CN116186498A - Earthquake signal noise suppression method and system based on self-supervision learning - Google Patents
Earthquake signal noise suppression method and system based on self-supervision learning Download PDFInfo
- Publication number
- CN116186498A CN116186498A CN202310147295.4A CN202310147295A CN116186498A CN 116186498 A CN116186498 A CN 116186498A CN 202310147295 A CN202310147295 A CN 202310147295A CN 116186498 A CN116186498 A CN 116186498A
- Authority
- CN
- China
- Prior art keywords
- noise
- seismic signal
- network
- seismic
- module
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 51
- 230000001629 suppression Effects 0.000 title claims abstract description 27
- 238000005070 sampling Methods 0.000 claims abstract description 57
- 238000012545 processing Methods 0.000 claims abstract description 46
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000002474 experimental method Methods 0.000 claims abstract description 18
- 238000005457 optimization Methods 0.000 claims abstract description 10
- 230000006870 function Effects 0.000 claims description 61
- 238000010606 normalization Methods 0.000 claims description 24
- 238000000926 separation method Methods 0.000 claims description 16
- 238000010276 construction Methods 0.000 claims description 10
- 230000008569 process Effects 0.000 claims description 8
- 238000011176 pooling Methods 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 6
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 5
- 238000013507 mapping Methods 0.000 abstract description 2
- 238000004590 computer program Methods 0.000 description 15
- 238000003860 storage Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 8
- 238000013527 convolutional neural network Methods 0.000 description 5
- 230000009977 dual effect Effects 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 238000001228 spectrum Methods 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 238000003384 imaging method Methods 0.000 description 3
- 238000004891 communication Methods 0.000 description 2
- 238000013461 design Methods 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 238000011478 gradient descent method Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000001427 coherent effect Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 238000013434 data augmentation Methods 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a self-supervision learning-based seismic signal noise suppression method and a self-supervision learning-based seismic signal noise suppression system, which are used for processing a single noisy seismic signal; after original seismic signals are read, data are normalized to a [0,1] interval; then carrying out Bernoulli sampling experiments for a plurality of times on the normalized seismic signals, and constructing data pairs of auxiliary tasks; constructing the whole structure of a denoising network, and determining the optimization target of the network; training the constructed Bernoulli sampling data pair by using a denoising network, and learning a mapping relation from a noise-containing seismic signal to a clean seismic signal; and after the network iteration converges, the parameters of the network model are saved, and the input single noisy seismic signal is recovered. The method effectively solves the problem that the clean seismic signal label is difficult to acquire, reduces the damage to useful signals as much as possible on the basis of effectively suppressing noise, and has good fidelity and practicability.
Description
Technical Field
The invention belongs to the technical field of seismic signal processing, and particularly relates to a seismic signal noise suppression method and system based on self-supervision learning.
Background
In field seismic exploration, a great amount of random noise is inevitably doped in the acquired seismic signals due to the influence of factors such as exploration technology, equipment, environment and the like. Noise can severely impact the resolution of the seismic signal, the accuracy of reservoir predictions, and subsequent interpretation of the seismic data. Therefore, the research of the method for suppressing the noise of the seismic signals is always a research hot spot in the field of seismic signal processing.
The traditional seismic signal noise suppression methods achieve good effects, and the traditional seismic signal noise suppression methods often need clear physical information when establishing a mathematical model, and then separate useful signals from random noise through mathematical optimization. Conventional methods of suppressing seismic signal noise can be broadly divided into the following three categories: the first is a filtering-based approach. Transforming useful signals and noise into an F-K domain through two-dimensional Fourier transform, and constructing a proper one-pass one-resistance region in the transform domain, so as to realize separation of the two signals; the second category is denoising methods based on matrix decomposition. According to the physical characteristics of useful signals and noise, decomposing the noisy signals into different components through local SVD, and selecting the effective components in the noisy signals to reconstruct the useful signals; the third class is based on sparse representation of the signal. Such methods typically find a suitable dictionary in the transform domain to sparsely represent the useful signal, but cannot sparsely represent random noise.
In recent years, as deep learning techniques typified by convolutional neural networks are successfully applied to the fields of image processing and computer vision. Deep learning also brings new elicitations to noise suppression of seismic signals. A supervised deep learning method represented by DnCNN and 3D-DnCNN is used for constructing a sample-tag data set and performing data augmentation, and then training an end-to-end convolutional neural network to enable the end-to-end convolutional neural network to have noise separation capability, so that a clean useful signal is successfully recovered.
The capability of the supervised denoising network to separate random noise is very dependent on a training data set, however, in actual seismic exploration, clean seismic signal labels are often difficult to acquire, and very large manpower, material resources and financial resources are required to be paid. It is often desirable to construct a dataset from processing of noisy seismic signals using conventional denoising methods, such that the performance of the denoising network is limited by the conventional denoising method of preprocessing. When preprocessing a large amount of seismic data, the processing speed is also very slow.
