CN117522149A - Tunnel security risk identification method and device and security management platform - Google Patents
Tunnel security risk identification method and device and security management platform Download PDFInfo
- Publication number
- CN117522149A CN117522149A CN202311707645.4A CN202311707645A CN117522149A CN 117522149 A CN117522149 A CN 117522149A CN 202311707645 A CN202311707645 A CN 202311707645A CN 117522149 A CN117522149 A CN 117522149A
- Authority
- CN
- China
- Prior art keywords
- risk
- construction
- image
- procedure
- features
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 268
- 238000010276 construction Methods 0.000 claims abstract description 170
- 230000008569 process Effects 0.000 claims abstract description 95
- 238000011156 evaluation Methods 0.000 claims abstract description 65
- 238000000605 extraction Methods 0.000 claims description 47
- 238000012545 processing Methods 0.000 claims description 35
- 238000004422 calculation algorithm Methods 0.000 claims description 33
- 238000007726 management method Methods 0.000 claims description 26
- 238000004140 cleaning Methods 0.000 claims description 25
- 238000007781 pre-processing Methods 0.000 claims description 25
- 238000012544 monitoring process Methods 0.000 claims description 23
- 238000010801 machine learning Methods 0.000 claims description 16
- 238000005516 engineering process Methods 0.000 claims description 13
- 238000012549 training Methods 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 11
- 238000005422 blasting Methods 0.000 claims description 11
- 238000013527 convolutional neural network Methods 0.000 claims description 10
- 238000005553 drilling Methods 0.000 claims description 10
- 238000002372 labelling Methods 0.000 claims description 10
- 230000003595 spectral effect Effects 0.000 claims description 10
- 238000009412 basement excavation Methods 0.000 claims description 9
- 230000007613 environmental effect Effects 0.000 claims description 7
- 239000011435 rock Substances 0.000 claims description 7
- 238000013526 transfer learning Methods 0.000 claims description 7
- 230000009467 reduction Effects 0.000 claims description 6
- 230000001133 acceleration Effects 0.000 claims description 5
- 238000006243 chemical reaction Methods 0.000 claims description 5
- 239000003086 colorant Substances 0.000 claims description 5
- 238000003066 decision tree Methods 0.000 claims description 5
- 238000003708 edge detection Methods 0.000 claims description 5
- 239000002360 explosive Substances 0.000 claims description 5
- 238000001914 filtration Methods 0.000 claims description 5
- 230000000977 initiatory effect Effects 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 5
- 238000003672 processing method Methods 0.000 claims description 5
- 238000007637 random forest analysis Methods 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 238000007619 statistical method Methods 0.000 claims description 5
- 238000012706 support-vector machine Methods 0.000 claims description 5
- 238000011426 transformation method Methods 0.000 claims description 5
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 claims description 5
- 238000005303 weighing Methods 0.000 claims description 5
- 238000010586 diagram Methods 0.000 description 8
- 230000006399 behavior Effects 0.000 description 7
- 238000013459 approach Methods 0.000 description 5
- 201000010099 disease Diseases 0.000 description 3
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 3
- 239000000428 dust Substances 0.000 description 3
- 238000003860 storage Methods 0.000 description 3
- 230000008859 change Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000006870 function Effects 0.000 description 2
- 230000004927 fusion Effects 0.000 description 2
- 238000012502 risk assessment Methods 0.000 description 2
- 238000012800 visualization Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000012098 association analyses Methods 0.000 description 1
- 238000009411 base construction Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000009933 burial Methods 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 230000007547 defect 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
- 238000009826 distribution Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000005484 gravity Effects 0.000 description 1
- 239000003673 groundwater Substances 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- 230000002265 prevention Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 239000004575 stone Substances 0.000 description 1
- 230000001960 triggered effect Effects 0.000 description 1
- 230000005641 tunneling Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0635—Risk analysis of enterprise or organisation activities
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/903—Querying
- G06F16/90335—Query processing
- G06F16/90344—Query processing by using string matching techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/904—Browsing; Visualisation therefor
-
- 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
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/096—Transfer learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0631—Resource planning, allocation, distributing or scheduling for enterprises or organisations
- G06Q10/06316—Sequencing of tasks or work
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/08—Construction
Landscapes
- Engineering & Computer Science (AREA)
- Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Human Resources & Organizations (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Databases & Information Systems (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Computational Linguistics (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Educational Administration (AREA)
- Quality & Reliability (AREA)
- Operations Research (AREA)
- Life Sciences & Earth Sciences (AREA)
- Software Systems (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Primary Health Care (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a tunnel security risk identification method, a tunnel security risk identification device and a security management platform, which solve the problem that the prior art cannot accurately and visually early warn the risk of a tunnel construction operation area, so that the security of tunnel construction personnel cannot be forcefully ensured. The method is used for tunnel security risk early warning, and comprises the steps of acquiring the number of items of risk at each process position for each process acquired in real time, and giving the process risk of each process and the evaluation value of each risk at the process position according to the number of items of risk at the process position; carrying out superposition summation according to respective weights based on the process risk of each process and the evaluation value of each risk at the process position to obtain the construction risk evaluation value of each process; and determining the risk evaluation grade of the area where each process is located according to the threshold range where the construction risk evaluation value of each process is located. The method is used for tunnel security risk early warning.
Description
Technical Field
A tunnel security risk identification method, a device and a security management platform are used for tunnel security risk early warning, and belong to the technical field of tunnel security risk identification and early warning.
Background
In recent years, the railway engineering construction of China tends to be in complicated and difficult areas such as plateaus, high and cold, and especially tunnel engineering has the characteristics of large burial depth, overlength, complex geology, and the like. The construction safety risk problems faced in the tunnel construction process are prominent, such as natural environment risks of landslide, groundwater piping, rock burst and the like, construction operation risks of blasting management inappropriateness, earth and rock excavation, mechanical overload operation and the like, and mechanical operation risks of equipment operation load, travel route, operation area, power consumption, mechanical operation irregularity and the like.
The traditional security risk management adopts a video monitoring mode, the management means is single, the effect is poor in dim dust environment, the monitoring blind area is easy to appear, the false alarm rate and the false alarm rate are high, the comprehensive management and control capability for 'people, machines, materials, methods and rings' multi-element security risk association analysis and dynamic change is lacked, and an omnibearing and efficient security risk dynamic identification and early warning prevention system is difficult to form.
The machine vision has the advantages of high automation degree, high informatization integration capability, high speed, high precision, high anti-interference capability and the like, thereby being applicable to dangerous working environments or occasions where manual vision is difficult to meet the requirements. The automatic and intelligent detection technology based on machine vision has been successfully applied to roads and tunnels and is primarily applied to bridges, but is mainly concentrated on the high-altitude concrete member apparent image acquisition technology with wide vision, and remains in the theoretical research stage in the aspect of automatic identification of risk diseases.
