CN112965113B - Method for improving signal-to-noise ratio of seismic data - Google Patents
Method for improving signal-to-noise ratio of seismic data Download PDFInfo
- Publication number
- CN112965113B CN112965113B CN202110138136.9A CN202110138136A CN112965113B CN 112965113 B CN112965113 B CN 112965113B CN 202110138136 A CN202110138136 A CN 202110138136A CN 112965113 B CN112965113 B CN 112965113B
- Authority
- CN
- China
- Prior art keywords
- data
- seismic
- denoising
- seismic data
- signal
- 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 76
- 238000012549 training Methods 0.000 claims abstract description 43
- 239000013307 optical fiber Substances 0.000 claims abstract description 42
- 238000013528 artificial neural network Methods 0.000 claims abstract description 35
- 230000006872 improvement Effects 0.000 claims abstract description 4
- 230000003287 optical effect Effects 0.000 claims description 12
- 238000012935 Averaging Methods 0.000 claims description 10
- 238000012545 processing Methods 0.000 claims description 8
- 238000004364 calculation method Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims description 2
- 230000009467 reduction Effects 0.000 abstract description 4
- 230000006870 function Effects 0.000 description 30
- 230000008569 process Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000009529 body temperature measurement Methods 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 230000005284 excitation Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 239000000654 additive Substances 0.000 description 1
- 230000000996 additive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 239000011800 void material Substances 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
- G01V1/36—Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/30—Noise handling
- G01V2210/32—Noise reduction
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a seismic data signal-to-noise ratio improving method, which comprises the following steps: s1: acquiring a plurality of data records of seismic data with noise for the same seismic signal by using a distributed optical fiber sensing seismograph; s2: training the neural network by using a noise2noise denoising method or a noise2void denoising method; s3: and denoising the plurality of data records by using the trained neural network to finish the improvement of the signal-to-noise ratio of the seismic data. The invention utilizes the advantages that the optical fiber sensing system with multiple channels, multiple wavelengths, multiple frequencies or multiple codes can record the same earthquake signal for multiple times, combines the noise2noise and noise2void denoising methods without clean signals, reduces the requirements on data samples when the signals are denoised, and can be combined with the traditional accumulated average denoising method to realize the efficient noise reduction of the earthquake data.
Description
Technical Field
The invention belongs to the technical field of seismic data denoising, and particularly relates to a seismic data signal-to-noise ratio improving method.
Background
Because of the characteristics of the distributed acoustic wave sensing technology and the characteristics of the optical fiber, which are very sensitive and easy to be interfered, the seismic data collected by the technology is often accompanied with more noise, so that the denoising link of the seismic data is an important bottleneck for the further development of the technology.
There are some problems with the traditional noise reduction of seismic data: the seismic data are huge, the algorithm iteration times are more, and the complexity is higher; or the local analysis is poor, and effective signals are distorted after denoising, so that the original amplitude characteristics of the seismic signals are destroyed, and the signal fidelity is reduced. However, the conventional data denoising method based on machine learning also needs to use clean data as target data for network training, but the acquisition of clean data is very difficult due to the characteristic problem of DAS technology, so that the method has certain limitation in realization and use.
Disclosure of Invention
The invention aims to solve the problems of poor universality and complex process of the existing denoising method, and provides a method for improving the signal-to-noise ratio of seismic data.
The technical scheme of the invention is as follows: the seismic data signal-to-noise ratio improving method comprises the following steps:
S1: acquiring a plurality of data records of seismic data with noise for the same seismic signal by using a distributed optical fiber sensing seismograph;
S2: training the neural network by using a noise2noise denoising method or a noise2void denoising method;
s3: and denoising the plurality of data records by using the trained neural network to finish the improvement of the signal-to-noise ratio of the seismic data.
The beneficial effects of the invention are as follows:
(1) The invention utilizes the advantages that the optical fiber sensing system with multiple channels, multiple wavelengths, multiple frequencies or multiple codes can record the same earthquake signal for multiple times, combines the advantages that the noise2noise and the noise2void denoising method do not need clean signals and the required sample data quantity is small, reduces the requirement on data samples when the signals are denoised, and can be combined with the traditional accumulated average denoising method to realize the efficient noise reduction of the earthquake data. The method is not only suitable for denoising the seismic data, but also suitable for denoising other data of the same type, and can be popularized to a distributed optical fiber temperature measurement and strain sensing system to improve the signal-to-noise ratio of the sensing data.
(2) According to the invention, on the basis of the traditional neural network-based data denoising method, the acquisition mode of seismic data is changed, the data acquisition of a network training set and the selection of a loss function are changed, the defect that the original neural network-based denoising method needs to acquire clean data in advance as a training target is overcome, the denoising step is simplified, the universality is further enhanced, and the denoising process of the seismic data is more accurate, efficient and intelligent. The noise2noise denoising method is used for denoising collected noisy signals of a determined target in different noise environments. The seismic data denoising by using the noise2void realizes denoising work of the time-varying signal under different noise environments.
Further, in step S1, the method for acquiring the seismic data with noise includes: laying an optical cable along the optical fiber logging, and receiving seismic signals by using the optical cable; and (3) carrying out seismic data acquisition on the seismic signals by adopting a distributed optical fiber sensing seismograph, and acquiring a group of seismic data pairs with noise for the same seismic signal.
Further, in step S1, the method for acquiring the seismic data for each seismic signal includes: a dual-channel optical fiber sensing seismometer is used for collecting two core optical fibers in an optical cable simultaneously; or using a dual-wavelength, dual-frequency or dual-code distributed optical fiber sensing seismograph to inject two wavelengths, frequencies and coded light into the same optical fiber, and acquiring each wavelength to obtain a pair of seismic data with noise; or a double-code distributed optical fiber sensing seismograph is used, two coded lights are injected into the same optical fiber, each wavelength is acquired, and a pair of seismic data with noise are obtained.
Further, in step S2, if the pair of seismic data corresponding to the seismic signal is noisy, the method for training the neural network is as follows: the objective function is trained using one of the seismic data (x i,yi) and the output results are averaged.
Further, in step S3, a pair of seismic data acquired for the same seismic signal is used as input data, one is used as target data, noise2noise denoising is used for denoising, or noise2void denoising is used for denoising two seismic data respectively, and the two processed data are averaged.
Further, in step S3, the method for acquiring the seismic data for each seismic signal includes: using a multi-channel, multi-wavelength, multi-frequency or multi-code distributed optical fiber sensing seismograph to collect n pieces of seismic data simultaneously for the same seismic signal, dividing the collected n pieces of seismic data into n/2 seismic data pairs, denoising each seismic data pair by using a noise2noise denoising method, and averaging the denoised data;
For the case of more channels, n/2 data pairs processed by a noise2noise denoising method are divided into n/4 data pairs, and n2n denoising processing is performed and then the average is performed.
The noise2void denoising method is used for denoising the acquired n seismic data respectively, and then averaging is carried out;
And exciting the seismic signals excited by the same mode at the same position, wherein the signals are basically the same as noise characteristics, exciting the seismic signals n times at one position, carrying out data acquisition on the seismic signals excited each time, denoising, and averaging the n denoised seismic data obtained after processing to obtain the final denoised seismic signals.
Further, in step S2, the noise2void denoising method is used to implement image denoising, and the method for training the neural network is as follows: and selecting a data block with N dimension from the seismic data pair as the input of the training neural network, replacing the central data with random data in the central data field to obtain target data, and training and optimizing the target function by using the target data to finish the denoising of the seismic data.
Drawings
FIG. 1 is a flow chart of a method of denoising seismic data.
Detailed Description
Embodiments of the present invention are further described below with reference to the accompanying drawings.
Before describing particular embodiments of the present invention, in order to make the aspects of the present invention more apparent and complete, abbreviations and key term definitions appearing in the present invention will be described first:
noise2noise: the abbreviation is noise to noise, which means training a neural network in the case where both input and output are noisy signals, abbreviated as noise2noise. One paper derived from ICML2018 was published jointly by researchers from the university of inflight, alto and MIT.
Noise2void: the abbreviation is meant noise to void, which refers to training a neural network in the presence of only one noisy signal, referred to as noise2void.
As shown in FIG. 1, the invention provides a method for improving the signal-to-noise ratio of seismic data, which comprises the following steps:
S1: acquiring a plurality of data records of seismic data with noise for the same seismic signal by using a distributed optical fiber sensing seismograph;
S2: training the neural network by using a noise2noise denoising method or a noise2void denoising method;
s3: and denoising the plurality of data records by using the trained neural network to finish the improvement of the signal-to-noise ratio of the seismic data.
In the embodiment of the present invention, as shown in fig. 1, in step S1, a method for acquiring seismic data with noise includes: laying an optical cable along the optical fiber logging, and receiving seismic signals by using the optical cable; and (3) carrying out seismic data acquisition on the seismic signals by adopting a distributed optical fiber sensing seismograph, and acquiring a group of seismic data pairs with noise for the same seismic signal.
In the embodiment of the present invention, as shown in fig. 1, in step S1, a method for acquiring seismic data for each seismic signal includes: a dual-channel optical fiber sensing seismometer is used for collecting two core optical fibers in an optical cable simultaneously; or using a dual-wavelength, dual-frequency or dual-code distributed optical fiber sensing seismograph to inject two wavelengths, frequencies and coded light into the same optical fiber, and acquiring each wavelength to obtain a pair of seismic data with noise; or a double-code distributed optical fiber sensing seismograph is used, two coded lights are injected into the same optical fiber, each wavelength is acquired, and a pair of seismic data with noise are obtained.
In the embodiment of the present invention, as shown in fig. 1, in step S2, if all the seismic data pairs corresponding to the seismic signals have noise, the method for training the neural network is as follows: the objective function is trained using one of the seismic data (x i,yi) and the output results are averaged.
In the embodiment of the present invention, as shown in fig. 1, in step S3, a pair of seismic data acquired for the same seismic signal is used as input data, one is used as target data, noise2noise denoising is used for denoising, or noise2void denoising is used for denoising two seismic data respectively, and the two processed data are averaged.
In the embodiment of the present invention, as shown in fig. 1, in step S3, the method for acquiring seismic data for each seismic signal includes: using a multi-channel, multi-wavelength, multi-frequency or multi-code distributed optical fiber sensing seismograph to collect n pieces of seismic data simultaneously for the same seismic signal, dividing the collected n pieces of seismic data into n/2 seismic data pairs, denoising each seismic data pair by using a noise2noise denoising method, and averaging the denoised data;
For the case of more channels, n/2 data pairs processed by a noise2noise denoising method are divided into n/4 data pairs, and n2n denoising processing is performed and then the average is performed.
The noise2void denoising method is used for denoising the acquired n seismic data respectively, and then averaging is carried out;
And exciting the seismic signals excited by the same mode at the same position, wherein the signals are basically the same as noise characteristics, exciting the seismic signals n times at one position, carrying out data acquisition on the seismic signals excited each time, denoising, and averaging the n denoised seismic data obtained after processing to obtain the final denoised seismic signals.
In the embodiment of the present invention, as shown in fig. 1, in step S3, if the seismic data pairs are noisy and the optimal solution of the loss function is at the arithmetic mean value, the training output result of the neural network is optimized averagely.
In the embodiment of the present invention, as shown in fig. 1, in step S2, image denoising is implemented by using a noise2void denoising method, and then the method for training the neural network is as follows: and selecting a data block with N dimension from the seismic data pair as the input of the training neural network, replacing the central data with random data in the central data field to obtain target data, and training and optimizing the target function by using the target data to finish the denoising of the seismic data.
In the invention, an optical cable is arranged along an optical fiber well logging, a seismic source generates seismic waves in a heavy hammer knocking mode, an excitation source is excited, the optical cable senses seismic signals to generate optical signals, and a distributed optical fiber sensing seismograph is used for data acquisition. For each seismic signal, the distributed optical fiber sensing seismograph can acquire N channels of data at one time, and a two-dimensional DAS seismic data is directly obtained and used without one-dimensional to two-dimensional processing. And, through the multichannel and multi-wavelength characteristic of distributed optical fiber sensing seismograph, can directly acquire the pair of noisy seismic data. According to this feature, the noise data may not be processed to obtain clean seismic data as target data for deep learning, but the image denoising method noise2noise without clean data may be applied to denoising of seismic data.
In the invention, training of the denoising network can be realized without obtaining a clean data set as a target data set through a complex data processing process. The data set acquired in the step of acquiring the seismic data is divided into two parts, namely an output signal training set and a target signal training set.
The characteristics of the seismic signals excited for a plurality of times at the same position are basically the same, so that the seismic signals after the denoising after the primary excitation can be performed, then the seismic signals after the denoising respectively are performed at the position and then the average treatment is performed, and the denoised seismic signals are obtained.
In the embodiment of the present invention, as shown in fig. 1, in step S3, if all the seismic data pairs corresponding to the seismic signals have noise, the noise2noise denoising method is used to implement image denoising, and the method for training the neural network is as follows: training an objective function by utilizing one of the seismic data (x i,yi), and carrying out average optimization on an output result; the expression of the objective function is:
Wherein argmin {.cndot }' represents the value operation of the corresponding variable when the objective function reaches the minimum value, Representing the mathematical expectation of the solving for x,The conditional expectation of y is expressed, L (·) is a loss function, f θ (x) is a network function, θ is a network parameter, x is input data, and y is output data.
In the embodiment of the present invention, as shown in fig. 1, in step S3, if the seismic data pairs are noisy and the optimal solution of the loss function is at the arithmetic mean value, the training output result of the neural network is optimized averagely, and the calculation formula is as follows:
where z represents noise-free data, y represents output data, Indicating the conditional expectation of y.
In the present invention, for a noisy signal, the corresponding noise-free signal has not only one possibility, but the output of the neural network is different due to the difference of training the loss function, and if the optimal solution of the loss function is at the arithmetic mean value, the result of the neural network output is the arithmetic mean value of all possible output results. At this time, y is replaced with y 2 whose noise is desired to be 0, so that the output is not changed, as long as the average value of y is unchanged.
The output data is ideal clean data when the data in a given training set is infinite, and the expected value of the output data is clean data when the training data is finite. As the amount of training data increases, the output results also more closely approximate noise-free data.
The above is the case when the optimal solution of the loss function is taken at the mean value, and the other loss functions are the same, and the specific choice of the loss function depends on the nature of the noise itself to be eliminated. The average value of the Gaussian white noise is zero, and the loss function is the best quality and has better denoising effect if the loss function is obtained at the average value.
In the embodiment of the present invention, as shown in fig. 1, in step S3, when the seismic signal corresponds to single seismic data, the expression of the objective function is:
Wherein argmin {. Cndot. } represents the value operation of the corresponding variable when the objective function reaches the minimum value, L (. Cndot.). Cndot.is the loss function, f (. Cndot.). Cndot.is the network function, θ is the network parameter, x is the input data, y is the output data, i is the first sum value, j is the second sum value, i and j are 1,2,3 … …, The input data is represented by a representation of the input data,Representing the target data.
In the present invention, if data of size n×n is taken as input from two-dimensional data, the network is trained with the center data of the data block as a target, and the network will learn to directly output the center value of the input data block. Therefore, in the use of the method, the central data of each input data block is replaced by random data in the neighborhood, so that the network is effectively prevented from learning the identity mapping. The noise signals are random signals, and under the condition that the seismic signals have correlation, the network cannot recover the noise signals at the central position from the noise in the neighborhood, but can recover the seismic data signals at the central position to a certain extent from the correlation of the seismic signals in the neighborhood. The input and target data may be derived from separate noisy seismic data. N data blocks are extracted from two-dimensional seismic data, and data in the center of the data blocks are replaced by surrounding random data to serve as input data.
In the embodiment of the invention, when the seismic signals are detected on site, the acquired seismic data are noisy seismic data, and in order to reduce noise of the noisy data and effectively suppress noise, proper loss functions are selected and optimized for different types of noise.
Gaussian noise belongs to a relatively common noise, and for gaussian noise, a loss function is selected because the average value of common additive white gaussian noise is 0: l (z, y) = (z-y) 2.
When the function is used as a loss function to train the neural network, the optimal solution of the function is obtained at the arithmetic mean value of y, and for the neural network trained by the loss function, the network learns the arithmetic mean value of all possible results, and the mean value of Gaussian white noise is 0, so that the noise of the output result of the neural network is theoretically 0, and clean seismic data is obtained. The result of this network training is ideal clean data given that there is an infinite number of data in the training set, and the expectation of the output data is clean given that there is a finite number of training data. As the amount of training data increases, the error in the output result will be smaller.
The working principle and the working process of the invention are as follows: firstly, a group of seismic data pairs with noise are collected by using a distributed optical fiber sensing seismograph; repeatedly acquiring a plurality of groups of seismic data pairs with noise, and taking the seismic data pairs as a data set for training a neural network; and training the neural network by utilizing a plurality of groups of seismic data pairs with noise to finish the denoising of the seismic data. Meanwhile, the seismic data signal-to-noise ratio improving method is not limited to a two-dimensional form.
The beneficial effects of the invention are as follows:
(1) The invention utilizes the advantages that the optical fiber sensing system with multiple channels, multiple wavelengths, multiple frequencies or multiple codes can record the same earthquake signal for multiple times, combines the advantages that the noise2noise and the noise2void denoising method do not need clean signals and the required sample data quantity is small, reduces the requirement on data samples when the signals are denoised, and can be combined with the traditional accumulated average denoising method to realize the efficient noise reduction of the earthquake data. The method is not only suitable for denoising the seismic data, but also suitable for denoising other data of the same type, and can be popularized to a distributed optical fiber temperature measurement and strain sensing system to improve the signal-to-noise ratio of the sensing data.
(2) According to the invention, on the basis of the traditional neural network-based data denoising method, the acquisition mode of seismic data is changed, the data acquisition of a network training set and the selection of a loss function are changed, the defect that the original neural network-based denoising method needs to acquire clean data in advance as a training target is overcome, the denoising step is simplified, the universality is further enhanced, and the denoising process of the seismic data is more accurate, efficient and intelligent. The noise2noise denoising method is used for denoising collected noisy signals of a determined target in different noise environments. The seismic data denoising by using the noise2void realizes denoising work of the time-varying signal under different noise environments.
Those of ordinary skill in the art will recognize that the embodiments described herein are for the purpose of aiding the reader in understanding the principles of the present invention and should be understood that the scope of the invention is not limited to such specific statements and embodiments. Those of ordinary skill in the art can make various other specific modifications and combinations from the teachings of the present disclosure without departing from the spirit thereof, and such modifications and combinations remain within the scope of the present disclosure.
Claims (3)
1. The method for improving the signal-to-noise ratio of the seismic data is characterized by comprising the following steps of:
S1: acquiring a plurality of data records of seismic data with noise for the same seismic signal by using a distributed optical fiber sensing seismograph;
S2: training the neural network by using a noise2noise denoising method or a noise2void denoising method;
S3: denoising the plurality of data records by using the trained neural network to finish the improvement of the signal-to-noise ratio of the seismic data;
In the step S1, the method for acquiring the seismic data with noise includes: laying an optical cable along the optical fiber logging, and receiving seismic signals by using the optical cable; the method comprises the steps that a distributed optical fiber sensing seismograph is adopted to collect seismic data of seismic signals, and a group of seismic data pairs with noise are collected and obtained for the same seismic signal;
In the step S1, the method for acquiring the seismic data for each seismic signal includes: a dual-channel optical fiber sensing seismometer is used for collecting two core optical fibers in an optical cable simultaneously; or using a dual-wavelength, dual-frequency or dual-code distributed optical fiber sensing seismograph to inject two wavelengths, frequencies and coded light into the same optical fiber, and acquiring each wavelength to obtain a pair of seismic data with noise; or a double-coding distributed optical fiber sensing seismograph is used, two coded lights are injected into the same optical fiber, each wavelength is collected, and a pair of seismic data with noise are obtained;
In the step S3, one pair of seismic data acquired for the same seismic signal is used as input data, one pair of seismic data is used as target data, noise2noise denoising is used for denoising, or noise2void denoising is used for denoising two seismic data respectively, and the two processed data are averaged;
In the step S3, the method for acquiring the seismic data for each seismic signal includes: using a multi-channel, multi-wavelength, multi-frequency or multi-code distributed optical fiber sensing seismograph to collect n pieces of seismic data simultaneously for the same seismic signal, dividing the collected n pieces of seismic data into n/2 seismic data pairs, denoising each seismic data pair by using a noise2noise denoising method, and averaging the denoised data;
for the case of more channels, dividing n/2 data pairs processed by a noise2noise denoising method into n/4 data pairs, and carrying out n2n denoising processing and averaging;
The noise2void denoising method is used for denoising the acquired n seismic data respectively, and then averaging is carried out;
exciting the seismic signals at the same position in the same mode, wherein the signals are basically the same as noise characteristics, exciting the seismic signals n times at one position, carrying out data acquisition on each excited seismic signal, denoising, and averaging the n denoised seismic data obtained after processing to obtain a final denoised seismic signal;
In the step S3, if the pair of seismic data corresponding to the seismic signal is noisy, the noise2noise denoising method is used to denoise the image, and the method for training the neural network is as follows: training an objective function by utilizing one of the seismic data (x i,yi), and carrying out average optimization on an output result; the expression of the objective function is:
Wherein argmin {.cndot }' represents the value operation of the corresponding variable when the objective function reaches the minimum value, Representing the mathematical expectation of the solving for x,Representing a conditional expectation of y, L (·) representing a loss function, f θ (x) representing a network function with respect to θ, θ representing a network parameter, x representing input data, y representing output data;
In the step S3, if the seismic data pairs are noisy and the optimal solution of the loss function is at the arithmetic mean value, the training output result of the neural network is optimized averagely, and the calculation formula is as follows:
Where z represents noise-free data, Representing the conditional expectation of y;
in the step S3, when the seismic signal corresponds to the single seismic data, the expression of the objective function is:
wherein f (·) represents a network function, i represents a first summation value, j represents a second summation value, i and j take 1,2,3 … …, Representing the input of a single seismic data item,Representing the target data.
2. The method for improving the signal-to-noise ratio of the seismic data according to claim 1, wherein in the step S2, if the pair of seismic data corresponding to the seismic signal is noisy, the method for training the neural network is as follows: the objective function is trained using one of the seismic data (x i,yi) and the output results are averaged.
3. The method for improving the signal-to-noise ratio of the seismic data according to claim 1, wherein in the step S2, the noise2void denoising method is used to implement image denoising, and the method for training the neural network is as follows: and selecting a data block with N dimension from the seismic data pair as the input of the training neural network, replacing the central data with random data in the central data field to obtain target data, and training and optimizing the target function by using the target data to finish the denoising of the seismic data.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110138136.9A CN112965113B (en) | 2021-02-01 | 2021-02-01 | Method for improving signal-to-noise ratio of seismic data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110138136.9A CN112965113B (en) | 2021-02-01 | 2021-02-01 | Method for improving signal-to-noise ratio of seismic data |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112965113A CN112965113A (en) | 2021-06-15 |
| CN112965113B true CN112965113B (en) | 2024-07-02 |
Family
ID=76272941
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110138136.9A Active CN112965113B (en) | 2021-02-01 | 2021-02-01 | Method for improving signal-to-noise ratio of seismic data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112965113B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113538260B (en) * | 2021-06-21 | 2022-04-12 | 复旦大学 | LDCT image denoising and classifying method for self-supervision and supervised combined training |
| CN113484913B (en) * | 2021-08-16 | 2023-06-16 | 成都理工大学 | Seismic data denoising method for multi-granularity feature fusion convolutional neural network |
| US20230105075A1 (en) * | 2021-10-06 | 2023-04-06 | Cgg Services Sas | Seismic interference noise attenuation using dnn |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110058305A (en) * | 2019-05-24 | 2019-07-26 | 吉林大学 | A kind of DAS seismic data noise-reduction method based on convolutional neural networks |
| DE102020101525A1 (en) * | 2019-01-24 | 2020-07-30 | Nvidia Corporation | BLIND-SPOT FOLDING ARCHITECTURES AND BAYESE IMAGE RECOVERY |
Family Cites Families (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8924158B2 (en) * | 2010-08-09 | 2014-12-30 | Schlumberger Technology Corporation | Seismic acquisition system including a distributed sensor having an optical fiber |
| CN108932480B (en) * | 2018-06-08 | 2022-03-15 | 电子科技大学 | Distributed optical fiber sensing signal feature learning and classifying method based on 1D-CNN |
| CN109782339A (en) * | 2019-01-14 | 2019-05-21 | 西安交通大学 | A kind of poststack three dimensional seismic data stochastic noise suppression method based on 3D-DnCNN network |
| US20200284937A1 (en) * | 2019-03-04 | 2020-09-10 | Chevron U.S.A. Inc. | System and method for displaying seismic events in distributed acoustic sensing data |
| US11802983B2 (en) * | 2019-05-24 | 2023-10-31 | Halliburton Energy Services, Inc. | Distributed acoustic sensing to geophone seismic data processing |
| CN110490823A (en) * | 2019-08-14 | 2019-11-22 | 北京大学深圳研究生院 | A kind of image de-noising method under true environment |
| CN110703316B (en) * | 2019-10-23 | 2021-08-10 | 电子科技大学 | Optical fiber ground seismic wave detection method and system |
| CN111539879B (en) * | 2020-04-15 | 2023-04-14 | 清华大学深圳国际研究生院 | Video blind denoising method and device based on deep learning |
| CN111580161A (en) * | 2020-05-21 | 2020-08-25 | 长江大学 | Earthquake random noise suppression method based on multi-scale convolution self-coding neural network |
| CN112130195A (en) * | 2020-10-13 | 2020-12-25 | 中油奥博(成都)科技有限公司 | Time-shifting VSP data acquisition system and method based on distributed optical fiber acoustic sensing |
-
2021
- 2021-02-01 CN CN202110138136.9A patent/CN112965113B/en active Active
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| DE102020101525A1 (en) * | 2019-01-24 | 2020-07-30 | Nvidia Corporation | BLIND-SPOT FOLDING ARCHITECTURES AND BAYESE IMAGE RECOVERY |
| CN110058305A (en) * | 2019-05-24 | 2019-07-26 | 吉林大学 | A kind of DAS seismic data noise-reduction method based on convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 采用无干净标签的深度学习方法衰减地震噪声;王函等;《2020年中国地球科学联合学术年会论文集（十四）》;第1336-1337页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112965113A (en) | 2021-06-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112965113B (en) | Method for improving signal-to-noise ratio of seismic data | |
| CN109858408B (en) | Ultrasonic signal processing method based on self-encoder | |
| CN108550116A (en) | The self-adapting random resonant denoising method of silicon monocrystal growth image under low signal-to-noise ratio | |
| CN115326783B (en) | Raman spectrum preprocessing model generation method, system, terminal and storage medium | |
| CN113484257B (en) | Water quality concentration calculating system and method based on neural network and full spectrum absorbance | |
| CN105445801A (en) | Processing method for eliminating random noises of two dimensional seismic data | |
| CN103364826A (en) | An earthquake blind source deconvolution method based on independent component analysis | |
| CN115905805A (en) | DAS data multi-scale noise reduction method based on global information judgment GAN | |
| CN116576956A (en) | Multisource vibration signal separation method based on distributed optical fiber acoustic wave sensing | |
| CN116662738A (en) | Distributed vibration sensing signal denoising method, electronic device and storage medium | |
| Harsuko et al. | Optimizing a Transformer-based network for a deep learning seismic processing workflow | |
| CN106950597B (en) | Mixing source data separation method based on the filtering of three sides | |
| CN113219539B (en) | Method and device for removing borehole wave interference in optical fiber acoustic sensing seismic data | |
| Shao et al. | Noise suppression of distributed acoustic sensing vertical seismic profile data based on time–frequency analysis | |
| CN112782763B (en) | Seismic quality factor estimation method, device, equipment and storage medium | |
| Dumont et al. | Deep learning on real geophysical data: A case study for distributed acoustic sensing research | |
| CN117972372A (en) | Intelligent DAS and method for separating aliasing signal orientation targets thereof | |
| CN112241024B (en) | Method for improving signal-to-noise ratio of seismic data, computer storage medium and system | |
| CN111323821A (en) | Shaping filtering method for sea surface fluctuation effect correction in marine seismic data | |
| Wang et al. | Self-Supervised Pre-Training Transformer for Seismic Data Denoising | |
| CN111856559A (en) | Multi-channel seismic spectrum inversion method and system based on sparse Bayes learning theory | |
| CN108627237A (en) | A kind of autocorrelation analysis signal processing method based on distributed optical fiber sensing system | |
| Tian et al. | Efficient SPSNet for downhole weak DAS signals recovery | |
| Mortazavi et al. | Adaptive tube wave suppression and elastic down-going and up-going separation in VSP data using pyramidal dual-tree directional filter bank | |
| CN113126163B (en) | Five-dimensional seismic data noise attenuation method and device |
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 |