CN109978032B - Bridge crack detection method based on space pyramid cavity convolution network - Google Patents
Bridge crack detection method based on space pyramid cavity convolution network Download PDFInfo
- Publication number
- CN109978032B CN109978032B CN201910198978.6A CN201910198978A CN109978032B CN 109978032 B CN109978032 B CN 109978032B CN 201910198978 A CN201910198978 A CN 201910198978A CN 109978032 B CN109978032 B CN 109978032B
- Authority
- CN
- China
- Prior art keywords
- network
- image
- data set
- bridge crack
- convolution 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.)
- Active
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- 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
- 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
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- Health & Medical Sciences (AREA)
- Quality & Reliability (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Image Analysis (AREA)
- Investigating Materials By The Use Of Optical Means Adapted For Particular Applications (AREA)
Abstract
The invention discloses a bridge crack detection method based on a space pyramid cavity convolution network, which comprises the following steps: 1. constructing a spatial pyramid cavity convolution network; 2. generating a training set, a label and a test set; 3. preprocessing each image in the training set and the test set; 4. training a spatial pyramid hole convolution network; 5. detecting the test set; 6. and calculating the test index of the bridge crack image. According to the method, the detection of the space pyramid cavity convolution network on the bridge crack image is guaranteed by building the space pyramid cavity convolution network, the space pyramid cavity convolution network is trained on a large data set by using a random gradient descent method, the robustness and the accuracy of the space pyramid cavity convolution network on the detection of the bridge crack image are guaranteed, and the training can be completed in a short time.
Description
Technical Field
The invention belongs to the technical field of physics, and further relates to a concrete bridge crack detection method based on a space pyramid cavity convolution network in the technical field of image processing. The method can be applied to detecting the cracks in the concrete bridge image.
Background
The index for measuring the damage degree of the concrete bridge consists of information such as crack length, width, number and the like. The common concrete bridge crack information acquisition method comprises the following steps: the magnifying glass measuring method or the crack measuring instrument is manually operated, so that the efficiency is low, the measuring precision is low, the cost is high, and certain danger can be caused to workers on the concrete bridge site; the deep learning detection method is an automatic operation method, has high efficiency, high measurement precision and low cost, is the most researched technology in the field of concrete bridge crack detection at present, and is characterized in that image acquisition is completed through a high-definition camera, and a computer processes and calculates the acquired concrete bridge crack image to obtain the length, width, number and other related information of cracks, so that the disease degree of the concrete bridge is evaluated according to the obtained information.
The patent document "concrete bridge surface crack detection method based on HOG characteristics" (patent application number: 2017102250202, application publication number: CN1071699 a) applied by the university of west-ann electronic technology proposes a method for detecting concrete bridge cracks by using HOG (Histogram of Oriented Gradient) characteristics. Inputting a color image H, converting the color image H into a gray image, denoising and filtering, then carrying out binarization processing on the filtered image, counting a histogram in a gradient direction, and carrying out primary processing and merging on a denoised gray binary image block according to a statistical result to obtain a complete binary image of the image H; next, calculating the HOG (Histogram of Oriented Gradient) feature of the image H, training an SVM classifier, calculating the HOG (Histogram of Oriented Gradient) feature of the color image N, inputting the HOG feature into the SVM classifier, classifying each pixel point of the image N, and labeling the classification result to obtain the detected crack map. Although the method can be used for classification training through HOG (Histogram of Oriented Gradient) features, the method still has the defects that the accuracy is reduced when the HOG (Histogram of Oriented Gradient) features are extracted due to more noise points of the concrete bridge crack images and quite sensitive to the noise points, and the accuracy is reduced when the SVM is classified by using the HOG features due to low HOG feature extraction accuracy, so that the accuracy is reduced when the concrete bridge crack detection accuracy is reduced.
The Nanjing university of Richardson who applies for the patent document, "pavement crack detection method based on multiple low-layer significant features" (patent application No. 201410141476.7, application publication No. 103903268A) proposes a pavement crack detection method based on multiple low-layer significant features. The method comprises the steps of collecting a pavement gray level image, dividing the image into a plurality of local blocks, calculating the rarity probability of each pixel, generating a crack characteristic map, extracting cracks through double neighborhood expansion of a Bayes theory, further extracting the cracks through region growing enhancement through self-adaptive threshold segmentation, and finally calculating specific parameters such as the area and the position of the cracks. Although the method can detect the cracks, the method has the disadvantage that due to the interference of background pixels of the crack images, the errors are large when the pixel sparsity probability is calculated, and therefore the resolution of the detected crack images is poor.
Disclosure of Invention
The invention aims to provide a concrete bridge crack detection method based on a space pyramid cavity convolution network aiming at the defects of the prior art, and the concrete bridge crack detection method is constructed in the space pyramid cavity convolution network, sets a detection state and quickly detects whether a concrete bridge image contains cracks.
To achieve the above object, the method of the present invention comprises the steps of:
(1) Constructing a spatial pyramid hole convolution network:
(1a) Constructing a space pyramid cavity convolution network formed by connecting a main network and a sub-network in parallel;
the main network consists of an input layer and a 1 st convolution layer;
the sub-network consists of four branches connected in parallel: the branch 1 is composed of an input layer of a main network and a 1 st cavity convolution layer; the branch 2 consists of an input layer of the main network and a 2 nd cavity convolution layer; the branch 3 is composed of an input layer of the main network and a 3 rd cavity convolution layer; the branch 4 consists of an input layer and a pooling layer of the main network;
(1b) The parameters of each layer of the spatial pyramid cavity convolution network are set as follows:
setting the number of feature maps of the input layer to 1024;
setting the number of the feature maps of the four convolutional layers to be 512, sequentially setting the sizes of the convolutional layers to be 1 × 1, 3 × 3 and 3 × 3, and sequentially setting the void rates of the 2 nd convolutional layer, the 3 rd convolutional layer and the 4 th convolutional layer to be 6, 12 and 18;
setting the pooling layer downsampling filter size to 2 x 2;
(2) Generating a training data set, a label data set and a test data set:
(2a) Collecting at least 1000 concrete bridge crack images in real time, wherein each concrete bridge crack image is provided with a concrete bridge crack, forming the collected pictures into a space pyramid cavity convolution network data set, randomly extracting 70% of pictures from the space pyramid cavity convolution network data set to form a training data set, and forming the rest 30% of pictures into a test data set;
(2b) Manually marking crack pixels in each concrete bridge crack image in the training data set to generate marked images to form a label data set;
(3) Preprocessing a concrete bridge crack image to be detected:
(3a) Performing convolution kernel operation on the concrete bridge crack image in the transverse direction and the longitudinal direction by using a gradient calculation method, calculating the gradient value of each pixel point, calculating the gradient direction angle according to the transverse gradient value and the longitudinal gradient value, and forming all the gradient values into a gradient image;
(3b) Normalizing the concrete bridge crack image matrix and the characteristic matrix, and normalizing the element values in the image to be between [0,1] to obtain a normalized image;
(4) Training a spatial pyramid cavity convolutional neural network:
inputting the training data set into the spatial pyramid cavity convolutional neural network, and continuously adjusting and optimizing network training parameters by adopting a gradient descent method until the loss value of the spatial pyramid cavity convolutional neural network is less than a preset value of 0.3 to obtain a trained spatial pyramid cavity convolutional neural network;
(5) And (3) detecting the test data set:
and inputting the test data set into the trained spatial pyramid cavity convolution neural network, and extracting crack pixels from the concrete bridge crack image.
Compared with the prior art, the invention has the following advantages:
firstly, the space pyramid cavity convolution network is constructed to detect the concrete bridge cracks, so that the problem of low detection accuracy rate caused by complex background of concrete bridge crack images in the prior art is solved, and the accuracy rate of concrete bridge crack detection is improved.
Secondly, the test set is input into the trained space pyramid cavity convolution network to detect the concrete bridge cracks, so that the problem that the multi-scale object of the image cannot be detected in the prior art is solved, and the resolution of the concrete bridge crack detection is improved.
Drawings
FIG. 1 is a flow chart of the present invention.
FIG. 2 is a diagram of simulation effect of the present invention.
Detailed Description
The invention is further described below with reference to the accompanying drawings.
The specific steps implemented by the present invention are described in further detail with reference to fig. 1.
To achieve the above object, the method of the present invention comprises the steps of:
step 1, constructing a spatial pyramid cavity convolution network:
constructing a space pyramid cavity convolution network formed by connecting a main network and a sub-network in parallel, wherein the main network consists of an input layer and a 1 st convolution layer;
the sub-network consists of four branches connected in parallel: the branch 1 is composed of an input layer of a main network and a 1 st cavity convolution layer; the branch 2 consists of an input layer of the main network and a 2 nd cavity convolution layer; the branch 3 is composed of an input layer of the main network and a 3 rd cavity convolution layer; the branch 4 consists of an input layer and a pooling layer of the main network;
the parameters of each layer of the spatial pyramid cavity convolution network are set as follows: setting the number of feature maps of the input layer to 1024; setting the number of the feature maps of the four convolutional layers to be 512, sequentially setting the sizes of the convolutional layers to be 1 × 1, 3 × 3 and 3 × 3, and sequentially setting the void rates of the 2 nd convolutional layer, the 3 rd convolutional layer and the 4 th convolutional layer to be 6, 12 and 18; setting the size of a downsampling filter of a pooling layer to be 2 multiplied by 2;
step 2, generating a training data set, a label data set and a test data set:
collecting at least 1000 concrete bridge crack images, wherein each concrete bridge crack image has a concrete bridge crack, forming a space pyramid hole convolution network data set by the collected pictures, randomly extracting 70% of the pictures from the space pyramid hole convolution network data set to form a training data set, and forming a testing data set by the rest 30% of the pictures; and manually marking crack pixels in each concrete bridge crack image in the training data set to generate marked images to form a label data set.
Step 3, preprocessing the concrete bridge crack image to be detected:
performing convolution kernel operation on the concrete bridge crack image in the transverse direction Gx and the longitudinal direction Gy by using a gradient calculation method, calculating the gradient value of each pixel point, calculating the gradient direction angle according to the transverse gradient value and the longitudinal gradient value, and combining all the gradient values into a gradient image, wherein the convolution kernel in the Gx direction and the convolution kernel in the Gy direction are as follows:
and I is an input 3X 3 matrix of the concrete bridge crack image.
Normalizing the concrete bridge crack image matrix, and normalizing the element values in the image matrix to be between [0,1] to obtain a normalized image matrix;
step 4, training the spatial pyramid cavity convolutional neural network:
loading the network parameters and the network structure of the deep learning network into a computer memory;
and inputting the samples of the training data set into the spatial pyramid cavity convolutional neural network, and iteratively updating the parameters of the spatial pyramid cavity convolutional neural network by adopting a random gradient descent method to obtain the trained spatial pyramid cavity convolutional neural network.
The steps of the gradient descent method are as follows:
step one, in the range of [10000,12000], arbitrarily selecting a value as the iteration number of the pyramid cavity convolution network, setting all 164778 parameter values of the 5-layer pyramid cavity convolution network into Gaussian random numbers, wherein the average value of the Gaussian random numbers is 0.0001, the variance is set to be 0.5, and the learning rate is set to be 0.005;
second, randomly selecting 10 images from the training data set of pyramid hole convolution
Thirdly, calculating the average value of the loss values of all the training data sets according to the following formula:
where L represents the average of all training data set loss values, Σ represents the summation operation, i represents the sequence number of each image in the training data set, i =1,2,3,4, y i Representing the expected output, y, of each image of the training data set i_ Representing the actual output of each image training of the training dataset.
Fourthly, updating each parameter in the pyramid hole convolution network according to the following formula:
wherein the content of the first and second substances,
represents the updated value of the x-th parameter in the pyramid hole convolution network, theta x Representing the value of the xth parameter in the pyramid hole convolution network before updating,
representing a derivation operation;
step five, judging whether the current iteration times are equal to the set iteration times of the pyramid hole convolution network, if so, executing a sixth step, otherwise, executing the second step;
and sixthly, obtaining the trained pyramid cavity convolution network.
And 5, detecting the test data set:
and inputting the test data set into the trained spatial pyramid cavity convolutional neural network, and outputting a detection image corresponding to each crack image in the test data set.
And 6, calculating the test indexes of the concrete bridge crack image.
The proportion of pixels detected in the sample image in the test data set to the label pixels, namely the accuracy is calculated and used as a detection index for measuring the invention.
The experimental simulation of the present invention is illustrated with reference to fig. 2:
1. experimental simulation conditions:
the computer used in the simulation experiment of the invention is configured as follows: the processor is an Intel Core i5CPU, the display card is NVIDIA GeForce GTX 950M, and the memory is 256GB. The computer system is windows 7, and a Keras frame deep learning network frame based on a Tenserflow rear end and a Pycarm software compiling environment are used for realizing a simulation experiment.
The training data set and the testing data set used by the invention are obtained from a plurality of different concrete bridge crack scenes, images are obtained through image acquisition equipment, the obtained images need to contain concrete bridge crack images with different changes such as uneven illumination, oil stain pollution and the like, and the testing data used by the simulation experiment of the invention is from the bottom surface of a concrete bridge in Baoji City of Shaanxi province.
2. Experiment simulation content:
the simulation experiment of the invention is that the method of the invention is combined with the U-net convolution neural network, 70% of crack images in the collected data set are randomly selected as a training data set, the rest 30% of cracks are used as a test data set, and the constructed space pyramid cavity convolution network is trained on the training data set to obtain the trained space pyramid cavity convolution network.
3. The experimental simulation effect is as follows:
fig. 2 is a simulation experiment result diagram of the present invention, fig. 2 (a) is a test sample image of a concrete bridge crack, the image is collected on the bottom surface of a concrete bridge in Baoji city, shaanxi province, and fig. 2 (b) is a result diagram obtained by detecting fig. 2 (a) according to the present invention.
As can be seen from fig. 2 (b), under complex background interference, the obtained detection result can clearly and accurately display the crack. In order to illustrate the accuracy of the method, the concrete bridge crack detection method based on the U-net convolution network is compared with the method disclosed by the invention for verification, and the comparison result of the accuracy and the recall rate is shown in the following table 1.
It can be seen from table 1 that the accuracy of the present invention is higher than that of the U-net network model. TABLE 1 comparison of concrete bridge crack pattern test results
| Algorithm | U-net network structure | Text network architecture |
| Evaluation criteria | Accuracy/recallRate of change | Accuracy/recall |
| First image | 71.9％/85.5％ | 87.3％/72.6％ |
| Second image | 78.5％/83.8％ | 92.7％/70.3％ |
| Third image | 76.8％/83.1％ | 89.6％/65.1％ |
Claims (4)
1. A bridge crack detection method based on a space pyramid hole convolution network is characterized by comprising the following specific steps of:
(1) Constructing a spatial pyramid hole convolution network:
(1a) Constructing a space pyramid cavity convolution network formed by connecting a main network and a sub-network in parallel;
the main network consists of an input layer and a 1 st convolution layer;
the sub-network consists of four branches connected in parallel: the branch 1 is composed of an input layer of a main network and a 1 st cavity convolution layer; the branch 2 consists of an input layer of the main network and a 2 nd cavity convolution layer; the branch 3 consists of an input layer of the main network and a 3 rd cavity convolution layer; the branch 4 consists of an input layer and a pooling layer of the main network;
(1b) The parameters of each layer of the spatial pyramid cavity convolution network are set as follows:
setting the number of feature maps of the input layer to 1024;
setting the number of the feature maps of the four convolutional layers to be 512, sequentially setting the sizes of the convolutional layers to be 1 × 1, 3 × 3 and 3 × 3, and sequentially setting the void rates of the 2 nd convolutional layer, the 3 rd convolutional layer and the 4 th convolutional layer to be 6, 12 and 18;
setting the pooling layer downsampling filter size to 2 x 2;
(2) Generating a training data set, a label data set and a test data set:
(2a) Collecting at least 1000 bridge crack images in real time, wherein each bridge crack image is provided with a bridge crack, forming the collected pictures into a space pyramid cavity convolution network data set, randomly extracting 70% of pictures from the space pyramid cavity convolution network data set to form a training data set, and forming the rest 30% of pictures into a test data set;
(2b) Manually marking crack pixels in each bridge crack image in the training data set to generate marked images to form a label data set;
(3) Preprocessing a bridge crack image to be detected:
(3a) Performing convolution kernel operation on the bridge crack image in the transverse direction and the longitudinal direction by using a gradient calculation method, calculating the gradient value of each pixel point, calculating the gradient direction angle according to the transverse gradient value and the longitudinal gradient value, and forming all the gradient values into a gradient image;
(3b) Normalizing the bridge crack image matrix and the characteristic matrix, and normalizing the element values in the image to be between [0,1] to obtain a normalized image;
(4) Training a spatial pyramid cavity convolution neural network:
inputting a training data set into the space pyramid cavity convolutional neural network, and continuously adjusting and optimizing network training parameters by adopting a gradient descent method until the loss value of the space pyramid cavity convolutional neural network is less than a preset value of 0.3 to obtain a trained space pyramid cavity convolutional neural network;
(5) And (3) detecting the test data set:
and inputting the test data set into the trained space pyramid cavity convolution neural network, and extracting crack pixels from the bridge crack image.
2. The method for detecting a bridge crack based on the spatial pyramid void convolution network as claimed in claim 1, wherein the gradient calculation method in step (3 a) specifically comprises the following steps:
the first step is as follows: respectively calculating the gradients of the bridge crack image in the x direction and the y direction;
the second step is that: calculating the gradient of the bridge crack image according to the following formula:
wherein G represents the gradient value of the pixel point of the crack image, G x Transverse difference, G, representing pixel points of the crack image y The longitudinal difference of the pixel points of the crack image is represented,
representing root number opening operation;
the third step: calculating the gradient angle of the bridge crack image according to the following formula:
wherein theta ^ represents the direction angle value of the gradient of the pixel point of the crack image, arctan represents the arctangent operation, G x Transverse difference, G, representing pixel points of the crack image y And expressing the longitudinal difference of the pixel points of the crack image.
3. The method for detecting the bridge crack based on the spatial pyramid hole convolution network as claimed in claim 1, wherein the step (3 b) of normalizing the bridge crack image matrix and the feature matrix comprises the following specific steps:
the method comprises the steps of firstly, respectively calculating the maximum value and the minimum value of each channel of a bridge crack image matrix and a feature matrix;
secondly, subtracting the minimum value of the channel pixel from all elements of each channel of the bridge crack image matrix, and dividing the subtracted pixel minimum value by the maximum value of the channel pixel to obtain a normalized bridge crack image matrix;
and thirdly, obtaining a normalized feature matrix by adopting the same method as the second step.
4. The method for detecting the bridge crack based on the spatial pyramid void convolution network as claimed in claim 1, wherein the gradient descent method in the step (4) comprises the following steps:
step one, in the range of [10000,12000], selecting a value arbitrarily as the iteration number of the pyramid hole convolution network, setting all 9472 parameter values of the 5-layer pyramid hole convolution network as Gaussian random numbers, wherein the mean value of the Gaussian random numbers is 0.0001, the variance is set as 1.00, and the learning rate is set as 0.005;
secondly, randomly selecting 10 images from the training data set of pyramid void convolution
Thirdly, calculating the average value of the loss values of all the training data sets according to the following formula:
where L represents the average of all training data set loss values, sigma represents the summation operation, i represents the sequence number of each image in the training data set, i =1,2,3,4, y i Representing the expected output, y, of each image of the training data set i_ Representing the actual output of each image of the training dataset;
fourthly, updating each parameter in the pyramid hole convolution network according to the following formula:
wherein the content of the first and second substances,
represents the updated value of the x-th parameter in the pyramid hole convolution network, theta x Representing the value of the xth parameter in the pyramid hole convolution network before updating,
representing a derivation operation;
fifthly, judging whether the current iteration number is equal to the set iteration number of the pyramid cavity convolution network, if so, executing a sixth step, otherwise, executing the second step;
and sixthly, obtaining the trained pyramid hole convolution network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910198978.6A CN109978032B (en) | 2019-03-15 | 2019-03-15 | Bridge crack detection method based on space pyramid cavity convolution network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910198978.6A CN109978032B (en) | 2019-03-15 | 2019-03-15 | Bridge crack detection method based on space pyramid cavity convolution network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109978032A CN109978032A (en) | 2019-07-05 |
| CN109978032B true CN109978032B (en) | 2022-12-06 |
Family
ID=67079128
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910198978.6A Active CN109978032B (en) | 2019-03-15 | 2019-03-15 | Bridge crack detection method based on space pyramid cavity convolution network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109978032B (en) |
Families Citing this family (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110503613B (en) * | 2019-08-13 | 2022-03-15 | 电子科技大学 | Single image-oriented rain removing method based on cascade cavity convolution neural network |
| CN110717527B (en) * | 2019-09-24 | 2023-06-27 | 东南大学 | Method for determining target detection model by combining cavity space pyramid structure |
| CN110738642A (en) * | 2019-10-08 | 2020-01-31 | 福建船政交通职业学院 | Mask R-CNN-based reinforced concrete crack identification and measurement method and storage medium |
| CN111126202B (en) * | 2019-12-12 | 2022-03-04 | 天津大学 | Optical remote sensing image target detection method based on void feature pyramid network |
| CN111080641A (en) * | 2019-12-30 | 2020-04-28 | 上海商汤智能科技有限公司 | Crack detection method and device, computer equipment and storage medium |
| CN111507185B (en) * | 2020-03-11 | 2020-11-24 | 杭州电子科技大学 | Tumble detection method based on stack cavity convolution network |
| CN112465776B (en) * | 2020-11-26 | 2023-10-31 | 常州信息职业技术学院 | Crack intelligent detection method based on wind turbine surface blurred image |
| CN112734734A (en) * | 2021-01-13 | 2021-04-30 | 北京联合大学 | Railway tunnel crack detection method based on improved residual error network |
| CN114441546B (en) * | 2022-04-08 | 2022-06-24 | 湖南万航科技有限公司 | Crack measurement endoscope |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108876780A (en) * | 2018-06-26 | 2018-11-23 | 陕西师范大学 | Bridge Crack image crack detection method under a kind of complex background |
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN109325534A (en) * | 2018-09-22 | 2019-02-12 | 天津大学 | A kind of semantic segmentation method based on two-way multi-Scale Pyramid |
-
2019
- 2019-03-15 CN CN201910198978.6A patent/CN109978032B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2018214195A1 (en) * | 2017-05-25 | 2018-11-29 | 中国矿业大学 | Remote sensing imaging bridge detection method based on convolutional neural network |
| CN108876780A (en) * | 2018-06-26 | 2018-11-23 | 陕西师范大学 | Bridge Crack image crack detection method under a kind of complex background |
| CN109325534A (en) * | 2018-09-22 | 2019-02-12 | 天津大学 | A kind of semantic segmentation method based on two-way multi-Scale Pyramid |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109978032A (en) | 2019-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109978032B (en) | Bridge crack detection method based on space pyramid cavity convolution network | |
| Ali et al. | Structural crack detection using deep convolutional neural networks | |
| CN109740603B (en) | Vehicle character recognition method based on CNN convolutional neural network | |
| CN110929607B (en) | Remote sensing identification method and system for urban building construction progress | |
| CN112232391B (en) | Dam crack detection method based on U-net network and SC-SAM attention mechanism | |
| CN109740639B (en) | Wind cloud satellite remote sensing image cloud detection method and system and electronic equipment | |
| CN110264448B (en) | Insulator fault detection method based on machine vision | |
| CN111257341B (en) | Underwater building crack detection method based on multi-scale features and stacked full convolution network | |
| Rahaman et al. | An efficient multilevel thresholding based satellite image segmentation approach using a new adaptive cuckoo search algorithm | |
| CN110751644B (en) | Road surface crack detection method | |
| CN110245697B (en) | Surface contamination detection method, terminal device and storage medium | |
| Feng et al. | Multi‐scale classification network for road crack detection | |
| CN104282027B (en) | Circle detecting method based on Hough transformation | |
| CN111369526B (en) | Multi-type old bridge crack identification method based on semi-supervised deep learning | |
| CN105512622B (en) | A kind of visible remote sensing image sea land dividing method based on figure segmentation and supervised learning | |
| Zhao et al. | Real‐time fabric defect detection based on multi‐scale convolutional neural network | |
| CN111161260A (en) | Hot-rolled strip steel surface defect detection method and device based on deep learning | |
| Li et al. | Pixel-level detection and measurement of concrete crack using faster region-based convolutional neural network and morphological feature extraction | |
| CN115170520A (en) | Metal mesh defect detection method based on structure contrast information lamination | |
| CN112489026A (en) | Asphalt pavement disease detection method based on multi-branch parallel convolution neural network | |
| CN112258495A (en) | Building wood crack identification method based on convolutional neural network | |
| CN113240620B (en) | Highly-adhesive and multi-size brain neuron automatic segmentation method based on point marking | |
| CN104820992B (en) | A kind of remote sensing images Semantic Similarity measure and device based on hypergraph model | |
| CN111445446B (en) | Concrete surface crack detection method based on improved U-net | |
| CN110349119B (en) | Pavement disease detection method and device based on edge detection neural network |
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 |