CN113516179A - Method and system for identifying water leakage performance of underground infrastructure - Google Patents
Method and system for identifying water leakage performance of underground infrastructure Download PDFInfo
- Publication number
- CN113516179A CN113516179A CN202110700803.8A CN202110700803A CN113516179A CN 113516179 A CN113516179 A CN 113516179A CN 202110700803 A CN202110700803 A CN 202110700803A CN 113516179 A CN113516179 A CN 113516179A
- Authority
- CN
- China
- Prior art keywords
- image
- water leakage
- underground infrastructure
- leakage
- water
- 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
-
- 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
Abstract
The invention relates to a method and a system for identifying the water leakage performance of underground infrastructure. The method comprises the following steps: acquiring underground infrastructure water leakage images under various environmental conditions; marking the underground infrastructure water leakage image to generate a marked underground infrastructure water leakage image; training a convolutional neural network according to the underground infrastructure water leakage image and the marked underground infrastructure water leakage image, and constructing a computer vision prediction model; and acquiring an underground infrastructure water leakage image to be predicted under any environmental condition, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted by utilizing the computer vision prediction model. The invention can improve the identification precision in complex and various underground environments.
Description
Technical Field
The invention relates to the crossing field of civil engineering structure abnormity detection and machine learning technology, in particular to a method and a system for identifying the water leakage performance of underground infrastructure.
Background
With the vigorous development of Chinese economy, the travel demand of people is greatly increased, resulting in the massive construction of underground infrastructures such as subways, tunnels, pipe galleries and the like. Underground infrastructure located in a complex geological section inevitably causes damage to tunnel walls due to leakage water caused by long-term use, and may last for months or even years, so that developing a reliable method for identifying the leakage water property of the tunnel is crucial to ensuring the safety of the tunnel. In the past few years, many empirical and semi-empirical models have been proposed to solve this problem, but the engineering is somewhat challenging due to the limitations of many internal and external factors, such as the complexity of the acquired image data, the insufficient expression capability of the adopted computer vision model, and so on. Therefore, a reasonable model is urgently needed to be established so as to adapt to complex working conditions and carry out high-precision water leakage property identification on the complex working conditions.
In recent years, with the rapid development of computer science, the real-time monitoring of the performance of underground infrastructure by using computer vision technology has attracted much attention. Convolutional neural networks are most widely applied to computer vision, because convolutional layers can perform hierarchical abstraction on image features by using local operations to acquire semantic information of different parts in an image. The method based on the convolutional neural network is used for semantically segmenting and predicting the image by utilizing the extracted local features of the image, and a deep learning network structure model based on the convolutional neural network has excellent performance in image processing in recent years. The computer vision model is very suitable and reliable for identifying the water leakage property, and is widely applied to projects represented by underground tunnels. However, although the conventional convolutional neural network structure achieves certain effects, the prediction accuracy is not ideal in the face of complex and various underground environments.
Disclosure of Invention
The invention aims to provide a method and a system for identifying the water leakage performance of an underground infrastructure, which aim to solve the problem of low prediction precision when a traditional convolutional neural network structure is used for identifying complex and various underground environments.
In order to achieve the purpose, the invention provides the following scheme:
a method for identifying the water leakage performance of underground infrastructure comprises the following steps:
acquiring underground infrastructure water leakage images under various environmental conditions; the underground infrastructure water leakage images under various environmental conditions comprise underground infrastructure water leakage images with multiple angles, different light intensities, different distances and different water leakage grades;
marking the underground infrastructure water leakage image to generate a marked underground infrastructure water leakage image;
training a convolutional neural network according to the underground infrastructure water leakage image and the marked underground infrastructure water leakage image, and constructing a computer vision prediction model; the computer vision prediction model is used for predicting the water leakage performance of the underground infrastructure water leakage image; the water leakage performance comprises a non-leakage abnormal state and a leakage abnormal state;
and acquiring an underground infrastructure water leakage image to be predicted under any environmental condition, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted by utilizing the computer vision prediction model.
Optionally, the marking of the underground infrastructure leakage water image is performed to generate a marked underground infrastructure leakage water image, which specifically includes:
screening the underground infrastructure water leakage image, removing non-leakage water abnormal data in the underground infrastructure water leakage image, and generating a screened underground infrastructure water leakage image;
cleaning the screened underground infrastructure water leakage image, removing fuzzified abnormal water leakage data in the screened underground infrastructure water leakage image, and generating a cleaned underground infrastructure water leakage image;
and annotating the cleaned underground infrastructure water leakage image, marking abnormal water leakage data in the cleaned underground infrastructure water leakage image, and generating an annotated underground infrastructure water leakage image.
Optionally, the training of the convolutional neural network according to the underground infrastructure leakage water image and the labeled underground infrastructure leakage water image to construct a computer vision prediction model specifically includes:
constructing an underground infrastructure water leakage image tensor according to the underground infrastructure water leakage image, and constructing an labeled underground infrastructure water leakage image tensor according to the labeled underground infrastructure water leakage image;
selecting hyper-parameters of a computer vision prediction model; the hyper-parameters comprise the number of training rounds, the learning rate and the data quantity of each training;
dividing the underground infrastructure water leakage image tensor and the marked underground infrastructure water leakage image tensor into a training set and a verification set;
based on the hyper-parameters, training the computer vision prediction model by using the underground infrastructure water leakage image tensor in the training set and the marked underground infrastructure water leakage image tensor, and outputting the prediction value of each pixel point in the underground infrastructure water leakage image; the predicted value of each pixel point is used for determining the water leakage prediction state of the underground infrastructure water leakage image;
and calculating the error between the predicted value and the true value by using a loss function, updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters are converged, and constructing the computer vision prediction model.
