CN114266716B - Method for detecting leakage water risk of shield tunnel wall based on deep learning - Google Patents
Method for detecting leakage water risk of shield tunnel wall based on deep learning Download PDFInfo
- Publication number
- CN114266716B CN114266716B CN202010967797.8A CN202010967797A CN114266716B CN 114266716 B CN114266716 B CN 114266716B CN 202010967797 A CN202010967797 A CN 202010967797A CN 114266716 B CN114266716 B CN 114266716B
- Authority
- CN
- China
- Prior art keywords
- classification
- risk
- model
- wall surface
- tunnel wall
- 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
- 238000000034 method Methods 0.000 title claims abstract description 52
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 title claims abstract description 45
- 238000013135 deep learning Methods 0.000 title claims abstract description 41
- 238000013145 classification model Methods 0.000 claims abstract description 59
- 230000011218 segmentation Effects 0.000 claims abstract description 53
- 238000012549 training Methods 0.000 claims abstract description 37
- 238000001514 detection method Methods 0.000 claims abstract description 33
- 238000004040 coloring Methods 0.000 claims abstract description 8
- 238000012545 processing Methods 0.000 claims abstract description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 21
- 238000013528 artificial neural network Methods 0.000 claims description 20
- 230000006870 function Effects 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 6
- 238000013526 transfer learning Methods 0.000 claims description 6
- 230000003287 optical effect Effects 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000004519 manufacturing process Methods 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 230000000694 effects Effects 0.000 abstract description 8
- 201000010099 disease Diseases 0.000 abstract description 4
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 abstract description 4
- 238000012423 maintenance Methods 0.000 abstract description 4
- 238000012544 monitoring process Methods 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 7
- 230000003321 amplification Effects 0.000 description 4
- 238000003199 nucleic acid amplification method Methods 0.000 description 4
- 238000004364 calculation method Methods 0.000 description 3
- 238000002790 cross-validation Methods 0.000 description 3
- 238000009826 distribution Methods 0.000 description 3
- 230000007547 defect Effects 0.000 description 2
- 230000000977 initiatory effect Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000003745 diagnosis Methods 0.000 description 1
- 230000036541 health Effects 0.000 description 1
- 238000011835 investigation Methods 0.000 description 1
- 238000002372 labelling Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000002269 spontaneous effect Effects 0.000 description 1
- 239000002352 surface water Substances 0.000 description 1
- 230000007306 turnover Effects 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention provides a method for detecting the leakage risk of a shield tunnel wall surface based on deep learning, which is used for detecting the leakage risk of the shield tunnel wall surface according to a detection image obtained by shooting the shield tunnel wall surface and comprises the following steps: and judging the classification type of the detected image through the classification model to obtain classification type containing no risk, processed risk and unprocessed risk, positioning and coloring the detection image with unprocessed risk in the classification type to obtain a colored label image, outputting the colored label image to provide a warning effect, realizing automatic positioning of the warning monitored image, and training according to the tunnel wall classification data set and the tunnel wall segmentation data set. Therefore, the demand of rapidly increasing tunnel maintenance can be slowed down, and the demand of the actual engineering on the automatic and batched processing of the water leakage disease image identification is met by sending out relevant warning and automatic positioning, so that the water leakage detection of the tunnel wall surface is more accurate and efficient.
Description
Technical Field
The invention relates to a shield tunnel wall surface leakage water risk detection method based on deep learning, and belongs to the technical field of tunnel risk detection.
Background
Along with the vigorous development of the economy in China, the convenience and diversified demands of the city on traffic are higher and higher, and the more subway tunnels and highway river-crossing tunnels are constructed. With the rapid increase of tunnel maintenance requirements, rapid and accurate identification and diagnosis of subway shield tunnel structure defects, particularly leakage water defects, are urgently needed. The health detection of the shield tunnel by using computer vision is a new trend at home and abroad in recent years, and the effect of higher robustness can be realized theoretically with very low cost, but the current identification effect of the water leakage disease image can not meet the requirements of actual engineering.
Deep learning is one of the major breakthroughs in the field of artificial intelligence in recent decades, and has achieved great success in the fields of speech recognition, natural language processing, computer vision, image and video analysis, multimedia and the like, and the maximum difference between the deep learning and the traditional mode recognition mode is that the deep learning is to automatically learn features from big data instead of manually designed features, so that the model has stronger expressive ability and higher efficiency.
The existing tunnel wall surface leakage water risk detection method is not used for manual detection or positioning by adopting a segmentation method, and no automatic implementation of warning and positioning can be realized.
Disclosure of Invention
In order to solve the problems, the invention provides a method for detecting the leakage water risk of the wall surface of a shield tunnel based on deep learning, which adopts the following technical scheme:
the invention provides a method for detecting the leakage risk of a shield tunnel wall surface based on deep learning, which is used for detecting the leakage risk of the shield tunnel wall surface according to a detection image obtained by shooting the shield tunnel wall surface, and is characterized by comprising the following steps: step 1-1, classifying the detected images through a pre-trained classification model to obtain classification categories of the corresponding detected images, wherein the classification categories comprise risk-free, risk-processed and risk-unprocessed; step 1-2, judging classification types; step 1-3, outputting a risk-free result when the classification category is judged to be risk-free and risk is processed; step 1-4, when the classification category is judged to be risk untreated, carrying out water leakage positioning on the detection image with the classification category being risk untreated through a pre-trained segmentation model, and processing the detection image to obtain a colored label image; step 1-5, outputting a coloring label graph to perform early warning, wherein the training process of the classification model and the segmentation model comprises the following steps: step 2-1, manufacturing a tunnel wall surface classification data set by downsampling a pre-acquired tunnel wall surface optical image to a uniform size; step 2-2, respectively splitting the tunnel wall surface classification data set into tunnel wall surface classification data sets which contain risk unprocessed, risk processed and risk-free, marking whether the risk exists or not on the part which is not processed through a segmentation marking tool, and further taking the marked part as the tunnel wall surface segmentation data set; step 2-3, training based on the tunnel wall classification data set to obtain a classification model; and 2-4, training based on the tunnel wall segmentation data set to obtain a classification model.
The method for detecting the leakage water risk of the shield tunnel wall surface based on deep learning provided by the invention can also have the technical characteristics that the method comprises the following sub-steps in the step 2-3: 2-3-1, calling a convolutional neural network and model parameters of the convolutional neural network, wherein the convolutional neural network is obtained by acquiring a large-scale open source data set based on transfer learning; step 2-3-2, modifying the final full connection layer of the convolutional neural network into a global pooling module, a convolutional kernel structure of 1*1 and a classifier as a classification model to be trained; step 2-3-3, calculating a tunnel wall surface classification data set according to a classifier in the classification model to be trained to obtain the classification probability of the classification model to be trained; step 2-3-4, taking the category corresponding to the maximum classification probability as the classification category; step 2-3-5, updating model parameters of the classification model to be trained by using a cross entropy loss function in a random gradient descent mode; 2-3-6, calculating the accuracy of the classification category obtained by the classification model and the model training speed; step 2-3-7, training the classification model to be trained by adjusting super parameters of the classification model to be trained and by a cross verification method; and step 2-3-8, repeating the steps 2-3-3 to 2-3-7, and outputting the model parameter with the highest accuracy and the to-be-trained classification model corresponding to the super parameter as the classification model.
The method for detecting the leakage water risk of the shield tunnel wall surface based on deep learning provided by the invention can also have the technical characteristics that the method comprises the following sub-steps in the steps 2-4: 2-4-1, calling a full convolution neural network and model parameters of the full convolution neural network, wherein the full convolution neural network and the model parameters of the full convolution neural network are obtained by acquiring a large-scale open source data set based on transfer learning; step 2-4-2, removing the last full connection layer of the full convolution neural network, applying model parameters of the full convolution neural network to the full convolution neural network, and judging the category of each pixel through a sigmoid activation function of an output layer so as to obtain a segmentation model to be trained; step 2-4-3, training a segmentation model to be trained through a tunnel wall segmentation data set, and updating model parameters by adopting a random gradient descent strategy through a cross entropy loss function; step 2-4-4, adjusting the super parameters of the segmentation model to be trained to perform cross verification and calculating the accuracy of the segmentation model to be trained corresponding to the current super parameters; and step 2-4-5, repeating the step 2-4-3 and the step 2-4-4, and outputting the super-parameters with highest accuracy and the to-be-trained segmentation model corresponding to the model parameters as segmentation models in the cross verification.
The actions and effects of the invention
According to the method for detecting the leakage water risk of the shield tunnel wall surface based on deep learning, the classification type judgment is carried out on the monitoring image through the classification model, the classification type containing no risk, risk processed and risk unprocessed is obtained, the detection image with the classification type being risk unprocessed is further subjected to water leakage positioning and coloring through the segmentation model to obtain a colored label image, the colored label image is finally output to provide a warning effect and realize automatic positioning of the warning monitoring image, the classification model and the segmentation model manufacture the tunnel wall surface optical image into a tunnel wall surface classification data set through downsampling to a uniform size, the manufactured tunnel wall surface classification data set is divided into three parts including risk unprocessed, risk processed and no risk, and the part with the risk unprocessed is marked as the tunnel wall surface segmentation data set, so that training is carried out. Therefore, the method for detecting the leakage water risk of the shield tunnel wall based on deep learning can slow down the requirement of rapidly increasing tunnel maintenance, and can meet the requirement of actual engineering on the identification of the leakage water disease image by warning the condition that the risk is unprocessed in the monitoring image and automatically positioning the part of the monitoring image that the risk is unprocessed, thereby realizing the automatic learning function more accurately and more efficiently, and realizing the automatic and batched processing of the tunnel wall leakage water detection image.
Drawings
FIG. 1 is a flow chart of a method for shield tunnel wall surface leakage water risk detection based on deep learning in an embodiment of the invention;
FIG. 2 is a schematic diagram of three classification categories in a method for detecting the leakage water risk of a shield tunnel wall surface based on deep learning in an embodiment of the invention;
FIG. 3 is a flow chart of a training classification model of a method for shield tunnel wall surface leakage water risk detection based on deep learning in an embodiment of the invention;
fig. 4 is a flowchart of a training segmentation model of a method for shield tunnel wall surface leakage water risk detection based on deep learning in an embodiment of the invention.
FIG. 5 is a data distribution of a tunnel wall segmentation dataset based on a method for deep learning-based shield tunnel wall leakage risk detection in an embodiment of the present invention;
FIG. 6 is a model parameter case related to a classification model in a method for shield tunnel wall surface leakage water risk detection based on deep learning in an embodiment of the present invention;
FIG. 7 is a model-dependent hyper-parametric situation of a classification model in a method for deep learning-based shield tunnel wall surface leakage risk detection in an embodiment of the present invention;
FIG. 8 is a schematic diagram of test 1 results in a method for detecting the risk of water leakage on the wall surface of a shield tunnel based on deep learning in an embodiment of the invention; and
fig. 9 is a schematic diagram of test 2 results in a method for detecting water leakage risk of a shield tunnel wall surface based on deep learning in an embodiment of the invention.
Detailed Description
In order to make the technical means, creation characteristics, achievement purposes and effects realized by the method easy to understand, the method for detecting the leakage risk of the shield tunnel wall surface based on deep learning is specifically described below with reference to the embodiment and the attached drawings.
< example >
Fig. 1 is a flowchart of a method for detecting the water leakage risk of a shield tunnel wall surface based on deep learning in an embodiment of the invention.
As shown in FIG. 1, the method for detecting the leakage water risk of the wall surface of the shield tunnel based on deep learning specifically comprises the steps 1-1 to 1-5.
Step 1-1, classifying the detected images through a pre-trained classification model to obtain classification categories of the corresponding detected images, wherein the classification categories comprise risk-free, risk-processed and risk-free processing, and then entering step 1-2.
Fig. 2 is a schematic diagram of three classification categories in a method for detecting the leakage water risk of a shield tunnel wall surface based on deep learning in an embodiment of the invention.
In this embodiment, the detected image is an image of water leakage on the wall surface of the shield tunnel, which is obtained by photographing the wall surface of the shield tunnel through an instrument, the initiation of the photographing operation may be an operation instruction to be sent by a person, or may be a spontaneous initiation by performing periodic investigation on a system, and the classification class corresponding to the detected image is obtained by performing forward reasoning on the detected image through a classification model, as shown in fig. 2, the (a) graph of fig. 2 is a detected image with risk unprocessed, the (b) graph of fig. 2 is a detected image with risk processed in the classification class, and the (c) graph of fig. 2 is a detected image with risk free in the classification class.
And step 1-2, judging classification types. When the classification category is judged to be risk-free and risk-processed, the step 1-3 is entered; when the classification category is determined to be at risk, it then proceeds to steps 1-5.
And step 1-3, outputting a risk-free result, and entering an ending state.
In this embodiment, the outputted risk-free result indicates that the wall surface for performing the shield tunnel wall surface water leakage risk detection does not have water leakage, and no treatment is required for the wall surface.
And step 1-4, performing water leakage positioning on the detection images classified into risk-free processing through a pre-trained segmentation model, processing the detection images to obtain a colored label image, and then entering step 1-5.
The coloring label image is an image which can visually locate the leaking position by coloring the detection image according to the inherent label in the image.
The steps of the training process of the classification model and the segmentation model include steps 2-1 to 2-4.
And 2-1, performing downsampling on a pre-acquired tunnel wall surface optical image to a uniform size and performing sample amplification to manufacture a tunnel wall surface classification data set, and then entering the step 2-2.
In this embodiment, the size obtained by downsampling is 432px x 648px.
Fig. 4 is a data distribution of a tunnel wall classification data set based on a method for detecting a shield tunnel wall leakage water risk based on deep learning in an embodiment of the present invention.
Sample amplification of the tunnel wall classified data set is achieved through operations such as symmetry, turnover, noise adding and the like, and data distribution of the tunnel wall classified data set obtained through sample amplification is shown in fig. 4.
And 2-2, respectively splitting the tunnel wall surface classification data set into tunnel wall surface classification data sets which contain risk unprocessed, risk processed and risk-free and are subjected to sample amplification, marking the risk unprocessed part through a segmentation marking tool, further taking the marked part as the tunnel wall surface segmentation data set, and then entering the step 2-3.
In this embodiment, the segmentation marking tool used is Matlab, and the risk area of each image that is not processed in risk is marked and colored to form a generated single-channel image.
And 2-3, training based on the tunnel wall surface classification data set to obtain a classification model, and then entering step 2-4.
Fig. 2 is a flowchart of a training classification model of a method for shield tunnel wall surface leakage water risk detection based on deep learning in an embodiment of the invention.
As shown in FIG. 2, the training process of the classification model in step 2-3 includes steps 2-3-1 to 2-3-8.
And 2-3-1, calling a convolutional neural network and model parameters of the convolutional neural network, wherein the convolutional neural network and the model parameters of the convolutional neural network are obtained by acquiring a large-scale open source data set based on transfer learning, and then entering the step 2-3-2.
Fig. 5 is a model parameter condition related to a classification model in a method for detecting a shield tunnel wall surface leakage water risk based on deep learning in an embodiment of the present invention.
In this embodiment, the convolutional neural network is a keras deep learning framework and a Densenet classification network, the model parameters of the convolutional neural network are parameters of a pre-training model obtained by training on an ImageNet public data set, the model parameter setting of the convolutional neural network is shown in fig. 5, and the accuracy obtained by verifying the tunnel wall classification data set through the classification model under the model parameter setting is 98%.
And 2-3-2, modifying the final full connection layer of the convolutional neural network into a global pooling module, a convolutional kernel structure of 1*1 and a classifier as a classification model to be trained, and then entering the step 2-3-3.
And 2-3-3, calculating a tunnel wall surface classification data set according to a classifier in the classification model to be trained to obtain the classification probability of the classification model to be trained, and then entering the step 2-3-4.
In the embodiment, the classifier in the classification model to be trained is softmax, and n classification classes S are provided k k∈(0,n]The calculation formula of the softmax result is:
where i denotes a certain classification category among k classification categories, g i A value representing the classification category.
And 2-3-4, taking the class corresponding to the maximum classification probability as the classification class, and then entering the step 2-3-5.
And 2-3-5, updating model parameters of the classification model to be trained by using a cross entropy loss function in a random gradient descent mode, and then entering the step 2-3-6.
And 2-3-6, calculating the accuracy of the classification category obtained by the classification model to be trained, and then entering the step 2-3-7.
And 2-3-7, training the classification model to be trained by adjusting the super parameters of the classification model to be trained and by a cross-validation method, and then entering the step 2-3-8.
And step 2-3-8, repeating the steps 2-3-3 to 2-3-7, outputting the model parameters with the highest accuracy and the to-be-trained classification model corresponding to the super parameters as classification models, and then entering an ending state.
Fig. 6 is a relevant model hyper-parameter condition of a classification model in a method for detecting a shield tunnel wall surface leakage water risk based on deep learning in an embodiment of the invention.
The super-parameters were trained at the model parameter settings, and the relevant model super-parameters settings are shown in fig. 6, at which the classification model accuracy was 83%.
And 2-4, training based on the tunnel wall segmentation data set to obtain a classification model, and then entering an ending state.
Fig. 3 is a flowchart of a training segmentation model of a method for shield tunnel wall surface leakage water risk detection based on deep learning in an embodiment of the invention.
As shown in FIG. 3, the training process of the classification model in step 2-4 includes steps 2-4-1 to 2-4-5.
And 2-4-1, calling the full convolution neural network and model parameters of the full convolution neural network, wherein the full convolution neural network and the model parameters of the full convolution neural network are obtained by acquiring a large-scale open source data set based on transfer learning, and then entering the step 2-4-2.
In this embodiment, the invoked full convolutional neural network employs a keras deep learning framework and a deep bv3 segmentation network, and the employed open source data set is an ImageNet data set.
And 2-4-2, removing the last full connection layer of the full convolution neural network, applying the full convolution network model parameters to the full convolution neural network, judging the category of each pixel through a sigmoid activation function of an output layer so as to obtain a segmentation model to be trained, and then entering the step 2-4-3.
And 2-4-3, training the segmentation model to be trained through a tunnel wall segmentation data set, adopting a random gradient descent strategy to update model parameters by using a cross entropy loss function, and then entering the step 2-4-4.
In this embodiment, the tunnel wall segmentation dataset is labeled with a Matlab segmentation labeling tool to color a risk area with risk unprocessed of each picture so as to generate a single-channel picture, and such single-channel picture is converted into a colored label picture, and label names of the corresponding generated single-channel picture and label picture are as follows: A10.png-A10_label.png.
And 2-4-4, adjusting the super parameters of the segmentation model to be trained to perform cross verification, calculating the accuracy of the segmentation model corresponding to the current super parameters, and then entering the step 2-4-5.
And step 2-4-5, repeating the step 2-4-3 and the step 2-4-4, outputting the super parameter with the highest accuracy in the cross verification and the to-be-trained segmentation model corresponding to the model parameter as the segmentation model, and then entering an ending state.
And step 1-5, outputting a coloring label graph so as to perform early warning, and ending the flow.
Fig. 7 is a schematic diagram of test 1 results in a method for detecting water leakage risk of a shield tunnel wall surface based on deep learning in an embodiment of the invention.
Fig. 8 is a schematic diagram of test 2 results in a method for detecting water leakage risk of a shield tunnel wall surface based on deep learning in an embodiment of the invention.
As shown in fig. 7 and 8, fig. 7 and 8 are schematic diagrams of test results of the detected images by the method of shield tunnel wall leakage water risk detection based on deep learning, fig. 7 (a) and 8 (a) are test pictures with risk unprocessed, and fig. 7 (b) and 8 (b) are test pictures with risk unprocessed.
Example operation and Effect
According to the method for detecting the leakage risk of the shield tunnel wall surface based on deep learning, the monitoring image is subjected to classification through the classification model, classification categories including non-risk, risk processed and risk unprocessed are obtained, the detection image with the classification categories being the risk unprocessed is subjected to water leakage positioning and coloring through the segmentation model to obtain the colored label image, the colored label image is output to provide a warning effect and realize automatic positioning of the warning monitoring image, the classification model and the segmentation model enable the tunnel wall surface optical image to be manufactured into a tunnel wall surface classification data set through downsampling to a uniform size, the manufactured tunnel wall surface classification data set is divided into three parts including the risk unprocessed, the risk processed and the risk unprocessed, and the part with the risk unprocessed is marked as the tunnel wall surface segmentation data set, so that training is carried out. Therefore, the method for detecting the leakage water risk of the shield tunnel wall surface based on deep learning can slow down the requirement of rapidly increasing tunnel maintenance, and can enable a detector to process the related warning area in time more quickly by warning the condition that the risk is unprocessed in the monitoring image and automatically positioning the part of the monitoring image which is not processed, thereby meeting the requirement of actual engineering on the identification of the leakage water disease image, realizing the automatic learning function more accurately and more efficiently, and realizing the automatic and batched processing of the leakage water detection image of the tunnel wall surface.
In the embodiment, when training is performed on the classification model, training time and training cost can be reduced by taking a convolutional neural network and removing a final full-connection layer through a mode of model training initialization parameters, so that calculation cost can be reduced by carrying out loss calculation through a cross entropy loss function in a random gradient descending mode, the adjusted classification model can obtain a model with high precision and good generalization performance through a mode of adjusting super parameters, and therefore the model is more reliable and can be better applied to actual engineering, and finally, overfitting can be reduced to a certain extent through a cross-validation mode and effective information can be obtained from limited data as much as possible.
In the embodiment, when training is performed on the segmentation model, the full convolutional neural network is called, the model training initialization parameters are realized by removing the last full connection layer and modifying the sigmoid activation function of the output layer, so that the training time and the training cost are reduced, meanwhile, the method of training super-parameter setting and training sub-module structure adjustment is adjusted, so that a convolutional neural network classification model with higher efficiency, good generalization performance and high precision is obtained, the convolutional neural network classification model is more reliable and can be better applied to practical engineering, and finally, the overfitting can be reduced to a certain extent by a cross-validation method, and as much effective information as possible can be obtained from limited data.
The above examples are only for illustrating the specific embodiments of the present invention, and the present invention is not limited to the description scope of the above examples.
