CN116883709A - Carbonate fracture-cavity identification method and system based on channel attention mechanism - Google Patents
Carbonate fracture-cavity identification method and system based on channel attention mechanism Download PDFInfo
- Publication number
- CN116883709A CN116883709A CN202210313154.0A CN202210313154A CN116883709A CN 116883709 A CN116883709 A CN 116883709A CN 202210313154 A CN202210313154 A CN 202210313154A CN 116883709 A CN116883709 A CN 116883709A
- Authority
- CN
- China
- Prior art keywords
- attention mechanism
- model
- mark
- channel
- 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.)
- Pending
Links
- 230000007246 mechanism Effects 0.000 title claims abstract description 59
- 238000000034 method Methods 0.000 title claims abstract description 55
- BVKZGUZCCUSVTD-UHFFFAOYSA-L Carbonate Chemical compound [O-]C([O-])=O BVKZGUZCCUSVTD-UHFFFAOYSA-L 0.000 title claims abstract description 46
- 238000003384 imaging method Methods 0.000 claims abstract description 72
- 238000012549 training Methods 0.000 claims abstract description 44
- 238000012795 verification Methods 0.000 claims abstract description 39
- 238000010586 diagram Methods 0.000 claims abstract description 33
- 238000012360 testing method Methods 0.000 claims abstract description 28
- 230000006870 function Effects 0.000 claims description 15
- 230000004913 activation Effects 0.000 claims description 13
- 230000005284 excitation Effects 0.000 claims description 9
- 238000011176 pooling Methods 0.000 claims description 9
- 230000000694 effects Effects 0.000 claims description 7
- 238000001125 extrusion Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000010276 construction Methods 0.000 claims description 4
- 238000005457 optimization Methods 0.000 claims description 4
- 238000002360 preparation method Methods 0.000 claims description 2
- 238000013135 deep learning Methods 0.000 abstract description 9
- 238000003860 storage Methods 0.000 description 14
- 238000012545 processing Methods 0.000 description 13
- 238000005516 engineering process Methods 0.000 description 10
- 238000004590 computer program Methods 0.000 description 9
- 230000011218 segmentation Effects 0.000 description 8
- 238000013528 artificial neural network Methods 0.000 description 5
- 238000003708 edge detection Methods 0.000 description 3
- 230000003993 interaction Effects 0.000 description 3
- 239000011148 porous material Substances 0.000 description 2
- 230000000644 propagated effect Effects 0.000 description 2
- 239000011435 rock Substances 0.000 description 2
- 239000004215 Carbon black (E152) Substances 0.000 description 1
- 238000003491 array Methods 0.000 description 1
- -1 carbonate hydrocarbon Chemical class 0.000 description 1
- 238000002939 conjugate gradient method Methods 0.000 description 1
- 238000013527 convolutional neural network Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000011156 evaluation Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 229930195733 hydrocarbon Natural products 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 238000005070 sampling Methods 0.000 description 1
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
- G06N3/08—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)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
Carbonate fracture-cavity identification method and system based on channel attention mechanism, comprising the following steps: preparing three data sets of a training set, a verification set and a test set; constructing a Unet network model for introducing channel attention mechanism constraint based on a conventional Unet network model; inputting the well logging imaging images in the training set and the corresponding crack and hole marks into a constructed Unet network model which introduces the attention mechanism constraint of the channel for training; inputting the well logging imaging diagram in the verification set and the corresponding crack and hole marks into a constructed Unet network model for introducing the attention mechanism constraint of the channel; predictive marking of cracks and holes. According to the invention, deep learning is combined with carbonate fracture and hole recognition, accurate recognition of the carbonate fracture and hole is realized in an end-to-end mode, and the crack and hole recognition is performed based on the electric imaging and acoustic imaging diagram by using the method, so that the accuracy and the automation degree of crack and hole recognition can be improved.
Description
Technical Field
The invention belongs to the technical field of carbonate hydrocarbon reservoir development, and particularly relates to a carbonate fracture and tunnel recognition method and system based on a channel attention mechanism.
Background
The middle area of the Tarim basin tower is one of main battlefields for the oil and gas exploration of China sea-phase carbonate rock, and reservoirs develop complex and various reservoir space types and mainly develop secondary pores such as cracks, dissolving holes, karst cave and the like besides primary pores. The cracks, the solution holes and the solution holes are often effective reservoir spaces, so that the identification of the cracks, the solution holes and the solution holes in the carbonate rock is the key of reservoir evaluation.
Because the logging imaging graph can intuitively display the microscopic features of the stratum along the periphery of the well, the identification of cracks and holes is more accurate and effective. The cracks and holes have unique responses and image displays on the electric imaging and acoustic imaging diagrams. At present, the crack and hole identification technology based on the logging imaging diagram mainly adopts methods of human-computer interaction such as threshold segmentation and edge detection, and the like, so that the degree of automation is low, and the identification result is greatly influenced by human subjective judgment.
At present, the crack and hole identification technology based on the logging imaging diagram mainly adopts methods of human-computer interaction such as threshold segmentation and edge detection, and the like, so that the degree of automation is low, and the identification result is greatly influenced by human subjective judgment.
There is no report of deep learning semantic segmentation technology for crack and hole identification based on logging imaging graphs. The deep learning semantic segmentation technique is based on an encoder-decoder architecture. The encoder network is an alternate convolutional neural network, and comprises layers of convolution, pooling, nonlinear activation and the like, the output of each convolutional layer is characterized by different receptive fields, the spatial dimension of a characteristic image generated by the encoder network is smaller than that of an original image due to spatial pooling, and the decoder network expands the characteristic image into a final semantic segmentation result through up-sampling and anti-pooling. In conventional codec architecture, the top-level derived feature map of the encoder network is used as input to the decoder network, which contains high-level features that remain unchanged for small variations, which invariance is not ideal for the task of semantically segmenting dense image labels that require precise pixel information, as important relationships may be abstracted. In order to fully utilize the multi-level features extracted by the encoder, the U-net network introduces a 'skip connection' between the encoder and the decoder, as shown in fig. 2, the features extracted by each level encoder are duplicated to the decoders of the same level with identical weights, and the effect of doing so is that the neural network is equivalent processing to all the features of the image, and the importance of the features is not distinguished, and the processing mode is unfavorable for understanding the whole image by the network, so that the semantic segmentation is unfavorable.
Disclosure of Invention
The invention aims to provide a carbonate fracture-cavity recognition method and system based on a channel attention mechanism, so as to solve the problems that the traditional recognition method is low in automation degree and the recognition result is greatly influenced by artificial subjective judgment.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
a carbonate fracture-cavity identification method based on a channel attention mechanism comprises the following steps:
preparing three data sets of a training set, a verification set and a test set, wherein the training set is an image library of well logging imaging images of the cracks and the holes marked, the verification set is an image library of well logging imaging images, and the test set is an image library of well logging electric imaging images of the cracks and the holes to be identified;
constructing a Unet network model for introducing channel attention mechanism constraint based on a conventional Unet network model;
initializing model parameters, inputting a logging imaging diagram in a training set and corresponding crack and hole marks thereof into a constructed Unet network model for introducing the attention mechanism constraint of a channel to train;
inputting the well logging imaging diagram and the corresponding crack and hole marks in the verification set into a constructed Unet network model which introduces the attention mechanism constraint of the channel, wherein the output of the model is a predicted mark, and comparing the predicted mark with a real mark to verify;
the well logging imaging images in the test set are input into a trained and verified network, and the output of the network is the predictive mark of cracks and holes in the images.
Further, the training set, the verification set and the test set have the functions of training a model, adjusting the model and outputting an identification result, and the proportion of the training model to the verification set is 6:2:2.
further, each well logging electric imaging image of the training set and the verification set corresponds to a group of manually marked crack and hole marks, so that the well logging electric imaging images of the carbonate stratum which do not exist in the image library are increased, and the crack and hole marks are automatically carried out.
Further, the construction of the network model specifically includes: the method is improved on the basis of a conventional Unet network, and a Unet network which introduces the constraint of a channel attention mechanism is constructed, and the method comprises the following steps of: and adding an attention module after the convolution combination layer of the Unet encoder to realize the attention of different degrees to the characteristics of different channels, and then clipping the characteristics subjected to attention constraint and transmitting the characteristics to the corresponding convolution combination layer of the decoder for decoding.
Further, the channel attention mechanism module comprises three parts, namely extrusion, excitation and attention, wherein the extrusion part compresses characteristic information of the dimension H, W and C output by the convolution layer into a vector of the dimension 1, C by carrying out global average pooling or maximum pooling on the characteristics according to channels, wherein C is the number of characteristic channels;
the excitation part comprises a convolution layer 1-an activation layer 1-a convolution layer 2-an activation layer 2, wherein the convolution layer 1 reduces the characteristic dimension to 1/r originally, the activation layer 1 is a ReLu function to realize the nonlinearity of data, the convolution layer 2 increases the dimension back to 1 x C originally, the activation layer 2 is a Sigmoid function, excitation weight of 1 x C is obtained through the learning of convolution layer convolution kernel parameters, and the given weight range is between (0 and 1);
note that the part is to multiply the weight values with the feature map of the convolutional layer output.
Further, the model parameters are initialized as follows: selecting a model optimization algorithm, and initializing model parameters of initial values of batch size, iteration times and learning rate;
further, in the training model, the output of the model is a predicted mark, the predicted mark is compared with a real mark, meanwhile, the error of the predicted mark and the real mark is calculated, and if the error is larger than a set threshold value, the update parameter is reversely propagated; and repeating the iteration until the error is smaller than the set threshold value, stopping updating the parameters and saving the network parameters.
Further, the verification process is as follows: comparing the predicted mark with the real mark, calculating the error of the predicted mark and the real mark, and if the error is smaller than the set threshold value, proving that the trained network parameters are optimal, and identifying cracks and holes; if the error is larger than the set threshold, the trained network parameters are proved to be not optimal, and the network is returned to be retrained until a good identification effect can be obtained in the verification stage.
Further, the identification process is as follows: and inputting the images in the test set into a trained Unet network, and outputting the identification result.
Compared with the prior art, the invention has the following technical effects:
the invention provides a carbonate fracture and hole identification method based on a channel attention mechanism constraint Unet neural network, which combines deep learning with carbonate fracture and hole identification, realizes accurate identification of the carbonate fracture and hole in an end-to-end mode, and can improve the accuracy and the automation degree of fracture and hole identification by using the method to identify the fracture and hole based on electric imaging and acoustic imaging diagrams.
The carbonate fracture and hole recognition technology of the Unet network based on the channel attention mechanism constraint is used for improving the recognition quality of the carbonate fracture and hole, solving the problems that the traditional recognition method is low in automation degree and the recognition result is greatly influenced by artificial subjective judgment, realizing high-accuracy and automatic recognition of the carbonate fracture and hole, and providing support for the reliability and accuracy of subsequent logging geological interpretation.
Drawings
FIG. 1 is a block diagram of a U-net network constrained by the channel attention mechanism of the present invention
FIG. 2 is a diagram of a conventional U-net network architecture
FIG. 3 is a block diagram of a channel attention mechanism module
FIG. 4 is a flow chart of the invention for identifying a slot hole
FIG. 5 is a flowchart of a training step
FIG. 6 is a flowchart of an identification step
Detailed Description
The invention is further described below with reference to the accompanying drawings:
referring to fig. 1 to 6, a carbonate fracture-cavity recognition method based on a channel attention mechanism includes:
preparing three data sets of a training set, a verification set and a test set, wherein the training set is an image library of well logging imaging images of the cracks and the holes marked, the verification set is an image library of well logging imaging images, and the test set is an image library of well logging electric imaging images of the cracks and the holes to be identified;
constructing a Unet network model for introducing channel attention mechanism constraint based on a conventional Unet network model;
initializing model parameters, inputting a logging imaging diagram in a training set and corresponding crack and hole marks thereof into a constructed Unet network model for introducing the attention mechanism constraint of a channel to train;
inputting the well logging imaging diagram and the corresponding crack and hole marks in the verification set into a constructed Unet network model which introduces the attention mechanism constraint of the channel, wherein the output of the model is a predicted mark, and comparing the predicted mark with a real mark to verify;
the well logging imaging images in the test set are input into a trained and verified network, and the output of the network is the predictive mark of cracks and holes in the images.
The invention provides a carbonate fracture and hole identification method based on a channel attention mechanism constraint Unet neural network, which combines deep learning with carbonate fracture and hole identification, realizes accurate identification of the carbonate fracture and hole in an end-to-end mode, and can improve the accuracy and the automation degree of fracture and hole identification by using the method to identify the fracture and hole based on electric imaging and acoustic imaging diagrams.
Example 1:
a carbonate fracture-cavity identification method based on a channel attention mechanism comprises the following steps:
preparing three data sets of a training set, a verification set and a test set, wherein the training set is an image library of well logging imaging images of the cracks and the holes marked, the verification set is an image library of well logging imaging images, and the test set is an image library of well logging electric imaging images of the cracks and the holes to be identified;
constructing a Unet network model for introducing channel attention mechanism constraint based on a conventional Unet network model;
initializing model parameters, inputting a logging imaging diagram in a training set and corresponding crack and hole marks thereof into a constructed Unet network model for introducing the attention mechanism constraint of a channel to train;
inputting the well logging imaging diagram and the corresponding crack and hole marks in the verification set into a constructed Unet network model which introduces the attention mechanism constraint of the channel, wherein the output of the model is a predicted mark, and comparing the predicted mark with a real mark to verify;
the well logging imaging images in the test set are input into a trained and verified network, and the output of the network is the predictive mark of cracks and holes in the images.
Each well logging electric imaging image of the training set and the verification set corresponds to a group of manually marked crack and hole marks, and carbonate stratum well logging electric imaging images which do not exist in an image library are increased and crack and hole marks are carried out by oneself.
The construction of the network model specifically comprises the following steps: the method is improved on the basis of a conventional Unet network, and a Unet network which introduces the constraint of a channel attention mechanism is constructed, and the method comprises the following steps of: and adding an attention module after the convolution combination layer of the Unet encoder to realize the attention of different degrees to the characteristics of different channels, and then clipping the characteristics subjected to attention constraint and transmitting the characteristics to the corresponding convolution combination layer of the decoder for decoding.
In the training model, the output of the model is a predicted mark, the predicted mark is compared with a real mark, the error of the predicted mark and the real mark is calculated at the same time, and if the error is larger than a set threshold value, the update parameter is reversely propagated; and repeating the iteration until the error is smaller than the set threshold value, stopping updating the parameters and saving the network parameters.
The verification process is as follows: comparing the predicted mark with the real mark, calculating the error of the predicted mark and the real mark, and if the error is smaller than the set threshold value, proving that the trained network parameters are optimal, and identifying cracks and holes; if the error is larger than the set threshold, the trained network parameters are proved to be not optimal, and the network is returned to be retrained until a good identification effect can be obtained in the verification stage.
The identification flow is as follows: and inputting the images in the test set into a trained Unet network, and outputting the identification result.
The carbonate fracture and hole recognition technology of the Unet network based on the channel attention mechanism constraint is used for improving the recognition quality of the carbonate fracture and hole, solving the problems that the traditional recognition method is low in automation degree and the recognition result is greatly influenced by artificial subjective judgment, realizing high-accuracy and automatic recognition of the carbonate fracture and hole, and providing support for the reliability and accuracy of subsequent logging geological interpretation.
Example 2:
the invention discloses a fracture-cavity identification flow chart as shown in fig. 4, which comprises the following specific steps:
step 1: a data set is prepared.
Three data sets were prepared: training set, verification set and test set, the function of the three is training model respectively, adjustment model and output recognition result, and the proportion of the three is 6:2:2.
the training set is an image library of well logging imaging images marked with cracks and holes, each well logging imaging image corresponds to a group of manually marked crack and hole marks, and a user can also increase carbonate stratum well logging imaging images which do not exist in the image library and mark the cracks and holes by himself. The verification set is an image library of logging imaging images, and each logging imaging image corresponds to a group of manually marked crack and hole marks. The test set is an image library of logging electric imaging images of the cracks and holes to be identified.
Step 2: and constructing a network model.
Based on the conventional Unet network model, the method improves on the basis of the conventional Unet network, and constructs the Unet network for introducing the constraint of the channel attention mechanism.
The specific improvement method is shown in fig. 2:
and adding an attention module after the convolution combination layer of the Unet encoder to realize the attention of different degrees to the characteristics of different channels, and then clipping the characteristics subjected to attention constraint and transmitting the characteristics to the corresponding convolution combination layer of the decoder for decoding.
The channel attention mechanism module comprises three parts of extrusion, excitation and attention, and the structure of the channel attention mechanism module is shown in fig. 3.
The extrusion part compresses the characteristic information of the H-W-C dimension output by the convolution layer into a vector of the 1-C dimension by carrying out global average pooling or maximum pooling on the characteristics according to channels, wherein C is the number of characteristic channels. The excitation part comprises a convolution layer 1-an activation layer 1-a convolution layer 2-an activation layer 2, wherein the convolution layer 1 reduces the characteristic dimension to 1/r originally, the activation layer 1 is a ReLu function to realize the nonlinearity of data, the convolution layer 2 increases the dimension back to 1 x C originally, the activation layer 2 is a Sigmoid function, excitation weight of 1 x C is obtained through the learning of convolution layer convolution kernel parameters, and the given weight range is between (0 and 1). The attention part multiplies the weight value and the characteristic diagram output by the convolution layer to realize the weighting processing of the channel, thereby realizing the attention of important characteristics.
Step 3: parameters are initialized and the model is trained.
And selecting a proper model optimization algorithm, and initializing model parameters such as batch size, iteration times, learning rate initial values and the like.
Inputting the logging imaging graph and the corresponding crack and hole marks in the training set into the network model constructed in the step (2) for training, outputting the model as a predicted mark, comparing the predicted mark with a real mark, calculating the error of the predicted mark and the real mark, and if the error is larger than a set threshold (the threshold is set according to the actual situation), reversely transmitting the updated parameters.
And repeating the iteration until the error is smaller than the set threshold value, stopping updating the parameters and saving the network parameters.
And carrying out parameter solving by adopting optimization algorithms such as a gradient descent method, a conjugate gradient method and the like, and updating parameters according to error gradient information when error counter-propagates. A flowchart of the training model is shown in fig. 5.
Step 4: and (5) verifying the model.
Inputting the well logging imaging diagram in the verification set and the corresponding crack and hole marks into the network model constructed in the step (2), wherein the output of the model is a predicted mark, comparing the predicted mark with a real mark, calculating the error of the predicted mark and the real mark, and if the error is smaller than a set threshold (the threshold is set according to the actual situation), proving that the trained network parameters are optimal, and entering the step (5) to identify the crack and the hole.
If the error is larger than the set threshold, the trained network parameters are proved to be not optimal, and the step (3) is returned to for retraining the network until a good identification effect can be obtained in the verification stage.
Step 5: the method is applied to crack and hole identification of the actually measured logging imaging image. The well logging imaging images in the test set are input into the trained and verified network, and the output of the network is the predictive mark of cracks and holes in the images. The identified flowchart is shown in fig. 6.
The invention provides a carbonate fracture and hole recognition technology of a Unet network based on channel attention mechanism constraint, so as to improve the quality of carbonate fracture and hole recognition, solve the problems of low automation degree and large influence of artificial subjective judgment on recognition results of the traditional recognition method, realize high-accuracy and automatic recognition of the carbonate fracture and hole, and provide support for the reliability and accuracy of subsequent logging geological interpretation.
The method is suitable for a logging data processing flow, and because the logging imaging diagram can intuitively display the stratum microcosmic characteristics of the underground along the periphery of the well, when the electric imaging diagram and the acoustic imaging diagram are measured in the logging process, the provided technology can be used for identifying cracks and holes based on the electric imaging diagram and the acoustic imaging diagram, and the accuracy rate and the automation degree of crack and hole identification are improved.
Application prospect:
at present, intelligent algorithms based on deep learning are widely applied to various industries, and logging data processing interpretation algorithms combined with the deep learning technology also become important research directions in logging exploration. The crack and hole recognition technology based on the deep learning semantic segmentation can improve the recognition accuracy and the automation degree. Compared with the traditional identification methods such as threshold segmentation and edge detection adopting man-machine interaction, the technology combines with a deep learning method, has higher operation efficiency and more ideal reconstruction effect in the processing of a large amount of logging imaging data, has higher application value and ideal application prospect, and is expected to become one of powerful tools for improving the processing quality of the logging data.
The following are device embodiments of the present invention that may be used to perform method embodiments of the present invention. For details not disclosed in the apparatus embodiments, please refer to the method embodiments of the present invention.
In an embodiment of the present invention, a carbonate fracture-cavity recognition system based on a channel attention mechanism is provided, which can be used to implement the above-mentioned carbonate fracture-cavity recognition method based on the channel attention mechanism, and specifically, the system includes:
the data preparation module is used for preparing three data sets, namely a training set, a verification set and a test set, wherein the training set is an image library of well logging imaging images of the marked cracks and holes, the verification set is an image library of well logging imaging images, and the test set is an image library of well logging electric imaging images of the cracks and holes to be identified;
the network model construction module is used for constructing a Unet network model introducing the channel attention mechanism constraint based on a conventional Unet network model;
the training module is used for initializing model parameters, inputting a logging imaging graph in a training set and corresponding crack and hole marks into a constructed Unet network model which introduces the attention mechanism constraint of the channel for training;
the verification module is used for inputting the logging imaging diagram and the corresponding crack and hole marks in the verification set into a constructed Unet network model which introduces the attention mechanism constraint of the channel, outputting the model as a predicted mark, and comparing the predicted mark with a real mark to verify;
the identification module is used for inputting the logging imaging images in the test set into the trained and verified network, and the output of the network is the predictive mark of the cracks and holes in the images.
All relevant contents of each step involved in the foregoing embodiment of the carbonate fracture-cavity recognition method based on the channel attention mechanism constraint ultraviolet neural network may be cited in the functional description of the functional module corresponding to the carbonate fracture-cavity recognition system based on the channel attention mechanism constraint ultraviolet neural network in the embodiment of the present invention, which is not described herein.
The division of the modules in the embodiments of the present invention is schematically only one logic function division, and there may be another division manner in actual implementation, and in addition, each functional module in each embodiment of the present invention may be integrated in one processor, or may exist separately and physically, or two or more modules may be integrated in one module. The integrated modules may be implemented in hardware or in software functional modules.
In yet another embodiment of the present invention, a computer device is provided that includes a processor and a memory for storing a computer program including program instructions, the processor for executing the program instructions stored by the computer storage medium.
The processor may be a central processing unit (Central Processing Unit, CPU), but may also be other general purpose processors, digital signal processors (Digital Signal Processor, DSP), application specific integrated circuits (Application Specific Integrated Circuit, ASIC), off-the-shelf Programmable gate arrays (FPGAs) or other Programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc., which are the computational core and control core of the terminal adapted to implement one or more instructions, in particular adapted to load and execute one or more instructions within a computer storage medium to implement the corresponding method flow or corresponding functions; the processor disclosed by the embodiment of the invention can be used for the operation of the carbonate fracture-cavity identification method based on a channel attention mechanism.
In yet another embodiment of the present invention, a storage medium, specifically a computer readable storage medium (Memory), is a Memory device in a computer device, for storing a program and data. It is understood that the computer readable storage medium herein may include both built-in storage media in a computer device and extended storage media supported by the computer device. The computer-readable storage medium provides a storage space storing an operating system of the terminal. Also stored in the memory space are one or more instructions, which may be one or more computer programs (including program code), adapted to be loaded and executed by the processor. The computer readable storage medium herein may be a high-speed RAM memory or a non-volatile memory (non-volatile memory), such as at least one magnetic disk memory. One or more instructions stored in a computer-readable storage medium may be loaded and executed by a processor to implement the corresponding steps associated with the embodiments described above.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Finally, it should be noted that: the above embodiments are only for illustrating the technical aspects of the present invention and not for limiting the same, and although the present invention has been described in detail with reference to the above embodiments, it should be understood by those of ordinary skill in the art that: modifications and equivalents may be made to the specific embodiments of the invention without departing from the spirit and scope of the invention, which is intended to be covered by the claims.
Claims (10)
1. The carbonate fracture-cavity identification method based on the channel attention mechanism is characterized by comprising the following steps of:
preparing three data sets of a training set, a verification set and a test set, wherein the training set is an image library of well logging imaging images of the cracks and the holes marked, the verification set is an image library of well logging imaging images, and the test set is an image library of well logging electric imaging images of the cracks and the holes to be identified;
constructing a Unet network model for introducing channel attention mechanism constraint based on a conventional Unet network model;
initializing model parameters, inputting a logging imaging diagram in a training set and corresponding crack and hole marks thereof into a constructed Unet network model for introducing the attention mechanism constraint of a channel to train;
inputting the well logging imaging diagram and the corresponding crack and hole marks in the verification set into a constructed Unet network model which introduces the attention mechanism constraint of the channel, wherein the output of the model is a predicted mark, and comparing the predicted mark with a real mark to verify;
the well logging imaging images in the test set are input into a trained and verified network, and the output of the network is the predictive mark of cracks and holes in the images.
2. The carbonate fracture-cavity recognition method based on the channel attention mechanism according to claim 1, wherein the training set, the verification set and the test set have the functions of training a model, adjusting the model and outputting a recognition result, and the ratio of the training model to the verification set to the test set is 6:2:2.
3. the method for identifying carbonate fracture and vug based on a channel attention mechanism according to claim 1, wherein each of the well logging electric imaging images of the training set and the verification set corresponds to a group of manually marked fracture and vug marks, and the well logging electric imaging images of the carbonate stratum which do not exist in the image library are added and the fracture and the vug marks are carried out by themselves.
4. The carbonate fracture-cavity recognition method based on the channel attention mechanism of claim 1, wherein constructing the network model specifically comprises: the method is improved on the basis of a conventional Unet network, and a Unet network which introduces the constraint of a channel attention mechanism is constructed, and the method comprises the following steps of: and adding an attention module after the convolution combination layer of the Unet encoder to realize the attention of different degrees to the characteristics of different channels, and then clipping the characteristics subjected to attention constraint and transmitting the characteristics to the corresponding convolution combination layer of the decoder for decoding.
5. The method for identifying carbonate holes based on a channel attention mechanism according to claim 4, wherein the channel attention mechanism module comprises three parts of extrusion, excitation and attention, the extrusion part compresses characteristic information of dimension H x W x C output by the convolution layer into a vector of dimension 1 x C by global average pooling or maximum pooling of the characteristics by channels, where C is the number of characteristic channels;
the excitation part comprises a convolution layer 1-an activation layer 1-a convolution layer 2-an activation layer 2, wherein the convolution layer 1 reduces the characteristic dimension to 1/r originally, the activation layer 1 is a ReLu function to realize the nonlinearity of data, the convolution layer 2 increases the dimension back to 1 x C originally, the activation layer 2 is a Sigmoid function, excitation weight of 1 x C is obtained through the learning of convolution layer convolution kernel parameters, and the given weight range is between (0 and 1);
note that the part is to multiply the weight values with the feature map of the convolutional layer output.
6. The carbonate fracture-cavity recognition method based on the channel attention mechanism of claim 1, wherein model parameters are initialized as follows: and selecting a model optimization algorithm, and initializing model parameters of initial values of batch size, iteration times and learning rate.
7. The carbonate fracture-cavity recognition method based on the channel attention mechanism according to claim 1, wherein in the training model, the output of the model is a predicted mark, the predicted mark is compared with a real mark, the error of the predicted mark and the real mark is calculated at the same time, and if the error is greater than a set threshold, the update parameters are back-propagated; and repeating the iteration until the error is smaller than the set threshold value, stopping updating the parameters and saving the network parameters.
8. The carbonate fracture-cavity recognition method based on the channel attention mechanism of claim 1, wherein the verification process is as follows: comparing the predicted mark with the real mark, calculating the error of the predicted mark and the real mark, and if the error is smaller than the set threshold value, proving that the trained network parameters are optimal, and identifying cracks and holes; if the error is larger than the set threshold, the trained network parameters are proved to be not optimal, and the network is returned to be retrained until a good identification effect can be obtained in the verification stage.
9. The carbonate fracture-cavity recognition method based on the channel attention mechanism of claim 1, wherein the recognition flow is as follows: and inputting the images in the test set into a trained Unet network, and outputting the identification result.
10. Carbonate fracture-cave recognition system based on passageway attention mechanism, its characterized in that includes:
the data preparation module is used for preparing three data sets, namely a training set, a verification set and a test set, wherein the training set is an image library of well logging imaging images of the marked cracks and holes, the verification set is an image library of well logging imaging images, and the test set is an image library of well logging electric imaging images of the cracks and holes to be identified;
the network model construction module is used for constructing a Unet network model introducing the channel attention mechanism constraint based on a conventional Unet network model;
the training module is used for initializing model parameters, inputting a logging imaging graph in a training set and corresponding crack and hole marks into a constructed Unet network model which introduces the attention mechanism constraint of the channel for training;
the verification module is used for inputting the logging imaging diagram and the corresponding crack and hole marks in the verification set into a constructed Unet network model which introduces the attention mechanism constraint of the channel, outputting the model as a predicted mark, and comparing the predicted mark with a real mark to verify;
the identification module is used for inputting the logging imaging images in the test set into the trained and verified network, and the output of the network is the predictive mark of the cracks and holes in the images.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210313154.0A CN116883709A (en) | 2022-03-28 | 2022-03-28 | Carbonate fracture-cavity identification method and system based on channel attention mechanism |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210313154.0A CN116883709A (en) | 2022-03-28 | 2022-03-28 | Carbonate fracture-cavity identification method and system based on channel attention mechanism |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN116883709A true CN116883709A (en) | 2023-10-13 |
Family
ID=88259179
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210313154.0A Pending CN116883709A (en) | 2022-03-28 | 2022-03-28 | Carbonate fracture-cavity identification method and system based on channel attention mechanism |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116883709A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117649529A (en) * | 2024-01-30 | 2024-03-05 | 中国科学技术大学 | Logging data interpretation method based on multidimensional signal analysis neural network |
-
2022
- 2022-03-28 CN CN202210313154.0A patent/CN116883709A/en active Pending
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117649529A (en) * | 2024-01-30 | 2024-03-05 | 中国科学技术大学 | Logging data interpretation method based on multidimensional signal analysis neural network |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102302609B1 (en) | Neural Network Architecture Optimization | |
| US11501153B2 (en) | Methods and apparatus for training a neural network | |
| CN112487807B (en) | Text relation extraction method based on expansion gate convolutional neural network | |
| CN110023963A (en) | Use Processing with Neural Network text sequence | |
| US8090538B2 (en) | System and method for interpretation of well data | |
| CN105930277A (en) | Defect source code locating method based on defect report analysis | |
| CN104217216A (en) | Method and device for generating detection model, method and device for detecting target | |
| AU2020270516B2 (en) | Holographic quantum dynamics simulation | |
| CN106340297A (en) | Speech recognition method and system based on cloud computing and confidence calculation | |
| CN104933428A (en) | Human face recognition method and device based on tensor description | |
| CN108549817A (en) | A kind of software security flaw prediction technique based on text deep learning | |
| CN110349597A (en) | A kind of speech detection method and device | |
| CN110490320A (en) | Deep neural network structural optimization method based on forecasting mechanism and Genetic Algorithm Fusion | |
| CN113359212A (en) | Reservoir characteristic prediction method and model based on deep learning | |
| CN115393656B (en) | Automatic classification method for stratum classification of logging-while-drilling image | |
| CN111882042B (en) | Neural network architecture automatic search method, system and medium for liquid state machine | |
| CN110162972A (en) | A kind of UAF leak detection method based on sentence combined coding deep neural network | |
| CN115146700A (en) | Runoff prediction method based on Transformer sequence-to-sequence model | |
| CN115935372A (en) | Vulnerability detection method based on graph embedding and bidirectional gated graph neural network | |
| CN113920379B (en) | Zero sample image classification method based on knowledge assistance | |
| CN109800776A (en) | Material mask method, device, terminal and computer readable storage medium | |
| CN109214401A (en) | SAR image classification method and device based on stratification autocoder | |
| CN114155477A (en) | Semi-supervised video paragraph positioning method based on average teacher model | |
| CN112417890B (en) | Fine granularity entity classification method based on diversified semantic attention model | |
| CN117058716A (en) | Cross-domain behavior recognition method and device based on image pre-fusion |
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 |