CN111694053A - First arrival picking method and device - Google Patents
First arrival picking method and device Download PDFInfo
- Publication number
- CN111694053A CN111694053A CN201910193634.6A CN201910193634A CN111694053A CN 111694053 A CN111694053 A CN 111694053A CN 201910193634 A CN201910193634 A CN 201910193634A CN 111694053 A CN111694053 A CN 111694053A
- Authority
- CN
- China
- Prior art keywords
- seismic data
- arrival
- arrival picking
- training
- test sample
- 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
Links
- 238000000034 method Methods 0.000 title claims abstract description 54
- 238000012360 testing method Methods 0.000 claims abstract description 105
- 238000013528 artificial neural network Methods 0.000 claims abstract description 59
- 238000005070 sampling Methods 0.000 claims abstract description 23
- 230000008602 contraction Effects 0.000 claims abstract description 18
- 238000012549 training Methods 0.000 claims description 108
- 230000008859 change Effects 0.000 claims description 29
- 238000013524 data verification Methods 0.000 claims description 17
- 238000000605 extraction Methods 0.000 claims description 16
- 238000010606 normalization Methods 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 13
- 230000006870 function Effects 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 5
- 230000003595 spectral effect Effects 0.000 claims description 5
- 238000012544 monitoring process Methods 0.000 claims description 4
- 230000011218 segmentation Effects 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000009466 transformation Effects 0.000 claims description 3
- 239000000523 sample Substances 0.000 description 140
- 238000012937 correction Methods 0.000 description 20
- 238000001228 spectrum Methods 0.000 description 11
- 238000010586 diagram Methods 0.000 description 10
- 238000012545 processing Methods 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000010521 absorption reaction Methods 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000003384 imaging method Methods 0.000 description 2
- 238000011176 pooling Methods 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 239000007795 chemical reaction product Substances 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013502 data validation Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000002708 enhancing effect Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 230000001052 transient effect Effects 0.000 description 1
Images
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
-
- 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
- G01V1/362—Effecting static or dynamic corrections; Stacking
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01V—GEOPHYSICS; GRAVITATIONAL MEASUREMENTS; DETECTING MASSES OR OBJECTS; TAGS
- G01V2210/00—Details of seismic processing or analysis
- G01V2210/50—Corrections or adjustments related to wave propagation
- G01V2210/52—Move-out correction
Landscapes
- Engineering & Computer Science (AREA)
- Remote Sensing (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Acoustics & Sound (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- General Physics & Mathematics (AREA)
- Geophysics (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention discloses a first arrival picking method and a device, wherein the method comprises the following steps: extracting the characteristic attributes of the seismic data test sample, inputting the characteristic attributes of the seismic data test sample into a trained first arrival picking model, and obtaining a first arrival of the seismic data test sample, wherein the first arrival picking model is a U-net full convolution neural network-based first arrival picking model. The contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer. According to the method, the first arrival of the seismic data test sample is obtained by using the first arrival picking model based on the U-net full convolution neural network, and the first arrival picking efficiency and the first arrival picking accuracy can be improved.
Description
Technical Field
The invention relates to the technical field of seismic exploration, in particular to a first arrival picking method and a first arrival picking device.
Background
This section is intended to provide a background or context to the embodiments of the invention that are recited in the claims. The description herein is not admitted to be prior art by inclusion in this section.
In seismic exploration, a first-arrival wave is a seismic wave that propagates from a shot point through the earth's surface to a receiving point. The first-arrival time carries near-surface structure information, the imaging of the underground medium structure and the inversion precision of stratum elastic parameters are seriously influenced by the change of the near-surface structure, and the accurate first-arrival time is a key parameter of the near-surface inversion and plays an important role in improving the static correction precision, the imaging of the underground medium and the inversion of the stratum elastic parameters.
Early first-arrival pickups were manual, non-real-time analyses, and with the development of computer technology and seismic processing technology, the first-arrival pickups transitioned from early manual analyses to semi-automatic analyses of human-computer interaction. The conventional first arrival picking method needs to be modified manually, so that the first arrival picking efficiency is low, and the accuracy of the initial picking is not satisfactory.
Therefore, the existing first arrival picking method has the problems of low first arrival picking efficiency and poor accuracy.
Disclosure of Invention
The embodiment of the invention provides a first arrival picking method, which is used for improving the first arrival picking efficiency and the first arrival picking accuracy and comprises the following steps:
extracting the characteristic attribute of the seismic data test sample;
inputting the characteristic attributes of the seismic data test sample into a trained first arrival picking model to obtain a first arrival of the seismic data test sample; the first arrival picking model is based on a U-net full convolution neural network;
wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
The embodiment of the invention also provides a first arrival picking device, which is used for improving the first arrival picking efficiency and the first arrival picking accuracy and comprises the following components:
the characteristic attribute extraction module is used for extracting the characteristic attribute of the seismic data test sample;
the first arrival acquisition module is used for inputting the characteristic attributes of the seismic data test sample into the trained first arrival picking model and acquiring the first arrival of the seismic data test sample; the first arrival picking model is based on a U-net full convolution neural network;
wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
In the embodiment of the invention, the characteristic attribute of the seismic data test sample is extracted, the characteristic attribute of the seismic data test sample is input into a trained first arrival picking model, and a first arrival of the seismic data test sample is obtained, wherein the first arrival picking model is a first arrival picking model based on a U-net full convolution neural network; wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer. According to the embodiment of the invention, the first arrival picking model based on the U-net full convolution neural network has fewer network layers, fewer network parameters, higher network training speed, easier convergence and higher network prediction precision.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, it is obvious that the drawings in the following description are only some embodiments of the present invention, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts. In the drawings:
fig. 1 is a flowchart illustrating an implementation of a first arrival picking method according to an embodiment of the present invention;
FIG. 2 is a flowchart illustrating an implementation of training a first arrival picking model according to an embodiment of the present invention;
FIG. 3 is a functional block diagram of a first arrival pick-up device according to an embodiment of the present invention;
fig. 4 is a functional block diagram of training a first arrival picking model in the first arrival picking apparatus according to the embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the embodiments of the present invention are further described in detail below with reference to the accompanying drawings. The exemplary embodiments and descriptions of the present invention are provided to explain the present invention, but not to limit the present invention.
Although the present invention provides the method operation steps or apparatus structures as shown in the following embodiments or figures, more or less operation steps or module units may be included in the method or apparatus based on conventional or non-inventive labor. In the case of steps or structures which do not logically have the necessary cause and effect relationship, the execution order of the steps or the block structure of the apparatus is not limited to the execution order or the block structure shown in the embodiment or the drawings of the present invention. The described methods or modular structures, when applied in an actual device or end product, may be executed sequentially or in parallel according to embodiments or the methods or modular structures shown in the figures.
Aiming at the defects of low first arrival picking efficiency and poor accuracy of the first arrival picking method in the prior art, the applicant of the invention provides a first arrival picking method and a device, the characteristic attributes of the seismic data test sample are extracted and input into a trained first arrival picking model based on a U-net full convolution neural network to obtain the first arrival of the seismic data test sample.
Fig. 1 shows a flow of implementing the first arrival picking method provided by the embodiment of the present invention, and for convenience of description, only the parts related to the embodiment of the present invention are shown, and the detailed description is as follows:
as shown in fig. 1, the first arrival picking method includes:
wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
First arrival (also called first arrival time) refers to the time when the wavefront of seismic waves reaches a certain observation point and the vibration of particles detected by a detector on the observation point is called the first arrival time of the waves, which is called the first arrival for short. Specifically, the positions of the wave crest, the wave trough, the zero point and the like of the seismic wave can be picked up according to the excitation polarity of the seismic source.
The characteristic attribute of the seismic data refers to the geometrical, kinematic, dynamic or statistical characteristics of the seismic waves obtained by processing the pre-stack or post-stack seismic data. However, to date, there has been no recognized classification of seismic signature attributes. The kinematics and dynamics of seismic waves have been used to classify seismic characteristic attributes into: amplitude, frequency, phase, energy, waveform, wave impedance, wave velocity and correlation and ratio, each of which comprises several to twenty categories. Thus, the seismic data test sample has characteristic attributes that characterize the amplitude, energy, and phase of the seismic data test sample. It will be understood by those skilled in the art that the seismic data test sample may include, in addition to the above-described amplitude, energy, and phase characteristic attributes, single-pass normalization attributes, phase spectral variance attributes, and the like.
In the embodiment of the invention, the first arrival picking model is based on a U-net full convolution neural network, and before the embodiment of the invention is implemented, the first arrival picking model is trained to obtain the trained first arrival picking model, so that the first arrival of the seismic data test sample is picked. The first arrivals of the seismic data test sample can be effectively extracted through the trained first arrival picking model.
The U-net full convolution neural network is a variant of the convolution neural network, and the whole neural network mainly comprises a contraction path and an expansion path, and is shaped like a letter U, so that the name of the U-net full convolution neural network is obtained. The contraction path is mainly used for capturing context information of the seismic data test sample, and the expansion path is mainly used for accurately positioning a part needing to be segmented in the seismic data test sample. The traditional convolution neural network needs a large amount of training sample data, even a large amount of training sample data, and the U-net full convolution neural network utilizes data enhancement, can train a small amount of training sample data, and can obtain higher accuracy rate of first arrival picking, so that the efficiency of data processing can be greatly improved.
The method comprises the following steps that a contraction path of a first arrival picking model based on a U-net full convolution neural network comprises two units, wherein each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
According to the first arrival picking model based on the U-net full convolution neural network provided by the embodiment of the invention, the contraction path and the expansion path only comprise 2 units, and each unit only comprises one convolution layer.
According to the first arrival picking model based on the U-net full convolution neural network, the downsampling layer adopts the convolution layer with the convolution kernel size of 4x4 and the convolution preset step length of 2 to downsample the input; the loss of the traditional maximum pooling layer to input is avoided, and the accuracy of network prediction is improved.
According to the U-net full convolution neural network-based first arrival picking model provided by the embodiment of the invention, the up-sampling layer firstly performs up-sampling by adopting two-dimensional neighbor interpolation, and then performs characteristic representation by connecting the up-sampling layer with the convolution kernel with the size of 3x3 and the convolution preset step length of 1, so that a grid-shaped artificial artifact caused by zero filling when the deconvolution layer performs up-sampling is avoided, and the accuracy of network prediction is improved.
Therefore, the first arrival picking model based on the U-net full convolution neural network provided by the embodiment of the invention has the advantages of fewer network layers, fewer network parameters, higher network training speed, easier convergence and higher network prediction precision.
Therefore, the embodiment of the invention takes the characteristic attribute of the seismic data test sample as the input data of the first arrival pickup model, inputs the input data into the pre-trained first arrival pickup model, and obtains the first arrival of the seismic data test sample through the trained first arrival pickup model. The first arrival picking model provided by the embodiment of the invention can greatly improve the first arrival picking efficiency of the seismic data test sample and can also improve the first arrival picking accuracy of the seismic data test sample.
In one embodiment of the invention, the seismic data includes two-dimensional seismic data and three-dimensional seismic data.
In one embodiment of the invention, the seismic data test samples are seismic data in segy format. The seismic data is generally organized in units of seismic traces and stored in an SEG-Y file format. The SEG-Y file format is one of the standard tape data formats proposed by SEG, and the SEG-Y file format is one of the most common seismic data formats in the field of oil exploration.
In one embodiment of the invention, the seismic data test sample includes one or more of:
common shot gather seismic data, common geophone gather seismic data, and common offset gather seismic data.
Additionally, the seismic data may include two-dimensional or three-dimensional common shot gather seismic data, common geophone gather seismic data, and common offset gather seismic data.
The common shot gather seismic data refers to common shot gather seismic data of a shot point, wherein the common shot gather seismic data is formed by all seismic channels excited by the same shot point and received by different demodulator probes.
The common-geophone gather seismic data refers to gather seismic data of a common-receiving-point gather, which is formed by all seismic channels received by the same geophone point and excited by different shot points, and is also used as common-geophone gather seismic data.
The common offset gather seismic data refers to gathers extracted from different common shot gather seismic data according to the same offset, and the gathers have the same offset, so the gathers are called as common offset gathers.
In one embodiment of the invention, the seismic data test sample may also include seismic data other than the common shot gather seismic data, the common geophone gather seismic data, and the common offset gather seismic data described above, such as common reflection point gather seismic data and common conversion point gather seismic data.
The common reflection point gather seismic data refers to the reflection of each observed seismic channel from the same underground point, the reflection point is called the common reflection point of the seismic channels, and the gather formed by the seismic channels is the common reflection point gather of the reflection point.
The common conversion point gather seismic data is formed by performing converted wave exploration and is similar to the common reflection point gather, which is different in calculation method and is not described in detail herein.
In one embodiment of the invention, the characteristic attributes of the seismic data test sample include one or more of:
single-pass normalization attribute, phase spectral variance attribute, dynamic gain attribute, and instantaneous amplitude attribute.
Correspondingly, step 101, extracting the characteristic attributes of the seismic data test sample, including:
the method comprises the following steps: and extracting the single-channel normalization attribute of the seismic data test sample.
The method comprises the following steps: and extracting the phase spectrum variance attribute of the seismic data test sample.
The method comprises the following steps: and extracting the dynamic gain attribute of the seismic data test sample.
The method comprises the following steps: the instantaneous amplitude attribute of the seismic data test sample is extracted.
In view of the fact that the seismic data test sample may be affected by noise, environmental changes, near-surface influences or different acquisition instruments, and the like, so that characteristic attributes of seismic waves in the seismic data test sample, such as arrival time, amplitude, velocity, frequency, phase and the like, may differ, normalization processing needs to be performed on the seismic data test sample to obtain a normalization attribute, namely a single-channel normalization attribute, of a single seismic channel of the seismic data test sample, so that the seismic data test sample has reasonable identity and difference, and the single-channel normalization attribute of the seismic data test sample is extracted.
And when single-channel normalization is carried out on the seismic data test sample, dividing the amplitude value of each time point of each seismic channel by the maximum absolute amplitude value of the seismic channel. The range of the data amplitude after normalization processing is (-1, 1), the transverse amplitude among the seismic channels is balanced, and the first arrival demarcation point is clear. The single-channel normalization attribute of the seismic data test sample is utilized to obtain the first arrival of the seismic data test sample, and the accuracy of the first arrival picking of the seismic data test sample can be improved.
The phase spectrum of seismic waves in a seismic data test sample is one of the important characteristics of the seismic waves, as is the amplitude spectrum of the seismic waves. The phase values of the seismic waves in the seismic data test sample form a phase spectrum as a function of frequency. And after the seismic data test sample is analyzed by using a time-frequency analysis method of short-time Fourier transform, the variance of the phase spectrum is the phase spectrum variance of the seismic data test sample. The phase spectrum variance can eliminate the influence of the first arrival frequency difference caused by the effect of near-surface absorption attenuation, so that the first arrival of the seismic data test sample is obtained by utilizing the phase spectrum variance property of the seismic data test sample, and the accuracy of the first arrival picking of the seismic data test sample can also be improved.
And giving a time window length, sliding the time window along the time direction, calculating the root-mean-square amplitude in the time window to be used as a weighting factor, and weighting the input seismic channels to realize dynamic gain compensation. After the dynamic gain compensation is carried out on the seismic data test sample, the amplitude of the seismic data test sample tends to be balanced, and the energy of weak first-arrival waves in the seismic data test sample is relatively enhanced, so that the accuracy of first-arrival picking of the seismic data test sample can be improved by utilizing the dynamic gain attribute of the seismic data test sample.
After Hilbert (Hilbert) transformation is carried out on the seismic data test sample, the square root of the total energy of the real part and the imaginary part of the seismic data test sample is the instantaneous amplitude of the seismic data test sample, the energy change of the seismic data test sample on a time domain is mainly reflected, and a first arrival demarcation point can be highlighted.
In one embodiment of the invention, in the first arrival picking model based on the U-net full convolution neural network: and connecting an ELU activation function for nonlinear transformation after the convolution layers in the contraction path and the expansion path, wherein the convolution kernel size of the convolution layers is 3x3, and the preset convolution step length is 1. The ELU activation function can enable the neural network to converge more quickly, and meanwhile, the prediction accuracy of the neural network is improved.
The convolution kernel size of the downsampling layer in the contraction path is 4x4, and the convolution preset step size is 2; the up-sampling layer in the expansion path performs up-sampling by adopting two-dimensional neighbor interpolation, and then is connected with a convolution layer, wherein the size of the convolution kernel is 3x3, and the preset convolution step length is 1;
the jump connection combines the output of each convolution layer of the contraction path and the output of each up-sampling layer corresponding to the expansion path as the input of the convolution layer corresponding to the expansion path; the network output layer outputs a binary first arrival segmentation graph; and the position with the value of 1 on the first-break segmentation graph is the first break of the seismic data test sample.
In the embodiment of the invention, based on a first arrival picking model of a U-net full convolution neural network, the size of a down-sampling layer convolution kernel is 4x4, the preset convolution step length is 2, and down-sampling is carried out on input; the loss of the traditional maximum pooling layer to input is avoided, and the accuracy of network prediction is improved; the up-sampling layer adopts two-dimensional neighbor interpolation for up-sampling, then is connected with a convolution layer, the size of a convolution kernel is 3x3, the preset step length of convolution is 1, the latticed artificial artifact caused by zero filling when the deconvolution layer is up-sampled is avoided, and the accuracy of network prediction is improved.
In a further embodiment, in order to further improve the first arrival pickup efficiency and pickup accuracy, the seismic data is divided into:
the method comprises the steps of seismic data training samples, seismic data verification samples and seismic data testing samples.
The preset proportion is a preset proportion, and in view of the fact that the U-net full convolution neural network can accurately acquire the first arrival of the seismic data test sample with high precision only by a small amount of training sample data, the preset proportion can be preset to be 5% and the like, and those skilled in the art can understand that the preset proportion can also be preset to be 4%, 6% and the like. In addition, the preset ratio may be preset to 3%, which is not particularly limited in the embodiment of the present invention.
When the seismic data are selected to form the seismic data training sample, the seismic data can be selected according to the preset step length. Since the prestack seismic database contains various types of seismic data, selecting the seismic data according to the preset step length represents the diversity of the prestack seismic data.
The preset step length is a preset step length, and in view of the fact that the U-net full convolution neural network can accurately acquire the first arrival of the seismic data test sample with high precision only by a small amount of training sample data, the preset step length can be preset according to actual requirements and specific conditions, and the embodiment of the invention does not specially limit the preset step length.
The method comprises the steps of obtaining pre-stack seismic data, and dividing the pre-stack seismic data into seismic data training samples, seismic data verification samples and seismic data testing samples according to a preset proportion or a preset step length. The preset proportion is a preset proportion, and in view of the fact that the improved U-net full convolution neural network can acquire the first arrival of the seismic data with high precision only by a small amount of training sample data, the seismic data can be divided according to the proportion of 5%, 5% and 90%, the preset step length can be that every 100 samples are taken as seismic data training samples, every 100 samples are taken as seismic data verification samples, and the rest are taken as test samples.
In the embodiment of the invention, the seismic data are divided into the seismic data training sample, the seismic data verification sample and the seismic data test sample according to the preset proportion or the preset step length, and the first arrival pickup efficiency can be improved by considering that the seismic data training sample is a small amount of seismic data selected from a pre-stack seismic database, and the small amount of seismic data represents the diversity of the pre-stack seismic data.
Fig. 2 illustrates an implementation flow of training the first arrival picking model provided by the embodiment of the present invention, and for convenience of description, only the relevant parts related to the embodiment of the present invention are shown, and the following details are described below:
in a further embodiment, in order to further improve the first arrival picking efficiency and picking accuracy, as shown in fig. 2, the process of training the first arrival picking model comprises:
and step 205, stopping training if the accuracy change of the first arrival picking model is not greater than the preset change, and obtaining the trained first arrival picking model.
The characteristic attributes of the seismic data training samples represent the characteristics of the seismic data training samples such as amplitude, phase and frequency, and therefore the characteristic attributes of the seismic data training samples are extracted and serve as the input of the whole first arrival picking model. In addition, the first arrivals of the seismic data training sample and the seismic data verification sample are picked up, and the first arrivals of the seismic data training sample and the seismic data verification sample are used as the label data of the first arrival picking model. The first arrivals of the seismic data training sample and the seismic data verification sample can be manually picked up, the positions of the first arrivals are set to be 1, and other positions are set to be 0, so that binary first arrival label data with the same size as the corresponding seismic data are manufactured. And training the first arrival picking model by using the characteristic attributes and the label data of the extracted seismic data training samples. And simultaneously, verifying the sample characteristic attribute and monitoring the accuracy of the first arrival picking model by using the seismic data.
The network parameters of the first arrival picking model mainly comprise the number of network layers, the number of convolution kernels, iteration times, the type of an optimizer, weight parameters, bias parameters and the like. During the training, network parameters of the first arrival picking model are adjusted so that the first arrival picking model meets the requirements on accuracy.
In an embodiment of the present invention, the variation of the accuracy of the first arrival picking model comprises: (1) the accuracy corresponding to the first arrival picking model after the current iterative training and the relative change value of the accuracy of the first arrival picking model after the previous iterative training; or (2) rate of change of accuracy. The change rate of the accuracy is equal to the ratio of the accuracy corresponding to the first arrival picking model after the current iterative training, the relative change value of the accuracy of the first arrival picking model after the previous iterative training and the accuracy of the first arrival picking model after the previous iterative training.
That is, (1) the relative variation value of the accuracy can be determined by the following formula:
ΔL＝Lat present-LThe previous time；
Wherein, Δ L represents the accuracy corresponding to the first arrival picking model after the current iterative training, and the relative variation value of the accuracy of the first arrival picking model after the previous iterative training, LAt presentRepresenting the corresponding accuracy, L, of the first arrival pickup model after the current iterative trainingThe previous timeAnd the accuracy of the first arrival picking model after the previous iterative training is shown.
Alternatively, (2) the rate of change of accuracy can be determined by the following equation:
α＝ΔL/Lthe previous time＝(LAt present-LThe previous time)/LThe previous time；
Wherein α represents the change rate of the accuracy, Δ L represents the accuracy corresponding to the first arrival picking model after the current iterative training, and the relative change value of the accuracy of the first arrival picking model after the previous iterative training, LAt presentRepresenting the corresponding accuracy, L, of the first arrival pickup model after the current iterative trainingThe previous timeAnd the accuracy of the first arrival picking model after the previous iterative training is shown.
In an embodiment of the present invention, for example, when the change of the accuracy of the first arrival picking model is a relative change value of the accuracy, the preset change may be set to 0.01 or 0.05, and it can be understood by those skilled in the art that the preset change may also be set to 0.03 or 0.04 according to actual needs, which is not particularly limited by the embodiment of the present invention.
In another embodiment of the present invention, for example, when the change of the accuracy of the first arrival picking model is a change rate of the accuracy, the preset change may be set to 1% or 2%, and it can be understood by those skilled in the art that the preset change may also be set to 3% or 4% according to actual needs, which is not particularly limited by the embodiment of the present invention.
In the process of training the first arrival picking model, network parameters of the first arrival picking model are continuously adjusted, training is stopped until the change of the accuracy of the first arrival picking model is not larger than the preset change, the trained first arrival picking model is obtained, the first arrival of the seismic data test sample is obtained by using the trained first arrival picking model, and the efficiency and the accuracy of the first arrival picking of the seismic data test sample can be improved.
In the embodiment of the invention, the characteristic attribute of the seismic data training sample is extracted, the first arrival of the seismic data training sample is picked up to be used as the label data, the first arrival picking model is trained by using the characteristic attribute of the seismic data training sample and the label data, the accuracy of the first arrival picking model is supervised by using the characteristic attribute of the seismic data training sample and the label data, the training is stopped if the variation of the accuracy of the first arrival picking model is not more than the preset variation, the trained first arrival picking model is obtained, and the first arrival picking efficiency and the picking accuracy can be further improved.
In one embodiment of the invention, similar to the seismic data test sample, the characteristic attributes of the seismic data training sample also include a single-channel normalization attribute, a phase spectral variance attribute, a dynamic gain attribute, and an instantaneous amplitude attribute. Correspondingly, step 201, extracting the characteristic attributes of the seismic data training sample, including:
the method comprises the following steps: and extracting the single-channel normalized attribute of the seismic data training sample.
The method comprises the following steps: and extracting the phase spectrum variance attribute of the seismic data training sample.
The method comprises the following steps: and extracting the dynamic gain attribute of the seismic data training sample.
The method comprises the following steps: transient amplitude attributes of seismic data training samples are extracted.
In the embodiments of the present invention, the feature attribute of the seismic data training sample is similar to the feature attribute of the seismic data test sample, and the feature attribute of the seismic data training sample is similar to the feature attribute of the seismic data test sample.
In a further embodiment, in order to further improve the first arrival picking efficiency and picking accuracy, the process of training the first arrival picking model further includes:
and step, performing linear dynamic correction on the seismic data training sample.
Correspondingly, step 201, extracting the characteristic attributes of the seismic data training sample, including:
and step two, extracting the characteristic attribute of the seismic data training sample after linear motion correction.
The linear dynamic correction means that the arrival time of reflected waves from the same interface and the same point on each seismic channel with different offset is corrected to be the echo time of a common central point, namely normal time difference correction, and the aim is to realize the homodromous superposition. After linear dynamic correction is carried out, the seismic data in a fixed short-time window are extracted, the time window takes the first arrival wave as the center, the sample size can be reduced through the linear dynamic correction, and the network computing speed is improved.
In the embodiment of the invention, the seismic data training sample is subjected to linear motion correction, and the characteristic attribute of the seismic data training sample after linear motion correction is extracted, so that the first arrival picking efficiency and the picking accuracy can be further improved.
In an embodiment of the present invention, the first arrival picking method further includes, on the basis of the above method steps:
performing linear dynamic correction on the seismic data test sample;
the characteristic attributes of the seismic data test sample are extracted by the following steps:
extracting characteristic attributes of the seismic data test sample after linear motion correction;
for linear dynamic correction of the seismic data test sample and linear dynamic correction of the seismic data training sample, please refer to the description of the relevant parts of the above embodiments, which is not described herein in detail.
The embodiment of the invention also provides a first arrival picking device, which is described in the following embodiment. Since the principle of these devices to solve the problem is similar to the first arrival picking method, the implementation of these devices can be referred to the implementation of the method, and the repeated descriptions are omitted.
Fig. 3 shows functional modules of a first arrival picking apparatus provided in an embodiment of the present invention, and for convenience of description, only the parts related to the embodiment of the present invention are shown, and detailed descriptions are as follows:
referring to fig. 3, each module included in the first arrival pickup apparatus is configured to perform each step in the embodiment corresponding to fig. 1, and specific reference is made to fig. 1 and the related description in the embodiment corresponding to fig. 1, which are not repeated herein. In the embodiment of the present invention, the first arrival picking apparatus includes a feature attribute extracting module 301 and a first arrival obtaining module 302.
And the characteristic attribute extraction module 301 is used for extracting the characteristic attributes of the seismic data test sample.
A first arrival obtaining module 302, configured to input the feature attributes of the seismic data test sample into the trained first arrival picking model, and obtain a first arrival of the seismic data test sample; the first arrival picking model is based on a U-net full convolution neural network;
wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
In the embodiment of the present invention, the first arrival obtaining module 302 obtains the first arrival of the seismic data test sample by using the first arrival pickup model based on the U-net full convolution neural network, and the first arrival pickup model based on the U-net full convolution neural network provided in the embodiment of the present invention has fewer network layers, fewer network parameters, faster network training speed, easier convergence, and higher network prediction precision.
