CN117970448A - Mixed source seismic data separation method and device based on U-Net++ network - Google Patents
Mixed source seismic data separation method and device based on U-Net++ network Download PDFInfo
- Publication number
- CN117970448A CN117970448A CN202311803223.7A CN202311803223A CN117970448A CN 117970448 A CN117970448 A CN 117970448A CN 202311803223 A CN202311803223 A CN 202311803223A CN 117970448 A CN117970448 A CN 117970448A
- Authority
- CN
- China
- Prior art keywords
- separation
- data
- pseudo
- mixed
- seismic
- 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
- 238000000926 separation method Methods 0.000 title claims abstract description 174
- 238000000034 method Methods 0.000 claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 27
- 238000012545 processing Methods 0.000 claims abstract description 21
- 238000013135 deep learning Methods 0.000 claims abstract description 9
- 238000004891 communication Methods 0.000 claims description 19
- 230000006870 function Effects 0.000 claims description 14
- 238000004590 computer program Methods 0.000 claims description 13
- 238000003860 storage Methods 0.000 claims description 7
- 230000001934 delay Effects 0.000 claims description 6
- 230000004913 activation Effects 0.000 claims description 4
- 230000008569 process Effects 0.000 abstract description 14
- 230000036961 partial effect Effects 0.000 abstract description 10
- 230000004927 fusion Effects 0.000 abstract description 7
- 238000010586 diagram Methods 0.000 description 13
- 238000005070 sampling Methods 0.000 description 7
- 238000012986 modification Methods 0.000 description 5
- 230000004048 modification Effects 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 4
- 238000011176 pooling Methods 0.000 description 4
- 230000001629 suppression Effects 0.000 description 4
- 230000009286 beneficial effect Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 238000002156 mixing Methods 0.000 description 3
- 230000002829 reductive effect Effects 0.000 description 3
- 230000007704 transition Effects 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 238000009826 distribution Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 230000002093 peripheral effect Effects 0.000 description 2
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 description 1
- 230000004075 alteration Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000002238 attenuated effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 230000003247 decreasing effect Effects 0.000 description 1
- 230000003111 delayed effect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000006185 dispersion Substances 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 238000004880 explosion Methods 0.000 description 1
- 238000010304 firing Methods 0.000 description 1
- 230000000670 limiting effect Effects 0.000 description 1
- 230000014759 maintenance of location Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 238000005065 mining Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000004088 simulation Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Landscapes
- Geophysics And Detection Of Objects (AREA)
Abstract
The application provides a method and a device for separating seismic data of a mixed seismic source based on a U-Net++ network, wherein after the method acquires the seismic data of the mixed seismic source to be separated, pseudo-separation processing is carried out on the seismic data of the mixed seismic source to be separated to obtain pseudo-separation data; and (3) adopting a sparse inversion separation method to separate partial pseudo-separation data to obtain a separation result of the partial data. And converting the pseudo separation data and the corresponding separation result into a common-detector gather, making a tag data pair, training a deep learning mixed seismic source seismic data separation model by using the tag data pair, inputting the pseudo separation data of the common-detector gather into the trained mixed seismic source seismic data separation model to obtain separated mixed seismic source seismic data, wherein the data separation model adopts a U-Net++ network. The method solves the problems of partial feature loss and insufficient precision of the feature fusion part caused by downsampling in the network model training process.
Description
Technical Field
The application relates to the technical field of seismic data separation, in particular to a method and a device for separating seismic data of a mixed seismic source based on a U-Net++ network.
Background
Efficient acquisition of seismic data is often based on economic and high quality. Conventional seismic data acquisition often requires a sufficiently large firing interval to avoid mutual interference between different sources, greatly consuming labor and material costs.
The mixed seismic source acquisition technology is favored by the vast seismic prospecting workers since the establishment, and a plurality of seismic sources are excited at different positions in a fixed coding mode at the same time or in a delayed mode, so that mixed seismic source seismic data are obtained, the sampling efficiency is improved, but a large amount of aliasing noise is doped in the process of separating mixed data, the signal-to-noise ratio of seismic records is reduced, and the subsequent data processing is influenced, so that the requirement for a high-precision mixed data separation method is urgent. The existing method for separating seismic data of a mixed seismic source based on data driving mainly adopts a U-Net network, partial characteristics are lost in the original U-Net downsampling process, the precision of the characteristic fusion part is insufficient, more effective signals are removed in a separation result, and the separation signal-to-noise ratio is low.
Disclosure of Invention
The embodiment of the application aims to provide a method and a device for separating seismic data of a mixed seismic source based on a U-Net++ network, which are used for solving the problems of partial feature loss and insufficient precision of feature fusion parts in the sampling process.
In a first aspect, a method for separating seismic data of a mixed source based on a U-Net++ network is provided, and the method may include:
Acquiring mixed source seismic data to be separated; the mixed seismic source seismic data to be separated are obtained by exciting a plurality of seismic sources with different time delays and collecting the seismic data, wherein the mixed seismic source comprises at least two seismic sources;
Performing pseudo-separation processing on the mixed source seismic data to be separated to obtain pseudo-separation data, wherein the pseudo-separation data comprises pseudo-separation data of a first part and pseudo-separation data of a second part;
adopting a sparse inversion separation method to separate the pseudo-separation data of the first part to obtain a separation result of the pseudo-separation data of the first part;
Converting the pseudo separation data of the first part and the corresponding separation result into a first common detector point gather, making a tag data pair, and training a separation model of the deep learning mixed source seismic data by using the tag data pair;
And inputting the pseudo separation data of the second part corresponding to the second common-wave-point gather into a trained separation model of the mixed seismic source seismic data to obtain separated mixed seismic source seismic data, wherein the separation model adopts a U-Net++ network.
In one possible implementation, the separation model of the hybrid source seismic data uses the formula: y=net (x, epsilon);
Wherein x is the pseudo separation data of the second part, y is the output of the separation model of the mixed source seismic data, epsilon is different parameters required to be optimized in the separation model of the mixed source seismic data, and Net represents the network architecture of the separation model of the mixed source seismic data.
In one possible implementation, the different parameters in ε include a learning rate, a batch data size of a separation model that trains the hybrid source seismic data, and an activation function.
In a second aspect, a hybrid source seismic data separation device based on a U-Net++ network is provided, the device may include:
The acquisition unit is used for acquiring the mixed source seismic data to be separated; the mixed seismic source seismic data to be separated are obtained by exciting a plurality of seismic sources with different time delays and collecting the seismic data, wherein the mixed seismic source comprises at least two seismic sources;
the processing unit is used for carrying out pseudo separation processing on the mixed seismic source seismic data to be separated to obtain pseudo separation data, wherein the pseudo separation data comprises pseudo separation data of a first part and pseudo separation data of a second part;
and performing separation treatment on the pseudo separation data of the first part by adopting a sparse inversion separation method to obtain a separation result of the pseudo separation data of the first part;
The training unit is used for converting the pseudo separation data of the first part and the corresponding separation result into a first common detector gather, making a tag data pair, and training a separation model of the deep learning mixed source seismic data by using the tag data pair;
The acquisition unit is further configured to input pseudo-separation data of a second portion corresponding to the second common-detector gather into a trained separation model of the mixed source seismic data, so as to obtain separated mixed source seismic data, where the separation model adopts a U-net++ network.
In a third aspect, an electronic device is provided, the electronic device comprising a processor, a communication interface, a memory, and a communication bus, wherein the processor, the communication interface, and the memory are in communication with each other via the communication bus;
a memory for storing a computer program;
A processor for implementing the method steps of any one of the above first aspects when executing a program stored on a memory.
In a fourth aspect, a computer-readable storage medium is provided, in which a computer program is stored which, when being executed by a processor, carries out the method steps of any of the first aspects.
The application provides a method for separating seismic data of a mixed seismic source based on a U-Net++ network, which comprises the steps of after the seismic data of the mixed seismic source to be separated are obtained, performing pseudo-separation processing on the seismic data of the mixed seismic source to be separated to obtain pseudo-separation data; and (3) adopting a sparse inversion separation method to separate partial pseudo-separation data to obtain a separation result of the partial data. And converting the pseudo separation data and the corresponding separation result into a common-wave-point gather, making a tag data pair, training a deep learning mixed seismic source seismic data separation model by using the tag data pair, inputting the other part of pseudo separation data of the common-wave-point gather into the trained mixed seismic source seismic data separation model to obtain separated mixed seismic source seismic data, wherein the data separation model adopts a U-Net++ network. The method solves the problems of partial characteristic loss and insufficient precision of the characteristic fusion part in the sampling process.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the embodiments of the present application will be briefly described below, it should be understood that the following drawings only illustrate some embodiments of the present application and should not be considered as limiting the scope, and other related drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic structural diagram of a U-Net network according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a DenseNet network according to an embodiment of the present application;
FIG. 3 is a schematic diagram of a U-Net++ network according to an embodiment of the present application;
FIG. 4 is a schematic flow chart of a method for separating seismic data of a mixed seismic source based on a U-Net++ network, which is provided by the embodiment of the application;
FIG. 5 is a schematic diagram showing discrete distribution of single source data, aliasing data and aliasing noise in a common detector gather according to an embodiment of the present application;
FIG. 6 is a graph showing a loss curve during training using different network models according to an embodiment of the present application;
FIG. 7 is a graph comparing the separation results of different models provided in the examples of the present application;
FIG. 8 is a comparison chart of common shot gather aliasing noise suppression according to an embodiment of the present application;
FIG. 9 is a graph comparing signal to noise ratios of single shots of a common shot gather according to an embodiment of the present application;
FIG. 10 is a graph showing a comparison of common-detector gathers for different network models according to an embodiment of the present application;
FIG. 11 is a comparison chart of a common-detector gather aliasing noise suppression according to an embodiment of the present application;
FIG. 12 is a graph showing a comparison of signal to noise ratios of a single channel for a common detector gather according to an embodiment of the present application;
FIG. 13 is a schematic structural diagram of a seismic data separation device for a hybrid seismic source based on a U-Net++ network according to an embodiment of the present application;
fig. 14 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and completely with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some embodiments, but not all embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments of the application without making any inventive effort, are intended to be within the scope of the application.
As can be seen from fig. 1, the U-Net network is considered as a symmetrical network architecture, with downsampling on the left and upsampling on the right, which may be functionally referred to as the coding layer and upsampling as the decoding layer. The input set of the network is the same as the previous layer, the coding layer part is used for extracting image characteristics, and the coding layer part consists of a plurality of downsampling modules, each module mainly comprises a convolution layer and a pooling layer, the convolution layer comprises the repeated application of two 3x3 convolutions (filled with 0), each convolution is activated by adopting a ReLU function, the maximum pooling operation of 2x2 is carried out, and the number of characteristic channels is doubled in each downsampling operation step. The decoding layer is used for recovering the image size and classifying pixels and consists of a plurality of up-sampling modules, each module mainly comprises a convolution layer and a deconvolution layer, after each deconvolution layer, the size of the feature map is enlarged to be 2 times of the original size, and the number of the feature maps is opposite to be 1/2 of the original size; likewise, there are four skip links between encoding and decoding, the purpose of which is to connect the up-sampling result and the output of a sub-module to the encoder with the same resolution as the next sub-module input to the decoder, to preserve more location information and context information. Each 64 component feature vector is mapped to the desired class using a 1x1 convolution at the last layer, the network having a total of 23 convolutions.
As shown in fig. 2, the DenseNet network uses DenseBlock +transition structure, and in the DenseBlock layers, the feature maps of the layers are uniform in size and can be connected in the channel dimension. DenseBlock consists of BN+ReLU+3×3Conv structures. For the Transition layer, its main function is to connect two neighbors DenseBlock and reduce the feature map size. The Transition layer includes a1×1 convolution and 2×2 average pooling, consisting of bn+relu+1×1conv+2×2AvgPooling.
The U-Net++ network combines the U-Net network with DenseNet network, which not only can make up the defects of partial characteristic loss, inaccurate precision of characteristic fusion part and the like in the down sampling process, but also can introduce a deep supervision mechanism to fill the defect that the middle layer cannot be counter-propagated. The network structure is shown in fig. 3. The network consists of a four-layer U-Net structure, and consists of two main parts of an encoder and a decoder. The encoder consists of a plurality of convolution layers and a pooling layer and is used for extracting the characteristic information of the input image layer by layer; the decoder gradually restores the characteristic information extracted by the encoder to the original image size through upsampling and bilinear interpolation, and a fusion module is introduced into each decoder layer. The black part is an original U-Net network and comprises a long connecting part; green and blue represent densely connected blocks; a deep supervision mechanism is introduced into each layer of U-Net output, and the superposition of the loss functions is carried out by supervising the output of each U-Net network with different depth. The specific implementation operation is to add a 1×1 convolution kernel after X 0,1、X0,2、X0,3、X0,4 in the figure, which is equivalent to unsupervised output of each branch U-Net.
The U-Net++ network has the following improvements over the original U-Net network and DenseNet network: (1) precision improvement. The method integrates the advantages of different features, and the X 0,4 adds the output features of all layers, so that the network output effect is better. (2) flexible construction. By using a deep supervision mechanism (L part in fig. 3), training conditions of different levels of U-Net networks can be observed, and the different levels of U-Net networks can be selected for testing according to results of a training set and a verification set, so that parameter quantity is reduced, and training speed is increased.
The preferred embodiments of the present application will be described below with reference to the accompanying drawings of the specification, it being understood that the preferred embodiments described herein are for illustration and explanation only, and not for limitation of the present application, and embodiments of the present application and features of the embodiments may be combined with each other without conflict.
Fig. 4 is a flow chart of a method for separating seismic data of a mixed seismic source based on a U-net++ network according to an embodiment of the present application. As shown in fig. 4, the method may include:
step S410, obtaining the mixed source seismic data to be separated.
The seismic data of the mixed seismic sources to be separated are obtained by carrying out seismic data mixed acquisition on at least two seismic sources, namely the mixed seismic sources comprise at least two seismic sources.
And S420, performing pseudo-separation processing on the seismic data of the mixed seismic source to be separated to obtain pseudo-separation data.
The pseudo-split data includes a first portion of pseudo-split data and a second portion of pseudo-split data.
And S430, performing separation processing on the pseudo separation data of the first part by adopting a sparse inversion separation method to obtain a separation result of the pseudo separation data of the first part.
Step S440, the pseudo separation data of the first part and the corresponding separation result are converted into a first common detection point gather, a label data pair is manufactured, and the label data pair is used for training a separation model of the deep learning mixed source seismic data.
The separation model may employ a U-Net++ network.
The separation model of the seismic data of the mixed source adopts the following formula: y=net (x, epsilon);
Where x is the pseudo-split data of the second portion, y is the output of the split model of the blended source seismic data, ε is the different parameters that need to be optimized in the split model of the blended source seismic data, and Net represents the network architecture of the split model of the blended source seismic data.
For different networks, its training process is also referred to as the optimization process for parameters.
In the network structure, the different parameters represented by epsilon mainly comprise learning rate, batch data size of a training seismic data denoising model, an activation function and the like. The learning rate can be dynamically adjusted in a stepwise decreasing manner. Proper selection of batch data sizes may enable more accurate updating toward extrema. In order to prevent gradient explosion or disappearance, reLU function solution can be introduced, gradient still cannot be greatly reduced through multi-layer counter propagation, and gradient dispersion problem is solved from the front.
And S450, inputting the pseudo separation data of the second part corresponding to the second common-wave-detecting point gather into a trained separation model of the mixed source seismic data to obtain separated mixed source seismic data.
In some embodiments, during training of the data separation model, the acquired seismic data for the plurality of sources may be as follows: the sampling interval is 4ms, the observation time is 4s,100 detectors, 100 seismic sources, and 100 single shot data are obtained through conventional acquisition.
In fig. 5, fig. 5 (a) shows 50 th single-source data, first, a mixed source operator is set to synthesize aliasing data with a mixing degree of 2, as in fig. (b), the mixing mode is that 1 st single-source and 51 st single-source form 1 st mixed-source, 2 nd single-source and 52 nd single-source form 2 nd mixed-source, and so on, so as to obtain aliasing noise shown in fig. (c) in discrete distribution in a common-detection-point gather.
In the embodiment of the application, in the training process, the patch size is 128×128, the number is 8652, the data block size is 32, the training times are 160, the initial learning rate is 0.001, and the learning rate is attenuated to 0.8 of the previous item every 10 times. The experiment adopts an optimizer as Adam, so that the learning rate can be dynamically adjusted, and the convergence speed is faster.
The loss function used may be the Mean Square Error (MSE) calculated as shown in the following equation.
Wherein n is the number of patches in a single patch; x i and y i are the output prediction data and tag data of the ith patch, respectively.
The evaluation index is the signal-to-noise ratio (SNR), and the calculation method is shown in the following formula.
Wherein x is data denoised by the seismic data denoising model, and y is corresponding effective signal tag data.
In embodiments of the application, the blended source seismic data may be written as a linear combination of conventional source data:
Ubl＝ΓblU
where U bl represents the blended source seismic data, U represents the single shot data to be separated (single source data), Γ bl represents the blended source operator.
Assuming that there are m single sources, n detectors, and m single sources combined into g blended sources on a two-dimensional survey line, the blended source operator Γ bl may be expressed in the following form:
Wherein the method comprises the steps of T jk represents the delay of the kth single shot in the jth blended source that participates in the blending. Let ρ jk =0 if the jth blended source does not contain the kth single source.
Typically, in actual seismic acquisition, the number of blended sources is less than Shan Zhenyuan, i.e., g < m, so Γ bl is an underdetermined matrix,Is not available, and cannot be directly obtained, but the pseudo-inverse can be constructed by the following formula:
wherein H represents the conjugate transpose of the matrix, resulting in pseudo-split data < U > -for the hybrid source seismic data:
The aliased noise in the pseudo-split data is distributed in incoherent form in the non-common shot gather, at which time the mixed source seismic data separation problem translates into a random noise suppression-like problem.
Analysis of experimental results:
The network model is trained by adopting the parameters, wherein fig. 6 shows a loss curve in the training process by utilizing different network models, the upper part of fig. 6 shows a loss curve of the original U-Net network training, and the lower part of fig. 5 shows a loss curve of the U-Net++ network training. From the convergence, the first iteration is larger in training error because no parameters can be used for reference. In the 30 th to 50 th iterations, the original U-Net and U-Net++ networks have small amplitude fluctuation, and the reason for the occurrence of the situation is probably due to the overlarge learning rate, but the U-Net++ network is better than the original U-Net network from the aspect of overall convergence.
Fig. 7 is a comparison graph of separation results (common shot gather) of different models, wherein fig. a is pseudo separation data of the common shot gather, fig. b is a separation result of a sparse inversion method, fig. c is a separation result of an original U-Net network, and fig. d is a separation result of a U-net++ network.
Fig. 8 is a comparison diagram of common shot gather aliasing noise suppression, where (a) is aliasing noise suppressed by a sparse inversion method, (b) is aliasing noise suppressed by an original U-Net network, and (c) is aliasing noise suppressed by a U-net++ network. It can be seen that the separation result of the U-Net++ network is better than that of the sparse inversion method and the original U-Net network, and the effective signals are removed less.
Fig. 9 is a graph comparing signal to noise ratios of single shots of a common shot gather, and it can be seen that the separation signal to noise ratio of the U-net++ network is the highest.
Fig. 10 is a comparison graph of separation results (co-detector gather) of different network models, where (a) is pseudo-separated co-detector gather data, (b) is a sparse inversion method separation result, (c) is an original U-Net network separation result, and (d) is a U-net++ network separation result. It can be seen that the effect obtained by the method is better and the aliasing noise removal degree is higher.
Fig. 11 is a comparison graph of common-detector gather aliasing noise suppression, where (a) is aliasing noise suppressed by a sparse inversion method, (b) is aliasing noise suppressed in the original U-Net network, and (c) is aliasing noise suppressed by the U-net++ network.
Fig. 12 is a graph comparing signal-to-noise ratios of a common-detector gather single-channel. It can be seen from fig. 9 and 12 that the U-net++ network separation signal-to-noise ratio is higher relative to other methods.
In summary, aiming at the problems of partial feature loss and insufficient precision of feature fusion parts caused by the original U-Net downsampling process, a dense connection and deep supervision mechanism is added, a U-Net++ network is introduced to train a network model, and finally the trained network is applied to realize mixed mining data separation. The experimental result of the simulation data shows that the U-Net++ network used by the application has better separation effect and higher retention of effective signals compared with the original U-Net.
Corresponding to the method, the embodiment of the application also provides a device for separating the seismic data of the mixed seismic source based on the U-Net++ network, as shown in FIG. 13, which comprises the following steps:
An acquisition unit 1310 for acquiring mixed source seismic data to be separated; the mixed seismic source seismic data to be separated are obtained by exciting a plurality of seismic sources with different time delays and collecting the seismic data, wherein the mixed seismic source comprises at least two seismic sources;
A processing unit 1320, configured to perform pseudo-separation processing on the mixed source seismic data to be separated to obtain pseudo-separation data, where the pseudo-separation data includes pseudo-separation data of a first portion and pseudo-separation data of a second portion;
and performing separation treatment on the pseudo separation data of the first part by adopting a sparse inversion separation method to obtain a separation result of the pseudo separation data of the first part;
A training unit 1330, configured to convert the pseudo separation data of the first portion and the corresponding separation result into a first common-detector gather, and make a tag data pair, and train a separation model of the deep learning hybrid source seismic data using the tag data pair;
The obtaining unit 1310 is further configured to input the pseudo-split data of the second portion corresponding to the second common-detector gather into a trained split model of the mixed source seismic data, so as to obtain split mixed source seismic data, where the split model adopts a U-net++ network.
The functions of each functional unit of the mixed seismic source seismic data separation device based on the U-Net++ network provided by the embodiment of the application can be realized through the steps of the method, so that the specific working process and beneficial effects of each unit in the mixed seismic source seismic data separation device based on the U-Net++ network provided by the embodiment of the application are not repeated herein.
The embodiment of the present application further provides an electronic device, as shown in fig. 14, including a processor 1410, a communication interface 1420, a memory 1430, and a communication bus 1440, where the processor 1410, the communication interface 1420, and the memory 1430 complete communication with each other through the communication bus 1440.
A memory 1430 for storing a computer program;
The processor 1410, when executing the program stored in the memory 1430, performs the following steps:
Acquiring mixed source seismic data to be separated; the mixed seismic source seismic data to be separated are obtained by exciting a plurality of seismic sources with different time delays and collecting the seismic data, wherein the mixed seismic source comprises at least two seismic sources;
Performing pseudo-separation processing on the mixed source seismic data to be separated to obtain pseudo-separation data, wherein the pseudo-separation data comprises pseudo-separation data of a first part and pseudo-separation data of a second part;
adopting a sparse inversion separation method to separate the pseudo-separation data of the first part to obtain a separation result of the pseudo-separation data of the first part;
Converting the pseudo separation data of the first part and the corresponding separation result into a first common detector point gather, making a tag data pair, and training a separation model of the deep learning mixed source seismic data by using the tag data pair;
And inputting the pseudo separation data of the second part corresponding to the second common-wave-point gather into a trained separation model of the mixed seismic source seismic data to obtain separated mixed seismic source seismic data, wherein the separation model adopts a U-Net++ network.
The communication bus mentioned above may be a peripheral component interconnect standard (PERIPHERAL COMPONENT INTERCONNECT, PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, or the like. The communication bus may be classified as an address bus, a data bus, a control bus, or the like. For ease of illustration, the figures are shown with only one bold line, but not with only one bus or one type of bus.
The communication interface is used for communication between the electronic device and other devices.
The Memory may include random access Memory (Random Access Memory, RAM) or may include Non-Volatile Memory (NVM), such as at least one disk Memory. Optionally, the memory may also be at least one memory device located remotely from the aforementioned processor.
The processor may be a general-purpose processor, including a central processing unit (Central Processing Unit, CPU), a network processor (Network Processor, NP), etc.; but may also be a digital signal processor (DIGITAL SIGNAL Processing, DSP), application SPECIFIC INTEGRATED Circuit (ASIC), field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA) or other Programmable logic device, discrete gate or transistor logic device, discrete hardware components.
Since the implementation manner and the beneficial effects of the solution to the problem of each device of the electronic apparatus in the foregoing embodiment may be implemented by referring to each step in the embodiment shown in fig. 4, the specific working process and the beneficial effects of the electronic apparatus provided by the embodiment of the present application are not repeated herein.
In yet another embodiment of the present application, a computer readable storage medium is provided, in which instructions are stored, which when run on a computer, cause the computer to perform the method for U-net++ network-based hybrid source seismic data separation as described in any of the above embodiments.
In yet another embodiment of the present application, a computer program product containing instructions that, when run on a computer, cause the computer to perform the method of U-Net++ network-based hybrid source seismic data separation as set forth in any one of the above embodiments is also provided.
It will be appreciated by those skilled in the art that embodiments of the present application may be provided as methods, systems, or computer program products. Accordingly, embodiments of the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, embodiments of the present application may take the form of a computer program product on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) having computer-usable program code embodied therein.
Embodiments of the present application are described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations 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.
While preferred embodiments of the present application have been described, additional variations and modifications in those embodiments may occur to those skilled in the art once they learn of the basic inventive concepts. It is therefore intended that the following claims be interpreted as including the preferred embodiment and all such alterations and modifications as fall within the scope of the embodiments of the application.
It will be apparent to those skilled in the art that various modifications and variations can be made in the embodiments of the present application without departing from the spirit or scope of the embodiments of the application. Thus, if such modifications and variations of the embodiments of the present application fall within the scope of the claims and the equivalents thereof, it is intended that such modifications and variations be included in the embodiments of the present application.
Claims (8)
1. A method for separating seismic data of a mixed seismic source based on a U-net++ network, the method comprising:
Acquiring mixed source seismic data to be separated; the mixed seismic source seismic data to be separated are obtained by exciting a plurality of seismic sources with different time delays and collecting the seismic data, wherein the mixed seismic source comprises at least two seismic sources;
Performing pseudo-separation processing on the mixed source seismic data to be separated to obtain pseudo-separation data, wherein the pseudo-separation data comprises pseudo-separation data of a first part and pseudo-separation data of a second part;
adopting a sparse inversion separation method to separate the pseudo-separation data of the first part to obtain a separation result of the pseudo-separation data of the first part;
Converting the pseudo separation data of the first part and the corresponding separation result into a first common detector point gather, making a tag data pair, and training a separation model of the deep learning mixed source seismic data by using the tag data pair;
And inputting the pseudo separation data of the second part corresponding to the second common-wave-point gather into a trained separation model of the mixed seismic source seismic data to obtain separated mixed seismic source seismic data, wherein the separation model adopts a U-Net++ network.
2. The method of claim 1, wherein the separation model of the blended source seismic data employs the formula: y=net (x, epsilon);
Wherein x is the pseudo separation data of the second part, y is the output of the separation model of the mixed source seismic data, epsilon is different parameters required to be optimized in the separation model of the mixed source seismic data, and Net represents the network architecture of the separation model of the mixed source seismic data.
3. The method of claim 2, wherein the different parameters in epsilon include a learning rate, a batch data size for training a separation model of the blended source seismic data, and an activation function.
4. A hybrid source seismic data separation device based on a U-net++ network, the device comprising:
The acquisition unit is used for acquiring the mixed source seismic data to be separated; the mixed seismic source seismic data to be separated are obtained by exciting a plurality of seismic sources with different time delays and collecting the seismic data, wherein the mixed seismic source comprises at least two seismic sources;
the processing unit is used for carrying out pseudo separation processing on the mixed seismic source seismic data to be separated to obtain pseudo separation data, wherein the pseudo separation data comprises pseudo separation data of a first part and pseudo separation data of a second part;
and performing separation treatment on the pseudo separation data of the first part by adopting a sparse inversion separation method to obtain a separation result of the pseudo separation data of the first part;
The training unit is used for converting the pseudo separation data of the first part and the corresponding separation result into a first common detector gather, making a tag data pair, and training a separation model of the deep learning mixed source seismic data by using the tag data pair;
The acquisition unit is further configured to input pseudo-separation data of a second portion corresponding to the second common-detector gather into a trained separation model of the mixed source seismic data, so as to obtain separated mixed source seismic data, where the separation model adopts a U-net++ network.
5. The apparatus of claim 4, wherein the separation model of the blended source seismic data employs the formula: y=net (x, epsilon);
Wherein x is the pseudo separation data of the second part, y is the output of the separation model of the mixed source seismic data, epsilon is different parameters required to be optimized in the separation model of the mixed source seismic data, and Net represents the network architecture of the separation model of the mixed source seismic data.
6. The apparatus of claim 5, wherein the different parameters in epsilon include a learning rate, a batch data size for training a separation model of the blended source seismic data, and an activation function.
7. An electronic device, characterized in that the electronic device comprises a processor, a communication interface, a memory and a communication bus, wherein the processor, the communication interface and the memory are in communication with each other through the communication bus;
a memory for storing a computer program;
A processor for implementing the method of any of claims 1-3 when executing a program stored on a memory.
8. A computer readable storage medium, characterized in that the computer readable storage medium has stored therein a computer program which, when executed by a processor, implements the method of any of claims 1-3.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311803223.7A CN117970448A (en) | 2023-12-26 | 2023-12-26 | Mixed source seismic data separation method and device based on U-Net++ network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311803223.7A CN117970448A (en) | 2023-12-26 | 2023-12-26 | Mixed source seismic data separation method and device based on U-Net++ network |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117970448A true CN117970448A (en) | 2024-05-03 |
Family
ID=90852177
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311803223.7A Pending CN117970448A (en) | 2023-12-26 | 2023-12-26 | Mixed source seismic data separation method and device based on U-Net++ network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117970448A (en) |
-
2023
- 2023-12-26 CN CN202311803223.7A patent/CN117970448A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10929720B2 (en) | Image processing method, image processing apparatus, and computer storage medium | |
| CN112861729A (en) | Real-time depth completion method based on pseudo-depth map guidance | |
| CN111553867B (en) | Image deblurring method and device, computer equipment and storage medium | |
| KR102313604B1 (en) | Learning method, learning device with multi feeding layers and test method, test device using the same | |
| CN110580727B (en) | Depth V-shaped dense network imaging method with increased information flow and gradient flow | |
| CN116152591B (en) | Model training method, infrared small target detection method and device and electronic equipment | |
| CN111815526B (en) | Rain image rainstrip removing method and system based on image filtering and CNN | |
| CN111551988B (en) | Seismic data anti-alias interpolation method combining deep learning and prediction filtering | |
| CN113642675B (en) | Underground rock stratum distribution imaging acquisition method, system, terminal and readable storage medium based on full waveform inversion and convolutional neural network | |
| CN114299358A (en) | Image quality evaluation method and device, electronic equipment and machine-readable storage medium | |
| CN112464725A (en) | First arrival picking method and device based on deep learning network | |
| CN114167500B (en) | Method and device for separating adjacent gun interference based on convolutional neural network | |
| CN117970448A (en) | Mixed source seismic data separation method and device based on U-Net++ network | |
| CN114236610B (en) | Iterative seismic data unmixing method and system based on depth convolution network prior | |
| CN115267899B (en) | DnCNN mixed source seismic data separation method and system based on boundary preservation | |
| CN111724410A (en) | Target tracking method based on residual attention | |
| CN114296134B (en) | Deep convolutional network seismic data unmixing method and system | |
| CN111105364A (en) | Image restoration method based on rank-one decomposition and neural network | |
| CN116091893A (en) | Method and system for deconvolution of seismic image based on U-net network | |
| CN116051532A (en) | Deep learning-based industrial part defect detection method and system and electronic equipment | |
| CN115082371A (en) | Image fusion method and device, mobile terminal equipment and readable storage medium | |
| CN115735224A (en) | Non-extraction image processing method and device | |
| CN117173552B (en) | Underwater target detection method, system, electronic equipment and storage medium | |
| CN112950638A (en) | Image segmentation method and device, electronic equipment and computer readable storage medium | |
| CN115222790B (en) | Single photon three-dimensional reconstruction method, system, equipment and storage medium |
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 |