Mining area production safety monitoring system based on target detection
By adopting target detection technology in the mining area production safety monitoring system, collecting data, building image detection models and risk assessment systems, the problem that the existing system cannot accurately monitor miners' safety and risk identification is solved, and the mining area operation process is realized Accurate monitoring and risk assessment ensure safety production and alarm notifications.
Patent Information
- Application Number
- CN202411105100.0
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-08-13
- Publication Date
- 2024-11-26
- Estimated Expiration
- Not applicable · inactive patent
AI Technical Summary
The existing mining area production safety monitoring system cannot accurately grasp the personal safety of miners, and cannot deal with emergencies during the mining area work, and the risk identification is incomplete.
The mining area production safety monitoring system based on target detection is adopted, and data is collected through the mining area inspection image acquisition unit, the monitoring trajectory of the camera gimbal is formulated, and the monitoring points are updated through reinforcement learning; the image recognition and analysis unit constructs the target detection model and performs image detection; The risk level judgment unit constructs a risk evaluation system through common word analysis and correlation analysis, and the monitoring and alarm unit issues alarm signals of the corresponding level.
Accurate monitoring and risk assessment of mining operation processes is realized, ensuring that the camera gimbal can be automatically adjusted, obtain the accuracy of video images, provide data support for spatial position safety detection, and issue warning notifications and alarm video recordings based on risk levels.
Smart Images

Figure CN119027872A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of safety management, and in particular to a mining area production safety monitoring system based on target detection. Background Art
[0002] The mining area includes several mines or open-pit mines, with complete production processes, ground transportation, power supply, communication dispatching, production management and life service facilities. The scope often depends on the scale of the ore deposit. In the Chinese patent application number 201910178126.0, "a mining area safety monitoring system is disclosed, including a personnel management module, a mining area monitoring module, a case analysis module, and a recruitment module; characterized in that: the personnel management module includes a personnel information unit, an age analysis unit, a physical strength calculation unit, a data display unit, and a fixed-point punching unit; the personnel information unit is used to record specific personnel information; the age analysis module is used to filter and record the age information of personnel; the physical strength calculation unit is used to calculate the physical strength of people of different age groups; the data display unit is used to announce the physical strength of people of different age groups; the fixed-point punching unit is used to control people to punch in on time; the present invention is mainly used to solve the problem that the existing mining area has insufficient management of miners and cannot accurately grasp the personal safety of miners"; This existing technology only solves the problem that the existing mining areas have insufficient management of miners and cannot accurately grasp the personal safety of miners. It does not take into account that in the actual mining work process, the camera pan-tilt cruise route preset based on the mining work arrangement information cannot cope with the emergencies in the actual mining work process. At the same time, the risk situation of the mining area obtained only by risk identification of the mining work area is not comprehensive. It is necessary to further consider the detection of the spatial position relationship of each detection target in the mining area based on production requirements, and to issue an alarm based on the detection results. Summary of the invention
[0003] The purpose of the present invention is to provide a mining area production safety monitoring system based on target detection to solve the problems raised in the above background technology.
[0004] To achieve the above-mentioned purpose, the present invention provides the following technical solutions: a mine production safety monitoring system based on target detection, comprising a mine patrol image acquisition unit, wherein the mine patrol image acquisition unit collects relevant data of the mine to be monitored, and formulates the operation trajectory of the staff in each mining operation area based on the relevant information of the mine operation, and then determines the monitoring trajectory of the camera pan platform, and sets the cruise monitoring point of the camera pan platform on the monitoring trajectory of the camera pan platform, and in the actual mine operation process, through reinforcement learning of the camera pan platform, the camera pan platform can update the position of the actual monitoring point according to the actual operation trajectory of the staff, and collect relevant video images of the operation process; A mining area image recognition and analysis unit, wherein the mining area image recognition and analysis unit constructs an image target detection model and a sample data set for training the image target detection model, performs model training and accuracy verification operations on the image target detection model through the image data set, and when the accuracy of the image target detection model does not meet the use requirements, performs model optimization on the image target detection model to ensure the accuracy of the image target detection model, and then detects the collected image through the image target detection model that meets the use requirements, and obtains the detection result; A mining area risk level judgment unit, wherein the mining area risk level judgment unit obtains relevant production risk factors in the mining area production operation process by performing co-word analysis on the text related to the open-pit mine production safety risk factors, and screens the obtained production risk factors by performing correlation analysis on the mining area production safety and the mining area production risk factors, and then constructs a risk evaluation system based on the screened production risk factors, thereby evaluating the risk level of the mining area production operation process; A mining area monitoring and alarm unit, which issues an alarm warning signal of a corresponding level based on the risk level of the mining area's production operation process, and simultaneously performs an alarm warning signal notification of a corresponding level, and also includes processing image detection results, performing alarm detection based on the processing results and in combination with mining area production safety requirements, and then performing alarm notification and alarm video recording operations based on the alarm detection results.
[0005] Preferably, the mine inspection image acquisition unit includes a mine data acquisition module and a cruise monitoring point setting module, and the mine data acquisition module collects relevant data of the mine to be monitored, including the mine's geographic information, environmental information, operator information, operators' daily work area information, operating equipment information and operation cycle scheduling information, specifically acquiring the mine's geographic information through sanitary remote sensing data, collecting environmental information through environmental monitoring equipment, acquiring the mine's operator information, operators' daily work area information, operating equipment information and operation cycle scheduling information by asking the mine management personnel, and dividing the mine into a mining work area and a mining equipment maintenance area, and dividing the mining work area into different and adjacent operating areas based on the operation cycle scheduling information, and sorting each adjacent operating area from top to bottom and from left to right. The order on the right is a naming operation of adjacent and non-overlapping numbers. The cruise monitoring point setting module is electrically connected to the mining area data acquisition module. The cruise monitoring point setting module is based on the operation cycle arrangement information corresponding to each mining operation area, and further based on the working area information of the mining staff in the mining operation area every day within the preset operation cycle, obtains the operation trajectory information of the mining staff in the mining operation area within the corresponding operation cycle, and then determines the cruise monitoring point of the camera pan-tilt head based on the operation trajectory information. The specific operation is to set the same camera pan-tilt head monitoring trajectory at a distance of 200m from the operation trajectory based on the operation trajectory of the mining staff in the mining operation area, and set multiple camera pan-tilt head cruise monitoring points at the same distance on the camera pan-tilt head monitoring trajectory, thereby realizing the monitoring of the mining operation process.
[0006] Preferably, the mining area inspection image acquisition unit also includes a monitoring video tracking module and a mining area image acquisition module. The monitoring video tracking module is electrically connected to the cruise monitoring point setting module. The monitoring video tracking module performs reinforcement learning on the inspection process of the camera pan-tilt head, so that in the actual mining area operation process, through the automatic adjustment of the camera pan-tilt head, the images of the mining area operation process subsequently collected are located at a suitable position in the picture. The specific operation is to use GYM to simulate and model the scene of the mining area operation area camera pan-tilt head monitoring operation process, adopt a modeling method with a target frame as the state space, and design the state space, action space and reward function, wherein the design of the state space is through In Pygame, the pixel point in the upper left corner of the screen is used as the origin and a coordinate system is established to determine the coordinate information of the four points of the target frame in the screen. By using the target frame position coordinate information as the environmental information of reinforcement learning, the impact of the environment on the reinforcement learning process can be reduced, and the generalization of reinforcement learning results can be improved by further randomly generating the size and initial position of the target frame. The design of the action space is based on the four action spaces of up, down, left and right that can be used by the camera pan, and the corresponding 0, 1, 2 and 3 step units are generated respectively. The design of the reward function is used to perform process rewards and victory rewards based on the exploration process of the camera pan in the simulation space. The function expression is as follows: ; In the formula, Represents the reward score of a single-step exploration process, represents the number of exploration steps, Indicates the distance between the center of the target frame and the center of the screen during the previous exploration. Indicates the distance between the center of the target frame and the center of the picture during the new exploration step. It represents the recognition range of the center of the screen, and sets the final value of the number of exploration steps to 200. When the number of exploration steps reaches the final value, the process rewards corresponding to each exploration step are accumulated to obtain a victory reward, including a negative cumulative reward value and a positive cumulative reward value slightly larger than the negative cumulative reward value. When the number of exploration steps has not reached the final value but the victory reward reaches the corresponding value, the exploration is also stopped. The mining area image acquisition module is electrically connected to the monitoring video tracking module. The mining area image acquisition module collects video images of the operation process in the mining operation area through the camera pan-tilt during the actual cruising process.