At present, a disease method based on machine vision exists, and a detection system based on machine vision is adopted to efficiently identify the same crack in different periods, grasp disease development conditions and formulate corresponding correction measures, and the method comprises the following steps: (1) measuring an initial position of a scan crack; (2) identifying fracture inflection points; (3) false inflection point rejection; and (4) crack generation and type discrimination. However, the following technical problems exist: the method is only used for crack type identification, and the application range is narrow; the application of the machine learning algorithm is not outstanding, and more feature recognition technology is active; the algorithm can only identify the type and the length of the crack, can not accurately judge the width and the acceptance direction of the crack, and has weak guidance;
aiming at the problems of large delay in human behavior recognition, limited monitoring of construction behaviors and construction processes of workers in the prior art, for each frame of characteristics, the significance of spatial information of the characteristics is considered, the significance of the characteristics on a time sequence is considered, and the similarity between a weighted graph and a characteristic sequence is measured by combining the weighted graph and a measuring method, so that the problems caused by the differences of unstable skeleton points and the length of the behavior sequence can be effectively processed, and the characteristics can be compressed, thereby reducing the recognition time, realizing real-time human behavior recognition and normalizing the construction behaviors and the construction processes of the workers. Comprising the following steps: acquiring construction state parameters, wherein the construction state parameters comprise environment parameters and human body parameters in a tunnel; performing dimension reduction on the construction state parameters, and performing standardized processing on information of different dimensions to obtain main components for reflecting the construction state of workers; inputting the obtained principal component into a BP neural network, and outputting a plurality of values as coordinates to judge and classify the construction state of the worker. The invention fully utilizes the inherent connection between the construction process and the environmental parameters, constructors and construction instruments, and solves the problems of higher dust concentration, low visibility, difficult acquisition of the joint point coordinates of the part of the skeleton of the worker in the construction process and the like. However, the following technical problems exist: the method is only used for human behaviors, and the application range is narrow; the dust problem is solved through machine vision, the on-site personnel behavior can be assisted and standardized, but the safety of tunnel personnel is not forcefully ensured, and the risk identification is not contributed.
In summary, the prior art has the following technical problems:
1. the risk of the tunnel construction operation area cannot be accurately and visually pre-warned, so that the safety of tunnel construction personnel cannot be powerfully ensured;
2. the process risk and the process related risk cannot be comprehensively considered, so that the problem of poor risk prediction precision is caused.
Disclosure of Invention
The invention aims to provide a tunnel security risk identification method, a tunnel security risk identification device and a tunnel security management platform, which solve the problem that the prior art cannot perform visual early warning on the risk of a tunnel construction operation area, so that the security of tunnel construction personnel cannot be effectively ensured.
In order to achieve the above purpose, the technical scheme adopted by the invention is as follows:
a tunnel security risk identification method acquires required evaluation procedures, carries out security risk identification on each acquired procedure, and comprises the following steps:
step S1: for each obtained procedure, obtaining the number of items of risk at each procedure position, and giving the procedure risk of each procedure and the evaluation value of each risk at the procedure position according to the number of items of risk at the procedure position, wherein the procedure comprises one or more of advanced support, drilling and blasting construction, primary support, lining structure, portal structure, stepping, groove, water-proof component and open cut backfill, and each risk at the procedure position comprises natural environment risk, construction operation risk and large-scale mechanical equipment risk;
Step S2: carrying out superposition summation according to respective weights based on the process risk of each process and the evaluation value of each risk at the process position to obtain the construction risk evaluation value of each process;
step S3: determining the risk evaluation level of the area where each working procedure is located according to the threshold range where the construction risk evaluation value of each working procedure is located;
step S4: and generating a security risk four-color map according to the risk evaluation grades of the working procedures.
Further, in the step S1, the specific implementation steps for acquiring the risk at each process and the process position in real time are as follows:
aiming at the acquisition of the working procedure, the specific steps are as follows:
acquiring image and video data of a construction site by using a machine vision technology to acquire corresponding working procedures;
the specific steps for acquiring the natural environment risk are as follows:
cleaning and preprocessing the collected natural risk images, texts and voices, wherein the texts comprise the humidity and the temperature of the environment, and the voices comprise noise;
the cleaning and preprocessing of the natural risk image sequentially comprises noise removal, image size adjustment and color space processing; the method comprises the following specific steps:
removing noise by Gaussian filtering: replacing each pixel value with a weighted average of its neighborhood;
The nearest neighbor interpolation is adopted to adjust the size of the image: setting the pixel value in the target image as the value of the nearest neighbor pixel in the source image to adjust the image size, wherein the source image refers to a natural risk image before the image is not adjusted in size, and the target image refers to a natural risk image after the image is adjusted in size;
color space processing refers to converting a color space from HSV to RGB, and the conversion formula is as follows: red component= (hue 150+30)% 120, green component= (hue 200+50)% 120, blue component= (hue 250+70)% 120, where the RGB color space consists of three components: red, green and blue, the HSV color space consists of hue, saturation and brightness;
the text is cleaned and preprocessed by the following steps:
removing special characters and punctuations in the text by using a regular expression or character string processing method, and only retaining effective text content;
for the reserved valid text content, removing common nonsensical words including 'yes' and 'yes' by using a stop word list;
removing repeated content by comparing adjacent sentences or paragraphs in the text after removing stop words;
the specific steps of cleaning and preprocessing the voice are as follows:
Noise reduction processing is performed on the voice signal by using spectral subtraction in digital signal processing;
performing enhancement processing on the noise-reduced voice signal by using spectral subtraction;
after the adding process, a voice segmentation algorithm is used for separating a voice part from a noise part in the voice signal and removing noise;
extracting natural risk features, text features and voice features of the natural risk images, texts and voices obtained through cleaning and preprocessing by adopting a feature extraction method A, wherein the natural risk features comprise low-level features and high-level features, the low-level features comprise color features, texture features and shape features, the high-level features comprise semantic features, specifically comprise geological structures, ground surface morphology, vegetation coverage and hydrogeological information, and the feature extraction method A comprises a convolutional neural network, a Fourier transformation method, an edge detection method, a color feature extraction method, a word vector feature extraction method and a voice feature extraction method in a machine vision algorithm;
performing time and space alignment on the natural risk features, the text features and the voice features, and performing labeling to obtain a labeled data set;
training the constructed recognition module by the marked data set, and optimizing the recognition module by a transfer learning or increment learning method to obtain a trained recognition model, wherein the recognition model comprises a plurality of encoders corresponding to different types of input data, a convolution layer for splicing the outputs of the encoders, a cooperative encoder for compressing and encoding the output of the convolution layer and an output layer for acquiring feature vector output from feature codes output by the cooperative encoder;
Extracting key features of natural risk images, texts and voices to be identified by adopting a trained identification model to obtain natural environment risks, wherein the natural environment risks comprise one or more of poor geology, water burst and rock burst;
the specific steps for acquiring the construction operation risk are as follows:
based on the obtained initial construction operation image, performing risk point monitoring to obtain monitoring data, including construction operation images and construction records, and cleaning and preprocessing the construction operation images and the construction records;
carrying out construction operation risk feature extraction on the monitoring data by adopting a feature extraction method B, wherein the feature extraction method comprises an image feature and a data feature, the image feature comprises textures, colors and shapes, the data feature comprises time and geographic positions, and the feature extraction method B comprises a convolutional neural network and a statistical method in a machine vision algorithm;
according to the extracted construction operation risk characteristics, carrying out data marking on the images in the monitoring data;
training a machine learning algorithm based on the data set obtained after labeling, wherein the machine learning algorithm comprises a support vector machine, a decision tree and a random forest;
identifying construction operation images to be identified based on a trained machine learning algorithm to obtain construction operation risks, wherein the construction operation risks comprise one or more of drilling and blasting management, collapse in a tunnel, initiating explosive device use, worker posture, equipment state and environmental conditions;
The specific steps for acquiring the risk of the large-scale mechanical equipment are as follows:
accurate position information of the large-scale mechanical equipment entering the field is obtained through a GPS, and simultaneously, the posture of the mechanical equipment, including an inclination angle and a rotation angle, is obtained based on a sensor and an inertial measurement unit;
measuring the movement speed and acceleration of the large-scale mechanical equipment through a speed sensor or an accelerometer;
measuring the load condition of the large-scale mechanical equipment and the pressure exerted on the equipment by a pressure sensor or a weighing sensor;
judging whether the parameters of the large-scale mechanical equipment obtained by each sensor meet the requirements of entering corresponding working procedures or not, if so, entering the corresponding working procedures by the large-scale mechanical equipment, otherwise, not entering the corresponding working procedures;
or (b)
And acquiring image and video data of a construction site by utilizing a machine vision technology, and detecting, identifying and tracking large-scale mechanical equipment to judge whether the large-scale mechanical equipment enters a corresponding procedure or not.