Claims (3)
1. The method for detecting the leakage risk of the wall surface of the shield tunnel based on the deep learning is used for detecting the leakage risk of the wall surface of the shield tunnel according to a detection image obtained by shooting the wall surface of the shield tunnel, and is characterized by comprising the following steps:
step 1-1, classifying the detection images through a pre-trained classification model to obtain classification categories corresponding to the detection images, wherein the classification categories comprise risk-free, risk-processed and risk-unprocessed;
step 1-2, judging the classification category;
step 1-3, outputting a risk-free result when the classification category is judged to be risk-free and risk-processed;
step 1-4, when the classification category is judged to be risk untreated, carrying out water leakage positioning on the detection image with the classification category being risk untreated through a pre-trained segmentation model, and processing the detection image to obtain a color-marked label image;
step 1-5, outputting the coloring label graph to perform early warning,
the training process of the classification model and the segmentation model comprises the following steps:
step 2-1, manufacturing a tunnel wall surface classification data set by downsampling a pre-acquired tunnel wall surface optical image to a uniform size;
step 2-2, respectively splitting the tunnel wall surface classification data set into tunnel wall surface classification data sets which contain risk unprocessed, risk processed and no risk, marking the part with risk unprocessed through a segmentation marking tool, and further taking the marked part as a tunnel wall surface segmentation data set;
step 2-3, training based on the tunnel wall classification data set to obtain the classification model;
and 2-4, training based on the tunnel wall segmentation data set to obtain the segmentation model.
2. The method for detecting the leakage water risk of the wall surface of the shield tunnel based on deep learning according to claim 1, wherein the method comprises the following steps:
wherein, the step 2-3 comprises the following substeps:
2-3-1, calling a convolutional neural network and model parameters of the convolutional neural network, wherein the convolutional neural network is obtained by acquiring a large-scale open source data set based on transfer learning;
step 2-3-2, modifying the final full connection layer of the convolutional neural network into a global pooling module, a convolutional kernel structure of 1*1 and a classifier as a classification model to be trained;
step 2-3-3, calculating the tunnel wall surface classification data set according to the classifier in the classification model to be trained to obtain the classification probability of the classification model to be trained;
step 2-3-4, taking the category corresponding to the maximum classification probability as the classification category;
step 2-3-5, updating the model parameters of the classification model to be trained by using a cross entropy loss function in a random gradient descent mode;
2-3-6, calculating the accuracy of the classification category obtained by the classification model;
step 2-3-7, training the classification model to be trained by adjusting super parameters of the classification model to be trained and by a cross verification method;
and step 2-3-8, repeating the steps 2-3-3 to 2-3-7, and outputting the model parameters with highest accuracy and the to-be-trained classification model corresponding to the super parameters as classification models.
3. The method for detecting the leakage water risk of the wall surface of the shield tunnel based on deep learning as claimed in claim 1, wherein the method comprises the following steps:
wherein, the steps 2-4 comprise the following substeps:
2-4-1, calling a full convolution neural network and the model parameters of the full convolution neural network, wherein the full convolution neural network and the model parameters of the full convolution neural network are obtained by acquiring a large-scale open source data set based on transfer learning;
step 2-4-2, removing the last full connection layer of the full convolution neural network, applying the model parameters of the full convolution neural network to the full convolution neural network, and judging the category of each pixel through a sigmoid activation function of an output layer so as to obtain a segmentation model to be trained;
step 2-4-3, training the segmentation model to be trained through the tunnel wall segmentation data set, and updating model parameters by adopting a random gradient descent strategy through a cross entropy loss function;
step 2-4-4, adjusting the super parameters of the segmentation model to be trained to perform cross verification and calculating the accuracy of the segmentation model to be trained corresponding to the current super parameters;