In one embodiment of the invention, the seismic data test sample includes one or more of:
common shot gather seismic data, common geophone gather seismic data, and common offset gather seismic data.
In one embodiment of the invention, the characteristic attributes of the seismic data test sample include one or more of:
single-pass normalization attribute, phase spectral variance attribute, dynamic gain attribute, and instantaneous amplitude attribute.
Accordingly, the feature attribute extraction module 301 includes:
and the first characteristic attribute extraction unit is used for extracting the single-channel normalized attribute of the seismic data test sample.
And the second characteristic attribute extraction unit is used for extracting the phase spectrum variance attribute of the seismic data test sample.
And the third characteristic attribute extraction unit is used for extracting the dynamic gain attribute of the seismic data test sample.
And the fourth characteristic attribute extraction unit is used for extracting the instantaneous amplitude attribute of the seismic data test sample.
Fig. 4 shows functional modules for training a first arrival picking model in a first arrival picking apparatus provided in an embodiment of the present invention, and for convenience of description, only the parts related to the embodiment of the present invention are shown, which are detailed as follows:
in a further embodiment, in order to further improve the first arrival picking efficiency and the picking accuracy, referring to fig. 4, each module included in the first arrival picking model trained in the first arrival picking apparatus is used to execute each step in the embodiment corresponding to fig. 2, which is specifically referred to fig. 2 and the related description in the embodiment corresponding to fig. 2, and is not repeated herein. In the embodiment of the present invention, the training of the first arrival picking model in the first arrival picking apparatus includes an extraction module 401, a tag data obtaining module 402, a training module 403, a supervision module 404, and a stopping module 405.
The extraction module 401 is configured to extract feature attributes of the seismic data training sample and the seismic data verification sample.
And a tag data acquisition module 402, configured to pick up first arrivals of the seismic data training sample and the seismic data verification sample as tag data.
A training module 403, configured to train a first arrival picking model using the feature attributes and the label data of the seismic data training samples;
a supervision module 404 for supervising the accuracy of the first arrival picking model using the seismic data validation sample characteristic attributes and the tag data;
a stopping module 405, configured to stop training if the change in the accuracy of the first arrival picking model is not greater than a preset change, and obtain a trained first arrival picking model.
In the embodiment of the invention, an extraction module 401 extracts characteristic attributes of seismic data training samples, a label data acquisition module 402 picks up first arrivals of the seismic data training samples as label data, a training module 403 trains a first arrival picking model by using the characteristic attributes and the label data of the seismic data training samples, and a supervision module 404 is used for supervising the accuracy of the first arrival picking model by using the characteristic attributes and the label data of the seismic data verification samples; and the stopping module 405 is used for stopping training if the change of the accuracy of the first arrival picking model is not greater than the preset change, obtaining the trained first arrival picking model, and further improving the first arrival picking efficiency and picking accuracy.
In a further embodiment, in order to further improve the first arrival picking efficiency and picking accuracy, in an embodiment of the present invention, on the basis of the above module structure, the first arrival picking apparatus trains a first arrival picking model, and further includes a linear motion correction module.
And the linear motion correction module is used for performing linear motion correction on the seismic data training sample.
Accordingly, the extraction module 401 comprises a first extraction unit.
And the first extraction unit is used for extracting the characteristic attribute of the seismic data training sample after linear motion correction.
In the embodiment of the invention, the linear motion correction module performs linear motion correction on the seismic data training sample, and the first extraction unit extracts the characteristic attribute of the seismic data training sample after the linear motion correction, so that the first arrival picking efficiency and the picking accuracy can be further improved.
The embodiment of the invention also provides computer equipment which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein the processor executes the computer program to realize the first arrival picking method.
An embodiment of the present invention further provides a computer-readable storage medium, where a computer program for executing the first arrival pickup method is stored in the computer-readable storage medium.
The first arrival picking method and the device provided by the embodiment of the invention have the following beneficial effects:
(1) the first-arrival picking model based on the U-net full convolution neural network can obtain a high-precision first-arrival picking result under the condition of a small amount of training sample data, and the generalization capability of the model is very strong. The robust first-arrival picking model can be established only by a few seismic data training samples (dozens of pre-stack seismic gathers), accurate first-arrival picking can be carried out on various types of seismic data, and especially the first-arrival picking of complex earth surface low signal-to-noise ratio seismic data can be carried out.
(2) The first arrival picking model based on the U-net full convolution neural network is a convolution layer, does not comprise a full connection layer, has few network parameters and is high in training speed. The earthquake first arrival picking of the first arrival picking model based on the U-net full convolution neural network has higher efficiency than that of the conventional convolution neural network.
(3) The first arrival picking model based on the U-net full convolution neural network has no limit on the size of input data, the size of a seismic data training sample does not need to be consistent with the size of a seismic data test sample, and the method is suitable for first arrival picking of pre-stack seismic data of various types and different samples.
(4) The method has the advantages that the attribute which can better represent the first arrival position characteristics is extracted from the input data (namely seismic data training samples) of the first arrival picking model based on the U-net full convolution neural network, the signal difference before and after the first arrival is sharpened, the accuracy of recognizing the first arrival in the seismic data test sample is improved, and therefore various types of first arrivals are picked accurately. Single channel normalization can balance the energy difference of each seismic channel in the transverse direction, dynamic gain compensation is used for enhancing the first arrival with weak far-offset energy, and the weak signal first arrival pickup precision is improved; the phase spectrum variance can weaken the influence of random noise above the first arrival and improve the accuracy of the first arrival pickup of the seismic data with low signal-to-noise ratio; on the other hand, the influence of the frequency difference of the first arrival waves caused by the effect of near-surface absorption attenuation is eliminated, and the first arrival pickup precision of the composite waves, the low-frequency signals, the high-frequency signals and the like is improved.
