CN112149614B - Pre-stack well-seismic combined intelligent denoising method - Google Patents

Pre-stack well-seismic combined intelligent denoising method Download PDF

Info

Publication number
CN112149614B
CN112149614B CN202011085246.5A CN202011085246A CN112149614B CN 112149614 B CN112149614 B CN 112149614B CN 202011085246 A CN202011085246 A CN 202011085246A CN 112149614 B CN112149614 B CN 112149614B
Authority
CN
China
Prior art keywords
data
sample
gather
denoising
prestack
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
Application number
CN202011085246.5A
Other languages
Chinese (zh)
Other versions
CN112149614A (en
Inventor
何文渊
宋明水
毕建军
曹佳佳
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Beijing Zhongheng Lihua Petroleum Technology Research Institute
Original Assignee
Beijing Zhongheng Lihua Petroleum Technology Research Institute
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Beijing Zhongheng Lihua Petroleum Technology Research Institute filed Critical Beijing Zhongheng Lihua Petroleum Technology Research Institute
Priority to CN202011085246.5A priority Critical patent/CN112149614B/en
Publication of CN112149614A publication Critical patent/CN112149614A/en
Application granted granted Critical
Publication of CN112149614B publication Critical patent/CN112149614B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
    • G06F2218/02—Preprocessing
    • G06F2218/04—Denoising
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/217—Validation; Performance evaluation; Active pattern learning techniques

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Artificial Intelligence (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Signal Processing (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

The invention discloses a prestack well-seismic combined intelligent denoising method, which comprises the following steps: firstly, carrying out data segmentation on a forward modeling gather on a well and forming all segmented sample data into a sample training data set; randomly extracting partial data from the sample training data set, and performing a dictionary learning algorithm of overcomplete dictionary and sparse coefficient iterative learning on the partial data to obtain an overcomplete dictionary; thirdly, performing data segmentation on the seismic prestack gather to obtain a prestack gather prediction sample data set; randomly extracting part of sample data from the prediction sample data set, solving a sparse representation coefficient of each sample data by using an over-complete dictionary, establishing a quadratic term objective function, and solving to obtain a pre-stack denoising gather; and fifthly, traversing all the gathers according to the third step and the fourth step to finish denoising. The invention achieves the goal of denoising by carrying out amplitude-preserving reconstruction on the effective signal, and the denoising of the prestack gather operated by the algorithm not only improves the signal-to-noise ratio, but also considers the AVO trend, thereby being beneficial to the inversion of prestack elastic parameters.

Description

Pre-stack well-seismic combined intelligent denoising method
Technical Field
The invention relates to the technical field of oil and gas exploration, in particular to a prestack well-seismic combined intelligent denoising method.
Background
The prestack inversion is an important mode for explaining oil gas detection and geological model parameters at present, the theoretical basis of the prestack inversion is a Zoeppritz equation and an approximate expression thereof, the basic form of the prestack inversion is to simultaneously invert longitudinal waves, transverse waves and density by utilizing a common reflection point prestack gather under the theoretical constraint of the Zeoppritz equation, and then, by utilizing the three basic rock physical parameters, other elastic parameter data bodies representing hydrocarbon detection and physical properties are obtained through calculation.
The quality of the data of the prestack gather determines the prestack inversion effect, the data quality of the prestack gather is reflected in the aspects of signal-to-noise ratio, wide angle and resolution, wherein the signal-to-noise ratio has the largest influence on the stability of an inversion algorithm, and the inversion result becomes unstable because the signal-to-noise ratio influences the condition number of an inversion coefficient matrix. The noise of the prestack gather can be divided into two categories of linear noise and random noise, at present, the removal of the random noise of the prestack gather can be filtered by using a filter of a time domain aiming at Gaussian distribution random hypothesis or a frequency domain filter, and the removal of the linear noise can be filtered by using wave field transformation separation or two-dimensional Fourier transformation denoising. The objective of prestack denoising is amplitude-preserving denoising of effective signals, but in actual data, the effective signals and noise signals of a prestack gather are mixed together, the conditions of approximation and aliasing appear in the amplitude range, the frequency range and even the statistical attribute range, and the objective of preserving the effective signals while filtering noise is difficult to achieve by using a conventional noise filter.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provides a prestack well-seismic combined intelligent denoising method.
The purpose of the invention is realized by the following technical scheme:
a prestack well-seismic combined intelligent denoising method specifically comprises the following steps:
the method comprises the following steps: performing data segmentation of a uniform size on the forward gather on the well, wherein one piece of segmented block data is used as sample data, and all the sample data form a sample training data set;
step two: randomly extracting partial data from the sample training data set, and performing a dictionary learning algorithm on the partial sample data about an over-complete dictionary and sparse coefficient iterative learning to obtain an over-complete dictionary capable of representing the two-dimensional structure characteristics of the pre-stack gather;
step three: carrying out data segmentation on the seismic prestack gather to obtain a prestack gather prediction sample data set;
step four: randomly extracting part of sample data from the prediction sample data set, solving a sparse representation coefficient of each sample data by using the over-complete dictionary obtained in the step two, then establishing a quadratic term objective function taking the denoising prestack gather as a target, and solving to obtain the prestack denoising gather;
step five: traversing all the gather according to the sequence of the third step and the fourth step to complete the denoising of all the pre-stack gathers.
Specifically, the data segmentation process in the first step includes segmenting data by using a segmentation window with a size of M × M, where data in one segmentation window is a sample, the size of the segmentation window M is one wavelength, and data at a non-integral multiple window length position in the segmentation process is discarded.
Specifically, the second step specifically includes the following substeps:
s201, randomly taking K samples from a pre-stack trace set training sample set to form a training sample set
Figure GDA0003144221330000021
The learning objective is a two-dimensional structure comprising a prestack gather, each sample having thereon a sparsely represented overcomplete dictionary D and sparsely represented coefficients aiEstablishing an objective function expression as shown in the following formula:
Figure GDA0003144221330000022
wherein λ is a sparse constraint factor;
s202, fixing the overcomplete dictionary D, and solving a sparse coefficient of each training sample under the condition that D is known, wherein the sparse coefficient is shown as the following formula:
Figure GDA0003144221330000023
wherein i is 1,2, … K
Solving to obtain the sparse coefficient a of each sample by using an optimization algorithm of linear programmingiThe sparse coefficients of all samples form a sparse matrix a ═ a (a)1,a2,…,aK)；
S203, deducing and expanding an object function formula related to D according to the object function expression as shown in the following formula, and performing iterative solution by using a projection gradient method to obtain an over-complete dictionary D
Specifically, the third step specifically includes: performing data segmentation on the seismic prestack gather by adopting two-dimensional windows with the same size, unfolding two-dimensional segmentation data into one-dimensional data serving as a prediction sample, and forming a prediction sample set by all segmentation data; when the data of the seismic prestack gather is segmented, a non-integral multiple window can be filled in a symmetrical continuation mode, and the filling data is cut after denoising operation.
Specifically, the fourth step specifically includes the following substeps:
s401, randomly extracting K sample data from the prediction sample set, sparsely reconstructing K prediction samples by using an over-complete dictionary, and meanwhile, calculating to obtain a denoising prestack gather of the whole effective signal, and establishing a target function about K prediction sample sparse representation coefficients and the denoising prestack gather by using an expression as follows:
Figure GDA0003144221330000024
in the above formula, the first and second carbon atoms are,
Figure GDA0003144221330000031
is a denoised prestack gather, I is an original prestack seismic gather, I and j refer to two-dimensional serial numbers of samples, RijI is the ijth sample data, RijIs a sample decimation matrix, DaijIs a reconstructed sub-sample, uijIs sparsity constrained sparsity;
s402, solving by using a solving expression for each sample
Figure GDA0003144221330000032
Solving the expression is shown as follows:
Figure GDA0003144221330000033
s403, solving sparse coefficient a of K sample data by using solving expressionijSolving a denoising prestack gather matrix based on the obtained solution to obtain a denoising prestack gather matrix
Figure GDA0003144221330000034
Expression ofCalculating to obtain a denoising prestack gather matrix by using a quadratic matrix equation solution algorithm
Figure GDA0003144221330000035
Wherein denoising the prestack gather matrix
Figure GDA0003144221330000036
Is represented by the following formula:
Figure GDA0003144221330000037
the invention has the beneficial effects that: the method comprises the steps of establishing an overcomplete atom library through an on-well forward modeling gather, and denoising a seismic prestack gather according to the sparsity of effective signals in the overcomplete atom library. The traditional denoising method mainly removes noise according to the spectrum distribution rule of effective signals and noise signals, but in many cases, the spectrum regions of the effective signals and the noise signals are mixed together and are difficult to separate, the effective signals are damaged while the noise is removed, and the uncertainty of pre-stack inversion is increased due to the distortion of the effective signals. Because the forward gather does not contain random noise and coherent noise, and the forward gather can reflect the accurate AVO trend, the method adopts the forward gather to establish an over-complete dictionary which can represent the two-dimensional structure characteristics of the pre-stack gather, effective signals can be obtained by the sparse combination addition of the over-complete dictionary, and noise signals do not have the characteristics, so that the dictionary sparse reconstruction of the seismic pre-stack gather can be carried out, and the de-noised pre-stack gather data can be obtained; the denoising process is different from the traditional thought of signal-noise separation, the method achieves the denoising target by preserving the amplitude of the effective signal and reconstructing the target, and the denoising of the prestack gather operated by the algorithm not only improves the signal-to-noise ratio, but also considers the AVO trend and is beneficial to the inversion of prestack elastic parameters.
Drawings
Fig. 1 is a flow diagram of the method of the present invention.
Figure 2 is a diagram of a near-well Zeoppritz forward course set of the present invention.
FIG. 3 is an overcomplete dictionary diagram for forward gather batch dictionary learning of the present invention.
Fig. 4 (a) to (c) show the comparison of the denoising effect of the present invention, where (a) is an original angle trace set diagram, (b) is an angle trace set diagram after denoising of the present invention, and (c) is a residual error of the present invention.
Detailed Description
In order to more clearly understand the technical features, objects, and effects of the present invention, embodiments of the present invention will now be described with reference to the accompanying drawings.
In this embodiment, as shown in fig. 1, a method for pre-stack well-seismic joint intelligent denoising specifically includes the following steps:
the method comprises the following steps: performing data segmentation of a uniform size on the forward gather on the well, wherein one segmented block data is used as sample data, and all the sample data form a sample training data set:
as shown in fig. 2, the prestack gather is subjected to data segmentation with a size of a black frame in the graph, the segmentation window is (M × M), two-dimensional data of one segmentation window is expanded into one-dimensional data as one sample data, and all the prestack gather is traversed to obtain a training sample data set.
The method takes two-dimensional segmentation data as sample data, and aims to learn and obtain the over-complete dictionary with both waveform and horizontal AVO characteristics. The size M of the division window is preferably one wavelength length, and data at non-integral multiple window length positions in the division process is discarded.
Step two: randomly extracting partial data from the sample training data set, and performing a dictionary learning algorithm on the partial sample data to obtain an overcomplete dictionary capable of representing the two-dimensional structure characteristics of the prestack gather, wherein the dictionary learning algorithm is related to the overcomplete dictionary and sparse coefficient iterative learning:
the invention adopts a machine learning method of dictionary learning to establish an over-complete dictionary model represented by the waveform characteristics of a prestack gather, and firstly, K samples are randomly taken out from a prestack gather training sample set to form a training sample set
Figure GDA0003144221330000041
The learning objective is a two-dimensional structure comprising prestack gathers, anAn overcomplete dictionary D with a sparse representation on each sample and coefficients a of the sparse representationiAn objective function is established as shown in the following equation (1-1):
Figure GDA0003144221330000042
where λ is the sparsity constraint factor.
Since the dictionary D and the sparse coefficient a are unknown, an iterative approximation strategy of fixing one and then solving the other can be adopted for solving, and if the overcomplete dictionary D is fixed first, the initial D is set as the cosine transform dictionary base.
(1) On the basis of the over-complete dictionary D, solving a sparse coefficient a:
d is fixed, and the sparse coefficient of each training sample under the condition that D is known is solved, as represented by formula (1-2):
Figure GDA0003144221330000043
(1-2) solving by using an optimization algorithm of linear programming to obtain a sparse coefficient a of each sampleiThe sparse coefficients of all samples form a sparse matrix a ═ a (a)1,a2,…,aK)。
(2) On the basis of the sparse matrix a, solving a dictionary D:
the formula (1-3) of the objective function related to D is derived from the formula (1-1)
Figure GDA0003144221330000044
The formula (1-3) is developed to obtain (1-4)
Figure GDA0003144221330000051
The formula (1-4) belongs to a quadratic programming problem, and can be solved by a projection gradient method generally, that is, the overcomplete dictionary D can be obtained by performing iterative solution through the formula (1-5).
Figure GDA0003144221330000052
Through the iterative operations of (1-2), (1-4) and (1-5), an overcomplete dictionary D of the prestack gather representation two-dimensional waveform structure can be obtained.
As shown in fig. 3, which represents an overcomplete dictionary learned using the forward gather of traces on the well.
Step three: carrying out data segmentation on the seismic prestack gather to obtain a prestack gather prediction sample data set:
and (3) performing data segmentation on the seismic prestack gather by adopting two-dimensional windows with the same size, unfolding the two-dimensional segmentation data into one-dimensional data serving as a prediction sample, and forming a prediction sample set by all the segmentation data.
When the data of the seismic prestack gather is segmented, a non-integral multiple window can be filled in a symmetrical continuation mode, and the filling data is cut after the denoising operation.
Step four: randomly extracting part of sample data from the prediction sample data set, solving the sparse representation coefficient of each sample data by using the over-complete dictionary obtained in the step two, then establishing a quadratic term objective function taking the denoising prestack gather as a target, and solving to obtain the prestack denoising gather:
randomly extracting K sample data from the prediction sample set, carrying out sparse reconstruction on the K prediction samples by utilizing an over-complete dictionary, simultaneously calculating to obtain a denoising prestack gather of the whole effective signal, and establishing a target function about K prediction sample sparse representation coefficients and the denoising prestack gather by utilizing (2-1):
Figure GDA0003144221330000053
in the above formula
Figure GDA0003144221330000054
Is a denoised prestack gather, IIs the original pre-stack seismic gather, i, j refer to the two-dimensional serial number of the sample, RijI is the ijth sample data, RijIs a sample decimation matrix, DaijIs a reconstructed sub-sample, uijIs sparsity constrained sparsity.
(1) For each sample, solving
Figure GDA0003144221330000055
As shown in formula (2-2):
Figure GDA0003144221330000061
(2) solving sparse coefficient a of K sample data by using (2-2)ijThen (2-1) is carried in to solve the denoising prestack gather matrix
Figure GDA0003144221330000062
To obtain the formula (2-3):
Figure GDA0003144221330000063
calculating by using quadratic matrix equation to obtain denoising prestack gather matrix
Figure GDA0003144221330000064
Fig. 4 (a) to (c) show the comparison of the denoising effect of the present invention, where (a) is an original angle trace set diagram, (b) is an angle trace set diagram after denoising of the present invention, and (c) is a residual error of the present invention.
Step five: traversing all the gather according to the sequence of the third step and the fourth step to complete the denoising of all the pre-stack gathers.
The foregoing shows and describes the general principles and broad features of the present invention and advantages thereof. It will be understood by those skilled in the art that the present invention is not limited to the embodiments described above, which are described in the specification and illustrated only to illustrate the principle of the present invention, but that various changes and modifications may be made therein without departing from the spirit and scope of the present invention, which fall within the scope of the invention as claimed. The scope of the invention is defined by the appended claims and equivalents thereof.

Claims (2)

1. A prestack well-seismic combined intelligent denoising method is characterized by comprising the following steps:
step S1: performing data segmentation of a uniform size on the forward gather on the well, wherein one piece of segmented block data is used as sample data, and all the sample data form a sample training data set; the data segmentation process in step S1 includes: dividing data by using a division window with the size of M by M, wherein the data in one division window is a sample, the size of the division window M is one wavelength, and the data at the position of the non-integral multiple window length in the division process is discarded;
step S2: randomly extracting partial data from the sample training data set, and performing a dictionary learning algorithm on the partial sample data about an over-complete dictionary and sparse coefficient iterative learning to obtain an over-complete dictionary capable of representing the two-dimensional structure characteristics of the pre-stack gather;
step S3: carrying out data segmentation on the seismic prestack gather to obtain a prestack gather prediction sample data set;
step S4: randomly extracting part of sample data from the prediction sample data set, solving a sparse representation coefficient of each sample data by using the over-complete dictionary obtained in the step two, then establishing a quadratic term objective function taking the denoising prestack gather as a target, and solving to obtain the prestack denoising gather;
step S5: traversing all the gathers according to the sequence of the steps S3 and S4 to complete the denoising of all the pre-stack gathers;
the step S2 specifically includes the following sub-steps:
step S201, randomly taking K samples from a pre-stack trace set training sample set to form a training sample set
Figure FDA0003144221320000011
The learning objective is a two-dimensional structure comprising prestack gathers, and each sampleOvercomplete dictionary D with sparse representation on it and sparse representation coefficients aiEstablishing an objective function expression as shown in the following formula:
Figure FDA0003144221320000012
wherein λ is a sparse constraint factor;
step S202, fixing the overcomplete dictionary D, and solving a sparse coefficient of each training sample under the condition that D is known, wherein the sparse coefficient is shown as the following formula:
Figure FDA0003144221320000013
wherein i is 1,2, … K
Solving to obtain the sparse coefficient a of each sample by using an optimization algorithm of linear programmingiThe sparse coefficients of all samples form a sparse matrix a ═ a (a)1,a2,…,aK)；
Step S203, deducing and expanding an object function formula related to D according to the object function expression as shown in the following formula, and performing iterative solution by using a projection gradient method to obtain an over-complete dictionary D;
Figure FDA0003144221320000014
the step S4 specifically includes the following sub-steps:
step S401, randomly extracting K sample data from the prediction sample set, sparsely reconstructing K prediction samples by using an over-complete dictionary, simultaneously calculating to obtain a denoising prestack gather of the whole effective signal, and establishing a target function about K prediction sample sparse representation coefficients and the denoising prestack gather by using an expression as follows:
Figure FDA0003144221320000021
in the above formula, the first and second carbon atoms are,
Figure FDA0003144221320000022
is a denoised prestack gather, I is an original prestack seismic gather, lambda is a sparse constraint factor, I and j refer to two-dimensional serial numbers of samples, RijI is the ijth sample data, RijIs a sample decimation matrix, DaijIs a reconstructed sub-sample, uijIs a sparse constraint coefficient;
step S402, solving for each sample by using a solving expression
Figure FDA0003144221320000023
Solving the expression is shown as follows:
Figure FDA0003144221320000024
step S403, solving sparse coefficient a of K sample data by using solving expressionijSolving a denoising prestack gather matrix based on the obtained solution to obtain a denoising prestack gather matrix
Figure FDA0003144221320000025
And calculating to obtain a denoising prestack gather matrix by utilizing a quadratic matrix equation solution algorithm
Figure FDA0003144221320000026
Wherein denoising the prestack gather matrix
Figure FDA0003144221320000027
Is represented by the following formula:
Figure FDA0003144221320000028
2. the method for pre-stack well-seismic joint intelligent denoising of claim 1, wherein the step S3 specifically comprises: performing data segmentation on the seismic prestack gather by adopting two-dimensional windows with the same size, unfolding two-dimensional segmentation data into one-dimensional data serving as a prediction sample, and forming a prediction sample set by all segmentation data; when the data of the seismic prestack gather is segmented, a non-integral multiple window can be filled in a symmetrical continuation mode, and the filling data is cut after denoising operation.
CN202011085246.5A 2020-10-12 2020-10-12 Pre-stack well-seismic combined intelligent denoising method Active CN112149614B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202011085246.5A CN112149614B (en) 2020-10-12 2020-10-12 Pre-stack well-seismic combined intelligent denoising method

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202011085246.5A CN112149614B (en) 2020-10-12 2020-10-12 Pre-stack well-seismic combined intelligent denoising method

Publications (2)

Publication Number Publication Date
CN112149614A CN112149614A (en) 2020-12-29
CN112149614B true CN112149614B (en) 2021-09-07

Family

ID=73951490

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202011085246.5A Active CN112149614B (en) 2020-10-12 2020-10-12 Pre-stack well-seismic combined intelligent denoising method

Country Status (1)

Country Link
CN (1) CN112149614B (en)

Families Citing this family (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113156498B (en) * 2021-02-26 2024-01-26 中海石油(中国)有限公司 Pre-stack AVO three-parameter inversion method and system based on homotopy continuation

Family Cites Families (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101598812B (en) * 2008-06-04 2011-04-20 中国石油天然气集团公司 Method for removing abnormal noise in single-point reception of seismic record by digital detector
WO2014194482A1 (en) * 2013-06-05 2014-12-11 中国科学院微电子研究所 Gas recognition method based on compressive perception theory

Also Published As

Publication number Publication date
CN112149614A (en) 2020-12-29

Similar Documents

Publication Publication Date Title
Richardson et al. Seismic data denoising and deblending using deep learning
CN111596366B (en) Wave impedance inversion method based on seismic signal optimization processing
CN112904426A (en) Decoupling elastic wave reverse time migration method, system and application
CN111505719B (en) Diffracted multiple suppression method based on wave field decomposition
CN114418886B (en) Robust denoising method based on depth convolution self-encoder
CN113885077A (en) Multi-seismic-source seismic data separation method based on deep learning
CN111368247A (en) Sparse representation regularization prestack AVO inversion method based on fast orthogonal dictionary
CN112149614B (en) Pre-stack well-seismic combined intelligent denoising method
CN111399057B (en) Seismic data noise suppression method based on non-convex sparse constraint
CN110895348A (en) Method, system and storage medium for extracting low-frequency information of seismic elastic impedance
CN110687597B (en) Wave impedance inversion method based on joint dictionary
CN113608259A (en) Seismic thin layer detection method based on ICEEMDAN constraint generalized S transformation
Jurkevics et al. A critique of seismic deconvolution methods
FR2616920A1 (en) INVERSION OF A VERTICAL SEISMIC PROFILE BY MINIMIZING A FUNCTION OF THE ENTROPY TYPE
Liu et al. A dictionary learning method with atom splitting for seismic footprint suppression
CN114325821A (en) Method and system for suppressing strong scattering noise in pre-stack seismic data based on 3D-SNACNN network
CN112147678B (en) High-precision time-varying wavelet inversion method
Yang et al. Deep learning with fully convolutional and dense connection framework for ground roll attenuation
CN114740528A (en) Pre-stack multi-wave joint inversion method based on ultramicro Laplace block constraint
CN111239805B (en) Block constraint time-lapse seismic difference inversion method and system based on reflectivity method
Zhou et al. Coherent noise attenuation by kurtosis-guided adaptive dictionary learning based on variational sparse representation
CN108387928B (en) Data construction method based on seismic feature transformation space
Liu et al. Seismic random noise suppression based on deep image prior and total variation
CN110687605A (en) Improved K-SVD algorithm-based algorithm analysis application in seismic signal processing
CN110398773B (en) Recovery and reconstruction method for partial missing seismic data

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