CN114708590A - Deep learning-based composite solidified soil microstructure identification and analysis method and system - Google Patents

Deep learning-based composite solidified soil microstructure identification and analysis method and system Download PDF

Info

Publication number
CN114708590A
CN114708590A CN202210630904.7A CN202210630904A CN114708590A CN 114708590 A CN114708590 A CN 114708590A CN 202210630904 A CN202210630904 A CN 202210630904A CN 114708590 A CN114708590 A CN 114708590A
Authority
CN
China
Prior art keywords
microstructure
network model
neural network
solidified soil
composite solidified
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
Application number
CN202210630904.7A
Other languages
Chinese (zh)
Inventor
李秉宜
钱彬
陈永辉
沈峰
蒋明镜
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources
Original Assignee
Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources filed Critical Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources
Priority to CN202210630904.7A priority Critical patent/CN114708590A/en
Publication of CN114708590A publication Critical patent/CN114708590A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06F—ELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00—Pattern recognition
    • G06F18/20—Analysing
    • G06F18/24—Classification techniques
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G06N3/045—Combinations of networks
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02W—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO WASTEWATER TREATMENT OR WASTE MANAGEMENT
    • Y02W30/00—Technologies for solid waste management
    • Y02W30/50—Reuse, recycling or recovery technologies
    • Y02W30/91—Use of waste materials as fillers for mortars or concrete

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Artificial Intelligence (AREA)
  • General Physics & Mathematics (AREA)
  • General Engineering & Computer Science (AREA)
  • Computing Systems (AREA)
  • Software Systems (AREA)
  • Molecular Biology (AREA)
  • Computational Linguistics (AREA)
  • Biophysics (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • General Health & Medical Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Evolutionary Biology (AREA)
  • Image Analysis (AREA)
  • Analysing Materials By The Use Of Radiation (AREA)

Abstract

The invention relates to a composite solidified soil microstructure identification and analysis method based on deep learning, which comprises the steps of S1, obtaining microstructure images of a plurality of composite solidified soil samples, and labeling pores and hydration products in each image to form an image data set; s2, constructing a neural network model, setting the initial learning rate of the neural network model, and calculating the loss value of the neural network model; changing the learning rate to realize iterative training of the neural network model until loss values are converged or unchanged after multiple iterations to obtain an optimal neural network model; s3, identifying pores and hydration products in the microstructure image by using the optimal neural network model, and generating a plurality of identification frames on the image; s4, clustering and segmenting the images in the identification frame, extracting pixel areas of pores and hydration products, and calculating the amount of the pores and the hydration products; s5 obtaining the hydration degree and cementation degree of the composite solidified soil based on the pore space and the amount of each hydration product.

Description

Deep learning-based composite solidified soil microstructure identification and analysis method and system
Technical Field
The invention relates to the technical field of composite solidified soil application, in particular to a method and a system for identifying and analyzing a microstructure of composite solidified soil based on deep learning.
Background
With the increase of the demand of infrastructure construction in China, the discharge amount of waste soil of buildings keeps high-level increase, and at present, an economical and effective method is to utilize industrial waste residues to be cooperated with materials such as cement, lime and the like to compound and solidify the waste residue soil to form industrial waste residue compound solidified soil which can be utilized in engineering.
The indexes of the composite solidified soil which are concerned in the application process are strength, compression coefficient and the like, but the strength increase and various performance evolutions of the composite solidified soil in the external environment are determined by the internal structure, wherein various complex reactions are involved. Therefore, a rapid and accurate quantitative analysis method for the microstructure is needed to analyze the evolution process of the microstructure of the composite solidified soil more accurately, so as to establish the relationship between the change of the microstructure and the macroscopic characteristics such as strength and the like.
At present, the method simply utilizes microstructure images of the composite solidified soil obtained by tests such as a Scanning Electron Microscope (SEM) test and the like to manually label pores, calcium silicate hydrate, ettringite and the like in the image soil, has low efficiency and questionable accuracy, and cannot quantitatively analyze the microstructure of the composite solidified soil.
Disclosure of Invention
The invention aims to solve the technical problem of providing a method and a system for identifying and analyzing a microstructure of composite solidified soil based on deep learning, which can identify and segment pores and various hydration products in a microstructure image of the composite solidified soil by using a trained neural network model, calculate the quantity of each structure by using the model, and further analyze the hydration degree and the cementation degree of the composite solidified soil; the microstructure of the composite soil can be quantitatively measured and analyzed, and the subsequent connection between the change of the microstructure and the macroscopic characteristics such as strength and the like is conveniently established.
In order to solve the technical problems, the invention provides a composite solidified soil microstructure identification and analysis method based on deep learning, which comprises the following steps: s1, acquiring microstructure images of a plurality of composite solidified soil samples, and labeling the pores and hydration products in each image to form an image data set; s2, constructing a neural network model, setting an initial learning rate of the neural network model, and calculating a loss value of the neural network model based on the image data set; changing the learning rate to realize iterative training of the neural network model until loss values are converged or unchanged after multiple iterations, thereby obtaining an optimal neural network model; s3, identifying pores and hydration products in the microstructure image by using the optimal neural network model, and generating a plurality of identification frames on the microstructure image; s4, clustering and segmenting the images in the identification frame, extracting pixel areas of pores and hydration products, and calculating to obtain the amount of the pores and the hydration products; and S5, obtaining the hydration degree and the cementation degree of the composite solidified soil based on the pore space in the microstructure of the composite solidified soil and the amount of each hydration product.
Preferably, the method for calculating the loss value of the neural network model comprises: obtaining a loss function of the neural network model, the loss function comprising:
loss of coordinates:
Figure 480360DEST_PATH_IMAGE001
；
frame loss:
Figure 353639DEST_PATH_IMAGE002
；
classification loss:
Figure 468225DEST_PATH_IMAGE003
；
confidence loss:
Figure 390045DEST_PATH_IMAGE004
；
wherein the content of the first and second substances,
Figure 906477DEST_PATH_IMAGE005
as a contributing parameter, for the coordination error; s is the number of grids, B represents that B anchor frames are generated by each grid;
Figure 153918DEST_PATH_IMAGE006
whether the jth anchor frame in the ith grid frames to the object or not is represented, if yes, the jth anchor frame is 1, and if not, the jth anchor frame is 0;
Figure 365588DEST_PATH_IMAGE007
is 0;
Figure 153415DEST_PATH_IMAGE008
the coordinates of the center point of the object detected for the ith mesh,
Figure 524354DEST_PATH_IMAGE009
、
Figure 677118DEST_PATH_IMAGE010
representing the coordinates of the center point of the corresponding actual object;
Figure 969559DEST_PATH_IMAGE011
and
Figure 92236DEST_PATH_IMAGE012
respectively representing the width and height of the prediction box;
Figure 458626DEST_PATH_IMAGE013
represents the width of the mark frame;
Figure 313450DEST_PATH_IMAGE014
represents the high of the marker box;
Figure 624345DEST_PATH_IMAGE015
respectively representing the probability that the prediction box belongs to the class c and the probability that the mark box belongs to the class c;
Figure 753975DEST_PATH_IMAGE016
respectively representing the probability value and the true value of the target object,
Figure 974872DEST_PATH_IMAGE017
and
Figure 594072DEST_PATH_IMAGE018
are all parameter values; and calculating loss values of the neural network model under different learning rates based on the loss function.
Preferably, in S1, after the microstructure images of the multiple composite solidified soil samples are acquired, the images are enhanced by using inversion, rotation, whitening and gaussian noise strategies.
Preferably, the microstructure image is obtained by a scanning electron microscope test.
Preferably, the neural network model is iteratively trained using the Keras and tensoroflow deep learning framework.
Preferably, the neural network model is an R-CNN network model, a U-net network model or a YOLO network model.
Compound solidification soil microstructure identification analysis system based on degree of depth study, its characterized in that includes: the image data set acquisition module is used for acquiring microstructure images of a plurality of composite solidified soil samples and carrying out frame selection and marking on pores and hydration products in each image to form an image data set; the optimal neural network model forming module is used for presetting a neural network model, setting the initial learning rate of the neural network model, and calculating the loss value of the neural network model based on the image data set; changing the learning rate to realize iterative training of the neural network model until loss values are converged or unchanged after multiple iterations, thereby obtaining an optimal neural network model; the microstructure identification module is used for identifying pores and hydration products in the microstructure image and generating a plurality of identification frames on the microstructure image; the microstructure quantitative characterization module is used for clustering and segmenting the image in the identification frame to extract pixel regions of pores and hydration products so as to obtain the amount of the pores and the hydration products; and obtaining the hydration degree and the cementation degree of the composite solidified soil based on the amount of the pores and the hydration products in the microstructure of the composite solidified soil.
Preferably, the composite solidified soil sample is subjected to multi-magnification image acquisition to generate a microstructure image.
Preferably, the hydration products include calcium silicate and ettringite.
A computer readable storage medium having stored therein instructions that, when executed by a processor, perform the method for deep learning-based composite solidified-soil microstructure identification analysis.
Compared with the prior art, the technical scheme of the invention has the following advantages:
according to the invention, the neural network model is arranged to rapidly segment and identify the pores and hydration products in the microstructure image, iterative training can be further carried out on the neural network model until the optimal neural network model is obtained, and the pores and the hydration products of the soil in the image can be accurately segmented and identified through the optimal neural network model, so that the amount of the pores and the hydration products is calculated, and the purpose of quantitatively representing the microstructure of the composite solidified soil is achieved. Accurate identification, high working efficiency and strong generalization capability.
Drawings
In order that the present disclosure may be more readily and clearly understood, reference is now made to the following detailed description of the present disclosure taken in conjunction with the accompanying drawings, in which:
FIG. 1 is a schematic flow diagram of the present invention;
FIG. 2 is a schematic diagram of the deep learning-based identification and analysis of the microstructure of the composite solidified soil according to the present invention;
FIG. 3 is an example of an image dataset of a composite solidified soil microstructure of the present invention;
FIG. 4 is a diagram illustrating loss values of a neural network model according to an embodiment of the present invention;
FIG. 5 is a schematic view of microstructure recognition of solidified soil according to the present invention.
Detailed Description
The present invention is further described below in conjunction with the following figures and specific examples so that those skilled in the art may better understand the present invention and practice it, but the examples are not intended to limit the present invention.
The invention discloses a composite solidified soil microstructure identification and analysis method based on deep learning, which is shown in figure 1 and comprises the following steps:
acquiring microstructure images of a large number of composite solidified soil samples, and labeling the pores and hydration products in each image to form an image data set.
In one preferred embodiment of the present invention, an example of an image dataset of a composite solidified soil microstructure is shown with reference to fig. 3.
The microstructure image can be obtained by methods including, but not limited to, scanning electron microscopy.
Among them, the hydration products include, but are not limited to, calcium silicate hydrate (C-S-H), ettringite (AFt), and the like. After microstructure images of a plurality of composite solidified soil samples are obtained, images are subjected to enhancement processing by using overturning, rotating, whitening and Gaussian noise strategies, so that an image data set is stronger.
Further, the image data set is divided into a training set, a verification set and a test set according to a preset proportion.
And step two, constructing a neural network model, and adopting various common deep convolution neural network models to identify the microstructure of the composite solidified soil and improve and utilize the microstructure. The model to be selected is R-CNN, U-net or YOLO, wherein R-CNN is two-stage target recognition, and YOLO is single-stage target implementation.
Taking the YOLO model as an example, the Darknet-53 network structure composed of the convolutional network incorporating the active layers and the multilayer residual blocks extracts the characteristics of the porosity and hydration products of the composite solidified soil from the input image dataset. The YOLO model forms a multi-branch network and a Concat layer through a route structure, simultaneously uses a Biliner Upsampling layer to expand characteristics to form 3 branch networks for positioning three target areas with different scales, and obtains a characteristic matrix of a composite solidified soil microstructure image through the branch networks. After the feature matrix is obtained, the feature matrix is input into a detection network to obtain a positioning matrix, the YOLO model outputs three positioning matrices with different dimensions, and the three positioning matrices can preferably respectively correspond to a 13 × 13 grid, a 26 × 26 grid and a 52 × 52 grid for positioning a target area.
Setting an initial learning rate of the neural network model, calculating a loss value of the neural network model based on the training set and the verification set data, and changing the learning rate to realize iterative training of the neural network model until the loss value is converged or unchanged after multiple iterations, thereby obtaining the optimal neural network model.
The loss function includes:
loss of coordinates:
Figure 267630DEST_PATH_IMAGE001
；
frame loss:
Figure 200951DEST_PATH_IMAGE019
；
classification loss:
Figure 400988DEST_PATH_IMAGE003
；
confidence loss:
Figure 800877DEST_PATH_IMAGE004
；
wherein the content of the first and second substances,
Figure 351944DEST_PATH_IMAGE005
as a contributing parameter, for the coordination error; s is the number of grids, B represents that B anchor frames are generated by each grid;
Figure 292218DEST_PATH_IMAGE006
whether the jth anchor frame in the ith grid frames to the object or not is represented, if yes, the jth anchor frame is 1, and if not, the jth anchor frame is 0;
Figure 550024DEST_PATH_IMAGE007
is 0;
Figure 120814DEST_PATH_IMAGE008
the coordinates of the center point of the object detected for the ith mesh,
Figure 159177DEST_PATH_IMAGE009
、
Figure 106404DEST_PATH_IMAGE010
representing the coordinates of the center point of the corresponding actual object;
Figure 218717DEST_PATH_IMAGE011
and
Figure 616200DEST_PATH_IMAGE012
respectively representing the width and height of the prediction box;
Figure 813963DEST_PATH_IMAGE013
represents the width of the mark frame;
Figure 564882DEST_PATH_IMAGE014
represents the height of the marker box;
Figure 328438DEST_PATH_IMAGE015
respectively representing the probability that the prediction box belongs to the class c and the probability that the mark box belongs to the class c;
Figure 772189DEST_PATH_IMAGE016
respectively representing the probability value and the true value of the target object,
Figure 457248DEST_PATH_IMAGE017
and
Figure 136491DEST_PATH_IMAGE018
are all parameter values.
The training environment of the neural network model is as follows: ubuntu 20.04 operating system, Python 3.7 compiled language, Keras 2.2.4 framework and tensrflow 1.18 deep learning framework.
Step three, target detection: and identifying the pores and the hydration products in the microstructure image by using the optimal neural network model, and generating a plurality of identification frames on the microstructure image.
Referring to FIG. 2, the identified pores and hydrates are identified in the box. In practical applications, the recognition boxes representing different objects can be set to different colors, such as a green box recognizing pores, a red box recognizing calcium silicate hydrate (C-S-H), and other colors recognizing other objects.
And step four, clustering and segmenting the image center points in the identification frame, and extracting pixel regions of the pores and the hydration products by matching with a plurality of image processing methods, so as to calculate the amount of the pores and the hydration products.
And step five, obtaining the hydration degree and the cementation degree of the composite solidified soil based on the pores in the composite solidified soil microstructure and the quantity of each hydration product, quantitatively characterizing and analyzing the composite solidified soil microstructure, and facilitating the subsequent establishment of the relation between the change of the microstructure and the macroscopic characteristics such as strength. The same composite solidified soil sample needs to be determined by carrying out batch quantitative analysis and averaging on a plurality of images.
In one preferred embodiment, the initial learning rate of the neural network model is set to 0.0001, with a sample size of training (batch size) of 16 for 100 rounds (epoch). Referring to FIG. 4, the losses in the training set and validation set begin to converge after 60 rounds of training.
Referring to fig. 5, as a recognition result, a recognized target object and a confidence (probability of being considered as the target object) are identified in the recognition frame. The confidence coefficient threshold value preset for the identification is 0.5, and the confidence coefficient greater than 0.5 is displayed. According to the actual situation, the accuracy of the model can be improved by increasing the data set and the optimization algorithm in the later period, and other microscopic means are matched for verification.
Step six, further improving the composite solidified soil microstructure identification and analysis method based on deep learning:
the improvement direction mainly comprises: more scales of identification, feature maps with larger sizes (such as 104 × 104) are added according to requirements to identify smaller targets, and the identification capability is improved. In the case where some category data is relatively lacking, a larger amount of data is obtained using the migration learning, and the like.
The method can identify and segment pores and various hydration products in the microstructure image of the composite solidified soil by using the trained neural network model, calculate the quantity of each structure by using the model, and further analyze the hydration degree and the cementation degree of the composite solidified soil; the microstructure of the composite soil can be quantitatively measured and analyzed, and the subsequent connection between the change of the microstructure and the macroscopic characteristics such as strength and the like is conveniently established.
Further, the invention also provides a composite solidified soil microstructure identification and analysis system based on deep learning, which comprises:
and the image data set acquisition module is used for acquiring microstructure images of a plurality of composite solidified soil samples and carrying out frame selection and labeling on the pores and the hydration products in each image to form an image data set.
Preferably, the composite solidified soil sample is subjected to multi-magnification image acquisition to generate a microstructure image.
And the optimal neural network model forming module is used for presetting the neural network model, setting the initial learning rate of the neural network model and calculating the loss value of the neural network model based on the image data set. Changing the learning rate to realize iterative training of the neural network model until loss values are converged or unchanged after multiple iterations, thereby obtaining an optimal neural network model;
and the microstructure identification module is used for identifying the pores and the hydration products in the microstructure image and generating a plurality of identification frames on the microstructure image.
And the microstructure quantitative characterization module is used for clustering and segmenting the image in the identification frame so as to extract pixel regions of the pores and the hydration products, thereby obtaining the amount of the pores and the hydration products. And obtaining the hydration degree and the cementation degree of the composite solidified soil based on the amount of the pores and the hydration products in the microstructure of the composite solidified soil. The change of the microstructure is convenient to be subsequently linked with macroscopic characteristics such as strength and the like.
The invention also discloses a computer readable storage medium, wherein instructions are stored, and when the instructions are executed by a processor, the method for identifying and analyzing the microstructure of the composite solidified soil based on deep learning is executed.
As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, system, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application 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 so forth) having computer-usable program code embodied therein.
The present application is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the application. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams 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.
It should be understood that the above examples are only for clarity of illustration and are not intended to limit the embodiments. Other variations and modifications will be apparent to persons skilled in the art in light of the above description. And are neither required nor exhaustive of all embodiments. And obvious variations or modifications of the invention may be made without departing from the spirit or scope of the invention.

Claims (10)

1. The method for identifying and analyzing the microstructure of the composite solidified soil based on deep learning is characterized by comprising the following steps of:
s1, acquiring microstructure images of a plurality of composite solidified soil samples, and labeling the pores and hydration products in each image to form an image data set;
s2, constructing a neural network model, setting an initial learning rate of the neural network model, and calculating a loss value of the neural network model based on the image data set;
changing the learning rate to realize iterative training of the neural network model until loss values are converged or unchanged after multiple iterations, thereby obtaining an optimal neural network model;
s3, identifying pores and hydration products in the microstructure image by using the optimal neural network model, and generating a plurality of identification frames on the microstructure image;
s4, clustering and segmenting the images in the identification frame, extracting pixel areas of pores and hydration products, and calculating to obtain the amount of the pores and the hydration products;
and S5, obtaining the hydration degree and the cementation degree of the composite solidified soil based on the pore space in the microstructure of the composite solidified soil and the amount of each hydration product.
2. The method for identifying and analyzing the microstructure of the composite solidified soil based on deep learning of claim 1, wherein the method for calculating the loss value of the neural network model comprises the following steps:
obtaining a loss function of the neural network model, the loss function comprising:
loss of coordinates:
Figure 452949DEST_PATH_IMAGE001
；
frame loss:
Figure 67601DEST_PATH_IMAGE002
；
classification loss:
Figure 36694DEST_PATH_IMAGE003
；
confidence loss:
Figure 457311DEST_PATH_IMAGE004
；
wherein the content of the first and second substances,
Figure 336406DEST_PATH_IMAGE005
as a contributing parameter, for the coordination error; s is the number of grids, B represents that B anchor frames are generated by each grid;
Figure 184276DEST_PATH_IMAGE006
whether the jth anchor frame in the ith grid frames to the object or not is represented, if yes, the jth anchor frame is 1, and if not, the jth anchor frame is 0;
Figure 906244DEST_PATH_IMAGE007
is 0;
Figure 864973DEST_PATH_IMAGE008
the coordinates of the center point of the object detected for the ith mesh,
Figure 598574DEST_PATH_IMAGE009
、
Figure 414083DEST_PATH_IMAGE010
representing the coordinates of the center point of the corresponding actual object;
Figure 561031DEST_PATH_IMAGE011
and
Figure 261133DEST_PATH_IMAGE012
respectively representing the width and height of the prediction box;
Figure 177137DEST_PATH_IMAGE013
represents the width of the mark frame;
Figure 429127DEST_PATH_IMAGE014
represents the height of the marker box;
Figure 469895DEST_PATH_IMAGE015
respectively representing the probability that the prediction box belongs to the class c and the probability that the mark box belongs to the class c;
Figure 770426DEST_PATH_IMAGE016
respectively representing the probability value and the true value of the target object,
Figure 868832DEST_PATH_IMAGE017
and
Figure 229406DEST_PATH_IMAGE018
are all parameter values;
and calculating loss values of the neural network model under different learning rates based on the loss function.
3. The method for identifying and analyzing the microstructure of the composite solidified soil based on the deep learning of claim 1, wherein in the step S1, after the microstructure images of a plurality of composite solidified soil samples are obtained, the images are enhanced by using inversion, rotation, whitening and gaussian noise strategies.
4. The deep learning-based composite solidified soil microstructure identification and analysis method as claimed in claim 1, wherein the microstructure image acquisition method comprises a scanning electron microscope test.
5. The deep learning-based composite solidified soil microstructure recognition analysis method as claimed in claim 1, wherein the neural network model is iteratively trained using Keras and Tensorflow deep learning frameworks.
6. The deep learning-based composite solidified soil microstructure recognition analysis method as claimed in claim 1, wherein the neural network model is an R-CNN network model, a U-net network model or a YOLO network model.
7. Compound solidification soil microstructure identification analysis system based on degree of depth study, its characterized in that includes:
the image data set acquisition module is used for acquiring microstructure images of a plurality of composite solidified soil samples and performing frame selection and marking on pores and hydration products in each image to form an image data set;
the optimal neural network model forming module is used for presetting a neural network model, setting the initial learning rate of the neural network model, and calculating the loss value of the neural network model based on the image data set; changing the learning rate to realize iterative training of the neural network model until loss values are converged or unchanged after multiple iterations, thereby obtaining an optimal neural network model;
the microstructure identification module is used for identifying pores and hydration products in the microstructure image and generating a plurality of identification frames on the microstructure image;
the microstructure quantitative characterization module is used for clustering and segmenting the image in the identification frame to extract pixel regions of pores and hydration products so as to obtain the amount of the pores and the hydration products; and obtaining the hydration degree and the cementation degree of the composite solidified soil based on the amount of the pores and the hydration products in the microstructure of the composite solidified soil.
8. The deep learning based composite solidified soil microstructure identification and analysis system of claim 7, wherein the composite solidified soil sample is subjected to multi-magnification image acquisition to generate a microstructure image.
9. The deep learning-based composite solidified-soil microstructure recognition analysis system of claim 7, wherein the hydration products comprise calcium silicate and ettringite.
10. A computer readable storage medium having stored therein instructions which, when executed by a processor, perform the method for deep learning-based composite solidified-soil microstructure identification analysis according to any one of claims 1-6.
CN202210630904.7A 2022-06-06 2022-06-06 Deep learning-based composite solidified soil microstructure identification and analysis method and system Pending CN114708590A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210630904.7A CN114708590A (en) 2022-06-06 2022-06-06 Deep learning-based composite solidified soil microstructure identification and analysis method and system

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210630904.7A CN114708590A (en) 2022-06-06 2022-06-06 Deep learning-based composite solidified soil microstructure identification and analysis method and system

Publications (1)

Publication Number Publication Date
CN114708590A true CN114708590A (en) 2022-07-05

Family

ID=82177928

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210630904.7A Pending CN114708590A (en) 2022-06-06 2022-06-06 Deep learning-based composite solidified soil microstructure identification and analysis method and system

Country Status (1)

Country Link
CN (1) CN114708590A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115147399A (en) * 2022-07-29 2022-10-04 济南大学 Method and system for predicting element content of cement microscopic image pixel points

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113920443A (en) * 2021-09-29 2022-01-11 海南大学 Yoov 5-based remote sensing directed target detection method
CN114298145A (en) * 2021-11-22 2022-04-08 三峡大学 Permeable concrete pore intelligent identification and segmentation method based on deep learning

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113920443A (en) * 2021-09-29 2022-01-11 海南大学 Yoov 5-based remote sensing directed target detection method
CN114298145A (en) * 2021-11-22 2022-04-08 三峡大学 Permeable concrete pore intelligent identification and segmentation method based on deep learning

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115147399A (en) * 2022-07-29 2022-10-04 济南大学 Method and system for predicting element content of cement microscopic image pixel points

Similar Documents

Publication Publication Date Title
CN108961235B (en) Defective insulator identification method based on YOLOv3 network and particle filter algorithm
CN110705457B (en) Remote sensing image building change detection method
CN108647585B (en) Traffic identifier detection method based on multi-scale circulation attention network
CN108846835B (en) Image change detection method based on depth separable convolutional network
CN109977997B (en) Image target detection and segmentation method based on convolutional neural network rapid robustness
CN110084817B (en) Digital elevation model production method based on deep learning
CN111402227B (en) Bridge crack detection method
CN110991444B (en) License plate recognition method and device for complex scene
Li et al. Automatic bridge crack identification from concrete surface using ResNeXt with postprocessing
CN111429418A (en) Industrial part detection method based on YO L O v3 neural network
Bhugra et al. Deep convolutional neural networks based framework for estimation of stomata density and structure from microscopic images
CN109145836A (en) Ship target video detection method based on deep learning network and Kalman filtering
CN116843999B (en) Gas cylinder detection method in fire operation based on deep learning
CN114781514A (en) Floater target detection method and system integrating attention mechanism
CN114742820A (en) Bolt looseness detection method and system based on deep learning and storage medium
CN113762396A (en) Two-dimensional image semantic segmentation method
CN112836614A (en) High-resolution remote sensing image classification method based on residual error network and transfer learning
CN111879972A (en) Workpiece surface defect detection method and system based on SSD network model
CN114708590A (en) Deep learning-based composite solidified soil microstructure identification and analysis method and system
CN114581789A (en) Hyperspectral image classification method and system
CN117853722A (en) Steel metallographic structure segmentation method integrating superpixel information
CN116385425A (en) YOLOv5 fabric defect detection method for improving CA attention mechanism
CN114639013A (en) Remote sensing image airplane target detection and identification method based on improved Orient RCNN model
CN113901947A (en) Intelligent identification method for tire surface flaws under small sample
JP2007139421A (en) Morphological classification device and method, and variation region extraction device and method

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination