CN114707856B - Risk identification analysis and early warning system based on computer vision - Google Patents
Risk identification analysis and early warning system based on computer vision Download PDFInfo
- Publication number
- CN114707856B CN114707856B CN202210351010.4A CN202210351010A CN114707856B CN 114707856 B CN114707856 B CN 114707856B CN 202210351010 A CN202210351010 A CN 202210351010A CN 114707856 B CN114707856 B CN 114707856B
- Authority
- CN
- China
- Prior art keywords
- risk
- personnel
- degree
- comprehensive
- equipment
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/18—Complex mathematical operations for evaluating statistical data, e.g. average values, frequency distributions, probability functions, regression analysis
-
- 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
-
- 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
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Data Mining & Analysis (AREA)
- Strategic Management (AREA)
- Mathematical Physics (AREA)
- Entrepreneurship & Innovation (AREA)
- Operations Research (AREA)
- Software Systems (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Quality & Reliability (AREA)
- Marketing (AREA)
- Pure & Applied Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Computational Mathematics (AREA)
- Development Economics (AREA)
- Game Theory and Decision Science (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Educational Administration (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Artificial Intelligence (AREA)
- Probability & Statistics with Applications (AREA)
- Health & Medical Sciences (AREA)
- Algebra (AREA)
Abstract
The invention provides a risk identification analysis and early warning system based on computer vision, which comprises: the risk identification module is used for acquiring a comprehensive risk distribution map of the production site; the risk analysis module is used for acquiring a neighborhood risk coefficient of the path; the dangerous degree acquisition module is used for acquiring the dangerous degree from one person to the other person in the binary group; and the risk early warning module is used for comprehensively predicting the risk degree of the position of the personnel to be analyzed by combining the risk degree of other personnel to the position of the personnel to be analyzed, and carrying out risk early warning according to the comprehensive predicted risk degree. The invention ensures that the risk identification and the early warning result are more reasonable and accurate, and reduces the consequences of dangerous accidents.
Description
Technical Field
The application relates to the field of big data safety production, in particular to a risk identification analysis and early warning system based on computer vision.
Background
Enterprise security is critical to the production and development of enterprises. The production, storage, and materials required, etc. of an enterprise are potential sources of risk in an enterprise's production environment. When a potential risk source is in an accident, the potential risk source not only brings loss to enterprises, but also influences the life health of production staff of the enterprises. Therefore, the monitoring and identification of potential risk sources of enterprises and the corresponding equipment risk and personnel risk early warning are very important links for enterprise safety supervision.
Disclosure of Invention
In order to solve the above problems, an object of the present invention is to provide a risk identification analysis and early warning system based on computer vision, the system comprising:
the risk identification module is used for acquiring a comprehensive risk distribution map of the production site;
the risk analysis module is used for forming any two persons in the monitoring area into a binary group and acquiring a path between the two persons; for any coordinate point on a path, generating a window by taking the coordinate point as a center, capturing an image from a comprehensive risk distribution map by using the obtained window to obtain a sub-image, calculating a gray level co-occurrence matrix of the sub-image, and obtaining a neighborhood risk coefficient of the coordinate point according to each element value and the position of each element value in the gray level co-occurrence matrix; summing the neighborhood risk coefficients of all coordinate points on the path to obtain the neighborhood risk coefficient of the path;
the dangerous degree acquisition module is used for acquiring a first set corresponding to the personnel in the binary group, wherein a dangerous area of the equipment in the first set comprises the predicted position of the personnel, and the dangerous degree from one personnel to the other personnel in the binary group is obtained according to the unit vector of the connection line of the positions of the two personnel in the binary group, the accident propagation vector of the equipment in the first set corresponding to the personnel and the neighborhood risk coefficient of the corresponding path of the binary group;
and the risk early warning module is used for comprehensively predicting the risk degree of the position of the personnel to be analyzed by combining the risk degree of other personnel to the position of the personnel to be analyzed, and carrying out risk early warning according to the comprehensive predicted risk degree.
Preferably, the risk analysis module further comprises: the operation activity analysis unit is used for acquiring personnel position information according to operation activity data of a production site, processing a time-dependent change sequence of personnel positions by utilizing a neural network to obtain a position sequence of personnel in future time, superposing hot spots generated at each position of the position sequence to obtain personnel prediction heat distribution, and obtaining personnel prediction positions according to the personnel prediction heat distribution.
Preferably, the comprehensive risk profile is specifically: and acquiring an area taking the equipment of the production site as a dangerous area of the equipment, assigning a value to the dangerous area of the equipment according to the risk degree of the equipment to obtain the risk distribution of the equipment, and superposing the risk distribution of all the equipment to obtain a comprehensive risk distribution map.
Preferably, the step of integrating the risk levels of other people to the positions of the people to be analyzed to obtain the integrated predicted risk level of the positions of the people to be analyzed includes:
acquiring a dangerous degree directed graph according to the position relation of all the personnel; on the dangerous degree directed graph, for any node, obtaining an arrival risk coefficient corresponding to the neighbor node according to the ratio of the edge weight of the neighbor node to the weighted outages of the neighbor node, and calculating the sum of the arrival risk coefficients corresponding to all the neighbor nodes to obtain the comprehensive arrival risk coefficient of the node; and obtaining comprehensive prediction risk degree of the position of the personnel to be analyzed according to the comprehensive arrival risk coefficient of the node and the size of the node, and carrying out risk early warning according to the comprehensive prediction risk degree.
Preferably, the acquiring the hazard level directed graph specifically includes: setting personnel as nodes, wherein the size of the nodes is the personal risk degree of the personnel, obtaining edge weights among the nodes according to the risk degree of the binary groups, and obtaining a risk degree directed graph according to the position relation and the edge weights of all the personnel.
Preferably, the risk analysis module further comprises:
the accident propagation vector obtaining unit is used for obtaining a first set from a device set containing a personnel prediction position in a dangerous area corresponding to the device, obtaining a second set from a neighbor node device set of the device in the first set according to a first-order association risk graph of the device, and obtaining an accident propagation vector of the personnel corresponding to the device in the first set according to a minimum weighted path between the device node of the first set and the device node of the second set and the risk degree of the first set.
Preferably, the first-order association risk graph of the device is specifically:
setting a device as a node for any two devices: if the dangerous areas of the two devices have intersection, the two device nodes are connected by edges, the edge weight is the intersection ratio of the dangerous areas of the two devices, and the connection relation between all the devices and the edge weight thereof are obtained to obtain a first-order association risk graph of the devices.
Preferably, the risk identification module is further configured to analyze the operation parameters of the device by using a neural network, so as to obtain the risk degree of the device.
Preferably, the risk early warning according to the comprehensive prediction risk degree comprises: and if the comprehensive prediction risk degree is greater than a preset threshold value, carrying out early warning reminding.
The beneficial effects of the invention are as follows:
according to the invention, the risk identification is carried out by combining computer vision and artificial intelligence technology with equipment risk and personnel operation activity analysis in the scene, so that the accuracy of the risk identification result is improved.
According to the risk identification method, the personal risk degree of the personnel is obtained according to the risk identification result, and the early warning result is obtained by analyzing the risk degree of the binary group, so that the early warning result is more accurate and reasonable, and the loss and influence of the accident can be reduced when the safety accident occurs.
Drawings
Fig. 1 is a system block diagram.
Detailed Description
The invention will be described in further detail with reference to the drawings and the specific examples.
Embodiment one:
the embodiment provides a risk identification analysis and early warning system based on computer vision, and a system frame is shown in fig. 1.
And the risk identification module is used for acquiring a comprehensive risk distribution map of the production site.
First, monitoring and identification of potential risk sources in an enterprise production environment is required. The potential risk sources in the present invention mainly refer to production equipment, storage equipment, etc. And then, based on a risk identification result, acquiring an area taking the equipment as a center as a dangerous area of the equipment, assigning a value to the dangerous area of the equipment according to the risk degree of the equipment to obtain the risk distribution of the equipment, and superposing the risk distribution of all the equipment to obtain a comprehensive risk distribution map.
The risk identification module is also used for analyzing the equipment operation parameters by utilizing the neural network to obtain the risk degree of the equipment.
And installing cameras in a production area of a chemical plant, obliquely viewing the camera with a view angle facing downwards, enabling adjacent cameras to have partial coincident views, acquiring images acquired by all cameras, and splicing and fusing the images to obtain a whole ground overlooking spliced image in the area.
Since there are numerous production devices in a high-risk production area, these devices are prone to generate safety accidents once they fail or the parameters of the devices are abnormal, the risk levels of the safety accidents caused by different devices are inconsistent, and each device is assigned a risk level, where the risk level is used to represent the risk level of the device, and this is determined by the type and function of the device, and the risk levels are classified into ten levels in this embodiment: a higher ranking of 0.1,0.2, … …,1.0 indicates a greater risk level. The identification of the risk degree of the equipment can be realized based on a neural network or can be distributed based on experience. Specifically, if the risk identification is realized through a neural network, the operation parameters of the equipment need to be acquired, and the neural network analyzes the operation parameters of the equipment to obtain a risk identification result of the equipment.
The method comprises the steps of acquiring an image acquired by a camera, inputting the image into YOLOv4 to acquire a bounding box of each device in the image, further acquiring the position of the device in the image, and transforming the position of the device in the image into a ground overlook mosaic by using the affine change matrix to acquire the position coordinates of the device in the ground overlook mosaic. Each device is divided into a circular danger zone by taking the device as a center, and the danger zone represents the range of influence when the device is in danger.
On a ground overlook splice graph, generating a Gaussian hot spot with the same size as a corresponding dangerous area for one equipment position, wherein the Gaussian hot spot generating method comprises the following steps: a gaussian convolution kernel, which is a gaussian hot spot, is generated centered on the location of the device and the diameter of the hazardous area. The different positions on the Gaussian hot spot correspond to a heat value, and the heat value of all positions except the Gaussian hot spot is 0.
The heat value of each location is multiplied by the risk level of the device, the multiplication result representing the magnitude of the risk of each location. Thus far, for one device, assuming that the device a is the device a, the risk of each location is obtained, and the risk of all locations is referred to as a risk distribution of the device a, which is used to characterize the risk existing around the device. And (3) superposing the risk distribution of all the equipment to obtain a comprehensive risk distribution diagram, wherein the result is represented on a ground overlook splice diagram, and the process is the process of acquiring data of dangerous places.
And the risk analysis module comprises a working activity analysis unit and is used for processing the time-dependent change sequence of the position of the personnel by utilizing the neural network to obtain a position sequence of the future time of the personnel, superposing the hot spots generated at each position of the position sequence to obtain personnel prediction heat distribution, and obtaining the personnel prediction position according to the personnel prediction heat distribution.
For one person, the previous K of the current moment is obtained 1 Within a time unit (in this embodiment, K is a time unit every 2 seconds) 1 =20, also determinable by the implementer) sequence of position changes over time. Inputting the sequence into TCN network to output the future K of the person 2 Within a time unit (K in this embodiment) 1 =5) (the TCN network described implements the conventional timing prediction task). Generating a hot spot for each position in the position sequence, and carrying out forgetting coefficient superposition on the hot spots, wherein the superposition result is the personnel predicted heat distribution; and each position on the superposition result corresponds to a heat value, and the centroid position (or the expected position and the average position) of the superposition result is obtained by processing the superposition result softargmax, wherein the centroid position is the predicted position of the person. The predicted heat distribution of the personnel is used for reflecting the motion state or motion speed of the personnel for a period of time in the future, and the predicted position of the personnel represents the expected position of the personnel for the period of time in the future.
An accident propagation vector obtaining unit, configured to set a device as a node, for any two devices: if the dangerous areas of the two devices have intersection, the nodes of the two devices are connected by edges, the edge weight is the intersection ratio of the dangerous areas of the two devices, and the connection relation between all the devices and the edge weight thereof are obtained to obtain a first-order association risk graph of the devices; acquiring a first set of equipment sets of equipment corresponding to dangerous areas and containing personnel prediction positions, acquiring a second set of neighbor node equipment sets of equipment in the first set according to a first-order associated risk graph of the equipment, and acquiring an accident propagation vector of the equipment corresponding to the first set according to a minimum weighted path between the equipment nodes of the first set and the equipment nodes of the second set and the risk degree of the equipment of the first set.
The invention acquires a predicted position of a person c, and analyzes the propagation characteristics of a risk event if the predicted position is at risk. The present invention uses the event propagation vector to describe this feature by:
when the dangerous areas of the two devices are intersected, edges are connected between the nodes represented by the two devices, the edge weight is equal to the intersection ratio of the two dangerous areas, and no edge exists when the two areas are not intersected. Acquiring connection relations among all devices and edge weights of all devices to form a first-order association risk graph G of the devices 1 The larger the edge weight value is, the larger the influence of the equipment is, and the domino effect is easy to generate when two pieces of equipment have safety accidents.
The predicted position of the person c may be a point in a plurality of dangerous areas, and the devices corresponding to the dangerous areas are assembled into a first set S 1 I.e. first set S 1 The dangerous area of each device contains the predicted location of the person c.
Acquiring a first set S 1 In G 1 Obtaining a K-order neighborhood node set of a to obtain a second set S 2 In this embodiment, k=4, and the practitioner may obtain the devices a to S according to the actual implementation scene adjustment 2 The shortest path of any device b, represented by a vector, called the incident propagation vector for device a; the vector size is the product of all edge weights on the path multiplied by theta, wherein theta is the risk degree of the equipment a, and the product of the edge weights represents the probability of the equipment a to spread an accident; the direction of the vector is the direction of the displacement vector from the a device position to the b device position, the tableAnd a, showing the propagation direction of the accident when the safety accident occurs.
Traversal S 1 All values of a in (a) are traversed simultaneously S 2 When all b of the plurality are valued, a first set S corresponding to the person c is obtained 1 The accident propagation vector of the device, this set of vectors is denoted V. The accident propagation vector set V characterizes the accident propagation situation after the occurrence of a dangerous accident.
The risk analysis module is used for forming any two persons in the monitoring area into a binary group and acquiring a path between the two persons; for any coordinate point on a path, generating a window by taking the coordinate point as a center, capturing an image from a comprehensive risk distribution map by using the obtained window to obtain a sub-image, calculating a gray level co-occurrence matrix of the sub-image, and obtaining a neighborhood risk coefficient of the coordinate point according to each element value and the position of each element value in the gray level co-occurrence matrix; and summing the neighborhood risk coefficients of all coordinate points on the path to obtain the neighborhood risk coefficient of the path.
Any two persons in the monitoring area form a binary group, for one binary group, one person is c, the other person is d, a linear path between the person c and the other person d is obtained, and a position coordinate set passed by the linear path is set as S 3 Acquiring S 3 One coordinate point of (2) is p 1 . On the comprehensive risk distribution map of the equipment, taking the comprehensive risk distribution map as a coordinate point p 1 A window is generated at the center, preferably the window has a size of 11×11 in the present embodiment, the sub-image is obtained by capturing the image with the obtained window, the gray level co-occurrence matrix M of the sub-image is calculated, and the coordinate point p is calculated 1 The neighborhood risk coefficient of (2) is the sum of index values of all elements in the gray level co-occurrence matrix, wherein the index value is the product of the element row and column number average value and the element value, and the index value is expressed as follows by a formula:
wherein (x, y) is a coordinate point on the gray level co-occurrence matrix M, f (x, y) represents a value at a (x, y) position on the gray level co-occurrence matrix M, and f (x, y) represents a gray level and co-occurrenceThe larger the probability of the position with larger pixel value on the sub-image is, the larger the probability of the position with larger pair occurrence is, the larger the dangerous density of the position is, S is calculated 3 Obtaining the neighborhood risk coefficient of the path by summing the neighborhood risk coefficients of all coordinate points in the path
The dangerous degree acquisition module is used for acquiring a first set corresponding to the personnel in the binary group, a dangerous area of the equipment in the first set comprises the predicted position of the personnel, and the dangerous degree from one personnel to the other personnel in the binary group is obtained according to the unit vector of the connection line of the positions of the two personnel in the binary group, the accident propagation vector of the equipment in the first set corresponding to the personnel and the neighborhood risk coefficient of the corresponding path of the binary group. Specifically, the risk level of one person in the binary group to another person's location is: and calculating the inner product of the unit vector of the two-person position connecting line and the accident propagation vector of the equipment in the first set corresponding to the person to be saved, and multiplying the obtained inner product by the neighborhood risk coefficient of the corresponding path of the binary group to obtain the risk degree.
Expressed in terms of a formula, the risk level of person d to person c in the binary group is expressed as:
N dc ＝∑ v∈V relu(v cd v T )×F
wherein V represents that person c corresponds to the first set S 1 The accident propagation vector set of the device V represents a vector element (V is a row vector) in the set V cd A unit vector (which is a row vector) v representing the position of person c pointing to the position of person d cd v T Representing the inner product of two vectors (essentially representing vector v at vector v) cd Projection length on) of the projection light beam,
∑ v∈V relu(v cd v T ) The larger the person d is, the more likely the person d is to be affected by a chain accident or accident transmission when the person d arrives at the person c, and the more likely the danger is; the larger F represents person d to person cThe potential risk on the path traversed by the location is high, i.e. N dc The greater the difficulty of indicating the d to c position; the risk level of all the tuples is obtained.
And the risk early warning module is used for comprehensively predicting the risk degree of the position of the personnel to be analyzed by combining the risk degree of other personnel to the position of the personnel to be analyzed, and carrying out risk early warning according to the comprehensive predicted risk degree.
The personal risk degree of the personnel is specifically the product of the heat value of the personnel position on the personnel prediction heat distribution and the risk value of the corresponding position in the comprehensive risk distribution diagram.
Setting personnel as nodes, wherein the size of the nodes is the personal risk degree of the personnel, obtaining edge weights among the nodes according to the risk degree of the binary groups, and obtaining a risk degree directed graph according to the position relation and the edge weights of all the personnel.
On the dangerous degree directed graph, for any node, obtaining an arrival risk coefficient corresponding to the neighbor node according to the ratio of the edge weight of the neighbor node to the weighted outages of the neighbor node, and calculating the sum of the arrival risk coefficients corresponding to all the neighbor nodes to obtain the comprehensive arrival risk coefficient of the node; and obtaining comprehensive prediction risk degree of the position of the personnel to be analyzed according to the comprehensive arrival risk coefficient of the node and the size of the node, and carrying out risk early warning according to the comprehensive prediction risk degree.
And if the comprehensive prediction risk degree is greater than a preset threshold value, carrying out early warning reminding.
In particular, in the process of safety production of personnel, it is most important to ensure the safety of personnel, so that the personal risk degree of the personnel needs to be calculated, and the specific method is as follows: acquiring a predicted thermal profile S of a person 4 Setting a personnel position on the predicted heat distribution as q, setting the corresponding heat degree as f (q), obtaining a risk value of the position as g (q) according to the comprehensive risk distribution map, and setting the personnel risk degree as
Setting personnel as nodes, wherein the size of the nodes is the person of the personnelThe side weight of the person d to the other person c in the binary group is N dc The edge weight of the person c to the other person d is N cd Constructing a hazard degree directed graph G 2 ；G 2 The risk level of all personnel and other different personnel locations is characterized.
For G 2 The second order aggregation is carried out on each personnel node, and the approximate method of aggregation is as follows: and obtaining a personnel node c, obtaining an arrival risk coefficient corresponding to the neighbor node according to the ratio of the edge weight value from the neighbor node to the node c to the weighted outages of the neighbor node, and calculating the sum of the arrival risk coefficients corresponding to all the neighbor nodes to obtain the comprehensive arrival risk coefficient of the node. And obtaining the comprehensive prediction risk degree according to the comprehensive arrival risk coefficient of the node and the size of the node.
The specific method for polymerization comprises the following steps: let the neighbor node set of personnel node c be S 5 The result of the polymerization is
Node d is S 5 One element of (a), N d Indicating the sum of the risk levels of person d to all other person locations. />
Smaller (less)>
The larger the representation d is more prone to the location of person c; g C Representing the overall arrival risk factor for person c, the greater this value is indicative of the greater the probability that person c will be at the location where person c is located when there is a risk for that person. Let Y c ＝D c -δ×G c ，D c Representing the size of the person node c, i.e. the degree of personal risk of the person c, δ is a hyper-parameter, preferably δ=0.3 in this embodiment. Y is Y c The comprehensive prediction of the risk degree of the personnel c is used for indicating the risk of the personnel c in a future period of time. And comparing the comprehensive predicted risk degree of the personnel with a preset threshold value, and carrying out alarm reminding when the comprehensive predicted risk degree of the personnel is larger than the threshold value.
The above embodiments are merely preferred embodiments of the present invention and are not intended to limit the present invention, and any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of the present invention.
Claims (5)
1. A risk identification analysis and early warning system based on computer vision, the system comprising:
the risk identification module is used for acquiring a comprehensive risk distribution map of the production site;
the risk analysis module is used for forming any two persons in the monitoring area into a binary group and acquiring a path between the two persons; for any coordinate point on a path, generating a window by taking the coordinate point as a center, capturing an image from a comprehensive risk distribution map by using the obtained window to obtain a sub-image, calculating a gray level co-occurrence matrix of the sub-image, and obtaining a neighborhood risk coefficient of the coordinate point according to each element value and the position of each element value in the gray level co-occurrence matrix; summing the neighborhood risk coefficients of all coordinate points on the path to obtain the neighborhood risk coefficient of the path;
the dangerous degree acquisition module is used for acquiring a first set corresponding to the personnel in the binary group, wherein a dangerous area of the equipment in the first set comprises the predicted position of the personnel, and the dangerous degree from one personnel to the other personnel in the binary group is obtained according to the unit vector of the connection line of the positions of the two personnel in the binary group, the accident propagation vector of the equipment in the first set corresponding to the personnel and the neighborhood risk coefficient of the corresponding path of the binary group;
the risk early warning module is used for comprehensively predicting the risk degree of the position of the personnel to be analyzed according to the risk degree of other personnel to the position of the personnel to be analyzed, and carrying out risk early warning according to the comprehensive predicted risk degree;
the risk analysis module further includes:
the accident propagation vector acquisition unit is used for acquiring a device set of a dangerous area corresponding to the device and containing a predicted position of a person to obtain a first set, acquiring a neighbor node device set of the device in the first set to obtain a second set according to a first-order association risk graph of the device, and acquiring an accident propagation vector of the person corresponding to the device in the first set according to a minimum weighted path between the device nodes of the first set and the device nodes of the second set and the risk degree of the first set;
the comprehensive prediction risk degree of the positions of the personnel to be analyzed obtained by integrating the risk degree of other personnel to the positions of the personnel to be analyzed comprises the following steps:
the personal risk degree of the personnel is specifically the product of the heat value of the personnel position on the personnel prediction heat distribution and the risk value of the corresponding position in the comprehensive risk distribution diagram;
setting personnel as nodes, wherein the size of the nodes is the personal risk degree of the personnel, obtaining edge weights among the nodes according to the risk degree of the binary groups, and obtaining a risk degree directed graph according to the position relation and the edge weights of all the personnel;
the first-order associated risk graph of the equipment specifically comprises the following steps:
setting a device as a node for any two devices: if the dangerous areas of the two devices have intersection, the nodes of the two devices are connected by edges, the edge weight is the intersection ratio of the dangerous areas of the two devices, and the connection relation between all the devices and the edge weight thereof are obtained to obtain a first-order association risk graph of the devices;
on the dangerous degree directed graph, for any node, obtaining an arrival risk coefficient corresponding to the neighbor node according to the ratio of the edge weight of the neighbor node to the weighted outages of the neighbor node, and calculating the sum of the arrival risk coefficients corresponding to all the neighbor nodes to obtain the comprehensive arrival risk coefficient of the node; and obtaining comprehensive prediction risk degree of the position of the personnel to be analyzed according to the comprehensive arrival risk coefficient of the node and the size of the node, and carrying out risk early warning according to the comprehensive prediction risk degree.
2. The system of claim 1, wherein the risk analysis module further comprises:
the operation activity analysis unit is used for acquiring personnel position information according to operation activity data of a production site, processing a time-dependent change sequence of personnel positions by utilizing a neural network to obtain a position sequence of personnel in future time, superposing hot spots generated at each position of the position sequence to obtain personnel prediction heat distribution, and obtaining personnel prediction positions according to the personnel prediction heat distribution.
3. The system according to claim 1, wherein the integrated risk profile is in particular:
and acquiring an area taking the equipment of the production site as a dangerous area of the equipment, assigning a value to the dangerous area of the equipment according to the risk degree of the equipment to obtain the risk distribution of the equipment, and superposing the risk distribution of all the equipment to obtain a comprehensive risk distribution map.
4. The system of claim 1, wherein the risk identification module is further configured to analyze the device operating parameters using a neural network to obtain a risk level of the device.
5. The system of claim 1, wherein the risk pre-warning based on the comprehensively predicted risk level comprises: and if the comprehensive prediction risk degree is greater than a preset threshold value, carrying out early warning reminding.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210351010.4A CN114707856B (en) | 2022-04-02 | 2022-04-02 | Risk identification analysis and early warning system based on computer vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210351010.4A CN114707856B (en) | 2022-04-02 | 2022-04-02 | Risk identification analysis and early warning system based on computer vision |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114707856A CN114707856A (en) | 2022-07-05 |
| CN114707856B true CN114707856B (en) | 2023-06-30 |
Family
ID=82172785
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210351010.4A Active CN114707856B (en) | 2022-04-02 | 2022-04-02 | Risk identification analysis and early warning system based on computer vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114707856B (en) |
Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108378450A (en) * | 2018-03-08 | 2018-08-10 | 公安部天津消防研究所 | A kind of perception of blast accident and risk profile early warning Intelligent fire-fighting helmet implementation method |
Family Cites Families (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9680855B2 (en) * | 2014-06-30 | 2017-06-13 | Neo Prime, LLC | Probabilistic model for cyber risk forecasting |
| CN112435469B (en) * | 2020-11-10 | 2024-04-02 | 腾讯科技（深圳）有限公司 | Vehicle early warning control method and device, computer readable medium and electronic equipment |
| CN112966590B (en) * | 2021-03-03 | 2022-04-19 | 河南鑫安利安全科技股份有限公司 | Real-time early warning method and device based on AIOT and enterprise safety index |
| CN113112151B (en) * | 2021-04-12 | 2023-12-08 | 河南鑫安利安全科技股份有限公司 | Intelligent wind control evaluation method and system based on multidimensional sensing and enterprise data quantification |
| CN113642631B (en) * | 2021-08-10 | 2022-07-12 | 沭阳协润电子有限公司 | Dangerous area electronic fence generation method and system based on artificial intelligence |
| CN113408968B (en) * | 2021-08-19 | 2021-11-16 | 南通市立新机械制造有限公司 | Safety production area personnel management method and system based on big data analysis |
| CN113762473B (en) * | 2021-08-25 | 2024-04-12 | 江苏大学 | Complex scene driving risk prediction method based on multi-time space diagram |
| CN114067271A (en) * | 2021-11-20 | 2022-02-18 | 河南鑫安利安全科技股份有限公司 | Safety risk early warning method based on AIOT and personnel trajectory analysis |
-
2022
- 2022-04-02 CN CN202210351010.4A patent/CN114707856B/en active Active
Patent Citations (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108378450A (en) * | 2018-03-08 | 2018-08-10 | 公安部天津消防研究所 | A kind of perception of blast accident and risk profile early warning Intelligent fire-fighting helmet implementation method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114707856A (en) | 2022-07-05 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Asadzadeh et al. | Sensor-based safety management | |
| Kim et al. | Vision-based object-centric safety assessment using fuzzy inference: Monitoring struck-by accidents with moving objects | |
| Khan et al. | Utilizing safety rule correlation for mobile scaffolds monitoring leveraging deep convolution neural networks | |
| Xu et al. | Dynamic safety prewarning mechanism of human–machine–environment using computer vision | |
| CN110602449A (en) | Intelligent construction safety monitoring system method in large scene based on vision | |
| Zhafran et al. | Computer vision system based for personal protective equipment detection, by using convolutional neural network | |
| Karim et al. | Modeling and simulation of a robotic bridge inspection system | |
| Goh et al. | Management of safe distancing on construction sites during COVID-19: A smart real-time monitoring system | |
| CN115797864B (en) | Security management system applied to intelligent community | |
| Huang et al. | Skeleton-based automatic assessment and prediction of intrusion risk in construction hazardous areas | |
| Li et al. | Fast safety distance warning framework for proximity detection based on oriented object detection and pinhole model | |
| Xu et al. | Construction worker safety prediction and active warning based on computer vision and the gray absolute decision analysis method | |
| CN114707856B (en) | Risk identification analysis and early warning system based on computer vision | |
| Kulinan et al. | Advancing construction site workforce safety monitoring through BIM and computer vision integration | |
| Rivera et al. | Work safety assessment through contextual analysis with computer vision | |
| Chang et al. | Safety risk assessment of electric power operation site based on variable precision rough set | |
| KR20240040530A (en) | System for infectious disease prevention based on deep learning | |
| KR20230097854A (en) | Method and system for recognizing dangerous behavior of workers in power plant | |
| Daud et al. | Conceptual design of human detection via deep learning for industrial safety enforcement in manufacturing site | |
| Nugraha et al. | Convolutional Neural Network for Identification of Personal Protective Equipment Usage Compliance in Manufacturing Laboratory | |
| Almadhoun et al. | Artificial Intelligence Aims to Save Lives in Offshore Marine Vessels | |
| Chen et al. | A novel system for automated proper use identification of personal protective equipment in decommissioning site of fukushima daiichi nuclear power station | |
| Park et al. | Transforming Construction Site Safety with iSAFE: An Automated Safety Management Platform | |
| Tran et al. | Development of Computer Vision and BIM-cloud based Automated Status Updating for Construction Safety Monitoring | |
| Fraser et al. | Reducing Viral Transmission through AI-based Crowd Monitoring and Social Distancing Analysis |
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 |