Further, the formula of the construction risk evaluation value in step S2 is:
B nJ ＝B n1 +B n2 +B n3 +…+B nj
C nK ＝C n1 +C n2 +C n3 +…+C nk
D nL ＝D n1 +D n2 +D n3 +…+D nl
wherein M is n A represents the construction risk evaluation value of the nth step n Representing the process risk score in the nth process, B nJ Sum of scores representing J natural environment risks in the nth step, C nK Representing the sum of the scores of the risks of K construction operations in the nth step, D nL The sum of the scores representing the risk of the L large mechanical equipment in the nth procedure,and->The weights of the working procedure risk, the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk in each given working procedure are respectively expressed.
A tunnel security risk identification device, comprising:
the evaluation value acquisition module: for each obtained procedure, obtaining the number of items of risk at each procedure position, and giving the procedure risk of each procedure and the evaluation value of each risk at the procedure position according to the number of items of risk at the procedure position, wherein the procedure comprises one or more of advanced support, drilling and blasting construction, primary support, lining structure, portal structure, stepping, groove, water-proof component and open cut backfill, and each risk at the procedure position comprises natural environment risk, construction operation risk and large-scale mechanical equipment risk;
construction risk evaluation value acquisition module: carrying out superposition summation according to respective weights based on the process risk of each process and the evaluation value of each risk at the process position to obtain the construction risk evaluation value of each process;
risk evaluation grade acquisition module: determining the risk evaluation level of the area where each working procedure is located according to the threshold range where the construction risk evaluation value of each working procedure is located;
Four-color map generation module: and generating a security risk four-color map according to the risk evaluation grades of the working procedures.
Further, in the evaluation value acquisition module, the specific implementation steps of acquiring the risk at each process and the process position in real time are as follows:
aiming at the acquisition of the working procedure, the specific steps are as follows:
acquiring image and video data of a construction site by using a machine vision technology to acquire corresponding working procedures;
the specific steps for acquiring the natural environment risk are as follows:
cleaning and preprocessing the collected natural risk images, texts and voices, wherein the texts comprise the humidity and the temperature of the environment, and the voices comprise noise;
the cleaning and preprocessing of the natural risk image sequentially comprises noise removal, image size adjustment and color space processing; the method comprises the following specific steps:
removing noise by Gaussian filtering: replacing each pixel value with a weighted average of its neighborhood;
the nearest neighbor interpolation is adopted to adjust the size of the image: setting the pixel value in the target image as the value of the nearest neighbor pixel in the source image to adjust the image size, wherein the source image refers to a natural risk image before the image is not adjusted in size, and the target image refers to a natural risk image after the image is adjusted in size;
Color space processing refers to converting a color space from HSV to RGB, and the conversion formula is as follows: red component= (hue 150+30)% 120, green component= (hue 200+50)% 120, blue component= (hue 250+70)% 120, where the RGB color space consists of three components: red, green and blue, the HSV color space consists of hue, saturation and brightness;
the text is cleaned and preprocessed by the following steps:
removing special characters and punctuations in the text by using a regular expression or character string processing method, and only retaining effective text content;
for the reserved valid text content, removing common nonsensical words including 'yes' and 'yes' by using a stop word list;
removing repeated content by comparing adjacent sentences or paragraphs in the text after removing stop words;
the specific steps of cleaning and preprocessing the voice are as follows:
noise reduction processing is performed on the voice signal by using spectral subtraction in digital signal processing;
performing enhancement processing on the noise-reduced voice signal by using spectral subtraction;
after the adding process, a voice segmentation algorithm is used for separating a voice part from a noise part in the voice signal and removing noise;
Extracting natural risk features, text features and voice features of the natural risk images, texts and voices obtained through cleaning and preprocessing by adopting a feature extraction method A, wherein the natural risk features comprise low-level features and high-level features, the low-level features comprise color features, texture features and shape features, the high-level features comprise semantic features, specifically comprise geological structures, ground surface morphology, vegetation coverage and hydrogeological information, and the feature extraction method A comprises a convolutional neural network, a Fourier transformation method, an edge detection method, a color feature extraction method, a word vector feature extraction method and a voice feature extraction method in a machine vision algorithm;
performing time and space alignment on the natural risk features, the text features and the voice features, and performing labeling to obtain a labeled data set;
training the constructed recognition module by the marked data set, and optimizing the recognition module by a transfer learning or increment learning method to obtain a trained recognition model, wherein the recognition model comprises a plurality of encoders corresponding to different types of input data, a convolution layer for splicing the outputs of the encoders, a cooperative encoder for compressing and encoding the output of the convolution layer and an output layer for acquiring feature vector output from feature codes output by the cooperative encoder;
Extracting key features of natural risk images, texts and voices to be identified by adopting a trained identification model to obtain natural environment risks, wherein the natural environment risks comprise one or more of poor geology, water burst and rock burst;
the specific steps for acquiring the construction operation risk are as follows:
based on the obtained initial construction operation image, performing risk point monitoring to obtain monitoring data, including construction operation images and construction records, and cleaning and preprocessing the construction operation images and the construction records;
carrying out construction operation risk feature extraction on the monitoring data by adopting a feature extraction method B, wherein the feature extraction method comprises an image feature and a data feature, the image feature comprises textures, colors and shapes, the data feature comprises time and geographic positions, and the feature extraction method B comprises a convolutional neural network and a statistical method in a machine vision algorithm;
according to the extracted construction operation risk characteristics, carrying out data marking on the images in the monitoring data;
training a machine learning algorithm based on the data set obtained after labeling, wherein the machine learning algorithm comprises a support vector machine, a decision tree and a random forest;
identifying construction operation images to be identified based on a trained machine learning algorithm to obtain construction operation risks, wherein the construction operation risks comprise one or more of drilling and blasting management, collapse in a tunnel, initiating explosive device use, worker posture, equipment state and environmental conditions;
The specific steps for acquiring the risk of the large-scale mechanical equipment are as follows:
accurate position information of the large-scale mechanical equipment entering the field is obtained through a GPS, and simultaneously, the posture of the mechanical equipment, including an inclination angle and a rotation angle, is obtained based on a sensor and an inertial measurement unit;
measuring the movement speed and acceleration of the large-scale mechanical equipment through a speed sensor or an accelerometer;
measuring the load condition of the large-scale mechanical equipment and the pressure exerted on the equipment by a pressure sensor or a weighing sensor;
judging whether the parameters of the large-scale mechanical equipment obtained by each sensor meet the requirements of entering corresponding working procedures or not, if so, entering the corresponding working procedures by the large-scale mechanical equipment, otherwise, not entering the corresponding working procedures;
or (b)
And acquiring image and video data of a construction site by utilizing a machine vision technology, and detecting, identifying and tracking large-scale mechanical equipment to judge whether the large-scale mechanical equipment enters a corresponding procedure or not.
