CN113705654A - FFPN model-based microseism first-arrival intelligent pickup method, system, equipment and storage medium - Google Patents
FFPN model-based microseism first-arrival intelligent pickup method, system, equipment and storage medium Download PDFInfo
- Publication number
- CN113705654A CN113705654A CN202110970928.2A CN202110970928A CN113705654A CN 113705654 A CN113705654 A CN 113705654A CN 202110970928 A CN202110970928 A CN 202110970928A CN 113705654 A CN113705654 A CN 113705654A
- Authority
- CN
- China
- Prior art keywords
- model
- arrival
- ffpn
- microseism
- picking
- 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
- 238000000034 method Methods 0.000 title claims abstract description 50
- 238000012549 training Methods 0.000 claims abstract description 52
- 230000006870 function Effects 0.000 claims description 19
- 238000004364 calculation method Methods 0.000 claims description 11
- 238000012360 testing method Methods 0.000 claims description 9
- 238000004590 computer program Methods 0.000 claims description 7
- 238000007781 pre-processing Methods 0.000 claims description 5
- 230000015654 memory Effects 0.000 claims description 4
- 230000000694 effects Effects 0.000 abstract description 11
- 230000004927 fusion Effects 0.000 abstract description 6
- 238000005457 optimization Methods 0.000 abstract description 5
- 230000008569 process Effects 0.000 abstract description 4
- 238000001514 detection method Methods 0.000 description 10
- 238000004422 calculation algorithm Methods 0.000 description 7
- 238000013136 deep learning model Methods 0.000 description 6
- 238000013527 convolutional neural network Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 238000013528 artificial neural network Methods 0.000 description 3
- 238000012545 processing Methods 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 3
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000003993 interaction Effects 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 238000013526 transfer learning Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000007796 conventional method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000002245 particle Substances 0.000 description 1
- 230000000644 propagated effect Effects 0.000 description 1
- 230000006403 short-term memory Effects 0.000 description 1
Images
Classifications
-
- 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
- G06F18/211—Selection of the most significant subset of features
-
- 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
- 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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
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 discloses a microseism first-arrival intelligent picking method, a microseism first-arrival intelligent picking system, microseism first-arrival intelligent picking equipment and a storage medium based on an FFPN model, belongs to the field of exploration geophysics, and the model provided by the invention keeps the multi-scale feature fusion capability of an original FPN network model and is suitable for the first-arrival intelligent picking problem and can achieve a more detailed first-arrival output result. The network model selects the output of two scale characteristics as the optimization condition of the model, calculates more detailed first arrival picking output through the low-level characteristics with rich detail characteristics, calculates two classification outputs through the high-level characteristics with larger receptive field, and performs back propagation in the training process to achieve the optimization of the network model by calculating the loss of the two output parts, thereby obtaining more accurate two classification results (first arrival picking results). Moreover, the point-aware loss function provided by the model can calculate the loss of the low-level feature output, and the effect of optimizing the high-level feature output is achieved by optimizing the low-level feature.
Description
Technical Field
The invention belongs to the field of exploration geophysics, and relates to an FFPN model-based microseism first-arrival intelligent pickup method, system, equipment and storage medium.
Background
First arrival pickup plays an important role in microseismic processing and imaging. With the increasing amount of received data, it becomes very time-consuming to manually pick up the first arrivals of the micro-earthquakes, and manual interpretation of the first arrivals of the micro-earthquakes relies on the experience of an interpreter, which is prone to human errors and interference. Over the past several decades, researchers have struggled to achieve automatic/semi-automatic microseismic first arrival pickup.
STA/LTA is a traditional microseism first arrival picking method which is realized by calculating the moving average energy ratio of a short time window and a long time window. The STA/LTA method is simple in calculation and simple and convenient to operate, and is widely applied to application such as microseism first-break explanation. However, the STA/LTA approach has two major drawbacks: first, due to the effect of the short window length, the first arriving waveform cannot always be received accurately; secondly, the microseism data often contains strong interference noise, which affects the accuracy and stability of the microseism first-arrival picking result. On the basis, various automatic/semi-automatic microseism first arrival picking methods are proposed, such as local linear embedding and improved particle swarm optimization clustering, adaptive multi-band selection, a kurtosis ratio method, a rapid AIC method and the like. The traditional microseism first arrival picking method has good performance on microseism data with strong peak amplitude and stable noise, and can accurately explain the microseism first arrival. However, in most practical cases, noise and effective wave energy are lost very much, and these conventional methods often exhibit instability, i.e. the inability to accurately pick up microseismic first arrivals.
Deep learning is a branch of machine learning, and has attracted attention in various fields as computing power has developed. Convolutional Neural Networks (CNN) proposed by LeCun et al (1989) is one of the most popular and widely used deep learning algorithms. The typical CNN model is initially used for image classification tasks, i.e. converting image structure into some kind of feature information, the classification task is concerned with overall features, and finally content description of the whole picture is given. The last layer of the classification network is generally a fully connected layer for predicting target labels, and the classical CNN network models provide a research basis for other tasks. In addition, extending to the field of target detection, a technical target needs to pay attention to a specific object target, and requires to obtain category information and position information of the target at the same time, and the output of the type detection network is a list, that is, each item includes the category and position of a detection frame. Some typical networks are Regions with CNN features (RCNN), You Only Look Once (YOLO), Feature Pyramid Network (FPN), etc. With the development of research and the improvement of demand, classification tasks are gradually refined to semantic segmentation tasks in the field of image processing, which are units of analysis by using pixels, and to solve the problem, network models such as full volumetric networks (FCN), SegNet, deep, and the like are proposed successively.
In recent years, deep learning has achieved good results in many fields, and with the continuous exploration and practice of researchers, many deep learning-based algorithms and models are used for microseismic data seismic processing and interpretation, such as Long Short-Term Memory network (LSTM). With the development of these deep Learning models, many methods for microseismic first arrival picking have appeared, such as Atomic Energy Network (AEnet), Encoder-decoder Network (Encoder-decoder Network), Transfer Learning (Transfer Learning), and Human-Computer Interaction Learning (Human-Computer Interaction Learning). The application prospect of the deep learning models in the micro earthquake first arrival picking application is shown by the deep learning models and the successful application of the deep learning models in the micro earthquake first arrival picking. However, although these deep learning models show good results in microseismic first arrival pickup applications, there is still a need for further improvement and promotion in generalization and computational efficiency.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention aims to provide a microseism first-arrival intelligent pickup method, a system, equipment and a storage medium based on an FFPN model, and aims to solve the technical problems of low microseism first-arrival pickup generalization and low calculation efficiency in the prior art.
The invention provides a microseism first arrival intelligent pickup method of an FFPN model, which comprises the following steps:
s1, preprocessing the actual micro-seismic data set, establishing a sample data set and a label data set, randomly selecting sample data in a certain proportion and label data corresponding to the sample data as a training set, and selecting the rest sample data and the label data corresponding to the sample data as a test set;
s2, reserving a basic structure of the original FPN model, replacing the two-dimensional convolution with one-dimensional convolution, building an initial one-dimensional feature pyramid first arrival picking network model FPN, and training the one-dimensional feature pyramid first arrival picking network model FPN to obtain a one-dimensional feature pyramid first arrival picking network training model;
s3, performing microseism first arrival prediction according to the one-dimensional characteristic pyramid initial quality picking network training model obtained in S2, building an improved characteristic pyramid model FFPN with double output double losses, introducing a specific loss function into the improved pyramid model FFPN, training the improved pyramid model FFPN by adopting the training set obtained in S1, and obtaining actual microseism first arrival intelligent picking.
Preferably, in S3, the FFPN network models all use one-dimensional convolution, the second layer output of the FPN network model is cancelled, the first layer output is modified to calculate the first arrival pick-up output, and the binary output is calculated by the high layer feature with the receptive field.
Preferably, in S3, the training data randomly selected in S1 is input into the improved pyramid model FFPN, and the improved pyramid model FFPN is supervised by using the two classification results corresponding to the first arrival picking tags, so as to obtain an improved pyramid training model through training.
Preferably, the two-class output loss calculation is as shown in equation (1):
lcla＝CrossEntropy(mpred,mg) (1)
where Cross Encopy is the cross entropy loss, mpredIs the classification output of the network, mgIs the true value of the classification.
Preferably, a specific loss function is added, namely point-aware loss, and the loss function is used for calculating the output error of the microseismic first arrival picked point.
Preferably, the calculation method of the loss function point-aware loss is as shown in formula (2):
lp＝CrossEntropy(ppred,pg) (2)
wherein Cross Encopy is the cross entropy loss, ppredIs the first arrival pick-up output of the network, pgIs a first arrival pick-up truth.
Preferably, the method for calculating the total loss function of the modified feature pyramid model FFPN is as shown in formula (3):
l＝lcla+λlp (3)
where λ is used as a hyper-parameter to equalize the gradients of the two loss functions.
The invention also discloses a first arrival picking system of the intelligent first arrival picking method for the micro earthquake based on the FFPN model, which comprises the following steps:
the data acquisition module is used for preprocessing actual micro-seismic data to acquire a sample data set and a tag data set, randomly selecting a certain proportion of actual micro-seismic data and corresponding tag data as a training set, and using the rest as a test set;
the model acquisition module is used for building and training a one-dimensional characteristic pyramid initial quality picking network model and acquiring a one-dimensional characteristic pyramid initial quality picking network training model;
and the first arrival picking module is used for carrying out microseism first arrival prediction on the one-dimensional characteristic pyramid first quality picking network training model, building an improved characteristic pyramid model and training the improved characteristic pyramid model, and acquiring the first arrival intelligent picking of actual microseism data.
The invention provides computer equipment which comprises a memory and a processor, wherein the memory stores a computer program, and the processor realizes the steps of the FFPN model-based microseism first-arrival intelligent pickup method when executing the computer program.
The invention proposes a computer-readable storage medium, in which a computer program is stored which, when being executed by a processor, implements the steps of a microseismic first arrival intelligent pickup method based on an FFPN model.
Compared with the prior art, the invention has the following beneficial effects:
the invention provides an FFPN model-based microseism first-arrival intelligent picking method, which combines the actual microseism data characteristics to adjust and improve the structure of an initial FPN network model and provides an improved characteristic pyramid (FFPN) model-based microseism first-arrival intelligent picking method. The model reserves the multi-scale feature fusion capability of the original FPN network model, is suitable for the problem of first-arrival intelligent pickup, and achieves a more detailed first-arrival output result. The core advantage of the original one-dimensional FPN network lies in the multi-scale feature fusion capability, and the model can fully utilize the rich detail information of the low-level features and the rich semantic information of the high-level network, so that the multi-scale feature extraction and learning are carried out. The FPN model also directly calculates model loss using the output of multiple scale features, thereby optimizing the network model for more accurate feature extraction and learning. The invention provides a first-arrival intelligent picking method based on an improved characteristic pyramid model with double output and double loss, which is obtained by combining actual micro-seismic data characteristics and first-arrival characteristics to deform on the basis of an original one-dimensional FPN network. The FFPN model for the intelligent pickup of the first arrival with double output and double loss is finally obtained, and the FFPN model trained by the method can be used for realizing accurate and efficient intelligent pickup of the first arrival of the microseism.
Furthermore, the problem that classification fuzzy is shown at the first arrival boundary when an original FPN network model is applied to the first arrival picking problem is solved, the FFPN network model provided by the invention selects the outputs of two scale features as the optimization conditions (o1, o3) of the model, the more refined first arrival picking output (o1) is calculated only through the low-level features with rich detail features, the binary output (o3) is calculated through the high-level features with larger receptive fields, and the two parts of output loss are calculated and are propagated reversely in the training process to achieve the optimization of the network model, so that more accurate binary results (first arrival picking results) are obtained.
Further, the invention provides a point-aware loss function, which only calculates the output error of the first arrival picked point, specifically calculates the loss (o1) of the output of the low-level feature, and optimizes the output of the high-level feature by optimizing the low-level feature.
The invention also provides a first-arrival picking system of the intelligent first-arrival picking method for the micro earthquake based on the FFPN model, which realizes first-arrival picking by adopting a modular idea, so that the modules are independent from each other, and the unified management of the modules is convenient to realize.
Drawings
FIG. 1 is actual microseismic data and its labels used in the present invention;
FIG. 2 shows training data and labels used in the present invention (the red portion is the training set);
FIG. 3 is a schematic diagram of a raw FPN deep neural network structure;
FIG. 4 is a schematic diagram of the FFPN deep neural network structure of the present invention;
FIG. 5 is a first-arrival result predicted from actual microseismic data by using the model (red dots), the encoding-decoding model (blue dots) and the STA/LTA method (green dots) according to the present invention;
FIG. 6 is a first-arrival result predicted by the present invention using FFPN (red dots), code-decode model (blue dots) and STA/LTA method (green dots) for microseismic data with a SNR of 0.50;
FIG. 7 shows the first-arrival results of the microseismic data prediction of the present invention using FFPN (red dots), code-de-code model (blue dots) and STA/LTA method (green dots) with a SNR of 0.30;
FIG. 8 is a first-arrival result predicted by the present invention using FFPN (Red dot), code-decode model (blue dot) and STA/LTA method (Green dot) for microseismic data with a SNR of 0.15;
FIG. 9 shows the first-arrival results of the present invention predicted from actual microseismic data of other work areas using the trained FFPN model.
Detailed Description
In order to make the technical solutions of the present invention better understood, the technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and in the drawings described above are used for distinguishing between similar elements and not necessarily for describing a particular sequential or chronological order. It is to be understood that the data so used is interchangeable under appropriate circumstances such that the embodiments of the invention described herein are capable of operation in sequences other than those illustrated or described herein. Furthermore, the terms "comprises," "comprising," and "having," and any variations thereof, are intended to cover a non-exclusive inclusion, such that a process, method, system, article, or apparatus that comprises a list of steps or elements is not necessarily limited to those steps or elements expressly listed, but may include other steps or elements not expressly listed or inherent to such process, method, article, or apparatus.
The invention is described in further detail below with reference to the accompanying drawings:
the invention provides a microseism first arrival intelligent pickup method based on an FFPN model, and the method is characterized in that the method refers to FIG. 1 and is used for actual microseism data and a first arrival pickup label thereof; randomly selecting 80 data from the left rectangular frame as a training set, and using the rest as a test set; referring to fig. 2, the diagram of microseism data in a left rectangular frame in fig. 1 is shown, wherein a red part is randomly selected partial data and is used as a training set used by a training model, and the microseism first arrival intelligent pickup method comprises the following steps:
s1, performing artificial first arrival interpretation according to the actual microseism data set, establishing a sample data set and a label data set for the intelligent pickup of the subsequent microseism first arrival, randomly selecting part of actual microseism data and label data corresponding to the actual microseism data as a training set, and using the rest of the sample data and the label data corresponding to the sample data as a test set;
s2, reserving a basic structure of the original FPN model, replacing the two-dimensional convolution with one-dimensional convolution, building an initial one-dimensional Feature Pyramid (FPN) first arrival picking Network model, and training the one-dimensional Feature Pyramid model FPN to obtain a one-dimensional Feature Pyramid training model;
s3, optimizing and adjusting an initial one-dimensional Feature Pyramid Network model FPN, building an improved Feature Pyramid model (FFPN) with double output and double loss, performing improved Feature Pyramid model training again by utilizing part of randomly selected actual micro-seismic data sets and label data sets, and performing first-arrival intelligent picking of actual micro-seismic data by utilizing the trained FFPN model.
Referring to fig. 4, which is a schematic diagram of the FFPN deep neural network structure of the present invention, wherein the left image is a seismic shot gather, each time the network inputs a piece of data schematically shown in a rectangular frame, and outputs one column of the two classified output images; the rightmost image represents the auxiliary output of the network, the bar in the figure represents the characteristic graph of the output of each layer of the network, the arrow represents the convolution operation, and the detailed operation steps from S1 to S3 are as follows:
s1, establishing a sample data set and a label data set for the microseism first-arrival intelligent pickup model, randomly selecting actual microseism data and the corresponding label data set in a certain proportion as a training set, and using the residual sample data and the label data corresponding to the sample data as a test set; wherein, the proportion of the training set to the testing set is 1: 5.4.
the method comprises the steps of utilizing a supervised deep learning model to carry out microseism first-arrival intelligent picking to obtain a sample data set and a label data set, carrying out artificial first-arrival interpretation according to an actual microseism data set, and establishing the sample data set and the label data set for a subsequent microseism first-arrival intelligent picking model.
S2, reserving a basic structure of an original FPN model, replacing two-dimensional convolution with one-dimensional convolution, building an initial one-dimensional Feature Pyramid first arrival picking Network model (FPN), training the initial one-dimensional Feature Pyramid first arrival picking Network model FPN, and analyzing a Network structure and an actual microseism data first arrival picking effect;
the original FPN network proposed the original goal of solving the conventional target detection problem, and the model has a network structure of multi-scale fusion and multi-scale outputs (multi-scale and multi-level outputs). Due to the fact that target scales in a target detection task are different greatly, a conventional network model often has a poor effect on detection of small objects; the high-level features have a large receptive field, so that more semantic information can be captured, but the original image area corresponding to the unit pixel is large, so that the small object is not accurately mapped in the high-level features. The microseism first arrival has very fine structural characteristics and is difficult to accurately position in high-level features, so that an initial one-dimensional FPN first arrival picking model is built on the basis of the multi-scale feature fusion of an original FPN network.
Through training and initializing the FPN network model and analyzing the network structure characteristics and the first arrival picking effect in detail, the classification effect of the network model near the first arrival is found to be poor. The reason for the poor performance of this network model is as follows: the original FPN model proposes the purpose of solving the problem of target detection of a conventional image, wherein the target detection involves feature analysis of each region of the whole image, so that the feature extraction is not targeted, and therefore, the features near the first arrival extracted by the model have no uniqueness and pertinence. The first arrival expression is very fine in the microseism data, so that a network model is forced to enhance the extraction and learning of data characteristics near the first arrival so as to achieve a better classification effect (namely a first arrival picking effect).
For an original FPN network model, although the network model uses multi-scale feature fusion and fully utilizes detail information of low-level features, data features around first arrivals are not extracted in a targeted mode, and therefore the first arrival features learned by the network model are not enough to better optimize classification results near the first arrivals.
S3, optimizing and adjusting an initial one-dimensional Feature Pyramid first arrival picking Network model FPN, building an improved Feature Pyramid model (FFPN) with double output and double loss, introducing a specific loss function to the FFPN of the improved Pyramid model, performing model training again by using part of randomly selected microseismic data and label data, and performing actual microseismic data first arrival intelligent picking by using the trained FFPN model.
The invention provides an improved FPN network structure by re-adjusting the topological structure of an initial FPN network according to actual micro-seismic data characteristics, first arrival data characteristics and initial FPN network characteristics.
The core of the method is to solve the problem that the original FPN model has no pertinence in feature extraction, and can not accurately extract the data features near the first arrival of the microseism, so that the classification effect near the first arrival is poor. Firstly, the FFPN network model provided by the invention cancels the second layer output (o2) of the original FPN network, modifies the first layer output into a first arrival picking output (o1) with more detailed calculation, and calculates the binary output (o3) through the high-level characteristics with larger receptive field; inputting training data randomly selected by S1 into the improved pyramid model FFPN, and supervising the improved pyramid model FFPN by using the two classification results corresponding to the first arrival picking labels to train to obtain an improved pyramid training model; and the output loss calculation for the second class is shown in equation (1):
lcla＝CrossEntropy(mpred,mg) (1)
where Cross Encopy is the cross entropy loss, mpredIs the classification output of the network, mgIs the true value of the classification.
Secondly, a point-aware loss function is proposed, the loss function only calculates the output error of the first arrival picking point, specifically, the loss of the low-level feature output is calculated (o1), and the effect of optimizing the high-level feature output is achieved by optimizing the low-level feature, and the calculation method of the loss function is shown as formula (2):
lp＝CrossEntropy(ppred,pg) (2)
wherein Cross Encopy is the cross entropy loss, ppredIs the first arrival pick-up output of the network, pgIs a first arrival pick-up truth.
The method for calculating the total loss function of the FFPN of the improved characteristic pyramid model is shown as a formula (3):
l＝lcla+λlp (3)
where λ is used as a hyper-parameter to equalize the gradients of the two loss functions.
And then, training the network provided by the invention by using the actual micro-seismic data and the label data to obtain an FFPN model.
The invention also discloses a first arrival picking system of the intelligent first arrival picking method for the micro earthquake based on the FFPN model, which comprises the following steps:
the data acquisition module is used for preprocessing actual micro-seismic data to acquire a sample data set and a tag data set, randomly selecting a certain proportion of actual micro-seismic data and corresponding tag data as a training set, and using the rest as a test set;
the model acquisition module is used for building and training a one-dimensional characteristic pyramid initial quality picking network model and acquiring a one-dimensional characteristic pyramid initial quality picking network training model;
and the first arrival picking module is used for carrying out microseism first arrival prediction on the one-dimensional characteristic pyramid first quality picking network training model, building an improved characteristic pyramid model and training the improved characteristic pyramid model, and acquiring the first arrival intelligent picking of actual microseism data.
Referring to fig. 3, which is a top-down network corresponding to a characteristic pyramid of a characteristic graph pyramid network, if there are targets with different sizes in the original image in fig. 3, but different targets have different characteristics, simple targets can be distinguished by using the characteristics of a shallow layer; complex objects can be distinguished by using deep features. From top to bottom in fig. 3, the layer 1 outputs the example segmentation result of the larger target, the layer 2 outputs the example detection result of the next larger target, and the layer 3 outputs the example segmentation result of the smaller target. The same is true for detection, with simple targets output at layer 1, more complex targets output at layer 2, and complex targets output at layer 3. The enlarged area in fig. 3 is the lateral connection and the use of 1 x 1 convolution kernels has the main effect of reducing the number of convolution kernels.
And applying the trained FFPN Network model to the actual microseismic data for first arrival intelligent pickup, and comparing with an Encoder-decoder Network (Encoder-decoder Network) model and an STA/LTA algorithm. From the initial interpretation results of the original microseismic data in fig. 5, it can be found that the model (red dots) provided by the invention can obtain more accurate microseismic initial interpretation results, and the network model of the encoder and decoder (blue dots) and the STA/LTA algorithm (green dots) cannot obtain accurate initial interpretation results at certain positions, as shown by yellow ellipses. In order to further verify the stability of the model provided by the invention, random noise with a signal-to-noise ratio of 0.50 is added into the original microseism data, and the results in fig. 6 show that the model provided by the invention (red dots) can still accurately explain microseism first arrivals under the noise-containing condition, and the network model of a coder decoder (blue dots) and the STA/LTA algorithm (green dots) have poor stability, so that an accurate first arrival explanation result cannot be obtained, as shown by a yellow ellipse; random noise with the signal-to-noise ratio of 0.30 is added into the original microseism data, and the result in fig. 7 shows that the model (red dots) provided by the invention can still accurately explain microseism first arrivals under the noise-containing condition, the network model (blue dots) of a coder-decoder and the STA/LTA algorithm (green dots) have poor stability, and accurate first arrival explanation results cannot be obtained, as shown by a yellow ellipse; random noise with a signal-to-noise ratio of 0.15 is added to the original microseism data, and the result in fig. 8 shows that the model (red dots) provided by the invention can still accurately explain microseism first arrivals under the noise-containing condition, and the network model (blue dots) of the coder-decoder and the STA/LTA algorithm (green dots) are poor in stability, so that an accurate first arrival explanation result cannot be obtained, as shown by a yellow ellipse. In order to further verify the generalization of the model provided by the invention, the model is used for carrying out first-arrival intelligent explanation on the microseism data of other work areas, and the result of the figure 9 shows that the model can still accurately explain the first-arrival of the microseism of different work areas, thereby verifying the generalization of the model.
The above-mentioned contents are only for illustrating the technical idea of the present invention, and the protection scope of the present invention is not limited thereby, and any modification made on the basis of the technical idea of the present invention falls within the protection scope of the claims of the present invention.
Claims (10)
1. A microseism first arrival intelligent pickup method based on an FFPN model is characterized by comprising the following steps:
s1, preprocessing actual micro-seismic data, establishing a sample data set and a label data set, randomly selecting sample data in a certain proportion and label data corresponding to the sample data from the actual micro-seismic data as a training set, and taking the rest sample data and the label data corresponding to the sample data as a test set;
s2, reserving a basic structure of the original FPN model, replacing the two-dimensional convolution with one-dimensional convolution, building a one-dimensional feature pyramid initial quality picking network model FPN, and training the one-dimensional feature pyramid initial quality picking network model FPN to obtain a one-dimensional feature pyramid initial quality picking network training model;
s3, performing microseism first arrival prediction according to the one-dimensional characteristic pyramid initial quality picking network training model obtained in S2, building an improved characteristic pyramid model FFPN with double output double losses, introducing a specific loss function into the improved pyramid model FFPN, training the improved pyramid model FFPN by adopting the training set obtained in S1, and obtaining actual microseism first arrival intelligent picking.
2. The FFPN model-based microseismic first arrival intelligent pickup method as claimed in claim 1, wherein in S3, the FFPN network model uses all one-dimensional convolution, the second layer output of the FPN network model is cancelled, the first layer output is modified to be the calculation first arrival pickup output, and the binary output is calculated through the high layer feature with receptive field.
3. The FFPN model-based microseism first arrival intelligent picking method as claimed in claim 2, wherein in S3, the training data randomly selected in S1 is input into an improved pyramid model FFPN, and the improved pyramid model FFPN is supervised by using the binary classification result corresponding to the first arrival picking tag, so as to obtain an improved pyramid training model through training.
4. The FFPN model-based microseism first-arrival intelligent pickup method according to claim 2, wherein the two-classification output loss calculation is shown as formula (1):
lcla＝CrossEntropy(mpred,mg) (1)
where Cross Encopy is the cross entropy loss, mpredIs the classification output of the network, mgIs the true value of the classification.
5. The FFPN model-based microseism first arrival intelligent pickup method according to claim 4, wherein a specific loss function is added, and the loss function is point-aware loss and is used for calculating the output error of a microseism first arrival pickup point.
6. The FFPN model-based microseism first arrival intelligent pickup method as claimed in claim 5, wherein the calculation method of the loss function point-aware loss is shown as formula (2):
lp＝CrossEntropy(ppred,pg) (2)
whereinCross Entrophy is the cross entropy loss, ppredIs the first arrival pick-up output of the network, pgIs a first arrival pick-up truth.
7. The FFPN model-based microseism first-arrival intelligent pickup method as claimed in claim 6, wherein the overall loss function calculation method of the FFPN of the improved feature pyramid model is shown as formula (3):
l＝lcla+λlp (3)
where λ is used as a hyper-parameter to equalize the gradients of the two loss functions.
8. The system for realizing the FFPN model-based microseism first-arrival intelligent pickup method comprises the following steps:
the data acquisition module is used for preprocessing actual micro-seismic data to acquire a sample data set and a tag data set, randomly selecting a certain proportion of actual micro-seismic data and corresponding tag data as a training set, and using the rest as a test set;
the model acquisition module is used for building and training a one-dimensional characteristic pyramid initial quality picking network model and acquiring a one-dimensional characteristic pyramid initial quality picking network training model;
and the first arrival picking module is used for carrying out microseism first arrival prediction on the one-dimensional characteristic pyramid first quality picking network training model, building and training an improved characteristic pyramid model FFPN, and acquiring the first arrival intelligent picking of actual microseism data.
9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor when executing the computer program implements the steps of the FFPN model-based microseismic first-arrival intelligent picking method of any of claims 1 to 7.
10. A computer readable storage medium storing a computer program, wherein the computer program when executed by a processor implements the steps of the FFPN model-based microseismic first arrival intelligent picking method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110970928.2A CN113705654B (en) | 2021-08-23 | 2021-08-23 | FFPN model-based micro-seismic first-arrival intelligent pickup method, system, equipment and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110970928.2A CN113705654B (en) | 2021-08-23 | 2021-08-23 | FFPN model-based micro-seismic first-arrival intelligent pickup method, system, equipment and storage medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113705654A true CN113705654A (en) | 2021-11-26 |
| CN113705654B CN113705654B (en) | 2024-04-05 |
Family
ID=78654276
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110970928.2A Active CN113705654B (en) | 2021-08-23 | 2021-08-23 | FFPN model-based micro-seismic first-arrival intelligent pickup method, system, equipment and storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113705654B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027452A (en) * | 2019-12-05 | 2020-04-17 | 中国矿业大学(北京) | Microseismic signal arrival time and seismic phase identification method and system based on deep neural network |
| CN111695413A (en) * | 2020-04-28 | 2020-09-22 | 长江大学 | Signal first arrival pickup method and device combining U-Net and Temporal encoding |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
-
2021
- 2021-08-23 CN CN202110970928.2A patent/CN113705654B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111027452A (en) * | 2019-12-05 | 2020-04-17 | 中国矿业大学(北京) | Microseismic signal arrival time and seismic phase identification method and system based on deep neural network |
| WO2021139069A1 (en) * | 2020-01-09 | 2021-07-15 | 南京信息工程大学 | General target detection method for adaptive attention guidance mechanism |
| CN111695413A (en) * | 2020-04-28 | 2020-09-22 | 长江大学 | Signal first arrival pickup method and device combining U-Net and Temporal encoding |
Non-Patent Citations (2)
| Title |
|---|
| 周创;居兴国;李子昂;刘小民;: "基于深度卷积生成对抗网络的地震初至拾取", 石油物探, no. 05 * |
| 郑晶;曹子原;姜天琪;沈帅帅;: "基于深度信念神经网络的微震波到时拾取方法", 矿业科学学报, no. 06 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113705654B (en) | 2024-04-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108985268B (en) | Inductive radar high-resolution range profile identification method based on deep migration learning | |
| CN109117793B (en) | Direct-push type radar high-resolution range profile identification method based on deep migration learning | |
| CN109635744A (en) | A kind of method for detecting lane lines based on depth segmentation network | |
| CN107292298A (en) | Ox face recognition method based on convolutional neural networks and sorter model | |
| CN111695457B (en) | Human body posture estimation method based on weak supervision mechanism | |
| Wang et al. | Distilling knowledge from an ensemble of convolutional neural networks for seismic fault detection | |
| CN111598167B (en) | Small sample image identification method and system based on graph learning | |
| CN109635726B (en) | Landslide identification method based on combination of symmetric deep network and multi-scale pooling | |
| CN113344045B (en) | Method for improving SAR ship classification precision by combining HOG characteristics | |
| CN108805102A (en) | A kind of video caption detection and recognition methods and system based on deep learning | |
| Liu et al. | Microseismic first-arrival picking using fine-tuning feature pyramid networks | |
| CN112580458A (en) | Facial expression recognition method, device, equipment and storage medium | |
| CN115659254A (en) | Power quality disturbance analysis method for power distribution network with bimodal feature fusion | |
| CN112883931A (en) | Real-time true and false motion judgment method based on long and short term memory network | |
| CN115578568A (en) | Noise correction algorithm driven by small-scale reliable data set | |
| CN110135435B (en) | Saliency detection method and device based on breadth learning system | |
| CN111882057A (en) | Multi-stage deep learning method oriented to space-time sequence data modeling and application | |
| Ye et al. | PlantBiCNet: A new paradigm in plant science with bi-directional cascade neural network for detection and counting | |
| CN110188621A (en) | A kind of three-dimensional face expression recognition methods based on SSF-IL-CNN | |
| CN105205476A (en) | Face recognition hardware framework based on LBP characteristics | |
| CN112966815A (en) | Target detection method, system and equipment based on impulse neural network | |
| CN116681921A (en) | Target labeling method and system based on multi-feature loss function fusion | |
| CN113705654A (en) | FFPN model-based microseism first-arrival intelligent pickup method, system, equipment and storage medium | |
| CN114821337B (en) | Semi-supervised SAR image building area extraction method based on phase consistency pseudo tag | |
| CN110674845A (en) | Dish identification method combining multi-receptive-field attention and characteristic recalibration |
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 |