Optionally, the calculating an error between the predicted value and the true value by using a loss function, and updating a neural network parameter in the computer vision prediction model by using a gradient descent method until the neural network parameter converges, so as to construct the computer vision prediction model, and then further comprising:
selecting different hyper-parameters and constructing different computer vision prediction models;
and verifying different computer vision prediction models by using the underground infrastructure water leakage image tensor in the verification set and the marked underground infrastructure water leakage image tensor, determining an optimal computer vision prediction model, and recording the hyper-parameters corresponding to the optimal computer vision prediction model.
Optionally, the obtaining an image of leakage water of the underground infrastructure to be predicted under any environmental condition, and predicting the water leakage behavior of the image of leakage water of the underground infrastructure to be predicted by using the computer vision prediction model specifically include:
marking the underground infrastructure water leakage image to be predicted to generate a marked underground infrastructure water leakage image to be predicted;
constructing a tensor of the underground infrastructure water leakage image to be predicted according to the underground infrastructure water leakage image to be predicted, and constructing an annotated tensor of the underground infrastructure water leakage image to be predicted according to the annotated underground infrastructure water leakage image to be predicted;
and inputting the tensor of the underground infrastructure water leakage image to be predicted and the annotated tensor of the underground infrastructure water leakage image to be predicted into the computer vision prediction model, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted.
A system for identifying the behavior of underground infrastructure water seepage, comprising:
the system comprises an underground infrastructure water leakage image acquisition module, a data processing module and a data processing module, wherein the underground infrastructure water leakage image acquisition module is used for acquiring underground infrastructure water leakage images under various environmental conditions; the underground infrastructure water leakage images under various environmental conditions comprise underground infrastructure water leakage images with multiple angles, different light intensities, different distances and different water leakage grades;
the marking module is used for marking the underground infrastructure water leakage image to generate a marked underground infrastructure water leakage image;
the computer vision prediction model building module is used for training a convolutional neural network according to the underground infrastructure water leakage image and the marked underground infrastructure water leakage image to build a computer vision prediction model; the computer vision prediction model is used for predicting the water leakage performance of the underground infrastructure water leakage image; the water leakage performance comprises a non-leakage abnormal state and a leakage abnormal state;
and the water leakage performance prediction module is used for acquiring an underground infrastructure water leakage image to be predicted under any environmental condition and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted by utilizing the computer vision prediction model.
Optionally, the labeling module specifically includes:
the screening unit is used for screening the underground infrastructure water leakage image, removing non-leakage water abnormal data in the underground infrastructure water leakage image and generating a screened underground infrastructure water leakage image;
a cleaning unit, configured to clean the screened underground infrastructure leakage water image, remove blurred leakage water abnormal data in the screened underground infrastructure leakage water image, and generate a cleaned underground infrastructure leakage water image;
and the first marking unit is used for annotating the cleaned underground infrastructure leakage water image, marking the abnormal leakage water data in the cleaned underground infrastructure leakage water image and generating the marked underground infrastructure leakage water image.
Optionally, the computer vision prediction model building module specifically includes:
the first tensor construction unit is used for constructing the tensor of the underground infrastructure leakage water image according to the underground infrastructure leakage water image and constructing the marked underground infrastructure leakage water image matrix according to the marked underground infrastructure leakage water image;
the hyper-parameter selection unit is used for selecting hyper-parameters of the computer vision prediction model; the hyper-parameters comprise the number of training rounds, the learning rate and the data quantity of each training;
the dividing unit is used for dividing the underground infrastructure water leakage image tensor and the marked underground infrastructure water leakage image matrix into a training set and a verification set;
a predicted value output unit, configured to train the computer vision prediction model by using the tensor of the underground infrastructure leakage water image in the training set and the labeled underground infrastructure leakage water image matrix based on the hyper-parameter, and output a predicted value of whether leakage occurs to each pixel point in the underground infrastructure leakage water image; the predicted value of each pixel point is used for determining the water leakage prediction state of the underground infrastructure water leakage image;
and the computer vision prediction model construction unit is used for calculating the error between the predicted value and the true value by using a loss function, updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters are converged, and constructing the computer vision prediction model.
Optionally, the method further includes:
the different computer vision prediction model building units are used for selecting different hyper-parameters and building different computer vision prediction models;
and the optimal computer vision prediction model determining unit is used for verifying different computer vision prediction models by utilizing the underground infrastructure water leakage image tensor in the verification set and the marked underground infrastructure water leakage image matrix, determining an optimal computer vision prediction model and recording the hyper-parameters corresponding to the optimal computer vision prediction model.
Optionally, the water leakage performance prediction module specifically includes:
the second labeling unit is used for labeling the underground infrastructure leakage water image to be predicted and generating a labeled underground infrastructure leakage water image to be predicted;
the second tensor construction unit is used for constructing the tensor of the underground infrastructure leakage water image to be predicted according to the underground infrastructure leakage water image to be predicted and constructing the marked underground infrastructure leakage water image matrix to be predicted according to the marked underground infrastructure leakage water image to be predicted;
and the water leakage performance prediction unit is used for inputting the tensor of the underground infrastructure water leakage image to be predicted and the marked underground infrastructure water leakage image matrix to be predicted into the computer vision prediction model to predict the water leakage performance of the underground infrastructure water leakage image to be predicted.
