CN114898343A - Soil compaction degree detection method, system and medium based on image recognition processing - Google Patents

Soil compaction degree detection method, system and medium based on image recognition processing Download PDF

Info

Publication number
CN114898343A
CN114898343A CN202210417102.8A CN202210417102A CN114898343A CN 114898343 A CN114898343 A CN 114898343A CN 202210417102 A CN202210417102 A CN 202210417102A CN 114898343 A CN114898343 A CN 114898343A
Authority
CN
China
Prior art keywords
image
layer
soil
sampling
neural 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
Application number
CN202210417102.8A
Other languages
Chinese (zh)
Inventor
李秉宜
钱彬
占鑫杰
唐译
陆勇
沈玉为
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.)
Suzhou University of Science and Technology
Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources
Original Assignee
Suzhou University of Science and Technology
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 Suzhou University of Science and Technology, Nanjing Hydraulic Research Institute of National Energy Administration Ministry of Transport Ministry of Water Resources filed Critical Suzhou University of Science and Technology
Priority to CN202210417102.8A priority Critical patent/CN114898343A/en
Publication of CN114898343A publication Critical patent/CN114898343A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/60Type of objects
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/84Systems specially adapted for particular applications
    • GPHYSICS
    • G01MEASURING; TESTING
    • G01NINVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N9/00Investigating density or specific gravity of materials; Analysing materials by determining density or specific gravity
    • G01N9/24Investigating density or specific gravity of materials; Analysing materials by determining density or specific gravity by observing the transmission of wave or particle radiation through the material
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/04Architecture, e.g. interconnection topology
    • G06N3/045Combinations of networks
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06NCOMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00Computing arrangements based on biological models
    • G06N3/02Neural networks
    • G06N3/08Learning methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/40Extraction of image or video features
    • G06V10/44Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00Arrangements for image or video recognition or understanding
    • G06V10/70Arrangements for image or video recognition or understanding using pattern recognition or machine learning
    • G06V10/82Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Computing Systems (AREA)
  • Artificial Intelligence (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Chemical & Material Sciences (AREA)
  • Biophysics (AREA)
  • Mathematical Physics (AREA)
  • Molecular Biology (AREA)
  • Biomedical Technology (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Pathology (AREA)
  • Immunology (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Medical Informatics (AREA)
  • Databases & Information Systems (AREA)
  • Image Analysis (AREA)

Abstract

The invention relates to a soil compaction degree detection method based on image recognition processing, which comprises the steps of S1 obtaining an image of a soil sample, and carrying out interception, gray scale calibration and offset field removal processing on the image; extracting local information of the processed image to obtain a local sample image; s2, constructing a convolution neural network, wherein a convolution layer performs convolution operation on the local sample image to map the local sample image into a plurality of feature map images; the sampling layer performs down-sampling operation on the feature mapping image to generate a corresponding shrinkage rate graph; the full connection layer combines the shrinkage rate maps to form global information; the output layer calculates the global information to obtain a calculation result; s3 error calculation, and adjusting the weight of the convolution neural network. The method is based on an image recognition processing technology, and can accurately measure the density, the water content and the compaction degree value of the soil body in time by obtaining the input of a local sample image; the operation is more quick, simple and convenient, and it is higher to measure the precision of result, alleviates manpower and material resources.

Description

Soil compaction degree detection method, system and medium based on image recognition processing
Technical Field
The invention relates to the field of soil body detection, in particular to a soil body compaction degree detection method, a soil body compaction degree detection system and a soil body compaction degree detection medium based on image recognition processing.
Background
In the field of civil infrastructure, the artificial intelligence technology is deeply integrated with the whole life cycle of planning, designing, building and maintaining of the civil infrastructure, and the development of civil engineering is deeply changed.
Soil body compactness indexes in the field of geotechnical engineering can reflect soil body compression degree, and engineering quality is controlled by controlling foundation soil compactness indexes in municipal, water conservancy and port engineering at present, for example, JTS237-2017 detection technical rules of foundation tests of water transport engineering for in-situ density and compactness of soil bodies of medium foundations, the detection methods mainly comprise a cutting ring method, a sand filling method, a water filling method and a nuclear ray method.
However, the ring knife method, the sand-irrigation method, the water-irrigation method and the nuclear ray method all have certain defects: (1) the ring cutter method soil body compactness detects the cycle length, and the test result easily receives experimenter's influence, because different experimenters have different experimental operation habits and different operation proficiency, takes back a soil sample every time and returns the laboratory and probably have different experimenters operation, probably have certain error nature to the result. (2) The sand irrigation method and the water irrigation method need to solve density tests on site, sand (water) filling and sand (water) emptying operations need to be repeated in the test process, the test operation is relatively complex, and the test result greatly depends on the proficiency of testers. (3) The nuclear ray method instrument is a nuclear density instrument, the price of the instrument is expensive, and the instrument has certain radioactivity.
The common prior art, such as publication number CN113326659A, discloses a method for rapidly detecting the compaction degree of red clay, which utilizes PFWD equipment to lift the equipment hammer to a certain height for free fall, obtains a series of data related to the dynamic stress response of the soil body in the impact process, and calculates the compaction degree information through the established relationship between deflection or dynamic modulus and compaction degree. The PFWD equipment needs to be additionally arranged, the test operation is relatively complex, and the detection working efficiency is low.
Disclosure of Invention
The invention aims to provide a method, a system and a medium for quickly detecting soil compaction indexes, which can be used for effectively and accurately measuring the actual density, water content and compaction parameter results of on-site filling earthwork by combining an artificial intelligent convolutional neural network technology and acquiring the input of a local sample image based on an image recognition processing technology on the basis of a traditional geotechnical engineering compaction detection method.
Compared with the traditional compaction degree detection method, the method has the advantages that the operation is faster, simpler and more convenient, the measured result precision is more accurate, the error influence of manual operation of different testers is solved, and the human resource and material resource cost is greatly reduced.
In order to solve the technical problems, the invention provides a method, a system and a medium for quickly detecting soil compaction indexes, which comprise the following steps: s1, acquiring an image of the soil sample, and intercepting, calibrating gray scale and removing an offset field of the image; extracting local information of the processed image to obtain a local sample image; s2, constructing a convolutional neural network, which comprises a convolutional layer, a sampling layer, a full connection layer and an output layer; the convolution layer convolves the local sample images such that the local sample images are mapped into a plurality of feature map images; the sampling layer performs down-sampling operation on the feature mapping image to generate a shrinkage ratio map corresponding to the feature mapping image; the full connection layer combines the shrinkage rate maps of the feature map images to form global information; the output layer calculates the global information to obtain a calculation result, wherein the calculation result comprises the water content, the density and the compaction degree of the soil sample; s3, comparing the calculation result with the actual water content, density and compactness of the soil sample to obtain a relative error; if the relative error is within a preset range, outputting the calculation result; and if the relative error is not in the preset range, adjusting the weight of the convolutional neural network, and returning to the step S2.
Preferably, the convolution formula of the convolutional layer is as follows:
Figure BDA0003606481100000021
wherein f is an activation function, i and j are subscripts of rows and columns of the feature map image, k is a convolution kernel element, b is a bias term, l is the number of input layers, and s is a weight component; the activation function is:
Figure BDA0003606481100000031
preferably, in S2, the convolutional layer performs a convolution operation on the local sample image, specifically: and carrying out convolution operation processing on the local sample image in a preset step size from left to right and from top to bottom in sequence, so that the local sample image is mapped into a plurality of feature mapping images.
Preferably, in S2, the sampling layer performs a down-sampling operation on the feature map image, specifically: the sampling layer abstractly represents the feature mapping image output by the convolutional layer so as to reduce the size of the feature mapping image; the feature map image is further mapped using a mean convolution kernel to generate a scaled map corresponding to the feature map image.
Preferably, the weight of the convolutional neural network is adjusted by using an error propagation learning algorithm; the error propagation learning algorithm comprises an error term recursion calculation formula: delta (l-1) =(W (l) ) T δ (l-1) ⊙f′(u (l-1) ) Wherein W is a weight matrix, T is a matrix transposition operation, u is a temporary variable, and f is an activation function; the activation function is:
Figure BDA0003606481100000032
preferably, the convolutional neural network comprises n convolutional layers and n sampling layers, wherein n is larger than or equal to 1.
Preferably, S1 further includes: sampling on site by using a ring cutter method to obtain a plurality of soil sample samples; and carrying out an indoor test on each soil sample in a laboratory to obtain the actual water content, density and compactness of each soil sample.
The soil compaction degree detection system based on image recognition processing preferably comprises: the image acquisition module is used for acquiring an image of a soil sample, and intercepting, gray calibration and offset field removal processing are carried out on the image; extracting local information of the processed image to obtain a local sample image; the convolutional neural network module comprises a convolutional layer, a sampling layer, a full connection layer and an output layer; the convolution layer convolves the local sample images such that the local sample images are mapped into a plurality of feature map images; the sampling layer performs down-sampling operation on the feature mapping image to generate a shrinkage ratio map corresponding to the feature mapping image; the full connection layer combines the shrinkage rate maps of the feature map images to form global information; the output layer calculates the global information to obtain a calculation result; the error calculation module is used for comparing the calculation result output by the convolutional neural network with the actual water content, density and compactness of the soil sample to obtain a relative error value; and the weight value adjusting module is used for adjusting the weight value of the convolutional neural network so that the relative error is in a preset range.
Preferably, the output layer outputs three calculation results, and the calculation results comprise the water content, the density and the compaction degree of the soil sample.
A computer readable storage medium having stored therein instructions which, when executed by a processor, perform the method for detecting soil compaction based on image recognition processing.
Compared with the prior art, the technical scheme of the invention has the following advantages:
1. the method comprises the steps of obtaining an image of a soil sample, and extracting local information of the processed image to obtain a local sample image. Local sample images are used as input files of the convolutional neural network instead of whole information, and a link of extracting characteristic values is omitted; compared with the traditional neural network, the convolutional neural network model has the advantages that the complexity of the structure and the number of weights are reduced, the efficiency is improved, and certain advantages are realized when a large-scale image is processed.
2. Compared with the traditional cutting ring method, irrigation method and sand filling method, the method provided by the invention neglects the influence of manual operation errors, is simple to operate, and better saves human resources and material resources.
3. The method can better control the construction quality, is simple to operate, can be implemented on a construction site, can obtain the density, water content and compactness parameter results by taking the local sample image as an input file and calculating through a convolutional neural network, and ensures the accuracy of the detection result while ensuring the detection working speed.
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 chart of a soil compaction degree detection method according to the present invention;
FIG. 2 is a schematic diagram of the structure of a convolutional neural network of the present invention;
FIG. 3 is a schematic diagram of an image processing grid of a soil sample from a top view;
fig. 4 is a schematic diagram of an image processing grid of a soil sample under a cross-sectional view angle.
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 soil compaction degree detection method based on image recognition processing, which is shown in figure 1 and comprises the following steps:
step one, carrying out field soil sample sampling work by using a cutting ring method to obtain a soil sample. Wherein, the number of the soil sample can be a plurality of, and the soil sample is numbered. Carrying out indoor tests on the soil samples in a laboratory to obtain result parameters of the soil samples, wherein the result parameters comprise: actual water content, density and maximum dry density parameters, summarizing result parameter information of each soil sample: actual moisture content, density and compaction.
And photographing and imaging each soil sample to obtain an image of each soil sample. Referring to fig. 3 and 4, the image is automatically cut to a predetermined size, preferably 60 × 60 mm. And carrying out gray scale calibration and offset field removal processing on each intercepted image so as to remove gray scale deviation and gray scale unevenness, and endowing result parameters of the soil sample indoor test to the corresponding soil sample image.
And extracting local information of the image subjected to interception, gray scale calibration and de-offset field processing to obtain a local sample image.
In the construction process, the soil body is rolled and compacted, and the pores of the soil body are smaller, so that in the whole soil sample image, the remote pixel relation is weaker, the local pixels are closely related, and the global information can be obtained through local information extraction. According to the invention, the local sample image is used as the input file instead of the whole image, so that the extraction link of the characteristic value is saved, and compared with the traditional compaction degree detection method, the operation is faster, simpler and more convenient.
Step two, constructing a convolutional neural network, wherein the convolutional neural network comprises the following steps: the convolutional neural network comprises a convolutional layer, a sampling layer, a full-link layer and an output layer, wherein the convolutional neural network comprises n convolutional layers and n sampling layers, and n is more than or equal to 1. The specific operation method comprises the following steps:
an input operation: and inputting the local sample image into a convolutional neural network.
And (3) convolutional layer operation: the convolution layer performs a convolution operation on the local sample image such that the local sample image is mapped into a plurality of feature map images. Specifically, the convolution kernel performs convolution processing on a local sample image according to a sequence from left to right and from top to bottom by a specific step length, each soil sample image is mapped into a plurality of different abstract images, the convolution layer is activated by utilizing a tanh activation function, a plurality of new feature mapping images are obtained after the operation processing is finished, and the plurality of feature mapping images are used as the input of the sampling layer.
Wherein, the convolution formula of the convolutional layer is as follows:
Figure BDA0003606481100000061
f is the activation function of the convolutional layer, i and j are the row and column subscripts of the feature map image, k is the convolutional kernel element, b is the bias term, l is the number of input layers, s is the weight component, and the activation function of the convolutional layer is as follows:
Figure BDA0003606481100000062
sampling layer operation: the sampling layer performs down-sampling operation on the feature map image to generate a scaling map corresponding to the feature map image. Specifically, the sampling layer further abstractly represents the feature map image output by the convolutional layer, performs down-sampling operation, reduces the size of the feature map image to reduce the number of fully-connected nerve clouds, and maps the input feature map image by using a mean value convolutional kernel to generate a scaling map corresponding to the feature map image, thereby achieving the purpose of scaling down the feature map image.
Full connection layer: which is used to combine the scaled maps of the eigen-map images output by the sampling layer to form global information.
An output layer: and the output layer calculates the global information and obtains three calculation results, wherein the calculation results comprise the water content, the density and the compaction degree of the soil sample.
The method comprises the steps of obtaining an image of a soil sample, and extracting local information of the processed image to obtain a local sample image. Local sample images are used as input files of the convolutional neural network instead of whole information, and a link of extracting characteristic values is omitted; compared with the traditional neural network, the convolutional neural network model has the advantages that the complexity of the structure and the number of weights are reduced, the efficiency is improved, and certain advantages are achieved when large-scale images are processed.
And step three, comparing the water content, the density and the compactness of the soil sample calculated by the convolutional neural network with the actual water content, the density and the compactness obtained by the indoor test to obtain a relative error value. If the relative error is in a preset range, outputting a calculation result, and if the relative error is not in the preset range, outputting: and adjusting the weight of the convolutional neural network, and returning to the step two.
Wherein, the compaction degree calculation formula of the soil sample is as follows:
Figure BDA0003606481100000071
in the formula: rho d Is dry density, omega is water content, rho is density, rho dmax Is the maximum dry density.
And adjusting the weight of the convolutional neural network by using an error propagation learning algorithm, wherein the error propagation learning algorithm comprises an error item recursion calculation formula: delta. for the preparation of a coating (l-1) =(W (l) ) T δ (l-1) ⊙f′(u (l-1) ) Wherein W is a weight matrix, T is a matrix transpose operation, u is a temporary variable, f is an activation function, and the activation function is:
Figure BDA0003606481100000072
compared with the traditional cutting ring method, irrigation method and sand filling method, the invention neglects the influence of artificial operation errors, has simple operation and better saves human resources and material resources cost.
Referring to fig. 2, in another preferred embodiment, the image of each soil sample is automatically cut to a predetermined size, which is 60 x 60 mm. And carrying out gray scale calibration and de-offset field processing on each intercepted image, and intercepting the image with the size of 30 x 30mm again.
And extracting local information of the image with the size of 30 x 30mm to obtain a local sample image.
Determining two convolutional layers and two sampling layers respectively, wherein the convolutional neural network has the structure as follows: input-convolutional layer-sampling layer-full-link layer-output layer. And determining the convolution kernel size of all convolution layers of the convolution neural network to be 3 x 3 and the size of the sampling layer to be 2 x 2 based on the convolution neural network algorithm.
The sizes of the images of each of the convolutional layer-sampling layer-convolutional layer-sampling layer are respectively 30 × 30mm, 28 × 28mm, 24 × 24mm, 20 × 20mm and 16 × 16mm, the number of feature maps of the first convolutional layer is set to be 16, and the number of feature maps of the second convolutional layer is set to be 8. The output layer outputs 3 results: density, moisture content and compaction.
The method can better control the construction quality, is simple to operate and can be implemented on a construction site. The local sample image is used as an input file, and the density, moisture content and compactness parameter results can be obtained through the calculation of the convolutional neural network, so that the accuracy of the detection result is ensured while the detection working speed is ensured.
Based on the soil compaction degree detection method based on the image recognition processing, the invention also provides a soil compaction degree detection system based on the image recognition processing, which comprises the following steps:
the image acquisition module is used for acquiring an image of a soil sample, performing interception, gray scale calibration and offset field removal processing on the image according to a preset size, and extracting local information of the processed image to obtain a local sample image.
The convolutional neural network module comprises a convolutional layer, a sampling layer, a full connection layer and an output layer.
Wherein the convolution layer performs a convolution operation on the local sample image such that the local sample image is mapped into a plurality of feature map images; the sampling layer performs down-sampling operation on the feature mapping image to generate a shrinkage rate graph corresponding to the feature mapping image; the full connection layer combines the shrinkage rate maps of the feature mapping images to form global information, and the output layer calculates the global information to obtain a calculation result.
And the error calculation module is used for comparing the calculation result output by the convolutional neural network with the actual water content, density and compactness of the soil sample to obtain a relative error value.
And the weight value adjusting module is used for adjusting the weight value of the convolutional neural network so that the relative error is in a preset range.
The invention also includes a computer readable storage medium having instructions stored therein, which when executed by a processor, perform the method for rapidly detecting soil compaction index.
Furthermore, in the filling and compacting construction of a newly-built dike, a soil sample sampling work of a cutting ring method is carried out after the soil of the first layer is compacted. Setting the volume of the cutting ring to be 100cm 3 The bottom area is 20cm 2 The height is 5cm, and the number of the obtained soil samples is 900 groups of soil samples. Numbering the soil sample one by one, numbering No. 1-900, obtaining an image of the soil sample, setting No. 1-800 soil samples as training samples, and performing weight fine adjustment by convolutional neural network training and an error discovery propagation algorithm to obtain an optimized convolutional neural network; and setting 801 and 900 numbers as test samples to test and debug the convolutional neural network.
Preferably, the relative error control index is set to be less than 2% in the convolutional neural network model.
The test is carried out, the first layer of soil layer is filled for sampling, the second layer, the third layer and the fourth layer are filled for sampling and comparing, the comparison result is shown in the following table 1, wherein, the maximum dry density is 1.66g/cm 3
TABLE 1 summary of actual parameters and calculation results of convolutional neural networks
Figure BDA0003606481100000091
As can be seen from Table 1, the relative error of each parameter is less than 2%, and the set requirements are met, so that the result obtained by the method is accurate and can be applied to the practical engineering application.
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 the like) 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 soil compaction degree detection method based on image recognition processing is characterized by comprising the following steps of:
s1, acquiring an image of the soil sample, and extracting local information of the processed image to obtain a local sample image;
s2, constructing a convolutional neural network, which comprises a convolutional layer, a sampling layer, a full connection layer and an output layer;
the convolution layer convolves the local sample images such that the local sample images are mapped into a plurality of feature map images;
wherein, the convolution formula of the convolutional layer is as follows:
Figure FDA0003606481090000011
f is the activation function of the convolutional layer, i and j are the row and column subscripts of the feature map image, k is the convolution kernel element, b is the bias term, l is the number of input layers, s is the weight component, and the activation function is:
Figure FDA0003606481090000012
the sampling layer performs down-sampling operation on the feature mapping image to generate a shrinkage ratio map corresponding to the feature mapping image;
the full connection layer combines the shrinkage rate maps of the feature map images to form global information;
the output layer calculates the global information to obtain a calculation result, wherein the calculation result comprises the water content, the density and the compaction degree of the soil sample;
s3, comparing the calculation result with the actual water content, density and compactness of the soil sample; if the relative error is within a preset range, outputting the calculation result; and if the relative error is not in the preset range, returning to the step S2, and adjusting the weight of the convolutional neural network.
2. The method of claim 1, wherein in step S1, the image of the soil sample is subjected to clipping, gray scale calibration and de-skewing field processing.
3. The method for detecting soil compaction degree based on image recognition processing according to claim 1, wherein in S2, the convolution layer performs convolution operation on the local sample image, specifically:
and carrying out convolution operation processing on the local sample image in a preset step size from left to right and from top to bottom in sequence, so that the local sample image is mapped into a plurality of feature mapping images.
4. The method for detecting soil compaction degree based on image recognition processing according to claim 1, wherein in S2, the sampling layer performs a down-sampling operation on the feature map image, specifically:
the sampling layer abstractly represents the feature mapping image output by the convolutional layer so as to reduce the size of the feature mapping image;
the feature map image is further mapped using a mean convolution kernel to generate a scaled map corresponding to the feature map image.
5. The image recognition processing-based soil compaction degree detection method according to claim 1, wherein the weight of the convolutional neural network is adjusted by using an error propagation learning algorithm;
the error propagation learning algorithm comprises an error term recursion calculation formula:
δ (l-1) =(W (l) ) T δ (l-1) ⊙f′(u (l-1) ) Wherein W is a weight matrix, u is a temporary variable, T is a matrix transposition operation, and f is an activation function of the convolutional layer;
the activation function is:
Figure FDA0003606481090000021
6. the image recognition processing-based soil compaction degree detection method according to claim 1, wherein the convolutional neural network comprises n convolutional layers and n sampling layers, wherein n is greater than or equal to 1.
7. The method of detecting soil compaction based on image recognition processing as claimed in claim 1 wherein said step S1 is preceded by the step of:
sampling on site by using a ring cutter method to obtain a plurality of soil sample samples;
and carrying out an indoor test on each soil sample in a laboratory to obtain the actual water content, density and compactness of each soil sample.
8. Soil body compactness detecting system based on image identification handles, its characterized in that includes:
the image acquisition module is used for acquiring an image of a soil sample, and intercepting, gray calibration and offset field removal processing are carried out on the image; extracting local information of the processed image to obtain a local sample image;
the convolutional neural network module comprises a convolutional layer, a sampling layer, a full connection layer and an output layer;
the convolution layer convolves the local sample images such that the local sample images are mapped into a plurality of feature map images; the sampling layer performs down-sampling operation on the feature mapping image to generate a shrinkage ratio map corresponding to the feature mapping image; the full connection layer combines the shrinkage rate maps of the feature map images to form global information; the output layer calculates the global information to obtain a calculation result;
the error calculation module is used for comparing the calculation result output by the convolutional neural network with the actual water content, density and compactness of the soil sample to obtain a relative error value;
and the weight value adjusting module is used for adjusting the weight value of the convolutional neural network so that the relative error is in a preset range.
9. The image recognition processing-based soil compaction degree detection system of claim 8, wherein the output layer outputs three calculation results, wherein the calculation results comprise water content, density and compaction degree of the soil sample.
10. A computer readable storage medium having stored therein instructions which, when executed by a processor, perform the method of soil compaction based image recognition processing of claims 1-7.
CN202210417102.8A 2022-04-20 2022-04-20 Soil compaction degree detection method, system and medium based on image recognition processing Pending CN114898343A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210417102.8A CN114898343A (en) 2022-04-20 2022-04-20 Soil compaction degree detection method, system and medium based on image recognition processing

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210417102.8A CN114898343A (en) 2022-04-20 2022-04-20 Soil compaction degree detection method, system and medium based on image recognition processing

Publications (1)

Publication Number Publication Date
CN114898343A true CN114898343A (en) 2022-08-12

Family

ID=82717313

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210417102.8A Pending CN114898343A (en) 2022-04-20 2022-04-20 Soil compaction degree detection method, system and medium based on image recognition processing

Country Status (1)

Country Link
CN (1) CN114898343A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116678783A (en) * 2023-06-05 2023-09-01 河北圣燕检验检测服务有限公司 Weighing end data automatic acquisition method, acquisition device, equipment and medium

Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007010568A (en) * 2005-07-01 2007-01-18 Hitachi Constr Mach Co Ltd Method for measuring soil dry density and method and apparatus for determining degree of soil compaction
CN108896296A (en) * 2018-04-18 2018-11-27 北京信息科技大学 A kind of wind turbine gearbox method for diagnosing faults based on convolutional neural networks
CN109468918A (en) * 2018-10-31 2019-03-15 北京龙马智行科技有限公司 A kind of subgrade and pavement intelligence compacting decision system
JP2020041290A (en) * 2018-09-07 2020-03-19 学校法人金沢工業大学 Optimal compaction determination construction system for concrete
CN111259602A (en) * 2020-01-16 2020-06-09 长沙理工大学 Rebound modulus determination method of BP neural network based on multi-population genetic algorithm optimization
US20200242422A1 (en) * 2019-01-29 2020-07-30 Boe Technology Group Co., Ltd. Method and electronic device for retrieving an image and computer readable storage medium
CN113326659A (en) * 2021-06-16 2021-08-31 长沙理工大学 Rapid detection method for red clay compactness

Patent Citations (7)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JP2007010568A (en) * 2005-07-01 2007-01-18 Hitachi Constr Mach Co Ltd Method for measuring soil dry density and method and apparatus for determining degree of soil compaction
CN108896296A (en) * 2018-04-18 2018-11-27 北京信息科技大学 A kind of wind turbine gearbox method for diagnosing faults based on convolutional neural networks
JP2020041290A (en) * 2018-09-07 2020-03-19 学校法人金沢工業大学 Optimal compaction determination construction system for concrete
CN109468918A (en) * 2018-10-31 2019-03-15 北京龙马智行科技有限公司 A kind of subgrade and pavement intelligence compacting decision system
US20200242422A1 (en) * 2019-01-29 2020-07-30 Boe Technology Group Co., Ltd. Method and electronic device for retrieving an image and computer readable storage medium
CN111259602A (en) * 2020-01-16 2020-06-09 长沙理工大学 Rebound modulus determination method of BP neural network based on multi-population genetic algorithm optimization
CN113326659A (en) * 2021-06-16 2021-08-31 长沙理工大学 Rapid detection method for red clay compactness

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
KATALIN SZILAGYI等: "Effect of compaction, curing and surface moisture content on the rebound hardness of concrete", 《CONCRETE STRUCTURES》, 31 January 2014 (2014-01-31), pages 27 - 34 *
王佳俊: "心墙堆石坝坝面碾压智能监控方法与应用研究", 《中国博士学位论文全文数据库工程科技Ⅱ辑》, no. 06, 15 June 2021 (2021-06-15), pages 037 - 8 *

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116678783A (en) * 2023-06-05 2023-09-01 河北圣燕检验检测服务有限公司 Weighing end data automatic acquisition method, acquisition device, equipment and medium

Similar Documents

Publication Publication Date Title
Shahzad et al. TecDEM: A MATLAB based toolbox for tectonic geomorphology, Part 2: Surface dynamics and basin analysis
CN102855806B (en) System and method for quickly building physical model
CN102081030B (en) Geotechnical model test system based on macro-micro mechanics as well as elaboration test method
CN105678757B (en) A kind of ohject displacement measuring method
CN104809362B (en) It is a kind of to include heterogeneous deformation speckle pattern preparation method based on numerical computations
CN110415265A (en) Terraced fields extraction method based on unmanned plane high accuracy DEM data
Wang et al. Morphological approach to quantifying soil cracks: Application to dynamic crack patterns during wetting‐drying cycles
CN108414343A (en) A method of measuring sample rubber-oil-pipe sensor amount in coarse-grained soil Large-scale Triaxial Experiments
CN114898343A (en) Soil compaction degree detection method, system and medium based on image recognition processing
Baily et al. Comparative assessment of analytical and digital photogrammetric methods in the construction of DEMs of geomorphological forms
CN104165892B (en) Soil Crack Parameters on-line measurement system and use this system to realize soil Crack Parameters extracting method
CN106018238A (en) Method for determining pore structure of soil profile
CN117763926A (en) digital twin information driven high-reliability structure deformation monitoring method
Merel et al. The monitoring of soil surface development using analytical photogrammetry
CN107452059A (en) A kind of three-dimensional interpolation modeling method, equipment and its storage device
CN110470275A (en) A method of withered riverbed bed ripples morphological parameters are measured based on UAV aerial survey terrain data
CN110599477B (en) Method and system for detecting effective porosity of grass planting concrete
CN112730102A (en) Method for measuring evolution of particle material shear band based on standard deviation ellipse
CN110532969B (en) Slope unit dividing method based on multi-scale image segmentation
CN116152654A (en) Soil body water content identification method and device based on convolutional neural network
CN111257374A (en) Soil water content and nitrogen content monitoring method, device, equipment and storage medium
CN112884739B (en) Deep learning network-based method for rapidly detecting filling compactness of rock-fill body
CN115809599A (en) Dam body permeability coefficient inversion method and device, electronic equipment and storage medium
Bruce et al. Soil variables and interactions affecting prediction of crop yield pattern
CN114492206A (en) Calculation method for determining tunnel excavation footage based on fractured rock mass joint model

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