CN116520429A - Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension - Google Patents

Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension Download PDF

Info

Publication number
CN116520429A
CN116520429A CN202310411730.XA CN202310411730A CN116520429A CN 116520429 A CN116520429 A CN 116520429A CN 202310411730 A CN202310411730 A CN 202310411730A CN 116520429 A CN116520429 A CN 116520429A
Authority
CN
China
Prior art keywords
data
low
frequency
band
self
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202310411730.XA
Other languages
Chinese (zh)
Inventor
王本锋
莫侗桐
孙雪奕
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.)
Tongji University
Original Assignee
Tongji University
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 Tongji University filed Critical Tongji University
Priority to CN202310411730.XA priority Critical patent/CN116520429A/en
Publication of CN116520429A publication Critical patent/CN116520429A/en
Pending legal-status Critical Current

Links

Classifications

    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28Processing seismic data, e.g. for interpretation or for event detection
    • G01V1/36Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy
    • G01V1/364Seismic filtering
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28Processing seismic data, e.g. for interpretation or for event detection
    • G01V1/32Transforming one recording into another or one representation into another
    • G01V1/325Transforming one representation into another
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V1/00Seismology; Seismic or acoustic prospecting or detecting
    • G01V1/28Processing seismic data, e.g. for interpretation or for event detection
    • G01V1/34Displaying seismic recordings or visualisation of seismic data or attributes
    • G01V1/345Visualisation of seismic data or attributes, e.g. in 3D cubes
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/10Pre-processing; Data cleansing
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V2210/00Details of seismic processing or analysis
    • G01V2210/20Trace signal pre-filtering to select, remove or transform specific events or signal components, i.e. trace-in/trace-out
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V2210/00Details of seismic processing or analysis
    • G01V2210/40Transforming data representation
    • G01V2210/45F-x or F-xy domain
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01VGEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
    • G01V2210/00Details of seismic processing or analysis
    • G01V2210/70Other details related to processing
    • G01V2210/74Visualisation of seismic data

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Physics & Mathematics (AREA)
  • Remote Sensing (AREA)
  • Theoretical Computer Science (AREA)
  • Data Mining & Analysis (AREA)
  • Geology (AREA)
  • General Life Sciences & Earth Sciences (AREA)
  • Geophysics (AREA)
  • Evolutionary Computation (AREA)
  • Artificial Intelligence (AREA)
  • Environmental & Geological Engineering (AREA)
  • General Engineering & Computer Science (AREA)
  • Acoustics & Sound (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Geophysics And Detection Of Objects (AREA)

Abstract

The invention relates to a self-supervision seismic data interpolation reconstruction method, a system and equipment based on band extension, wherein the method comprises the following steps: sparse observation data containing space aliasing is input, and low-frequency observation data without space aliasing is obtained through low-pass filtering processing; the method comprises the steps of obtaining encrypted low-frequency data in an analytic mode from low-frequency data without space aliasing; splitting the encrypted low frequency data according to the observation channel and the missing channel, and constructing a self-adaptive training set and a testing set for band prolongation; training the U-net network using the adaptive training set; inputting the test set data into a trained U-net network to obtain full-band data on a missing channel; and recombining the reconstructed missing channel full-band data with the original sparse observation data to obtain a final interpolation reconstruction result. Compared with the prior art, the method and the system for achieving the rule missing seismic data interpolation reconstruction are independent of tag data, high-efficiency and high-precision completion of the rule missing seismic data interpolation reconstruction is achieved in a self-supervision mode, and a training network has certain expansibility.

Description

Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension
Technical Field
The invention relates to the technical field of self-supervision data reconstruction, in particular to a frequency band continuation-based self-supervision seismic data interpolation reconstruction method applicable to rule deletion.
Background
The high-density seismic data without space aliasing has important significance for acquiring a high-quality migration imaging section, a high-precision inversion result and accurately describing oil reservoir distribution in subsequent seismic data processing. Conventional seismic data acquisition is limited by limited acquisition cost and complex construction environment, and the acquired data has large track spacing and often contains space aliasing. Therefore, efficient and high-precision interpolation reconstruction of seismic data becomes an important part of seismic data processing.
The conventional seismic data interpolation reconstruction method can be divided into a prediction filtering method, a low-rank method, a compressed sensing method based on sparse transformation and a wave equation method. However, for the situation of rule missing, the low-rank and compressed sensing methods are not applicable, and the algorithm is increased with the increase of the data scale, so that the calculation cost is remarkably increased. With the development of computer science and technology in recent years, the deep learning algorithm is widely focused in the seismic data interpolation reconstruction field, but the supervised deep learning method is based on the fact that tag data is used as a training sample, and the tag data is difficult to obtain in the actual seismic acquisition process; the network for simulating data training generally has the problems of low generalization performance, low interpolation reconstruction precision and the like in the actual data application process.
Disclosure of Invention
The invention aims to overcome the defects of the prior art and provide a self-supervision seismic data interpolation reconstruction method based on band extension.
The aim of the invention can be achieved by the following technical scheme:
as a first aspect of the present invention, there is provided a self-supervised seismic data interpolation reconstruction method based on band extension, the method comprising the steps of:
sparse observation data containing space aliasing is input, and low-frequency observation data without space aliasing is obtained through low-pass filtering processing;
based on Nyquist sampling law, acquiring encrypted low-frequency data from the low-frequency data without spatial aliasing in an analytic mode;
splitting the encrypted low frequency data according to the observation channel and the missing channel, and constructing a self-adaptive training set and a testing set for band prolongation;
training the U-net network using the constructed adaptive training set such that the network has a nonlinear mapping capability that maps low frequency data to full frequency band data;
inputting the low frequency data in the test set into a trained U-net network to obtain full frequency band data on a missing channel;
and recombining the reconstructed missing channel full-band data with the original sparse observation data to obtain a final interpolation reconstruction result.
Further, the step of obtaining the low-frequency observation data without spatial aliasing comprises the following steps:
conversion of sparsely acquired observation data D (x, t) to f-k domain D (f, k) using two-dimensional Fourier transform x );
Selecting a spatial aliasing starting frequency f in the f-k domain aliase And setting the initial frequency f of Hamming window cut A frequency low pass filter is constructed as follows:
performing frequency domain low-pass filtering on the sparsely acquired data D to obtain low-frequency data without spatial aliasing
In the formula, +..
Further, the step of obtaining the encrypted low frequency data includes:
based on Nyquist sampling law, the f-k spectrum corresponding to the encrypted low frequency data is obtainedAnalytical characterization can be as follows:
in the method, in the process of the invention,representing low frequency data without spatial aliasing, k x Represents horizontal wavenumber, k Nyquist Representing Nyquist wavenumbers;
for a pair ofPerforming two-dimensional inverse Fourier transform to obtain encrypted low-frequency space-free artificial frequency data +.>
Further, the adaptive training set includes:
the observation channel low frequency data of the encrypted low frequency data is taken as a sample input;
and the original acquired full-band sparse observation data is used as a label.
Further, the test set includes unlabeled low frequency data on the missing tracks.
Further, the expected output of the U-net network is full-band data, and the following loss function is constructed to perform network training:
Loss=||d obs -f(d low ,θ)|| 1
wherein d obs Is the observation data of the whole frequency band, d low For the input of low frequency data to the network, f (·) is the designed U-net network, θ is a parameter of the network and, I.I 1 Is L 1 And (5) norm constraint.
Further, the U-net network quantitatively evaluates the reconstruction result by adopting the following signal-to-noise ratio formula:
wherein y is tag data,for the result of the network reconstruction, I.I F Is the F-norm.
As a second aspect of the present invention, there is provided a self-supervising seismic data interpolation reconstruction system based on band extension, the system comprising:
the low-pass filtering processing module: the method comprises the steps of obtaining low-frequency observation data without space aliasing by low-pass filtering processing on sparse observation data with space aliasing;
low frequency data encryption module: based on Nyquist sampling law, low-frequency data without space aliasing is obtained in an analytic mode;
the data set construction module: the method comprises the steps of splitting an encrypted low-frequency data according to an observation channel and a missing channel, and constructing an adaptive training set and a test set for band extension;
u-net network training module: training the U-net network using the constructed adaptive training set such that the network has a nonlinear mapping capability that maps low frequency data to full frequency band data;
and the missing channel full-band data reconstruction module is used for: inputting the low frequency data in the test set into a trained U-net network to obtain full frequency band data on a missing channel;
interpolation reconstruction module: and the method is used for recombining the reconstructed missing channel full-band data with the original sparse observation data to obtain a final interpolation reconstruction result.
As a third aspect of the present invention, there is provided an electronic apparatus comprising:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the self-supervised seismic data interpolation reconstruction method based on band extension as described above.
As a fourth aspect of the present invention, there is provided a computer readable storage medium storing a computer program which, when executed by a processor, implements the steps of a self-supervised seismic data interpolation reconstruction method based on band extension as described above.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides a self-supervision seismic data interpolation reconstruction method based on band extension, which is suitable for rule missing seismic data interpolation reconstruction. The method converts the problem of seismic data interpolation reconstruction into two sub-problems of low-frequency component reconstruction and band prolongation, and firstly analyzes and acquires encrypted low-frequency seismic data through low-pass filtering and Nyquist sampling law; and acquiring full-band data on the missing channel in a self-supervision learning mode, thereby completing seismic data interpolation reconstruction. The scheme provided by the invention is independent of label data, the rule missing seismic data interpolation reconstruction is efficiently and highly accurately completed in a self-supervision mode, and the training network has certain expansibility.
Drawings
FIG. 1 is a flow chart of a method for self-monitoring seismic data interpolation reconstruction based on band extension of the present invention;
FIG. 2 is a flow chart of the adaptive training set construction and data separation operation of the present invention;
FIG. 3 is a schematic diagram of a U-net network framework for band extension according to an embodiment of the present invention;
FIG. 4 is a diagram showing the interpolation result of the low frequency component of the analog data according to the present invention; (a) complete data; (b) regularly missing 2/3 data; (c) low frequency integrity data; (d) observing low frequency data on the track; (e-h) represents the f-k spectrum corresponding to the (a-d) data;
FIG. 5 is a converging curve of the network training set validation set loss function and the reconstructed signal-to-noise ratio of the present invention;
FIG. 6 is a graph showing the comparison of the results of interpolation reconstruction of the 80 th shot data according to an embodiment of the present invention; (a) raw complete data; (b) raw missing data; (c) the method of this patent reconstructs the result (snr=25.3 dB); (d) residual error corresponding to figure (c); (e) DIP method reconstruction result (snr=15.8 dB); (f) residual error corresponding to the graph (e).
Detailed Description
The invention will now be described in detail with reference to the drawings and specific examples. The present embodiment is implemented on the premise of the technical scheme of the present invention, and a detailed implementation manner and a specific operation process are given, but the protection scope of the present invention is not limited to the following examples.
Aiming at the interpolation reconstruction problem of the seismic data with rule missing, the invention provides a self-supervision interpolation reconstruction method based on band extension. As shown in fig. 1, the interpolation reconstruction method is implemented by:
(1) And (3) inputting sparse observation data containing space aliasing, and obtaining low-frequency observation data without space aliasing through low-pass filtering processing.
(2) Based on Nyquist sampling law, the encrypted low-frequency data is acquired in an analytic mode by the low-frequency data without space aliasing.
(3) Splitting the reconstructed low frequency data according to the observation channel and the missing channel.
(4) After splitting, the low-frequency data of the observation channel and the full-frequency data of the original observation channel are respectively used as sample input and labels to construct a self-adaptive training set for band extension, and the low-frequency data of the missing channel is used as a test set.
(5) Training the U-net network using the adaptive training set constructed in step (4) such that the network has a nonlinear mapping capability that maps low frequency data to full frequency band data.
(6) And inputting the low frequency data in the test set into a trained U-net network, thereby acquiring the full-band data on the missing channel.
(7) And recombining the reconstructed missing channel full-band data with the original sparse observation data, and further obtaining a final interpolation reconstruction result.
The method comprises the following specific steps:
low-frequency reconstruction based on low-pass filtering and Nyquist sampling law
Based on Shannon's sampling theorem, assuming that the seismic data trace spacing is deltax, its corresponding Nyquist wavenumber is shown in equation (1),
when the horizontal wave number is k x Greater than k Nyquist When the corresponding wave number component is folded, aliasing occurs, and spatial aliasing occurs in the f-k domain of the frequency-wave number domain. Because the channel interval deltax of the sparsely acquired seismic data is larger, the Nyquist wavenumber is smaller, and the sparse acquired seismic data has stronger spatial aliasing in the f-k domain. Since the apparent velocity of the seismic data has a certain spread range and the wave number k x The relation between the frequency f and the apparent velocity v is k x =f/v. Under the condition of a certain visual velocity, the wave number corresponding to the low-frequency component of the seismic data is smaller, and the spatial aliasing is not contained generally, so that the seismic data without the spatial aliasing can be obtained by carrying out low-pass filtering on the sparse acquisition data. Assuming sparsely acquired observation data is D (x, t), D is transformed into f-k domain D (f, k) using a two-dimensional Fourier transform x ),Representing a two-dimensional Fourier transform,
selecting a spatial aliasing starting frequency f in the f-k domain aliase And setting Hamming windowInitial frequency f cut And (3) constructing a frequency low-pass filter according to the formula (3) to avoid the Gibbs phenomenon.
Performing frequency domain low-pass filtering on the sparsely acquired data D according to a formula (4) to obtain low-frequency data without spatial aliasingAs indicated by the dot product operator,
based on Nyquist sampling law, any discrete signal without aliasing can obtain a continuous signal corresponding to the discrete signal in an analytic mode, so as to obtain an encrypted discrete signal. Assuming that the space between the encrypted seismic data channels is reduced to delta x/n, the corresponding Nyquist wave number becomes nk Nyquist . Based on Nyquist sampling law, the f-k spectrum corresponding to the encrypted low frequency data is obtainedIt can be analytically characterized as,
for a pair ofPerforming two-dimensional inverse Fourier transform to obtain encrypted low-frequency space-free artificial frequency data +.>
Band extension based on self-supervision deep learning
And obtaining the encrypted low-frequency space-free artificial-frequency seismic data with high precision in an analytic mode based on low-pass filtering and Nyquist sampling law, namely completing the interpolation reconstruction of the low-frequency component. After the interpolation of the low-frequency components is completed, the second processing step of the method is to acquire the encrypted full-frequency band data by using a self-supervision deep learning method in a band extension mode, namely, the interpolation reconstruction of the sparse seismic data is completed.
For the deep learning algorithm, constructing a reasonable training set is a necessary precondition for ensuring the effectiveness of the deep learning algorithm. For the problem of band extension, the encrypted low-frequency data is split according to the observation channel and the missing channel, the low-frequency data on the observation channel and the originally acquired full-band sparse observation data are respectively used as sample input and labels to construct a self-adaptive training set for band extension, and the unlabeled low-frequency data on the missing channel is used as a test set.
Training is performed by adopting the U-net network architecture designed by the figure 3, the input of the network is sparse low-frequency data, the expected output of the network is full-frequency band data, the following loss function is constructed for performing network training,
Loss=||d obs -f(d low ,θ)|| 1 . (7)
wherein d is obs Observation data (tag) of full frequency band d low For the input of low frequency data to the network, f (·) is the designed U-net network, θ is a parameter of the network and, I.I 1 Is L 1 Norms. Quantitatively evaluating the reconstruction result by adopting the following signal-to-noise ratio formula:
where y is the label data and where y is the label data,for the result of the network reconstruction, I.I F Is the F-norm.
After the network training is completed, the low-frequency data (test set) on the missing channel is input into the network, so that the full-frequency band data corresponding to the missing channel is output. And recombining the full-band data on the missing channel with the original observed data, and further obtaining encrypted seismic data, namely finishing interpolation reconstruction. The technical flow of the invention is shown in figure 1.
Experimental simulation results
256 co-shot gather are simulated by adopting a scalar acoustic wave equation, each shot gather comprises 240 channels of seismic data, each channel of seismic data comprises 401 time sampling points, the channel spacing is 12.5m, and the time sampling rate is 4ms. In order to simulate the condition of rule missing, 2/3 seismic channels are regularly missing for each common shot point gather, the channel spacing is enlarged to 37.5m, and the data after the rule missing are adopted as the original observation data for testing. Fig. 4 (a), 4 (b) show sections before and after the absence of seismic data, respectively; fig. 4 (e), 4 (f) respectively correspond to f-k spectra of the seismic data before and after the deletion, after the regular deletion, the track spacing of the data is enlarged, and the f-k spectra show spatial aliasing.
And (3) carrying out low-frequency component interpolation reconstruction on the regular missing seismic data containing the space aliasing based on the low-pass filtering and Nyquist sampling law to obtain complete low-frequency seismic data (fig. 4 (c) and 4 (g)). Separating the low frequency seismic data on the observation path from the missing path to obtain the low frequency seismic data on the observation path (fig. 4 (d)), and constructing the training pair for band extension from the full-band data on the original observation path (fig. 4 (a)).
The 256 common shot point gathers are used for constructing 256 training pairs, 236 common shot point gathers are uniformly selected as training sets to train the U-net network, and the rest 20 common shot point gathers are used as verification sets to verify the trained U-net network. In the training process, the reconstructed signal-to-noise ratio and the loss function convergence curve are shown in fig. 5, and after 250 iterations, the average signal-to-noise ratios of the training set and the verification set are respectively converged to 27.8dB and 25.7dB. And inputting the low frequency data on the missing channel into a trained network, outputting full-band seismic data reconstructed by the missing channel, and recombining the full-band seismic data with the observed data on the original observed channel, thereby completing the interpolation reconstruction of the seismic data. FIG. 6 shows the result of interpolation reconstruction of the 80 th shot data in the dataset for detailed analysis and comparison with Depth Image Prior (DIP) deep learning methods. FIG. 6 (a) shows the complete data without the missing, and FIG. 6 (b) shows the seismic data after 2/3 of the regular missing; after the processing of the patent method, the interpolation reconstruction result and the corresponding residual error are shown in fig. 6 (c), 6 (d), and the reconstruction signal-to-noise ratio reaches 25.3dB; the processing result and residual error of the DIP method are shown in fig. 6 (e), 6 (f), the reconstruction signal-to-noise ratio is 15.8dB, which is lower than that of the present patent method, and the corresponding residual error is larger than that of the present patent method, so that the validity of the present patent method is further verified, and the seismic data after interpolation reconstruction is favorable for improving the precision of the subsequent migration imaging and inversion methods.
Example 2
As a second aspect of the present invention, there is also provided a self-supervised seismic data interpolation reconstruction system based on band extension, the system applying the self-supervised interpolation reconstruction method based on band extension as described in the above embodiment, which comprises:
the low-pass filtering processing module: the method comprises the steps of obtaining low-frequency observation data without space aliasing by low-pass filtering processing on sparse observation data with space aliasing;
low frequency data encryption module: based on Nyquist sampling law, low-frequency data without space aliasing is obtained in an analytic mode;
the data set construction module: the method comprises the steps of splitting an encrypted low-frequency data according to an observation channel and a missing channel, and constructing an adaptive training set and a test set for band extension;
u-net network training module: training the U-net network using the constructed adaptive training set such that the network has a nonlinear mapping capability that maps low frequency data to full frequency band data;
and the missing channel full-band data reconstruction module is used for: inputting the low frequency data in the test set into a trained U-net network to obtain full frequency band data on a missing channel;
interpolation reconstruction module: and the method is used for recombining the reconstructed missing channel full-band data with the original sparse observation data to obtain a final interpolation reconstruction result.
Example 3
As a third aspect of the present invention, the present application further provides an electronic device, including: one or more processors; a memory for storing one or more programs; the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the self-supervised interpolation reconstruction method based on band extension as described above. In addition to the processor, the memory, and the interface, any device with data processing capability in the embodiments generally includes other hardware according to the actual function of the any device with data processing capability, which will not be described herein.
Example 4
As a fourth aspect of the present invention, there is also provided a computer readable storage medium having stored thereon computer instructions which, when executed by a processor, implement a self-supervised interpolation reconstruction method based on band extension as described above. The computer readable storage medium may be an internal storage unit, such as a hard disk or a memory, of any of the data processing enabled devices described in any of the previous embodiments. The computer readable storage medium may also be an external storage device, such as a plug-in hard disk, a Smart Media Card (SMC), an SD Card, a Flash memory Card (Flash Card), or the like, provided on the device. Further, the computer readable storage medium may include both internal storage units and external storage devices of any device having data processing capabilities. The computer readable storage medium is used for storing the computer program and other programs and data required by the arbitrary data processing apparatus, and may also be used for temporarily storing data that has been output or is to be output.
The foregoing describes in detail preferred embodiments of the present invention. It should be understood that numerous modifications and variations can be made in accordance with the concepts of the invention by one of ordinary skill in the art without undue burden. Therefore, all technical solutions which can be obtained by logic analysis, reasoning or limited experiments based on the prior art by the person skilled in the art according to the inventive concept shall be within the scope of protection defined by the claims.

Claims (10)

1. The self-supervision seismic data interpolation reconstruction method based on the band extension is characterized by comprising the following steps:
sparse observation data containing space aliasing is input, and low-frequency observation data without space aliasing is obtained through low-pass filtering processing;
based on Nyquist sampling law, acquiring encrypted low-frequency data from the low-frequency data without spatial aliasing in an analytic mode;
splitting the encrypted low frequency data according to the observation channel and the missing channel, and constructing a self-adaptive training set and a testing set for band prolongation;
training the U-net network using the constructed adaptive training set such that the network has a nonlinear mapping capability that maps low frequency data to full frequency band data;
inputting the low frequency data in the test set into a trained U-net network to obtain full frequency band data on a missing channel;
and recombining the reconstructed missing channel full-band data with the original sparse observation data to obtain a final interpolation reconstruction result.
2. The method for reconstructing self-monitoring seismic data interpolation based on band extension according to claim 1, wherein the step of obtaining the low-frequency observation data without spatial aliasing comprises:
conversion of sparsely acquired observation data D (x, t) to f-k domain D (f, k) using two-dimensional Fourier transform x );
Selecting a spatial aliasing starting frequency f in the f-k domain aliase And setting the initial frequency f of Hamming window cut A frequency low pass filter is constructed as follows:
frequency domain low pass filtering of sparsely acquired data DObtaining the low frequency data without space aliasing
In the formula, +..
3. The method for reconstructing self-monitoring seismic data interpolation based on band extension according to claim 1, wherein the step of obtaining encrypted low frequency data comprises:
based on Nyquist sampling law, the f-k spectrum corresponding to the encrypted low frequency data is obtainedAnalytical characterization can be as follows:
in the method, in the process of the invention,representing low frequency data without spatial aliasing, k x Represents horizontal wavenumber, k Nyquist Representing Nyquist wavenumbers;
for a pair ofPerforming two-dimensional inverse Fourier transform to obtain encrypted low-frequency space-free artificial frequency data +.>
4. The method for self-monitoring seismic data interpolation reconstruction based on band extension according to claim 1, wherein the adaptive training set comprises:
the observation channel low frequency data of the encrypted low frequency data is taken as a sample input;
and the original acquired full-band sparse observation data is used as a label.
5. The method for self-monitoring seismic data interpolation reconstruction based on band extension according to claim 1, wherein the test set comprises unlabeled low frequency data on missing channels.
6. The self-supervision seismic data interpolation reconstruction method based on band extension according to claim 1, wherein the expected output of the U-net network is full-band data, and the following loss function is constructed for network training:
Loss=||d obs -f(d low ,θ)|| 1
wherein d obs Is the observation data of the whole frequency band, d low For the input of low frequency data to the network, f (·) is the designed U-net network, θ is a parameter of the network and, I.I 1 Is L 1 And (5) norm constraint.
7. The self-supervision seismic data interpolation reconstruction method based on band extension according to claim 1, wherein the U-net network quantitatively evaluates a reconstruction result by adopting a signal-to-noise ratio formula:
wherein y is tag data,for the result of the network reconstruction, I.I F Is the F-norm.
8. A self-supervising seismic data interpolation reconstruction system based on band extension, the system comprising:
the low-pass filtering processing module: the method comprises the steps of obtaining low-frequency observation data without space aliasing by low-pass filtering processing on sparse observation data with space aliasing;
low frequency data encryption module: based on Nyquist sampling law, low-frequency data without space aliasing is obtained in an analytic mode;
the data set construction module: the method comprises the steps of splitting an encrypted low-frequency data according to an observation channel and a missing channel, and constructing an adaptive training set and a test set for band extension;
u-net network training module: training the U-net network using the constructed adaptive training set such that the network has a nonlinear mapping capability that maps low frequency data to full frequency band data;
and the missing channel full-band data reconstruction module is used for: inputting the low frequency data in the test set into a trained U-net network to obtain full frequency band data on a missing channel;
interpolation reconstruction module: and the method is used for recombining the reconstructed missing channel full-band data with the original sparse observation data to obtain a final interpolation reconstruction result.
9. An electronic device, comprising:
one or more processors;
a memory for storing one or more programs;
the one or more programs, when executed by the one or more processors, cause the one or more processors to implement the band-extension-based self-supervising seismic data interpolation reconstruction method as set forth in any one of claims 1-7.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the steps of the band-extension-based self-supervising seismic data interpolation reconstruction method as claimed in any one of claims 1 to 7.
CN202310411730.XA 2023-04-17 2023-04-17 Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension Pending CN116520429A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202310411730.XA CN116520429A (en) 2023-04-17 2023-04-17 Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202310411730.XA CN116520429A (en) 2023-04-17 2023-04-17 Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension

Publications (1)

Publication Number Publication Date
CN116520429A true CN116520429A (en) 2023-08-01

Family

ID=87389550

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202310411730.XA Pending CN116520429A (en) 2023-04-17 2023-04-17 Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension

Country Status (1)

Country Link
CN (1) CN116520429A (en)

Similar Documents

Publication Publication Date Title
CN109903259B (en) Magnetic resonance spectrum reconstruction method based on deep learning
US10169293B2 (en) Fractional scaling digital filters and the generation of standardized noise and synthetic data series
CN111046824A (en) Time series signal efficient denoising and high-precision reconstruction modeling method and system
CN112596107B (en) Elastic parameter inversion method and device
CN110174702B (en) Method and system for recovering low-frequency weak signals of marine seismic data
CN111881858B (en) Microseismic signal multi-scale denoising method and device and readable storage medium
CN113723171B (en) Electroencephalogram signal denoising method based on residual error generation countermeasure network
Durall et al. Deep diffusion models for seismic processing
CN113777650B (en) Sparse time-frequency spectrum decomposition method, device and equipment based on mixed norm and wavelet transform and storage medium
CN115795282B (en) Shock tube dynamic pressure reconstruction method and device, electronic equipment and storage medium
CN111783363A (en) Ionized layer prediction method based on SSA and RBF neural network model
CN111273353A (en) Intelligent seismic data de-aliasing method and system based on U-Net network
Du et al. Study on optical fiber gas-holdup meter signal denoising using improved threshold wavelet transform
Cicone et al. Jot: a variational signal decomposition into jump, oscillation and trend
CN112213782A (en) Processing method and device for sub-phase seismic data and server
CN116520429A (en) Self-supervision seismic data interpolation reconstruction method, system and equipment based on band extension
CN117031539A (en) Low-frequency reconstruction method and system for self-supervision deep learning seismic data
CN110598579A (en) Hypercomplex number magnetic resonance spectrum reconstruction method based on deep learning
Brzostowski Sparse reconstruction for enhancement of the empirical mode decomposition-based signal denoising
CN114236610B (en) Iterative seismic data unmixing method and system based on depth convolution network prior
CN106338769A (en) Seismic data denoising method and system
US5873050A (en) Method of analyzing capabilities of multiple-suppression computer seismic data processing software
CN112379415A (en) Multi-scale full-waveform inversion method and device for reconstructed low-frequency data based on downsampling
Aburakhia et al. On the Intersection of Signal Processing and Machine Learning: A Use Case-Driven Analysis Approach
CN116610937B (en) Method and device for carrying out low-frequency information continuation in implicit space and electronic equipment

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