Disclosure of Invention
The invention aims to solve the technical problems of providing a self-supervision learning-based method and a self-supervision learning-based system for suppressing seismic signal noise, which are used for solving the technical problems that clean seismic signals are difficult to acquire in actual seismic exploration.
The invention adopts the following technical scheme:
a self-supervision learning-based seismic signal noise suppression method comprises the following steps:
carrying out normalization processing on a single seismic signal, processing the seismic signal obtained by normalization by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments;
constructing a denoising network based on a coding and decoding structure;
determining an optimized objective function of a denoising network;
training a denoising network to be converged based on the optimized objective function;
and (3) learning the constructed Bernoulli sampling data in an input denoising network, recovering a single noise-containing seismic signal, and reconstructing to obtain a clean useful signal.
Specifically, the data pair of auxiliary tasks
The following are provided:
wherein ,mi Mask used for the ith Bernoulli sample, +.N is the number of Bernoulli sampling data pairs, y norm Is normalized and processed noise-containing seismic signals.
Further, the bernoulli experiment was repeated 100 times.
Specifically, the denoising network based on the coding and decoding structure comprises a data processing module, an encoder, a decoder and a residual noise separation module; the data processing module comprises normalization operation and Bernoulli sampling; the two-dimensional seismic signal encoded by the encoder becomes the size of the input actual seismic signal
Recovering the two-dimensional seismic signals passing through the decoder to the original size; the residual noise separation module is used for calculating the difference value between the input noisy seismic signal and the clean useful signal predicted by the network to obtain the noise separated by the network, and taking the condition that the prior value and the average value of the noise are 0 as regularization constraint.
Further, the encoder has 5 coding modules, each coding module including a partial convolution, a hole convolution, a residual error learning unit, and a maximum pooling; the decoder has 5 decoding modules, each comprising up-sampling with a scale factor of 2, a skip-join and a standard convolutional layer with Dropout.
Specifically, the optimization objective function L of the denoising network total The method comprises the following steps:
L total ＝L target +αL zm +βL tv
wherein ,Ltarget For the target loss function, alpha is the weight coefficient of the noise zero mean loss function, L zm Is a noise zero-mean loss function, beta is a weight coefficient of a total variation loss function, L tv Is the total variation loss function.
Specifically, the training denoising network specifically comprises:
optimizing the objective function L for the entire denoising network using Adam gradient descent algorithm total Optimizing, wherein the initial learning rate is 0.0001, the epoch is set to 15000, dropout is opened in the training process, and the denoising network is stored after the convergence of the optimized objective functionIs a parameter of (a).
Specifically, the single input noisy seismic signal is predicted, dropout is also opened and the test is repeated for N times during the prediction, and finally, the average value of each test result is selected as the final result.
Further, the average value x' of each experimental result is calculated specifically as follows:
wherein ,
and (5) recovering the noise-containing seismic signals by the denoising network in the ith Bernoulli experiment.
In a second aspect, an embodiment of the present invention provides a seismic signal noise suppression system based on self-supervised learning, including:
the data module is used for carrying out normalization processing on the single seismic signals, processing the seismic signals obtained through the normalization processing by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments;
the construction module is used for constructing a denoising network based on the coding and decoding structure;
the function module is used for determining an optimized objective function of the denoising network obtained by the construction module;
the training module trains the denoising network obtained by the construction module to be converged based on the optimized objective function obtained by the function module;
and the suppression module is used for learning the Bernoulli sampling data obtained by the data module into the denoising network obtained by the training module, recovering a single noisy seismic signal and reconstructing the single noisy seismic signal to obtain a clean useful signal.
Compared with the prior art, the invention has at least the following beneficial effects:
a self-supervision learning-based seismic signal noise suppression method is characterized in that all acquired seismic signals are mapped to a [0,1] interval through normalization operation in order to keep consistency of data dimension due to the fact that the data range of actual seismic signals is large. Meanwhile, in order to solve the problem that clean seismic signals are difficult to acquire, an auxiliary task data pair is constructed through Bernoulli sampling, an original problem which cannot be directly solved is converted into a dual problem, and as the original problem and the dual problem have the same solution, an optimized objective function can be constructed and a denoising network is built to train on the premise that a clean label is unknown, blind denoising of single seismic signals is achieved, when a structure of the denoising network is constructed, as the encoding and decoding network can extract high-level semantic features of data through an encoder, a decoder converts the extracted features into objective task data through up-sampling and jump connection, and a denoising network based on the encoding and decoding structure is constructed to optimally solve the objective function, and finally useful signals and random noise are separated at an output end through iteration for a certain number of times.
Further, a Bernoulli sampling data pair is constructed
With the Bernoulli sampling data pair, the original problem can be converted into the dual problem, and the optimization target of the self-supervision network is obtained. Meanwhile, bernoulli sampling also solves the problem of insufficient training samples of single seismic data.
Further, n=100 bernoulli samples are repeated to construct data pairs
On one hand, the problem of insufficient training samples of single seismic data is solved; on the other hand, the denoising network averages the processed result of each Bernoulli sampling data, so that the accuracy of the network prediction result can be improved.
Furthermore, the whole denoising network adopts a structure based on encoding and decoding, and comprises a data processing module, an encoder, a decoder and a residual noise separation module, wherein the data processing module constructs Bernoulli sampling data pairs through normalization and Bernoulli sampling operation
The encoder can effectively extract the high-level semantic features of the input seismic data through the convolution and pooling operation step by step; the decoder fully fuses the characteristics extracted by the deep and shallow networks through splicing and upsampling operations, realizes the conversion of the high-level semantic characteristics into data of a target task, and the residual noise separation module obtains separated noise by calculating residual errors between an input actual seismic signal and a network pre-calculated useful signal and carries out constraint by using the prior information that the noise is independent and the mean value is zero. The introduction of the residual noise module can also avoid network overfitting to a certain extent.
Further, the encoder is intended to extract high-level semantic features of the actual seismic data, each coding module comprising a partial convolution, a hole convolution, a residual learning unit and a max pooling operation. Wherein, partial convolution can selectively acquire context information through a mask (mask) to realize the restoration of imaging results; the cavity convolution introduces cavities on the basis of common convolution, so that the receptive field of a network can be increased; meanwhile, in order to prevent network degradation, a residual error learning unit is introduced; the pooling operation may compress the imaging and extract semantic features. The decoder comprises up-sampling with a scale factor of 2, and jump connection can merge information of a shallow network and a deep network, so that details and sizes of the target seismic imaging can be gradually restored.
Further, at network optimization objective L target The regular term is introduced on the basis, so that overfitting can be prevented, the fidelity of recovered clean seismic signals is improved, and L is minimized target On the one hand, the predicted value of the denoising network is enabled
As close as possible to the true value x i The method comprises the steps of carrying out a first treatment on the surface of the On the other hand minimize noise n i Is a function of the energy of the (c). L (L) zm The noise of denoising network separation is guaranteed to meet zero mean value, useful signals are prevented from being doped in the noise of network separation, and the fidelity of useful signals is improved. L (L) tv By constraining the gradient changes in the horizontal and vertical directions, noise can be suppressed to some extent. Together, these three loss functions make up the wholeAnd the network optimization target is used for improving the denoising performance of the whole network.
Further, the loss function of the whole network is optimized by using an Adam gradient descent method, and the initial learning rate is 0.0001. The gradient descent method aims at finding the minimum value of the network loss function and determining the optimal parameters of the whole denoising network. In addition, adam can automatically adjust the learning rate for each input variable of the loss function and update the variables by using moving averages that exponentially decrease the gradient.
Further, the single input noise-containing seismic signal is predicted, N Bernoulli sampling is repeatedly carried out, data pairs are constructed and sent to a denoising network for learning, the average value of the predicted results is taken as a final result, and multiple experiments expand the data on one hand and improve the accuracy of the network on the other hand.
Further, the multiple bernoulli sampling experiments are equivalent to training multiple denoising networks, and the prediction result of each denoising network on a single noisy seismic signal is averaged to obtain a final result. Thus, the accuracy of the network prediction of the clean useful signal can be improved.
It will be appreciated that the advantages of the second aspect may be found in the relevant description of the first aspect, and will not be described in detail herein.
In summary, the invention converts the problem which can not be solved directly originally by Bernoulli sampling according to the mathematical characteristics of noise, and obtains the approximate optimal solution of the original problem by solving the dual problem. The invention saves a great deal of time and cost required by constructing the manually marked clean seismic signals, and solves the problem that the clean seismic signals are difficult to acquire in actual seismic exploration.
The technical scheme of the invention is further described in detail through the drawings and the embodiments.
Drawings
FIG. 1 is a flow chart of the invention;
FIG. 2 is a block diagram of the entire denoising neural network of the present invention;
FIG. 3 is a graph of the processing results of actual seismic data, wherein (a) is the actual acquired seismic signal, (b) is the recovered useful seismic signal of the network of the invention, and (c) is the noise of the network separation of the invention;
FIG. 4 is a graph of the F-K spectrum of FIG. 3, wherein (a) is the F-K spectrum corresponding to the actual seismic signal, (b) is the F-K spectrum of the useful seismic signal, and (c) is the spectrum of noise;
FIG. 5 is a graph of the processing results of actual seismic data, wherein (a) is the actual acquired seismic signal, (b) is the useful seismic signal recovered by the inventive network, and (c) is the noise separated by the inventive network.
Detailed Description
The following description of the embodiments of the present invention will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are some, but not all embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In the description of the present invention, it will be understood that the terms "comprises" and "comprising," when used in this specification, specify the presence of stated features, integers, steps, operations, elements, and/or components, but do not preclude the presence or addition of one or more other features, integers, steps, operations, elements, components, and/or groups thereof.
It is also to be understood that the terminology used in the description of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It should be further understood that the term "and/or" as used in the present specification and the appended claims refers to any and all possible combinations of one or more of the associated listed items, and includes such combinations, e.g., a and/or B, may represent: a exists alone, A and B exist together, and B exists alone. In addition, the character "/" herein generally indicates that the front and rear associated objects are an "or" relationship.
It should be understood that although the terms first, second, third, etc. may be used to describe the preset ranges, etc. in the embodiments of the present invention, these preset ranges should not be limited to these terms. These terms are only used to distinguish one preset range from another. For example, a first preset range may also be referred to as a second preset range, and similarly, a second preset range may also be referred to as a first preset range without departing from the scope of embodiments of the present invention.
Depending on the context, the word "if" as used herein may be interpreted as "at … …" or "at … …" or "in response to a determination" or "in response to detection". Similarly, the phrase "if determined" or "if detected (stated condition or event)" may be interpreted as "when determined" or "in response to determination" or "when detected (stated condition or event)" or "in response to detection (stated condition or event), depending on the context.
Various structural schematic diagrams according to the disclosed embodiments of the present invention are shown in the accompanying drawings. The figures are not drawn to scale, wherein certain details are exaggerated for clarity of presentation and may have been omitted. The shapes of the various regions, layers and their relative sizes, positional relationships shown in the drawings are merely exemplary, may in practice deviate due to manufacturing tolerances or technical limitations, and one skilled in the art may additionally design regions/layers having different shapes, sizes, relative positions as actually required.
The invention provides a self-supervision learning-based seismic signal noise suppression method, which does not need to construct a clean label before processing by using an end-to-end convolutional neural network, but uses an auxiliary task to mine mathematical characteristics of data as supervision information (not an original task label but a constructed auxiliary task label), so that the convolutional neural network has the capability of separating useful signals and noise. Due to the maldistribution of the seismic signals, the input seismic signals are first normalized to the [0,1] interval using a normalization operation. The bernoulli sampling experiment is then repeated multiple times on the normalized seismic signal using the bernoulli sampling to construct a signature of the auxiliary task. And building an end-to-end deep learning network based on the coding and decoding structure, and training the data pairs constructed by Bernoulli sampling. When the network iterates to a certain number of times to converge, the parameters of the network model are saved, and the mapping from the single noisy seismic signal to the clean useful signal is realized.
Referring to fig. 1, the method for suppressing the seismic signal noise based on self-supervision learning of the present invention comprises the following steps:
s1, carrying out normalization operation on an input single seismic signal to a [0,1] interval;
firstly, reading actual seismic data y, and carrying out data normalization processing, wherein the data normalization processing is shown in the following formula (1):
wherein ,ymin and ymax Representing the minimum and maximum values of the two-dimensional seismic signal, respectively. Normalization converts the incoming noisy seismic signal to [0,1]]Interval.
S2, processing the seismic signals by using Bernoulli sampling and repeating the experiment for a plurality of times to construct data pairs of auxiliary tasks;
multiple Bernoulli sampling is carried out on the normalized seismic signals, and data pairs are constructed
The following is shown: />
wherein ,mi Mask for the ith Bernoulli sample, +. norm Is normalized and processed noise-containing seismic signals.
Bernoulli sampling corresponds to a binary mask consisting of 0 and 1 only, which acts on the two-dimensional seismic signals, and the probability of each position on the mask to be reserved is 0.7, so that a Bernoulli sampling data pair is constructed in a one-pass one-resistance mode. Meanwhile, in order to solve the defect of the data volume of a single noise-containing seismic signal, 100 Bernoulli experiments are repeatedly carried out.
S3, constructing a denoising network based on a coding and decoding structure;
referring to fig. 2, the denoising network based on the codec structure is similar to the conventional codec network UNet, and mainly includes a data processing module, an encoder, a decoder, and a residual noise separation module. Wherein the data processing module includes normalization operations and bernoulli sampling. The encoder has 5 encoding modules, and each encoding module comprises a partial convolution, a hole convolution, a residual error learning unit and maximum pooling. Through the encoder, the encoded two-dimensional seismic signal finally becomes an input size
The decoder also has 5 decoding modules, each module comprising up-sampling with a scale factor of 2, a skip-join and a standard convolutional layer with Dropout. Through the decoder, the two-dimensional seismic signal is finally restored to its original size. And finally, a residual noise separation module calculates the difference between the input noisy seismic signal and the clean signal predicted by the network to obtain the noise of network separation, and takes the condition that the prior value and the mean value of the noise are 0 as regularization constraint to prevent the network from being overfitted, so that the denoising network can be used for optimizing and solving the objective function.
S4, determining an optimized objective function of the denoising network obtained in the step S3;
taking into account clean seismic signals x i Is unknown, so that the network optimization solution formula (4) cannot be directly utilized:
the data pair constructed for step 02 bernoulli sampling and the noise independent and mean 0 property, convert equation (4) to equations (5 a) and (5 b):
to avoid doping the separated noise with any useful signal, the noise independent and mean value 0 a priori information is used as part of the optimization objective in the form of a regular penalty term. Meanwhile, the total variation loss function is introduced as a regularization term, and noise is further suppressed by constraining gradient changes in the horizontal and vertical directions. Finally, an optimized objective function of the whole denoising network is determined, wherein the optimized objective function is represented by the following formulas (6 a) - (6 d):
L total ＝L target +aL zm +βL tv (6a)
wherein ,Ltarget L is the target loss function zm Is a noise zero-mean loss function, alpha is L zm Weight coefficient of L tv As a total variation loss function, beta is L tv Is the height of the two-dimensional seismic data (sampling point number), W is the width of the two-dimensional seismic data (channel number), F θ (. Cndot.) is denoisingThe network is configured to provide a network,
and ,
Data pairs constructed for Bernoulli sampling, m i The mask used for the ith Bernoulli sample, θ is the parameter of the denoising network, n' h，w The value of noise n ' at the (h, w) position, x ' separated for the denoising network ' h，w The value of the clean seismic signal x' at the (h, w) location recovered for the denoising network.
S5, training the denoising network obtained in the step S3 based on the optimization objective function obtained in the step S4 until convergence, and storing parameters of the denoising network;
and (3) optimizing the objective function formula (6 a) by using an Adam gradient descent algorithm, setting the initial learning rate to 0.0001, setting the epoch size to 15000, opening Dropout in the training process, and storing parameters of a network model after the objective function converges.
S6, inputting the noise-containing seismic signals into the denoising network obtained in the step S5 for learning, recovering the single noise-containing seismic signals, and reconstructing clean useful signals.
And predicting the single input noise-containing seismic signal, opening Dropout during prediction, repeating the test for N times, and finally selecting the average value of each test result as a final result, wherein the average value is as follows:
wherein ,
and (5) recovering the noise-containing seismic signals by the denoising network in the ith Bernoulli experiment.
The denoising network can process the input noisy seismic signals end to end, excavate the characteristics of random noise, and finally map the noisy seismic signals into clean useful signals.
In still another embodiment of the present invention, a self-supervised learning-based seismic signal noise suppression system is provided, which can be used to implement the self-supervised learning-based seismic signal noise suppression method described above, and specifically, the self-supervised learning-based seismic signal noise suppression system includes a data module, a construction module, a function module, a training module, and a suppression module.
The data module is used for carrying out normalization processing on the single seismic signals, processing the seismic signals obtained through the normalization processing by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments;
the construction module is used for constructing a denoising network based on the coding and decoding structure;
the function module is used for determining an optimized objective function of the denoising network obtained by the construction module;
the training module trains the denoising network obtained by the construction module to be converged based on the optimized objective function obtained by the function module;
and the suppression module is used for learning the Bernoulli sampling data obtained by the data module into the denoising network obtained by the training module, recovering a single noisy seismic signal and reconstructing the single noisy seismic signal to obtain a clean useful signal.
In yet another embodiment of the present invention, a terminal device is provided, the terminal device including a processor and a memory, the memory for storing a computer program, the computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium. The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processor, digital signal processor (Digital Signal Processor, DSP), application specific integrated circuit (Application Specific Integrated Circuit, ASIC), off-the-shelf programmable gate array (Field-Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components, etc., which are the computational core and control core of the terminal adapted to implement one or more instructions, in particular to load and execute one or more instructions to implement the corresponding method flow or corresponding functions; the processor according to the embodiment of the invention can be used for the operation of the self-supervision learning-based earthquake signal noise suppression method, and comprises the following steps:
carrying out normalization processing on a single seismic signal, processing the seismic signal obtained by normalization by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments; constructing a denoising network based on a coding and decoding structure; determining an optimized objective function of a denoising network; training a denoising network to be converged based on the optimized objective function; and (3) learning the constructed Bernoulli sampling data in an input denoising network, recovering a single noise-containing seismic signal, and reconstructing to obtain a clean useful signal.
In a further embodiment of the present invention, the present invention also provides a storage medium, in particular, a computer readable storage medium (Memory), which is a Memory device in a terminal device, for storing programs and data. It will be appreciated that the computer readable storage medium herein may include both a built-in storage medium in the terminal device and an extended storage medium supported by the terminal device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium may be a high-speed RAM Memory or a Non-Volatile Memory (Non-Volatile Memory), such as at least one magnetic disk Memory.
One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the respective steps of the above-described embodiments with respect to a self-supervised learning-based seismic signal noise suppression method; one or more instructions in a computer-readable storage medium are loaded by a processor and perform the steps of:
carrying out normalization processing on a single seismic signal, processing the seismic signal obtained by normalization by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments; constructing a denoising network based on a coding and decoding structure; determining an optimized objective function of a denoising network; training a denoising network to be converged based on the optimized objective function; and (3) learning the constructed Bernoulli sampling data in an input denoising network, recovering a single noise-containing seismic signal, and reconstructing to obtain a clean useful signal.
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. The components of the embodiments of the present invention generally described and illustrated in the figures herein may be arranged and designed in a wide variety of different configurations. Thus, the following detailed description of the embodiments of the invention, as presented in the figures, is not intended to limit the scope of the invention, as claimed, but is merely representative of selected embodiments of the invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
Referring to fig. 3, fig. 3 (a) shows a two-dimensional seismic signal actually acquired, for a total of 500 channels, each channel having 800 sampling points, and the sampling time interval being 2ms. It can be seen that the horizontal in-phase axis is strongly disturbed by random noise, which seriously affects the continuity and smoothness of the useful signal in-phase axis, and also reduces the signal-to-noise ratio. The invention relates to a self-supervision learning-based seismic signal noise suppression method for processing an input signal.
Fig. 3 (b) and fig. 3 (c) are respectively a useful signal reconstructed by the denoising network and separated random noise, and it can be seen that the denoising network has a remarkable suppression effect on the random noise, and the continuity and smoothness of the same phase axis of the seismic signal are recovered. And by observing the noise of the network separation, we cannot see any in-phase structure. I.e. our network fidelity is better, will not basically damage the useful signal. Further, looking at fig. 3 (c), the present invention suppresses coherent noise in the oblique direction to some extent. To further analyze the effectiveness of the denoising network of the present invention, fig. 3 (a), 3 (b) and 3 (c) are transformed into the F-K domain, as shown in fig. 4 (a), 4 (b) and 4 (c), respectively. It can be seen that the energy of the useful signal is mainly concentrated in the middle part, while the random noise is mainly distributed around it. The useful signal and random noise separated by the denoising network are also well separated in the transform domain, which also indirectly indicates the effectiveness of the network of the present invention.
Fig. 5 (a), 5 (b) and 5 (c) are another set of actual seismic data, a noisy seismic signal, a recovered useful signal of the invention and separated noise, respectively. The actual seismic signals are 1000 channels in total, each channel has 3001 sampling points, and the sampling time interval is 2ms. In fig. 5 (b), it can be seen that the denoising network of the present invention can suppress noise well, and ensure continuity of the same phase axis of the seismic signal, regardless of whether the data area contains weak noise or strong noise. And the structure of the horizontal same phase axis cannot be seen by observing fig. 5 (c), which shows that the recovered useful signal has no energy residue basically and good fidelity.
In summary, according to the self-supervised learning-based method and system for suppressing the seismic signal noise, the problem which cannot be directly solved originally is converted through Bernoulli sampling according to the mathematical characteristics of the noise, and the approximate optimal solution of the original problem is obtained through a dual problem solving mode. Thus, the problem that clean seismic signals are difficult to acquire in actual seismic exploration is solved. In addition, when the denoising network is constructed, in order to avoid the occurrence of over fitting of a single input signal in the training process, a residual error learning unit is introduced, so that the performance of the network is improved. And to better mine the characteristics of the seismic data, hole convolution is used to increase the receptive field of the entire network. And finally, a residual noise separation module is added, random noise is separated by calculating the difference value between the noise-containing seismic signal and the useful signal which is predicted to be clean by the network, and the characteristic that the noise is independent and the mean value is zero is used as regularization constraint to avoid the relevant information of the useful signal doped in the separated noise, so that the whole network has fidelity. Experimental results show that the method can effectively remove random noise, basically has no damage to the recovered useful signals, and has good fidelity and practicability.
It will be apparent to those skilled in the art that, for convenience and brevity of description, only the above-described division of the functional units and modules is illustrated, and in practical application, the above-described functional distribution may be performed by different functional units and modules according to needs, i.e. the internal structure of the apparatus is divided into different functional units or modules to perform all or part of the above-described functions. The functional units and modules in the embodiment may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit, where the integrated units may be implemented in a form of hardware or a form of a software functional unit. In addition, specific names of the functional units and modules are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present application. The specific working process of the units and modules in the above system may refer to the corresponding process in the foregoing method embodiment, which is not described herein again.
In the foregoing embodiments, the descriptions of the embodiments are emphasized, and in part, not described or illustrated in any particular embodiment, reference is made to the related descriptions of other embodiments.
Those of ordinary skill in the art will appreciate that the various illustrative elements and algorithm steps described in connection with the embodiments disclosed herein may be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present invention.
In the embodiments provided in the present invention, it should be understood that the disclosed apparatus/terminal and method may be implemented in other manners. For example, the apparatus/terminal embodiments described above are merely illustrative, e.g., the division of the modules or units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed may be an indirect coupling or communication connection via interfaces, devices or units, which may be in electrical, mechanical or other forms.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
In addition, each functional unit in the embodiments of the present invention may be integrated in one processing unit, or each unit may exist alone physically, or two or more units may be integrated in one unit. The integrated units may be implemented in hardware or in software functional units.
The integrated modules/units, if implemented in the form of software functional units and sold or used as stand-alone products, may be stored in a computer readable storage medium. Based on such understanding, the present invention may implement all or part of the flow of the method of the above embodiment, or may be implemented by a computer program to instruct related hardware, where the computer program may be stored in a computer readable storage medium, and when the computer program is executed by a processor, the computer program may implement the steps of each of the method embodiments described above. Wherein the computer program comprises computer program code which may be in source code form, object code form, executable file or some intermediate form etc. The computer readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a U disk, a removable hard disk, a magnetic disk, an optical disk, a computer Memory, a Read-Only Memory (ROM), a random access Memory (RandomAccess Memory, RAM), an electrical carrier wave signal, a telecommunications signal, a software distribution medium, etc., it should be noted that the computer readable medium may contain content that is appropriately increased or decreased according to the requirements of jurisdictions and patent practices, such as in certain jurisdictions, according to the jurisdictions and patent practices, the computer readable medium does not contain electrical carrier wave signals and telecommunications signals.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations 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 above is only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited by this, and any modification made on the basis of the technical scheme according to the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (10)
1. The earthquake signal noise suppression method based on self-supervision learning is characterized by comprising the following steps of:
carrying out normalization processing on a single seismic signal, processing the seismic signal obtained by normalization by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments;
constructing a denoising network based on a coding and decoding structure;
determining an optimized objective function of a denoising network;
training a denoising network to be converged based on the optimized objective function;
and (3) learning the constructed Bernoulli sampling data in an input denoising network, recovering a single noise-containing seismic signal, and reconstructing to obtain a clean useful signal.
2. The method for suppressing seismic signal noise based on self-supervised learning as recited in claim 1, wherein the pair of data for the auxiliary tasks
The following are provided:
wherein ,mi Mask used for the ith Bernoulli sample, +.Hadamard product, +.Amount, y norm Is normalized and processed noise-containing seismic signals.
3. The method for suppressing seismic signal noise based on self-supervised learning as recited in claim 2, wherein the bernoulli experiment is repeated 100 times.
4. The self-supervised learning based seismic signal noise suppression method of claim 1, wherein the codec structure based denoising network comprises a data processing module, an encoder, a decoder, and a residual noise separation module; the data processing module comprises normalization operation and Bernoulli sampling; the two-dimensional seismic signal encoded by the encoder becomes the size of the input actual seismic signal
Recovering the two-dimensional seismic signals passing through the decoder to the original size; the residual noise separation module is used for calculating the difference value between the input noisy seismic signal and the clean useful signal predicted by the network to obtain the noise separated by the network, and taking the condition that the prior value and the average value of the noise are 0 as regularization constraint.
5. The method for suppressing seismic signal noise based on self-supervised learning as recited in claim 4, wherein the encoder has 5 encoding modules, each encoding module including partial convolution, hole convolution, residual learning unit, and maximum pooling; the decoder has 5 decoding modules, each comprising up-sampling with a scale factor of 2, a skip-join and a standard convolutional layer with Dropout.
6. The method for suppressing seismic signal noise based on self-supervised learning as recited in claim 1, wherein the denoising network is optimized by an objective function L total The method comprises the following steps:
L total ＝L target +αL zm +βL tv
wherein ,Ltarget As a target loss function, alphaWeight coefficient, L, of zero mean loss function of noise zm Is a noise zero-mean loss function, beta is a weight coefficient of a total variation loss function, L tv Is the total variation loss function.
7. The self-supervised learning based seismic signal noise suppression method of claim 1, wherein the training denoising network is specifically:
optimizing the objective function L for the entire denoising network using Adam gradient descent algorithm total And (3) optimizing, wherein the initial learning rate is 0.0001, the epoch is set to 15000, dropout is opened in the training process, and parameters of the denoising network are stored after the optimization objective function converges.