and 2-4-5, repeating the steps 2-4-3 and 2-4-4, and outputting the super-parameters with the highest accuracy in cross verification and the segmentation model to be trained corresponding to the model parameters as the segmentation model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010967797.8A CN114266716B (en) | 2020-09-15 | 2020-09-15 | Method for detecting leakage water risk of shield tunnel wall based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010967797.8A CN114266716B (en) | 2020-09-15 | 2020-09-15 | Method for detecting leakage water risk of shield tunnel wall based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114266716A CN114266716A (en) | 2022-04-01 |
| CN114266716B true CN114266716B (en) | 2024-04-02 |
Family
ID=80824141
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010967797.8A Active CN114266716B (en) | 2020-09-15 | 2020-09-15 | Method for detecting leakage water risk of shield tunnel wall based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114266716B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116935016B (en) * | 2023-07-28 | 2024-03-08 | 上海济目科技有限公司 | Integrated tunnel construction site risk identification and laser positioning method |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019136946A1 (en) * | 2018-01-15 | 2019-07-18 | 中山大学 | Deep learning-based weakly supervised salient object detection method and system |
| CN110909657A (en) * | 2019-11-19 | 2020-03-24 | 上海勘察设计研究院(集团)有限公司 | Method for identifying apparent tunnel disease image |
| CN111507990A (en) * | 2020-04-20 | 2020-08-07 | 南京航空航天大学 | Tunnel surface defect segmentation method based on deep learning |
| CN111612758A (en) * | 2020-05-19 | 2020-09-01 | 南京航空航天大学 | Tunnel surface defect detection method based on convolutional neural network |
-
2020
- 2020-09-15 CN CN202010967797.8A patent/CN114266716B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2019136946A1 (en) * | 2018-01-15 | 2019-07-18 | 中山大学 | Deep learning-based weakly supervised salient object detection method and system |
| CN110909657A (en) * | 2019-11-19 | 2020-03-24 | 上海勘察设计研究院(集团)有限公司 | Method for identifying apparent tunnel disease image |
| CN111507990A (en) * | 2020-04-20 | 2020-08-07 | 南京航空航天大学 | Tunnel surface defect segmentation method based on deep learning |
| CN111612758A (en) * | 2020-05-19 | 2020-09-01 | 南京航空航天大学 | Tunnel surface defect detection method based on convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| 基于深度学习的地铁隧道衬砌病害检测模型优化;薛亚东;高健;李宜城;黄宏伟;;湖南大学学报(自然科学版);20200721(07);全文 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114266716A (en) | 2022-04-01 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109086756B (en) | Text detection analysis method, device and equipment based on deep neural network | |
| CN110287849B (en) | Lightweight depth network image target detection method suitable for raspberry pi | |
| CN110427875B (en) | Infrared image target detection method based on deep migration learning and extreme learning machine | |
| Siddhartha et al. | COVIDLite: A depth-wise separable deep neural network with white balance and CLAHE for detection of COVID-19 | |
| CN108399406A (en) | The method and system of Weakly supervised conspicuousness object detection based on deep learning | |
| CN112381788B (en) | Part surface defect increment detection method based on double-branch matching network | |
| CN111429418A (en) | Industrial part detection method based on YO L O v3 neural network | |
| Yu et al. | Research of image main objects detection algorithm based on deep learning | |
| CN114998220B (en) | Tongue image detection and positioning method based on improved Tiny-YOLO v4 natural environment | |
| CN112613428B (en) | Resnet-3D convolution cattle video target detection method based on balance loss | |
| JP2020004405A (en) | Information processing method, and information processing device | |
| CN110659601B (en) | Depth full convolution network remote sensing image dense vehicle detection method based on central point | |
| CN115631369A (en) | Fine-grained image classification method based on convolutional neural network | |
| CN112949510A (en) | Human detection method based on fast R-CNN thermal infrared image | |
| CN114266716B (en) | Method for detecting leakage water risk of shield tunnel wall based on deep learning | |
| WO2020119624A1 (en) | Class-sensitive edge detection method based on deep learning | |
| Singh et al. | CNN based approach for traffic sign recognition system | |
| US20220092869A1 (en) | System and method for generating a mask for object instances in an image | |
| CN111160282A (en) | Traffic light detection method based on binary Yolov3 network | |
| CN113705531B (en) | Identification method of alloy powder inclusions based on microscopic imaging | |
| Klusek et al. | Supporting the process of sewer pipes inspection using machine learning on embedded devices | |
| CN115458174A (en) | Method for constructing intelligent diagnosis model of diabetic retinopathy | |
| An | Xception network for weather image recognition based on transfer learning | |
| Lv et al. | Automated nanoparticle count via modified blendmask instance segmentation on sem images | |
| Dehaerne et al. | Ensemble deep learning-based defect classification and detection in sem images |
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 |