Further, the formula of the construction risk evaluation value in the construction risk evaluation value obtaining module is as follows:
B nJ ＝B n1 +B n2 +B n3 +…+B nj
C nK ＝C n1 +C n2 +C n3 +…+C nk
D nL ＝D n1 +D n2 +D n3 +…+D nl
wherein M is n A represents the construction risk evaluation value of the nth step n Representing the process risk score in the nth process, B nJ Sum of scores representing J natural environment risks in the nth step, C nK Representing the sum of the scores of the risks of K construction operations in the nth step, D nL The sum of the scores representing the risk of the L large mechanical equipment in the nth procedure,and->The weights of the working procedure risk, the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk in each given working procedure are respectively expressed.
A tunnel security risk identification security management platform, comprising:
and (3) a data arrangement system: the method comprises the steps of collecting tunnel construction operation data and sorting, and obtaining a tunnel excavation construction operation flow, tunnel engineering construction operation content and a high risk circulation process risk category list after sorting;
basic reserve system: creating a knowledge base based on the result obtained by the data arrangement module;
machine vision analysis system: generating a security risk four-color map by adopting a tunnel security risk identification method based on the current procedure;
analysis processing system: and generating a processing scheme based on the knowledge base and the security risk four-color map, and obtaining security early warning pushing, risk four-color map displaying and emergency scheme pushing.
Compared with the prior art, the invention has the advantages that:
1. the invention provides a more effective, more timely and more intelligent safety management method for the safety management of tunnel construction based on the construction of the machine vision safety risk four-color map; the concrete implementation is as follows:
According to the invention, the acquired natural risk images, texts and voices are subjected to multi-mode fusion training through the constructed recognition model, natural environment risk items of all working procedures are accurately distinguished after transfer learning or increment learning is performed and optimized, meanwhile, the recognized construction operation risk items, the number of large-sized mechanical equipment in all working procedures and the recognized working procedure risk items are combined to accurately generate a security risk four-color map according to given weights, so that more effective, more timely and more intelligent security management is provided for the security management of tunnel construction;
2. according to the invention, the process risk and the process related risk are comprehensively considered, and the risk assessment consideration factors are more comprehensive, so that the early warning is more accurate;
3. the method realizes the visualization of the risk of the tunnel construction operation area, ensures the update of the tunnel construction risk in real time, and ensures the early warning of the tunnel construction risk in time;
4. the invention provides a comprehensive early warning scheme comprising the functions of twins visualization database construction, safety risk diagnosis, safety risk comprehensive prediction and the like.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings that are needed in the embodiments will be briefly described below, it being understood that the following drawings only illustrate some embodiments of the present invention and should not be considered limiting the scope, and that other related drawings can be obtained according to these drawings without inventive effort for a person skilled in the art.
FIG. 1 is a schematic flow chart of construction safety risk four-color thermodynamic diagram generation in the invention;
FIG. 2 is a schematic diagram of tunnel risk thermodynamic diagram update in accordance with the present invention;
FIG. 3 is a schematic diagram of a machine vision analysis flow chart in the present invention;
FIG. 4 is a schematic view of process risk color partitioning according to the present invention;
FIG. 5 is a schematic diagram of a machine vision analysis process according to the present invention.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present invention more apparent, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention, and it is apparent that the described embodiments are some embodiments of the present invention, but not all embodiments of the present invention. All other embodiments, which can be made by those skilled in the art based on the embodiments of the invention without making any inventive effort, are intended to be within the scope of the invention.
In this embodiment, please refer to fig. 1, fig. 1 is a schematic flow chart of a tunnel security risk identification method provided in this case, and the tunnel security risk identification method obtains required evaluation procedures, performs security risk identification on each obtained procedure, and includes the following steps:
Step S1: and for each acquired procedure, acquiring the number of items of risk at each procedure position, and giving the procedure risk of each procedure and the evaluation value of each risk at the procedure position according to the number of items of risk at the procedure position, wherein the procedure comprises one or more of advanced support, drilling and blasting construction, primary support, lining structure, portal structure, stepping, groove, water-proof and drainage member and open cut backfill, and each risk at the procedure position comprises natural environment risk, construction operation risk and large-scale mechanical equipment risk. Through the effective classification of the procedures and the risk items, the efficiency of identifying the risk items in the subsequent steps can be effectively improved.
The specific implementation steps for acquiring risks at each process and process position in real time are as follows:
for the acquisition of the working procedure, the specific steps comprise:
and acquiring image and video data of a construction site by using a machine vision technology so as to acquire corresponding working procedures. In order to accurately acquire different types of risks, different acquisition methods can be used for identifying and processing different types of risk features.
As shown in fig. 3, the specific steps for acquiring the natural environment risk are as follows:
Cleaning and preprocessing the collected natural risk images, texts and voices, wherein the texts comprise the humidity and the temperature of the environment, and the voices comprise noise;
the cleaning and preprocessing of the natural risk image sequentially comprises noise removal, image size adjustment and color space processing; the method comprises the following specific steps:
removing noise by Gaussian filtering: replacing each pixel value with a weighted average of its neighborhood;
the nearest neighbor interpolation is adopted to adjust the size of the image: setting the pixel value in the target image as the value of the nearest neighbor pixel in the source image to adjust the image size, wherein the source image refers to a natural risk image before the image is not adjusted in size, and the target image refers to a natural risk image after the image is adjusted in size;
color space processing refers to converting a color space from HSV to RGB, and the conversion formula is as follows: red component= (hue 150+30)% 120, green component= (hue 200+50)% 120, blue component= (hue 250+70)% 120, where the RGB color space consists of three components: red, green and blue, the HSV color space consists of hue, saturation and brightness;
the text is cleaned and preprocessed by the following steps:
removing special characters and punctuations in the text by using a regular expression or character string processing method, and only retaining effective text content;
For the reserved effective text content, removing common nonsensical words including 'yes' by using a stop word list so as to reduce noise influence;
after the stop words are removed, the repeated content is removed by comparing adjacent sentences or paragraphs in the text, so that noise and redundant information are reduced;
the specific steps of cleaning and preprocessing the voice are as follows:
noise reduction processing is carried out on the voice signal by using spectral subtraction in digital signal processing, so that the influence of noise is reduced;
using spectral subtraction to enhance the noise-reduced voice signal, and improving the quality and definition of voice;
after the adding processing, a voice segmentation algorithm is used for separating a voice part from a noise part in the voice signal, and removing noise, so that the noise can be better processed;
and extracting natural risk features, text features and voice features of the natural risk images, texts and voices obtained through cleaning and preprocessing by adopting a feature extraction method A, wherein the natural risk features comprise low-level features and high-level features, the low-level features comprise color features, texture features and shape features, the high-level features comprise semantic features, and specifically comprise geological structures, ground surface morphology, vegetation coverage and hydrogeological information, and the feature extraction method A comprises a convolutional neural network, a Fourier transformation method, an edge detection method, a color feature extraction method, a word vector feature extraction method and a voice feature extraction method in a machine vision algorithm.
In the scheme, the image, the text and the voice realize comprehensive judgment of the characteristics by a multi-mode characteristic extraction method, and the defects of inaccurate characteristic identification and low identification efficiency of the existing image are overcome. By combining the text features and the voice features serving as one of the features of risk recognition with the conventional image features, the recognition accuracy and recognition efficiency of natural risks can be effectively improved.
In order to enable the risk feature to be identified more efficiently by using the identification method, an identification module may be constructed by using the identification method, and the specific steps may include:
performing time and space alignment on the natural risk features, the text features and the voice features, and performing labeling to obtain a labeled data set;
training the constructed recognition module by the marked data set, and optimizing the recognition module by a transfer learning or increment learning method to obtain a trained recognition model, wherein the recognition model comprises a plurality of encoders corresponding to different types of input data, a convolution layer for splicing the outputs of the encoders, a cooperative encoder for compressing and encoding the output of the convolution layer and an output layer for acquiring feature vector output from feature codes output by the cooperative encoder;
And extracting key features of the natural risk image, text and voice to be identified by adopting the trained identification model to obtain natural environment risks including one or more of poor geology, water burst and rock burst. The image, text and voice are subjected to the multi-mode feature extraction method to realize the risk identification method for comprehensively judging the features, and the risk identification method for construction operation risks and large-scale mechanical equipment risks are also applied, so that the beneficial effects and application schemes can refer to the natural risk identification method, and are not repeated later.
As shown in fig. 3, the specific steps for acquiring the risk of the construction work are as follows:
based on the obtained initial construction operation image, performing risk point monitoring to obtain monitoring data, including construction operation images and construction records, and cleaning and preprocessing the construction operation images and the construction records;
carrying out construction operation risk feature extraction on the monitoring data by adopting a feature extraction method B, wherein the feature extraction method comprises an image feature and a data feature, the image feature comprises textures, colors and shapes, the data feature comprises time and geographic positions, and the feature extraction method B comprises a convolutional neural network and a statistical method in a machine vision algorithm;
According to the extracted construction operation risk characteristics, carrying out data marking on the images in the monitoring data;
training a machine learning algorithm based on the data set obtained after labeling, wherein the machine learning algorithm comprises a support vector machine, a decision tree and a random forest;
identifying construction operation images to be identified based on a trained machine learning algorithm to obtain construction operation risks, wherein the construction operation risks comprise one or more of drilling and blasting management, collapse in a tunnel, initiating explosive device use, worker posture, equipment state and environmental conditions;
as shown in fig. 3, the specific steps for acquiring the risk of the large-scale mechanical equipment are as follows:
accurate position information of the large-scale mechanical equipment entering the field is obtained through a GPS, and simultaneously, the posture of the mechanical equipment, including an inclination angle and a rotation angle, is obtained based on a sensor and an inertial measurement unit;
measuring the movement speed and acceleration of the large-scale mechanical equipment through a speed sensor or an accelerometer;
measuring the load condition of the large-scale mechanical equipment and the pressure exerted on the equipment by a pressure sensor or a weighing sensor;
judging whether the parameters of the large-scale mechanical equipment obtained by each sensor meet the requirements of entering corresponding working procedures or not, if so, entering the corresponding working procedures by the large-scale mechanical equipment, otherwise, not entering the corresponding working procedures;
Or (b)
And acquiring image and video data of a construction site by utilizing a machine vision technology, and detecting, identifying and tracking large-scale mechanical equipment to judge whether the large-scale mechanical equipment enters a corresponding procedure or not.
Step S2: carrying out superposition summation according to respective weights based on the process risk of each process and the evaluation value of each risk at the process position to obtain the construction risk evaluation value of each process; the formula of the construction risk evaluation value is as follows:
B nJ ＝B n1 +B n2 +B n3 +…+B nj
C nK ＝C n1 +C n2 +C n3 +…+C nk
D nL ＝D n1 +D n2 +D n3 +…+D ni
wherein M is n A represents the construction risk evaluation value of the nth step n Representing the process risk score in the nth process, B nJ Sum of scores representing J natural environment risks in the nth step, C nK Representing the sum of the scores of the risks of K construction operations in the nth step, D nL The sum of the scores representing the risk of the L large mechanical equipment in the nth procedure,and->The weights of the working procedure risk, the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk in each given working procedure are respectively expressed.
Step S3: determining the risk evaluation level of the area where each working procedure is located according to the threshold range where the construction risk evaluation value of each working procedure is located;
step S4: according to the risk evaluation grades of all the working procedures, a safety risk four-color map (four-color map for short) is generated, namely a red, orange, yellow and blue four-color risk thermodynamic map is generated, the four-color map is based on the risk grade of the tunnel construction working procedure, the large, general and low-risk working procedures are identified by red, orange and yellow Lan Si, and detailed information such as procedure names, risk descriptions and management and control measures is built in the four-color map, so that cyclic tunneling along with the tunnel working procedures is realized, and the risk grade of a construction area is dynamically adjusted.
Examples
As shown in fig. 4, the method for determining the evaluation value of the process risk is as follows: determining an evaluation value according to the process color in the WBS process exploded view, wherein the evaluation value is 100 minutes in red, 80 minutes in orange, 60 minutes in yellow and 40 minutes in blue;
the risk evaluation value determination schemes at the process positions are as follows: the initial values of the risk evaluation values of the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk are 0, 10 is added to one risk in each occurrence of the references, namely the corresponding evaluation value is 10 minutes, for example, 10 natural environment risks in one working procedure are added, namely the sum of all natural environment risk scores in the working procedure is obtained by adding 10 natural environment risks, the maximum value is not set, and calculation is carried out according to practical requirements.
After judging and summarizing the specific gravity of each risk in each risk item by organizing relevant technicians, determining the weight distribution of each risk item as follows:
| risk item | Risk of process | Risk of natural environment | Risk of construction work | Risk of large-scale mechanical equipment |
| Weighting of | 0.63 | 0.12 | 0.1 | 0.15 |
Total risk score = 0.63 per process + sum of all natural environmental risk scores per process 0.12 per process + sum of all construction work risk scores per process 0.1 per process + sum of all large machinery risk scores per process 0.15. Therefore, through accurate scoring and weight division, the risk level of each risk can be accurately evaluated, and the generated risk four-color chart can be ensured to have stronger guiding efficacy.
It should be noted that, the determination and the acquisition of the weight values are determined by integrating the field experience and the expert group scoring conclusion result, so that the guiding effectiveness of the risk four-color chart on the safety risk management of the construction field can be further ensured.
For example, assuming that the process is a yellow process in fig. 4, the process risk is divided into 60 points, and the natural environment risk, the construction operation risk and the large mechanical equipment risk each have 1 risk item, the initial points of the natural environment risk, the construction operation risk and the large mechanical equipment risk are all 10 points, and the total risk score is:
60*0.63+10*0.12+10*0.1+10*0.15＝41.5
the corresponding color is determined according to the threshold range in the following table.
At this time, the color in the risk four-color chart is yellow.
Assuming that the process is a yellow process in fig. 4, the process risk is divided into 60 points, and the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk each have 3 risk items, and then the initial points of the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk are all 30 points, and the total risk scores are:
60*0.63+30*0.12+30*0.1+30*0.15＝48.9
at this time, the process color in the risk four-color chart is displayed as orange.
In this case, the risk four-color thermodynamic diagram is continuously updated along with the tunnel circulation procedure, and the specific process includes:
the machine vision algorithm identifies the tunnel construction process change, the four-color thermodynamic map of the tunnel risk is updated through the construction risk evaluation value obtained in real time, and fig. 2 is a schematic diagram of updating the thermodynamic diagram of the tunnel risk.
The tunnel circulation process is continuously advanced, after the excavation of the tunnel body is finished, the original excavation area of the tunnel body is converted into a supporting area, the non-excavation area is converted into an advanced operation undetermined area and a tunnel body excavation undetermined area, and after equipment and personnel approach is detected, the undetermined area is converted into a formal tunnel body excavation area and an advanced operation undetermined area.
Identifying tunnel construction process variations includes: personnel approach recognition, equipment approach recognition and tunnel procedure feature recognition.
Personnel approach identification includes: the personnel wear the identification; whether wearing a safety helmet is identified; and (5) identifying the personnel position.
The device approach identification includes: entering a large-scale device; the small and medium-sized equipment enters the ground.
The tunnel process feature identification includes: a device location; the equipment use condition; the operator position and status.
In summary, the construction of the machine vision-based security risk four-color map provides a more effective, more timely and more intelligent security management method for the security management of tunnel construction; the method comprises the steps of carrying out multi-mode fusion training on the acquired natural risk images, texts and voices through a constructed recognition model, carrying out tuning and optimizing through transfer learning or increment learning, accurately identifying natural environment risk items of all working procedures, and simultaneously combining the recognized construction operation risk items, the number of large-sized mechanical equipment in all working procedures and the recognized working procedure risk items to accurately generate a security risk four-color map according to given weights so as to provide more effective, more timely and more intelligent security management for the security management of tunnel construction; according to the invention, the process risk and the process related risk are comprehensively considered, and the risk assessment consideration factors are more comprehensive, so that the early warning is more accurate.
As shown in fig. 5, a tunnel security risk identification security management platform includes:
and (3) a data arrangement system: the method comprises the steps of collecting tunnel construction operation data and sorting, and obtaining a tunnel excavation construction operation flow, tunnel engineering construction operation content and a high risk circulation process risk category list after sorting;
basic reserve system: creating a knowledge base based on the result obtained by the data arrangement module;
machine vision analysis system: generating a security risk four-color map by adopting a tunnel security risk identification method based on the current procedure;
analysis processing system: and generating a processing scheme based on the knowledge base and the security risk four-color map, and obtaining security early warning pushing, risk four-color map displaying and emergency scheme pushing. The knowledge base construction is the basis of the management platform, and the knowledge base is constructed through the data arrangement module and the basic storage module. The data sorting system is a front-end module of the basic storage module, and has the functions of manually uploading and calling information such as construction flow, operation content, risk list and the like from a related platform; the machine vision analysis system and the analysis processing system behind the basic storage module serve as the basis of the knowledge base to provide corresponding information. And determining the risk level and the risk item through an analysis processing system, finding out the corresponding solution and responsibility level in a knowledge base after identifying the risk item, and pushing the risk item and the solution to a responsible person. In another possible implementation manner, when it is recognized that the worker does not wear the helmet and the tunnel has various risks such as a falling stone risk, the updating adjustment of the four-color chart is triggered.
Claims (7)
1. The tunnel security risk identification method is characterized by obtaining required evaluation procedures and carrying out security risk identification on each obtained procedure, and comprises the following steps:
step S1: for each obtained procedure, obtaining the number of items of risk at each procedure position, and giving the procedure risk of each procedure and the evaluation value of each risk at the procedure position according to the number of items of risk at the procedure position, wherein the procedure comprises one or more of advanced support, drilling and blasting construction, primary support, lining structure, portal structure, stepping, groove, water-proof component and open cut backfill, and each risk at the procedure position comprises natural environment risk, construction operation risk and large-scale mechanical equipment risk;
step S2: carrying out superposition summation according to respective weights based on the process risk of each process and the evaluation value of each risk at the process position to obtain the construction risk evaluation value of each process;
step S3: determining the risk evaluation level of the area where each working procedure is located according to the threshold range where the construction risk evaluation value of each working procedure is located;
step S4: and generating a security risk four-color map according to the risk evaluation grades of the working procedures.
2. The tunnel security risk identification method according to claim 1, wherein in the step S1, the specific implementation steps of acquiring the risk at each process and the process position in real time are as follows:
Aiming at the acquisition of the working procedure, the specific steps are as follows:
acquiring image and video data of a construction site by using a machine vision technology to acquire corresponding working procedures;
the specific steps for acquiring the natural environment risk are as follows:
cleaning and preprocessing the collected natural risk images, texts and voices, wherein the texts comprise the humidity and the temperature of the environment, and the voices comprise noise;
the cleaning and preprocessing of the natural risk image sequentially comprises noise removal, image size adjustment and color space processing; the method comprises the following specific steps:
removing noise by Gaussian filtering: replacing each pixel value with a weighted average of its neighborhood;
the nearest neighbor interpolation is adopted to adjust the size of the image: setting the pixel value in the target image as the value of the nearest neighbor pixel in the source image to adjust the image size, wherein the source image refers to a natural risk image before the image is not adjusted in size, and the target image refers to a natural risk image after the image is adjusted in size;
color space processing refers to converting a color space from HSV to RGB, and the conversion formula is as follows: red component= (hue 150+30)% 120, green component= (hue 200+50)% 120, blue component= (hue 250+70)% 120, where the RGB color space consists of three components: red, green and blue, the HSV color space consists of hue, saturation and brightness;
The text is cleaned and preprocessed by the following steps:
removing special characters and punctuations in the text by using a regular expression or character string processing method, and only retaining effective text content;
for the reserved valid text content, removing common nonsensical words including 'yes' and 'yes' by using a stop word list;
removing repeated content by comparing adjacent sentences or paragraphs in the text after removing stop words;
the specific steps of cleaning and preprocessing the voice are as follows:
noise reduction processing is performed on the voice signal by using spectral subtraction in digital signal processing;
performing enhancement processing on the noise-reduced voice signal by using spectral subtraction;
after the adding process, a voice segmentation algorithm is used for separating a voice part from a noise part in the voice signal and removing noise;
extracting natural risk features, text features and voice features of the natural risk images, texts and voices obtained through cleaning and preprocessing by adopting a feature extraction method A, wherein the natural risk features comprise low-level features and high-level features, the low-level features comprise color features, texture features and shape features, the high-level features comprise semantic features, specifically comprise geological structures, ground surface morphology, vegetation coverage and hydrogeological information, and the feature extraction method A comprises a convolutional neural network, a Fourier transformation method, an edge detection method, a color feature extraction method, a word vector feature extraction method and a voice feature extraction method in a machine vision algorithm;
Performing time and space alignment on the natural risk features, the text features and the voice features, and performing labeling to obtain a labeled data set;
training the constructed recognition module by the marked data set, and optimizing the recognition module by a transfer learning or increment learning method to obtain a trained recognition model, wherein the recognition model comprises a plurality of encoders corresponding to different types of input data, a convolution layer for splicing the outputs of the encoders, a cooperative encoder for compressing and encoding the output of the convolution layer and an output layer for acquiring feature vector output from feature codes output by the cooperative encoder;
extracting key features of natural risk images, texts and voices to be identified by adopting a trained identification model to obtain natural environment risks, wherein the natural environment risks comprise one or more of poor geology, water burst and rock burst;
the specific steps for acquiring the construction operation risk are as follows:
based on the obtained initial construction operation image, performing risk point monitoring to obtain monitoring data, including construction operation images and construction records, and cleaning and preprocessing the construction operation images and the construction records;
carrying out construction operation risk feature extraction on the monitoring data by adopting a feature extraction method B, wherein the feature extraction method comprises an image feature and a data feature, the image feature comprises textures, colors and shapes, the data feature comprises time and geographic positions, and the feature extraction method B comprises a convolutional neural network and a statistical method in a machine vision algorithm;
According to the extracted construction operation risk characteristics, carrying out data marking on the images in the monitoring data;
training a machine learning algorithm based on the data set obtained after labeling, wherein the machine learning algorithm comprises a support vector machine, a decision tree and a random forest;
identifying construction operation images to be identified based on a trained machine learning algorithm to obtain construction operation risks, wherein the construction operation risks comprise one or more of drilling and blasting management, collapse in a tunnel, initiating explosive device use, worker posture, equipment state and environmental conditions;
the specific steps for acquiring the risk of the large-scale mechanical equipment are as follows:
accurate position information of the large-scale mechanical equipment entering the field is obtained through a GPS, and simultaneously, the posture of the mechanical equipment, including an inclination angle and a rotation angle, is obtained based on a sensor and an inertial measurement unit;
measuring the movement speed and acceleration of the large-scale mechanical equipment through a speed sensor or an accelerometer;
measuring the load condition of the large-scale mechanical equipment and the pressure exerted on the equipment by a pressure sensor or a weighing sensor;
judging whether the parameters of the large-scale mechanical equipment obtained by each sensor meet the requirements of entering corresponding working procedures or not, if so, entering the corresponding working procedures by the large-scale mechanical equipment, otherwise, not entering the corresponding working procedures;
Or (b)
And acquiring image and video data of a construction site by utilizing a machine vision technology, and detecting, identifying and tracking large-scale mechanical equipment to judge whether the large-scale mechanical equipment enters a corresponding procedure or not.
3. The method for identifying the security risk of the tunnel according to claim 2, wherein the formula of the construction risk evaluation value in the step S2 is:
B nJ ＝B n1 +B n2 +B n3 +…+B nj
C nK ＝C n1 +C n2 +C n3 +…+C nk
D nL ＝D n1 +D n2 +D n3 +…+D nl
wherein M is n A represents the construction risk evaluation value of the nth step n Representing the process risk score in the nth process, B nJ Sum of scores representing J natural environment risks in the nth step, C nK Representing the sum of the scores of the risks of K construction operations in the nth step, D nL The sum of the scores representing the risk of the L large mechanical equipment in the nth procedure,andthe weights of the working procedure risk, the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk in each given working procedure are respectively expressed.
4. A tunnel security risk identification device, comprising:
the evaluation value acquisition module: for each obtained procedure, obtaining the number of items of risk at each procedure position, and giving the procedure risk of each procedure and the evaluation value of each risk at the procedure position according to the number of items of risk at the procedure position, wherein the procedure comprises one or more of advanced support, drilling and blasting construction, primary support, lining structure, portal structure, stepping, groove, water-proof component and open cut backfill, and each risk at the procedure position comprises natural environment risk, construction operation risk and large-scale mechanical equipment risk;
Construction risk evaluation value acquisition module: carrying out superposition summation according to respective weights based on the process risk of each process and the evaluation value of each risk at the process position to obtain the construction risk evaluation value of each process;
risk evaluation grade acquisition module: determining the risk evaluation level of the area where each working procedure is located according to the threshold range where the construction risk evaluation value of each working procedure is located;
four-color map generation module: and generating a security risk four-color map according to the risk evaluation grades of the working procedures.
5. The tunnel security risk identification device according to claim 4, wherein the specific implementation steps of acquiring the risk at each process and each process position in real time in the evaluation value acquisition module are as follows:
aiming at the acquisition of the working procedure, the specific steps are as follows:
acquiring image and video data of a construction site by using a machine vision technology to acquire corresponding working procedures;
the specific steps for acquiring the natural environment risk are as follows:
cleaning and preprocessing the collected natural risk images, texts and voices, wherein the texts comprise the humidity and the temperature of the environment, and the voices comprise noise;
the cleaning and preprocessing of the natural risk image sequentially comprises noise removal, image size adjustment and color space processing; the method comprises the following specific steps:
Removing noise by Gaussian filtering: replacing each pixel value with a weighted average of its neighborhood;
the nearest neighbor interpolation is adopted to adjust the size of the image: setting the pixel value in the target image as the value of the nearest neighbor pixel in the source image to adjust the image size, wherein the source image refers to a natural risk image before the image is not adjusted in size, and the target image refers to a natural risk image after the image is adjusted in size;
color space processing refers to converting a color space from HSV to RGB, and the conversion formula is as follows: red component= (hue 150+30)% 120, green component= (hue 200+50)% 120, blue component= (hue 250+70)% 120, where the RGB color space consists of three components: red, green and blue, the HSV color space consists of hue, saturation and brightness;
the text is cleaned and preprocessed by the following steps:
removing special characters and punctuations in the text by using a regular expression or character string processing method, and only retaining effective text content;
for the reserved valid text content, removing common nonsensical words including 'yes' and 'yes' by using a stop word list;
removing repeated content by comparing adjacent sentences or paragraphs in the text after removing stop words;
The specific steps of cleaning and preprocessing the voice are as follows:
noise reduction processing is performed on the voice signal by using spectral subtraction in digital signal processing;
performing enhancement processing on the noise-reduced voice signal by using spectral subtraction;
after the adding process, a voice segmentation algorithm is used for separating a voice part from a noise part in the voice signal and removing noise;
extracting natural risk features, text features and voice features of the natural risk images, texts and voices obtained through cleaning and preprocessing by adopting a feature extraction method A, wherein the natural risk features comprise low-level features and high-level features, the low-level features comprise color features, texture features and shape features, the high-level features comprise semantic features, specifically comprise geological structures, ground surface morphology, vegetation coverage and hydrogeological information, and the feature extraction method A comprises a convolutional neural network, a Fourier transformation method, an edge detection method, a color feature extraction method, a word vector feature extraction method and a voice feature extraction method in a machine vision algorithm;
performing time and space alignment on the natural risk features, the text features and the voice features, and performing labeling to obtain a labeled data set;
Training the constructed recognition module by the marked data set, and optimizing the recognition module by a transfer learning or increment learning method to obtain a trained recognition model, wherein the recognition model comprises a plurality of encoders corresponding to different types of input data, a convolution layer for splicing the outputs of the encoders, a cooperative encoder for compressing and encoding the output of the convolution layer and an output layer for acquiring feature vector output from feature codes output by the cooperative encoder;
extracting key features of natural risk images, texts and voices to be identified by adopting a trained identification model to obtain natural environment risks, wherein the natural environment risks comprise one or more of poor geology, water burst and rock burst;
the specific steps for acquiring the construction operation risk are as follows:
based on the obtained initial construction operation image, performing risk point monitoring to obtain monitoring data, including construction operation images and construction records, and cleaning and preprocessing the construction operation images and the construction records;
carrying out construction operation risk feature extraction on the monitoring data by adopting a feature extraction method B, wherein the feature extraction method comprises an image feature and a data feature, the image feature comprises textures, colors and shapes, the data feature comprises time and geographic positions, and the feature extraction method B comprises a convolutional neural network and a statistical method in a machine vision algorithm;
According to the extracted construction operation risk characteristics, carrying out data marking on the images in the monitoring data;
training a machine learning algorithm based on the data set obtained after labeling, wherein the machine learning algorithm comprises a support vector machine, a decision tree and a random forest;
identifying construction operation images to be identified based on a trained machine learning algorithm to obtain construction operation risks, wherein the construction operation risks comprise one or more of drilling and blasting management, collapse in a tunnel, initiating explosive device use, worker posture, equipment state and environmental conditions;
the specific steps for acquiring the risk of the large-scale mechanical equipment are as follows:
accurate position information of the large-scale mechanical equipment entering the field is obtained through a GPS, and simultaneously, the posture of the mechanical equipment, including an inclination angle and a rotation angle, is obtained based on a sensor and an inertial measurement unit;
measuring the movement speed and acceleration of the large-scale mechanical equipment through a speed sensor or an accelerometer;
measuring the load condition of the large-scale mechanical equipment and the pressure exerted on the equipment by a pressure sensor or a weighing sensor;
judging whether the parameters of the large-scale mechanical equipment obtained by each sensor meet the requirements of entering corresponding working procedures or not, if so, entering the corresponding working procedures by the large-scale mechanical equipment, otherwise, not entering the corresponding working procedures;
Or (b)
And acquiring image and video data of a construction site by utilizing a machine vision technology, and detecting, identifying and tracking large-scale mechanical equipment to judge whether the large-scale mechanical equipment enters a corresponding procedure or not.
6. The tunnel security risk identification device according to claim 5, wherein the formula of the construction risk evaluation value in the construction risk evaluation value obtaining module is:
B nJ ＝B n1 +B n2 +B n3 +…+B nj
C nK ＝C n1 +C n2 +C n3 +…+C nk
D nL ＝D n1 +D n2 +D n3 +…+D nl
wherein M is n A represents the construction risk evaluation value of the nth step n Representing the process risk score in the nth process, B nJ Sum of scores representing J natural environment risks in the nth step, C nK Representing the sum of the scores of the risks of K construction operations in the nth step, D nL The sum of the scores representing the risk of the L large mechanical equipment in the nth procedure,andthe weights of the working procedure risk, the natural environment risk, the construction operation risk and the large-scale mechanical equipment risk in each given working procedure are respectively expressed.
7. A tunnel security risk identification security management platform, comprising:
and (3) a data arrangement system: the method comprises the steps of collecting tunnel construction operation data and sorting, and obtaining a tunnel excavation construction operation flow, tunnel engineering construction operation content and a high risk circulation process risk category list after sorting;
Basic reserve system: creating a knowledge base based on the result obtained by the data arrangement module;
machine vision analysis system: generating a security risk four-color map by adopting a tunnel security risk identification method based on the current procedure;
analysis processing system: and generating a processing scheme based on the knowledge base and the security risk four-color map, and obtaining security early warning pushing, risk four-color map displaying and emergency scheme pushing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311707645.4A CN117522149A (en) | 2023-12-12 | 2023-12-12 | Tunnel security risk identification method and device and security management platform |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311707645.4A CN117522149A (en) | 2023-12-12 | 2023-12-12 | Tunnel security risk identification method and device and security management platform |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117522149A true CN117522149A (en) | 2024-02-06 |
Family
ID=89753159
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311707645.4A Pending CN117522149A (en) | 2023-12-12 | 2023-12-12 | Tunnel security risk identification method and device and security management platform |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117522149A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117877735A (en) * | 2024-03-12 | 2024-04-12 | 中南大学 | Tunnel constructor thermal risk monitoring and early warning system and method |
| CN117877735B (en) * | 2024-03-12 | 2024-06-04 | 中南大学 | Tunnel constructor thermal risk monitoring and early warning system and method |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115187033A (en) * | 2022-06-30 | 2022-10-14 | 江苏鑫昇腾科技发展股份有限公司 | Intelligent construction site construction safety risk early warning system and method thereof |
| CN115375154A (en) * | 2022-08-29 | 2022-11-22 | 黑龙江农业工程职业学院 | Building construction quality safety online risk management system |
-
2023
- 2023-12-12 CN CN202311707645.4A patent/CN117522149A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115187033A (en) * | 2022-06-30 | 2022-10-14 | 江苏鑫昇腾科技发展股份有限公司 | Intelligent construction site construction safety risk early warning system and method thereof |
| CN115375154A (en) * | 2022-08-29 | 2022-11-22 | 黑龙江农业工程职业学院 | Building construction quality safety online risk management system |
Non-Patent Citations (1)
| Title |
|---|
| 陈礼彪: ""山岭随道施工期安全风险评价方法研究"", 《博士电子期刊》, no. 7, 15 July 2018 (2018-07-15), pages 120 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117877735A (en) * | 2024-03-12 | 2024-04-12 | 中南大学 | Tunnel constructor thermal risk monitoring and early warning system and method |
| CN117877735B (en) * | 2024-03-12 | 2024-06-04 | 中南大学 | Tunnel constructor thermal risk monitoring and early warning system and method |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108416985B (en) | Geological disaster monitoring and early warning system and method based on image recognition | |
| CN107808133B (en) | Unmanned aerial vehicle line patrol-based oil and gas pipeline safety monitoring method and system and software memory | |
| CN112884747B (en) | Automatic bridge crack detection system integrating cyclic residual convolution and context extractor network | |
| CN112149512A (en) | Helmet wearing identification method based on two-stage deep learning | |
| CN111144656A (en) | Disaster evaluation analysis method based on GIS | |
| CN114581764B (en) | Underground structure crack disease discriminating method based on deep learning algorithm | |
| CN115330664A (en) | Image recognition-based surrounding rock weathering degree full-automatic recognition method and device | |
| Zhang et al. | Deep learning‐based automatic detection of muck types for earth pressure balance shield tunneling in soft ground | |
| CN115272826A (en) | Image identification method, device and system based on convolutional neural network | |
| CN114973320A (en) | Underground coal mine personnel detection method based on depth information | |
| CN117726991B (en) | High-altitude hanging basket safety belt detection method and terminal | |
| JP2007035017A (en) | Disaster prevention operation plan support system | |
| CN110889834A (en) | Road tunnel surrounding rock grading method based on cloud computing | |
| CN116647651B (en) | Unmanned aerial vehicle construction monitoring method and system based on Beidou satellite | |
| CN116343071B (en) | Surface mining crack unmanned aerial vehicle image recognition method based on deep learning algorithm | |
| CN117522149A (en) | Tunnel security risk identification method and device and security management platform | |
| CN116363502B (en) | Mining land multidimensional information acquisition method and device integrating multisource geographic big data | |
| CN114972906A (en) | Soil quality type identification method for excavation surface of soil pressure balance shield | |
| CN111507423B (en) | Engineering quantity measuring method for cleaning transmission line channel | |
| CN112883889A (en) | Method suitable for detecting violation behaviors in power construction scene | |
| CN110827264A (en) | Evaluation system for apparent defects of concrete member | |
| CN113807724B (en) | Site selection method for slag disposal site based on comprehensive risk evaluation | |
| Siljander et al. | L and use/land cover classification for the iron mining site of Kishushe, Kenya: A feasibility study of traditional and machine learning algorithms | |
| CN115082865B (en) | Bridge crane intrusion dangerous behavior early warning method and system based on visual image recognition | |
| CN116824513B (en) | Drilling process automatic identification supervision method and system based on deep learning |
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 |