CN110361778A - A kind of Reconstruction of seismic data method based on generation confrontation network - Google Patents
A kind of Reconstruction of seismic data method based on generation confrontation network Download PDFInfo
- Publication number
- CN110361778A CN110361778A CN201910599289.6A CN201910599289A CN110361778A CN 110361778 A CN110361778 A CN 110361778A CN 201910599289 A CN201910599289 A CN 201910599289A CN 110361778 A CN110361778 A CN 110361778A
- Authority
- CN
- China
- Prior art keywords
- data
- seismic data
- reconstruction
- algorithm
- confrontation 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 33
- 238000012549 training Methods 0.000 claims abstract description 20
- 238000005457 optimization Methods 0.000 claims abstract description 11
- 230000004913 activation Effects 0.000 claims description 9
- 238000004364 calculation method Methods 0.000 claims description 4
- 238000003475 lamination Methods 0.000 claims description 3
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 abstract description 17
- 230000000694 effects Effects 0.000 abstract description 7
- 230000009286 beneficial effect Effects 0.000 abstract description 2
- 230000009466 transformation Effects 0.000 description 16
- 230000006870 function Effects 0.000 description 12
- 238000004458 analytical method Methods 0.000 description 6
- 238000009826 distribution Methods 0.000 description 5
- 238000002474 experimental method Methods 0.000 description 4
- 230000006835 compression Effects 0.000 description 3
- 238000007906 compression Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000004807 localization Effects 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000001149 cognitive effect Effects 0.000 description 2
- 230000004069 differentiation Effects 0.000 description 2
- 238000013507 mapping Methods 0.000 description 2
- 238000005259 measurement Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- BWSIKGOGLDNQBZ-LURJTMIESA-N (2s)-2-(methoxymethyl)pyrrolidin-1-amine Chemical compound COC[C@@H]1CCCN1N BWSIKGOGLDNQBZ-LURJTMIESA-N 0.000 description 1
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 238000003339 best practice Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000001788 irregular Effects 0.000 description 1
- 230000009191 jumping Effects 0.000 description 1
- 230000000704 physical effect Effects 0.000 description 1
- 238000011176 pooling Methods 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 238000003860 storage Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000032258 transport Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V1/00—Seismology; Seismic or acoustic prospecting or detecting
- G01V1/28—Processing seismic data, e.g. for interpretation or for event detection
- G01V1/282—Application of seismic models, synthetic seismograms
-
- 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
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- General Engineering & Computer Science (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Remote Sensing (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention belongs to oilfield quake big data reconstruction technique field more particularly to a kind of Reconstruction of seismic data methods based on generation confrontation network, comprising: use and be cut into the seismic slice data for unifying size as training set；Confrontation network is generated using depth convolution to be trained training set, and generates the training judging quota of model as seismic data using Wasserstein distance；It generates model using seismic data to rebuild seismic data, using back-propagation algorithm and based on the optimization algorithm of normal gradients come the gradient of optimization object function, so that the difference for rebuilding data and missing data minimizes.Beneficial effects of the present invention: solve the problems, such as that traditional Reconstruction of seismic data algorithm needs to meet the limitation of Nyquist sampling thheorem；Solve the problems, such as that rebuilding seismic data sparse basis using compressed sensing algorithm is difficult to select；It is bad to solve the problems, such as that compressed sensing algorithm and traditional reconstruction algorithm rebuild effect in extremely low sample rate.
Description
Technical field
The invention belongs to oilfield quake big data reconstruction technique field more particularly to a kind of ground based on generation confrontation network
Shake data re-establishing method.
Background technique
Data reconstruction is the pith of data processing.In signal field, since environment, equipment and the factors such as artificial are adopted
The signal data collected is not necessarily complete.If carrying out data explanation and analysis using incomplete data,
Analysis result can have biggish deviation, so needing to rebuild data before data interpretive analysis.In addition for ground
In the biggish collecting work of this data volume of seismic exploration, a large amount of data can be generated in links such as acquisition, storage and transports
Huge cost.Therefore on the one hand wish to reduce collected data as far as possible, on the other hand wish that the data for reconstructing are most
It is possibly accurate.
The conventional seismic data method of sampling is that had based on Nyquist sampling thheorem to the sampling interval of seismic signal
Certain requirement will appear fake frequency phenomenon if sample frequency is too low, influence the reconstruction of data.And compressive sensing theory table
It is bright: sparsity signal-based, in the case where being lower than Nyquist lack sampling, to a small number of sampled points by suitably rebuilding
Method remains to accurately reconstruction signal.Usually seismic signal is sparse in some transform domain, to utilize compressive sensing theory
It rebuilds seismic data and provides possibility.
Traditional Reconstruction of seismic data method point three classes: first kind method is the method based on predictive filtering, that is, uses and divide
Frequency forecast reason predicts high-frequency information by low-frequency information.Such methods usually by irregular sampling data as regular data at
Reason, and interpolation is carried out by Gaussian window, relatively it is easily introduced error.Second class method is the method based on wave equation, that is, passes through DMO
Or AMO Forward Modeling and Inversion operator iteratively solves an indirect problem, such methods rebuild seismic wave field using the physical property that wave is propagated,
But the prior information of underground structure is needed, and calculation amount is very big.Third class is the method based on certain transformation, i.e., first to earthquake number
According to certain transformation is carried out, then rebuild in transform domain.Such methods are answered extensively since principle is intuitive, calculated result is steady
With.
Traditional method is all to compare stubborn problem in face of low sampling rate and Non uniform sampling data reconstruction.And it is traditional
Regular uniform sampling is limited by Nyquist sampling thheorem.And the compressive sensing theory that new development is got up is thought even if sampling
Frequency is lower than the Nyquist limit, it is also possible to recover the partial data for meeting certain required precision.Compressed sensing technology is first
It is required that signal is sparse or compressible, but most of signal itself is not sparse.But, if it is in some transform domain
Inside meet this condition, is equally applicable to compressive sensing theory.Since the theoretical frame is suggested, common transform method master
There are discrete cosine transform, Fourier transformation, wavelet transformation and warp wavelet, and the learning-oriented super complete redundancy gradually adopted
Dictionary etc..
Discrete cosine transform (DCT) is one of most common transformation of field of signal processing, but dct transform is a kind of overall situation
Transformation, can not the local feature to image effectively identified.The application field of most of compressed sensing is all selected in Fu
Leaf transformation is as sparse transformation base, but Fourier transformation is the integral in entire time-domain, is a kind of transformation of overall situation, no
The spectrum signature of some local time can be portrayed well, so, in the processing this spy for having obvious jumping phenomenon of seismic data
When sign, Fourier transformation is not optimal selection.The Short Time Fourier Transform that Gabor is proposed, can preferably portray letter
Number local feature, extract spectrum information of the signal in local time interval.Its basic thought is by way of adding window
Signal is divided into many small time intervals, then does Fourier analysis in each time window, when identifying this to reach
Between be spaced in local frequencies purpose.
It is substantially the analysis side with single resolution ratio although this method realizes localization to a certain extent
Method.But for the sophisticated signal as seismic data, changed greatly in the waveform of different moments, Short Time Fourier Transform when
Frequency localization ability or limited.The localization thought of Short Time Fourier Transform, window are then inherited and developed to wavelet analysis
Mouth size is fixed, but shape can be converted with the variation of frequency, according to the difference of frequency come adjustment time resolution ratio, be made up
The window size and shape of Short Time Fourier Transform cannot disadvantages varying with frequency.But, wavelet transformation does not have direction
Recognition capability can only capture a Strange properties.Later, a kind of transformation referred to as Curvelet (Qu Bo) was developed,
Transformation base is made of the curve packetized elementary in different sizes and direction, has multiple dimensioned and multi-direction recognition capability, it is considered to be ground
Shake one of the best practice of Sparse expression.The Shearlet transformation developed in the recent period has more sensitive directionality, compares
Curvelet transformation can carry out more sparse expression to seismic signal, make compressed sensing based Reconstruction of seismic data effect more
It is good, but still there is a problem of cannot be adaptively selected according to pending data.
Summary of the invention
In view of the above-mentioned problems, the invention proposes and it is a kind of based on generate confrontation network Reconstruction of seismic data method, packet
It includes:
Step 1: using and be cut into the seismic slice data for unifying size as training set；
Step 2: confrontation network being generated using depth convolution, training set is trained, and use Wasserstein distance
The training judging quota of model is generated as seismic data；
Step 3: model being generated using seismic data, seismic data is rebuild, using back-propagation algorithm and based on mark
The optimization algorithm of quasi- gradient carrys out the gradient of optimization object function, so that the difference for rebuilding data and missing data minimizes.
The depth convolution generates confrontation Web vector graphic convolutional layer and substitutes the primary full articulamentum in confrontation network.
The depth convolution generates confrontation network and specifically includes:
Pond layer is replaced using the convolutional layer with step-length in arbiter model, is carried out in generator using warp lamination
Sampling；In addition to the output layer of generator and the input layer of arbiter, other network layers all use batch normalization layer；In addition to connecting entirely
Layer is connect, directly using the input layer of convolutional layer connection generator and the output layer of discriminator；In generator, in addition to output layer makes
With Tanh activation primitive, other layers all use ReLU activation primitive；All using Leaky ReLU activation primitive in arbiter.
The objective function is defined as:
Wherein, loss (z) is objective function, and G (z) is that depth convolution generates confrontation network function, and z is that the input of G (z) is made an uproar
Sound, A are calculation matrix, and y is the missing data detected,For regular terms, λ is the coefficient of regular terms.
Beneficial effects of the present invention:
1, solve the problems, such as that traditional Reconstruction of seismic data algorithm needs to meet the limitation of Nyquist sampling thheorem；
2, solve the problems, such as that rebuilding seismic data sparse basis using compressed sensing algorithm is difficult to select；
3, solving compressed sensing algorithm and traditional reconstruction algorithm, that effect is rebuild in extremely low sample rate is bad
Problem.
Detailed description of the invention
Fig. 1 is original GAN frame diagram.
Fig. 2 is DCGAN Maker model structure chart.
Fig. 3 is the generation result figure after embodiment training.
Fig. 4 is the seismic slice datagram of embodiment.
Fig. 5 is using the conventional compression cognitive method that SP algorithm is iterative algorithm to data reconstruction result map.
Fig. 6 is using algorithm of the invention to data reconstruction result map.
Fig. 7 is using algorithm of the invention to the SNR schematic diagram after data reconstruction.
Fig. 8 is using algorithm of the invention to the PSNR schematic diagram after data reconstruction.
Specific embodiment
The invention proposes and it is a kind of based on the Reconstruction of seismic data method for generating confrontation network, comprising: using being reduced
At the seismic slice data of unified size as training set；Confrontation network is generated using depth convolution to be trained training set,
And the training judging quota of model is generated as seismic data using Wasserstein distance；Mould is generated using seismic data
Type rebuilds seismic data, using back-propagation algorithm and based on the optimization algorithm of normal gradients come optimization object function
Gradient, so that the difference for rebuilding data and missing data minimizes.
Original GAN (generating confrontation network) frame is as shown in Fig. 1.The differentiation network D of original GAN can regard as by
Input sample is mapped to function D:D (x) → (0,1) for differentiating probability.The generator G fixed for one, can train differentiation
Device D is that (true, (false, probability is close to 0) close to self-generator G 1) is still carried out for probability from training data for differentiating input sample.
If arbiter D is trained, to current optimal state, it will be unable to be spoofed, and at this moment generator G needs continue to train
To reduce the accuracy rate of arbiter D.If generator G distribution is enough the distribution of perfect matching truthful data, arbiter is by nothing
Method tell the true and false of input sample and all inputs are provided 0.5 probability value.
The cost of the training of original GAN can be assessed with a cost function V (G, D), wherein contain generator and
The parameter of discriminator.It is formulated as follows:
Wherein x is the sample data of input, pdata(x) probability being distributed for x from truthful data, pG(x) come for x spontaneous
It grows up to be a useful person and exports the probability of sample.
Convolutional neural networks are more suitable for image data.DCGAN (depth convolution generates confrontation network) is replaced using convolutional layer
For the full articulamentum in original GAN, specifically include:
1. replacing pond pooling layers using the convolutional layer with step-length in arbiter model, warp is used in generator
Lamination is up-sampled.
2. other network layers are all that batch has been used to standardize in addition to the output layer of generator and the input layer of arbiter
Batch Normalization.Convergence is helped speed up using BN layers, stablizes study, prevents over-fitting.
3. full articulamentum is eliminated, directly using the input layer of convolutional layer connection generator and the output layer of discriminator.
4., in addition to output layer uses Tanh activation primitive, other layers all use ReLU activation primitive in generator；Sentencing
All using Leaky ReLU activation primitive in other device.
The DCGAN Maker model structure that the present invention designs is as shown in Fig. 2.
Furthermore the present invention is also modified the objective function of model, using Wasserstein Distance,
The mathematic(al) representation of Wasserstein Distance are as follows:
Wherein PdataAnd PGThe respectively distribution of training sample and generation sample.D (x) indicates discriminator to the defeated of discriminator
Enter the x output of sample.PpenaltyRepresent the distribution of input x.By PdataAnd PGThe point sampled in data is attached, then will
Obtained through stochastical sampling put is used as P on linepenaltyPoint.In this way, PGP can be pulled todata, increased penalty term can guarantee that D is
Smooth variation.Ideal D is in PdataNearby should be big as far as possible, in PGIt is small as far as possible nearby.More become
It is better to be bordering on 1.Compared with the JS Divergence in original GAN, Wasserstein Distance is a better distance
Measurement, it may finally be converted into optimization problem.
Generating confrontation network can be from low-dimensional representation space z ∈ RkIt is mapped to higher-dimension sample space G (z) ∈ Rn.It was training
This mapping can be motivated to generate the output sample for being similar to training data in journey.Therefore, generator study trained in advance is arrived
Be truthful data distribution, the input of generator be it is this output sample low-dimensional mapping.
If x*∈RnIt is desirable to reconstruct the vector come.Enable A ∈ Rm×nFor calculation matrix, η ∈ RmFor noise vector.It observes
Vector y=Ax*+η.Given y and A, our task is found out close to x*Reconstruct vector x.
Confrontation generates network model and is indicated by the function G determined.This method is to find a vector in generating space, is made
It matches with the measured value observed.Objective function can be with is defined as:
By using series of optimum algorithm, the loss loss (z) of z can be made to minimize.Network mould is fought due to generating
Type G can be micro-, therefore back-propagation algorithm can be used and calculated based on the optimization algorithm of normal gradients about loss (z)
Gradient, and obtain one optimizationTo x*Reconstruction result beMeasurement error, which can be defined, isReconstruction error is
During the experiment, a regularization term L (z) is added in objective function by discovery can make reconstructed results more preferable.
Therefore minimizing objective function is:
Whereinλ=0.1 is all taken in experiment.
The training set used is the seismic slice data for being tailored to 512*512 size, and each picture provides 512*512=
262144 input dimensions.Each pixel value is zoomed in and out, so that all values are all within the section [- 1:1].We use this
A training set has trained a DCGAN frame, and uses WassersteinDistance as the loss of arbiter.Setting life
The vector for complying with standard normal state input that the input grown up to be a useful person is dimension k=100, the result of generation are also that a size is 512*
512 seismic slice data.The model each update cycle once updates discriminator, carries out twice more to generator
Newly.Each more capable to use Adam optimizer, wherein trained batch parameter setting is 16, learning rate is set as 0.0002.Instruction
Generation result after white silk is as shown in Fig. 3.
This experiment is sampled seismic slice data as shown in Fig. 4 as initial data on its basis, and is made
It is rebuild with different algorithm for reconstructing.
5% sampling is carried out to earthquake slice of data, is the conventional compression cognitive method logarithm of iterative algorithm using SP algorithm
According to reconstruction, as shown in Fig. 5, reuses the algorithm that the present invention designs and rebuild, as shown in Fig. 6.
In order to verify the feasibility and validity of the algorithm for reconstructing that the present invention designs, true seismic data is counted
Value experiment.The index for measuring data reconstruction effect is respectively signal-to-noise ratio (SNR) and Y-PSNR (PSNR), it may be assumed that
In formula: y is initial data；For the data after reconstruction；MSE is initial data and the mean square error for rebuilding data.By
For formula (1,2) it is found that the bigger Representative errors of SNR and PSNR are smaller, the effect of reconstruction is better.
In order to study the overall trend that inventive algorithm changes with sample rate, to initial data carry out 10%-80% with
Machine sampling, is rebuild using SP, SAMP algorithm and the compressed sensing algorithm based on DCGAN respectively.Compression sense based on DCGAN
Know that algorithm does not add using addition regular terms and two methods of regular terms and rebuild to data respectively.It carries out ten times counting after rebuilding
It calculates average SNR and PSNR and maps respectively, as a result as shown in attached drawing 7 and attached drawing 8.
It can be obtained from experimental result, the compressed sensing of (sample rate is lower than 30%) based on DCGAN in the case where low sampling rate
Algorithm effect is significantly better than other algorithms, and the lower compressed sensing algorithm superiority based on DCGAN of sample rate is more obvious.When
The compressed sensing algorithm effect based on DCGAN has reached certain bottleneck when sample rate is higher than 40%, will not mentioning with sample rate
It rises and is obviously improved.
This embodiment is merely preferred embodiments of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with scope of protection of the claims
Subject to.
Claims (4)
1. a kind of based on the Reconstruction of seismic data method for generating confrontation network characterized by comprising
Step 1: using and be cut into the seismic slice data for unifying size as training set；
Step 2: confrontation network being generated using depth convolution, training set is trained, and made using Wasserstein distance
The training judging quota of model is generated for seismic data；
Step 3: model being generated using seismic data, seismic data is rebuild, using back-propagation algorithm and based on standard ladder
The optimization algorithm of degree carrys out the gradient of optimization object function, so that the difference for rebuilding data and missing data minimizes.
2. Reconstruction of seismic data method according to claim 1, which is characterized in that the depth convolution, which generates confrontation network, to be made
The primary full articulamentum in confrontation network is substituted with convolutional layer.
3. Reconstruction of seismic data method according to claim 2, which is characterized in that the depth convolution generates confrontation network tool
Body includes:
Pond layer is replaced using the convolutional layer with step-length in arbiter model, adopt using warp lamination in generator
Sample；In addition to the output layer of generator and the input layer of arbiter, other network layers all use batch normalization layer；In addition to connecting entirely
Layer, directly using the input layer of convolutional layer connection generator and the output layer of discriminator；In generator, in addition to output layer uses
Tanh activation primitive, other layers all use ReLU activation primitive；All using Leaky ReLU activation primitive in arbiter.
4. any Reconstruction of seismic data method according to claim 1~3, which is characterized in that the definition of the objective function
Are as follows:
Wherein, loss (z) is objective function, and G (z) is that depth convolution generates confrontation network function, and z is the input noise of G (z),
For calculation matrix, y is the missing data detected,For regular terms, λ is the coefficient of regular terms.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910599289.6A CN110361778B (en) | 2019-07-04 | 2019-07-04 | Seismic data reconstruction method based on generation countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910599289.6A CN110361778B (en) | 2019-07-04 | 2019-07-04 | Seismic data reconstruction method based on generation countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110361778A true CN110361778A (en) | 2019-10-22 |
| CN110361778B CN110361778B (en) | 2020-10-13 |
Family
ID=68217961
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910599289.6A Active CN110361778B (en) | 2019-07-04 | 2019-07-04 | Seismic data reconstruction method based on generation countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110361778B (en) |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111007566A (en) * | 2019-12-27 | 2020-04-14 | 西南石油大学 | Curvature-driven diffusion full-convolution network seismic data bad channel reconstruction and denoising method |
| CN111224938A (en) * | 2019-11-08 | 2020-06-02 | 吉林大学 | Wireless seismograph network compressed data transmission method |
| CN111681156A (en) * | 2020-06-16 | 2020-09-18 | 南开大学 | Deep compressed sensing image reconstruction method applied to wireless sensor network |
| CN112700372A (en) * | 2021-01-11 | 2021-04-23 | 河北工业大学 | Seismic data interpolation method combining Gabor feature extraction and support vector regression |
| CN113484908A (en) * | 2021-08-25 | 2021-10-08 | 成都理工大学 | Missing seismic data reconstruction method with partial convolution and attention mechanism fused with deep learning network |
| CN113552615A (en) * | 2020-04-23 | 2021-10-26 | 中国石油天然气股份有限公司 | Seismic data interpolation method and device based on generation countermeasure network |
| CN113945974A (en) * | 2021-06-10 | 2022-01-18 | 中国矿业大学(北京) | Seismic data reconstruction method, device, equipment and medium |
| CN114063168A (en) * | 2021-11-16 | 2022-02-18 | 电子科技大学 | Artificial intelligence noise reduction method for seismic signals |
| CN115062139A (en) * | 2022-05-10 | 2022-09-16 | 电子科技大学 | Automatic searching method for dialogue text abstract model |
| CN116736372A (en) * | 2023-06-05 | 2023-09-12 | 成都理工大学 | Seismic interpolation method and system for generating countermeasure network based on spectrum normalization |
| CN116756515A (en) * | 2023-07-07 | 2023-09-15 | 广州大学 | Artificial seismic wave generation method based on deep convolution generation countermeasure network |
| CN118052558A (en) * | 2024-04-15 | 2024-05-17 | 万联易达物流科技有限公司 | Wind control model decision method and system based on artificial intelligence |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109146868A (en) * | 2018-08-27 | 2019-01-04 | 北京青燕祥云科技有限公司 | 3D Lung neoplasm generation method, device and electronic equipment |
| US20190064389A1 (en) * | 2017-08-25 | 2019-02-28 | Huseyin Denli | Geophysical Inversion with Convolutional Neural Networks |
| CN109447249A (en) * | 2018-12-17 | 2019-03-08 | 中国科学院计算技术研究所 | A kind of confrontation neural network log data method for reconstructing based on depth convolution |
| CN109490957A (en) * | 2018-11-28 | 2019-03-19 | 华北电力大学 | A kind of Reconstruction of seismic data method based on space constraint compressed sensing |
| CN109544656A (en) * | 2018-11-23 | 2019-03-29 | 南京信息工程大学 | A kind of compressed sensing image rebuilding method and system based on generation confrontation network |
| US20190170888A1 (en) * | 2017-12-06 | 2019-06-06 | Chevron U.S.A. Inc. | Systems and methods for refining estimated parameter values in seismic imaging |
-
2019
- 2019-07-04 CN CN201910599289.6A patent/CN110361778B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190064389A1 (en) * | 2017-08-25 | 2019-02-28 | Huseyin Denli | Geophysical Inversion with Convolutional Neural Networks |
| US20190170888A1 (en) * | 2017-12-06 | 2019-06-06 | Chevron U.S.A. Inc. | Systems and methods for refining estimated parameter values in seismic imaging |
| CN109146868A (en) * | 2018-08-27 | 2019-01-04 | 北京青燕祥云科技有限公司 | 3D Lung neoplasm generation method, device and electronic equipment |
| CN109544656A (en) * | 2018-11-23 | 2019-03-29 | 南京信息工程大学 | A kind of compressed sensing image rebuilding method and system based on generation confrontation network |
| CN109490957A (en) * | 2018-11-28 | 2019-03-19 | 华北电力大学 | A kind of Reconstruction of seismic data method based on space constraint compressed sensing |
| CN109447249A (en) * | 2018-12-17 | 2019-03-08 | 中国科学院计算技术研究所 | A kind of confrontation neural network log data method for reconstructing based on depth convolution |
Cited By (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111224938A (en) * | 2019-11-08 | 2020-06-02 | 吉林大学 | Wireless seismograph network compressed data transmission method |
| CN111007566A (en) * | 2019-12-27 | 2020-04-14 | 西南石油大学 | Curvature-driven diffusion full-convolution network seismic data bad channel reconstruction and denoising method |
| CN113552615A (en) * | 2020-04-23 | 2021-10-26 | 中国石油天然气股份有限公司 | Seismic data interpolation method and device based on generation countermeasure network |
| CN111681156A (en) * | 2020-06-16 | 2020-09-18 | 南开大学 | Deep compressed sensing image reconstruction method applied to wireless sensor network |
| CN112700372B (en) * | 2021-01-11 | 2022-03-01 | 河北工业大学 | Seismic data interpolation method combining Gabor feature extraction and support vector regression |
| CN112700372A (en) * | 2021-01-11 | 2021-04-23 | 河北工业大学 | Seismic data interpolation method combining Gabor feature extraction and support vector regression |
| CN113945974A (en) * | 2021-06-10 | 2022-01-18 | 中国矿业大学(北京) | Seismic data reconstruction method, device, equipment and medium |
| CN113484908A (en) * | 2021-08-25 | 2021-10-08 | 成都理工大学 | Missing seismic data reconstruction method with partial convolution and attention mechanism fused with deep learning network |
| CN113484908B (en) * | 2021-08-25 | 2023-07-14 | 成都理工大学 | Missing seismic data reconstruction method for deep learning network by combining partial convolution and attention mechanism |
| CN114063168A (en) * | 2021-11-16 | 2022-02-18 | 电子科技大学 | Artificial intelligence noise reduction method for seismic signals |
| CN115062139A (en) * | 2022-05-10 | 2022-09-16 | 电子科技大学 | Automatic searching method for dialogue text abstract model |
| CN115062139B (en) * | 2022-05-10 | 2024-06-11 | 电子科技大学 | Automatic searching method for dialogue text abstract model |
| CN116736372A (en) * | 2023-06-05 | 2023-09-12 | 成都理工大学 | Seismic interpolation method and system for generating countermeasure network based on spectrum normalization |
| CN116736372B (en) * | 2023-06-05 | 2024-01-26 | 成都理工大学 | Seismic interpolation method and system for generating countermeasure network based on spectrum normalization |
| CN116756515A (en) * | 2023-07-07 | 2023-09-15 | 广州大学 | Artificial seismic wave generation method based on deep convolution generation countermeasure network |
| CN116756515B (en) * | 2023-07-07 | 2024-02-23 | 广州大学 | Artificial seismic wave generation method based on deep convolution generation countermeasure network |
| CN118052558A (en) * | 2024-04-15 | 2024-05-17 | 万联易达物流科技有限公司 | Wind control model decision method and system based on artificial intelligence |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110361778B (en) | 2020-10-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110361778A (en) | A kind of Reconstruction of seismic data method based on generation confrontation network | |
| CN101303764B (en) | Method for self-adaption amalgamation of multi-sensor image based on non-lower sampling profile wave | |
| CN109490957A (en) | A kind of Reconstruction of seismic data method based on space constraint compressed sensing | |
| CN106384092B (en) | Online low-rank anomalous video event detecting method towards monitoring scene | |
| CN109410114B (en) | Compressed Sensing Image Reconstruction Algorithm Based on Deep Learning | |
| CN109165660A (en) | A kind of obvious object detection method based on convolutional neural networks | |
| CN111046737B (en) | Efficient intelligent sensing acquisition method for microseism signal detection | |
| CN103325092B (en) | A kind of method generating two-dimensional phase disentanglement quality picture and device | |
| CN107729926A (en) | A kind of data amplification method based on higher dimensional space conversion, mechanical recognition system | |
| CN107179550B (en) | A kind of seismic signal zero phase deconvolution method of data-driven | |
| CN113935249B (en) | Upper-layer ocean thermal structure inversion method based on compression and excitation network | |
| CN115238550B (en) | Land electric field numerical simulation calculation method for landslide rainfall of self-adaptive unstructured grid | |
| CN107944353A (en) | SAR image change detection based on profile ripple BSPP networks | |
| CN105607122A (en) | Seismic texture extraction and enhancement method based on total variation seismic data decomposition model | |
| CN111612906A (en) | Method and system for generating three-dimensional geological model and computer storage medium | |
| CN109859131A (en) | A kind of image recovery method based on multi-scale self-similarity Yu conformal constraint | |
| CN115114841A (en) | U-net frame-based subsurface temperature field reconstruction method for high spatial-temporal resolution of edge sea | |
| Chen et al. | Robust interpolation of DEMs from LiDAR-derived elevation data | |
| CN115238565A (en) | Resistivity model reconstruction network training method, electromagnetic inversion method and device | |
| CN116091492B (en) | Image change pixel level detection method and system | |
| CN105931184B (en) | SAR image super-resolution method based on combined optimization | |
| Wang et al. | Accelerated reconstruction of electrical impedance tomography images via patch based sparse representation | |
| Gallego et al. | Variational stereo imaging of oceanic waves with statistical constraints | |
| CN109782346B (en) | Acquisition footprint pressing method based on morphological component analysis | |
| CN112242001B (en) | Model parameter disturbance method for wavelet transformation |
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 |