CN117890978A - Seismic velocity image generation method based on visual transducer - Google Patents
Seismic velocity image generation method based on visual transducer Download PDFInfo
- Publication number
- CN117890978A CN117890978A CN202410302937.8A CN202410302937A CN117890978A CN 117890978 A CN117890978 A CN 117890978A CN 202410302937 A CN202410302937 A CN 202410302937A CN 117890978 A CN117890978 A CN 117890978A
- Authority
- CN
- China
- Prior art keywords
- seismic
- seismic velocity
- velocity image
- data
- visual transducer
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000000007 visual effect Effects 0.000 title claims abstract description 55
- 238000000034 method Methods 0.000 title claims abstract description 43
- 238000012549 training Methods 0.000 claims abstract description 17
- 230000006870 function Effects 0.000 claims abstract description 15
- 238000007781 pre-processing Methods 0.000 claims abstract description 6
- 238000012360 testing method Methods 0.000 claims abstract description 5
- 238000012545 processing Methods 0.000 claims description 7
- 230000004927 fusion Effects 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 230000008014 freezing Effects 0.000 claims description 4
- 238000007710 freezing Methods 0.000 claims description 4
- 238000010586 diagram Methods 0.000 claims description 3
- 238000011423 initialization method Methods 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 3
- 230000008447 perception Effects 0.000 claims description 3
- 238000009877 rendering Methods 0.000 claims description 3
- 238000012800 visualization Methods 0.000 claims description 3
- 230000003993 interaction Effects 0.000 abstract description 2
- 238000013473 artificial intelligence Methods 0.000 description 5
- 238000013528 artificial neural network Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 238000005516 engineering process Methods 0.000 description 3
- 229910052799 carbon Inorganic materials 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 238000012067 mathematical method Methods 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- OKTJSMMVPCPJKN-UHFFFAOYSA-N Carbon Chemical compound [C] OKTJSMMVPCPJKN-UHFFFAOYSA-N 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 230000007246 mechanism Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000011218 segmentation Effects 0.000 description 1
- 230000009919 sequestration Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000003245 working effect Effects 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Landscapes
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention provides a seismic velocity image generation method based on a visual transducer. Generating a seismic velocity image using a vision transducer-based seismic velocity image generation system, comprising: acquiring seismic data, and preprocessing the seismic data to obtain a seismic data set; dividing the seismic dataset into a training set and a testing set; initializing parameters of a seismic velocity image generation system; selecting an optimizer, and setting Beta parameter values in the optimizer; training the seismic velocity image generation system by using a combined function of loss functions such as mean square error loss, average absolute error loss and perceived loss as a total loss function; selecting the number of the seismic velocity images to be generated as input data; outputting data of the seismic velocity and visualizing the data; according to the method, the visual transducer is introduced into the full waveform inversion problem, global information in the seismic data graph can be captured, remote interaction of information in the seismic data is realized, and a seismic velocity image with higher quality is generated.
Description
Technical Field
The invention relates to the technical field of artificial intelligence, in particular to a seismic velocity image generation method based on a visual transducer.
Background
Today, with rapid development of technology, artificial intelligence has become one of the most innovative technologies, and plays an increasingly important role in various fields of education, medical treatment, transportation, and the like with unique advantages. Deep neural networks are one of the most successful artificial intelligence models. Theoretically, as long as the data is enough, the model architecture is reasonable, and the deep neural network can approximate any nonlinear function.
Full waveform inversion is a seismic imaging technique aimed at estimating physical parameter distribution and geophysical properties of subsurface geologic structures from full waveform information, playing an important role in various subsurface applications such as subsurface energy exploration, carbon capture and sequestration. Essentially, full waveform inversion can be viewed as a nonlinear inversion operator that can convert the relevant seismic data into corresponding geologic information, such as generating a seismic velocity image. The deep neural network may extract useful features from the seismic data to approximate this inversion operator, thereby generating a seismic velocity image. The method does not depend on a physical model behind the inversion problem, is a promising method for obtaining the relation between the seismic data and the velocity image by using a machine learning method from the data technology, and can skillfully bypass various problems faced by a physical equation driving method. Many deep neural network based models have achieved outstanding results in inversion problems. The current research shows that the inversion model based on the convolutional neural network and the generated countermeasure network has the advantages of high prediction speed, high precision and good generation quality of the speed image.
The success of convolutional neural networks in most visual tasks makes them the first choice for seismic velocity image generation. Most of the models available for seismic velocity image generation tasks are based on convolutional networks. However, the convolutional network works in a similar way to the local attention mechanism with sliding kernels on the image, which breaks the long-range relationship. In addition, the common practice of the deep convolutional neural network is to continuously abstract an original picture to obtain a highly abstract characteristic of an input picture, and the characteristic often lacks some important detail information for the image generation problem, so that the generated geological image is easy to be inaccurate.
Transformer is an emerging model in the field of artificial intelligence, which exhibits great dominance in many application fields, and existing large models of artificial intelligence are basically constructed based on the Transformer. With the advent of visual transducer, transducer has achieved very good results in the visual field. The transducer has the greatest advantages of capturing global information, realizing interaction of long-range information and being capable of parallel computation. Studies have shown that visual convertors are fully adequate and have good results in terms of the generation of problems such as object detection, semantic segmentation, contour recognition.
Disclosure of Invention
According to the technical problem, a method for generating a seismic velocity image based on a visual transducer is provided. The invention mainly uses a transducer to construct the feature extractor, can extract the features of global and abundant detail information, and can generate more accurate earthquake velocity images.
The invention adopts the following technical means:
a method for generating a seismic velocity image based on a visual transducer, which generates a seismic velocity image by using a seismic velocity image generation system based on the visual transducer, comprises the following steps:
s1, acquiring seismic data, and preprocessing the seismic data to obtain a seismic data set;
s2, dividing the preprocessed seismic data set into a training set and a testing set;
s3, initializing parameters of a seismic velocity image generation system based on a visual transducer;
s4, selecting an optimizer, and setting Beta parameter values in the optimizer;
s5, training the TFWI model by using a combined function of the mean square error loss, the average absolute error loss and the perception loss function as a total loss function;
s6, after training is completed, selecting the number of the seismic velocity images to be generated, inputting data into a trained seismic velocity image generation system, and processing the data by the seismic velocity image generation system to obtain four-dimensional seismic velocity images, wherein the first dimension is the number of pictures, the second dimension is the number of channels of the data, the third dimension is the height of the images, and the fourth dimension is the width of the images;
s7, visualizing the seismic velocity image output by the seismic velocity image generation system by utilizing a visualization library and color rendering;
the visual transducer-based seismic velocity image generation system comprises a TFWI model, wherein the TFWI model comprises a visual transducer encoder and a decoder, and the visual transducer encoder and the decoder are used for generating the visual transducer-based seismic velocity image, wherein:
the visual transducer encoder adopts a non-convoluting patch mode to extract global information in the seismic data graph and process the global information;
the decoder adopts a 1*1 convolution kernel to realize the fusion of the features in the channel dimension, and adds a jumper wire connection to realize the fusion of different features, gradually decodes the extracted global information, and finally generates a corresponding seismic velocity image.
Further, the method further comprises:
and S8, adding content loss, and optimizing the trained TFWI model.
Further, in the step S1, the preprocessing of the seismic data specifically includes:
and performing maximum and minimum normalization processing on the obtained seismic data, and converting the value range of the seismic data into [ -1,1 ].
Further, in the step S3, the initializing the parameters of the seismic velocity image generating system based on the visual transducer specifically includes:
the convolutional layer is initialized by using a Kaiming method, the linear layer is initialized by using an Xavier method, the BN layer is initialized by using a default initialization method, a weight attenuation factor is set to 0.0001, an initialization learning rate is 0.002, 465 batches are preheated by the learning rate, a training time period is set to 200 epochs, and Dropout of the encoder part is set to 0.1.
Further, in the step S8, content loss is added to optimize the trained seismic velocity image generating system based on the visual transducer, which specifically includes:
generating corresponding seismic data by the generated seismic velocity data through deep, measuring the distance between the generated seismic data and the original seismic data by using Euclidean distance, further optimizing parameters of a TFWI model, freezing parameters of a visual transducer encoder, optimizing parameters of a decoder, freezing parameters of the decoder, and optimizing parameters of the visual transducer encoder.
Compared with the prior art, the invention has the following advantages:
1. according to the seismic velocity image generation method based on the visual transducer, the visual transducer is introduced into the full waveform inversion problem, so that global information of seismic data can be obtained, and the image generation quality and effect are improved.
2. According to the earthquake velocity image generation method based on the visual transducer, the earthquake velocity image generation system based on the visual transducer is very fast in prediction speed, so that large-scale earthquake velocity image prediction can be realized, and the efficiency is improved.
3. According to the seismic velocity image generation method based on the visual transducer, the corresponding seismic data are only required to be directly sent into the model for training, the corresponding prediction can be directly carried out after the training is finished, the knowledge and mathematical methods in the geophysical aspect are not required to be provided, and for a person who is completely layed in full waveform inversion, the trained model can be used for generating the seismic velocity image.
4. The seismic velocity image generation method based on the visual transducer can be used for generating seismic velocity images on other data sets, and a trained model is used, so that fine adjustment can be performed only by using corresponding data, a good working effect can be generated by the model, and generalization capability is improved.
5. The seismic velocity image generation method based on the visual transducer provided by the invention has the advantages that the seismic velocity image generation system can be used for generating the seismic velocity image in full-waveform inversion, can be used for generating other images in full-waveform inversion, and has expandability, universality and the like.
In conclusion, the invention brings improvement of performance, precision, efficiency, simplicity, generalization capability and expandability. The invention has wide application prospect and commercial value in the generation of the earthquake velocity image due to the technical effects and the improvement of main performance indexes.
For the reasons, the method can be widely popularized in the fields of image processing and the like.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required in the embodiments or the description of the prior art will be briefly described, and it is obvious that the drawings in the following description are some embodiments of the present invention, and other drawings may be obtained according to the drawings without inventive effort to a person skilled in the art.
FIG. 1 is a schematic diagram of a system for generating seismic velocity images based on visual transducers according to the present invention.
FIG. 2 illustrates a non-convolved patch approach provided by an embodiment of the present invention.
FIG. 3 is a comparison of a seismic velocity image generated by a TFWI model provided by an embodiment of the invention with a true image.
Detailed Description
In order that those skilled in the art will better understand the present invention, a technical solution in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in which it is apparent that the described embodiments are only some embodiments of the present invention, not all embodiments. All other embodiments, which can be made by those skilled in the art based on the embodiments of the present invention without making any inventive effort, shall fall within the scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and the claims of the present invention and the above figures are used for distinguishing between similar objects and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used may be interchanged where appropriate such that the embodiments of the invention described herein may be implemented in sequences other than those illustrated or otherwise described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention provides a method for generating a seismic velocity image based on a visual transducer, which utilizes a seismic velocity image generating system based on the visual transducer to generate the seismic velocity image, namely: in extracting features from a seismic dataset Curve to generate a seismic velocity image, the method of the invention can be used to generate a higher quality velocity image, including:
s1, acquiring seismic data, and preprocessing the seismic data to obtain a seismic data set; in this embodiment, preprocessing the seismic data specifically includes: and performing maximum and minimum normalization processing on the obtained seismic data, and converting the value range of the seismic data into [ -1,1 ].
S2, dividing the preprocessed seismic data set into a training set and a testing set; in this embodiment, the training set is 24000 data, and the test set uses the remaining data.
S3, initializing parameters of a seismic velocity image generation system based on a visual transducer; in this embodiment, the method specifically includes: the convolutional layer is initialized by using a Kaiming method, the linear layer is initialized by using an Xavier method, the BN layer is initialized by using a default initialization method, a weight attenuation factor is set to 0.0001, an initialization learning rate is 0.002, 465 batches are preheated by the learning rate, a training time period is set to 200 epochs, and Dropout of the encoder part is set to 0.1.
S4, selecting an optimizer, and setting Beta parameter values in the optimizer; in this embodiment AdamW is used as the optimizer, or Adam optimizer is used, with Beta set to 0.9 and 0.999.
S5, training a seismic velocity image generation system based on a vision transducer by using a combined function of loss functions such as mean square error loss, average absolute error loss and perception loss as a total loss function;
s6, after training is completed, selecting the number of the seismic velocity images to be generated, inputting data into a trained seismic velocity image generation system, and processing the data by the seismic velocity image generation system to obtain four-dimensional seismic velocity images, wherein the first dimension is the number of pictures, the second dimension is the number of channels of the data, the third dimension is the height of the images, and the fourth dimension is the width of the images; for example: if B pictures are desired to be generated, the shape of the input data to the vision transducer based seismic velocity image generation system should be (B, C, H, W).
S7, visualizing the seismic velocity image output by the seismic velocity image generation system by utilizing a visualization library and color rendering;
and S8, adding content loss, optimizing the trained seismic velocity image generation system based on the visual transducer, and further improving the model performance. In this embodiment, specifically, a full-waveform forward tool, such as deep wave, is used to generate corresponding seismic data from the generated seismic velocity data through the deep wave, the euclidean distance is used to measure the distance between the generated seismic data and the original seismic data, parameters of a seismic velocity image generating system based on visual transducer are further optimized, encoder parameters in the seismic velocity image generating system based on visual transducer are frozen, decoder parameters are optimized, and then the decoder parameters are frozen, so that the encoder is optimized.
The seismic velocity image generation system based on the visual transducer, as shown in fig. 1, comprises a TFWI model, wherein the TFWI model comprises a visual transducer encoder and a decoder, and the TFWI model comprises the following components:
the visual transducer encoder adopts a non-convoluting patch mode to extract global information in the seismic data diagram, and processes the global information; such as the left hand portion of fig. 1. Convolution is very common in the field of vision to obtain a patch, but in full waveform inversion of a seismic, convolution kernel parameters of the patch interfere with the ability of an encoder to extract and process seismic data, so that the method adopts a non-convolved patch mode in the encoder. As shown in fig. 2.
The decoder adopts a 1*1 convolution kernel to realize the fusion of the features in the channel dimension, and adds a jumper wire connection to realize the fusion of different features, gradually decodes the extracted global information, and finally generates a corresponding seismic velocity image. As in the right part of fig. 1, the decoding capability of the decoder is improved.
InversionNet, velocityGAN in the following table is a current more advanced model, a model based on convolutional neural network, TFWI is the seismic velocity model proposed by the present invention.
Table I comparison of results indexes of different models
Therefore, as shown in fig. 3 (a) and (b), by using the velocity generation model of the present invention, a high quality seismic velocity image is obtained, and a large-scale prediction can be performed with a high prediction speed as compared with a method based on a physical equation. Compared with the existing advanced model, the model has stronger resistance to noise, and the velocity generation model can still generate high-quality seismic velocity images even under Gaussian noise with larger variance.
In conclusion, the method has the advantages of good picture quality, high prediction speed and simple model use, and does not need a user to have related knowledge in the aspect of geophysics or grasp a corresponding mathematical method. The invention can conduct large-scale prediction, can directly migrate to other data sets by using the seismic velocity generation model, and can obtain better effects by fine tuning.
Finally, it should be noted that: the above embodiments are only for illustrating the technical solution of the present invention, and not for limiting the same; although the invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical scheme described in the foregoing embodiments can be modified or some or all of the technical features thereof can be replaced by equivalents; such modifications and substitutions do not depart from the spirit of the invention.
Claims (5)
1. A method for generating a seismic velocity image based on a visual transducer, the method comprising generating a seismic velocity image using a seismic velocity image generation system based on a visual transducer, comprising:
s1, acquiring seismic data, and preprocessing the seismic data to obtain a seismic data set;
s2, dividing the preprocessed seismic data set into a training set and a testing set;
s3, initializing parameters of a seismic velocity image generation system based on a visual transducer;
s4, selecting an optimizer, and setting Beta parameter values in the optimizer;
s5, training a seismic velocity image generation system based on a vision transducer by using a combined function of a mean square error loss function, a mean absolute error loss function and a perception loss function as a total loss function;
s6, after training is completed, selecting the number of the seismic velocity images to be generated, inputting data into a trained seismic velocity image generation system, and processing the data by the seismic velocity image generation system to obtain four-dimensional seismic velocity images, wherein the first dimension is the number of pictures, the second dimension is the number of channels of the data, the third dimension is the height of the images, and the fourth dimension is the width of the images;
s7, visualizing the seismic velocity image output by the seismic velocity image generation system by utilizing a visualization library and color rendering;
the visual transducer-based seismic velocity image generation system comprises a TFWI model, wherein the TFWI model comprises a visual transducer encoder and a decoder, and the visual transducer encoder and the decoder are used for generating the visual transducer-based seismic velocity image, wherein:
the visual transducer encoder adopts a non-convoluting patch mode to extract global information in the seismic data diagram, and processes the global information;
the decoder adopts a 1*1 convolution kernel to realize the fusion of the features in the channel dimension, and adds a jumper wire connection to realize the fusion of different features, gradually decodes the extracted global information, and finally generates a corresponding seismic velocity image.
2. The method of generating a visual transducer-based seismic velocity image of claim 1, further comprising:
and S8, adding content loss, and optimizing the trained TFWI model.
3. The method for generating a seismic velocity image based on a visual transducer according to claim 1, wherein in step S1, the seismic data is preprocessed, specifically comprising:
and performing maximum and minimum normalization processing on the obtained seismic data, and converting the value range of the seismic data into [ -1,1 ].
4. The method for generating a seismic velocity image based on a visual transducer according to claim 1, wherein in step S3, parameters of the seismic velocity image generating system based on the visual transducer are initialized, specifically comprising:
the convolutional layer is initialized by using a Kaiming method, the linear layer is initialized by using an Xavier method, the BN layer is initialized by using a default initialization method, a weight attenuation factor is set to 0.0001, an initialization learning rate is 0.002, 465 batches are preheated by the learning rate, a training time period is set to 200 epochs, and Dropout of the encoder part is set to 0.1.
5. The method for generating a seismic velocity image based on a visual transducer according to claim 2, wherein in step S8, content loss is added to optimize the trained seismic velocity image generating system based on a visual transducer, specifically comprising:
generating corresponding seismic data by the generated seismic velocity data through deep, measuring the distance between the generated seismic data and the original seismic data by using Euclidean distance, further optimizing parameters of a TFWI model, freezing parameters of a visual transducer encoder, optimizing parameters of a decoder, freezing parameters of the decoder, and optimizing parameters of the visual transducer encoder.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410302937.8A CN117890978B (en) | 2024-03-18 | 2024-03-18 | Seismic velocity image generation method based on visual transducer |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410302937.8A CN117890978B (en) | 2024-03-18 | 2024-03-18 | Seismic velocity image generation method based on visual transducer |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117890978A true CN117890978A (en) | 2024-04-16 |
| CN117890978B CN117890978B (en) | 2024-05-10 |
Family
ID=90643164
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410302937.8A Active CN117890978B (en) | 2024-03-18 | 2024-03-18 | Seismic velocity image generation method based on visual transducer |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117890978B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5999489A (en) * | 1997-03-21 | 1999-12-07 | Tomoseis Inc. | High vertical resolution crosswell seismic imaging |
| CN105089657A (en) * | 2015-06-15 | 2015-11-25 | 中国石油天然气股份有限公司 | Physical simulation method and physical simulation experimental device for oil and gas charging of fracture-cave carbonate reservoir |
| CN116520394A (en) * | 2022-01-21 | 2023-08-01 | 中国石油化工股份有限公司 | Method for directly predicting porosity based on seismic logging double-drive fusion multi-scale information |
| CN117521763A (en) * | 2024-01-04 | 2024-02-06 | 大连海事大学 | Artificial intelligent model compression method integrating regularized pruning and importance pruning |
-
2024
- 2024-03-18 CN CN202410302937.8A patent/CN117890978B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5999489A (en) * | 1997-03-21 | 1999-12-07 | Tomoseis Inc. | High vertical resolution crosswell seismic imaging |
| CN105089657A (en) * | 2015-06-15 | 2015-11-25 | 中国石油天然气股份有限公司 | Physical simulation method and physical simulation experimental device for oil and gas charging of fracture-cave carbonate reservoir |
| CN116520394A (en) * | 2022-01-21 | 2023-08-01 | 中国石油化工股份有限公司 | Method for directly predicting porosity based on seismic logging double-drive fusion multi-scale information |
| CN117521763A (en) * | 2024-01-04 | 2024-02-06 | 大连海事大学 | Artificial intelligent model compression method integrating regularized pruning and importance pruning |
Non-Patent Citations (1)
| Title |
|---|
| 孙建孟，等: "数字岩心和数字井筒技术研究与应用进展", 石油物探, vol. 62, no. 5, 30 September 2023 (2023-09-30), pages 806 - 819 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117890978B (en) | 2024-05-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Fang et al. | Seismic data interpolation based on U-net with texture loss | |
| CN111050219B (en) | Method and system for processing video content using a spatio-temporal memory network | |
| CN110969124A (en) | Two-dimensional human body posture estimation method and system based on lightweight multi-branch network | |
| Oliveira et al. | Improving seismic data resolution with deep generative networks | |
| He et al. | Seismic data consecutively missing trace interpolation based on multistage neural network training process | |
| Huang et al. | Self-supervised deep learning to reconstruct seismic data with consecutively missing traces | |
| Zeng et al. | InversionNet3D: Efficient and scalable learning for 3-D full-waveform inversion | |
| CN111797891A (en) | Unpaired heterogeneous face image generation method and device based on generation countermeasure network | |
| Liu et al. | A super resolution algorithm based on attention mechanism and srgan network | |
| Porter et al. | A deep learning approach to selecting representative time steps for time-varying multivariate data | |
| GB2579262A (en) | Space-time memory network for locating target object in video content | |
| CN115049921A (en) | Method for detecting salient target of optical remote sensing image based on Transformer boundary sensing | |
| Wang et al. | Underwater image super-resolution using multi-stage information distillation networks | |
| Zhang et al. | Nonconvex log-sum function-based majorization–minimization framework for seismic data reconstruction | |
| Wang et al. | Adapting the residual dense network for seismic data denoising and upscaling | |
| Lan et al. | Seismic data recovery using deep targeted denoising priors in an alternating optimization framework | |
| Song et al. | Weighted envelope correlation-based waveform inversion using automatic differentiation | |
| Goyes-Peñafiel et al. | Coordinate-based seismic interpolation in irregular land survey: a deep internal learning approach | |
| Krishnan et al. | A novel underwater image enhancement technique using ResNet | |
| CN117890978B (en) | Seismic velocity image generation method based on visual transducer | |
| Li et al. | Unsupervised neural rendering for image hazing | |
| Guan et al. | DiffWater: Underwater image enhancement based on conditional denoising diffusion probabilistic model | |
| CN113284150B (en) | Industrial quality inspection method and industrial quality inspection device based on unpaired industrial data | |
| Zhou et al. | Deep learning with fault prior for 3-D seismic data super-resolution | |
| Wei et al. | Hybrid Loss-Guided Coarse-to-Fine Model for Seismic Data Consecutively Missing Trace Reconstruction |
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 |