CN110210296A - A kind of microseism useful signal detection method of combination U-net network and DenseNet network - Google Patents
A kind of microseism useful signal detection method of combination U-net network and DenseNet network Download PDFInfo
- Publication number
- CN110210296A CN110210296A CN201910336622.4A CN201910336622A CN110210296A CN 110210296 A CN110210296 A CN 110210296A CN 201910336622 A CN201910336622 A CN 201910336622A CN 110210296 A CN110210296 A CN 110210296A
- Authority
- CN
- China
- Prior art keywords
- signal
- network
- net
- data collection
- net network
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
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. analysis, for interpretation, for correction
- G01V1/282—Application of seismic models, synthetic seismograms
-
- 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. analysis, for interpretation, for correction
- G01V1/30—Analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
- G06F18/24147—Distances to closest patterns, e.g. nearest neighbour classification
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Remote Sensing (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Environmental & Geological Engineering (AREA)
- Geology (AREA)
- General Life Sciences & Earth Sciences (AREA)
- Geophysics (AREA)
- Data Exchanges In Wide-Area Networks (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
Abstract
The present invention provides the microseism useful signal detection methods of a kind of combination U-net network and DenseNet network, comprising: (1) generates raw data set；(2) nominal data collection；(3) MSNet network is constructed；(4) network parameter is adjusted；(5) Onset point is demarcated after signal learning.The present invention also provides the microseism useful signal detection systems of a kind of combination U-net network and DenseNet network, it include: raw data set generation module, data set demarcating module, MSNet network struction module, network parameter adjusts module, Onset point demarcating module after study.Study and extraction the present invention is based on U-net network to CNN network in signal characteristic achieve the purpose that the study for advanced optimizing signal characteristic and extraction in conjunction with the DenseNet network.The present invention has the characteristics that deeper time feature extraction, the segmentation of finer signal in micro-seismic monitoring, can be widely applied to underground state monitoring field.
Description
Technical field
The present invention relates to the geophysical techniques of underground state monitoring, more particularly to a kind of microseismic.
Technical background
Microseism data useful signal energy is weaker, and noise is relatively low, effectively detects that signal is more difficult, traditional letter
Number detection technique include by Fast Fourier Transform (FFT) to signal carry out spectrum analysis, small echo, Qu Bo and shearing wave conversion into
The means such as row time-frequency convert retain useful signal to achieve the purpose that remove noise.If but traditional method directly apply to it is micro-
Seismic data can not often obtain satisfied effect, and this will directly affect the quality and precision of micro-seismic monitoring.
The extensive concern of people, main cause are gradually received in recent years based on the signal monitoring technology that deep learning is done
It is that it has the characteristics that parameter is more, capacity is big, so that its network possesses mass data powerful processing capacity.CNN net
The feature for the strong learning ability that network has is widely used in it in study and extraction of signal characteristic, but due to traditional
CNN network carries out will lead to important feature loss in part when feature extraction, and U-net network --- one kind is based on the FCN network architecture
New neural network, can more efficiently restore some material particular features that CNN loses in learning characteristic so that
Will not lossing signal time and space characteristics can not extract more profound spy but since its network number of plies is less
Sign.
Summary of the invention
The present invention provides the microseism useful signal detection method of a kind of combination U-net network and DenseNet network,
It more efficient, more can accurately extract the feature of microseism signal.
The embodiment of the present invention in a first aspect, provide a kind of combination U-net network and DenseNet network micro-ly
Useful signal detection method is shaken, this method comprises:
Step 1, the analog signal under Different Strata model is generated using finite difference forward modeling, is adopted with using wave detector is practical
The actual formation data collected collectively forms raw data set；
Step 2, first break pickup is carried out to the raw data set with algorithm, the signal waveform at constituency first arrival and non-first arrival
And it is demarcated respectively；
Step 3, two Denseblock are added in original U-net network；
Step 4, calibrated signal data collection is inputted in MSNet network and is learnt, to the signal exported after study
Make the calculating of Softmax function, calculated result and calibrated data set are made into cross entropy and calculate loss function, minimizes loss letter
Number is to adjust network parameter.
Step 5, the output signal after Softmax is calculated is adjusted by network parameter, the position where gained probability peak point
Set, as input signal learnt by MSNet after the Onset point demarcated.
In the second aspect of the embodiment of the present invention, provide a kind of combination U-net network and DenseNet network micro-ly
Useful signal detection system is shaken, which includes:
Raw data set generation module: the analog signal under Different Strata model for combining finite difference forward modeling to generate
The real data arrived with actual acquisition, collectively forms raw data set；
Data set demarcating module: for carrying out first break pickup to the raw data set, at constituency first arrival and non-first arrival
Signal waveform is simultaneously demarcated respectively；
MSNet network struction module: for adding two Denseblock in original U-net network；
Network parameter adjusts module: learning for inputting calibrated signal data collection in MSNet network, to
The signal exported after habit makees the calculating of Softmax function, and calculated result and calibrated data set are made cross entropy and calculate loss letter
Number minimizes loss function to adjust network parameter.
Onset point demarcating module after study: the output signal after Softmax is calculated is adjusted by network parameter, gained probability
The Onset point that position where peak point, as input signal are demarcated after being learnt by MSNet.
As can be seen from the above technical solutions, the embodiment of the present invention has the advantage that
The purpose of the present invention is overcoming the shortcomings of above-mentioned background technique, joined on the basis of original U-net network
Denseblock constructs the new network of entitled MSNet, to realize deeper feature extraction, to be suitable for signal
Carry out finer segmentation.
The microseism useful signal detection method of combination U-net network and DenseNet network of the invention, using in U-
Denseblock layers are added in net network, mainly there is following benefit:
1, this method can extract the profound feature of signal data distribution, to facilitate the accurate extraction realized to signal.
2, due to characteristic that U-net network itself has, it is possible to largely restore to lose in convolution process
Spatial information.
3, the number of plies having due to primary U-net network itself is less, can not sufficiently extract the spy of signal data distribution
Sign, so the block layer being added is very suitable for the further feature being distributed to signal data.
Detailed description of the invention
Fig. 1 is the totality for the microseism useful signal detection method that the present invention combines U-net network and DenseNet network
Techniqueflow schematic diagram.
Fig. 2 is that the present invention combines step 2 in the microseism useful signal detection method of U-net network and DenseNet network
The schematic diagram of data set calibration.
Fig. 3 is that the present invention combines step 4 in the microseism useful signal detection method of U-net network and DenseNet network
Construct the flow chart of MSNet network.
Fig. 4 is the network for the microseism useful signal detection method that the present invention combines U-net network and DenseNet network
The probability schematic diagram of parameter adjustment.
Specific embodiment
Combined with specific embodiments below, the present invention is furture elucidated, it should be understood that these embodiments are merely to illustrate the present invention
Rather than it limits the scope of the invention.
Fig. 1 is that the microseism of a kind of combination U-net network and DenseNet network that one embodiment of the invention provides is effective
The general technical flow diagram of signal detecting method, comprises the following processes:
Step 1, the analog signal under Different Strata model is generated using finite difference forward modeling, is adopted with using wave detector is practical
The actual formation data collected collectively forms raw data set.
Step 2, first break pickup is carried out to the raw data set with algorithm, the signal waveform at constituency first arrival and non-first arrival
And it is demarcated respectively.
The first break pickup is to be calculated using STA algorithm signal, chooses and picks up corresponding point work at peak of curve
For Onset point, one-hot label is made, first arrival site is set and is designated as 1, representing the probability of Onset point is 1.
Step 3, two Denseblock are added in original U-net network.
The U-net network its can more efficiently restore that CNN network loses in learning characteristic it is some important thin
Save feature so that will not lossing signal time and space characteristics can not sufficiently extract but since its network number of plies is less
The feature of signal data distribution.Two Denseblock modules of the addition, being can in order to solve one Denseblock of addition
Can feature extraction amount to signal it is smaller, be added three Denseblock again may over-fitting the problem of.It can make U-net net
The deeper extraction signal characteristic of network, to be suitable for carrying out signal finer segmentation.
Step 4, calibrated signal data collection is inputted in MSNet network and is learnt, to the signal exported after study
Make the calculating of Softmax function, calculated result and calibrated data set are made into cross entropy and calculate loss function, minimizes loss letter
Number is to adjust network parameter.
Optionally, the step 4 may particularly include following steps:
Step 4.1, which inputs signal data collection in MSNet network, to be learnt.
The signal exported after step 4.2 pair study makees the calculating of Softmax function, calculates output probability with Softmax function,
Two class probabilities of each point are obtained, 0 is classified as non-first arrival, and 1 is classified as first arrival, and the probability peak for only choosing first arrival classification is
Onset point.
Step 4.3 Softmax is calculated after output signal as the prediction label of function, by first break pickup production
For one-hot label as true tag, loss function is calculated as cross entropy in prediction label and true tag.
Step 4.4 is with Adam algorithmic minimizing loss function to adjust network parameter.
The Softmax function formula that the step 4.2 calculates probability distribution indicates are as follows:
Wherein, k=1~6001 indicates number of sampling points, qk(x) indicate that each point is the probability of first arrival.
It should be understood that the value of above-mentioned formula k is not limited to 1~6001, specifically depending on number of sampling points.
The formula of the step 4.3 adjustment network parameter loss function indicates are as follows:
Wherein, H (p, q) indicates that loss function, pi (x) indicate the true tag (mark of one-hot made by first break pickup
Label), qi (x) indicates prediction label (the predictive marker distribution of the model after training), by minimizing loss function to obtain most
Excellent network parameter, the true tag are that step 2 demarcates gained signal, and the prediction label is that step 4 is known by e-learning
Not Chu Lai first arrival position.
Step 5, the output signal after Softmax is calculated is adjusted by network parameter, the position where gained probability peak point
Set, as input signal learnt by MSNet after the Onset point demarcated.
It in practice, as the Onset point position of label is demarcated by traditional algorithm, this algorithm is more mature at present, and one
As Onset point can facilitate inspection, but this algorithm calibration speed is very slow by visually seeing substantially.In this hair
In bright embodiment, accelerated to demarcate speed with U-net neural network to realize that batch is demarcated.Since U-net neural network one is opened
The weight of beginning and biasing are set at random, and before not a large amount of iteration, the predicted value of propagated forward output often and is really marked
Label value difference is larger, and the method by calculating loss and gradient decline, U-net neural network can be with can be minimized loss
Those weights replace original weight to improve the output accuracy of network, thus achieve the purpose that network parameter adjusts.
Fig. 2 is that the microseism of a kind of combination U-net network and DenseNet network that one embodiment of the invention provides is effective
The structural schematic diagram of the building MSNet network of signal detecting method, the specific implementation process of specific introduction step 3 are as follows in detail:
Two Denseblock modules are added in original U-net network, the MSNet network comprises the steps of:
1) pass through U-net constricted path: input signal, signal data collection passes through the constricted path of U-net network, for obtaining
The contextual information for the number of winning the confidence；
2) pass through Denseblock1: signal data collection passes through the first Dense block mould added in U-net network
Block, the process do not interfere with the quantity of signal data collection feature, can only reinforce the utilization to feature, then pass through U-net net
The up-sampling structure that network had originally, can be realized the recovery of original signal feature vector, final output and input signal vector
The same vector, to realize end-to-end identification；
3) pass through convolution: signal data collection is slipped on signal by convolutional layer, convolution kernel, obtains the feature in different channels
Figure, the i.e. feature of learning signal；
4) pass through pond: signal data collection carries out down-sampling by pond layer, to signal, the dimension of feature is reduced, to subtract
Few unnecessary network parameter；
5) pass through Denseblock2: signal data collection passes through the second Dense block mould added in U-net network
Block, and described identical by Denseblock1 method, can be avoided a Denseblock may be to data set feature extraction amount
Deficiency, increase data set signal feature extraction amount；
6) pass through convolution: signal data collection is slipped on signal by convolutional layer, convolution kernel, obtains the feature in different channels
Figure, the i.e. feature of learning signal；
7) pass through pond: signal data collection carries out down-sampling by pond layer, to signal, the dimension of feature is reduced, to subtract
Few unnecessary network parameter；
8) carried out by the sequence of U-net path expander: signal data collection passes through the path expander of U-net network, logarithm
Precise positioning, output signal are carried out according to the part split required for concentration；
Fig. 3 is that the microseism of a kind of combination U-net network and DenseNet network that one embodiment of the invention provides is effective
The schematic diagram of the data set calibration of signal detecting method, is embodied as follows:
First break pickup is carried out to raw data set signal using STA algorithm, is chosen by output information and picks up peak of curve
Locate corresponding point as Onset point, make one-hot label, it is described first that first arrival site, which is set, and is designated as the vertical thick line of 1, Fig. 2
To point.
Fig. 4 is that the microseism of a kind of combination U-net network and DenseNet network that one embodiment of the invention provides is effective
The schematic diagram of the adjustment network parameter of signal detecting method, is embodied as follows:
Output signal after being calculated using Softmax is adjusted by network parameter, the position where gained probability peak point,
As input signal learnt by MSNet after the Onset point demarcated.Fig. 4 abscissa indicates that sampled point, ordinate indicate probability, figure
The sampled point that middle probability is 1 is the probability peak, this sampled point is Onset point.
The above, the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although referring to before
Stating embodiment, invention is explained in detail, those skilled in the art should understand that: it still can be to preceding
Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features；And these
It modifies or replaces, the spirit and scope for technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution.
Claims (9)
1. the microseism useful signal detection method of a kind of combination U-net network and DenseNet network, which is characterized in that packet
It includes:
Step 1, the analog signal under Different Strata model is generated using finite difference forward modeling, is arrived with using wave detector actual acquisition
Actual formation data collectively form raw data set；
Step 2, first break pickup is carried out to the raw data set with algorithm, the signal waveform at constituency first arrival and non-first arrival is simultaneously divided
It is not demarcated；
Step 3, two Denseblock are added in original U-net network；
Step 4, calibrated signal data collection is inputted in MSNet network and is learnt, the signal exported after study is made
Softmax function calculates, and calculated result and calibrated data set are made cross entropy and calculate loss function, minimizes loss function
To adjust network parameter；
Step 5, the output signal after Softmax is calculated is adjusted by network parameter, the position where gained probability peak point, i.e.,
The Onset point demarcated after being learnt for input signal by MSNet.
2. the microseism useful signal detection method of combination U-net network according to claim 1 and DenseNet network,
It is characterized in that, step 2 detailed process are as follows:
First break pickup is carried out to signal using STA algorithm, chooses and picks up corresponding at peak of curve put as Onset point, production
First arrival site is set and is designated as 1 by one-hot label.
3. the microseism useful signal detection method of combination U-net network according to claim 1 and DenseNet network,
Detailed process is as follows for the step 4:
Step 4.1: signal data collection is inputted in MSNet network and is learnt, the learning procedure specifically:
1) pass through U-net constricted path: input signal, signal data collection passes through the constricted path of U-net network, for obtaining letter
Number contextual information；
2) pass through Denseblock1: signal data collection passes through the first Dense block module added in U-net network, then
The up-sampling structure having originally by U-net network realizes that the recovery of original signal feature vector, final output and input are believed
The vector of number vector striking resemblances, realizes end-to-end identification；
3) pass through convolution: signal data collection is slipped on signal by convolutional layer, convolution kernel, obtains the characteristic pattern in each channel,
That is the feature of learning signal；
4) pass through pond: signal data collection carries out down-sampling by pond layer, to signal, the dimension of feature is reduced, to reduce net
Network parameter；
5) pass through Denseblock2: signal data collection passes through the second Dense block module added in U-net network, with
It is described identical by Denseblock1 method；
6) pass through convolution: signal data collection is slipped on signal by convolutional layer, convolution kernel, obtains the characteristic pattern in each channel,
That is the feature of learning signal；
7) pass through pond: signal data collection carries out down-sampling by pond layer, to signal, the dimension of feature is reduced, to reduce net
Network parameter；
8) carried out by the sequence of U-net path expander: signal data collection passes through the path expander of U-net network, to data set
In required for the part that splits carry out precise positioning, output signal；
Step 4.2: Softmax function being made to the signal exported after study and is calculated, calculates output probability with Softmax function, obtains
It is classified as non-first arrival to two class probabilities of each point, 0,1 is classified as first arrival, and the probability peak for choosing first arrival classification is first arrival
Point；
Step 4.3: the output signal after Softmax is calculated is as the prediction label of function, the one- that first break pickup is made
For hot label as true tag, loss function is calculated as cross entropy in prediction label and true tag；
Step 4.4: with Adam algorithmic minimizing loss function to adjust network parameter.
4. the microseism useful signal detection method of combination U-net network according to claim 3 and DenseNet network,
In the step 4.2 and step 4.3, the Softmax function formula for calculating probability distribution is indicated are as follows:
Wherein, k=1~6001 indicates number of sampling points, qk(x) indicate that each point is the probability of first arrival.
5. the microseism useful signal detection method of combination U-net network according to claim 3 and DenseNet network,
In the step 4.3 and step 4.4, the formula of adjustment network parameter loss function is indicated are as follows:
Wherein, H (p, q) indicates that loss function, pi (x) indicate that true tag, qi (x) indicate prediction label, damage by minimizing
It loses function and obtains optimal network parameter, the true tag is one-hot label made by step 2 first break pickup, described pre-
Mark label are the first arrival position that step 4 is identified by e-learning.
6. the microseism useful signal detection system of a kind of combination U-net network and DenseNet network, which is characterized in that packet
It includes:
Raw data set generation module: the analog signal and reality under Different Strata model for combining finite difference forward modeling to generate
The collected real data in border, collectively forms raw data set；
Data set demarcating module: for carrying out first break pickup to the raw data set, the signal at constituency first arrival and non-first arrival
Waveform is simultaneously demarcated respectively；
MSNet network struction module: for adding two Denseblock in original U-net network；
Network parameter adjusts module: learning for inputting calibrated signal data collection in MSNet network, after study
The signal of output makees the calculating of Softmax function, and calculated result and calibrated data set are made cross entropy and calculate loss function, most
Smallization loss function is to adjust network parameter；
Onset point demarcating module after study: the output signal after Softmax is calculated is adjusted by network parameter, gained probability peak
The Onset point that position where point, as input signal are demarcated after being learnt by MSNet.
7. the microseism useful signal detection system of U-net network and DenseNet network is combined as claimed in claim 6,
It is characterized in that, the data set demarcating module specifically includes:
First break pickup is carried out to signal using STA algorithm, chooses and picks up corresponding at peak of curve put as Onset point, production
First arrival site is set and is designated as 1 by one-hot label.
8. a kind of microseism useful signal detection system of combination U-net network and DenseNet network as claimed in claim 6
System, which is characterized in that the MSNet network struction module specifically includes:
U-net constricted path unit: it is used for input signal, signal data collection to pass through the constricted path of U-net network, to obtain letter
Number contextual information；
Denseblock1 unit: signal data collection passes through the first Dense block module added in U-net network, then leads to
The up-sampling structure that U-net network had originally is crossed, realizes the recovery of original signal feature vector, final output and input signal
The vector of vector striking resemblances, realizes end-to-end identification；
Convolution unit: for signal data collection by convolutional layer, convolution kernel slips on signal, obtains the feature in each channel
Figure, the i.e. feature of learning signal；
Pond unit: for signal data collection by pond layer, down-sampling is carried out to signal, reduces the dimension of feature, to reduce
Network parameter；
Denseblock2 unit: passing through the second Dense block module added in U-net network for signal data collection,
With it is described identical by Denseblock1 method；
Convolution unit: for signal data collection by convolutional layer, convolution kernel slips on signal, obtains the feature in each channel
Figure, the i.e. feature of learning signal；
Pond unit: for signal data collection by pond layer, down-sampling is carried out to signal, reduces the dimension of feature, to reduce
Network parameter；
U-net path expander unit: passing through the path expander of U-net network for signal data collection, required for concentrating to data
The part split carries out precise positioning, output signal.
9. the microseism useful signal detection system of U-net network and DenseNet network is combined as claimed in claim 6,
It is characterized in that, the network parameter adjustment module specifically includes:
Unit: signal data collection is inputted in MSNet network and is learnt；
Probability distribution computing unit: according to formulaSoftmax function is made to the signal exported after study
It calculates, wherein k=1~6001 indicates number of sampling points, qk(x) indicate that each point is the probability of first arrival, with Softmax function
Output probability is calculated, two class probabilities of each point are obtained, 0 is classified as non-first arrival, and 1 is classified as first arrival, chooses first arrival classification
Probability peak be Onset point；
Cross entropy computing unit: the output signal after Softmax is calculated makes first break pickup as the prediction label of function
One-hot label as true tag, according to formulaBy prediction label and
Loss function is calculated wherein as cross entropy in true tag, and H (p, q) indicates that loss function, pi (x) indicate true tag (just
The one-hot label made to pickup), qi (x) indicates prediction label；
Loss function computing unit: with Adam algorithmic minimizing loss function to adjust network parameter.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910336622.4A CN110210296B (en) | 2019-04-25 | 2019-04-25 | Microseism effective signal detection method combining U-net network and DenseNet network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910336622.4A CN110210296B (en) | 2019-04-25 | 2019-04-25 | Microseism effective signal detection method combining U-net network and DenseNet network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110210296A true CN110210296A (en) | 2019-09-06 |
| CN110210296B CN110210296B (en) | 2021-09-17 |
Family
ID=67786287
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910336622.4A Active CN110210296B (en) | 2019-04-25 | 2019-04-25 | Microseism effective signal detection method combining U-net network and DenseNet network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110210296B (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111505705A (en) * | 2020-01-19 | 2020-08-07 | 长江大学 | Microseism P wave first arrival pickup method and system based on capsule neural network |
| CN111505706A (en) * | 2020-04-28 | 2020-08-07 | 长江大学 | Microseism P wave first arrival pickup method and device based on depth T-Net network |
| CN111523258A (en) * | 2020-01-17 | 2020-08-11 | 长江大学 | MS-Net network-based microseism effective signal first arrival pickup method and system |
| US20210117737A1 (en) * | 2019-10-18 | 2021-04-22 | Korea University Research And Business Foundation | Earthquake event classification method using attention-based convolutional neural network, recording medium and device for performing the method |
| CN112711604A (en) * | 2019-10-25 | 2021-04-27 | 中国石油天然气股份有限公司 | Geophysical prospecting training data set construction method and device |
| CN113570041A (en) * | 2021-07-29 | 2021-10-29 | 中国地质大学（北京） | Neural network and method for suppressing seismic data noise of marine optical fiber streamer by using neural network |
| CN113960654A (en) * | 2020-07-20 | 2022-01-21 | 中国石油天然气股份有限公司 | Seismic data processing method and system |
| CN113960657A (en) * | 2020-07-20 | 2022-01-21 | 中国石油天然气股份有限公司 | Seismic data feature acquisition method and system |
| CN113570041B (en) * | 2021-07-29 | 2024-05-10 | 中国地质大学（北京） | Neural network and method for pressing seismic data noise of marine optical fiber towing cable by using same |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1074858A1 (en) * | 1999-07-28 | 2001-02-07 | Gaz De France (Service National) | Method of analysing recorded signals by automatically selecting from them at least one significant time event |
| CN103837891A (en) * | 2014-03-26 | 2014-06-04 | 宋维琪 | High-precision pickup method of first arrival of microseism |
| CN106896407A (en) * | 2017-03-28 | 2017-06-27 | 吉林大学 | A kind of microseism signal first break pickup method based on approximate negentropy |
| 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 |
-
2019
- 2019-04-25 CN CN201910336622.4A patent/CN110210296B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1074858A1 (en) * | 1999-07-28 | 2001-02-07 | Gaz De France (Service National) | Method of analysing recorded signals by automatically selecting from them at least one significant time event |
| CN103837891A (en) * | 2014-03-26 | 2014-06-04 | 宋维琪 | High-precision pickup method of first arrival of microseism |
| 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 |
| CN106896407A (en) * | 2017-03-28 | 2017-06-27 | 吉林大学 | A kind of microseism signal first break pickup method based on approximate negentropy |
Non-Patent Citations (4)
| Title |
|---|
| GAO HUANG，ET AL: "《Densely Connected Convolutional Networks》", 《ARXIV：1608.06993V5》 * |
| ZONGWEI ZHOU，ET AL: "《UNet++:A Nested U-Net Architecture for Medical Image Segmentation》", 《ARXIV：1807.10165V1》 * |
| 巩佳琦: "《基于Shearle变换与AIC的微地震初至提取》", 《中国优秀硕士学位论文全文数据库 基础科学辑》 * |
| 盛冠群，等: "《基于小波分解与高阶统计量的微地震初至拾取方法研究》", 《石油物探》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US11947061B2 (en) * | 2019-10-18 | 2024-04-02 | Korea University Research And Business Foundation | Earthquake event classification method using attention-based convolutional neural network, recording medium and device for performing the method |
| US20210117737A1 (en) * | 2019-10-18 | 2021-04-22 | Korea University Research And Business Foundation | Earthquake event classification method using attention-based convolutional neural network, recording medium and device for performing the method |
| CN112711604A (en) * | 2019-10-25 | 2021-04-27 | 中国石油天然气股份有限公司 | Geophysical prospecting training data set construction method and device |
| CN112711604B (en) * | 2019-10-25 | 2023-10-31 | 中国石油天然气股份有限公司 | Geophysical prospecting training data set construction method and device |
| CN111523258A (en) * | 2020-01-17 | 2020-08-11 | 长江大学 | MS-Net network-based microseism effective signal first arrival pickup method and system |
| CN111523258B (en) * | 2020-01-17 | 2023-06-13 | 长江大学 | Microseism effective signal first arrival pickup method and system based on MS-Net network |
| CN111505705B (en) * | 2020-01-19 | 2022-08-02 | 长江大学 | Microseism P wave first arrival pickup method and system based on capsule neural network |
| CN111505705A (en) * | 2020-01-19 | 2020-08-07 | 长江大学 | Microseism P wave first arrival pickup method and system based on capsule neural network |
| CN111505706B (en) * | 2020-04-28 | 2023-04-18 | 长江大学 | Microseism P wave first arrival pickup method and device based on deep T-Net network |
| CN111505706A (en) * | 2020-04-28 | 2020-08-07 | 长江大学 | Microseism P wave first arrival pickup method and device based on depth T-Net network |
| CN113960657A (en) * | 2020-07-20 | 2022-01-21 | 中国石油天然气股份有限公司 | Seismic data feature acquisition method and system |
| CN113960654A (en) * | 2020-07-20 | 2022-01-21 | 中国石油天然气股份有限公司 | Seismic data processing method and system |
| CN113570041A (en) * | 2021-07-29 | 2021-10-29 | 中国地质大学（北京） | Neural network and method for suppressing seismic data noise of marine optical fiber streamer by using neural network |
| CN113570041B (en) * | 2021-07-29 | 2024-05-10 | 中国地质大学（北京） | Neural network and method for pressing seismic data noise of marine optical fiber towing cable by using same |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110210296B (en) | 2021-09-17 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110210296A (en) | A kind of microseism useful signal detection method of combination U-net network and DenseNet network | |
| CN109993280B (en) | Underwater sound source positioning method based on deep learning | |
| CN108984745A (en) | A kind of neural network file classification method merging more knowledge mappings | |
| CN114092832B (en) | High-resolution remote sensing image classification method based on parallel hybrid convolutional network | |
| CN109597043A (en) | Radar Signal Recognition method based on quantum particle swarm convolutional neural networks | |
| CN109165585A (en) | A kind of improved ship target detection method based on YOLO V2 | |
| CN110490230A (en) | The Acoustic Object recognition methods of confrontation network is generated based on depth convolution | |
| CN109767776B (en) | Deception voice detection method based on dense neural network | |
| CN108696331A (en) | A kind of signal reconfiguring method based on generation confrontation network | |
| CN111080021B (en) | Sand body configuration CMM neural network prediction method based on geological information base | |
| CN111582236B (en) | LPI radar signal classification method based on dense convolutional neural network | |
| CN107870321B (en) | Radar one-dimensional range profile target identification method based on pseudo-label learning | |
| Kong et al. | Radar emitter identification based on deep convolutional neural network | |
| CN103077408B (en) | Method for converting seabed sonar image into acoustic substrate classification based on wavelet neutral network | |
| CN111931820A (en) | Water target radiation noise LOFAR spectrogram spectrum extraction method based on convolution residual error network | |
| CN112884134A (en) | Time domain-based convolutional neural network model for seismic facies identification and application | |
| CN115327616B (en) | Automatic positioning method for mine microseism focus driven by massive data | |
| CN113283409A (en) | Airplane detection method in aerial image based on EfficientDet and Transformer | |
| CN115598697A (en) | Thin-layer structure high-resolution seismic inversion method, device, medium and equipment | |
| CN113780346B (en) | Priori constraint classifier adjustment method, system and readable storage medium | |
| CN112818777B (en) | Remote sensing image target detection method based on dense connection and feature enhancement | |
| CN113111786B (en) | Underwater target identification method based on small sample training diagram convolutional network | |
| CN113869289B (en) | Multi-channel ship radiation noise feature extraction method based on entropy | |
| CN115964667A (en) | River-lake lithofacies well logging identification method based on deep learning and resampling | |
| CN113671564B (en) | NARX dynamic neural network-based microseism effective event automatic pickup 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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |