CN115908887A - Rock image classification method, system, equipment and medium - Google Patents
Rock image classification method, system, equipment and medium Download PDFInfo
- Publication number
- CN115908887A CN115908887A CN202210676726.1A CN202210676726A CN115908887A CN 115908887 A CN115908887 A CN 115908887A CN 202210676726 A CN202210676726 A CN 202210676726A CN 115908887 A CN115908887 A CN 115908887A
- Authority
- CN
- China
- Prior art keywords
- rock
- rock image
- image classification
- classification model
- target
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Landscapes
- Image Analysis (AREA)
Abstract
The invention discloses a rock image classification method, a system, equipment and a medium, wherein the rock image classification method comprises the following steps: constructing a rock image classification model; training the rock image classification model; and acquiring a target rock image, inputting the target rock image to a trained rock image classification model, and predicting the rock class. The method can more comprehensively extract the characteristics of the rock images, improve the accuracy of rock image classification and improve the detail distinguishing capability on the rock images.
Description
Technical Field
The invention relates to the technical field of image processing, in particular to a rock image classification method, system, equipment and medium based on a residual error network fused with an attention mechanism or a residual error network fused with a convolutional neural network.
Background
The identification and classification of the rocks are an important field of geological research and an important content of geological investigation work, and the accurate classification of the rocks is beneficial to dividing the stratigraphic times and establishing a reliable geological time scale. Rock type identification also plays a certain role in the fields of oil and gas exploration, mineral resource exploration and the like.
The traditional rock classification method mainly comprises a physical test method and a mathematical analysis method. The physical test method is to research and analyze physical properties of rock, such as density, hardness, permeability, magnetism, electrical property, thermal conductivity, spectral characteristics and the like, so as to achieve the purpose of rock classification. For example, the land-based salinized lake basin fine particle sedimentary rock is classified according to the rock sedimentary formation characteristics and the rock composition. The physical test method has great workload, needs professional instruments and equipment and requires researchers to have higher professional literacy. The mathematical analysis method refers to analyzing rock features using a mathematical method to classify the rock. The data analysis method has the advantages of large calculated amount, high requirement on research personnel and low efficiency.
Due to the fact that the types of rocks are various and the characteristics are similar, the traditional rock classification method is high in workload, depends on professional instruments and equipment, and has high requirements on researchers; the traditional machine learning rock image classification accuracy is limited, the feature extraction has larger uncertainty, and the existing rock classification model constructed based on the deep learning cannot accurately distinguish the rock types in local details.
Disclosure of Invention
In view of the above, the invention provides a rock image classification method, system, device and medium based on a residual error network fused with an attention mechanism, which can focus on important features of rock images and improve the detail distinguishing accuracy of a classification model on the rock images.
The invention provides a rock image classification method in a first aspect, which comprises the following steps: constructing a rock image classification model; training a rock image classification model; and acquiring a target rock image, inputting the target rock image to a trained rock image classification model, and predicting the rock class.
Further, the step of constructing the rock image classification model comprises: selecting a residual error network Resnet50 as a backbone network; and (5) fusing an attention mechanism in a residual error network Resnet50 to obtain a rock image classification model.
Further, the step of inputting the target rock image into a trained rock image classification model and predicting the rock class comprises: extracting a plurality of characteristic maps in the target rock image by using a residual error network; learning the weight of each feature map by adopting an attention mechanism; and multiplying the weight of each characteristic diagram in the target rock image by the original characteristics of the characteristic diagram to obtain various rock characteristics.
Further, the step of constructing the rock image classification model comprises: selecting a residual error network ResNet-18 as a main network; and (4) fusing a convolutional neural network in the residual error network ResNet-18 to obtain a rock image classification model.
Further, the step of inputting the target rock image into a trained rock image classification model and predicting the rock class comprises: extracting a plurality of characteristic maps in the target rock image by using a residual error network; and extracting rock characteristics in each characteristic diagram by adopting a convolutional neural network method to obtain various types of rocks.
Further, the step of extracting a plurality of feature maps in the target rock image by using the residual error network comprises: inputting the target rock image into a residual error network for convolution processing to obtain a plurality of characteristic graphs; and carrying out batch normalization processing on the plurality of feature maps, and linearly correcting the plurality of feature maps.
Further, the step of training the rock image classification model includes: establishing a rock image sample library, wherein the rock image sample library comprises a plurality of rock sample image data; splitting the rock image sample library to form a training set and a testing set, wherein the training set is the first part of rock sample image data; the test set is a second portion of rock sample image data; training the rock image classification model according to the training set to obtain a pre-trained rock image classification model; inputting the test set to a pre-trained rock image classification model, and determining evaluation parameters; determining that the evaluation parameters meet preset requirements; and determining a trained rock image classification model based on the condition that the evaluation parameters meet the preset requirements.
A second aspect of the invention provides a rock image classification system comprising: the model building module is used for building a rock image classification model; the model training module is used for training a rock image classification model; and the classification module is used for acquiring a target rock image, inputting the target rock image to a trained rock image classification model and predicting the rock class.
A third aspect of the invention provides a rock image classification apparatus comprising: a memory for storing a computer program; a processor for implementing the steps of the rock image classification method as described above when executing the computer program.
A fourth aspect of the invention provides a computer readable storage medium having stored thereon a computer program which, when executed by a processor, carries out the steps of the rock image classification method as described above.
According to the rock image classification method and system, the rock image classification model is constructed by adopting the residual error network fused with the attention mechanism or the convolutional neural network, the rock image classification model is trained and evaluated by utilizing various rock sample image data, the target rock image is obtained, the rock type is predicted by utilizing the rock image classification model, the characteristics of the rock image can be extracted more comprehensively, the accuracy of rock image classification is improved, and the detail distinguishing capability on the rock image is improved.
Drawings
For purposes of illustration and not limitation, the present invention will now be described in accordance with its preferred embodiments, particularly with reference to the accompanying drawings, in which:
fig. 1 is a flowchart of a rock image classification method based on a residual error network of a fusion attention mechanism according to an embodiment of the present invention;
FIG. 2 is a flowchart of a rock image classification method based on a residual error network fused with a convolutional neural network according to a second embodiment of the present invention;
fig. 3 is a schematic structural diagram of a rock image classification system based on a residual error network of a fusion attention mechanism according to a third embodiment of the present invention;
fig. 4 is a schematic structural diagram of a rock image classification system based on a residual error network of a fusion attention mechanism according to a fourth embodiment of the present invention.
Detailed Description
In order that the above objects, features and advantages of the present invention can be more clearly understood, a detailed description of the present invention will be given below with reference to the accompanying drawings and specific embodiments. It should be noted that the embodiments of the present invention and features of the embodiments may be combined with each other without conflict.
In the following description, numerous specific details are set forth to provide a thorough understanding of the present invention, and the described embodiments are merely a subset of the embodiments of the present invention, rather than a complete embodiment. 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.
Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs. The terminology used in the description of the invention herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention.
Example one
The residual network is characterized by easy optimization and can improve accuracy by adding considerable depth. The inner residual block uses jump connection, and the problem of gradient disappearance caused by depth increase in a deep neural network is relieved.
The attention mechanism may enable the neural network to focus on a subset of its inputs (or features): a particular input is selected. Attention may be applied to any type of input regardless of its shape. In the case of limited computing power, an attention mechanism (attention mechanism) is a resource allocation scheme of a main means for solving the information overload problem, and computing resources are allocated to more important tasks.
Attention is generally divided into two categories: one is conscious attention from top to bottom, called focused (focus) attention. Focused attention refers to attention that has a predetermined purpose, is task dependent, and is actively focused consciously on a subject; the other is a bottom-up unconscious attention called salience-based attention. Attention based on significance is attention driven by external stimuli, does not require active intervention, and is also task independent. If the stimulation information of a subject differs from its surroundings, an unconscious "winner-take-all" or gating (gating) mechanism may divert attention to the subject. Regardless of whether such attention is intended or unintended, most human brain activities require attention, such as memorizing information, reading or thinking, and the like.
Rock classification means that geologic bodies formed by gathering rock-making minerals according to a certain structure become rocks, and can be divided into three major types of rock pulp rock, sedimentary rock and metamorphic rock according to the causes of the geologic bodies. Magma rock is also called igneous rock and is formed by that magma under the crust rises along the weak zone of the crust to invade the crust or is sprayed out of the ground surface and then is condensed.
The method aims at the problems that the accuracy rate of the traditional rock image classification method based on machine learning is limited, and the feature extraction has larger uncertainty and the like; the invention provides a rock image classification method based on a residual error network with a fusion attention mechanism, which can focus on important features of rock images, improve the detail distinguishing capability of a model on the rock images and improve the classification accuracy.
Fig. 1 is a flowchart of a rock image classification method based on a residual error network of a fusion attention mechanism according to an embodiment of the present invention. The rock image classification method adopts a residual error network fused with an attention mechanism to construct a rock image classification model, utilizes image data of various rock samples to train and evaluate the rock image classification model, obtains a target rock image, and utilizes the rock image classification model to predict rock classes.
Referring to fig. 1, the rock image classification method based on the residual error network of the fusion attention mechanism includes the following steps:
and S100, constructing a rock image classification model based on a residual error network of the fusion attention mechanism.
The traditional convolutional neural network realizes feature extraction through multilayer stacking, and the feature extraction capability can be improved by increasing the depth of the convolutional neural network. However, increasing the number of network layers when reaching a certain depth cannot further improve the model effect, but reduces the convergence rate of the model, and may cause the situation of gradient disappearance, and meanwhile, an excessively deep network may also cause the accuracy of model classification to decrease. The residual error network is characterized in that a jump connection and identity mapping mechanism are added between weight layers, so that the problems of gradient disappearance, model degradation and the like caused by too deep traditional convolutional neural networks can be effectively solved. Therefore, the residual error network is selected as the skeleton network, and the attention mechanism is fused in the residual error network.
The basic unit of the residual error network is a residual error unit, and jump connection and identity mapping are added to the residual error unit on the basis of the traditional convolutional neural network. The formula for the residual unit is as follows:
y＝F(x)+x
where y represents the output of the residual unit; x represents the input of the residual unit; f (x) denotes residual mapping.
The residual Block is a stack of residual units, and has two types, namely an identity Block (ID Block) and a convolution Block (CONV Block). The output and the input of the constant block have the same shape, and the length, the width and the dimension of the feature extracted by convolution cannot be changed. The convolution block adds a convolution operation to the jump connection to adjust the shape of the input matrix.
The Attention Mechanism (Attention Mechanism) is to give different weights to different inputs, so that feature selection or region selection is emphasized. The attention mechanism applies human attention to the machine, and lets the machine learn to perceive important and unimportant parts of the data. Adding a focus mechanism to the network allows the network to ignore irrelevant information and focus on important information.
The method comprises the steps of automatically learning the importance degree of each channel, then improving the characteristics useful for the current task according to the importance degree, and inhibiting the characteristics which are not useful for the current task. The set is formed by a collection of SE modules, which are mainly divided into two operations, namely squeezing (Squeeze) and Excitation (Excitation). The extrusion operation is to compress the features along the spatial dimension, match the output dimension with the number of the input channel features, convert each two-bit channel feature into a scalar, and enable each feature map to acquire global information. The excitation operation is to fuse channel information through two full-connection layers, obtain the nonlinear relation between channels by using a Sigmoid gate function, and perform excitation operation F ex The formula of (1) is as follows:
e C ＝F ex (s，W)＝σ(g(s，W))＝σ(W 2 δ(W 1 s))
where, σ represents a gate function,
δ represents the linear activation function operation.
And finally multiplying the learned activation value of each channel with the original characteristic of the characteristic diagram to obtain a final output result.
In step S100, the method for constructing the rock image classification model includes:
s101, selecting a residual error network Resnet50 as a backbone network, wherein the residual error network Resnet50 comprises a plurality of residual error blocks.
In this embodiment, a residual error network Resnet50 is selected as a backbone network, and the residual error network Resnet50 is composed of 49 convolutional layers and 1 full link layer. In order to solve the problems of gradient disappearance, gradient explosion and the like after data convolution, the residual error network Resnet50 performs Batch Normalization (BN) processing after each convolution, and adjusts the data to normal distribution with a mean value of 0 and a variance of 1. The final fully connected layer is then adjusted according to the classification class number using a modified Linear Unit (ReLU) as the activation function (zeroing out part of the output data, making gradient descent and back propagation more efficient).
And S102, fusing an attention mechanism in a residual error network Resnet50 to obtain a rock image classification model.
In this embodiment, an attention mechanism (squeize and Excitation Networks, send) is fused in the Residual Network Resnet50 to obtain a Residual Network model (Residual Network With squeize-and-Excitation, resnet-SE) of the attention mechanism, and the Residual Network model of the attention mechanism is used as a rock image classification model.
Specifically, the specific implementation manner of merging the attention mechanism in the residual error network Resnet50 in step S102 is as follows:
after the SE module in the attention mechanism is added to each residual block, recalibration is carried out on different network channel characteristics, more weight is given to important characteristics, and the classification capability of the classification model on the rock image is improved.
The embodiment integrates an attention mechanism in a residual error network, can effectively prevent the problems of degradation and the like of a classification model in the rock image training process, protects the information integrity of the rock image, simplifies the learning difficulty of the classification model on the rock image, and enables the classification model to extract the characteristics of the rock image more completely.
And S200, training a rock image classification model.
In this embodiment, the concrete implementation method for training the rock image classification model in step S200 is as follows:
s201, establishing a rock image sample library.
The embodiment collects the image data of 15 common rock samples, and establishes a rock image sample library according to the image data of the 15 common rock samples.
S202, splitting a rock image sample library to form a training set and a testing set, wherein the training set is the image data of the first part of rock samples; the test set is a second portion of rock sample image data.
In the embodiment, a part of rock sample image data in a rock image sample library is used as a training set for training a rock image classification model; and taking the other part of the rock sample image data as a test set for evaluating the rock image classification model.
And S203, training the rock image classification model according to the training set to obtain a pre-trained rock image classification model.
And S204, inputting the test set to a pre-trained rock image classification model, and determining evaluation parameters.
And S205, determining that the evaluation parameters meet preset requirements.
And S206, determining the trained rock image classification model based on the evaluation parameters meeting the preset requirements.
In the embodiment, image data of 15 common rock samples are adopted to train and evaluate the rock image classification model so as to improve the identification accuracy of the rock image classification model.
S300, acquiring a target rock image, inputting the target rock image into a trained rock image classification model, predicting the rock classification, and obtaining a rock classification result.
In this embodiment, in step S300, the target rock image is input into a trained rock image classification model, and the step of predicting the rock class includes:
s301, extracting a plurality of characteristic maps in the target rock image by using a residual error unit in a residual error network;
s302, learning the weight of each feature map by adopting an SE module of a channel attention mechanism;
and S303, multiplying the weight of each feature map in the target rock image by the original features of the feature maps to obtain various rock features, and realizing rock classification.
In this embodiment, a target rock image to be classified is acquired, the acquired target rock image is input to the trained rock image classification model obtained in step S200, rock characteristics are calculated through the trained rock image classification model, and a rock class prediction result is output, so that automatic and accurate classification of rock images such as grayish black mudstone, black coal, gray argillaceous siltstone, gray fine sandstone, light gray fine sandstone, dark gray siltstone mudstone, and dark gray mudstone is achieved.
According to the rock image classification method based on the residual error network with the fusion attention mechanism, the rock image classification model is built through the residual error network with the fusion attention mechanism, the rock image classification model is trained and evaluated through the multiple rock sample image data, the target rock image is obtained, the rock type is predicted through the rock image classification model, the characteristics of the rock image can be extracted more comprehensively, the accuracy of rock image classification is improved, and the detail distinguishing capability on the rock image is improved.
Example two
Convolutional Neural Networks (CNN) are a class of feed forward Neural Networks (fed forward Neural Networks) that contain convolution computations and have a deep structure, and are one of the representative algorithms for deep learning (deep learning). Convolutional Neural Networks have a representation learning (representation learning) capability, and are capable of performing Shift-Invariant classification (Shift-Invariant classification) on input information according to their hierarchical structure, and are therefore also called "Shift-Invariant Artificial Neural Networks (SIANN).
The study of convolutional neural networks began in the 80 to 90 s of the twentieth century, with time delay networks and LeNet-5 being the earliest convolutional neural networks that emerged; after the twenty-first century, with the introduction of deep learning theory and the improvement of numerical computing equipment, convolutional neural networks have been rapidly developed and applied to the fields of computer vision, natural language processing, and the like.
The convolutional neural network is constructed by imitating a visual perception (visual perception) mechanism of a living being, can perform supervised learning and unsupervised learning, and has the advantages that the convolutional neural network can learn grid-like topologic features such as pixels and audio with small calculation amount, has stable effect and has no additional feature engineering (feature engineering) requirement on data due to the fact that convolutional kernel parameter sharing in an implicit layer and sparsity of connection between layers.
The method aims at the problems that the accuracy rate of the traditional rock image classification method based on machine learning is limited, and the feature extraction has larger uncertainty and the like; the invention provides a rock image classification method based on a residual error network fused with a convolutional neural network, which can focus on important features of rock images, improve the detail distinguishing capability of a classification model on the rock images and improve the classification accuracy.
Fig. 2 is a flowchart of a rock image classification method based on a residual error network fused with a convolutional neural network according to a second embodiment of the present invention. The rock image classification method adopts a residual error network fused with a convolutional neural network to construct a rock image classification model, utilizes various rock sample image data to train and evaluate the rock image classification model, obtains a target rock image, and utilizes the rock image classification model to predict the rock category.
Referring to fig. 2, the rock image classification method based on the residual error network fused with the convolutional neural network includes the following steps:
s400, building a rock image classification model based on a residual error network fused with the convolutional neural network.
In step S400, the method for constructing the rock image classification model includes:
s401, selecting a residual error network ResNet-18 as a backbone network, wherein the residual error network ResNet-18 comprises a plurality of residual error blocks.
The residual error network adopted by the embodiment can effectively prevent the problems of degradation and the like of the classification model in the rock image training process, protect the information integrity of the rock image, simplify the learning difficulty of the classification model on the rock image, and enable the classification model to extract the characteristics of the rock image more completely.
S402, fusing the convolution neural network in the residual error network ResNet-18 to obtain a rock image classification model.
In the embodiment, the convolutional neural network is fused in the residual error network ResNet-18 to obtain a residual error network model fused with the convolutional neural network, and the residual error network model fused with the convolutional neural network is used as a rock image classification model.
Specifically, the specific implementation manner of fusing the convolutional neural network in the residual error network ResNet-18 in step S402 is as follows:
after the convolutional neural network is added to each residual block, recalibration is carried out on different network channel characteristics, more weights are given to important characteristics, and the classification capability of the classification model on the rock images is improved.
The analysis of the characteristic diagram generated in the learning process of the convolutional neural network successfully extracts minerals in various types of rocks, such as quartz, feldspar, mica and other minerals in granite, amphibole, plagioclase and other minerals in amphibole, sericite and other minerals in phyllite, so that the convolutional neural network method can effectively extract the mineral component characteristics of the rocks, the effectiveness of the convolutional neural network method on rock identification is also illustrated, the rock identification according to the mineral components is facilitated, the rock identification is effective, and the accuracy rate of more than 70% can be achieved.
And S500, training a rock image classification model.
In this embodiment, the concrete implementation method for training the rock image classification model in step S500 is as follows:
s501, establishing a rock image sample library.
In this embodiment, image data of 5 common rock samples, such as quartz, amphibole, biotite, garnet, olivine, are collected, and a rock image sample library is established according to the image data of the 5 common rock samples.
S502, splitting a rock image sample library to form a training set and a test set, wherein the training set is the image data of the first part of rock samples; the test set is a second portion of rock sample image data.
In the embodiment, a part of rock sample image data in a rock image sample library is used as a training set for training a rock image classification model; and taking the other part of the rock sample image data as a test set for evaluating the rock image classification model.
And S503, training the rock image classification model according to the training set to obtain a pre-trained rock image classification model.
And S504, inputting the test set to a pre-trained rock image classification model, and determining evaluation parameters.
And S505, determining that the evaluation parameters meet preset requirements.
And S506, determining the trained rock image classification model based on the evaluation parameters meeting the preset requirements.
In this embodiment, image data of 5 kinds of common rock samples are adopted to train and evaluate the rock image classification model, so as to improve the recognition accuracy of the rock image classification model.
S600, acquiring a target rock image, inputting the target rock image into a trained rock image classification model, predicting the rock classification, and obtaining a rock classification result.
In this embodiment, in step S600, the target rock image is input to a trained rock image classification model, and the step of predicting the rock class includes:
s601, extracting a plurality of characteristic maps in the target rock image by using a residual error unit in a residual error network;
and S602, extracting rock features in each feature map by adopting a convolutional neural network method to obtain multiple types of rocks.
In this embodiment, a target rock image to be classified is acquired, the acquired target rock image is input to the trained rock image classification model obtained in step S500, rock characteristics are calculated through the trained rock image classification model, and a rock class prediction result is output, so that seven types of rock images, namely, a gray black mudstone, black coal, gray argillaceous siltstone, gray fine sandstone, light gray fine sandstone, dark gray siltstone, and dark gray mudstone, are automatically and accurately classified.
According to the rock image classification method based on the residual error network fused with the convolutional neural network, the residual error network fused with the convolutional neural network is adopted to construct a rock image classification model, the rock image classification model is trained and evaluated by utilizing various rock sample image data, a target rock image is obtained, the rock type is predicted by utilizing the rock image classification model, the characteristics of the rock image can be extracted more comprehensively, the accuracy of rock image classification is improved, and the detail distinguishing capability on the rock image is improved.
EXAMPLE III
Fig. 3 is a block diagram of a rock image classification system 20 according to a third embodiment of the present invention.
In this embodiment, the rock image classification system 20 may be applied to a computer device, and the rock image classification system 20 may include a plurality of functional modules composed of program code segments. The program code of the various program segments in the rock image classification system 20 may be stored in a memory of a computer device and executed by at least one processor of the computer device to implement (see detailed description of fig. 1) the rock image classification function.
In this embodiment, the rock image classification system 20 may be divided into a plurality of functional modules according to the functions performed by the system. The functional module may include: a model building module 201, a model training module 202, and a classification module 203. The module referred to herein is a series of computer program segments capable of being executed by at least one processor and capable of performing a fixed function and is stored in memory. In the present embodiment, the functions of the modules will be described in detail in the following embodiments.
The model construction module 201 is used for constructing a rock image classification model; the model training module 202 is used for training a rock image classification model; and the classification module 203 is used for acquiring a target rock image, inputting the target rock image to a trained rock image classification model, and predicting the rock class.
Example four
Corresponding to the above method embodiment, referring to fig. 4, fig. 4 is a schematic diagram of a rock image classification apparatus provided by the present invention, and the apparatus 30 may include:
a memory 31 for storing a computer program;
the processor 32, when executing the computer program stored in the memory 31, may implement the following steps:
constructing a rock image classification model; training a rock image classification model; and acquiring a target rock image, inputting the target rock image to a trained rock image classification model, and predicting the rock class.
For the introduction of the device provided by the present invention, please refer to the above method embodiment, which is not described herein again.
Corresponding to the above method embodiment, the present invention further provides a computer-readable storage medium, on which a computer program is stored, where the computer program, when executed by a processor, can implement the following steps:
constructing a rock image classification model; training a rock image classification model; and acquiring a target rock image, inputting the target rock image to a trained rock image classification model, and predicting the rock class.
The computer-readable storage medium may include: a U-disk, a removable hard disk, a Read-only Memory (ROM), a Random Access Memory (RAM), a magnetic disk, an optical disk, or other various media capable of storing program codes.
For the introduction of the computer-readable storage medium provided by the present invention, please refer to the above method embodiments, which are not described herein again.
The embodiments are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same or similar parts among the embodiments are referred to each other. The device, the apparatus and the computer-readable storage medium disclosed in the embodiments correspond to the method disclosed in the embodiments, so that the description is simple, and the relevant points can be referred to the description of the method.
The above-described embodiments should not be construed as limiting the scope of the invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions can occur, depending on design requirements and other factors. Any modification, equivalent replacement, and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. A rock image classification method, comprising:
constructing a rock image classification model;
training the rock image classification model;
and acquiring a target rock image, inputting the target rock image to a trained rock image classification model, and predicting the rock class.
2. The rock image classification method according to claim 1, characterized in that the step of constructing a rock image classification model comprises:
selecting a residual error network Resnet50 as a backbone network;
and (5) fusing an attention mechanism in a residual error network Resnet50 to obtain a rock image classification model.
3. The rock image classification method according to claim 2, wherein the step of inputting the target rock image into a trained rock image classification model and predicting the rock class comprises:
extracting a plurality of characteristic maps in the target rock image by using a residual error network;
learning the weight of each feature map by adopting an attention mechanism;
and multiplying the weight of each characteristic diagram in the target rock image by the original characteristics of the characteristic diagram to obtain various rock characteristics.
4. The rock image classification method according to claim 1, characterized in that the step of constructing a rock image classification model comprises:
selecting a residual error network ResNet-18 as a main network;
and (4) fusing a convolutional neural network in the residual error network ResNet-18 to obtain a rock image classification model.
5. The method for classifying rock images according to claim 4, wherein the step of inputting the target rock image into a trained rock image classification model and predicting rock classes comprises:
extracting a plurality of characteristic maps in the target rock image by using a residual error network;
and extracting rock characteristics in each characteristic diagram by adopting a convolutional neural network method to obtain various types of rocks.
6. A rock image classification method according to claim 3 or 5, wherein the step of extracting a plurality of feature maps in the target rock image using a residual network comprises:
inputting the target rock image into the residual error network for convolution processing to obtain a plurality of characteristic graphs;
and carrying out batch normalization processing on the plurality of feature maps, and linearly correcting the plurality of feature maps.
7. The method of rock image classification of claim 1, wherein the step of training a rock image classification model comprises:
establishing a rock image sample library, wherein the rock image sample library comprises a plurality of rock sample image data;
splitting the rock image sample library to form a training set and a testing set, wherein the training set is the first part of rock sample image data; the test set is a second portion of rock sample image data;
training the rock image classification model according to the training set to obtain a pre-trained rock image classification model;
inputting the test set to a pre-trained rock image classification model, and determining evaluation parameters;
determining that the evaluation parameters meet preset requirements;
and determining a trained rock image classification model based on the condition that the evaluation parameters meet the preset requirements.
8. A rock image classification system, comprising:
the model building module is used for building a rock image classification model;
the model training module is used for training a rock image classification model;
and the classification module is used for acquiring a target rock image, inputting the target rock image to a trained rock image classification model and predicting the rock class.
9. A rock image classification device, comprising:
a memory for storing a computer program;
a processor for implementing the steps of the rock image classification method according to any one of claims 1 to 7 when executing the computer program.
10. A computer-readable storage medium, characterized in that a computer program is stored thereon, which computer program, when being executed by a processor, carries out the steps of the rock image classification method according to any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210676726.1A CN115908887A (en) | 2022-06-15 | 2022-06-15 | Rock image classification method, system, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210676726.1A CN115908887A (en) | 2022-06-15 | 2022-06-15 | Rock image classification method, system, equipment and medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115908887A true CN115908887A (en) | 2023-04-04 |
Family
ID=86471416
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210676726.1A Pending CN115908887A (en) | 2022-06-15 | 2022-06-15 | Rock image classification method, system, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908887A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117474956A (en) * | 2023-12-25 | 2024-01-30 | 浙江优众新材料科技有限公司 | Light field reconstruction model training method based on motion estimation attention and related equipment |
-
2022
- 2022-06-15 CN CN202210676726.1A patent/CN115908887A/en active Pending
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117474956A (en) * | 2023-12-25 | 2024-01-30 | 浙江优众新材料科技有限公司 | Light field reconstruction model training method based on motion estimation attention and related equipment |
| CN117474956B (en) * | 2023-12-25 | 2024-03-26 | 浙江优众新材料科技有限公司 | Light field reconstruction model training method based on motion estimation attention and related equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Buscombe | SediNet: A configurable deep learning model for mixed qualitative and quantitative optical granulometry | |
| Boisvert et al. | Geometallurgical modeling at Olympic dam mine, South Australia | |
| Behrens et al. | Multiscale contextual spatial modelling with the Gaussian scale space | |
| Pérez et al. | Verifying the high-order consistency of training images with data for multiple-point geostatistics | |
| Koch et al. | Automated drill core mineralogical characterization method for texture classification and modal mineralogy estimation for geometallurgy | |
| Xu et al. | Deep learning of rock microscopic images for intelligent lithology identification: Neural network comparison and selection | |
| CN117171602B (en) | Method and system for monitoring biodiversity protection area | |
| Emery et al. | Truncated Gaussian simulation of discrete-valued, ordinal coregionalized variables | |
| CN115908887A (en) | Rock image classification method, system, equipment and medium | |
| CN110222832A (en) | Entrance of Changjiang River salt marshes macrobenthos habitat simulation prediction technique | |
| CN105277979B (en) | The optimization method and device of a kind of seismic properties | |
| CN114139819A (en) | Geochemical variable space prediction method based on geostatistical weighted random forest | |
| Jacq et al. | Sedimentary structure discrimination with hyperspectral imaging in sediment cores | |
| CN112417267A (en) | User behavior analysis method and device, computer equipment and storage medium | |
| CN115655612A (en) | Tunnel blasting charging structure identification recording method, device, equipment and medium | |
| CN115147615A (en) | Rock image classification method and device based on metric learning network | |
| CN114359617A (en) | Method for identifying lithology of rock based on lightweight convolutional neural network | |
| Mountrakis et al. | Assessing reference dataset representativeness through confidence metrics based on information density | |
| Jalilian et al. | Stability assessment of open spans in underground entry-type excavations by focusing on data mining methods | |
| CN110622042A (en) | Analysis device, stratum generation device, analysis method, stratum generation method, and program | |
| Waite et al. | Landscape‐scale drivers of liana load across a Southeast Asian forest canopy differ to the Neotropics | |
| Babu et al. | Enhancing infrastructure safety: A uav-based approach for crack detection | |
| Wang et al. | MATLAB code to estimate landslide volume from single remote sensed image using genetic algorithm and imagery similarity measurement | |
| Hui | A Bayesian solution to the modifiable areal unit problem | |
| Alshawabkeh et al. | Automated Pavement Crack Detection Using Deep Feature Selection and Whale Optimization Algorithm. |
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 |