According to the specific embodiment provided by the invention, the invention discloses the following technical effects: the invention provides a method and a system for identifying the water leakage performance of underground infrastructure, which are used for acquiring underground infrastructure water leakage images under various environmental conditions to enable image data to have diversity, meanwhile, marking the underground infrastructure water leakage images to generate marked underground infrastructure water leakage images so as to mark abnormal water leakage data, combining the abnormal water leakage data in the underground infrastructure and the external environmental conditions thereof, training a convolutional neural network by using the combined image data, and generating a computer vision prediction model to identify and predict the water leakage performance. Due to the diversity of image data and the computer vision prediction model constructed by combining the abnormal water leakage data in the underground infrastructure, the computer vision prediction model can have high identification precision even when the complex and diverse underground environment identification is carried out.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings needed in the embodiments will be briefly described below, and it is obvious that the drawings in the following description are only some embodiments of the present invention, and it is obvious for those skilled in the art to obtain other drawings without creative efforts.
FIG. 1 is a flow chart of a method for identifying the water leakage status of an underground infrastructure according to the present invention;
FIG. 2 is a flow chart of computer vision prediction model construction;
FIG. 3 is a schematic diagram of a computer vision model constructed based on a convolutional neural network;
FIG. 4 is an overall flowchart of the method for identifying the water leakage behavior of the underground infrastructure in an actual application scenario;
fig. 5 is a structural diagram of an underground infrastructure water leakage behavior identification system provided by the present invention.
Detailed Description
The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, and not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The invention aims to provide a method and a system for identifying the water leakage performance of an underground infrastructure, which can improve the identification precision in complex and various underground environments.
In order to make the aforementioned objects, features and advantages of the present invention comprehensible, embodiments accompanied with figures are described in further detail below.
Fig. 1 is a flowchart of a method for identifying a water leakage performance of an underground infrastructure according to the present invention, and as shown in fig. 1, the method for identifying a water leakage performance of an underground infrastructure includes:
step 101: acquiring underground infrastructure water leakage images under various environmental conditions; the underground infrastructure water leakage images under various environmental conditions comprise underground infrastructure water leakage images with multiple angles, different light intensities, different distances and different water leakage levels.
Step 102: and marking the underground infrastructure water leakage image to generate a marked underground infrastructure water leakage image.
The step 102 specifically includes: screening the underground infrastructure water leakage image, removing non-leakage water abnormal data in the underground infrastructure water leakage image, and generating a screened underground infrastructure water leakage image; cleaning the screened underground infrastructure water leakage image, removing fuzzified abnormal water leakage data in the screened underground infrastructure water leakage image, and generating a cleaned underground infrastructure water leakage image; and annotating the cleaned underground infrastructure water leakage image, marking abnormal water leakage data in the cleaned underground infrastructure water leakage image, and generating an annotated underground infrastructure water leakage image.
Step 103: training a convolutional neural network according to the underground infrastructure water leakage image and the marked underground infrastructure water leakage image, and constructing a computer vision prediction model; the computer vision prediction model is used for predicting the water leakage performance of the underground infrastructure water leakage image; the water leakage behavior comprises a no leakage abnormal state and a leakage abnormal state.
The step 103 specifically includes: constructing an underground infrastructure water leakage image tensor according to the underground infrastructure water leakage image, and constructing an labeled underground infrastructure water leakage image matrix according to the labeled underground infrastructure water leakage image; selecting hyper-parameters of a computer vision prediction model; the hyper-parameters comprise the number of training rounds, the learning rate and the data quantity of each training; dividing the tensor of the underground infrastructure leakage water image and the marked underground infrastructure leakage water image matrix into a training set and a verification set; based on the hyper-parameters, training the computer vision prediction model by using the tensor of the underground infrastructure leakage water image in the training set and the marked underground infrastructure leakage water image matrix, and outputting the prediction value of each pixel point in the underground infrastructure leakage water image; the predicted value of each pixel point is used for determining the water leakage prediction state of the underground infrastructure water leakage image; and calculating the error between the predicted value and the true value by using a loss function, updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters are converged, and constructing the computer vision prediction model.
Calculating an error between the predicted value and the true value by using a loss function, updating a neural network parameter in the computer vision prediction model by using a gradient descent method until the neural network parameter is converged, and constructing the computer vision prediction model, and then: selecting different hyper-parameters and constructing different computer vision prediction models; and verifying different computer vision prediction models by using the underground infrastructure water leakage image tensor in the verification set and the marked underground infrastructure water leakage image matrix, determining an optimal computer vision prediction model, and recording the hyper-parameters corresponding to the optimal computer vision prediction model.
Step 104: and acquiring an underground infrastructure water leakage image to be predicted under any environmental condition, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted by utilizing the computer vision prediction model.
The step 104 specifically includes: marking the underground infrastructure water leakage image to be predicted to generate a marked underground infrastructure water leakage image to be predicted; constructing a tensor of the underground infrastructure water leakage image to be predicted according to the underground infrastructure water leakage image to be predicted, and constructing an annotated underground infrastructure water leakage image matrix to be predicted according to the annotated underground infrastructure water leakage image to be predicted; and inputting the tensor of the underground infrastructure water leakage image to be predicted and the marked underground infrastructure water leakage image matrix to be predicted into the computer vision prediction model, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted.
The method is suitable for detecting the water leakage performance of various underground infrastructures such as tunnels, subways, underground buildings and the like, and the data to be detected are image data of the underground infrastructures. Through screening, cleaning and labeling of the original image data, the computer vision prediction model can learn the image characteristics and semantic information of the water leakage data, and identify the water leakage performance of the underground infrastructure.
