CN111523543A - Tunnel surface defect positioning method based on learning - Google Patents
Tunnel surface defect positioning method based on learning Download PDFInfo
- Publication number
- CN111523543A CN111523543A CN202010319611.8A CN202010319611A CN111523543A CN 111523543 A CN111523543 A CN 111523543A CN 202010319611 A CN202010319611 A CN 202010319611A CN 111523543 A CN111523543 A CN 111523543A
- Authority
- CN
- China
- Prior art keywords
- defect
- positioning
- data
- model
- tunnel surface
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/24—Aligning, centring, orientation detection or correction of the image
- G06V10/245—Aligning, centring, orientation detection or correction of the image by locating a pattern; Special marks for positioning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/084—Backpropagation, e.g. using gradient descent
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Multimedia (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a tunnel surface defect positioning method based on deep learning, which comprises the following specific steps of S1: acquiring data, namely acquiring tunnel surface data; s2: constructing a data set, namely constructing a tunnel surface defect positioning data set by using the acquired tunnel surface data; s3: constructing a model, namely constructing a tunnel surface defect positioning model by utilizing a deep learning technology; s4: training a model, inputting the acquired data set into a tunnel surface defect positioning model constructed by S3, and training the model; s5: and (4) defect positioning, namely positioning the defects by using the tunnel surface defect positioning model trained in S4 with new data as input. The invention provides an efficient automatic method for positioning the defects on the surface of the tunnel, which not only can reduce the labor intensity of professionals and improve the efficiency, but also can more effectively avoid false detection and missing detection caused by human subjectivity, and provides an efficient means for the maintenance and management of the tunnel.
Description
Technical Field
The invention relates to the field of intelligent detection, in particular to a tunnel surface defect positioning method based on deep learning.
Background
Nowadays, the traffic transportation construction business of China is greatly developed, and due to the particularity of the terrain structure, the tunnel structure is widely applied to construction, especially the underground rail transit industry. In the operation process of the tunnel structure, defects of different degrees are inevitably generated due to surrounding construction, long construction time, untimely maintenance, subway vehicle operation vibration and the like. The existence of defects will certainly cause great threat to the tunnel operation safety and the safety of people's lives and properties. For this reason, a method is needed for effective defect localization.
The tunnel surface detection in the traditional mode is mainly completed by manual inspection in a skylight period. The manual detection often has a plurality of defects such as high labor intensity, low detection efficiency, poor precision, high labor cost and the like. At present, the development of artificial intelligence technology is fierce, and deep learning technology is taken as a branch of the artificial intelligence technology, which has been successfully applied in many industrial fields, and fully explains the great potential of the technology.
Therefore, the invention applies the deep learning technology to the defect positioning of the tunnel surface, provides the tunnel surface defect positioning method based on the deep learning, saves the labor cost and improves the efficiency and the precision of the defect positioning.
Disclosure of Invention
The invention aims to solve the technical problems of high labor intensity, low detection efficiency, poor precision and high labor cost in tunnel surface defect detection in the prior art by providing a tunnel surface defect positioning method based on deep learning.
In order to solve the technical problems, the technical scheme of the invention is as follows: the method for positioning the tunnel surface defects based on deep learning has the innovation points that: the method specifically comprises the following steps:
s1: data acquisition: collecting tunnel surface data by using a data collecting device;
s2: and (3) construction of a data set: constructing a tunnel surface defect positioning data set by using the acquired tunnel surface data;
s3: constructing a model: constructing a tunnel surface defect positioning model by utilizing a deep learning technology;
s4: training of the model: inputting the acquired data set into the tunnel surface defect positioning model constructed in the step S3, and training the model;
s5: defect positioning: and taking new data as input, and positioning the tunnel surface defects by using the trained tunnel surface defect positioning model.
Further, the data acquisition device in step S1 includes a high-resolution line camera, the data acquisition device needs to illuminate through a high-luminance line light source to shoot image data on the surface of the tunnel to acquire the data when acquiring the data, the data acquired by the data acquisition device is a shot picture of the surface of the tunnel, the acquisition device runs on a track at a constant speed in the acquisition process, the data acquired by the line camera is cut at regular time so as to obtain image data with a proper resolution, and the cutting width is determined according to the storage space of the device and the running speed of the device.
Further, the specific method for constructing the tunnel surface defect location data set in step S2 is as follows:
(1) processing the photo collected in the step S1;
(2) and marking defects of the processed photos.
Further, the processing method for the acquired photo in step (1) includes denoising, filtering, clipping, resolution adjustment and brightness adjustment.
Further, the specific method for labeling the defects of the processed photo in the step (2) is as follows: and marking the defective part in the photo by using a rectangle surrounding the defect, marking the obtained information of each defect as a 5-dimensional vector including whether the defect is the defect, the coordinate of the defect and the size of the defect, and not marking the photo on the surface of the tunnel without the defect.
Further, the specific steps of constructing the deep learning model in step S3 are as follows:
(1) selecting a feature extraction network: the feature extraction network converts an input tunnel surface picture into a feature matrix (feature map) of a high-dimensional space;
(2) constructing a defect detection model: the constructed defect detection model is a single-class target detection network, is used for detecting and positioning defects, comprises 16 convolution layers and 2 full-connection layers, and is used as follows: measuring 3 target frames with different length-width ratios in each dimension direction of a feature diagram output by the feature extraction network in the channel direction, classifying whether the target frames contain defects or not by adopting an SVM classifier, inputting each target frame into an auto-encoder to perform regression on the target frames to obtain a more accurate detection result, and taking the classified result and the regression result as the output of the network for detecting and positioning the defects;
(3) design of loss: the loss is constructed as L ═ Lcls+LposWherein L isclsA loss function representing the existence of defect-free classification and responsible for the accuracy of the classification; l isposThe loss function, which represents the defect location, is responsible for the accuracy of the defect location.
Further, the method for training the model in step S4 includes: inputting the data set obtained in the step S2 into the defect positioning model constructed in the step S3, selecting hyper-parameters to train a network model, saving one model for each iteration for a certain number of times, testing all finally obtained models on the other part of reserved data set which does not participate in the training, and saving the model with the highest testing accuracy for the step S5, wherein the condition for stopping the training of the models is that the loss is reduced to a specific order of magnitude or the training reaches the maximum iteration number, and the condition for selecting the optimal model is that the model reaches the highest accuracy on the test data.
Further, the specific step of defect location in step S5 is:
(1) collecting and processing new data: the acquisition of new data is the same as the data acquisition operation in step S1, and the new data processing is in accordance with the data processing operation in step S2;
(2) and (3) positioning the defects: inputting the collected new data into the defect positioning model trained in step S4, and the defect positioning model evaluates the input data and outputs a defect positioning result.
Further, the defect positioning result is an n × 5 matrix, where n is related to a network structure and a data size of an input network, and 5 denotes (cls, x, y, w, h), where cls indicates whether a detection result includes a defect or not, and (x, y, w, h) denotes position information of the defect, the collected new data is input to a model trained in S4, and then an output matrix is traversed, when a confidence of cls is greater than a specific threshold θ, it denotes that a corresponding position includes a defect, and if all are less than θ, it denotes that the input data does not include a defect.
Compared with the prior art, the invention has the following beneficial effects:
the tunnel surface defect positioning method based on deep learning provides an efficient automatic method for positioning the tunnel surface defects, not only can reduce the labor intensity of professionals and improve the efficiency, but also can more effectively avoid false detection and missing detection caused by human subjectivity, and provides an efficient means for maintenance and management of tunnels.
Drawings
In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings needed to be used in the embodiments are briefly described below, and it is obvious that the drawings in the following description are only some embodiments described in the present invention, and it is obvious for those skilled in the art that other drawings can be obtained according to these drawings without creative efforts.
Fig. 1 is a flowchart of a tunnel surface defect positioning method based on deep learning according to the present invention.
Fig. 2 is a schematic diagram of a collection device employed in an embodiment of the present invention.
FIG. 3 is an exemplary diagram of an original image after cropping, brightness adjustment and labeling according to the present invention.
Fig. 4 is a diagram of a network architecture employed by an embodiment of the present invention.
FIG. 5 is a diagram illustrating a defect location result according to an embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to the following drawings and specific embodiments, it being understood that the specific embodiments described herein are merely illustrative of the invention and are not to be construed as limiting the invention.
The invention provides a tunnel surface defect positioning method based on deep learning, which aims to solve the problem of positioning tunnel surface defects, save labor cost and improve efficiency and precision of defect positioning. The specific flow chart is shown in fig. 1, and the implementation of the defect positioning method includes the following steps:
s1: data acquisition: collecting tunnel surface data by using a data collecting device; the data acquisition device comprises a high-resolution linear array camera, a specific structure diagram of the data acquisition device is shown in fig. 2, the data acquisition device comprises a linear array camera 1 provided with 8 12288 x1 linear array cameras, 8 high-brightness light sources 2, an industrial personal computer 3, a support 4 and moving platforms 5 and 6 which are tracks, when the data acquisition device acquires data, the data acquisition device needs to illuminate through the high-brightness linear light sources to shoot image data on the surface of a tunnel for data acquisition, the data acquired by the data acquisition device is a shot picture of the surface of the tunnel, the data acquired by the linear array cameras are cut at regular time to obtain image data with the resolution of 12288 x n, n is determined according to the memory space of equipment and the running speed of the equipment, and n is 18000, namely the acquired original image resolution is 12288 x 18000.
S2: and (3) construction of a data set: the method comprises the following steps of constructing a tunnel surface defect positioning data set by using collected tunnel surface data, and specifically comprises the following steps:
(1) processing the photo collected in step S1: processing the collected photos comprises denoising, filtering, cutting, resolution adjustment and brightness adjustment; the resolution of the collected original image is too high, the original image is cropped to be an image with the size of 512 x 375 due to the performance limit of equipment and the feasibility of implementation of subsequent steps, and the brightness of the collected image is too dark due to the limited light input amount of the linear array camera.
(2) And (3) marking defects of the processed photos: the specific method for marking the defects of the processed photo comprises the following steps: the method comprises the steps of marking the defective part in a photo by using a rectangle surrounding the defect, marking each defect by using a 5-dimensional vector, wherein the vector comprises the defect, the coordinate of the defect and the size of the defect, and not marking the photo on the surface of the tunnel without the defect.
S3: constructing a model: constructing a tunnel surface defect positioning model by utilizing a deep learning technology; the specific steps for constructing the deep learning model are as follows:
(1) selecting a feature extraction network, wherein the feature extraction network can convert an input tunnel surface picture into a feature matrix (feature map) of a high-dimensional space, and the feature extraction network selects VGG 16;
(2) constructing a defect detection model: the constructed defect detection model is a single-class target detection network and is used for detecting and positioning defects, as shown in fig. 4, the defect detection model comprises 16 convolutional layers and 2 full-connection layers, and the defect detection model is used as follows: measuring 3 target frames with different length-width ratios in each dimension direction of a VGG16 feature diagram output by a feature extraction network in a channel direction, classifying whether the target frames contain defects or not by adopting an SVM classifier, inputting each target frame into an auto-encoder to perform regression on the target frames to obtain a more accurate detection result, and taking the classification result and the regression result as the output of the network for detecting and positioning the defects;
(3) design of loss: the loss is constructed as L ═ Lcls+LposWherein L isclsA loss function representing the existence of defect-free classification and responsible for the accuracy of the classification; l isposIndication lackThe loss function of the defect location is responsible for the accuracy of the defect location. L in the inventionpos1-IOU, wherein the IOU is an intersection ratio, namely the ratio of the intersection area of the prediction result and the position of the real target to the combination area of the two.
S4: training of the model: and inputting the acquired data set into the tunnel surface defect positioning model constructed in the step S3, and training the model. The specific method for training the model comprises the following steps: inputting the data set obtained in the step S2 into the defect positioning model constructed in the step S3, selecting hyper-parameters to train a network model, saving one model for each iteration for a certain number of times, testing all finally obtained models on the other part of reserved data set which does not participate in the training, and saving the model with the highest testing accuracy for the step S5, wherein the condition for stopping the training of the models is that the loss is reduced to a specific order of magnitude or the training reaches the maximum iteration number, and the condition for selecting the optimal model is that the model reaches the highest accuracy on the test data.
In the embodiment of the invention, the pre-trained weight on ImageNet is used for initializing the layer related to VGG16 in the network, the other layers adopt a random initialization strategy, the initial learning rate is 0.01, the attenuation coefficient is 0.9, the maximum iteration time is 500000, the learning rate adopts an exponential attenuation strategy after 100 iterations, the current model is stored, and when the training Loss is reduced to 10-5And stopping training when the magnitude order is reached, testing all the stored models, reserving the model with the highest testing accuracy as a final training result for the step S5, wherein in the embodiment, the equipment used for training is Dell XPS8930, the CPU is i7-8700K, the GPU is GTX1080, the memory is 16G, and the training time is about 20 hours.
S5: defect positioning: the method comprises the following steps of taking new data as input, and positioning the tunnel surface defects by using a trained tunnel surface defect positioning model, wherein the defect positioning comprises the following specific steps:
(1) collecting and processing new data: the acquisition of new data is the same as the data acquisition operation in step S1, and the new data processing is in accordance with the data processing operation in step S2;
(2) and (3) positioning the defects: inputting the collected new data into the defect localization model trained in step S4, where the defect localization model evaluates the input data and outputs a defect localization result, and the specific defect localization result of the present invention is shown in fig. 5.
The original output result of the defect positioning network is an n x 5 matrix, wherein n is related to the network structure and the data size of the input network, 5 represents (cls, x, y, w, h), wherein cls indicates whether the detection result contains a defect or not, and (x, y, w, h) represents the position information of the defect, the collected new data is input into a model trained in S4, the output matrix is traversed, when the confidence coefficient of cls is greater than a specific threshold theta, the corresponding position contains a defect, and if all the confidence coefficients are less than theta, the input data does not contain the defect. According to the invention, the result of defect positioning can be obtained by taking 0.5 as theta and taking 576 as n, and performing traversal judgment on the output 576 × 5 matrix and displaying the result on the original image.
The tunnel surface defect positioning method based on deep learning provides an efficient automatic method for positioning the tunnel surface defects, reduces the labor intensity of professionals, improves the efficiency, can effectively avoid false detection and missed detection caused by human subjectivity, and provides an efficient means for maintenance and management of the tunnel.
It should be noted that, in this document, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising an … …" does not exclude the presence of other like elements in a process, method, article, or apparatus that comprises the element.
Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a mobile terminal (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method according to the embodiments of the present invention.
While the present invention has been described with reference to the embodiments illustrated in the drawings, the present invention is not limited to the embodiments, which are illustrative rather than restrictive, and it will be apparent to those skilled in the art that many more modifications and variations can be made without departing from the spirit of the invention and the scope of the appended claims.
Claims (9)
1. A tunnel surface defect positioning method based on deep learning is characterized in that: the method specifically comprises the following steps:
s1: data acquisition: collecting tunnel surface data by using a data collecting device;
s2: and (3) construction of a data set: constructing a tunnel surface defect positioning data set by using the acquired tunnel surface data;
s3: constructing a model: constructing a tunnel surface defect positioning model by utilizing a deep learning technology;
s4: training of the model: inputting the acquired data set into the tunnel surface defect positioning model constructed in the step S3, and training the model;
s5: defect positioning: and taking new data as input, and positioning the tunnel surface defects by using the tunnel surface defect positioning model trained in S4.
2. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 1, wherein: the data acquisition device in the step S1 includes a high-resolution linear array camera, and the data acquisition device illuminates the surface of the tunnel to be shot through a high-brightness light source when acquiring data, so as to acquire image data, the data acquired by the data acquisition device is a shot picture of the surface of the tunnel, and the data acquired by the linear array camera is cut at regular time according to the performance of the equipment and the storage space, so as to obtain image data with proper resolution.
3. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 1, wherein: the specific method for constructing the tunnel surface defect location data set in step S2 is as follows:
(1) processing the photo data collected in step S1;
(2) and marking defects of the processed photos.
4. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 3, wherein: the processing of the collected picture in the step (1) comprises denoising, filtering, clipping, resolution adjustment and brightness adjustment.
5. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 3, wherein: the specific method for labeling the defects of the processed photo in the step (2) is as follows: and marking the defective part in the photo by using a rectangle surrounding the defect, marking the obtained information of each defect as a 5-dimensional vector including whether the defect is the defect, the coordinate of the defect and the size of the defect, and not marking the photo on the surface of the tunnel without the defect.
6. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 1, wherein: the specific steps of constructing the deep learning model in step S3 are as follows:
(1) selecting a feature extraction network: the feature extraction network can convert an input tunnel surface picture into a feature matrix (feature map) of a high-dimensional space;
(2) constructing a defect detection model: the constructed defect detection model is a single-class target detection network and is used for detecting and positioning defects, and the defect detection model comprises 16 convolution layers and 2 full-connection layers. The use of the defect detection model is as follows: measuring 3 target frames with different length-width ratios in each dimension direction of a feature diagram output by the feature extraction network in the channel direction, classifying whether the target frames contain defects or not by adopting an SVM classifier, inputting each target frame into an auto-encoder to perform regression on the target frames to obtain a more accurate detection result, and taking the classified result and the regression result as the output of the network for detecting and positioning the defects;
(3) design of loss: the loss is constructed as L ═ Lcls+LposWherein L isclsA loss function representing the existence of defect-free classification and responsible for the accuracy of the classification; l isposThe loss function, which represents the defect location, is responsible for the accuracy of the defect location.
7. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 1, wherein: the method for training the model in step S4 includes: inputting the data set obtained in the step S2 into the defect localization model constructed in the step S3, selecting hyper-parameters to train a network model, saving one model for each iteration for a certain number of times, testing all finally obtained models on the other part of the reserved data set which does not participate in the training, and saving the model with the highest test accuracy for the step S5, wherein the condition for stopping training the models is that loss is reduced to a specific order of magnitude or the training reaches the maximum iteration number, and the condition for selecting the optimal model is that the model reaches the highest accuracy on the test data.
8. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 1, wherein: the specific step of defect location in step S5 is:
(1) collecting and processing new data: the acquisition of new data is the same as the data acquisition operation in step S1, and the new data processing is in accordance with the data processing operation in step S2;
(2) and (3) positioning the defects: inputting the collected new data into the defect positioning model trained in step S4, and the defect positioning model evaluates the input data and outputs a defect positioning result.
9. The method for positioning the tunnel surface defect based on the deep learning as claimed in claim 8, wherein: the defect positioning result is an n × 5 matrix, wherein n is related to a network structure and the size of data input into the network, and 5 represents (cls, x, y, w, h), wherein cls indicates whether a detection result contains a defect or not, and (x, y, w, h) represents position information of the defect, the collected new data is input into a model trained in S4, then the output matrix is traversed, when the confidence coefficient of cls is greater than a specific threshold theta, it represents that a corresponding position contains a defect, and if all the confidence coefficients are less than theta, it represents that the input data does not contain a defect.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010319611.8A CN111523543A (en) | 2020-04-21 | 2020-04-21 | Tunnel surface defect positioning method based on learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010319611.8A CN111523543A (en) | 2020-04-21 | 2020-04-21 | Tunnel surface defect positioning method based on learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111523543A true CN111523543A (en) | 2020-08-11 |
Family
ID=71903280
