CN110826691A - Intelligent seismic velocity spectrum pickup method based on YOLO and LSTM - Google Patents
Intelligent seismic velocity spectrum pickup method based on YOLO and LSTM Download PDFInfo
- Publication number
- CN110826691A CN110826691A CN201910973522.2A CN201910973522A CN110826691A CN 110826691 A CN110826691 A CN 110826691A CN 201910973522 A CN201910973522 A CN 201910973522A CN 110826691 A CN110826691 A CN 110826691A
- Authority
- CN
- China
- Prior art keywords
- yolo
- lstm
- velocity spectrum
- neural network
- seismic velocity
- 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
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Geophysics And Detection Of Objects (AREA)
Abstract
The invention provides an intelligent seismic velocity spectrum pickup method based on YOLO and LSTM, which specifically comprises the following steps: acquiring a seismic velocity spectrum and a corresponding manually picked time-velocity pair data set to form a training set; constructing a hybrid neural network model based on YOLO and LSTM; training the hybrid neural network model of YOLO and LSTM using "time-velocity" versus dataset; and automatically picking up an unknown seismic velocity spectrum by using the hybrid neural network model of the YOLO and the LSTM, and outputting a picked-up result by taking a time-velocity pair as a sequence. The invention has the beneficial effects that: the method realizes fully automatic speed picking, has better picking effect, and provides a new solution for the problem of automatic picking of the seismic velocity spectrum.
Description
Technical Field
The invention relates to the field of artificial intelligence, in particular to an intelligent seismic velocity spectrum picking method based on YOLO and LSTM.
Background
Velocity analysis is one of the important links in conventional seismic data processing, and is the basis of subsequent seismic processing, such as multiple suppression, migration imaging, inversion and the like. In order to obtain the result of the velocity analysis, it is currently necessary to manually perform velocity picking on the velocity spectrum, which takes a lot of manpower and time. In addition, the picking up velocity spectrum is not a simple mechanical manual labor, and is influenced by earthquake noise, multiple waves, lateral waves, diffracted waves, geological structure complexity, spatial variation of the geological structure and the like, so that a trained processor needs to perform mental labor to obtain correct and high-precision speed picking up results. Therefore, there is a need to develop a fast and accurate method for automatic velocity spectrum picking to reduce the mental and physical labor of seismic processing personnel.
In a conventional velocity analysis method, we make a velocity spectrum by fixing the two-pass reflection time of zero offset and stacking (or correlating) a common-center gather along hyperbolic traces defined by different velocities, which can be used as an image and can be mathematically expressed using a two-dimensional matrix.
We can look for energy peaks from the velocity spectrum and perform a pick-up of the pick-up stack velocity-obtaining a series of "time-velocity" pairs. These "time-velocity" pairs have a specific spatial relationship in the velocity spectrum, and the stacking velocity generally increases with time, indicating that the "time-velocity" pairs have distinct sequence features. Therefore, the picked-up superposition velocities can be regarded as a time series, and the picking-up process of the velocity spectrum image can be actually regarded as a mapping from a two-dimensional matrix to the time series.
Disclosure of Invention
The invention provides an intelligent seismic velocity spectrum picking method based on YOLO and LSTM, which aims to overcome the defects of the automatic picking velocity spectrum in the prior art, and the method comprises the steps of firstly training a mixed neural network model formed by YOLO and LSTM by using a large number of seismic velocity spectrum samples, and after training, mastering a velocity picking method by a mechanism simulating human brain learning by using the model, wherein the YOLO model has the advantages in the aspect of target detection to extract energy peak characteristics in the velocity spectrum, namely, a rough time-velocity pair is obtained, and a memory module of an LSTM hidden unit corrects the time-velocity pair by storing long interval information of the time-velocity pair and extracting the time sequence characteristics of the time-velocity pair to obtain an accurate recognition result.
The seismic velocity spectrum intelligent pickup method based on the YOLO and the LSTM comprises the following steps:
s101: acquiring a seismic velocity spectrum, and acquiring a time-velocity pair data set in the seismic velocity spectrum;
s102: constructing a hybrid neural network model based on YOLO and LSTM;
s103: taking the time-speed pair data set in the step S101 as a training set, training the mixed neural network model of the YOLO and the LSTM, and obtaining the trained mixed neural network model of the YOLO and the LSTM;
s104: and automatically picking up the seismic velocity spectrum which is not picked up by using the trained mixed neural network model of the YOLO and the LSTM, and outputting the picked-up result by taking a time-velocity pair as a sequence.
Further, step S101 is specifically as follows:
s201: extracting a plurality of common midpoint gathers from shot gather seismic data on the sea or on the land, and sequentially denoising and speed scanning the common midpoint gathers to obtain a corresponding seismic speed spectrum;
s202: time-velocity pairs are picked from the seismic velocity spectrum, resulting in a time-velocity pair dataset.
Further, step S102 specifically includes the following steps:
s301: constructing a YOLO model, wherein the adopted activation function is a LeakyReLU function, and the calculation formula is as follows:
s302: dividing the seismic velocity spectrum into n x n grids by using a YOLO model, and detecting 5 frames for each grid, wherein the number of the frames for detecting energy blobs in each input image is n x 5; n is specifically set according to the area proportion of the energy cliques in the input seismic velocity spectrum image; each box contains 5 information data; the 5 pieces of information data are x, y, w, h and confidence degrees respectively; where x and y are the coordinates of the center of the box, and w and h are the length and width of the box; the confidence coefficient indicates that the box has the pre-prediction of the velocity spectrum energy blobMeasuring the score, if not, setting the confidence coefficient to be 0, if existing, the predicted confidence coefficient is determined by
Determining; pr (object) represents the probability of the energy blob existing for the box,representing the intersection ratio of the real block and the prediction block;
s303: predicting the probability of energy cliques existing in each grid by using a YOLO model, obtaining n x 5 x (5+1) data, and identifying a box containing the energy cliques in the seismic velocity spectrum;
s304: constructing a time-speed-time sequence prediction module based on an LSTM model, wherein a network architecture of the prediction module comprises a plurality of LSTM layers; and selecting the central coordinate of the box with the highest confidence coefficient from the boxes of the energy cliques identified in the step S303 as the input of the LSTM model, wherein the dimensions of the input layer and the output layer and the number of the neuron nodes of the hidden layer are all n x 2, wherein 2 represents the abscissa and the ordinate of the energy cliques on the seismic velocity spectrum, and thus the construction of the YOLO and LSTM hybrid neural network model is completed.
Further, in the YOLO model constructed in step S301, the size of the main convolution kernel in all convolution layers is designed to be 3 × 3 or 1 × 1; and a plurality of convolution structures are connected in parallel on different convolution layers in sequence.
Further, in step S103, the "time-velocity" pair data set is used as a training set, and when the hybrid neural network model of YOLO and LSTM is trained, a stochastic gradient descent algorithm is used for optimization, and a self-adaptive learning rate method is used to update the learning rate of the parameters of the hybrid neural network model of YOLO and LSTM; and when the loss value of the mixed neural network of the YOLO and the LSTM is not reduced any more and the precision is not improved any more or reaches the preset training times, obtaining and storing the finally trained mixed neural network model of the YOLO and the LSTM.
In step S104, the trained mixed neural network model of YOLO and LSTM is used to automatically pick up an unreceived seismic velocity spectrum, and the picking result is output as a sequence of "time-velocity" pairs, specifically: and inputting the non-picked seismic velocity spectrum image into a trained mixed neural network model of YOLO and LSTM, and detecting and identifying to obtain a time-velocity pair sequence.
The technical scheme provided by the invention has the beneficial effects that: the method realizes fully automatic speed picking, has better picking effect, and provides a new solution for the problem of automatic picking of the seismic velocity spectrum.
Drawings
FIG. 1 is a flow chart of the seismic velocity spectrum intelligent picking method based on YOLO and LSTM according to the invention;
FIG. 2 is a schematic diagram of a YOLO structure of a hybrid neural network model in an intelligent seismic velocity spectrum picking method based on YOLO and LSTM according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of an LSTM structure of a hybrid neural network model in an intelligent seismic velocity spectrum picking method based on YOLO and LSTM according to an embodiment of the present invention;
FIG. 4 is a diagram showing the identification result of the YOLO + LSTM hybrid neural network model of two actual seismic velocity spectra according to the seismic velocity spectrum intelligent picking method based on YOLO and LSTM of the embodiment of the present invention;
FIG. 5 is a seismic stack time section obtained by velocity spectrum picking of marine actual seismic data by a YOLO and LSTM-based seismic velocity spectrum intelligent picking method according to an embodiment of the present invention; fig. 5(a) is a stacking section calculated using a stacking velocity of manual picking; fig. 5(b) is a stacking section calculated using the stacking velocity of the YOLO + LSTM auto-pickup.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be further described with reference to the accompanying drawings.
Referring to fig. 1 and 2, an embodiment of the present invention provides an intelligent pickup method for seismic velocity spectrum based on YOLO and LSTM:
s101: acquiring a seismic velocity spectrum, and acquiring a time-velocity pair data set in the seismic velocity spectrum;
s102: constructing a hybrid neural network model based on YOLO and LSTM;
s103: taking the time-speed pair data set in the step S101 as a training set, training the mixed neural network model of the YOLO and the LSTM, and obtaining the trained mixed neural network model of the YOLO and the LSTM;
s104: and automatically picking up the seismic velocity spectrum which is not picked up by using the trained mixed neural network model of the YOLO and the LSTM, and outputting the picked-up result by taking a time-velocity pair as a sequence.
Further, step S101 is specifically as follows:
s201: extracting a plurality of common midpoint gathers from shot gather seismic data on the sea or on the land, and sequentially denoising and speed scanning the common midpoint gathers to obtain a corresponding seismic speed spectrum;
s202: time-velocity pairs are picked from the seismic velocity spectrum, resulting in a time-velocity pair dataset.
Further, step S102 specifically includes the following steps:
s301: constructing a YOLO model, wherein the adopted activation function is a LeakyReLU function, and the calculation formula is as follows:
s302: dividing the seismic velocity spectrum into n x n grids by using a YOLO model, and detecting 5 frames for each grid, wherein the number of the frames for detecting energy blobs in each input image is n x 5; n is specifically set according to the area proportion of the energy clusters in the input seismic velocity spectrum image, and when the proportion of the energy clusters is larger, n is smaller, and when the proportion of the energy clusters is smaller, n is larger; each box contains 5 information data; the 5 pieces of information data are x, y, w, h and confidence degrees respectively; where x and y are the coordinates of the center of the box, and w and h are the length and width of the box; the confidence coefficient represents the predicted score of the velocity spectrum energy blob of the box, if not, the confidence coefficient is set to 0, if existing, the confidence coefficient is set to 0Confidence of prediction is given by
Determining; pr (object) represents the probability of the energy blob existing for the box,representing the intersection ratio of the real block and the prediction block;
s303: predicting the probability of energy cliques existing in each grid by using a YOLO model, obtaining n x 5 x (5+1) data, and identifying a box containing the energy cliques in the seismic velocity spectrum;
s304: constructing a time-speed-time sequence prediction module based on an LSTM model, wherein a network architecture of the prediction module comprises a plurality of LSTM layers; and selecting the central coordinate of the box with the highest confidence coefficient from the boxes of the energy cliques identified in the step S303 as the input of the LSTM model, wherein the dimensions of the input layer and the output layer and the number of the neuron nodes of the hidden layer are all n x 2, wherein 2 represents the abscissa and the ordinate of the energy cliques on the seismic velocity spectrum, and thus the construction of the YOLO and LSTM hybrid neural network model is completed.
Further, in the YOLO model constructed in step S301, the size of the main convolution kernel in all convolution layers is designed to be 3 × 3 or 1 × 1; and a plurality of convolution structures are connected in parallel on different convolution layers in sequence.
Further, in step S103, the "time-velocity" pair data set is used as a training set, and when the hybrid neural network model of YOLO and LSTM is trained, a stochastic gradient descent algorithm is used for optimization, and a self-adaptive learning rate method is used to update the learning rate of the parameters of the hybrid neural network model of YOLO and LSTM; when the loss value of the mixed neural network of the YOLO and the LSTM is not reduced any more and the precision is not improved any more, or the preset training times are reached, obtaining and storing a finally trained mixed neural network model of the YOLO and the LSTM; when the accuracy requirement is not met, the procedure needs to return to step S102 to readjust the model or optimize the training strategy. The readjusting of the hybrid neural network model or the optimization training strategy of the YOLO and the LSTM specifically includes: the decay rate parameters of the initial values, the base learning rate, and the learning rate of the hybrid neural network model including the training set, YOLO, and LSTM are changed.
In step S104, the trained mixed neural network model of YOLO and LSTM is used to automatically pick up an unreceived seismic velocity spectrum, and the picking result is output as a sequence of "time-velocity" pairs, specifically: and inputting the non-picked seismic velocity spectrum image into a trained mixed neural network model of YOLO and LSTM, and detecting and identifying to obtain a time-velocity pair sequence.
After the time-velocity pair sequence is obtained in step S104, the difference is also performed on the picked results of a plurality of seismic velocity spectra of the same survey line to obtain a stacking velocity profile, and the results of the manual picked velocity spectrum and the automatic picked spectrum by the method are evaluated through the stacking velocity profile.
Example 1
In this embodiment, a marine two-dimensional seismic survey line is taken as an example, and the method provided by the present invention is adopted to automatically pick up the seismic velocity spectrum of the survey line. The specific process for automatically picking up the seismic velocity spectrum by using the invention is as follows:
step 1: 7000 common midpoint gather sets are extracted from the marine two-dimensional seismic survey line, then preprocessing, denoising and velocity scanning are carried out on the common midpoint gather sets, corresponding seismic velocity spectrums are obtained, stacking velocity is manually picked up, and accordingly a seismic velocity spectrum data set is obtained, wherein 80% of the seismic velocity spectrums serve as a training set, and 20% of the seismic velocity spectrum data set serves as a testing set.
Step 2: the network model suitable for the embodiment is designed by utilizing the construction method of the artificial neural network model provided by the invention. The network model of the present embodiment is based on a hybrid neural network model of YOLO and LSTM.
Fig. 2 shows a network architecture of the YOLO model, which includes 26 convolutional layers, 7 pooling layers and 2 fully-connected layers. In all convolutional layers, the size of the main convolutional kernel is designed to be 3 × 3 or 1 × 1. 240 × 240 × 64, 120 × 120 × 128, 60 × 60 × 256, 30 × 30 × 512, and 15 × 15 × 1024 feature matrices are sequentially arranged on different convolutional layers, and 60 × 60, 30 × 30, and 15 × 15 convolutional structures are sequentially connected in parallel on the different convolutional layers. The activation function adopted in the YOLO model is a leak ReLU activation function, and the calculation formula is as follows:
fig. 3 shows the network architecture of the LSTM model. The model contains 2 layers of LSTM layers, where the output of the first layer is the input of the second layer, and the part of the input layer, the dimensions of the output layer, and the number of neuron nodes of the hidden layer are all 15 × 15 × 2 ═ 450.
And step 3: the training sample set is input into the mixed neural network model designed in the example, and 100 color seismic velocity spectrum images are randomly input each time. And optimizing by adopting a random gradient descent algorithm, and updating the learning rates of all model parameters by utilizing a self-adaptive learning rate algorithm. Wherein the basic learning rate is 0.01, the step length is 0.001, the exponential decay rate of the learning rate is 0.9, the learning rate is adjusted once every 1000 times of training, the training is performed for 1 ten thousand times in total, and finally the trained model is stored.
And 4, step 4: the non-picked seismic velocity spectrum image is input into a trained hybrid neural network model of YOLO and LSTM, the model identifies the position in the energy mass and obtains a 'time-velocity' pair, thereby automatically acquiring the stacking velocity. FIG. 4 is the results of the identification of the YOLO + LSTM model for two actual velocity spectra, where the square and star points represent the manual pick and the YOLO + LSTM model identification, respectively. And interpolating the picked results of a plurality of seismic velocity spectrums of the same survey line to obtain a stacking velocity profile, and performing dynamic correction and horizontal stacking on the common-center gather by using the stacking velocity profile to finally form the stacking profile. Figure 5 shows a stack profile derived from the stack velocities automatically picked using the manual pick and the YOLO + LSTM model. Fig. 5a shows the stacking velocity (350 common midpoint gathers) manually picked every 20 common midpoints, and fig. 5b shows the stacking velocity automatically picked from 7000 common midpoints by using the YOLO + LSTM model.
As can be seen from the above embodiments, with the method of the present invention, the stacking velocity of the automatic picking is matched with the stacking velocity of the manual picking, and the stacking section calculated by using the stacking velocity of the automatic picking is very close to the reflection in-phase axis of the manual picking on the whole, so that the method can completely replace the manual picking of the velocity spectrum, and greatly reduce the labor intensity of the earthquake processor.
The technical scheme provided by the invention has the beneficial effects that: the method realizes fully automatic speed picking, has better picking effect, and provides a new solution for the problem of automatic picking of the seismic velocity spectrum.
In this document, the terms front, back, upper and lower are used to define the positions of the devices in the drawings and the positions of the devices relative to each other, and are used for the sake of clarity and convenience in technical solution. It is to be understood that the use of the directional terms should not be taken to limit the scope of the claims.
The features of the embodiments and embodiments described herein above may be combined with each other without conflict.
The above description is only for the purpose of illustrating the preferred embodiments of the present invention and is not to be construed as limiting the invention, and any modifications, equivalents, improvements and the like that fall within the spirit and principle of the present invention are intended to be included therein.
Claims (6)
1. An intelligent seismic velocity spectrum pickup method based on YOLO and LSTM is characterized in that: the method specifically comprises the following steps:
s101: acquiring a seismic velocity spectrum, and acquiring a time-velocity pair data set in the seismic velocity spectrum;
s102: constructing a hybrid neural network model based on YOLO and LSTM;
s103: taking the time-speed pair data set in the step S101 as a training set, training the mixed neural network model of the YOLO and the LSTM, and obtaining the trained mixed neural network model of the YOLO and the LSTM;
s104: and automatically picking up the seismic velocity spectrum which is not picked up by using the trained mixed neural network model of the YOLO and the LSTM, and outputting the picked-up result by taking a time-velocity pair as a sequence.
2. The method for intelligently picking up seismic velocity spectrum based on YOLO and LSTM as claimed in claim 1, wherein: step S101 is specifically as follows:
s201: extracting a plurality of common midpoint gathers from shot gather seismic data on the sea or on the land, and sequentially denoising and speed scanning the common midpoint gathers to obtain a corresponding seismic speed spectrum;
s202: time-velocity pairs are picked from the seismic velocity spectrum, resulting in a time-velocity pair dataset.
3. The method for intelligently picking up seismic velocity spectrum based on YOLO and LSTM as claimed in claim 1, wherein: step S102 is specifically as follows:
s301: constructing a YOLO model, wherein the adopted activation function is a LeakyReLU function, and the calculation formula is as follows:
s302: dividing the seismic velocity spectrum into n x n grids by using a YOLO model, and detecting 5 frames for each grid, wherein the number of the frames for detecting energy blobs in each input image is n x 5; n is preset according to the area proportion of the energy cliques in the input seismic velocity spectrum image; each box contains 5 information data; the 5 pieces of information data are x, y, w, h and confidence degrees respectively; where x and y are the coordinates of the center of the box, and w and h are the length and width of the box; the confidence coefficient represents the predicted score of the velocity spectrum energy mass of the box, if the box exists, the confidence coefficient is set to be 0, and if the box exists, the predicted confidence coefficient is represented by
Determining; pr (object) represents the probability of the energy blob existing for the box,
representing the intersection ratio of the real block and the prediction block;
s303: predicting the probability of energy cliques existing in each grid by using a YOLO model, obtaining n x 5 x (5+1) data, and identifying a box containing the energy cliques in the seismic velocity spectrum;
s304: constructing a time-speed-time sequence prediction module based on an LSTM model, wherein a network architecture of the prediction module comprises a plurality of LSTM layers; and selecting the central coordinate of the box with the highest confidence coefficient from the boxes of the energy cliques identified in the step S303 as the input of the LSTM model, wherein the dimensions of the input layer and the output layer and the number of the neuron nodes of the hidden layer are all n x 2, wherein 2 represents the abscissa and the ordinate of the energy cliques on the seismic velocity spectrum, and thus the construction of the YOLO and LSTM hybrid neural network model is completed.
4. The method for intelligently picking up seismic velocity spectrum based on YOLO and LSTM as claimed in claim 3, wherein: in the YOLO model constructed in step S301, the size of the main convolution kernel in all convolution layers is designed to be 3 × 3 or 1 × 1; and a plurality of convolution structures are connected in parallel on different convolution layers in sequence.
5. The method for intelligently picking up seismic velocity spectrum based on YOLO and LSTM as claimed in claim 1, wherein: in step S103, the time-speed pair data set is used as a training set, a random gradient descent algorithm is adopted for optimization when the mixed neural network model of the YOLO and the LSTM is trained, and the learning rate of the parameters of the mixed neural network model of the YOLO and the LSTM is updated by using a self-adaptive learning rate method; and when the loss value of the mixed neural network of the YOLO and the LSTM is not reduced any more and the precision is not improved any more or reaches the preset training times, obtaining and storing the finally trained mixed neural network model of the YOLO and the LSTM.
6. The method for intelligently picking up seismic velocity spectrum based on YOLO and LSTM as claimed in claim 1, wherein: in step S104, the trained mixed neural network model of YOLO and LSTM is used to automatically pick up an unreceived seismic velocity spectrum, and the picking result is output as a sequence of "time-velocity" pairs, specifically: and inputting the non-picked seismic velocity spectrum image into a trained mixed neural network model of YOLO and LSTM, and detecting and identifying to obtain a time-velocity pair sequence.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910973522.2A CN110826691A (en) | 2019-10-14 | 2019-10-14 | Intelligent seismic velocity spectrum pickup method based on YOLO and LSTM |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910973522.2A CN110826691A (en) | 2019-10-14 | 2019-10-14 | Intelligent seismic velocity spectrum pickup method based on YOLO and LSTM |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110826691A true CN110826691A (en) | 2020-02-21 |
Family
ID=69549307
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910973522.2A Pending CN110826691A (en) | 2019-10-14 | 2019-10-14 | Intelligent seismic velocity spectrum pickup method based on YOLO and LSTM |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110826691A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111444821A (en) * | 2020-03-24 | 2020-07-24 | 西北工业大学 | Automatic identification method for urban road signs |
| CN111624653A (en) * | 2020-04-20 | 2020-09-04 | 中国石油天然气股份有限公司 | Automatic seismic velocity picking method and device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971230A (en) * | 2017-05-10 | 2017-07-21 | 中国石油大学(北京) | First break pickup method and device based on deep learning |
| CN108596327A (en) * | 2018-03-27 | 2018-09-28 | 中国地质大学（武汉） | A kind of seismic velocity spectrum artificial intelligence pick-up method based on deep learning |
| CN108932713A (en) * | 2018-07-20 | 2018-12-04 | 成都指码科技有限公司 | A kind of weld porosity defect automatic testing method based on deep learning |
-
2019
- 2019-10-14 CN CN201910973522.2A patent/CN110826691A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106971230A (en) * | 2017-05-10 | 2017-07-21 | 中国石油大学(北京) | First break pickup method and device based on deep learning |
| CN108596327A (en) * | 2018-03-27 | 2018-09-28 | 中国地质大学（武汉） | A kind of seismic velocity spectrum artificial intelligence pick-up method based on deep learning |
| CN108932713A (en) * | 2018-07-20 | 2018-12-04 | 成都指码科技有限公司 | A kind of weld porosity defect automatic testing method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| HAO ZHANG ET AL.: "Automatic velocity picking based on deep learning", 《SEG INTERNATIONAL EXPOSITION AND 89TH ANNUAL MEETING》 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111444821A (en) * | 2020-03-24 | 2020-07-24 | 西北工业大学 | Automatic identification method for urban road signs |
| CN111444821B (en) * | 2020-03-24 | 2022-03-25 | 西北工业大学 | Automatic identification method for urban road signs |
| CN111624653A (en) * | 2020-04-20 | 2020-09-04 | 中国石油天然气股份有限公司 | Automatic seismic velocity picking method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111310862B (en) | Image enhancement-based deep neural network license plate positioning method in complex environment | |
| CN106683048B (en) | Image super-resolution method and device | |
| CN111199214B (en) | Residual network multispectral image ground object classification method | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN112101278A (en) | Hotel point cloud classification method based on k nearest neighbor feature extraction and deep learning | |
| CN109165660A (en) | A kind of obvious object detection method based on convolutional neural networks | |
| CN109271990A (en) | A kind of semantic segmentation method and device for RGB-D image | |
| CN112347970B (en) | Remote sensing image ground object identification method based on graph convolution neural network | |
| CN109655815B (en) | Sonar target detection method based on SSD | |
| CN110826428A (en) | Ship detection method in high-speed SAR image | |
| CN109902715A (en) | A kind of method for detecting infrared puniness target based on context converging network | |
| CN109753996B (en) | Hyperspectral image classification method based on three-dimensional lightweight depth network | |
| CN114387520A (en) | Precision detection method and system for intensive plums picked by robot | |
| Guirado et al. | Deep-learning convolutional neural networks for scattered shrub detection with google earth imagery | |
| CN111783555A (en) | Deep learning satellite data cloud detection algorithm using unified samples | |
| CN114842208A (en) | Power grid harmful bird species target detection method based on deep learning | |
| CN106529458A (en) | Deep neural network space spectrum classification method for high-spectral image | |
| CN109919246A (en) | Pedestrian's recognition methods again based on self-adaptive features cluster and multiple risks fusion | |
| CN110826691A (en) | Intelligent seismic velocity spectrum pickup method based on YOLO and LSTM | |
| CN106845343A (en) | A kind of remote sensing image offshore platform automatic testing method | |
| CN113487600A (en) | Characteristic enhancement scale self-adaptive sensing ship detection method | |
| CN116524183A (en) | Camouflage target detection method based on multitask adapter fine adjustment | |
| CN116168240A (en) | Arbitrary-direction dense ship target detection method based on attention enhancement | |
| CN113989612A (en) | Remote sensing image target detection method based on attention and generation countermeasure network | |
| CN112766381A (en) | Attribute-guided SAR image generation method under limited sample |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20200221 |