The method for identifying the water leakage performance of the underground infrastructure, which is provided by the invention, takes a tunnel as an example and comprises the following specific steps:
step one, acquiring leakage water data of the underground tunnel.
The underground facility inspection personnel comprehensively inspect the underground facility and acquire the water leakage image data which comprise abnormal image data of multiple angles, different light intensities, different distances and different water leakage levels, thereby detecting and identifying the water leakage performance of the underground infrastructure.
And step two, acquiring trainable data. The trainable data refers to the operation of screening, cleaning and labeling the acquired original water leakage image data; the screening refers to removing abnormal data of non-leakage water; the cleaning is to remove the fuzzy abnormal data of the leakage water; the marking means that abnormal areas of abnormal water leakage data are indicated in a crowdsourcing mode; resulting in data that can be input as a computer vision prediction model.
Combining the original image data and the labeled data to construct a computer vision prediction model; the computer vision prediction model is characterized in that a convolutional neural network is used for extracting image features, and a semantic segmentation network is constructed through a residual error network structure, so that the water leakage performance is predicted and identified.
Image data of the facility is collected inside the underground infrastructure, and the health state of the facility is obtained through the data. However, the section where the underground infrastructure is located often includes more complicated climate and hydrological environments, and image acquisition can be disturbed by factors such as light, so the diversity of image data is very important to the prediction effect, and therefore the image data under various conditions can be acquired, and the accuracy of prediction is improved by combining with a computer vision model.
Fig. 2 is a flow chart of building a computer vision prediction model, as shown in fig. 2, the building steps of the computer vision prediction model are as follows:
step 1.1: obtaining the preprocessed original image data and the labeled data, and organizing the preprocessed original image data and the labeled data into a wholeh×w×cTensor ofXAndh×w of (2) matrixYWhereinhFor the length of the acquired image or images,win order to be wide in the acquired image,cfor harvestingNumber of channels of image, whereinXRepresenting original image data，YRepresents the annotation data, andXandYthe data at the same index position are respectively the original data and the labeled data of the same image.
Step 1.2: selecting three hyper-parameters of the model, training roundpControlling the number of executions of a model，Learning ratelControlling model parameter update rate, and data packet sizebControlling the amount of data per training.
Step 1.3: will input tensorXDividing the input matrix into a training set and a verification set, taking the first 70% of data items as the training set and the last 30% of data items as the verification setYThe same operation is done.
Step 1.4: and (3) training a computer vision prediction model by using data in the training set, extracting features in the image by using a Convolutional Neural Network (CNN), adding cavity space pyramid pooling (ASPP) after the convolutional layer, taking the pooled output as the input of a full-link layer, and outputting the prediction result of each pixel point.
Fig. 3 is a schematic diagram of a computer vision model constructed based on a convolutional neural network, as shown in fig. 3, the convolutional neural network has excellent performance in the aspects of extracting image features, obtaining image semantics, and the like, and can learn and identify the water leakage behavior, and the operation flow can be represented as the following steps:
and 1.4.1, randomly initializing all weight parameters of the model.
And step 1.4.2, inputting the training data into a convolutional neural network, and generating a predicted value after multiple times of convolution and pooling operations.
And step 1.4.3, comparing the predicted value with the true value, calculating a prediction error by using a loss function, and jumping to step 1.4.5 if the error is converged, or jumping to step 1.4.4 otherwise.
And 1.4.4, updating the parameters of the model by using a gradient descent algorithm, and jumping to the step 1.4.2.
And 1.4.5, saving the parameters of the model to the local, and exiting the training program.
Step 1.5: and calculating the error between the predicted value and the true value in the training set by using a loss function, updating each parameter of the computer vision prediction model by using a gradient descent method, and repeating the training until the parameters are converged, wherein the training is ended at the moment.
Step 1.6: and (3) testing the performance of the model on the verification set, skipping to the step 1.2, selecting other hyper-parameters for experiment until the hyper-parameter combination with the best effect is found out, and storing the parameters of the model.
The process of predicting and identifying the behaviour of the leakage water is as follows:
step 2.1: loading each weight parameter of the computer vision prediction model into a memory;
step 2.2: converting image data and label data into two input data in one-to-one correspondenceXAndY；
step 2.3: will be provided withXAndYinputting the image data into a computer vision prediction model, and outputting a predicted value of the image;
step 2.4: repeating the steps of 2.2 and 2.3 for a plurality of turns and a plurality of images to obtain the predicted value of a group of images, and calculating the corresponding prediction index.
And fourthly, forecasting the abnormal index of the leakage water of the underground infrastructure under various environmental conditions to realize the coverage test of the safety of the underground infrastructure, wherein the various boundary conditions refer to the combination of various different environmental factors to cover various possible conditions so as to realize the coverage test of the safety of the underground infrastructure.
The coverage test of the safety of the underground infrastructure is as follows:
step 3.1: loading each weight parameter of the computer vision prediction model into a memory;
step 3.2: for each environmental factor, collecting a group of water leakage image data, and combining to obtain a plurality of groups of water leakage image data;
step 3.3: and converting each collected possible environment data combination into an input tensor, inputting the input tensor into the computer vision prediction model to obtain the abnormal performance index prediction value under the condition, so that the range of the abnormal performance indexes of the underground infrastructure under all the environment condition combinations can be obtained, and the safety of the underground infrastructure is tested in a coverage manner.
After the computer vision prediction model is trained and adjusted on the validation set to the best parameters, the parameters of the model will be saved for subsequent application. By collecting and processing the image data of the underground facilities, the method can predict whether the underground facilities are damaged by water leakage, thereby realizing the health monitoring of the underground infrastructure.
The method is implemented based on computer science and various machine learning algorithms, needs a certain programming and machine learning and deep learning basis, and is realized based on Python programming language and an open-source machine learning library PyTorch. To verify the computer vision prediction model based on the convolutional neural network shown in fig. 2 and 3, experiments were performed using the water leakage behavior data collected from a certain tunnel. In order to evaluate the prediction capability of the model, three evaluation indexes, namely Cross Entropy error (Cross Entropy), an overlapping coefficient (IoU) and an F1 score, are used, the Cross Entropy is used for measuring the error between the predicted value and the true value of the monitoring index, the overlapping coefficient IoU is used for calculating the ratio of the intersection and the union of the predicted water leakage border and the true water leakage border, the ratio is a standard performance measurement index of the image segmentation problem, and the F1 score is used for measuring the accuracy of the water leakage prediction. In order to verify that the performance of the method is better than that of other models, other commonly used prediction models such as linear regression, support vector machines, multilayer perceptrons and the like are used for carrying out comparison experiments, and experiments prove that the prediction error of the method is the lowest and the prediction correlation is the highest.
Fig. 4 is an overall flowchart of the method for identifying the water leakage performance of the underground infrastructure in an actual application scenario. In the application scene, some visualization means are used for visually displaying the prediction result, and the charts are drawn by using a matplotlib library. In the application of the water leakage behavior identification of the underground infrastructure, the identification accuracy is expressed by comparing the coverage between the real water leakage area and the prediction area.
Fig. 5 is a structural diagram of an underground infrastructure water leakage performance identification system provided by the present invention, and as shown in fig. 5, an underground infrastructure water leakage performance identification system includes:
an underground infrastructure leakage water image acquisition module 501, configured to acquire an underground infrastructure leakage water image under multiple environmental conditions; the underground infrastructure water leakage images under various environmental conditions comprise underground infrastructure water leakage images with multiple angles, different light intensities, different distances and different water leakage levels.
And the marking module 502 is configured to mark the underground infrastructure leakage water image to generate a marked underground infrastructure leakage water image.
The labeling module 502 specifically includes: the screening unit is used for screening the underground infrastructure water leakage image, removing non-leakage water abnormal data in the underground infrastructure water leakage image and generating a screened underground infrastructure water leakage image; a cleaning unit, configured to clean the screened underground infrastructure leakage water image, remove blurred leakage water abnormal data in the screened underground infrastructure leakage water image, and generate a cleaned underground infrastructure leakage water image; and the first marking unit is used for annotating the cleaned underground infrastructure leakage water image, marking the abnormal leakage water data in the cleaned underground infrastructure leakage water image and generating the marked underground infrastructure leakage water image.
A computer vision prediction model construction module 503, configured to train a convolutional neural network according to the underground infrastructure leakage water image and the labeled underground infrastructure leakage water image, and construct a computer vision prediction model; the computer vision prediction model is used for predicting the water leakage performance of the underground infrastructure water leakage image; the water leakage behavior comprises a no leakage abnormal state and a leakage abnormal state.
The computer vision prediction model building module 503 specifically includes: the first tensor construction unit is used for constructing the tensor of the underground infrastructure leakage water image according to the underground infrastructure leakage water image and constructing the marked underground infrastructure leakage water image matrix according to the marked underground infrastructure leakage water image; the hyper-parameter selection unit is used for selecting hyper-parameters of the computer vision prediction model; the hyper-parameters comprise the number of training rounds, the learning rate and the data quantity of each training; the dividing unit is used for dividing the underground infrastructure water leakage image tensor and the marked underground infrastructure water leakage image matrix into a training set and a verification set; the predicted value output unit is used for training the computer vision prediction model by utilizing the tensor of the underground infrastructure leakage water image in the training set and the marked underground infrastructure leakage water image matrix based on the hyper-parameters and outputting the predicted value of whether leakage occurs to each pixel point in the underground infrastructure leakage water image; the predicted value of each pixel point is used for determining the water leakage prediction state of the underground infrastructure water leakage image; and the computer vision prediction model construction unit is used for calculating the error between the predicted value and the true value by using a loss function, updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters are converged, and constructing the computer vision prediction model.
The invention also includes: the different computer vision prediction model building units are used for selecting different hyper-parameters and building different computer vision prediction models; and the optimal computer vision prediction model determining unit is used for verifying different computer vision prediction models by utilizing the underground infrastructure water leakage image tensor in the verification set and the marked underground infrastructure water leakage image matrix, determining an optimal computer vision prediction model and recording the hyper-parameters corresponding to the optimal computer vision prediction model.
The water leakage behavior prediction module 504 is configured to obtain an image of water leakage from the underground infrastructure to be predicted under any environmental condition, and predict a water leakage behavior of the image of water leakage from the underground infrastructure to be predicted by using the computer vision prediction model.
The water leakage behavior prediction module 504 specifically includes: the second labeling unit is used for labeling the underground infrastructure leakage water image to be predicted and generating a labeled underground infrastructure leakage water image to be predicted; the second tensor construction unit is used for constructing the tensor of the underground infrastructure leakage water image to be predicted according to the underground infrastructure leakage water image to be predicted and constructing the marked underground infrastructure leakage water image matrix to be predicted according to the marked underground infrastructure leakage water image to be predicted; and the water leakage performance prediction unit is used for inputting the tensor of the underground infrastructure water leakage image to be predicted and the marked underground infrastructure water leakage image matrix to be predicted into the computer vision prediction model to predict the water leakage performance of the underground infrastructure water leakage image to be predicted.
Aiming at the problems of various geological conditions and complex internal structure of the position of the underground infrastructure, the method combines abnormal property data in the underground infrastructure with external environmental conditions, constructs an image segmentation method based on a convolutional neural network, and identifies and predicts the property of the water leakage in the underground infrastructure so as to improve the identification precision. The method comprises the steps of firstly screening, cleaning and labeling original image data acquired by an inspector to obtain image data of the same water leakage area under various environmental conditions and form data tensor. And then extracting corresponding image data and label data from the original data tensor to form an input tensor for each image, extracting image features by using a convolutional neural network, and taking the output of the pooling layer as the input of a fully-connected neural network, wherein the obtained output is an identification result. The invention is mainly applied to health monitoring of underground facilities, and realizes monitoring and early warning of the water leakage performance of the whole facility by acquiring the image data of the underground facility, thereby providing guarantee for the safe operation of the underground facility. In the task of predicting the performance health monitoring indexes of the underground infrastructure, the prediction method can achieve the effects of lower prediction error and higher prediction accuracy compared with the traditional prediction method.
The embodiments in the present description are described in a progressive manner, each embodiment focuses on differences from other embodiments, and the same and similar parts among the embodiments are referred to each other. For the system disclosed by the embodiment, the description is relatively simple because the system corresponds to the method disclosed by the embodiment, and the relevant points can be referred to the method part for description.
The principles and embodiments of the present invention have been described herein using specific examples, which are provided only to help understand the method and the core concept of the present invention; meanwhile, for a person skilled in the art, according to the idea of the present invention, the specific embodiments and the application range may be changed. In view of the above, the present disclosure should not be construed as limiting the invention.
Claims (10)
1. A method for identifying the property of water leakage of underground infrastructure is characterized by comprising the following steps:
acquiring underground infrastructure water leakage images under various environmental conditions; the underground infrastructure water leakage images under various environmental conditions comprise underground infrastructure water leakage images with multiple angles, different light intensities, different distances and different water leakage grades;
marking the underground infrastructure water leakage image to generate a marked underground infrastructure water leakage image;
training a convolutional neural network according to the underground infrastructure water leakage image and the marked underground infrastructure water leakage image, and constructing a computer vision prediction model; the computer vision prediction model is used for predicting the water leakage performance of the underground infrastructure water leakage image; the water leakage performance comprises a non-leakage abnormal state and a leakage abnormal state;
and acquiring an underground infrastructure water leakage image to be predicted under any environmental condition, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted by utilizing the computer vision prediction model.
2. The method for identifying the water leakage performance of the underground infrastructure according to claim 1, wherein the step of labeling the water leakage image of the underground infrastructure to generate a labeled water leakage image of the underground infrastructure comprises:
screening the underground infrastructure water leakage image, removing non-leakage water abnormal data in the underground infrastructure water leakage image, and generating a screened underground infrastructure water leakage image;
cleaning the screened underground infrastructure water leakage image, removing fuzzified abnormal water leakage data in the screened underground infrastructure water leakage image, and generating a cleaned underground infrastructure water leakage image;
and annotating the cleaned underground infrastructure water leakage image, marking abnormal water leakage data in the cleaned underground infrastructure water leakage image, and generating an annotated underground infrastructure water leakage image.
3. The method for identifying the water leakage property of the underground infrastructure according to claim 1, wherein the training of the convolutional neural network according to the water leakage image of the underground infrastructure and the labeled water leakage image of the underground infrastructure to construct the computer vision prediction model specifically comprises:
constructing an underground infrastructure water leakage image tensor according to the underground infrastructure water leakage image, and constructing an labeled underground infrastructure water leakage image matrix according to the labeled underground infrastructure water leakage image;
selecting hyper-parameters of a computer vision prediction model; the hyper-parameters comprise the number of training rounds, the learning rate and the data quantity of each training;
dividing the tensor of the underground infrastructure leakage water image and the marked underground infrastructure leakage water image matrix into a training set and a verification set;
based on the hyper-parameters, training the computer vision prediction model by using the tensor of the underground infrastructure leakage water image in the training set and the marked underground infrastructure leakage water image matrix, and outputting the prediction value of each pixel point in the underground infrastructure leakage water image; the predicted value of each pixel point is used for determining the water leakage prediction state of the underground infrastructure water leakage image;
and calculating the error between the predicted value and the true value by using a loss function, updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters are converged, and constructing the computer vision prediction model.
4. The method for identifying the water leakage behavior of the underground infrastructure as claimed in claim 3, wherein the calculating the error between the predicted value and the true value by using the loss function and updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters converge to construct the computer vision prediction model, and then further comprising:
selecting different hyper-parameters and constructing different computer vision prediction models;
and verifying different computer vision prediction models by using the underground infrastructure water leakage image tensor in the verification set and the marked underground infrastructure water leakage image matrix, determining an optimal computer vision prediction model, and recording the hyper-parameters corresponding to the optimal computer vision prediction model.
5. The method for identifying the water leakage performance of the underground infrastructure according to claim 4, wherein the obtaining the water leakage image of the underground infrastructure to be predicted under any environmental condition and predicting the water leakage performance of the water leakage image of the underground infrastructure to be predicted by using the computer vision prediction model specifically comprises:
marking the underground infrastructure water leakage image to be predicted to generate a marked underground infrastructure water leakage image to be predicted;
constructing a tensor of the underground infrastructure water leakage image to be predicted according to the underground infrastructure water leakage image to be predicted, and constructing an annotated underground infrastructure water leakage image matrix to be predicted according to the annotated underground infrastructure water leakage image to be predicted;
and inputting the tensor of the underground infrastructure water leakage image to be predicted and the marked underground infrastructure water leakage image matrix to be predicted into the computer vision prediction model, and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted.
6. A system for identifying the behavior of water leakage from an underground infrastructure, comprising:
the system comprises an underground infrastructure water leakage image acquisition module, a data processing module and a data processing module, wherein the underground infrastructure water leakage image acquisition module is used for acquiring underground infrastructure water leakage images under various environmental conditions; the underground infrastructure water leakage images under various environmental conditions comprise underground infrastructure water leakage images with multiple angles, different light intensities, different distances and different water leakage grades;
the marking module is used for marking the underground infrastructure water leakage image to generate a marked underground infrastructure water leakage image;
the computer vision prediction model building module is used for training a convolutional neural network according to the underground infrastructure water leakage image and the marked underground infrastructure water leakage image to build a computer vision prediction model; the computer vision prediction model is used for predicting the water leakage performance of the underground infrastructure water leakage image; the water leakage performance comprises a non-leakage abnormal state and a leakage abnormal state;
and the water leakage performance prediction module is used for acquiring an underground infrastructure water leakage image to be predicted under any environmental condition and predicting the water leakage performance of the underground infrastructure water leakage image to be predicted by utilizing the computer vision prediction model.
7. The system for identifying the water leakage property of the underground infrastructure as claimed in claim 6, wherein the labeling module comprises:
the screening unit is used for screening the underground infrastructure water leakage image, removing non-leakage water abnormal data in the underground infrastructure water leakage image and generating a screened underground infrastructure water leakage image;
a cleaning unit, configured to clean the screened underground infrastructure leakage water image, remove blurred leakage water abnormal data in the screened underground infrastructure leakage water image, and generate a cleaned underground infrastructure leakage water image;
and the first marking unit is used for annotating the cleaned underground infrastructure leakage water image, marking the abnormal leakage water data in the cleaned underground infrastructure leakage water image and generating the marked underground infrastructure leakage water image.
8. The system for identifying the water leakage behavior of underground infrastructure as claimed in claim 6, wherein the computer vision prediction model building module specifically comprises:
the first tensor construction unit is used for constructing the tensor of the underground infrastructure leakage water image according to the underground infrastructure leakage water image and constructing the marked underground infrastructure leakage water image matrix according to the marked underground infrastructure leakage water image;
the hyper-parameter selection unit is used for selecting hyper-parameters of the computer vision prediction model; the hyper-parameters comprise the number of training rounds, the learning rate and the data quantity of each training;
the dividing unit is used for dividing the underground infrastructure water leakage image tensor and the marked underground infrastructure water leakage image matrix into a training set and a verification set;
the predicted value output unit is used for training the computer vision prediction model by utilizing the tensor of the underground infrastructure leakage water image in the training set and the marked underground infrastructure leakage water image matrix based on the hyper-parameters and outputting a predicted value of whether each pixel point in the underground infrastructure leakage water image leaks or not; the predicted value of each pixel point is used for determining the water leakage prediction state of the underground infrastructure water leakage image;
and the computer vision prediction model construction unit is used for calculating the error between the predicted value and the true value by using a loss function, updating the neural network parameters in the computer vision prediction model by using a gradient descent method until the neural network parameters are converged, and constructing the computer vision prediction model.
9. The underground infrastructure water leakage performance identification system of claim 8, further comprising:
the different computer vision prediction model building units are used for selecting different hyper-parameters and building different computer vision prediction models;
and the optimal computer vision prediction model determining unit is used for verifying different computer vision prediction models by utilizing the underground infrastructure water leakage image tensor in the verification set and the marked underground infrastructure water leakage image matrix, determining an optimal computer vision prediction model and recording the hyper-parameters corresponding to the optimal computer vision prediction model.
10. The system for identifying the water leakage behavior of underground infrastructure as claimed in claim 9, wherein the water leakage behavior prediction module comprises:
the second labeling unit is used for labeling the underground infrastructure leakage water image to be predicted and generating a labeled underground infrastructure leakage water image to be predicted;
the second tensor construction unit is used for constructing the tensor of the underground infrastructure leakage water image to be predicted according to the underground infrastructure leakage water image to be predicted and constructing the marked underground infrastructure leakage water image matrix to be predicted according to the marked underground infrastructure leakage water image to be predicted;
and the water leakage performance prediction unit is used for inputting the tensor of the underground infrastructure water leakage image to be predicted and the marked underground infrastructure water leakage image matrix to be predicted into the computer vision prediction model to predict the water leakage performance of the underground infrastructure water leakage image to be predicted.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110700803.8A CN113516179B (en) | 2021-06-24 | 2021-06-24 | Method and system for identifying water leakage performance of underground infrastructure |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110700803.8A CN113516179B (en) | 2021-06-24 | 2021-06-24 | Method and system for identifying water leakage performance of underground infrastructure |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113516179A true CN113516179A (en) | 2021-10-19 |
| CN113516179B CN113516179B (en) | 2022-02-18 |
Family
ID=78065924
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110700803.8A Active CN113516179B (en) | 2021-06-24 | 2021-06-24 | Method and system for identifying water leakage performance of underground infrastructure |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113516179B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114383787A (en) * | 2021-11-19 | 2022-04-22 | 郑州大学 | Array type waterproof layer leakage detection system |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190244337A1 (en) * | 2018-02-05 | 2019-08-08 | Nec Laboratories America, Inc. | Unsupervised image-based anomaly detection using multi-scale context-dependent deep autoencoding gaussian mixture model |
| CN110598767A (en) * | 2019-08-29 | 2019-12-20 | 河南省收费还贷高速公路管理有限公司航空港分公司 | SSD convolutional neural network-based underground drainage pipeline defect identification method |
| CN110689042A (en) * | 2019-08-20 | 2020-01-14 | 中国矿业大学(北京) | Tunnel leakage grade identification method and device, storage medium and electronic device |
| CN111814835A (en) * | 2020-06-12 | 2020-10-23 | 理光软件研究所（北京）有限公司 | Training method and device of computer vision model, electronic equipment and storage medium |
| CN111899288A (en) * | 2020-06-01 | 2020-11-06 | 上海大学 | Tunnel leakage water area detection and identification method based on infrared and visible light image fusion |
| CN112347862A (en) * | 2020-10-20 | 2021-02-09 | 浙江新再灵科技股份有限公司 | Elevator door fault real-time detection method based on machine vision |
| CN112488119A (en) * | 2020-12-18 | 2021-03-12 | 山西省信息产业技术研究院有限公司 | Tunnel block falling or water seepage detection and measurement method based on double-depth learning model |
-
2021
- 2021-06-24 CN CN202110700803.8A patent/CN113516179B/en active Active
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190244337A1 (en) * | 2018-02-05 | 2019-08-08 | Nec Laboratories America, Inc. | Unsupervised image-based anomaly detection using multi-scale context-dependent deep autoencoding gaussian mixture model |
| CN110689042A (en) * | 2019-08-20 | 2020-01-14 | 中国矿业大学(北京) | Tunnel leakage grade identification method and device, storage medium and electronic device |
| CN110598767A (en) * | 2019-08-29 | 2019-12-20 | 河南省收费还贷高速公路管理有限公司航空港分公司 | SSD convolutional neural network-based underground drainage pipeline defect identification method |
| CN111899288A (en) * | 2020-06-01 | 2020-11-06 | 上海大学 | Tunnel leakage water area detection and identification method based on infrared and visible light image fusion |
| CN111814835A (en) * | 2020-06-12 | 2020-10-23 | 理光软件研究所（北京）有限公司 | Training method and device of computer vision model, electronic equipment and storage medium |
| CN112347862A (en) * | 2020-10-20 | 2021-02-09 | 浙江新再灵科技股份有限公司 | Elevator door fault real-time detection method based on machine vision |
| CN112488119A (en) * | 2020-12-18 | 2021-03-12 | 山西省信息产业技术研究院有限公司 | Tunnel block falling or water seepage detection and measurement method based on double-depth learning model |
Non-Patent Citations (2)
| Title |
|---|
| 于洪亮 等: "隧道渗漏水监测系统研究与设计", 《内蒙古公路与运输》 * |
| 李军 等: "地铁隧道结构机器视觉检测系统及应用分析", 《测绘通报》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114383787A (en) * | 2021-11-19 | 2022-04-22 | 郑州大学 | Array type waterproof layer leakage detection system |
| CN114383787B (en) * | 2021-11-19 | 2024-03-26 | 郑州大学 | Array waterproof layer leakage detection system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113516179B (en) | 2022-02-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| KR102026449B1 (en) | Simulation Data Preprocessing Technique for Development of Damage Detecting Method for Bridges Based on Convolutional Neural Network | |
| CN112528365B (en) | Method for predicting healthy evolution trend of underground infrastructure structure | |
| CN113518922A (en) | Improved method for assessing the condition of a pipeline | |
| CN112330591A (en) | Steel rail surface defect detection method and device capable of achieving sample-less learning | |
| CN110931112A (en) | Brain medical image analysis method based on multi-dimensional information fusion and deep learning | |
| Savino et al. | Automated classification of civil structure defects based on convolutional neural network | |
| Rosso et al. | Review on deep learning in structural health monitoring | |
| CN115546565A (en) | YOLOCBF-based power plant key area pipeline oil leakage detection method | |
| CN113516179B (en) | Method and system for identifying water leakage performance of underground infrastructure | |
| CN115761300A (en) | Method, system and detection device for dividing safety exit abnormity | |
| US11663815B2 (en) | System and method for inspection of heat recovery steam generator | |
| Roslan et al. | Real-time plastic surface defect detection using deep learning | |
| CN117422990A (en) | Bridge structure classification and evaluation method based on machine learning | |
| JP6257986B2 (en) | Interactive facility failure prediction method and interactive facility failure prediction system | |
| Giglioni et al. | Deep autoencoders for unsupervised damage detection with application to the Z24 benchmark bridge | |
| CN114550460B (en) | Rail transit anomaly detection method and device and storage medium | |
| Andersen et al. | Domain adapted probabilistic inspection using deep probabilistic segmentation | |
| Liang et al. | Deep Bayesian U-Nets for Efficient, Robust and Reliable Post-Disaster Damage Localization | |
| Kumar et al. | Leveraging visualization techniques to develop improved deep neural network architecture for sewer defect identification | |
| CN114139589A (en) | Fault diagnosis method, device, equipment and computer readable storage medium | |
| Shin et al. | Visualization for explanation of deep learning-based defect detection model using class activation map | |
| Zhu et al. | Research on recognition algorithm of tunnel leakage based on image processing | |
| CN117634006B (en) | BIM technology-based sleeve embedded engineering management system and method | |
| Panella et al. | Smart Image Based Technology and Deep Learning for Tunnel Inspection and Asset Management | |
| Li | Machine Learning Based Seismic Structural Health Monitoring and Reconnaissance |
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 |