large amount of label sample data for training; the MASK strategy is adopted in the method, starting from seismic data, randomly extracting a certain proportion of point-shaped, block-shaped and strip-shaped seismic data from different segmented seismic data blocks to perform MASK MASK processing, and using the data subjected to MASK processing as network input and original block-shaped data b _i And outputting the data as a network so as to train network model parameters. The conventional supervised learning method is abandoned, and the unsupervised learning mode is adopted, so that the requirement of synthesizing a large amount of mixed wave sample data is avoided, the mixed wave suppression processing is more flexible, and the seismic data range of the mixed wave suppression processing is wider.

setting initial network parameters of the neural network model;

Optionally, after the deep U-shaped network model is built, network model training parameters such as a training sample batch _ size, a training loop number epoch, a training learning rate l _ r, a loss function loss _ function, an optimization function optimizer and the like are set, and the parameters can be flexibly adjusted according to the characteristics of the seismic data.

Optionally, a certain chunk data b in the extended chunk data set is targeted in the training process _i And masking the seismic data b' _i And b _i Dividing the test sample into a training sample, a test sample and a verification sample according to a certain proportion; wherein, the training sample is used for training the network model and adjusting parameters; the test sample is used for evaluating the generalization performance of the training network model to determine whether to retrain the model or not and is not used for adjusting the model parameters; and the verification sample is used for verifying the performance of the network model, and the model hyper-parameters are adjusted in turn according to the verification result.

For example, since the seismic data before and after the mask are in one-to-one correspondence, such as 1000 mask processed b' _i Corresponding to 1000 masks before processing _i The data, that is, 1000 pairs of sample data in total, are divided into 800 pairs of training samples, 200 pairs of test samples and 100 pairs of verification samples according to the ratio of 8:1: 1.

And taking the seismic data obtained after masking as model input data, taking the seismic data before masking as model label data, continuously training the neural network model, and continuously decreasing loss functions of the training test set and the verification set data to be stable along with the continuous increase of training times. When the training result meets the preset requirement, stable network parameters can be obtained at the moment, and the stable network parameters can be used as follow-up mixed wave interference seismic data for processing.

S140, performing mixed wave suppression processing on the converted seismic data through the trained neural network model to obtain the seismic data of the target mixed wave interference seismic data after mixed waves are removed.

After the training of the neural network model is completed, the mixed wave interference seismic data to be processed is suppressed through the trained neural network model, and then the seismic data to be processed, from which the mixed waves are removed, corresponding to the mixed wave interference seismic data can be obtained.

The embodiment provides a method for conducting noise mixing suppression on seismic data through a U-shaped network based on a Mask strategy and unsupervised learning, firstly conducting data domain conversion on the seismic data needing noise mixing suppression, then conducting sliding cutting division on the seismic data to obtain tile seismic data, then utilizing a seismic data amplification technology to greatly improve the sample size of the seismic data and enrich and expand the sample diversity, then adopting different Mask technologies to conduct random Mask replacement on the tile data, abandoning a conventional supervised learning method, avoiding the requirement of synthesizing a large amount of mixed sample data by means of an unsupervised learning mode, enabling the seismic data range of mixed processing to be wider, using data before and after the Mask as deep network input and expected output, using a network model constructed based on Unet or Unet for input and output of the seismic data, and continuously training the network parameters of the model to extract the effective information characteristic structure of the seismic data. The unsupervised seismic attenuation network model driven by the characteristics of the seismic data is constructed and is directly used for the mixed wave noise interference seismic data, and the output of the network model is the seismic data after the mixed wave is removed.

When the mixed wave interference seismic data is subjected to mixed wave suppression through the mixed wave suppression method disclosed by the embodiment, a large amount of mixed wave sample seismic data does not need to be synthesized, an unsupervised training network model is directly performed according to the mixed wave seismic data to be processed, and the trained network model is directly used for mixed wave interference processing of the mixed wave seismic data.

Compared with the traditional sparse transform mixed wave suppression method, the method and the device have the advantages that the seismic data can be more effectively protected, the loss of effective seismic data information is avoided, the construction false image cannot be caused, and meanwhile, the signal to noise ratio is also improved.

Example two

The present embodiment provides a specific example, and in the present embodiment, the effectiveness of the mixed wave interference suppression method is verified. The method of the embodiment may include the steps of:

firstly, obtaining target mixed wave interference seismic data.

And secondly, performing data domain conversion on the target mixed wave interference seismic data to obtain converted seismic data.

In some embodiments, the data domain conversion comprises:

And thirdly, training a neural network model according to the converted seismic data to obtain the trained neural network model.

Optionally, a U-type network structure such as a unnet or unnet + + built based on a coding layer Encoder and a decoding layer Decoder is adopted, fig. 2 is a unnet network structure model diagram provided in an embodiment of the present application, for example, a unnet network structure model shown in fig. 2, a slice feature structure is extracted in the unnet network structure model through a multilayer deep network coding layer Encoder, and a compressed low-dimensional seismic data feature structure is obtained as the number of layers increases and a network goes deep continuously; and then, continuously reconstructing and recovering the high-dimensional seismic data through a Decoder layer Decoder to obtain the reconstructed and recovered seismic data.

In order to improve the suppression effect of mixed wave noise and simultaneously keep effective seismic signals, the coding layer and the decoding layer are connected by jump connection, effective seismic information extracted from the coding layer with the same height is directly spliced into the decoding layer on the same layer, so that the high-order characteristics of a depth network are not lost in the processes of down-sampling and up-sampling, and the effective information loss caused in the process of down-sampling can be recovered.

acquiring training data according to the expanded patch data set;

Optionally, the block cutting and dividing process includes the following steps:

setting the size of a block to be M x N, then sliding the block on the seismic data from top to bottom and from left to right, setting a step length x _ step along the space direction and a step length t _ step along the time direction, wherein the sizes of the block and the sliding step lengths in different directions are actually adjusted according to the section size of the seismic data. If the seismic data to be processed are large, the size of the slice and the sliding step length can be properly increased, and when the size of the seismic data is not large enough for slicing, zero filling processing needs to be carried out on the seismic data, so that the seismic data can be completely sliced and divided. Through the steps, one mixed wave interference seismic data section is changed into thousands of block data sets B with the same size and size.

the unsupervised mixed wave interference suppression technology is adopted, a large amount of seismic data and tag data do not need to be synthesized, and the block seismic data set obtained in the steps needs to be subjected to amplification and expansion processing for expanding the seismic data set. The seismic data augmentation method includes polarity inversion of the seismic data, rotation at different angles (for example, rotation by 90 degrees, 180 degrees, and 270 degrees counterclockwise), up-down or left-right inversion, and the like. Meanwhile, in order to further enhance the data diversity, the size of the seismic data can be enlarged by using a mirror image turning mode, for example, the size of 32 × 32 seismic data is mirror-expanded into 56 × 56, and sliding cutting is performed on the basis of the seismic data to obtain different types of 32 × 32 seismic data. By the aid of the amplification and expansion processing method, seismic data can be expanded by several times or even more than ten times to obtain a new block data set B', due to the fact that diversity of sample data is increased, a deep neural network cannot be trapped in a local extreme value trap at a later stage, and the problem of overfitting of a network model is avoided.

and taking the masked tile data set as the model input data.

and taking the extended patch data set as the model tag data.

acquiring the signal-to-noise ratio of the block data;

and under the condition that the signal-to-noise ratio of the tile data is not greater than a first preset threshold value and not less than a second preset threshold value, performing mask processing on the tiles in the expanded tile data set based on a block mask strategy.

randomly extracting one of the tiles b from the expanded tile data set _i Where i ═ 1, 2., N is the number of patch sample sets, the patch is subjected to a masking policy processing to obtain a masked patch b' _i As input to the deep network, original tile b _i And as expected output of the deep network, training network model parameters.

mse＝10*(b _i _mean/b _i _sd)

wherein, b _i Mean is chunk data b _i Average value of (a), b _i Sd is chunk data b _i Standard deviation of (1), mse is tile data b _i The signal-to-noise ratio result value of (c).

(1) dotted mask policy

For a certain tile b _i Performing dot mask processing, randomly selecting a certain proportion of p _ percent data points p _ point in the block, then setting a mask radius m _ r within a certain range, and randomly selecting other numbers within the set radius range around the p _ pointThe point value is replaced by the point value.

(2) Block mask policy

For a certain tile b _i And carrying out block mask processing, randomly selecting a data block b _ block of a certain proportion b _ percent in the blocks, then setting a mask radius m _ r in a certain range, and randomly selecting other data point values around the b _ block in the set radius range to replace the values in the data block.

The seismic data in the randomly extracted data block b _ block may also be zeroed out or filled with random noise.

(3) Striped mask policy

For a certain tile b _i Masking processing is carried out, strip seismic channels s _ trace with a certain proportion of s _ percent in a patch are randomly selected in the space direction, then mask radius m _ r in a certain range is set, and other strip seismic channels are randomly selected in the set radius range around s _ trace to replace the seismic channels.

It should be noted that, on the premise that the first preset threshold is greater than the second preset threshold, both the first preset threshold and the second preset threshold may be set according to the actual requirements of the user, and specific limitations are not specifically made here.

setting initial network parameters of the neural network model;

Optionally, a certain piece of data b in the extended piece of data set is targeted in the training process _i And masking the seismic data b' _i And b _i Dividing the test sample into a training sample, a test sample and a verification sample according to a certain proportion; wherein, the training sample is used for training the network model and adjusting parameters; the test sample is used for evaluating the generalization performance of the training network model to determine whether to retrain the model or not and is not used for adjusting the model parameters; and the verification sample is used for verifying the performance of the network model, and the model hyper-parameters are adjusted in turn according to the verification result.

For example, since the seismic data before and after the mask are in one-to-one correspondence, for example, 1000 mask processed b' _i Corresponding to 1000 masks before processing _i The data, that is, 1000 pairs of sample data in total, are divided into 800 pairs of training samples, 200 pairs of test samples and 100 pairs of verification samples according to the ratio of 8:1: 1.

And fourthly, performing mixed wave suppression processing on the converted seismic data through the trained neural network model to obtain the seismic data of the target mixed wave interference seismic data after the mixed waves are removed.

As shown in fig. 3, fig. 3 is a schematic diagram showing a suppression effect of mixed wave interference provided by an embodiment of the present application, where a left portion in fig. 3 is original seismic data containing mixed wave noise interference, which is serious in noise interference and cannot effectively identify a phase axis; the middle portion in FIG. 3 is seismic data after the blend has been removed using the present invention; the right part in fig. 3 is shown as the suppressed clutter interference, where no damage to the effective seismic reflection signal can be found.

The effectiveness of the mixed wave interference suppression method provided by the application is proved in the embodiment, when the mixed wave interference seismic data is subjected to mixed wave suppression through the mixed wave suppression method disclosed by the embodiment, a large amount of mixed wave sample seismic data does not need to be synthesized, an unsupervised training network model is directly performed according to the mixed wave seismic data to be processed, and then the trained network model is directly used for mixed wave interference processing of the mixed wave seismic data.

EXAMPLE III

The present embodiment provides a mixed wave interference suppression apparatus, which can be used to execute the method embodiment of the present application, and please refer to the method embodiment of the present application for details not disclosed in the embodiment of the present application. Fig. 4 is a schematic structural diagram of a mixed wave interference suppression device according to an embodiment of the present application, and as shown in fig. 4, a device 400 according to the embodiment includes:

an obtaining module 401, configured to obtain target mixed wave interference seismic data;

a conversion module 402, configured to perform data domain conversion on the target mixed wave interference seismic data to obtain converted seismic data;

a training module 403, configured to train a neural network model according to the converted seismic data, to obtain a trained neural network model;

and an interference suppression module 404, configured to perform mixed wave suppression processing on the converted seismic data through the trained neural network model, so as to obtain seismic data after the target mixed wave interference seismic data is removed from mixed waves.

In some embodiments, the data domain conversion comprises:

Optionally, a U-type network structure such as a pnet or a pnet + + built based on a coding layer Encoder and a decoding layer Decoder is adopted, fig. 2 is a network structure model diagram of the pnet provided in the embodiment of the present application, for example, a pnet network structure model shown in fig. 2, a slice feature structure is extracted in the pnet network structure model through a multi-layer deep network coding layer Encoder, and a compressed low-dimensional seismic data feature structure is obtained as the number of layers increases and the network continues to go deep. And then, continuously reconstructing and recovering the high-dimensional seismic data through a Decoder layer Decoder to obtain the reconstructed and recovered seismic data.

In order to improve the mixed wave noise suppression effect and simultaneously keep effective seismic signals, a Skip connection is utilized to connect a coding layer and a decoding layer, and effective seismic information extracted from the coding layer at the same height is directly spliced and collocated to the decoding layer at the same layer, so that the high-order characteristics of a depth network are not lost in the processes of down-sampling and up-sampling, and the effective information loss caused in the process of down-sampling can be recovered.

In order to improve the generalization capability of a network architecture and increase the diversity of a network model, the random drop Dropout is utilized to randomly inactivate neurons in a multilayer network, so that the network model can effectively retain seismic data characteristics and simultaneously discard noise characteristics.

In some embodiments, training module 403 includes: the device comprises a cutting and dividing unit, an expansion unit, an acquisition unit and a training unit; wherein,

the cutting and dividing unit is used for cutting and dividing the converted seismic data into blocks to obtain a block data set;

the expansion unit is used for carrying out amplification and expansion processing on the tile data set to obtain an expanded tile data set;

an obtaining unit, configured to obtain training data according to the extended tile data set;

and the training unit is used for training the neural network model according to the training data to obtain the trained neural network model.

Optionally, the tile cutting and dividing process includes the following steps:

setting the size of a tile as M × N, then sliding the tile on the seismic data from top to bottom and from left to right, setting a step length x _ step along the spatial direction and a step length t _ step along the time direction, wherein the sizes of the tiles and the sliding step lengths in different directions are actually adjusted according to the size of the section of the seismic data. If the seismic data to be processed are large, the size of the slice and the sliding step length can be properly increased, and when the size of the seismic data is not large enough for slicing, zero filling processing needs to be carried out on the seismic data, so that the seismic data can be completely sliced and divided. Through the steps, one mixed wave interference seismic data section is changed into thousands of block data sets B with the same size and size.

the unsupervised mixed wave interference suppression technology is adopted, a large amount of seismic data and tag data do not need to be synthesized, and the block seismic data set obtained in the steps needs to be subjected to amplification and expansion processing for expanding the seismic data set. The seismic data augmentation method includes polarity inversion of the seismic data, rotation at different angles (for example, rotation by 90 degrees, 180 degrees, and 270 degrees counterclockwise), up-and-down or left-and-right inversion, and the like. Meanwhile, in order to further enhance the data diversity, the size of the seismic data can be enlarged by using a mirror image turning mode, for example, the size of 32 × 32 seismic data is mirror-expanded into 56 × 56, and sliding cutting is performed on the basis of the seismic data to obtain different types of 32 × 32 seismic data. By the aid of the amplification and expansion processing method, seismic data can be expanded by several times or even more than ten times to obtain a new block data set B', due to the fact that diversity of sample data is increased, a deep neural network cannot be trapped in a local extreme value trap at a later stage, and the problem of overfitting of a network model is avoided.

In some embodiments, the training data comprises model input data, the obtaining unit comprises: a mask subunit, a first acknowledgement subunit; wherein,

a mask subunit, configured to perform mask processing on the tile data in the expanded tile data set based on a preset mask policy, respectively, to obtain a masked tile data set;

a first confirming subunit, configured to use the masked tile data set as the model input data.

In some embodiments, the training data further comprises model tag data, and the obtaining unit further comprises a second confirming subunit configured to use the augmented tile data set as the model tag data.

In some embodiments, the preset masking policy includes: at least one of a dotted mask policy, a blockmask policy, and a striped mask policy, the mask subunit comprising: an acquisition part, a first processing part, a second processing part and a third processing part; wherein,

an acquisition section for acquiring a signal-to-noise ratio of the tile data;

a first processing part, configured to perform mask processing on a tile in the extended tile data set based on a dotted mask policy when a signal-to-noise ratio of the tile data is greater than a first preset threshold;

a second processing part, configured to perform mask processing on tiles in the extended tile data set based on a striped mask policy when a signal-to-noise ratio of the tile data is smaller than a second preset threshold;

and the third processing part is used for performing mask processing on the tiles in the expanded tile data set based on a block mask strategy under the condition that the signal-to-noise ratio of the tile data is not greater than a first preset threshold and not less than a second preset threshold.

Optionally, the signal-to-noise ratio of the patch data, specifically, the patch data b containing the noise of the mixed wave seismic data, may be obtained by a mean variance analysis method _i Pre-evaluating the noise strength, firstly obtaining the block data b _i Then obtaining the block data b by the following formula _i Signal to noise ratio of (c):

mse＝10*(b _i _mean/b _i _sd)

(1) dotted mask policy

It should be noted that the random point extraction ratio p _ percent is selected from a range of a few tenths of a percent to a few tens of percent according to the characteristics of the seismic data.

(2) Block mask policy

For a certain tile b _i And carrying out block masking treatment, randomly selecting a data block b _ block of a certain proportion b _ percent in the blocks, then setting a masking radius m _ r within a certain range, and randomly selecting other data point values around the b _ block within the set radius range to replace the values in the data block.

(3) Striped mask policy

It should be noted that the random seismic trace extraction ratio s _ percentage is selected from a range of a few tenths of a percent to a few tens of percent according to the characteristics of the seismic data.

Optionally, when the interference energy of the mixed wave noise of the target mixed wave interference seismic data is weak and is close to the random noise, for example, when the signal-to-noise ratio is greater than a first preset threshold, a dot mask strategy is adopted; when the mixed wave noise interference energy of the target mixed wave interference seismic data is strong, for example, the signal-to-noise ratio is smaller than a second preset threshold value, a strip mask strategy is adopted; when the mixed wave noise interference energy of the target mixed wave interference seismic data is between weak and strong, for example, the signal-to-noise ratio is not greater than a first preset threshold and not less than a second preset threshold, a block mask strategy is adopted.

The conventional method for removing the mixed waves through deep learning needs to synthesize a large amount of label sample data for training; the MASK strategy is adopted in the method, starting from the seismic data, randomly extracting a certain proportion of point-shaped, block-shaped and strip-shaped seismic data from the segmented different pieces of seismic data to carry out MASK MASKProcessing, masking the processed data as network input, original block b _i And outputting the data as a network so as to train network model parameters. The conventional supervised learning method is abandoned, and the unsupervised learning mode is used, so that the requirement of synthesizing a large amount of mixed wave sample data is avoided, the mixed wave suppression processing is more flexible, and the seismic data range of the mixed wave suppression processing is wider.

In some embodiments, the training unit comprises: a setting subunit, a training subunit and a configuration subunit; wherein,

the setting subunit is used for setting initial network parameters of the neural network model;

the training subunit is used for training the neural network model according to model input data and model label data of the training data to obtain optimized network parameters;

and the configuration subunit is used for taking the optimized network parameters as the network parameters of the neural network model to obtain the trained neural network model.

Optionally, a certain piece of data b in the extended piece of data set is targeted in the training process _i And masking the seismic data b' _i And b _i Dividing the test sample into a training sample, a test sample and a verification sample according to a certain proportion; wherein, the training sample is used for training the network model and adjusting parameters; the test sample is used for evaluating the generalization performance of the training network model to determine whether to retrain the model or not and is not used for adjusting the model parameters; and the verification sample is used for testing the performance of the network model, and the model hyper-parameters are adjusted in turn according to the verification result.

And (3) taking the seismic data obtained after the mask as model input data and the seismic data before the mask as model label data, continuously training the neural network model, and continuously decreasing loss functions of the training test set and the verification set data to be stable along with the continuous increase of the training times. When the training result meets the preset requirement, stable network parameters can be obtained at the moment, and the stable network parameters can be used as follow-up mixed wave interference seismic data for processing.

Those skilled in the art will appreciate that the configuration shown in fig. 4 is not intended to be limiting of the devices of the embodiments of the present application and may include more or fewer components than shown, or some components may be combined, or a different arrangement of components.

It should be noted that each of the modules/units may be a functional module or a program module, and may be implemented by software or hardware. For the modules/units implemented by hardware, the above modules/units may be located in the same processor; or the modules/units can be respectively positioned in different processors in any combination.

The device disclosed in the embodiment comprises: an obtaining module 401, configured to obtain target mixed wave interference seismic data; a conversion module 402, configured to perform data domain conversion on the target mixed wave interference seismic data to obtain converted seismic data; a training module 403, configured to train a neural network model according to the converted seismic data, to obtain a trained neural network model; and an interference suppression module 404, configured to perform mixed wave suppression processing on the converted seismic data through the trained neural network model, so as to obtain seismic data after the target mixed wave interference seismic data is removed from mixed waves. When the device disclosed by the embodiment is used for mixed wave suppression, a large amount of seismic data containing mixed wave samples are not required to be synthesized, an unsupervised training network model is directly carried out according to the mixed wave seismic data to be processed, and the trained network model is directly used for mixed wave interference processing of the mixed wave seismic data. Compared with the traditional sparse transform mixed wave suppression method, the method has the advantages that the seismic data can be more effectively protected, the loss of effective seismic data information is avoided, the construction false image cannot be caused, and meanwhile, the signal-to-noise ratio is also improved.

Example four

The present embodiment also provides a computer-readable storage medium, in which a computer program is stored, and the computer program, when being executed by a processor, can implement all or part of the steps of the method for suppressing mixed wave interference in the first embodiment or the second embodiment.

The computer-readable storage medium may also include, among other things, a computer program, a data file, a data structure, etc., alone or in combination. The computer-readable storage medium or computer program may be specifically designed and understood by those skilled in the art of computer software, or the computer-readable storage medium may be known and available to those skilled in the art of computer software. Examples of computer-readable storage media include: magnetic media such as hard disks, floppy disks, and magnetic tape; optical media such as CDROM disks and DVDs; magneto-optical media, e.g., optical disks; and hardware devices, particularly configured to store and execute computer programs, such as Read Only Memory (ROM), Random Access Memory (RAM), flash memory; or a server, app application mall, etc. Examples of computer programs include both machine code, such as produced by a compiler, and files containing higher level code that may be executed by the computer using an interpreter. The described hardware devices may be configured to act as one or more software modules to perform the operations and methods described above, and vice versa. In addition, the computer readable storage medium may be distributed over network coupled computer systems so that program code or computer programs may be stored and executed in a distributed fashion.

EXAMPLE five

Fig. 5 is a connection block diagram of an electronic device according to an embodiment of the present application, and as shown in fig. 5, the electronic device 500 may include: one or more processors 501, memory 502, multimedia components 503, input/output (I/O) interfaces 504, and communication components 505.

Where the memory 502 is used to store various types of data, which may include, for example, instructions of any application or method in the electronic device, and application-related data, the one or more processors 501 are used to perform all or part of the steps of the method for suppressing mixed wave interference in embodiment one or embodiment two.

It should be noted that the one or more processors 501 may be implemented as an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a controller, a microcontroller, a microprocessor, or other electronic components, and may be configured to perform the methods described above.

The Memory 502 may be implemented by any type of volatile or non-volatile Memory device or combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read-Only Memory (EPROM), Programmable Read-Only Memory (PROM), Read-Only Memory (ROM), magnetic Memory, flash Memory, magnetic disk, or optical disk.

The multimedia component 503 may include a screen, which may be a touch screen, and an audio component for outputting and/or inputting audio signals. For example, the audio component may include a microphone for receiving external audio signals. The received audio signal may further be stored in a memory or transmitted through a communication component. The audio assembly also includes at least one speaker for outputting audio signals.

The I/O interface 504 provides an interface between the one or more processors 501 and other interface modules, such as a keyboard, mouse, buttons, and the like. These buttons may be virtual buttons or physical buttons.

The communication component 505 is used for wired or wireless communication between the electronic device 500 and other devices. The wired communication includes communication through a network port, a serial port and the like; the wireless communication includes: Wi-Fi, Bluetooth, Near Field Communication (NFC), 2G, 3G, 4G, 5G, or a combination of one or more of them. The corresponding communication component 505 may thus comprise: Wi-Fi module, bluetooth module, NFC module.

In summary, the present application discloses a method and an apparatus for suppressing mixed wave interference, a storage medium, and an electronic device. The method comprises the steps of firstly carrying out data domain conversion on seismic data needing mixed wave noise suppression, then carrying out sliding cutting and division on the seismic data to obtain block seismic data, then utilizing a seismic data amplification technology to greatly improve the sample size of the seismic data and enrich and expand the sample diversity, adopting different Mask technologies to carry out random Mask replacement on the block data, abandoning a conventional supervised learning method, avoiding the requirement of synthesizing a large amount of mixed wave sample data by means of an unsupervised learning mode, enabling the seismic data range of mixed wave processing to be wider, taking data before and after the Mask as deep network input and expected output, and using a network model constructed based on Unet or Unet + + for input and output of the seismic data, and continuously training the network parameters of the model to extract the effective information characteristic structure of the seismic data. An unsupervised seismic attenuation network model driven by the characteristics of seismic data is constructed and is directly used for mixed wave noise interference seismic data, and the output of the network model is the seismic data after mixed wave removal. The interference of mixed wave noise is removed through the unsupervised autonomous network model, a large amount of seismic data containing mixed wave samples do not need to be synthesized, the mixed wave seismic data are directly started, the trained model is directly used for processing the mixed wave seismic data through the unsupervised autonomous network seismic data characteristic structure, and the applicability is wider.

Compared with the traditional sparse transform mixed wave suppression method, the method and the device have the advantages that the seismic data can be more effectively protected, the loss of effective seismic information is avoided, the construction false image cannot be caused, and the signal-to-noise ratio is relatively improved.

It should be further understood that the method or system disclosed in the embodiments provided in the present application may be implemented in other ways. The method or system embodiments described above are merely illustrative, for example, the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of methods and apparatus according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, a computer program segment, or a portion of a computer program, which comprises one or more computer programs for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures, or indeed, may be executed substantially concurrently, or in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems that perform the specified functions or acts, or combinations of special purpose hardware and computer programs.

In this application, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other identical elements in a process, method, apparatus or device comprising the element; if the description to "first", "second", etc. is used for descriptive purposes only, it is not to be understood as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated; in the description of the present application, unless otherwise explicitly defined, terms such as "mixed wave interference seismic data", "data domain conversion", "slice segmentation and division", "augmentation", and "mask" should be understood in a broad sense, and those skilled in the art can reasonably determine the specific meaning of the terms in the present application in combination with the specific contents of the technical solutions. In addition, in the description of the present application, the terms "plurality" and "a plurality" mean at least two unless otherwise specified.

Finally, it should be noted that in the description of the present specification, reference to the description of "one embodiment," "some embodiments," "an example," "one example" or "some examples" or the like is intended to mean that a particular feature, structure, material, or characteristic described in connection with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the terms used above do not necessarily refer to the same embodiment or example. Furthermore, the particular features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

While embodiments of the present application have been shown and described above, it is to be understood that the above embodiments are exemplary and that the description is made only for the sake of understanding the present application and not intended to limit the present application. It will be understood by those skilled in the art that various changes in form and details may be made therein without departing from the spirit and scope of the disclosure as defined by the appended claims.

Claims

1. a mixed wave interference suppression method, is characterized in that, described method comprises:

Obtain the target mixed-wave interference seismic data;

performing data domain conversion on the target mixed-wave interference seismic data to obtain converted seismic data;

The neural network model is trained according to the converted seismic data to obtain a trained neural network model;

Mixing suppression processing is performed on the converted seismic data by using the trained neural network model, to obtain the seismic data after removing the mixing from the target mixed-wave interference seismic data.

2. The method according to claim 1, wherein the neural network model is trained according to the converted seismic data, and the trained neural network model is obtained, comprising:

The converted seismic data is sliced and divided to obtain a patch data set;

performing augmentation and expansion processing on the slice data set to obtain an expanded slice data set;

Obtain training data according to the expanded slice data set;

The neural network model is trained according to the training data to obtain the trained neural network model.

3. The method according to claim 2, wherein the training data comprises model input data, and the obtaining of the training data according to the expanded patch data set comprises:

Based on a preset mask strategy, masking is performed on the patch data in the expanded patch data set, respectively, to obtain a masked patch data set;

The masked patch dataset is used as the model input data.

4. The method according to claim 2, wherein the training data further comprises model label data, and the acquiring training data according to the expanded patch data set further comprises:

The expanded patch dataset is used as the model label data.

5. The method according to claim 3, wherein the preset mask strategy comprises: at least one of a dot mask strategy, a block mask strategy and a strip mask strategy, the The patch data in the expanded patch data set is masked based on a preset mask strategy, including:

obtaining the signal-to-noise ratio of the slice data;

In the case that the signal-to-noise ratio of the patch data is greater than the first preset threshold, mask processing is performed on the patches in the expanded patch data set based on a point mask strategy;

In the case that the signal-to-noise ratio of the slice data is less than the second preset threshold, mask processing is performed on the slices in the expanded slice data set based on a stripe mask strategy;

Under the condition that the signal-to-noise ratio of the patch data is not greater than the first preset threshold and not less than the second preset threshold, mask the patches in the expanded patch data set based on the block mask strategy code processing.

6. The method according to claim 1, wherein the data domain conversion comprises:

Any of co-detection point domain transformation, co-offset domain transformation and cross domain transformation.

7. The method according to claim 2, wherein the training of the neural network model according to the training data to obtain the trained neural network model comprises:

setting initial network parameters of the neural network model;

The neural network model is trained according to the model input data and model label data of the training data to obtain optimized network parameters;

Using the optimized network parameters as the network parameters of the neural network model, the trained neural network model is obtained.

8. A mixed wave interference suppression device is characterized in that, comprising:

The acquisition module is used to acquire the target mixed-wave interference seismic data;

a conversion module for performing data domain conversion on the target mixed-wave interference seismic data to obtain converted seismic data;

A training module for training the neural network model according to the converted seismic data to obtain the trained neural network model;

The interference suppression module is used for performing mixing suppression processing on the converted seismic data by using the trained neural network model to obtain the seismic data after the mixing of the target mixed-wave interference seismic data is removed.

9. A computer-readable storage medium, characterized in that, the computer program stored in the computer-readable storage medium can be executed by one or more processors, and is used to implement any one of claims 1 to 7. The mixing interference suppression method described above.

10. An electronic device, characterized in that it comprises a memory and one or more processors, a computer program is stored on the memory, and the memory and the one or more processors are communicatively connected to each other, and the computer When the program is executed by the one or more processors, the mixing interference suppression method according to any one of claims 1 to 7 is executed.