(5) The trained first arrival picking model can be migrated to other work areas to carry out first arrival picking, and seismic data of different first arrival types can be collected into a training set, so that the first arrival model accumulates knowledge and is more intelligent. If the first-motion wave characteristics of the pre-stack seismic data of other work areas are different from the first-motion wave characteristics of the training data of the first-motion pickup model, partial pre-stack seismic data can be selected from other work areas and incorporated into the training data set of the first-motion pickup model, and the new training data set is utilized to retrain the first-motion pickup model based on the U-net full convolution neural network, so that the first-motion pickup model with stronger robustness is obtained.
In summary, in the embodiments of the present invention, the feature attributes of the seismic data test sample are extracted, and the feature attributes of the seismic data test sample are input into the trained first arrival picking model, so as to obtain the first arrival of the seismic data test sample, where the first arrival picking model is a first arrival picking model based on a U-net full convolution neural network. According to the embodiment of the invention, the first arrival of the seismic data test sample is obtained by using the first arrival picking model based on the U-net full convolution neural network, so that the first arrival picking efficiency and the picking effect can be improved.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
The above-mentioned embodiments are intended to illustrate the objects, technical solutions and advantages of the present invention in further detail, and it should be understood that the above-mentioned embodiments are only exemplary embodiments of the present invention, and are not intended to limit the scope of the present invention, and any modifications, equivalent substitutions, improvements and the like made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (10)
1. A first arrival picking method, comprising:
extracting the characteristic attribute of the seismic data test sample;
inputting the characteristic attributes of the seismic data test sample into a trained first arrival picking model to obtain a first arrival of the seismic data test sample; the first arrival picking model is based on a U-net full convolution neural network;
wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
2. The method of claim 1, wherein the seismic data comprises one or more of:
common shot gather seismic data, common geophone gather seismic data, and common offset gather seismic data.
3. The method of claim 1, wherein the characteristic attributes of the seismic data test sample include one or more of:
single-pass normalization attribute, phase spectral variance attribute, dynamic gain attribute, and instantaneous amplitude attribute.
4. The method of claim 1, wherein in the U-net full convolution neural network based first arrival pick model:
connecting a convolution layer in the contraction path and the expansion path with an ELU activation function for nonlinear transformation, wherein the convolution kernel size of the convolution layer is 3x3, and the convolution preset step length is 1;
the convolution kernel size of the downsampling layer in the contraction path is 4x4, and the convolution preset step size is 2; the up-sampling layer in the expansion path performs up-sampling by adopting two-dimensional neighbor interpolation, and is connected with a convolution layer, the size of the convolution kernel is 3x3, and the preset convolution step length is 1;
the jump connection combines the output of each convolution layer of the contraction path and the output of each up-sampling layer corresponding to the expansion path as the input of the convolution layer corresponding to the expansion path;
the network output layer outputs a binary first arrival segmentation graph; and the position with the value of 1 on the first-break segmentation graph is the first break of the seismic data test sample.
5. The method of claim 1, wherein the seismic data is divided into:
the method comprises the steps of seismic data training samples, seismic data verification samples and seismic data testing samples.
6. The method of claim 1, wherein the process of training the first arrival picking model comprises:
extracting characteristic attributes of a seismic data training sample and a seismic data verification sample;
picking up first arrivals of the seismic data training sample and the seismic data verification sample as label data;
training a first arrival picking model by using the characteristic attributes and the label data of the seismic data training samples;
verifying the sample characteristic attribute and the label data by using the seismic data to supervise the accuracy of the first arrival picking model;
and if the change of the accuracy of the first arrival picking model is not more than the preset change, stopping training to obtain the trained first arrival picking model.
7. A first arrival pickup apparatus, comprising:
the characteristic attribute extraction module is used for extracting the characteristic attribute of the seismic data test sample;
the first arrival acquisition module is used for inputting the characteristic attributes of the seismic data test sample into the trained first arrival picking model and acquiring the first arrival of the seismic data test sample; the first arrival picking model is based on a U-net full convolution neural network;
wherein the contraction path of the first arrival picking model based on the U-net full convolution neural network comprises two units, and each unit comprises a convolution layer and a down-sampling layer; the expansion path of the first arrival picking model based on the U-net full convolution neural network comprises two units, wherein each unit comprises an upsampling layer and a convolution layer.
8. The apparatus of claim 7, wherein training the first arrival picking model comprises:
the extraction module is used for extracting the characteristic attributes of the seismic data training sample and the seismic data verification sample;
the tag data acquisition module is used for picking up the first arrivals of the seismic data training sample and the seismic data verification sample as tag data;
the training module is used for training a first arrival picking model by using the characteristic attributes and the label data of the seismic data training samples;
the monitoring module is used for monitoring the accuracy of the first arrival picking model by using the seismic data to verify the characteristic attribute of the sample and the label data;
and the stopping module is used for stopping training if the change of the accuracy of the first arrival picking model is not greater than the preset change, so as to obtain the trained first arrival picking model.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method of any one of claims 1 to 6 when executing the computer program.
10. A computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program for executing the method of any one of claims 1 to 6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910193634.6A CN111694053A (en) | 2019-03-14 | 2019-03-14 | First arrival picking method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910193634.6A CN111694053A (en) | 2019-03-14 | 2019-03-14 | First arrival picking method and device |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111694053A true CN111694053A (en) | 2020-09-22 |
Family
ID=72474505
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910193634.6A Pending CN111694053A (en) | 2019-03-14 | 2019-03-14 | First arrival picking method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111694053A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112464725A (en) * | 2020-10-30 | 2021-03-09 | 中国石油天然气集团有限公司 | First arrival picking method and device based on deep learning network |
| CN112464987A (en) * | 2020-10-30 | 2021-03-09 | 中国石油天然气集团有限公司 | First-arrival position prediction result evaluation method and device |
| WO2022062508A1 (en) * | 2020-09-28 | 2022-03-31 | 中国石油化工股份有限公司 | Static correction processing method and apparatus for seismic data, and computer device and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107807387A (en) * | 2017-10-31 | 2018-03-16 | 中国科学技术大学 | Acquisition methods when seismic first break based on neutral net is walked |
| US20180136351A1 (en) * | 2016-11-16 | 2018-05-17 | Korea Institute Of Geoscience And Mineral Resources | Method of first arrival picking of multi-channel seismic survey data considering sound source-receiver array |
| US10054699B1 (en) * | 2015-04-01 | 2018-08-21 | Peter J. Hutchinson | Method and apparatus for generating shear waves usable in determining distribution and orientation of geologic features in the subsurface among other uses |
| CN109143367A (en) * | 2017-06-28 | 2019-01-04 | 中国石油化工股份有限公司 | A kind of method of automatic first breaks pick-up |
-
2019
- 2019-03-14 CN CN201910193634.6A patent/CN111694053A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10054699B1 (en) * | 2015-04-01 | 2018-08-21 | Peter J. Hutchinson | Method and apparatus for generating shear waves usable in determining distribution and orientation of geologic features in the subsurface among other uses |
| US20180136351A1 (en) * | 2016-11-16 | 2018-05-17 | Korea Institute Of Geoscience And Mineral Resources | Method of first arrival picking of multi-channel seismic survey data considering sound source-receiver array |
| CN109143367A (en) * | 2017-06-28 | 2019-01-04 | 中国石油化工股份有限公司 | A kind of method of automatic first breaks pick-up |
| CN107807387A (en) * | 2017-10-31 | 2018-03-16 | 中国科学技术大学 | Acquisition methods when seismic first break based on neutral net is walked |
Non-Patent Citations (1)
| Title |
|---|
| LIANLIAN HU ET AL.: "U-net convolutional networks for first arrival picking", 《SOCIETY OF EXPLORATION GEOPHYCISTS》 * |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2022062508A1 (en) * | 2020-09-28 | 2022-03-31 | 中国石油化工股份有限公司 | Static correction processing method and apparatus for seismic data, and computer device and storage medium |
| CN112464725A (en) * | 2020-10-30 | 2021-03-09 | 中国石油天然气集团有限公司 | First arrival picking method and device based on deep learning network |
| CN112464987A (en) * | 2020-10-30 | 2021-03-09 | 中国石油天然气集团有限公司 | First-arrival position prediction result evaluation method and device |
| CN112464987B (en) * | 2020-10-30 | 2024-09-06 | 中国石油天然气集团有限公司 | First arrival position prediction result evaluation method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhu et al. | Seismic signal denoising and decomposition using deep neural networks | |
| Van den Ende et al. | A self-supervised deep learning approach for blind denoising and waveform coherence enhancement in distributed acoustic sensing data | |
| CN108387933B (en) | Method, device and system for determining formation quality factor | |
| CN108845352B (en) | Desert Denoising of Seismic Data method based on VMD approximate entropy and multi-layer perception (MLP) | |
| CN112946749B (en) | Method for suppressing seismic multiples based on data augmentation training deep neural network | |
| CN111694053A (en) | First arrival picking method and device | |
| CN104483704B (en) | Excess phase bearing calibration based on the constraint of AVO Exception Types | |
| GB2493088A (en) | Marine seismic wavefield separation using measurements vertical and radial components of particle velocity in the tau-apparent slowness domain | |
| US20160161619A1 (en) | Systems and methods for aligning a monitor seismic survey with a baseline seismic survey | |
| CN106199698A (en) | Frequency domain geological data reconstructing method based on many subwaves information | |
| CN102073064A (en) | Method for improving velocity spectrum resolution by using phase information | |
| CN116520419B (en) | Hot fluid crack channel identification method | |
| CN108226997A (en) | Seismic facies division method based on pre-stack seismic data | |
| CN117631028A (en) | Low-frequency reconstruction method for seismic data of multi-scale global information fusion neural network | |
| Li et al. | Distributed acoustic sensing vertical seismic profile data denoising based on multistage denoising network | |
| CN117031539A (en) | Low-frequency reconstruction method and system for self-supervision deep learning seismic data | |
| CN112578435A (en) | Rock ultrasonic testing first arrival picking method and system | |
| Cheng et al. | Multiscale recurrent-guided denoising network for distributed acoustic sensing-vertical seismic profile background noise attenuation | |
| CN109212609A (en) | Near surface Noise Elimination method based on wave equation continuation | |
| CN107479091B (en) | A method of extracting reverse-time migration angle gathers | |
| Xudong et al. | Pre-stack gather optimization technology based on an improved bidimensional empirical mode decomposition method | |
| CN113960660B (en) | Forward simulation based dynamic correction distortion area automatic identification and removal method | |
| Hu et al. | U-net convolutional networks for first arrival picking | |
| CN112526611A (en) | Method and device for extracting surface seismic wave quality factor | |
| CN112394393B (en) | CRP gather data volume reconstruction method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200922 |
|
| RJ01 | Rejection of invention patent application after publication |