CN111369526A - Multi-type old bridge crack identification method based on semi-supervised deep learning - Google Patents
Multi-type old bridge crack identification method based on semi-supervised deep learning Download PDFInfo
- Publication number
- CN111369526A CN111369526A CN202010138991.5A CN202010138991A CN111369526A CN 111369526 A CN111369526 A CN 111369526A CN 202010138991 A CN202010138991 A CN 202010138991A CN 111369526 A CN111369526 A CN 111369526A
- Authority
- CN
- China
- Prior art keywords
- image
- old bridge
- lbp
- fracture
- type
- 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
Links
Images
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
- 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
- G06F18/2155—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the incorporation of unlabelled data, e.g. multiple instance learning [MIL], semi-supervised techniques using expectation-maximisation [EM] or naïve labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- 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
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/187—Segmentation; Edge detection involving region growing; involving region merging; involving connected component labelling
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/40—Analysis of texture
- G06T7/41—Analysis of texture based on statistical description of texture
- G06T7/44—Analysis of texture based on statistical description of texture using image operators, e.g. filters, edge density metrics or local histograms
-
- 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)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Quality & Reliability (AREA)
- Probability & Statistics with Applications (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a multi-type old bridge crack identification method based on semi-supervised deep learning, which comprises the following steps: carrying out self-quotient filtering processing on the original old bridge fracture area image; carrying out binarization processing and connected domain analysis and screening to obtain a fracture image; obtaining a global LBP characteristic of the fracture image by calculating the LBP characteristic of the subarea; automatically labeling the image by using a labeling twin network; and training the fracture type recognition deep convolutional neural network by using the automatically labeled sample set so as to recognize the bridge fracture type. By using the method and the device, the fracture type identification precision and efficiency are improved and the manual marking workload is reduced in the identification of the old bridge fracture.
Description
Technical Field
The invention relates to the technical field of deep learning and computer vision, in particular to a multi-type old bridge crack identification method based on semi-supervised deep learning.
Background
With the rapid development of the social economy and the transportation industry of China, the bridge is built many years ago, and the traffic load of the bridge exceeds the designed load value by multiple times. At the beginning of bridge design, due to the defects in the aspects of process, materials, construction level and the like and the long-term overhaul, a plurality of bridge diseases appear on a plurality of old bridges. Most of the defects are concentrated on cracks, such as cracks caused by increased deflection, cracks caused by surface damage, cracks caused by corrosion of reinforcing steel bars caused by concrete carbonization, and the like.
At present, the detection of bridge cracks still stays on the basis of manual judgment. However, the bridges are numerous, the cost is high due to manual judgment, and the efficiency and the accuracy are difficult to meet the requirement of continuously monitoring a huge number of old bridge facilities.
The method has the remarkable advantages of non-contact, high efficiency, economy and the like based on a computer vision detection technology, and has wide application prospect in the field of automatic identification of various old bridge cracks. Therefore, foreign research institutions have proposed the detection of multiple types of old bridge cracks by using a computer vision method. However, the detection is under natural light conditions, the external environment is interfered more, and the feature selection is not accurate enough, so that the detection robustness is low, and the requirement of practical application cannot be met.
There have also been studies that propose to classify old bridge cracks using deep learning. As is known, deep learning requires a large amount of correct sample data to update parameters in a model, and an optimization method is used for iteration to obtain optimal parameter values. On the one hand, cracks are of various kinds; on the other hand, image data is numerous. These all require manual labeling to determine the true class of the current image. This approach is costly, inefficient, and difficult to apply on a large scale.
Therefore, the existing bridge crack identification technology has the problems of low crack type identification precision, low identification efficiency and large workload of manual marking.
Disclosure of Invention
The invention provides a multi-type old bridge crack identification method based on semi-supervised deep learning, which realizes the identification of old bridge cracks, improves the crack type identification precision and efficiency and reduces the workload of manual marking.
A multi-type old bridge crack identification method based on semi-supervised deep learning comprises the following steps:
firstly, carrying out self-quotient filtering processing on an acquired original old bridge fracture area image to obtain a filtering image;
secondly, carrying out binarization processing on the filtered image by adopting an Otsu threshold value method, and carrying out connected domain analysis screening on a binarization result to obtain a crack image;
dividing the crack image into a plurality of sub-regions, and calculating a binary LBP value for pixel points of each sub-region based on a template with a fixed size;
performing right cyclic shift operation on the obtained binary LBP value, calculating the decimal value of the binary LBP value after each shift operation is completed until all the shift operations are finished, and taking the binary LBP value corresponding to the minimum decimal value as the LBP value of the pixel point;
step five, calculating the LBP histogram of each subregion, and connecting the LBP histograms of the subregions to form a global LBP characteristic;
inputting global LBP characteristics of an original old bridge fracture area image with classification labeling information into a labeling twin network, and automatically labeling a sample set;
and seventhly, taking the automatically marked original old bridge fracture area image sample set as input, training a fracture type recognition deep convolution neural network, and reasoning the old bridge fracture area image by using the trained network so as to obtain the fracture type.
The first step comprises the following steps:
performing Gaussian blur processing on the collected original old bridge fracture area image;
carrying out self-quotient filtering on the original old bridge fracture area image:
wherein, G is an original old bridge fracture area image, I is an image subjected to Gaussian blur processing, and SQI is a filtering image.
The gaussian blur kernel size in step one is 3 x 3 with a gaussian radius of 1.5.
Template size in step three was 5x 5.
The sixth step comprises:
taking global LBP characteristics of an original old bridge region image with classification labeling information as input, and training and labeling a twin network according to a contrast loss function;
and automatically labeling the unclassified sample set images according to the trained labeling twin network.
And the deep convolutional neural network for crack type identification in the seventh step adopts VGG16 and comprises 13 convolutional layers, 5 pooling layers and 3 full-connection layers, wherein the last layer of the full-connection layer comprises M neurons, and M is the number of crack types.
The invention has the beneficial effects that:
1. according to the method, the condition of uneven illumination of the old bridge crack image is removed by adopting an autoquotient filtering method, and the robustness of subsequent crack detection can be improved.
2. The invention adopts an Otsu thresholding and connected domain analysis screening method to obtain the binary image of the crack, thereby removing interference information, obtaining an accurate crack image and improving the overall precision of the method.
3. The improved LBP characteristic is adopted, so that the local characteristic can be accurately expressed, the rotation invariance characteristic is realized, the subsequent automatic marking precision is improved, and the crack type detection precision is further improved.
4. According to the invention, a twin labeling network is adopted on the basis of the extracted LBP characteristics, and a large number of accurate samples are automatically generated for deep learning on the basis of a small number of labeled samples, so that semi-supervised learning is realized, and the workload of manual labeling is reduced.
5. On the basis of the automatically labeled sample set, the invention trains the sample set by adopting a deep convolutional neural network, and finally realizes a computer vision classification system with stronger robustness, stronger generalization capability and no need of parameter adjustment.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention will be described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.
The invention provides a multi-type old bridge crack identification method based on semi-supervised deep learning, and a flow chart is shown in figure 1. The following description will be made by way of specific examples.
The first embodiment is as follows:
the method for identifying the multi-type old bridge cracks based on semi-supervised deep learning comprises the following steps:
step one, carrying out self-quotient filtering processing on an original old bridge fracture area image collected by a camera to obtain a filtering image.
The method is used for identifying the old bridge crack based on the computer vision technology, and a camera is required to be used for collecting an original old bridge crack area image. Specifically, the camera can be used manually to collect the image of the old bridge crack area, and the unmanned aerial vehicle aerial photography method can also be used to obtain the image of the old bridge crack area. In this way, the original old bridge fracture area image G (x, y) can be obtained.
In order to realize accurate classification of the old bridge fracture, the fracture area needs to be intercepted from the whole image, so that the effect of considering interference information is achieved, and the classification accuracy is improved.
However, it is known that ambient lighting is not uniform and variable, and if a fixed threshold segmentation method is used, a wrong fracture region is obtained. Therefore, the invention firstly processes the uneven illumination of the image and adopts the self-quotient filtering to remove the uneven illumination.
According to the imaging principle, the following steps are known: an image produced by a physical optical process can be represented as the product of an illumination component and a reflection component:
G(x,y)＝I(x,y)×R(x,y)
wherein, G (x, y) is an original image, I (x, y) is an illumination component image, R (x, y) is a reflection component image, and x, y are pixel point coordinates.
As can be seen from the above formula, the illumination components of the image are reflection component multiplication relationships, and are difficult to separate in the signalization.
The reflection component image of the object is determined by the material, shape and posture lamp factors of the object and is irrelevant to illumination, so that the illumination nonuniformity phenomenon can be removed by adopting a self-quotient filtering mode, and the following formula is specifically adopted:
R(x,y)＝G(x,y)/I(x,y)
wherein, I (x, y) can be obtained as follows:
firstly, a Gaussian blur algorithm is adopted to perform blur processing on an image. The value rule of the Gaussian radius sigma is as follows: a larger σ means a higher degree of image blur. It should be noted that the gaussian fuzzy algorithm is well known, and is not described herein again, nor is it a protection content of the present invention. In the invention, sigma is 1.5, and the original image is subjected to Gaussian blur processing by a Gaussian convolution kernel by adopting a 3 x 3 template to obtain an image I.
Substituting the formula to obtain:
thus, the image SQI eliminating the illumination phenomenon can be obtained.
And step two, carrying out binarization processing on the filtered image by adopting an Otsu threshold value method, and carrying out connected domain analysis screening on a binarization result to obtain a crack image.
And carrying out binarization on the SQI by adopting an Otsu threshold value method to obtain a binary image of the crack. And further determining a rectangular surrounding frame of the crack area by adopting an outline searching mode, and cutting the large image according to the surrounding frame to obtain the crack image Z with the peripheral interference information removed. Specifically, contour searching is carried out on the image after binarization processing to obtain a fracture interesting region, then, an interference connected region is filtered according to a screening rule, and an accurate fracture image only containing fractures is obtained. The screening rule may be specifically defined according to implementation requirements, for example, the screening may be performed according to the connected domain area, the filling rate, and the aspect ratio, so that noise may be filtered out, and a fracture image may be obtained.
And step three, dividing the crack image into a plurality of sub-regions, and calculating a binary LBP value for pixel points of each sub-region based on a template with a fixed size.
A picture is divided into a plurality of sub-regions, and LBP characteristics are extracted from each pixel point in each sub-region. Further, extracting LBP characteristics of the fracture image Z, and using the characteristics to describe the fracture characteristics:
the entire image is traversed with a fixed-size template, and the present invention uses a 5X5 template, depending on the imaged size of the fracture.
For each template coverage area of 5X5, the initial local LBP features were calculated as follows:
25 points in the template are Pi(i is more than or equal to 0 and less than or equal to 24), wherein the center point of the template is P12If the gray value of the point is N, if the gray values of other points in the template are greater than N, the code value of the point is set to GiIs 1, otherwise is 0. The LBP feature of the center point is obtained as a one-dimensional binary code.
And fourthly, performing right cyclic shift operation on the obtained binary LBP value, calculating the decimal value of the binary LBP value after each shift operation is completed until all the shift operations are finished, and taking the binary LBP value corresponding to the minimum decimal value as the LBP value of the pixel point.
When the same image is rotated, the above feature description mode generates a completely different feature representation, that is, the above mode cannot adapt to the rotation invariance of the image.
Therefore, further processing of the above features is required. The concrete mode is as follows: the obtained binary LBP characteristics are carried outPerforming right cyclic shift operation, wherein the step length of the shift operation is 1, calculating decimal values of the shift operation once, obtaining 24 values until all the shift operations are finished, taking the minimum value, and obtaining the binary characteristic corresponding to the minimum value as the LBP characteristic f with rotation invariancer(p)：
fr(p)＝min(ROR(f(p),i))，0≤i≤23
Where ROR is the shift operation, i represents the shift index value, min represents the minimum value, and f (p) is the decimal value corresponding to the LBP binary code.
And step five, calculating the LBP histogram of each subregion, and connecting the LBP histograms of the subregions to form a global LBP characteristic.
A statistical histogram of the LBP features is built in each sub-region. And integrating the LBP histograms of each subarea to obtain the global LBP characteristic of the image. It should be noted that, the histogram integration method is well known, and is not described herein again, nor is it intended to be protected by the present invention.
And step six, inputting the global LBP characteristics of the original old bridge fracture area image with the classification labeling information into a labeling twin network, and automatically labeling the sample set.
Further, assuming that the old bridge fracture has M categories, there are a few pictures with labeled information in each category currently. The twin network can realize more accurate classification on small samples, and the twin network is adopted to automatically generate multi-class data samples based on the method.
Specifically, the LBP feature of the image is used as two inputs of the twin neural network, and the two inputs are respectively input into the two neural networks, and the weights of the two neural networks are shared. The twin network maps the input to a new space, respectively, by the encoder, forming a feature representation in the new space. And iteratively updating the network weight through a loss function, and evaluating the similarity of the two inputs.
And in the training stage, dividing every two groups of LBP characteristics of the existing labeled information data samples into a plurality of groups to train the twin network. It should be noted that each set of data may be data of the same category or data of different categories. Wherein the loss function of the twin network is a contrast loss function:
wherein the value of Y is 1 or 0. If the model predicts that the inputs are similar, i.e., that the two inputs belong to the same class of old bridge cracks, then Y has a value of 0, otherwise, if the model predicts that the two inputs are different classes of old bridge cracks, then Y is 1. DwEuclidean distance for a twin network:
training the network to obtain a sample similarity evaluation model, and predicting the unlabeled sample through the similarity evaluation model to obtain the labeled information of the unlabeled picture.
Therefore, accurate marking information of all unlabeled samples can be obtained according to the method, and the deep neural network is trained based on the marking information.
The LBP histogram feature of the crack image is utilized, the twin network is adopted to automatically label the acquired bridge area image, and the cost of manual labeling can be greatly reduced.
And seventhly, taking the automatically marked original old bridge fracture area image sample set as input, training a fracture type recognition deep convolution neural network, and reasoning the old bridge fracture area image by using the trained network so as to obtain the fracture type.
And constructing a convolutional neural network model VGG16, wherein the network model comprises 13 convolutional layers, 5 pooling layers and 3 full-connection layers. Wherein, the output of the last full-connection layer is changed into M, which represents M-type cracks. The loss function is:
wherein i is an index corresponding to the training set,
to predict value, yiFor true values, N is the number of samples.
At training time, all samples were run as 8: 1: 1, divided into training set, verification set and test set. The parameters are optimized by adopting a random gradient descent method, the target error is 0.00001, the initial learning rate is 0.001, and in order to prevent the model from falling into a local minimum value, the learning rate needs to be decreased progressively according to a mode of reducing every 100 iterations by 10 times
The implementer needs to observe the network convergence condition and stop training in due time.
According to the method, the condition of uneven illumination of the old bridge crack image is removed by adopting an autoquotient filtering method, and the robustness of subsequent crack detection can be improved. The invention adopts the great-fluid thresholding and contour searching method to obtain the binary image of the crack, thereby removing interference information, obtaining an accurate crack image and improving the overall precision of the method. The improved LBP characteristic is adopted, so that the local characteristic can be accurately expressed, the rotation invariance characteristic is realized, the subsequent automatic marking precision is improved, and the crack type detection precision is further improved. And a twin labeling network is adopted on the basis of the extracted LBP characteristics, and a large number of accurate samples are automatically generated for deep learning on the basis of a small number of labeled samples, so that semi-supervised learning is realized, and the workload of manual labeling is reduced. On the basis of the automatically labeled sample set, a deep convolutional neural network is adopted to train the sample set, and finally, a computer vision classification system with stronger robustness, stronger generalization capability and no need of parameter adjustment is realized.
The above embodiments are merely preferred embodiments of the present invention, which should not be construed as limiting the present invention, and any modifications, equivalents and improvements made within the spirit and principle of the present invention should be included in the scope of the present invention.
Claims (6)
1. A multi-type old bridge crack identification method based on semi-supervised deep learning is characterized by comprising the following steps:
firstly, carrying out self-quotient filtering processing on an acquired original old bridge fracture area image to obtain a filtering image;
secondly, carrying out binarization processing on the filtered image by adopting an Otsu threshold value method, and carrying out connected domain analysis screening on a binarization result to obtain a crack image;
dividing the crack image into a plurality of sub-regions, and calculating a binary LBP value for pixel points of each sub-region based on a template with a fixed size;
performing right cyclic shift operation on the obtained binary LBP value, calculating the decimal value of the binary LBP value after each shift operation is completed until all the shift operations are finished, and taking the binary LBP value corresponding to the minimum decimal value as the LBP value of the pixel point;
step five, calculating the LBP histogram of each subregion, and connecting the LBP histograms of the subregions to form a global LBP characteristic;
inputting global LBP characteristics of an original old bridge fracture area image with classification labeling information into a labeling twin network, and automatically labeling a sample set;
and seventhly, taking the automatically marked original old bridge fracture area image sample set as input, training a fracture type recognition deep convolution neural network, and reasoning the old bridge fracture area image by using the trained network so as to obtain the fracture type.
2. The method for identifying the multi-type old bridge cracks based on semi-supervised deep learning as claimed in claim 1, wherein the first step comprises:
performing Gaussian blur processing on the collected original old bridge fracture area image;
carrying out self-quotient filtering on the original old bridge fracture area image:
wherein, G is an original old bridge fracture area image, I is an image subjected to Gaussian blur processing, and SQI is a filtering image.
3. The method according to claim 1, wherein the size of the gaussian blur kernel in the first step is 3 x 3, and the gaussian radius is 1.5.
4. The method for identifying multi-type old bridge cracks based on semi-supervised deep learning as recited in claim 1, wherein the template size in the third step is 5x 5.
5. The method for identifying the multi-type old bridge cracks based on semi-supervised deep learning as recited in claim 1, wherein the sixth step comprises:
taking global LBP characteristics of an original bridge region image with classification labeling information as input, and training and labeling a twin network according to a contrast loss function;
and automatically labeling the unclassified sample set images according to the trained labeling twin network.
6. The method for identifying the multi-type old bridge cracks based on semi-supervised deep learning of claim 1, wherein the deep convolutional neural network for crack type identification in the seventh step adopts VGG16, which comprises 13 convolutional layers, 5 pooling layers and 3 fully-connected layers, the last layer of the fully-connected layer contains M neurons, and M is the number of crack types.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010138991.5A CN111369526B (en) | 2020-03-03 | 2020-03-03 | Multi-type old bridge crack identification method based on semi-supervised deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010138991.5A CN111369526B (en) | 2020-03-03 | 2020-03-03 | Multi-type old bridge crack identification method based on semi-supervised deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111369526A true CN111369526A (en) | 2020-07-03 |
| CN111369526B CN111369526B (en) | 2023-04-18 |
Family
ID=71212446
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010138991.5A Active CN111369526B (en) | 2020-03-03 | 2020-03-03 | Multi-type old bridge crack identification method based on semi-supervised deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111369526B (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112330670A (en) * | 2020-11-27 | 2021-02-05 | 中国铁建重工集团股份有限公司 | Tunnel drilling robot drilling method and device, drilling robot and medium |
| CN113838058A (en) * | 2021-10-11 | 2021-12-24 | 重庆邮电大学 | Automatic medical image labeling method and system based on small sample segmentation |
| CN114882039A (en) * | 2022-07-12 | 2022-08-09 | 南通透灵信息科技有限公司 | PCB defect identification method applied to automatic PCB sorting process |
| CN117875949A (en) * | 2024-03-13 | 2024-04-12 | 山东交通学院 | Intelligent bridge apparent disease detection method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106910186A (en) * | 2017-01-13 | 2017-06-30 | 陕西师范大学 | A kind of Bridge Crack detection localization method based on CNN deep learnings |
| US20170372470A1 (en) * | 2016-06-27 | 2017-12-28 | Sun Yat-Sen University | Method of separating, identifying and characterizing cracks in 3d space |
| CN108346144A (en) * | 2018-01-30 | 2018-07-31 | 哈尔滨工业大学 | Bridge Crack based on computer vision monitoring and recognition methods automatically |
| CN109871825A (en) * | 2019-03-13 | 2019-06-11 | 华南理工大学 | A kind of portrait identification method based on improved local 2D pattern |
| CN110135360A (en) * | 2019-05-17 | 2019-08-16 | 北京深醒科技有限公司 | A kind of parking stall recognition methods based on local binary patterns and support vector machines |
-
2020
- 2020-03-03 CN CN202010138991.5A patent/CN111369526B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170372470A1 (en) * | 2016-06-27 | 2017-12-28 | Sun Yat-Sen University | Method of separating, identifying and characterizing cracks in 3d space |
| CN106910186A (en) * | 2017-01-13 | 2017-06-30 | 陕西师范大学 | A kind of Bridge Crack detection localization method based on CNN deep learnings |
| CN108346144A (en) * | 2018-01-30 | 2018-07-31 | 哈尔滨工业大学 | Bridge Crack based on computer vision monitoring and recognition methods automatically |
| CN109871825A (en) * | 2019-03-13 | 2019-06-11 | 华南理工大学 | A kind of portrait identification method based on improved local 2D pattern |
| CN110135360A (en) * | 2019-05-17 | 2019-08-16 | 北京深醒科技有限公司 | A kind of parking stall recognition methods based on local binary patterns and support vector machines |
Non-Patent Citations (1)
| Title |
|---|
| 简丽琼;: "选择局部二值模式的纹理识别" * |
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112330670A (en) * | 2020-11-27 | 2021-02-05 | 中国铁建重工集团股份有限公司 | Tunnel drilling robot drilling method and device, drilling robot and medium |
| CN113838058A (en) * | 2021-10-11 | 2021-12-24 | 重庆邮电大学 | Automatic medical image labeling method and system based on small sample segmentation |
| CN113838058B (en) * | 2021-10-11 | 2024-03-19 | 重庆邮电大学 | Automatic medical image labeling method and system based on small sample segmentation |
| CN114882039A (en) * | 2022-07-12 | 2022-08-09 | 南通透灵信息科技有限公司 | PCB defect identification method applied to automatic PCB sorting process |
| CN114882039B (en) * | 2022-07-12 | 2022-09-16 | 南通透灵信息科技有限公司 | PCB defect identification method applied to automatic PCB sorting process |
| CN117875949A (en) * | 2024-03-13 | 2024-04-12 | 山东交通学院 | Intelligent bridge apparent disease detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111369526B (en) | 2023-04-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111369526B (en) | Multi-type old bridge crack identification method based on semi-supervised deep learning | |
| CN110264448B (en) | Insulator fault detection method based on machine vision | |
| CN111028217A (en) | Image crack segmentation method based on full convolution neural network | |
| CN108346144B (en) | Automatic bridge crack monitoring and identifying method based on computer vision | |
| CN108830196A (en) | Pedestrian detection method based on feature pyramid network | |
| CN112381788B (en) | Part surface defect increment detection method based on double-branch matching network | |
| CN111860459B (en) | Gramineae plant leaf pore index measurement method based on microscopic image | |
| CN106340016A (en) | DNA quantitative analysis method based on cell microscope image | |
| CN108932518B (en) | Shoe print image feature extraction and retrieval method based on visual bag-of-words model | |
| CN113673509B (en) | Instrument detection classification method based on image text | |
| CN111914902B (en) | Traditional Chinese medicine identification and surface defect detection method based on deep neural network | |
| CN112365497A (en) | High-speed target detection method and system based on Trident Net and Cascade-RCNN structures | |
| CN113221956B (en) | Target identification method and device based on improved multi-scale depth model | |
| CN111968171A (en) | Aircraft oil quantity measuring method and system based on artificial intelligence | |
| CN113435407B (en) | Small target identification method and device for power transmission system | |
| Yusof et al. | Automated asphalt pavement crack detection and classification using deep convolution neural network | |
| CN115631472B (en) | Intelligent detection method for pedestrian intrusion on expressway | |
| CN110751644A (en) | Road surface crack detection method | |
| CN116188756A (en) | Instrument angle correction and indication recognition method based on deep learning | |
| CN116953006A (en) | Casting material scanning electron microscope image defect identification and quantification method | |
| CN115170520A (en) | Metal mesh defect detection method based on structure contrast information lamination | |
| CN117522864B (en) | European pine plate surface flaw detection method based on machine vision | |
| CN110349119B (en) | Pavement disease detection method and device based on edge detection neural network | |
| CN111950556A (en) | License plate printing quality detection method based on deep learning | |
| Divya et al. | Segmentation of Defected Regions in Leaves using K-Means and OTSU's Method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| CB02 | Change of applicant information | ||
| CB02 | Change of applicant information |
Address after: 101121 1-3 / F, 21 Liyuan North Street, Tongzhou District, Beijing Applicant after: China Construction Second Engineering Bureau Civil Engineering Group Co.,Ltd. Address before: 101121 1-3 / F, 21 Liyuan North Street, Tongzhou District, Beijing Applicant before: CHINA CONSTRUCTION SECOND ENGINEERING BUREAU INFRASTRUCTURE CONSTRUCTION INVESTMENT Ltd. |
|
| GR01 | Patent grant | ||
| GR01 | Patent grant |