8. The method for suppressing the noise of the seismic signals based on the self-supervised learning as set forth in claim 1, wherein the single input seismic signal containing noise is predicted, dropout is also opened and the test is repeated N times during the prediction, and the average value of the results of each test is finally selected as the final result.
9. The method for suppressing seismic signal noise based on self-supervised learning as recited in claim 8, wherein calculating the average value x' of each experimental result is specifically:
wherein ,
and (5) recovering the noise-containing seismic signals by the denoising network in the ith Bernoulli experiment.
10. A self-supervised learning based seismic signal noise suppression system, comprising:
the data module is used for carrying out normalization processing on the single seismic signals, processing the seismic signals obtained through the normalization processing by using Bernoulli sampling, and constructing Bernoulli sampling data pairs of auxiliary tasks through repeated experiments;
the construction module is used for constructing a denoising network based on the coding and decoding structure;
the function module is used for determining an optimized objective function of the denoising network obtained by the construction module;
the training module trains the denoising network obtained by the construction module to be converged based on the optimized objective function obtained by the function module;
and the suppression module is used for learning the Bernoulli sampling data obtained by the data module into the denoising network obtained by the training module, recovering a single noisy seismic signal and reconstructing the single noisy seismic signal to obtain a clean useful signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310147295.4A CN116186498A (en) | 2023-02-21 | 2023-02-21 | Earthquake signal noise suppression method and system based on self-supervision learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310147295.4A CN116186498A (en) | 2023-02-21 | 2023-02-21 | Earthquake signal noise suppression method and system based on self-supervision learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116186498A true CN116186498A (en) | 2023-05-30 |
Family
ID=86441970
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310147295.4A Pending CN116186498A (en) | 2023-02-21 | 2023-02-21 | Earthquake signal noise suppression method and system based on self-supervision learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116186498A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117454094A (en) * | 2023-12-21 | 2024-01-26 | 天津大学 | Blind denoising method for radar echo signals |
-
2023
- 2023-02-21 CN CN202310147295.4A patent/CN116186498A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117454094A (en) * | 2023-12-21 | 2024-01-26 | 天津大学 | Blind denoising method for radar echo signals |
| CN117454094B (en) * | 2023-12-21 | 2024-04-19 | 天津大学 | Blind denoising method for radar echo signals |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109035142B (en) | Satellite image super-resolution method combining countermeasure network with aerial image prior | |
| Chen et al. | Image‐denoising algorithm based on improved K‐singular value decomposition and atom optimization | |
| CN112819732B (en) | B-scan image denoising method for ground penetrating radar | |
| CN111738954B (en) | Single-frame turbulence degradation image distortion removal method based on double-layer cavity U-Net model | |
| CN114723631B (en) | Image denoising method, system and device based on depth context prior and multi-scale reconstruction sub-network | |
| CN114283088B (en) | Low-dose CT image noise reduction method and device | |
| CN114429151B (en) | Method and system for identifying and reconstructing magnetotelluric signals based on depth residual error network | |
| CN116186498A (en) | Earthquake signal noise suppression method and system based on self-supervision learning | |
| CN111415323A (en) | Image detection method and device and neural network training method and device | |
| CN117171514A (en) | Seismic data denoising method based on multi-scale residual convolution | |
| CN110335196A (en) | A kind of super-resolution image reconstruction method and system based on fractal decoding | |
| CN116385281A (en) | Remote sensing image denoising method based on real noise model and generated countermeasure network | |
| CN115661655A (en) | Southwest mountain area cultivated land extraction method with hyperspectral and hyperspectral image depth feature fusion | |
| CN114460648A (en) | 3D convolutional neural network-based self-supervision 3D seismic data random noise suppression method | |
| CN115146667A (en) | Multi-scale seismic noise suppression method based on curvelet transform and multi-branch deep self-coding | |
| CN116823627A (en) | Image complexity evaluation-based oversized image rapid denoising method | |
| CN116719085B (en) | High-resolution processing method, device and equipment for seismic records and storage medium | |
| CN112285793B (en) | Magnetotelluric denoising method and system | |
| CN116405100B (en) | Distortion signal restoration method based on priori knowledge | |
| CN116310851B (en) | Remote sensing image change detection method | |
| CN117558288A (en) | Training method, device, equipment and storage medium of single-channel voice enhancement model | |
| CN116228576A (en) | Image defogging method based on attention mechanism and feature enhancement | |
| KR20200048002A (en) | Improvement Of Regression Performance Using Asymmetric tanh Activation Function | |
| Guo | Research on Mushroom Image Classification Algorithm Based on Deep Sparse Dictionary Learning | |
| CN118657093B (en) | Method and system for reconstructing missing information of damaged turbulent flow field |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |