CN114296134B - Deep convolutional network seismic data unmixing method and system - Google Patents
Deep convolutional network seismic data unmixing method and system Download PDFInfo
- Publication number
- CN114296134B CN114296134B CN202111605013.8A CN202111605013A CN114296134B CN 114296134 B CN114296134 B CN 114296134B CN 202111605013 A CN202111605013 A CN 202111605013A CN 114296134 B CN114296134 B CN 114296134B
- Authority
- CN
- China
- Prior art keywords
- aliasing
- data
- seismic data
- unmixing
- network
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 44
- 238000012549 training Methods 0.000 claims abstract description 49
- 239000011159 matrix material Substances 0.000 claims abstract description 45
- 238000012545 processing Methods 0.000 claims abstract description 14
- 238000013527 convolutional neural network Methods 0.000 claims description 64
- 230000006870 function Effects 0.000 claims description 11
- 238000004613 tight binding model Methods 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 4
- 239000000654 additive Substances 0.000 claims description 3
- 230000000996 additive effect Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 6
- 238000010586 diagram Methods 0.000 description 14
- 238000004590 computer program Methods 0.000 description 7
- 238000004422 calculation algorithm Methods 0.000 description 5
- 238000003860 storage Methods 0.000 description 4
- 238000012899 de-mixing Methods 0.000 description 3
- 238000013135 deep learning Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000002596 correlated effect Effects 0.000 description 1
- 230000000875 corresponding effect Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 230000002452 interceptive effect Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000002156 mixing Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 230000001360 synchronised effect Effects 0.000 description 1
Images
Landscapes
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a deep convolution network seismic data unmixing method and a system, which read aliasing acquisition seismic data and an aliasing matrix; carrying out unmixing by using the aliasing matrix to obtain pseudo unmixed data; for network prior, carrying out re-aliasing by using the acquired aliased common shot gather seismic data, taking the re-aliased pseudo-solution as input data, taking the aliased common shot gather data as output for training, and processing the pseudo-solution data by using the trained CNN network after the training is finished. The method utilizes the aliasing common shot gather structure training sample, can effectively mine the prior information of effective signals in the aliasing acquisition seismic data by utilizing the CNN network, selects the common shot gather structure training data according to the reciprocity theorem to train the data to be sent into the network, utilizes the CNN network obtained by training to process the aliasing acquisition pseudo-unmixing data, and can obtain the unmixing effective signals of the aliasing acquisition seismic data with high quality.
Description
Technical Field
The invention belongs to the technical field of seismic exploration data processing, and particularly relates to a deep convolution network seismic data unmixing method and system.
Background
In seismic exploration, synchronized source acquisition has become increasingly attractive over the past decade. In conventional seismic acquisition methods, interference between different sources is typically avoided by increasing the time interval between the sources. However, for situations where high source density or large offset coverage is required, such as wide azimuth offshore acquisition, the use of these methods would be very expensive. The method allows multiple sources to fire at overlapping time intervals and be recorded by a group of receivers. It has great potential to reduce acquisition costs by shortening the exploration time or to improve the quality of seismic signals by dense sampling and wide azimuth sources. However, this acquisition approach produces multi-source aliased signals containing severe interference noise, and in order for these data to be usable for industrial seismic exploration, it is necessary to develop algorithms to deal with the source interference.
The deep learning algorithm aims to automatically learn features and relations hidden in a large number of data sets, and is mainly used for regression, prediction and classification of the large data sets, such as face recognition, medical diagnosis and the like. As one of the most popular deep learning algorithms, convolutional Neural Networks (CNNs) use a shared local convolutional filter bank designed for images, which contains much fewer parameters than a fully connected multi-layer neural network, i.e., a full convolutional network (FCNN). As networks become deeper, or the size of the inputs becomes larger, FCNN may encounter computational and storage problems caused by a large number of parameters. FCNN ignores the structure of the input entirely. Seismic data have strong local structures that are highly correlated with neighboring samples. CNNs exploit correlation using shared local convolution filters, thereby avoiding the use of a large number of parameters. So far, the deep learning algorithm has achieved some achievements in the aspects of seismic data signal noise suppression and the like, but is still in a starting stage in the aspect of unmixing of aliasing acquisition seismic data.
The prior art is as follows:
a network seismic data unmixing method based on a deep convolutional network is disclosed. According to the method, a deep convolution network is introduced, pseudo-unmixed common receiving point gather seismic data are used as network input, effective signals are used as output to carry out weight optimization training on the network, and the trained network is used for processing aliasing acquisition seismic data, so that unmixing of the aliasing acquisition seismic data is realized.
The prior art has the following defects:
1. a great deal of additional common receiving point gather seismic data are needed to be used as a network training set, which often brings great cost to industrial application;
2. in the solving process, the parameters are selected, and if the values are not proper, the signal-to-noise ratio of the unmixing result is low or effective signals are damaged.
Disclosure of Invention
The technical problem to be solved by the present invention is to provide a method and a system for unmixing seismic data of a deep Convolutional Network, aiming at the defects in the prior art, which introduce pseudo unmixed data of aliasing acquisition seismic data into a CNN (Convolutional Neural Network) Network as a data prior regular term, and train the data sent into the CNN Network by using seismic data common shot gather structure training data, so as to adaptively obtain an unmixed result of the aliasing acquisition seismic data.
The invention adopts the following technical scheme:
a deep convolution network seismic data unmixing method comprises the following steps:
s1, reading aliasing common shot gather seismic data and an aliasing matrix;
s2, unmixing the aliasing matrix read in the step S1 to obtain pseudo unmixed data;
and S3, performing re-aliasing by using the aliasing common shot gather seismic data read in the step S1, taking the re-aliased pseudo-unmixing as input data of a convolutional neural network, taking the aliasing common shot gather data as output of the convolutional neural network, training the convolutional neural network, and processing the pseudo-unmixing data obtained in the step S2 by using the trained convolutional neural network to complete the unmixing of the deep convolutional network seismic data.
Specifically, in step S1, the aliasing acquisition seismic data b is a vector with a length of k, and is obtained by aliasing and stacking the data of the conventional common receiving point gather.
Further, the aliasing acquisition seismic data b is:
b＝Γd
where d is a vector representation of a conventional common receive point gather having n tracks and m sample points, and Γ is an aliasing matrix composed of n unit matrices of size m × m arranged in a stacked block diagonal configuration.
Specifically, in step S2, the aliasing matrix Γ is represented in a block form, and the pseudo-unmixed data is obtained by unmixing
The following:
wherein H n Is a block matrix [0 n1 I m 0 n2 ] T ，H n T Is H n Transpose of (I) m The unit matrix d with the size of m x m is the effective signal of the seismic data obtained after unmixing.
Specifically, in step S3, the structure of the convolutional neural network is as follows:
the input data of the convolutional neural network is pseudo-unmixed seismic data
And->
d represents an effective signal of the seismic data obtained after unmixing, and n represents additive noise; the output of the convolutional neural network is residual->
Learning and training the residual error by using a convolutional neural network to obtain a noise component in the pseudo-unmixed seismic data, and obtaining an effective signal (or greater than or equal to) of the seismic data after unmixing>
Further, effective signals of seismic data
Comprises the following steps:
where R denotes residual learning, Θ = { W, b } denotes network parameters, W denotes a weight matrix, and b denotes offset.
Further, the loss function l (Θ) of the convolutional neural network is:
wherein,
for N training data pairs, R is residual learning, Θ = { W, b } is a network parameter, W is a weight matrix, and F is a norm.
Further, according to the spatial reciprocity of the green function under certain initial and boundary conditions, a training set sample is constructed to train the convolutional neural network as follows:
G 1 (L r ,t；L s )＝G 2 (L s ,t；L r )
wherein G is 1 From the seismic source location L s To the detector location L r Green's function of wave propagation, G 2 For the slave detector position L r To the seismic source location L s Green's function of wave propagation, t is the time of wave propagation.
Specifically, in step S3, an aliasing common shot gather is used as a training set for convolutional neural network pre-training; using the collected aliasing seismic data common shot gather as clean data, constructing a new aliasing matrix by adding time jitter, and performing aliasing on the collected common shot gather; sending the constructed aliasing common shot point data and the corresponding pseudo unmixing into a convolutional neural network for learning; and after the training is finished, processing the pseudo-unmixing initial solution of the aliasing acquisition seismic data by using the convolutional neural network to finish the unmixing of the aliasing acquisition seismic data.
Another technical solution of the present invention is a deep convolutional network seismic data unmixing system, comprising:
the reading module is used for reading the aliasing common shot gather seismic data and the aliasing matrix;
the pseudo-unmixing module is used for unmixing the aliasing matrix read by the reading module to obtain pseudo-unmixed data;
and the unmixing module performs re-aliasing by using the aliasing common shot point gather seismic data read by the reading module, takes the re-aliased pseudo-unmixing as input data of the convolutional neural network, trains the convolutional neural network by using the aliasing common shot point gather data as the output of the convolutional neural network, processes the pseudo-unmixing data by using the trained convolutional neural network, and completes the unmixing of the deep convolutional network seismic data.
Compared with the prior art, the invention has at least the following beneficial effects:
the invention discloses a deep convolution network seismic data unmixing method which includes the steps of constructing a training sample by utilizing an aliasing common shot point gather, unmixing seismic data based on learnable CNN network priori, obtaining pseudo unmixing data of the seismic data by utilizing the characteristics of a model in the analyzed aliasing acquisition seismic data, performing aliasing structure training sample pair by utilizing the aliasing acquisition seismic data common shot point gather as clean data according to the reciprocity theorem, sending the aliasing structure training sample pair to a CNN network for training, and unmixing the aliasing acquisition seismic data by utilizing the CNN network as a priori regular term after full training, so that errors caused by manual selection of algorithm parameters are reduced.
Furthermore, the aliasing acquisition seismic data are represented as b = Γ d in the time domain, the noise component of the seismic data in the acquisition process is omitted, and the aliasing seismic data acquisition model is simplified.
Further, pseudo-unmixed data is obtained by using aliasing to acquire time domain representation
The problem of pseudo-solution mixing is converted into the problem of incoherent interference removal, the problem of the target to be solved is simplified, the complexity of network training is reduced, and the quality and the speed of the network training are greatly improved.
Further, a convolutional neural network is utilized to pair residual errors
Is learned and is indicated as->
The network training speed can be increased, and the problem of performance degradation of a deep network is solved.
Furthermore, according to the spatial reciprocity of the Green function under certain initial and boundary conditions, common shot point gather data can be reasonably utilized to construct a training sample under the condition of not additionally introducing common receiving point gather seismic data, the cost in industrial application is reduced, and the generalization capability of the method in the seismic data unmixing problem is improved.
In conclusion, the method can learn the parameters in a self-adaptive manner according to the seismic data, reduces the influence of the fixed parameters on the unmixing effect, trains the network under the condition of not additionally acquiring the seismic data, greatly reduces the industrial cost and improves the generalization capability of the method.
The technical solution of the present invention is further described in detail by the accompanying drawings and embodiments.
Drawings
FIG. 1 is a schematic diagram of an aliasing matrix;
FIG. 2 is a schematic diagram of simulated seismic data pseudo-unmixing;
FIG. 3 is a schematic diagram of a common shot gather of aliased acquisition simulated seismic data;
FIG. 4 is a schematic diagram of pseudo-unmixing obtained after artificial unmixing of a shot gather of the aliasing acquisition simulation seismic data;
FIG. 5 is a schematic diagram of artificial aliasing of a common shot gather;
FIG. 6 is a schematic diagram of time jitter introduced in artificial aliasing of common shot gathers;
FIG. 7 is a schematic diagram of aliasing acquisition effective signals, wherein (a) is aliasing acquisition simulated seismic data common receiving point gather clean effective signals, (b) is aliasing acquisition simulated seismic data common receiving point gather effective signals obtained by the method of the invention, and (c) is a difference value between the graph (a) and the graph (b);
fig. 8 is a CNN network structure;
FIG. 9 is a block flow diagram of the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
In the description of the present invention, it should be understood that the terms "comprises" and/or "comprising" indicate the presence of the described 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 herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the specification of the present invention 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 this specification and the appended claims refers to and includes any and all possible combinations of one or more of the associated listed items.
Various structural schematics according to the disclosed embodiments of the invention are shown in the drawings. The figures are not drawn to scale, wherein certain details are exaggerated and possibly omitted for clarity of presentation. The shapes of various regions, layers and their relative sizes and positional relationships shown in the drawings are merely exemplary, and deviations may occur in practice due to manufacturing tolerances or technical limitations, and a person skilled in the art may additionally design regions/layers having different shapes, sizes, relative positions, according to actual needs.
The invention provides a deep convolution network seismic data unmixing method, which analyzes the characteristics of a model in the aliasing acquisition seismic data, firstly obtains pseudo unmixing data of the seismic data, constructs an unmixing target and introduces a CNN network model as prior constraint, utilizes the common shot point gather data of the aliasing acquisition seismic data as clean data to perform artificial aliasing according to the reciprocity theorem, constructs a training sample and sends the training sample into a CNN network for training, and utilizes the CNN network to achieve the purpose of unmixing the aliasing acquisition seismic data after full training.
Referring to fig. 9, the method for de-mixing seismic data in a deep convolutional network according to the present invention constructs a training sample by using an aliasing common shot gather, and includes the following steps:
s1, reading aliasing acquisition seismic data and an aliasing matrix;
reading the aliased acquired seismic data and modeling the aliased acquired seismic data:
b＝Γd
wherein,
for a vector representation of a conventional common receiver gather having n channels and m sample points,
the unit matrix is an aliasing matrix formed by arranging n unit matrices with the size of m multiplied by m in a diagonal structure of overlapped blocks;
after aliasing, measurement aliasing is performed to acquire seismic data
Is a vector of length k which is an aliased superposition of conventional co-receiver gather data.
S2, performing unmixing by using an aliasing matrix to obtain pseudo-unmixing data;
the aliasing matrix Γ is represented blockwise:
Γ＝[H 1 ,...,H n ]
wherein,
is a matrix block, which can be expressed as follows:
wherein,
is an identity matrix of size m × m, 0 i1 Is a zero matrix of size c × m, 0 i2 Is a zero matrix of size (k-c-m) x m, c being an integer of size in the range 0 ≦ c ≦ k-m that varies depending on the aliasing order of the different positions and the given random jitter; the correspondence is as follows:
H i T H i ＝0 i1 T 0 i1 +I m 2 +0 i2 T 0 i2 ＝I m
further, obtaining:
further, obtaining:
wherein,
for effective seismic data pseudo-unmixing, the term on the right in the above equation is pseudo-unmixing interference, which is an unexcited domain transverse incoherent interference component based on a given random dithering sequence.
And S3, for network prior, carrying out re-aliasing by using the acquired aliasing common shot point gather seismic data, using the re-aliased pseudo-unmixing as input data, using the aliasing common shot point gather data as output for training, and after the training is finished, processing the pseudo-unmixing data by using the trained CNN network.
The structure of the a priori CNN network is as follows:
the input data of the CNN network is pseudo unmixed seismic data
And has->
Wherein d represents the effective signal of the seismic data obtained after unmixing, and n represents additive noise; pick the residual error>
As the output of the CNN network, the residual error is learned and trained by the network to obtain the noise component in the seismic data subjected to pseudo-unmixing, and further obtain the effective signal (or more than or equal to) of the seismic data obtained after unmixing>
Wherein R represents residual learning, Θ = { W, b } represents a network parameter, W represents a weight matrix, and b represents an offset;
in the case of a CNN network,
is represented as:
where NL =17 denotes the number of convolution layers, "+" denotes a convolution operation, and a nl Representing the instantaneous output at the nl-th level, reLU represents the activation function for max (0,), BN represents the batch normalization operation; the convolution filter size of each convolution layer of the CNN is set to be 3 x 3, and zero padding is adopted to keep the output size;
a small batch stochastic gradient descent method is used to optimize a loss function as follows:
wherein,
n training data pairs.
Constructing a training set sample, and according to the spatial reciprocity of the Green function under certain initial and boundary conditions:
G 1 (L r ,t；L s )＝G 2 (L s ,t；L r )
wherein, G 1 (G 2 ) To describe the location L from the seismic source s (Detector position L) r ) To the detector location L r (seismic source location L s ) The green function of wave propagation, t is the time of wave propagation.
Unmixing of the aliasing acquisition seismic data is realized in a common receiving point domain, but under a certain condition, a common shot point gather and a common receiving point gather of the seismic data have the same characteristics; they may have similar characteristics even if the conditions are not met in field situations; thus, an aliased common shot gather may be used as a training set for network pre-training; using the collected aliasing seismic data common shot gather as clean data, constructing a new aliasing matrix by adding time jitter, and performing aliasing on the collected common shot gather; sending the constructed aliasing common shot point data and the pseudo unmixing thereof into a CNN network for learning; and after the training is finished, processing the pseudo-unmixing initial solution of the aliasing acquisition seismic data by using the CNN network to finish the unmixing of the aliasing acquisition seismic data.
In another embodiment of the present invention, a deep convolutional network seismic data unmixing system is provided, which can be used to implement the deep convolutional network seismic data unmixing method described above, and specifically, the deep convolutional network seismic data unmixing system includes a reading module, a pseudo unmixing module, and an unmixing module.
The reading module is used for reading the aliasing common shot gather seismic data and the aliasing matrix;
the pseudo-unmixing module is used for unmixing the aliasing matrix read by the reading module to obtain pseudo-unmixed data;
and the unmixing module performs re-aliasing by using the aliasing common shot point gather seismic data read by the reading module, takes the re-aliased pseudo-unmixing as input data of the convolutional neural network, trains the convolutional neural network by using the aliasing common shot point gather data as the output of the convolutional neural network, processes the pseudo-unmixing data by using the trained convolutional neural network, and completes the unmixing of the deep convolutional network seismic data.
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some, but not all, embodiments of the present invention. The components of 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 present invention, 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 obtained by a person skilled in the art without inventive step based on the embodiments of the present invention, are within the scope of protection of the present invention.
The deep convolution network seismic data unmixing method based on the training sample constructed by the aliasing common shot gather is applied to the aliasing acquisition seismic data, and the method can realize effective separation of the aliasing acquisition seismic data and has higher fidelity of effective signals.
Referring to fig. 1, an aliasing matrix in an aliasing acquisition model is formed by arranging n unit matrices with size of m × m in a diagonal structure of overlapped blocks, wherein each slash represents a diagonal element equal to 1 in a unit block sub-matrix.
Referring to fig. 2, there are 128 channels of pseudo-unmixed data of a common receiving point channel set for aliasing acquisition of seismic data, the sampling interval is 2ms, and the interval between each channel is 12m. It can be seen that FIG. 2 contains a large number of interfering components, which significantly reduce the signal-to-noise ratio of the unmixed seismic data, affecting the usability of the data in subsequent interpretations.
Referring to FIG. 3, there are 128 common shot gathers for aliasing seismic data, with a sampling interval of 2ms and a per-trace interval of 12m. It can be seen that there are two different excitation sources and the seismic data from the two sources are aliased together.
Referring to fig. 4, the pseudo-unmixed data obtained by using the artificial aliasing of the seismic data shown in fig. 3 contains a large amount of interference components, and the characteristics of the interference components are similar to those of the pseudo-unmixed data of the common receiving point gather shown in fig. 2.
Referring to FIG. 5, there is shown an aliasing mode for the aliased acquisition seismic data shown in FIGS. 2, 3 and 4, in which the first source is excited from the left side and the second source is excited in the middle at a certain time to achieve the aliased acquisition of seismic data.
Referring to fig. 6, random time jitter introduced when training data is constructed by using the aliasing shown in fig. 3 to acquire seismic data common shot gather is randomly generated within 0 to 0.4s, so that the correlation during aliasing is reduced, and a condition is created for pseudo-unmixing of seismic data.
Referring to fig. 7, fig. 7 (a) shows the clean effective signal of the seismic data unmixed common receiving point gather, fig. 7 (b) shows the effective signal of the aliasing acquisition simulation seismic data common receiving point gather obtained by the method of the present invention, and fig. 7 (c) shows the difference between fig. 7 (a) and fig. 7 (b). Comparing fig. 7 (a) and fig. 7 (b), it can be seen that the result of the unmixing of the present invention can well eliminate the interference component in the pseudo-unmixing, and improve the accuracy and signal-to-noise ratio of the unmixing; as seen from fig. 7 (c), the effective signal is well protected by the unmixing result of the method of the present invention.
Please refer to fig. 8, which is a CNN prior network constructed by the method of the present invention, the network has 17 layers, the convolution filter size of each convolution layer is set to 3 × 3, zero padding is used to keep the output size, and the residual error is calculated
As the output of CNN network, the residual error is learned and trained by using the network to obtain the noise component in the pseudo unmixed seismic data, thus obtaining the effectiveness of the seismic data obtained after unmixingA signal.
The model and actual data calculation example shows that the deep convolution network seismic data unmixing method for constructing the training sample by using the aliasing common shot gather can realize high-quality unmixing of aliasing acquired seismic data, and effectively protects effective signals while suppressing the interference of pseudo-unmixed data.
In summary, the method and system for de-mixing deep convolutional network seismic data of the present invention have the following advantages:
1) The method uses the CNN network as a priori regular term for aliasing acquired seismic data, and can learn training parameters adaptively according to seismic data;
2) The invention has high fidelity to de-mixing effective signals and can effectively protect the effective signals;
3) The invention constructs the training data of the CNN prior network through the reciprocal theorem, and can train the network better at lower cost so as to meet the actual industrial requirements.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present 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 contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention should not be limited thereby, and any modification made on the basis of the technical idea proposed by the present invention falls within the protection scope of the claims of the present invention.
Claims (9)
1. A deep convolutional network seismic data unmixing method is characterized by comprising the following steps:
s1, reading aliasing common shot gather seismic data and an aliasing matrix;
s2, unmixing the aliasing matrix read in the step S1 to obtain pseudo unmixed data, performing block representation on the aliasing matrix gamma, and unmixing to obtain the pseudo unmixed data
The following were used:
where d is the effective signal of the seismic data and is a vector representation of a conventional common-receive-point gather having n channels and m sample points, 0 m Is a zero matrix of size mxm, H n Is a block matrix [0 n1 I m 0 n2 ] T ，H n T Is H n Transpose of (I) m Is a unit array of size m × m, 0 n1 And 0 n2 The seismic data acquisition system comprises zero matrixes with the sizes of n1 x m and n2 x m, wherein n1+ n2+ m is more than or equal to 0 and less than or equal to k, the sizes of n1 and n2 are changed along with different seismic aliasing strategies, and k is the length of the aliased common receiving point gather seismic data b;
and S3, multiplying the aliasing common shot gather seismic data read in the step S1 by using an aliasing matrix for aliasing again, taking the aliasing common shot gather data as input data of a convolutional neural network, taking the aliasing common shot gather data as output of the convolutional neural network, training the convolutional neural network, and processing the pseudo-unmixed data obtained in the step S2 by using the trained convolutional neural network to complete the unmixing of the seismic data of the deep convolutional network.
2. The method of claim 1, wherein in step S1, the aliased acquired seismic data b is a vector with a length k and is obtained by aliasing and stacking conventional common-receive-point gather data.
3. The method of deep convolutional network seismic data unmixing of claim 2, wherein the aliased acquisition seismic data b is:
b＝Γd
where Γ is an aliasing matrix formed by arranging a plurality of unit matrices of size m × m in an overlapped block diagonal configuration.
4. The deep convolutional network seismic data unmixing method of claim 1, wherein in step S3, the convolutional neural network has the following structure:
the input data of the convolutional neural network isPseudo unmixed seismic data
And->
noise represents additive noise; the output of the convolutional neural network is the residual->
Learning and training the residual error by using a convolutional neural network to obtain a noise component in the pseudo-unmixed seismic data, and obtaining an effective signal (or greater than or equal to) of the seismic data after unmixing>
5. The method of claim 4, wherein the significant signal of the seismic data is a signal of interest
Comprises the following steps:
where R denotes residual learning, Θ = { W, bias } denotes network parameters, W denotes a weight matrix, and bias denotes bias.
6. The deep convolutional network seismic data unmixing method of claim 4, wherein the loss function l (Θ) of the convolutional neural network is:
wherein,
for N training data pairs, d i For the ith label data, R is residual learning, Θ = { W, bias } is network parameter, W is weight matrix, bias represents bias, and F is norm.
7. The deep convolutional network seismic data unmixing method of claim 4, wherein a training set sample is constructed to train the convolutional neural network according to the spatial reciprocity of the green function under the initial and boundary conditions as follows:
G 1 (L r ,t；L s )＝G 2 (L s ,t；L r )
wherein, G 1 From the seismic source location L s To the detector location L r Green's function of wave propagation, G 2 For the slave detector position L r To the seismic source location L s The green function of wave propagation, t is the time of wave propagation.
8. The deep convolutional network seismic data unmixing method of claim 1, wherein in step S3, an aliasing common shot gather is used as a training set for convolutional neural network pre-training; using the collected aliasing seismic data common shot gather as clean data, constructing a new aliasing matrix by adding time jitter, and performing aliasing on the collected common shot gather; sending the constructed aliasing common shot point data and the corresponding pseudo unmixing into a convolutional neural network for learning; and after the training is finished, processing the pseudo-unmixing initial solution of the aliasing acquisition seismic data by using a convolutional neural network to finish the unmixing of the aliasing acquisition seismic data.
9. A deep convolutional network seismic data unmixing system, comprising:
the reading module is used for reading the aliasing common shot gather seismic data and the aliasing matrix;
a pseudo-unmixing module for unmixing the aliasing matrix read by the reading module to obtain a pseudo solutionMixing data, performing block representation on an aliasing matrix gamma, and performing unmixing to obtain pseudo unmixed data
The following:
wherein d is the effective signal of seismic data and is the vector representation of the conventional common receiving point gather with n channels and m sampling points, 0 m Is a zero matrix of size m × m, H n Is a block matrix [0 n1 I m 0 n2 ] T ，H n T Is H n Transpose of (I) m Is a unit array of size m × m, 0 n1 And 0 n2 The seismic data acquisition system comprises zero matrixes with the sizes of n1 x m and n2 x m, wherein n1+ n2+ m is more than or equal to 0 and less than or equal to k, the sizes of n1 and n2 are changed along with different seismic aliasing strategies, and k is the length of the aliased common receiving point gather seismic data b;
and the unmixing module is used for multiplying the aliasing common shot gather seismic data read by the reading module by using an aliasing matrix for performing reanalling, taking the unaliased pseudo-unmixing as input data of the convolutional neural network, taking the aliasing common shot gather data as the output of the convolutional neural network, training the convolutional neural network, processing the pseudo-unmixing data by using the trained convolutional neural network, and completing the unmixing of the deep convolutional network seismic data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111605013.8A CN114296134B (en) | 2021-12-24 | 2021-12-24 | Deep convolutional network seismic data unmixing method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111605013.8A CN114296134B (en) | 2021-12-24 | 2021-12-24 | Deep convolutional network seismic data unmixing method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114296134A CN114296134A (en) | 2022-04-08 |
| CN114296134B true CN114296134B (en) | 2023-03-31 |
Family
ID=80970542
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111605013.8A Active CN114296134B (en) | 2021-12-24 | 2021-12-24 | Deep convolutional network seismic data unmixing method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114296134B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114839673B (en) * | 2022-07-01 | 2022-09-23 | 中国海洋大学 | Separation method, separation system and computer equipment for multi-seismic-source efficient acquisition wave field |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113391354A (en) * | 2020-03-12 | 2021-09-14 | 中国石油天然气集团有限公司 | Aliasing noise suppression method and device |
Family Cites Families (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CA2913496C (en) * | 2013-08-23 | 2018-08-14 | Exxonmobil Upstream Research Company | Simultaneous sourcing during both seismic acquisition and seismic inversion |
| MX367935B (en) * | 2013-11-01 | 2019-09-12 | Cgg Services Sa | Hybrid deblending method and apparatus. |
| US10670757B2 (en) * | 2014-02-26 | 2020-06-02 | Pgs Geophysical As | Methods and systems for quantifying coherency and constraining coherency-based separation in simultaneous shooting acquisition |
| US11169293B2 (en) * | 2016-12-13 | 2021-11-09 | Cgg Services Sas | Device and method for model-based deblending |
| GB2558630A (en) * | 2017-01-12 | 2018-07-18 | Seismic Apparition Gmbh | Method for dealiasing data |
| CN108549106B (en) * | 2018-03-09 | 2019-10-11 | 中国石油天然气集团有限公司 | Aliasing noise drawing method and device |
| CN110244361B (en) * | 2019-06-21 | 2020-07-07 | 同济大学 | Seismic data separation method and system based on interpolation and de-aliasing simultaneous processing |
| US20220317322A1 (en) * | 2019-07-08 | 2022-10-06 | Bp Corporation North America Inc. | Method and apparatus for performing de-aliasing using deep learning |
| CN110568485B (en) * | 2019-09-06 | 2021-02-26 | 广州海洋地质调查局 | Neural network-based multi-channel seismic continuous recording and separating method |
| CN113156513B (en) * | 2021-04-14 | 2024-01-30 | 吉林大学 | Convolutional neural network seismic signal denoising method based on attention guidance |
| CN113743193B (en) * | 2021-07-22 | 2022-12-09 | 西安交通大学 | Pre-stack seismic data linear interference suppression method and system |
-
2021
- 2021-12-24 CN CN202111605013.8A patent/CN114296134B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113391354A (en) * | 2020-03-12 | 2021-09-14 | 中国石油天然气集团有限公司 | Aliasing noise suppression method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114296134A (en) | 2022-04-08 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110806602B (en) | Intelligent seismic data random noise suppression method based on deep learning | |
| CN108268931B (en) | Data processing method, device and system | |
| CN105137486B (en) | Anisotropic medium Elastic Wave reverse-time migration imaging method and its device | |
| CN104536044B (en) | The interpolation denoising method and system of a kind of geological data | |
| CN113885077B (en) | Multi-source seismic data separation method based on deep learning | |
| CN108416755A (en) | A kind of image de-noising method and system based on deep learning | |
| US20070274155A1 (en) | Coding and Decoding: Seismic Data Modeling, Acquisition and Processing | |
| CN114839673B (en) | Separation method, separation system and computer equipment for multi-seismic-source efficient acquisition wave field | |
| WO2008112036A1 (en) | Imaging of multishot seismic data | |
| CN110658557B (en) | Seismic data surface wave suppression method based on generation of countermeasure network | |
| CN114296134B (en) | Deep convolutional network seismic data unmixing method and system | |
| CN111551988B (en) | Seismic data anti-alias interpolation method combining deep learning and prediction filtering | |
| CN112231974A (en) | TBM rock breaking seismic source seismic wave field characteristic recovery method and system based on deep learning | |
| CN117111000A (en) | SAR comb spectrum interference suppression method based on dual-channel attention residual network | |
| CN116664419A (en) | InSAR phase unwrapping method of multi-scale feature fusion noise reduction CNN network | |
| CN114325821A (en) | Method and system for suppressing strong scattering noise in pre-stack seismic data based on 3D-SNACNN network | |
| CN104765069B (en) | A kind of method for suppressing the adjacent big gun interference of simultaneous shot collection | |
| CN115860113B (en) | Training method and related device for self-countermeasure neural network model | |
| CN114236610B (en) | Iterative seismic data unmixing method and system based on depth convolution network prior | |
| CN113191962A (en) | Underwater image color recovery method and device based on environment background light and storage medium | |
| CN114167500A (en) | Method and device for separating adjacent shot interference based on convolutional neural network | |
| Liu et al. | Aiding self-supervised coherent noise suppression by the introduction of signal segmentation using blind-spot networks | |
| Wang et al. | A physics-augmented deep learning method for seismic data deblending | |
| CN116319190A (en) | GAN-based large-scale MIMO system channel estimation method, device, equipment and medium | |
| CN115267899A (en) | DnCNN mixed seismic source seismic data separation method and system based on boundary preservation |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |