CN108346144A - Bridge Crack based on computer vision monitoring and recognition methods automatically - Google Patents

Bridge Crack based on computer vision monitoring and recognition methods automatically Download PDF

Info

Publication number
CN108346144A
CN108346144A CN201810089404.0A CN201810089404A CN108346144A CN 108346144 A CN108346144 A CN 108346144A CN 201810089404 A CN201810089404 A CN 201810089404A CN 108346144 A CN108346144 A CN 108346144A
Authority
CN
China
Prior art keywords
crack
subelement
image
sample
computer vision
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Granted
Application number
CN201810089404.0A
Other languages
Chinese (zh)
Other versions
CN108346144B (en
Inventor
李惠
徐阳
鲍跃全
李顺龙
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.)
Harbin Institute of Technology
Original Assignee
Harbin Institute of Technology
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 Harbin Institute of Technology filed Critical Harbin Institute of Technology
Priority to CN201810089404.0A priority Critical patent/CN108346144B/en
Publication of CN108346144A publication Critical patent/CN108346144A/en
Application granted granted Critical
Publication of CN108346144B publication Critical patent/CN108346144B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/0002—Inspection of images, e.g. flaw detection
    • G06T7/0004—Industrial image inspection
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M5/00—Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings
    • G01M5/0008—Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings of bridges
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
    • G01M5/00—Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings
    • G01M5/0033—Investigating the elasticity of structures, e.g. deflection of bridges or air-craft wings by determining damage, crack or wear
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/11—Region-based segmentation
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T7/00—Image analysis
    • G06T7/10—Segmentation; Edge detection
    • G06T7/136—Segmentation; Edge detection involving thresholding
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20081—Training; Learning
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/20—Special algorithmic details
    • G06T2207/20084—Artificial neural networks [ANN]
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T2207/00—Indexing scheme for image analysis or image enhancement
    • G06T2207/30—Subject of image; Context of image processing
    • G06T2207/30108—Industrial image inspection
    • G06T2207/30132—Masonry; Concrete

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Theoretical Computer Science (AREA)
  • Aviation & Aerospace Engineering (AREA)
  • Quality & Reliability (AREA)
  • Image Analysis (AREA)
  • Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)

Abstract

The present invention discloses a kind of Bridge Crack based on computer vision monitoring and recognition methods automatically, pass through structure training depth network model, it is input to shoot obtained image, by the operation of each hidden layer, final output obtains the tag along sort of image, it realizes crack identification, completes understanding of the computer to Input Image Content.Automatic monitoring and identification problem of the present invention for Bridge Crack, realize the full process automatization processing that the model training for the true steel box-girder crack image comprising complex background interference information, crack identification, result are shown.This method is convenient, accurate, improves the efficiency of Bridge Crack detection and accuracy and the stability of testing result.

Description

Bridge Crack based on computer vision monitoring and recognition methods automatically
Technical field
The present invention relates to civil engineerings to monitor field, and in particular to a kind of Bridge Crack based on computer vision is supervised automatically Survey and recognition methods.
Background technology
With the fast development of Chinese national economy construction, more and more Large Infrastructure Projects construction are played without proportion The effect wanted, especially large-scale steel box-girder bridge spanning the sea.Large-scale steel box-girder bridge spanning the sea due to bearing complicated vehicle lotus for a long time Load acts on, the commissure of steel box-girder often due to initial imperfection presence, cause different degrees of fatigue damage accumulation, and then shape At fatigue crack.Fatigue crack under the coupling of the disaster factors such as the long-term effect, fatigue effect and mutation effect of load, It can be extended along bead direction or to components such as top plate, diaphragm plates, bridge structure is caused to generate degradation resistance, it can under extreme case Disaster accident can be caused.Therefore, bridge management department can all invest a large amount of human and material resources, financial resources to steel box-girder inside every year Carry out manual inspection.Currently for steel box-girder crack detection mainly by patrol officer visually or by professional equipment, it is right Crack is positioned and is marked.Such detection is inefficient and inaccurate, detects the long time period of cost, and testing result The excessive subjective consciousness for depending on patrol officer.
With extensive use of the image processing method in civil engineering, there are some to be based on Threshold segmentation, shape at present The crack identification method of the traditional images Processing Algorithms such as state calculating.However these methods can not often obtain inside practical bridge To really effective application.This is because the internal environment of steel box-girder is extremely complex, and in the image taken, such as structure structure Part boundary, complicated body structure surface state (such as anti-corrosion spray painting, magnetic powder, local corrosion), uneven illumination condition etc., are steel The identification of fatigue crack brings great difficulty in box beam.It wherein influences maximum to be that often patrol officer is after having found crack With marking pen a mark line can be drawn along fracture strike, and write down around crack the cross section place residing for the crack and preliminary Dimension measurement result.In traditional images processing procedure, the identification band of these handmarkings and hand-written writing to true crack Huge interference is carried out.And the scale of fatigue crack is relatively small, and some fracture widths are only 10-1Mm grades, scheme in tradition As being easier that noise processed is taken as to fall in processing procedure.In addition, some recognition methods also require to provide the camera of image taking Inside and outside parameter (such as object distance, image distance, shooting angle), or need additional professional measuring apparatus.On the whole, traditional to split Seam recognition methods needs excessive manual intervention, and expensive.
Invention content
Based on the above shortcoming, the present invention provides a kind of Bridge Crack based on computer vision monitoring and identification automatically Method can be used for crack image identified off-line assessment, it can also be used to which crack monitors in real time.
The technology used in the present invention is as follows：A kind of Bridge Crack based on computer vision is automatic to be monitored and identification side Method, steps are as follows：
Step 1, training set make：Original input picture is cut into 64 × 64 × 3 subelement set, and therefrom with Machine extracts a certain proportion of sample, and sample size can determine as needed；The characteristics of image of these subelements is observed simultaneously, point It does not label, wherein number 1 represents Crack Element, 2 represent hand-written writing unit, and 3 represent background cell, after the completion, new to add Subelement set will be fused in former training set, each subelement corresponds to corresponding label, in order to consider injustice The influence of the three classes subelement sample size of weighing apparatus shows the quantity of three seed units at this time, and with the subelement number of minimum number On the basis of, the sample of identical quantity is randomly selected in remaining two class subelement sample, then, by each subelement sample inverse time Needle is rotated by 90 °, 180 degree, 270 degree, generate three new samples, complete data extending, the subelement sample each newly expanded possesses With identical label before rotation, so far, training set making finishes；
Step 2, the crack identification device training based on depth network model：Build the depth convolution god of fusion multi-stage characteristics It through network and completes to initialize, the size and function of each layer are as shown in table 1, are defeated in training set 64 × 64 × 3 subelement Entering, corresponding label is output, trains the parameter in the network, and the loss function in training process is softmaxloss functions, Optimization algorithm is the stochastic gradient descent algorithm with momentum, complete using the initial value of learning rate, momentum parameter and weight parameter The depth network obtained after is crack identification device；
The size and function of each layer in 1 depth network of table
Step 3, Crack Element image recognition：The subelement for being 64 × 64 × 3 by image cutting, and by each subelement It is input in crack identification device, output layer is corresponding label value, i.e., the subelement that label value is 1 is Crack Element, label value Subelement for 2 is writing unit, and the subelement that label value is 3 is background cell, and shows all types of identification knots respectively Fruit；
Step 4, post-processing output：Image segmentation, output two are carried out using optimal entropic threshold method to each crack subelement Value crack pixel recognition result, and according to length, the information of width in binaryzation crack pixel acquisition crack.
The present invention also has following technical characteristic：
1, step 2 as described above, the loss function in training process is softmaxloss functions, and formula is as follows：
In formula, L is loss function, and m is sample size, and C is classification quantity；1{y(i)=j } it is index function, as y(i)A sample classification is 1 when being jth class, is otherwise 0；bjFor weight to be updated and biasing, x(i)For input, λ is weight Parameter.
2, step 2 as described above, optimization algorithm are the stochastic gradient descent algorithm with momentum, and formula is as follows：
ν in formulaWFor weight renewal rate, αWFor weight learning rate, ηWFor weight momentum parameter, ▽WL(W；x(i),y(i)) It is loss function to the partial differential of weight；νbTo bias renewal rate, αbTo bias learning rate, ηbFor bias momentum parameter, ▽bL(W；x(i),y(i)) it is partial differential of the loss function to biasing.
3, step 4 as described above carries out image segmentation to each crack subelement using optimal entropic threshold method, public Formula is as follows：
In formula, piIndicate the ratio shared by the i-th rank gray scale, niIndicate that the quantity shared by the i-th rank gray scale, n indicate sum of all pixels Amount, PiIndicate the cumulative probability of the i-th rank gray scale, HP(t) foreground entropy, H are indicatedB(t) indicate that background entropy, H (t) indicate image total entropy, T indicates the intensity slicing value when image total entropy obtains maximum value.
4, step 4 as described above inputs pixel resolution to obtain after completing Threshold segmentation in user interface Obtain the actual length in crack, the information of width.
Automatic monitoring and identification problem of the present invention for Bridge Crack, realize for comprising complex background interference information The model training of true steel box-girder crack image, crack identification, result displaying full process automatization processing.This method is just It is prompt, accurate, improve the efficiency of Bridge Crack detection and accuracy and the stability of testing result.Entire crack identification process It is automatic business processing, significantly reduces the artificial participation during crack identification.The present invention can also meet crack and supervise online The real time data processing demand of early warning is surveyed, i.e., updates without training set, directly the image collected is identified, it is as a result defeated Going out delay can be down to second grade.The present invention improves the automating, is intelligent of Bridge Crack identification, accuracy and robustness, for soil The automatic monitoring of wood engineering Bridge Crack provides solution with identification.
Description of the drawings
Bridge Cracks of the Fig. 1 based on computer vision and deep learning monitors automatically and identification process figure
Fig. 2 merges the depth convolutional neural networks figure of multi-stage characteristics；
Fig. 3 is one long Crack Element recognition result comparison diagram；
Fig. 4 is many cracks unit recognition result comparison diagram；
Fig. 5 is crack enlarged drawing recognition result comparison diagram；
Fig. 6 is the binaryzation recognition result figure in a long crack；
Fig. 7 is the binaryzation recognition result figure of many cracks；
Fig. 8 is the binaryzation recognition result figure of crack enlarged drawing.
Specific implementation mode
Below according to Figure of description citing, the present invention will be further described：
Embodiment 1：
As shown in Figure 1, a kind of monitoring and the recognition methods automatically of Bridge Crack based on computer vision, based in MATLAB Environment is realized：
The first step, training set make：Original input picture is cut into 64 × 64 × 3 subelement set, and therefrom with Machine extracts a certain proportion of sample, and sample size can determine as needed；The characteristics of image of these subelements is observed simultaneously, point It does not label, wherein number 1 represents Crack Element, 2 represent hand-written writing unit, and 3 represent background cell, after the completion, new to add Subelement set will be fused in former training set, each subelement corresponds to corresponding label, in order to consider injustice The influence of the three classes subelement sample size of weighing apparatus shows the quantity of three seed units at this time, and with the subelement number of minimum number On the basis of, the sample of identical quantity is randomly selected in remaining two class subelement sample, then, by each subelement sample inverse time Needle is rotated by 90 °, 180 degree, 270 degree, generate three new samples, complete data extending, the subelement sample each newly expanded possesses With identical label before rotation.So far, training set making finishes.
Second step, the training of crack identification device.Build the depth convolutional neural networks of fusion multi-stage characteristics as shown in Figure 2 simultaneously Initialization is completed, the size and function of each layer are as shown in table 1.It is input in training set 64 × 64 × 3 subelement, accordingly Label is output, the parameter in the training network.Loss function in training process is softmaxloss functions (such as 1 institute of formula Show), optimization algorithm is the stochastic gradient descent algorithm (SGDM, as shown in formula 2) with momentum.Joined using learning rate, momentum The initial value of number and weight parameter, the depth network obtained after the completion is crack identification device.
In formula, L is loss function, and m is sample size, and C is classification quantity；1{y(i)=j } it is index function, as y(i)A sample classification is 1 when being jth class, is otherwise 0；bjFor weight to be updated and biasing, x(i)For input, λ is weight Parameter；
ν in formulaWFor weight renewal rate, αWFor weight learning rate, ηWFor weight momentum parameter, ▽WL(W；x(i),y(i)) It is loss function to the partial differential of weight；νbTo bias renewal rate, αbTo bias learning rate, ηbFor bias momentum parameter, ▽bL(W；x(i),y(i)) it is partial differential of the loss function to biasing；
The size and function of each layer in 1 depth network of table
Layer is not Highly Width Depth Operation Highly Width Depth Quantity Step pitch
L0 64 64 3 Convolutional layer 1-1 10 10 3 16 2
L1 28 28 16 Normalizing layer 1-1 - - - - -
L2 28 28 16 Active coating 1-1 - - - - -
L3 28 28 16 Pond layer 1-1 2 2 - - 2
L4 14 14 16 Convolutional layer 1-2 5 5 16 25 1
L5 10 10 25 Normalizing layer 1-2 - - - - -
L6 10 10 25 Active coating 1-2 - - - - -
L7 10 10 25 Pond layer 1-2 2 2 - - 2
L8 5 5 25 Full articulamentum 1 5 5 25 3 1
L9 14 14 16 Convolutional layer 2-1 7 7 16 25 1
L10 8 8 25 Normalizing layer 2-1 - - - - -
L11 8 8 25 Active coating 2-1 - - - - -
L12 8 8 25 Pond layer 2-1 2 2 - - 2
L13 4 4 25 Convolutional layer 2-2 4 4 25 36 1
L14 1 1 36 Normalizing layer 2-2 - - - - -
L15 1 1 36 Active coating 2-2 - - - - -
L16 1 1 36 Full articulamentum 2 1 1 36 3 1
L17 4 4 25 Full connection 3-1 4 4 25 36 1
L18 1 1 36 Active coating 3-1 - - - - -
L19 1 1 36 Lose layer - - - - -
L20 1 1 36 Full connection 3-2 1 1 36 3 1
L21 1 1 36 Full articulamentum 4 1 1 36 3 1
L22 1 1 3 Fused layer - - - - -
L23 1 1 3 Classification layer - - - - -
L24 1 1 1 Error layer - - - - -
Third walks, Crack Element image recognition：The subelement for being 64 × 64 × 3 by image cutting, and by each subelement It is input in crack identification device, output layer is corresponding label value, i.e., the subelement that label value is 1 is Crack Element, label value Subelement for 2 is writing unit, and the subelement that label value is 3 is background cell, and shows all types of identification knots respectively Fruit, as in Figure 3-5.
4th step, post-processing output：Image segmentation, such as formula are carried out using optimal entropic threshold method to each crack subelement Shown in 3, output binaryzation crack pixel recognition result as shown in figs 6-8, and obtains crack according to binaryzation crack pixel Length, the information of width.After completing Threshold segmentation, the true of crack is obtained by inputting pixel resolution (mm/pixel) The information of true length degree, width.
In formula, piIndicate the ratio shared by the i-th rank gray scale, niIndicate that the quantity shared by the i-th rank gray scale, n indicate sum of all pixels Amount, PiIndicate the cumulative probability of the i-th rank gray scale, HP(t) foreground entropy, H are indicatedB(t) indicate that background entropy, H (t) indicate image total entropy, T indicates the intensity slicing value when image total entropy obtains maximum value.
The present embodiment is implemented under MATLAB environment, can be directly applied for the crack pattern shot with consumer level general camera Picture does not need special shooting or detection device, and accuracy of identification is high, and speed is fast, at low cost, can be not only used for identified off-line assessment, It can also be used for monitoring in real time, improve the automating, is intelligent of steel box-girder fatigue crack identification, accuracy and robustness.

Claims (5)

1. a kind of monitoring and the recognition methods automatically of Bridge Crack based on computer vision, which is characterized in that method is as follows：
Step 1, training set make：Original input picture is cut into 64 × 64 × 3 subelement set, and therefrom random pumping Take a certain proportion of sample, sample size that can determine as needed；The characteristics of image for observing these subelements simultaneously, beats respectively Label, wherein number 1 represents Crack Element, 2 represent hand-written writing unit, and 3 represent background cell, after the completion, newly added son Unit set will be fused in former training set, each subelement corresponds to corresponding label, unbalanced in order to consider The influence of three classes subelement sample size shows the quantity of three seed units at this time, and using the subelement number of minimum number as base Standard, the sample that identical quantity is randomly selected in remaining two class subelement sample then revolve each subelement sample counterclockwise Turn 90 degrees, 180 degree, 270 degree, generate three new samples, complete data extending, the subelement sample each newly expanded possesses and revolve Identical label before turning, so far, training set making finish；
Step 2, the crack identification device training based on depth network model：Build the depth convolutional Neural net of fusion multi-stage characteristics Network simultaneously is completed to initialize, and the size and function of each layer are as shown in table 1, is input, phase in training set 64 × 64 × 3 subelement The label answered is output, trains the parameter in the network, and the loss function in training process is softmaxloss functions, optimization Algorithm is the stochastic gradient descent algorithm with momentum, using the initial value of learning rate, momentum parameter and weight parameter, after the completion Obtained depth network is crack identification device；
The size and function of each layer in 1 depth network of table
Step 3, Crack Element image recognition：Image cutting is 64 × 64 × 3 subelement, and each subelement is inputted Into crack identification device, output layer is corresponding label value, i.e., the subelement that label value is 1 is Crack Element, and label value is 2 Subelement is writing unit, and the subelement that label value is 3 is background cell, and shows all types of recognition results respectively；
Step 4, post-processing output：Image segmentation is carried out using optimal entropic threshold method to each crack subelement, exports binaryzation Crack pixel recognition result, and according to length, the information of width in binaryzation crack pixel acquisition crack.
2. a kind of Bridge Crack based on computer vision according to claim 1 monitoring and recognition methods automatically, special Sign is：Step 2, the loss function in training process is softmaxloss functions, and formula is as follows：
In formula, L is loss function, and m is sample size, and C is classification quantity；1{y(i)=j } it is index function, as y(i)It is a Sample classification is 1 when being jth class, is otherwise 0；bjFor weight to be updated and biasing, x(i)For input, λ is weight parameter.
3. a kind of Bridge Crack based on computer vision according to claim 1 monitoring and recognition methods automatically, special Sign is：Step 2, optimization algorithm are the stochastic gradient descent algorithm with momentum, and formula is as follows：
ν in formulaWFor weight renewal rate, αWFor weight learning rate, ηWFor weight momentum parameter,For loss The partial differential of function pair weight；νbTo bias renewal rate, αbTo bias learning rate, ηbFor bias momentum parameter,It is loss function to the partial differential of biasing.
4. a kind of Bridge Crack based on computer vision according to claim 1 monitoring and recognition methods automatically, special Sign is：Step 4 carries out image segmentation to each crack subelement using optimal entropic threshold method, and formula is as follows：
In formula, piIndicate the ratio shared by the i-th rank gray scale, niIndicate that the quantity shared by the i-th rank gray scale, n indicate total number of pixels, PiIndicate the cumulative probability of the i-th rank gray scale, HP(t) foreground entropy, H are indicatedB(t) indicate that background entropy, H (t) indicate image total entropy, T tables Show the intensity slicing value when image total entropy obtains maximum value.
5. a kind of Bridge Crack based on computer vision according to claim 1 monitoring and recognition methods automatically, special Sign is：Step 4 obtains the actual length in crack, width after completing Threshold segmentation by inputting pixel resolution Information.
CN201810089404.0A 2018-01-30 2018-01-30 Automatic bridge crack monitoring and identifying method based on computer vision Active CN108346144B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201810089404.0A CN108346144B (en) 2018-01-30 2018-01-30 Automatic bridge crack monitoring and identifying method based on computer vision

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201810089404.0A CN108346144B (en) 2018-01-30 2018-01-30 Automatic bridge crack monitoring and identifying method based on computer vision

Publications (2)

Publication Number Publication Date
CN108346144A true CN108346144A (en) 2018-07-31
CN108346144B CN108346144B (en) 2021-03-16

Family

ID=62960701

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201810089404.0A Active CN108346144B (en) 2018-01-30 2018-01-30 Automatic bridge crack monitoring and identifying method based on computer vision

Country Status (1)

Country Link
CN (1) CN108346144B (en)

Cited By (12)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109029381A (en) * 2018-10-19 2018-12-18 石家庄铁道大学 A kind of detection method of tunnel slot, system and terminal device
CN109376676A (en) * 2018-11-01 2019-02-22 哈尔滨工业大学 Highway engineering site operation personnel safety method for early warning based on unmanned aerial vehicle platform
CN109408985A (en) * 2018-11-01 2019-03-01 哈尔滨工业大学 The accurate recognition methods in bridge steel structure crack based on computer vision
CN109919942A (en) * 2019-04-04 2019-06-21 哈尔滨工业大学 Bridge Crack intellectualized detection method based on high-precision noise reduction theory
CN110020652A (en) * 2019-01-07 2019-07-16 新而锐电子科技(上海)有限公司 The dividing method of Tunnel Lining Cracks image
CN110569832A (en) * 2018-11-14 2019-12-13 安徽艾睿思智能科技有限公司 text real-time positioning and identifying method based on deep learning attention mechanism
CN111091554A (en) * 2019-12-12 2020-05-01 哈尔滨市科佳通用机电股份有限公司 Railway wagon swing bolster fracture fault image identification method
CN111369526A (en) * 2020-03-03 2020-07-03 中建二局基础设施建设投资有限公司 Multi-type old bridge crack identification method based on semi-supervised deep learning
CN111563888A (en) * 2020-05-06 2020-08-21 清华大学 Quantitative crack growth monitoring method
CN111832617A (en) * 2020-06-05 2020-10-27 上海交通大学 Engine cold state test fault diagnosis method
CN113406088A (en) * 2021-05-10 2021-09-17 同济大学 Fixed point type steel box girder crack development observation device
CN113935086A (en) * 2021-09-17 2022-01-14 哈尔滨工业大学 Intelligent structure design method based on computer vision and deep learning

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1236173A2 (en) * 1999-10-27 2002-09-04 Biowulf Technologies, LLC Methods and devices for identifying patterns in biological systems
CN105975968A (en) * 2016-05-06 2016-09-28 西安理工大学 Caffe architecture based deep learning license plate character recognition method
CN106384080A (en) * 2016-08-31 2017-02-08 广州精点计算机科技有限公司 Apparent age estimating method and device based on convolutional neural network
CN107133943A (en) * 2017-04-26 2017-09-05 贵州电网有限责任公司输电运行检修分公司 A kind of visible detection method of stockbridge damper defects detection
CN107301383A (en) * 2017-06-07 2017-10-27 华南理工大学 A kind of pavement marking recognition methods based on Fast R CNN
CN107403197A (en) * 2017-07-31 2017-11-28 武汉大学 A kind of crack identification method based on deep learning

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP1236173A2 (en) * 1999-10-27 2002-09-04 Biowulf Technologies, LLC Methods and devices for identifying patterns in biological systems
CN105975968A (en) * 2016-05-06 2016-09-28 西安理工大学 Caffe architecture based deep learning license plate character recognition method
CN106384080A (en) * 2016-08-31 2017-02-08 广州精点计算机科技有限公司 Apparent age estimating method and device based on convolutional neural network
CN107133943A (en) * 2017-04-26 2017-09-05 贵州电网有限责任公司输电运行检修分公司 A kind of visible detection method of stockbridge damper defects detection
CN107301383A (en) * 2017-06-07 2017-10-27 华南理工大学 A kind of pavement marking recognition methods based on Fast R CNN
CN107403197A (en) * 2017-07-31 2017-11-28 武汉大学 A kind of crack identification method based on deep learning

Cited By (17)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN109029381A (en) * 2018-10-19 2018-12-18 石家庄铁道大学 A kind of detection method of tunnel slot, system and terminal device
CN109376676A (en) * 2018-11-01 2019-02-22 哈尔滨工业大学 Highway engineering site operation personnel safety method for early warning based on unmanned aerial vehicle platform
CN109408985A (en) * 2018-11-01 2019-03-01 哈尔滨工业大学 The accurate recognition methods in bridge steel structure crack based on computer vision
CN110569832B (en) * 2018-11-14 2022-05-31 安徽省科亿信息科技有限公司 Text real-time positioning and identifying method based on deep learning attention mechanism
CN110569832A (en) * 2018-11-14 2019-12-13 安徽艾睿思智能科技有限公司 text real-time positioning and identifying method based on deep learning attention mechanism
CN110020652A (en) * 2019-01-07 2019-07-16 新而锐电子科技(上海)有限公司 The dividing method of Tunnel Lining Cracks image
CN109919942A (en) * 2019-04-04 2019-06-21 哈尔滨工业大学 Bridge Crack intellectualized detection method based on high-precision noise reduction theory
CN109919942B (en) * 2019-04-04 2020-01-14 哈尔滨工业大学 Bridge crack intelligent detection method based on high-precision noise reduction theory
CN111091554A (en) * 2019-12-12 2020-05-01 哈尔滨市科佳通用机电股份有限公司 Railway wagon swing bolster fracture fault image identification method
CN111369526A (en) * 2020-03-03 2020-07-03 中建二局基础设施建设投资有限公司 Multi-type old bridge crack identification method based on semi-supervised deep learning
CN111369526B (en) * 2020-03-03 2023-04-18 中建二局土木工程集团有限公司 Multi-type old bridge crack identification method based on semi-supervised deep learning
CN111563888A (en) * 2020-05-06 2020-08-21 清华大学 Quantitative crack growth monitoring method
CN111832617A (en) * 2020-06-05 2020-10-27 上海交通大学 Engine cold state test fault diagnosis method
CN111832617B (en) * 2020-06-05 2022-11-08 上海交通大学 Engine cold state test fault diagnosis method
CN113406088A (en) * 2021-05-10 2021-09-17 同济大学 Fixed point type steel box girder crack development observation device
CN113935086A (en) * 2021-09-17 2022-01-14 哈尔滨工业大学 Intelligent structure design method based on computer vision and deep learning
CN113935086B (en) * 2021-09-17 2022-08-02 哈尔滨工业大学 Intelligent structure design method based on computer vision and deep learning

Also Published As

Publication number Publication date
CN108346144B (en) 2021-03-16

Similar Documents

Publication Publication Date Title
CN108346144A (en) Bridge Crack based on computer vision monitoring and recognition methods automatically
Li et al. Automatic pavement crack detection by multi-scale image fusion
CN110084195B (en) Remote sensing image target detection method based on convolutional neural network
CN108319949A (en) Mostly towards Ship Target Detection and recognition methods in a kind of high-resolution remote sensing image
CN109919934A (en) A kind of liquid crystal display panel defect inspection method based on the study of multi-source domain depth migration
CN107564002A (en) Plastic tube detection method of surface flaw, system and computer-readable recording medium
CN105180890A (en) Rock structural surface occurrence measuring method integrated with laser-point cloud and digital imaging
CN107507170A (en) A kind of airfield runway crack detection method based on multi-scale image information fusion
CN110232379A (en) A kind of vehicle attitude detection method and system
CN109508881B (en) Sea island region classification and ecological resource value evaluation method
CN111339827A (en) SAR image change detection method based on multi-region convolutional neural network
CN109615604A (en) Accessory appearance flaw detection method based on image reconstruction convolutional neural networks
CN108520277A (en) Reinforced concrete structure seismic Damage automatic identification based on computer vision and intelligent locating method
CN108932724A (en) A kind of system automatic auditing method based on multi-person synergy image labeling
CN115115627B (en) Soil saline-alkali soil monitoring method based on data processing
CN111724358A (en) Concrete quality detection method and system based on image and convolutional neural network
CN111079773A (en) Gravel parameter acquisition method, device, equipment and storage medium based on Mask R-CNN network
CN107301649A (en) A kind of region merging technique SAR image coastline Detection Method algorithm based on super-pixel
CN113360587B (en) Land surveying and mapping equipment and method based on GIS technology
Ning et al. Research on surface defect detection algorithm of strip steel based on improved YOLOV3
Bayırlı et al. Determining different plant leaves' fractal dimensions: a new approach to taxonomical study of plants
CN116309155A (en) Image restoration method, model and device based on convolution and converter hybrid network
CN116415843A (en) Multi-mode remote sensing auxiliary mine ecological environment evaluation method for weak network environment
CN111275684A (en) Strip steel surface defect detection method based on multi-scale feature extraction
Bober et al. Synthetic Landscape Differentiation Index a Tool for Spatial Planning

Legal Events

Date Code Title Description
PB01 Publication
PB01 Publication
SE01 Entry into force of request for substantive examination
SE01 Entry into force of request for substantive examination
GR01 Patent grant
GR01 Patent grant