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010319611.8A Pending CN111523543A (en) | 2020-04-21 | 2020-04-21 | Tunnel surface defect positioning method based on learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111523543A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112113978A (en) * | 2020-09-22 | 2020-12-22 | 成都国铁电气设备有限公司 | Vehicle-mounted tunnel defect online detection system and method based on deep learning |
| CN113092494A (en) * | 2021-03-25 | 2021-07-09 | 中车青岛四方车辆研究所有限公司 | Inspection robot and intelligent detection method for train tunnel structure diseases |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105548205A (en) * | 2016-01-28 | 2016-05-04 | 北京工业大学 | Tunnel surface defect positioning method and positioning system |
| CN109374631A (en) * | 2018-09-30 | 2019-02-22 | 中国铁道科学研究院集团有限公司铁道建筑研究所 | A kind of tunnel state evaluating method |
| CN110726726A (en) * | 2019-10-30 | 2020-01-24 | 中南大学 | Quantitative detection method and system for tunnel forming quality and defects thereof |
| CN110992349A (en) * | 2019-12-11 | 2020-04-10 | 南京航空航天大学 | Underground pipeline abnormity automatic positioning and identification method based on deep learning |
-
2020
- 2020-04-21 CN CN202010319611.8A patent/CN111523543A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105548205A (en) * | 2016-01-28 | 2016-05-04 | 北京工业大学 | Tunnel surface defect positioning method and positioning system |
| CN109374631A (en) * | 2018-09-30 | 2019-02-22 | 中国铁道科学研究院集团有限公司铁道建筑研究所 | A kind of tunnel state evaluating method |
| CN110726726A (en) * | 2019-10-30 | 2020-01-24 | 中南大学 | Quantitative detection method and system for tunnel forming quality and defects thereof |
| CN110992349A (en) * | 2019-12-11 | 2020-04-10 | 南京航空航天大学 | Underground pipeline abnormity automatic positioning and identification method based on deep learning |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112113978A (en) * | 2020-09-22 | 2020-12-22 | 成都国铁电气设备有限公司 | Vehicle-mounted tunnel defect online detection system and method based on deep learning |
| CN113092494A (en) * | 2021-03-25 | 2021-07-09 | 中车青岛四方车辆研究所有限公司 | Inspection robot and intelligent detection method for train tunnel structure diseases |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102166458B1 (en) | Defect inspection method and apparatus using image segmentation based on artificial neural network | |
| CN113436169B (en) | Industrial equipment surface crack detection method and system based on semi-supervised semantic segmentation | |
| CN112598672A (en) | Pavement disease image segmentation method and system based on deep learning | |
| CN111784685A (en) | Power transmission line defect image identification method based on cloud edge cooperative detection | |
| CN109934163A (en) | A kind of aerial image vehicle checking method merged again based on scene priori and feature | |
| CN110909666A (en) | Night vehicle detection method based on improved YOLOv3 convolutional neural network | |
| CN112381175A (en) | Circuit board identification and analysis method based on image processing | |
| CN112686833A (en) | Industrial product surface defect detecting and classifying device based on convolutional neural network | |
| CN111523543A (en) | Tunnel surface defect positioning method based on learning | |
| CN110991447B (en) | Train number accurate positioning and identifying method based on deep learning | |
| CN114596278A (en) | Method and device for detecting hot spot defects of photovoltaic panel of photovoltaic power station | |
| CN113788051A (en) | Train on-station running state monitoring and analyzing system | |
| CN111597939B (en) | High-speed rail line nest defect detection method based on deep learning | |
| CN116597270A (en) | Road damage target detection method based on attention mechanism integrated learning network | |
| CN113298767A (en) | Reliable go map recognition method capable of overcoming light reflection phenomenon | |
| CN116071294A (en) | Optical fiber surface defect detection method and device | |
| CN111881914B (en) | License plate character segmentation method and system based on self-learning threshold | |
| CN116596895B (en) | Substation equipment image defect identification method and system | |
| CN113762247A (en) | Road crack automatic detection method based on significant instance segmentation algorithm | |
| CN115908952B (en) | High-speed railway tunnel fixture detection method based on improved YOLOv5 algorithm | |
| CN112329550A (en) | Weak supervision learning-based disaster-stricken building rapid positioning evaluation method and device | |
| CN115631197B (en) | Image processing method, device, medium, equipment and system | |
| CN115223114A (en) | End-to-end vehicle attitude estimation method based on bidirectional fusion feature pyramid | |
| CN111382645A (en) | Method and system for identifying expired buildings in electronic map | |
| CN113343977B (en) | Multipath automatic identification method for container terminal truck collection license plate |
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 | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200811 |