CN115184999B - Marchenko imaging focusing function correction method based on deep learning - Google Patents
Marchenko imaging focusing function correction method based on deep learning Download PDFInfo
- Publication number
- CN115184999B CN115184999B CN202210796555.6A CN202210796555A CN115184999B CN 115184999 B CN115184999 B CN 115184999B CN 202210796555 A CN202210796555 A CN 202210796555A CN 115184999 B CN115184999 B CN 115184999B
- Authority
- CN
- China
- Prior art keywords
- focusing function
- model
- marchenko
- taking
- imaging
- 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
- 230000006870 function Effects 0.000 title claims abstract description 88
- 238000003384 imaging method Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000012937 correction Methods 0.000 title claims abstract description 17
- 238000013135 deep learning Methods 0.000 title claims abstract description 17
- 238000012549 training Methods 0.000 claims abstract description 19
- 230000005012 migration Effects 0.000 claims abstract description 4
- 238000013508 migration Methods 0.000 claims abstract description 4
- 238000004613 tight binding model Methods 0.000 claims description 8
- 238000013526 transfer learning Methods 0.000 claims description 7
- 230000004044 response Effects 0.000 claims description 6
- 238000005070 sampling Methods 0.000 claims description 5
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000009499 grossing Methods 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000008901 benefit Effects 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification 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
- G01V1/36—Effecting static or dynamic corrections on records, e.g. correcting spread; Correlating seismic signals; Eliminating effects of unwanted energy
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)
- Radar Systems Or Details Thereof (AREA)
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention is suitable for the technical field of seismic exploration, and provides a Marchenko imaging focusing function correction method based on deep learning, which comprises the following steps: taking the seismic data obtained by the observation system and the direct wave obtained by the background model as input, and taking a focusing function obtained by Marchenko equation as original training data to be corrected; taking the accurate focusing function obtained by the reference model as tag data, and training the model according to the designed U-net architecture; the trained model is used to correct the error-containing focusing function, resulting in a more accurate Marchenko image. The method can effectively correct the focusing function in an unknown area, thereby accelerating Marchenko imaging process, and simultaneously, through migration learning, the model can effectively correct the focusing function of the direct wave under the condition that the travel time error is contained in the direct wave and the seismic data has missing information, without spending a great deal of time to reconstruct lost input data, thereby saving time and cost.
Description
Technical Field
The invention belongs to the technical field of seismic exploration, and particularly relates to a Marchenko imaging focusing function correction method based on deep learning.
Background
Complex geological structures of the overburden are often encountered in seismic exploration. The seismic source is excited from the earth's surface and the seismic wave is distorted by the complex overburden, thereby severely affecting the seismic record received by the receiver. Conventional virtual source methods may bypass the overburden to reconstruct the subsurface source to the subsurface receiver location. This method requires the presence of a physical receiver underground. However, in practice it may not be possible to place the receiver at a given location deep in the subsurface. Marchenko the imaging method is based on the concept of "autofocus", and only the surface reflection response and the subsurface focus-to-surface direct wave are needed to calculate the focus-to-surface green function, i.e., the method does not require placement of an actual receiver in the subsurface relative to the virtual source method, and can specify any location in the subsurface (i.e., at the focus) as a virtual receiving point. The resulting green's function is used as an input to the imaging, and the resulting subsurface imaging suppresses artifacts associated with the interbed multiples. The green function used by Marchenko imaging methods can more accurately interpret multiples than other conventional imaging methods.
The background model used by the conventional Marchenko method is estimated, and an inaccurate background model influences the estimation of the direct wave, so that an error is generated in the focusing function obtained through iteration, and finally imaging is focused at an error position. In addition, both the travel time error and the incompletely sampled surface reflection response can lead to errors in the focusing function and subsequent imaging.
Disclosure of Invention
The embodiment of the invention aims to provide a Marchenko imaging focusing function correction method based on deep learning, and aims to solve the problem that a background model used by a conventional Marchenko method is estimated, and an inaccurate background model influences the estimation of direct waves, so that an error is generated in an iteratively obtained focusing function, and imaging is finally focused at an error position. In addition, both the travel time error and the incompletely sampled surface reflection response can lead to problems with errors in the focusing function and subsequent imaging.
The embodiment of the invention is realized in such a way that the Marchenko imaging focusing function correction method based on deep learning comprises the following steps:
Step 1: taking the seismic data obtained by the observation system and the direct wave obtained by the background model as input, and taking a focusing function obtained by Marchenko equation as original training data to be corrected;
step 2: taking the accurate focusing function obtained by the reference model as tag data, and training the model according to the designed U-net architecture;
Step 3: the trained model is used to correct the error-containing focusing function, resulting in a more accurate Marchenko image.
According to a further technical scheme, the specific steps of the step 1 comprise:
Step 1.1: establishing an observation system through a reference model to obtain earth surface reflection response;
Step 1.2: the initial downlink focusing function is approximated by a time reversal of the direct arrival of the green function:
I.e., the focus-to-surface direct wave in the specified region estimated using the smoothing model.
The initial uplink focusing function is:
Step 1.3: solving a focusing function containing errors according to an iteration scheme:
Wherein the method comprises the steps of
K is the number of iterations and,Is the/>, of the kth iterationIs represented by a symbol representing a time domain convolution.
Step 1.4: and (3) calculating an accurate direct wave through a reference model, and substituting the accurate direct wave into the step (1.3) to obtain an accurate focusing function.
According to a further technical scheme, the specific steps of the step 2 comprise:
And designing a U-Net network architecture, taking the focusing function containing errors as training data, and taking the accurate focusing function as tag data to carry out model training. Using Adam optimizer, the initial learning rate was set to 0.0001 and the batch size was 16. Since the problem to be solved by this study is to predict a specific value, i.e. the regression problem, the loss function uses root Mean Square Error (MSE), which is defined as follows:
Where y i is the label value of the i-th element in the error-containing focusing function y, y i' is the predicted value of the i-th element of the neural network, and n is the total number of elements in the focusing function. Fig. 4 shows the effect of the contrast before and after correction of the focusing function.
According to a further technical scheme, the step 3 comprises the following specific steps:
step 3.1: taking the trained model as a pre-training model, and performing migration learning on the focusing function of the corresponding region;
step 3.2: the green function is obtained by Marchenko equation:
Step 3.3: marchenko imaging is carried out, and the imaging formula is as follows:
step 3.4: and respectively placing the focusing function containing the travel time error and the focusing function obtained by incomplete sampling into a model for transfer learning to obtain a corrected focusing function.
According to the Marchenko imaging focusing function correction method based on the deep learning, the deep learning is introduced to process the focusing function estimation error caused by the calculation error of the direct wave, and the transfer learning is utilized to enable the focusing function correction to be very effective in more unknown areas, so that the Marchenko imaging process is accelerated. Furthermore, by extending the application of this method to other scenarios of Marchenko method: time offset due to travel time error of the direct wave, and artifacts and gaps due to incomplete sampling. Through transfer learning, the model can effectively correct the focusing function of the direct wave under the condition that the travel time error is contained in the direct wave and the seismic data has missing information, and does not need to spend a great deal of time to reconstruct the missing input data, so that the benefits and wide application of deep learning and transfer learning in the Marchenko method are proved, and the time and the cost are saved.
Drawings
FIG. 1 is a flowchart of an algorithm implementation of a Marchenko imaging focusing function correction method based on deep learning according to an embodiment of the present invention;
FIG. 2 is a reference model diagram in a method for correcting a Marchenko imaging focusing function based on deep learning according to an example embodiment of the present invention;
FIG. 3 is a network architecture diagram in a Marchenko imaging focusing function correction method based on deep learning according to an embodiment of the present invention;
fig. 4 is a schematic structural diagram of focusing function contrast in a Marchenko imaging focusing function correction method based on deep learning according to an embodiment of the present invention;
FIG. 5 is a graph comparing imaging results in a Marchenko imaging focusing function correction method based on deep learning according to an embodiment of the present invention;
FIG. 6 is a graph showing a comparison of downlink focusing functions containing travel time error in a method for correcting Marchenko imaging focusing functions based on deep learning according to an embodiment of the present invention;
Fig. 7 is a comparison chart of a downlink focusing function obtained by incomplete sampling in a Marchenko imaging focusing function correction method based on deep learning according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the drawings and examples, in order to make the objects, technical solutions and advantages of the present invention more apparent. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the scope of the invention.
Specific implementations of the invention are described in detail below in connection with specific embodiments.
As shown in fig. 1, a depth learning-based Marchenko imaging focusing function correction method according to one embodiment of the present invention includes the following steps:
Step 1: taking the seismic data obtained by the observation system and the direct wave obtained by the background model as input, and taking a focusing function obtained by Marchenko equation as original training data to be corrected;
step 2: taking the accurate focusing function obtained by the reference model as tag data, and training the model according to the designed U-net architecture;
Step 3: the trained model is used to correct the error-containing focusing function, resulting in a more accurate Marchenko image.
As a preferred embodiment of the present invention, the specific steps of the step 1 include:
Step 1.1: establishing an observation system through a reference model to obtain earth surface reflection response;
Step 1.2: the initial downlink focusing function is approximated by a time reversal of the direct arrival of the green function:
I.e., the direct wave from the focus to the surface in region a in fig. 2 estimated using a smoothing model.
The initial uplink focusing function is:
Step 1.3: solving a focusing function containing errors according to an iteration scheme:
Wherein the method comprises the steps of
K is the number of iterations and,Is the/>, of the kth iterationIs represented by a symbol representing a time domain convolution.
Step 1.4: and (3) calculating an accurate direct wave through a reference model, and substituting the accurate direct wave into the step (1.3) to obtain an accurate focusing function.
As a preferred embodiment of the present invention, the specific steps of the step 2 include:
the U-Net network architecture (figure 3) is designed, the focusing function containing errors is used as training data, and the accurate focusing function is used as label data for model training. Using Adam optimizer, the initial learning rate was set to 0.0001 and the batch size was 16. Since the problem to be solved by this study is to predict a specific value, i.e. the regression problem, the loss function uses root Mean Square Error (MSE), which is defined as follows:
Where y i is the label value of the i-th element in the error-containing focusing function y, y i' is the predicted value of the i -th element of the neural network, and n is the total number of elements in the focusing function. Fig. 4 shows the effect of the contrast before and after correction of the focusing function.
As a preferred embodiment of the present invention, the step 3 includes the following specific steps:
Step 3.1: taking the trained model as a pre-training model, and performing migration learning on the focusing functions of the areas B-F in the graph 2;
step 3.2: the green function is obtained by Marchenko equation:
Step 3.3: marchenko imaging is carried out, and the imaging formula is as follows:
fig. 5 shows the comparative effect of the final imaging results.
Step 3.4: and respectively placing the focusing function containing the travel time error and the focusing function obtained by incomplete sampling into a model for transfer learning to obtain a corrected focusing function. Fig. 6 and 7 show training results of the transfer learning. In fig. 6, a is an uncorrected downlink focusing function, b is a corrected downlink focusing function, c is a reference downlink focusing function, and d is a single-pass enlarged contrast chart. In fig. 7, a is an uncorrected downlink focusing function, b is a corrected downlink focusing function, c is a reference downlink focusing function, and d is a single-pass enlarged contrast chart.
The foregoing description of the preferred embodiments of the invention is not intended to be limiting, but rather is intended to cover all modifications, equivalents, and alternatives falling within the spirit and principles of the invention.
Claims (1)
1. The Marchenko imaging focusing function correction method based on the deep learning is characterized by comprising the following steps of:
Step 1: taking the seismic data obtained by the observation system and the direct wave obtained by the background model as input, and taking a focusing function obtained by Marchenko equation as original training data to be corrected;
step 2: taking the accurate focusing function obtained by the reference model as tag data, and training the model according to the designed U-net architecture;
Step 3: the trained model is used for correcting the focusing function containing errors, and finally, more accurate Marchenko imaging is obtained;
the specific steps of the step 1 comprise:
Step 1.1: establishing an observation system through a reference model to obtain earth surface reflection response;
Step 1.2: the initial downlink focusing function is approximated by a time reversal of the direct arrival of the green function:
；
estimating a direct wave from a focus to the earth surface in a designated area by using a smoothing model;
the initial uplink focusing function is:
；
Step 1.3: solving a focusing function containing errors according to an iteration scheme:
；
Wherein the method comprises the steps of ；
For iteration number,/>Is/>Iteration/>Wake, sign/>Representing a time domain convolution;
step 1.4: calculating an accurate direct wave through a reference model, and substituting the accurate direct wave into the step 1.3 to obtain an accurate focusing function;
The specific steps of the step 2 include: designing a U-Net network architecture, taking a focusing function containing errors as training data, taking an accurate focusing function as tag data to perform model training, wherein the model training uses an Adam optimizer, the initial learning rate is set to be 0.0001, and the batch size is set to be 16; the loss function uses root mean square error, which is defined as follows:
；
wherein, Is a focusing function/>, containing errorsMiddle/>Tag value of element,/>Is the first/>, of the neural networkPredicted value of element,/>Is the total number of elements in the focusing function;
the step 3 comprises the following specific steps:
step 3.1: taking the trained model as a pre-training model, and performing migration learning on the focusing function of the corresponding region;
step 3.2: the green function is obtained by Marchenko equation:
；
Step 3.3: marchenko imaging is carried out, and the imaging formula is as follows:
；
step 3.4: and respectively placing the focusing function containing the travel time error and the focusing function obtained by incomplete sampling into a model for transfer learning to obtain a corrected focusing function.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210796555.6A CN115184999B (en) | 2022-07-06 | 2022-07-06 | Marchenko imaging focusing function correction method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210796555.6A CN115184999B (en) | 2022-07-06 | 2022-07-06 | Marchenko imaging focusing function correction method based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN115184999A CN115184999A (en) | 2022-10-14 |
| CN115184999B true CN115184999B (en) | 2024-04-19 |
Family
ID=83517394
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210796555.6A Active CN115184999B (en) | 2022-07-06 | 2022-07-06 | Marchenko imaging focusing function correction method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115184999B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118033746B (en) * | 2024-04-07 | 2024-07-09 | 吉林大学 | Object-oriented marine streamer seismic data Marchenko imaging method |
| CN118131322B (en) * | 2024-05-06 | 2024-08-13 | 吉林大学 | Micro-seismic source positioning method based on Marchenko imaging |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2015560A2 (en) * | 2007-07-13 | 2009-01-14 | Morpho Inc. | Image data processing method and imaging apparatus |
| CN107329233A (en) * | 2017-08-25 | 2017-11-07 | 西南大学 | A kind of droplet type PCR instrument Atomatic focusing method based on neutral net |
| CN113219535A (en) * | 2021-05-21 | 2021-08-06 | 东北石油大学 | One-step method interlayer multiple pressing method based on Marchenko theory |
| CN114428342A (en) * | 2020-10-10 | 2022-05-03 | 中国石油化工股份有限公司 | Markenko Green function reconstruction method and system based on MPI |
-
2022
- 2022-07-06 CN CN202210796555.6A patent/CN115184999B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP2015560A2 (en) * | 2007-07-13 | 2009-01-14 | Morpho Inc. | Image data processing method and imaging apparatus |
| CN107329233A (en) * | 2017-08-25 | 2017-11-07 | 西南大学 | A kind of droplet type PCR instrument Atomatic focusing method based on neutral net |
| CN114428342A (en) * | 2020-10-10 | 2022-05-03 | 中国石油化工股份有限公司 | Markenko Green function reconstruction method and system based on MPI |
| CN113219535A (en) * | 2021-05-21 | 2021-08-06 | 东北石油大学 | One-step method interlayer multiple pressing method based on Marchenko theory |
Non-Patent Citations (1)
| Title |
|---|
| 机载合成孔径雷达ETF成像算法研究;王国华;孙进平;;遥测遥控;20060620(第03期);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN115184999A (en) | 2022-10-14 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115184999B (en) | Marchenko imaging focusing function correction method based on deep learning | |
| CN107392097A (en) | A kind of 3 D human body intra-articular irrigation method of monocular color video | |
| CN109886884B (en) | Visual data complement method based on low-rank tensor estimation of limiting kernel norms | |
| US7440628B2 (en) | Method and system for motion correction in a sequence of images | |
| US7542604B2 (en) | System and method for image segmentation by solving an inhomogenous dirichlet problem | |
| CN105825200A (en) | High-spectrum abnormal object detection method based on background dictionary learning and structure sparse expression | |
| WO2021196528A1 (en) | Global ionospheric total electron content prediction method based on deep recurrent neural network | |
| US7565010B2 (en) | System and method for image segmentation by a weighted multigrid solver | |
| US6597816B1 (en) | Correcting distortion in an imaging system using parametric motion estimation | |
| US20210241428A1 (en) | Image Reconstruction using Artificial Intelligence (Ai) Modules Agnostic to Image Acquisition Settings | |
| CN102968765A (en) | Method for correcting infrared focal plane heterogeneity based on sigma filter | |
| US20230078158A1 (en) | High-resolution Seismic Fault Detection with Adversarial Neural Networks and Regularization | |
| CN116524062A (en) | Diffusion model-based 2D human body posture estimation method | |
| EP3229207A2 (en) | Warp models for registering multi-spectral imagery | |
| CN109145738B (en) | Dynamic video segmentation method based on weighted non-convex regularization and iterative re-constrained low-rank representation | |
| Cappé et al. | Simulation-based methods for blind maximum-likelihood filter identification | |
| CN114187598B (en) | Handwriting digital recognition method, handwriting digital recognition equipment and computer readable storage medium | |
| CN104065853A (en) | Infrared camera crosstalk eliminating method | |
| US20210287049A1 (en) | Method and system for processing an image and performing instance segmentation using affinity graphs | |
| US11689693B2 (en) | Video frame interpolation method and device, computer readable storage medium | |
| Labat et al. | Sparse blind deconvolution accounting for time-shift ambiguity | |
| KR102044626B1 (en) | Joint Filtering Device and Method Using Learning | |
| CN110647986A (en) | Road damage image generation method based on countermeasure type generation network | |
| CN112464989A (en) | Closed loop detection method based on target detection network | |
| US7962318B2 (en) | Methods for optimizing system models and the like using least absolute value estimations |
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 |