[0007] Preferably, the mining area image recognition and analysis unit includes an image detection model construction module and a sample data set construction module, wherein the image detection model construction module constructs an image target detection model based on the YOLOv5 network, and the image target detection model specifically includes four parts: an input end, a feature extraction network, a neck network, and an output end, wherein the input end is used to perform data enhancement operations on the input image through the Mosaic method, wherein the feature extraction network is used to improve the image target detection model's ability to extract target features through the Focus slicing module, wherein the neck network is used to obtain feature maps of different scales through the feature pyramid network and the pyramid attention network and perform feature enhancement processing, wherein the output end is used to convert the processed feature map into the required detection result, and the sample data set construction module obtains video images of the operation process in the mining operation area collected by the historical camera pan-tilt from the mining area management personnel, and extracts a frame of the video every 2 seconds to obtain a sample image, and annotates the obtained sample image through LabelImg, and automatically generates a corresponding txt file for each image during the image annotation process through LabelImg, and the txt file specifically contains the target category and the coordinate information of the target box in the sample image target box, and constructs a sample data set based on the annotated sample image.
[0008] Preferably, the mining area image recognition and analysis unit further includes a model training and verification module and a detection result acquisition module. The model training and verification module is electrically connected to the image detection model construction module and the sample data set construction module respectively. The model training and verification module is used to perform model training on the image target detection model through the image data set, and perform accuracy verification on the trained image target detection model through the model accuracy algorithm. The specific calculation formula is as follows: ; In the formula, Represents the accuracy value of the image target detection model, Indicates the number of sample data that are correctly detected. Indicates the number of sample data with detection errors. When the calculated result is greater than 0.95, it indicates that the accuracy of the image target detection model meets the requirements. When the calculated result is not greater than 0.95, it indicates that the accuracy of the image target detection model does not meet the requirements and the image target detection model needs to be optimized and adjusted. Specifically, the GIOU Loss loss function is selected as the optimization function of the image target detection model. The overlap between the detection box and the target box is detected by the GIOU Loss loss function, thereby optimizing the accuracy of the detection box. The detection result acquisition module is electrically connected to the model training and verification module. The detection result acquisition module detects the collected image through the image target detection model that meets the use requirements, obtains the detection result, and obtains the position information of the target box corresponding to each category of targets in the collected image.
[0009] Preferably, the mining area risk level judgment unit includes a production risk factor acquisition module and a production risk factor analysis module. The production risk factor acquisition module collects text related to open-pit mine production safety risk factors through an Octopus collector, and performs co-word analysis on the collected text related to open-pit mine production safety risk factors through Citespace software to obtain production risk factors in the mining area production operation process. The production risk factor analysis module and the production risk factor acquisition module are electrically connected. The production risk factor analysis module performs a correlation analysis between mining area production safety and mining area production risk factors on each production risk factor through SPSS data analysis software, and retains the top eight production risk factors based on the correlation analysis results between mining area production safety and mining area production risk factors, thereby realizing the screening of production risk factors and improving the accuracy of subsequent risk analysis.
[0010] Preferably, the mining area risk level judgment unit also includes a risk assessment system generation module and a risk level judgment module, the risk assessment system generation module is electrically connected to the production risk factor analysis module, the risk assessment system generation module constructs a risk assessment system based on the screened production risk factors, the risk level judgment module is electrically connected to the risk assessment system generation module, the risk level judgment module is based on the production risk factors in the risk assessment system, and the production risk factors are judged by the Delphi method in terms of importance, and the judgment results are further calculated by SuperDecisions3.0 software to obtain the weight value of the corresponding production risk factor, thereby collecting data on the corresponding production risk factors and analyzing the acquired data by Max-M in standardization method is used to standardize the data so that the distribution range of the collected data is between 0 and 1. In combination with the calculated weight values of the corresponding production risk factors, the standardized processing values of each production risk factor are multiplied by the weight values of the corresponding production risk factors, and each product result is accumulated to solve the risk value of production safety in the mining area. The risk level is judged based on the solution result. Specifically, when 0<solution result ≤0.2, it is judged that the risk level is very low; when 0.2<solution result ≤0.4, it is judged that the risk level is low; when 0.4<solution result ≤0.6, it is judged that the risk level is good; when 0.6<solution result ≤0.8, it is judged that the risk level is high; when 0.8<solution result ≤1, it is judged that the risk level is very high.
[0011] Preferably, the mining area monitoring alarm unit includes an alarm warning notification module, which is provided with alarm warning levels one to five from low to high corresponding to the risk level of the mining area, and sends an alarm warning signal of a corresponding level to the corresponding mining work area based on the risk level of the mining area production operation process obtained by the risk level judgment module, and notifies the alarm warning signal of the corresponding level on the large display screen set in the corresponding mining work area, and also includes calculating the spatial position relationship between the corresponding targets in the spatial dependency relationship based on the position information of the target frame corresponding to each category of targets in the acquired image, and generating an alarm signal when it is detected that the spatial position relationship of the corresponding targets does not meet the safety production requirements of the mining area. signal, otherwise no alarm signal is generated. Taking the tires and vehicle limiting devices of the mining vehicles in the mining equipment maintenance area as an example, when the overlap between the target frame corresponding to the mining vehicle tires and the target frame corresponding to the vehicle limiting device during the vehicle maintenance process is not greater than 0, it is considered that the vehicle limiting device is not correctly placed. When the overlap between the target frame corresponding to the mining vehicle tires and the target frame corresponding to the vehicle limiting device during the vehicle maintenance process is greater than 0, a coordinate system is established with the center position of the target frame corresponding to the mining vehicle tires as the coordinate origin. When it is detected that the target frame corresponding to the vehicle limiting device is above the center of the target frame corresponding to the mining vehicle tires, it is also considered that the vehicle limiting device is not correctly placed. Otherwise, it is considered that the vehicle limiting device is correctly placed.
[0012] Preferably, the mining area monitoring alarm unit also includes an alarm detection notification module, which is electrically connected to the alarm warning notification module. The alarm detection notification module receives the alarm signal generated by the alarm warning notification module through a signal receiver, and sends an alarm notification through a speaker at the same time, and controls the camera pan to save the current camera pan position information and zoom factor information, and then controls the camera pan to focus and advance, continuously acquires and tracks the position of the alarm target, ensures that when there are multiple target alarms in the picture, each alarm target is tracked in sequence without repetition, and ensures that the scale of the alarm target in the picture is appropriate through the camera pan after reinforcement learning, ends the tracking after the camera pan alarm video acquisition is completed and returns to the position before tracking, and saves and uploads the alarm video.
[0013] Compared with the prior art, the beneficial effects of the present invention at least include: the present invention proposes a mine production safety monitoring system based on target detection, and a mine inspection image acquisition unit is set up to formulate a cruise route of the camera pan head based on the mining work arrangement information and to set cruise points on the cruise route, and through reinforcement learning, the camera pan head can update the cruise route based on the actual mine operation process, so that the collected video image meets the use requirements, and the mine image recognition and analysis unit is set up to obtain the position information of the target frame corresponding to each category of targets in the collected video image, so as to provide data support for subsequent spatial position safety detection, and the mine risk level judgment unit is set up to evaluate the risk level of the mine based on the constructed risk evaluation system, and the mine monitoring alarm unit is set up to generate an alarm and make a corresponding level of warning level notification and alarm notification based on the risk level assessment result and the detection result of the spatial position relationship of the target with spatial dependence relationship based on the actual production requirements, and the camera pan head is controlled to save the alarm video. BRIEF DESCRIPTION OF THE DRAWINGS
[0014] Figure 1 It is a schematic diagram of the overall process of an embodiment of the present invention; Figure 2 This is a schematic diagram of the process of a mining area inspection image acquisition unit according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the process of a mining area image recognition and analysis unit according to an embodiment of the present invention; Figure 4 This is a schematic diagram of the process of a mining area risk level determination unit according to an embodiment of the present invention; Figure 5 The figure is a flowchart of a mining area monitoring and alarm unit according to an embodiment of the present invention.
[0015] In the figure: 100, mining area inspection image acquisition unit; 101, mining area data acquisition module; 102, patrol monitoring point setting module; 103, monitoring video tracking module; 104, mining area image acquisition module; 200, mining area image recognition and analysis unit; 201, image detection model construction module; 202, sample data set construction module; 203, model training and verification module; 204, detection result acquisition module; 300, mining area risk level judgment unit; 301, production risk factor acquisition module; 302, production risk factor analysis module; 303, risk assessment system generation module; 304, risk level judgment module; 400, mining area monitoring alarm unit; 401, alarm warning notification module; 402, alarm detection notification module. DETAILED DESCRIPTION
[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.
[0017] See also Figure 1-5 The present invention provides a technical solution: a mining area production safety monitoring system based on target detection, including a mining area patrol image acquisition unit 100, the mining area patrol image acquisition unit 100 collects relevant data of the mining area to be monitored, and formulates the operation trajectory of the staff in each mining operation area based on the relevant information of the mining area operation, and then determines the monitoring trajectory of the camera pan platform, and sets the cruise monitoring point of the camera pan platform on the monitoring trajectory of the camera pan platform, and in the actual mining area operation process, through the reinforcement learning of the camera pan platform, the camera pan platform can update the position of the actual monitoring point according to the actual operation trajectory of the staff, and collect relevant video images of the operation process; The mining area image recognition and analysis unit 200 constructs an image target detection model and a sample data set for training the image target detection model. The image target detection model is trained and precision verification is performed through the image data set. When the precision of the image target detection model does not meet the use requirements, the image target detection model is optimized to ensure the accuracy of the image target detection model. The collected image is then detected through the image target detection model that meets the use requirements, and the detection result is obtained. The mining area risk level judgment unit 300 obtains relevant production risk factors in the mining area production operation process by performing co-word analysis on the text related to the open-pit mine production safety risk factors, and screens the obtained production risk factors by performing correlation analysis on the mining area production safety and the mining area production risk factors, and then constructs a risk evaluation system based on the screened production risk factors, thereby evaluating the risk level of the mining area production operation process; The mining area monitoring and alarm unit 400 sends out an alarm warning signal of a corresponding level based on the risk level of the mining area production operation process, and at the same time, notifies the alarm warning signal of a corresponding level, and also includes processing the image detection results, performing alarm detection based on the processing results and in combination with the mining area production safety requirements, and then performing alarm notification and alarm video recording operations based on the alarm detection results.
[0018] The mine inspection image acquisition unit 100 includes a mine data acquisition module 101 and a patrol monitoring point setting module 102. The mine data acquisition module 101 collects relevant data of the mine to be monitored, including the geographic information, environmental information, operator information, daily work area information of the operator, operating equipment information and operation cycle arrangement information of the mine. Specifically, the geographic information of the mine is acquired through sanitary remote sensing data, and the environmental information is collected through environmental monitoring equipment. The mine operator information, daily work area information of the operator, operating equipment information and operation cycle arrangement information are acquired by asking the mine management personnel, and the mine is divided into a mining work area and a mining equipment maintenance area. Based on the operation cycle arrangement information, the mining work area is divided into different and adjacent operation areas, and each adjacent operation area is arranged from top to bottom and from left to right. The naming operation of the content is adjacent and non-overlapping numbers, the cruise monitoring point setting module 102 is electrically connected to the mining area data acquisition module 101, the cruise monitoring point setting module 102 is based on the operation cycle arrangement information corresponding to each mining operation area, and further based on the work area information of the mining area staff in the mining operation area every day within the preset operation cycle, obtains the operation trajectory information of the mining area staff in the mining operation area within the corresponding operation cycle, and then determines the cruise monitoring point of the camera pan-tilt head based on the operation trajectory information. The specific operation is to set the same camera pan-tilt head monitoring trajectory at a distance of 200m from the operation trajectory based on the operation trajectory of the mining area staff in the mining operation area, and evenly set multiple camera pan-tilt head cruise monitoring points at the same distance on the camera pan-tilt head monitoring trajectory, so as to realize the monitoring of the mining operation process; The mining inspection image acquisition unit 100 also includes a monitoring video tracking module 103 and a mining image acquisition module 104. The monitoring video tracking module 103 is electrically connected to the cruise monitoring point setting module 102. The monitoring video tracking module 103 performs reinforcement learning on the inspection process of the camera pan-tilt, so that in the actual mining operation process, through the automatic adjustment of the camera pan-tilt, the image of the mining operation process subsequently collected is located at a suitable position in the picture. The specific operation is to use GYM to simulate and model the scene of the mining operation area camera pan-tilt monitoring operation process, adopt a modeling method with the target frame as the state space, and design the state space, action space and reward function, wherein the state space The design is to establish a coordinate system in Pygame with the pixel point in the upper left corner of the screen as the origin, and then determine the coordinate information of the four points of the target frame in the screen. By using the target frame position coordinate information as the environmental information of reinforcement learning, the impact of the environment on the reinforcement learning process can be reduced, and the generalization of reinforcement learning results can be improved by further randomly generating the size and initial position of the target frame. The design of the action space is based on the four action spaces of up, down, left and right that the camera gimbal can use to generate corresponding 0, 1, 2 and 3 step units respectively. The design of the reward function is used to perform process rewards and victory rewards based on the exploration process of the camera gimbal in the simulation space. Its function expression is as follows: ; In the formula, Represents the reward score of a single-step exploration process, represents the number of exploration steps, Indicates the distance between the center of the target frame and the center of the screen during the previous exploration. Indicates the distance between the center of the target frame and the center of the picture during the new exploration step. Indicates the recognition range of the center of the screen, and sets the final value of the exploration steps to 200. When the exploration steps reach the final value, the process rewards corresponding to each exploration step are accumulated to obtain a victory reward, including a negative cumulative reward value and a positive cumulative reward value slightly larger than the negative cumulative reward value. When the exploration steps do not reach the final value but the victory reward reaches the corresponding value, the exploration is also stopped. The mining area image acquisition module 104 is electrically connected to the monitoring video tracking module 103. The mining area image acquisition module 104 uses the camera pan-tilt to collect video images of the operation process in the mining operation area during the actual cruise process; The mining area image recognition and analysis unit 200 includes an image detection model construction module 201 and a sample data set construction module 202. The image detection model construction module 201 constructs an image target detection model based on the YOLOv5 network, and the image target detection model specifically includes four parts: an input end, a feature extraction network, a neck network, and an output end. The input end is used to perform data enhancement operations on the input image through the Mosaic method, the feature extraction network is used to improve the image target detection model's ability to extract target features through the Focus slicing module, and the neck network is used to obtain feature maps of different scales through the feature pyramid network and the pyramid attention network. And perform feature enhancement processing, wherein the output end is used to convert the processed feature map into the required detection result, the sample data set construction module 202 obtains the video image of the operation process in the mining operation area collected by the historical camera pan-tilt from the mine area manager, and extracts a frame of the video every 2 seconds to obtain the sample image, and annotates the obtained sample image through LabelImg, and automatically generates a corresponding txt file for each image during the image annotation process through LabelImg, and the txt file specifically contains the target category and the coordinate information of the target frame in the sample image target frame, and constructs the sample data set based on the annotated sample image; The mining area image recognition and analysis unit 200 also includes a model training and verification module 203 and a detection result acquisition module 204. The model training and verification module 203 is electrically connected to the image detection model construction module 201 and the sample data set construction module 202 respectively. The model training and verification module 203 is used to perform model training on the image target detection model through the image data set, and to perform accuracy verification on the trained image target detection model through the model accuracy algorithm. The specific calculation formula is as follows: ; In the formula, Represents the accuracy value of the image target detection model, Indicates the number of sample data that are correctly detected. Indicates the number of sample data with detection errors. When the calculation result is greater than 0.95, it indicates that the accuracy of the image target detection model meets the requirements. When the calculation result is not greater than 0.95, it indicates that the accuracy of the image target detection model does not meet the requirements and the image target detection model needs to be optimized and adjusted. Specifically, the GIOU Loss loss function is selected as the optimization function of the image target detection model. The overlap between the detection frame and the target frame is detected by the GIOU Loss loss function, so as to optimize the accuracy of the detection frame. The detection result acquisition module 204 is electrically connected to the model training verification module 203. The detection result acquisition module 204 detects the collected image through the image target detection model that meets the use requirements, obtains the detection result, and obtains the position information of the target frame corresponding to each category of targets in the collected image. The mining area risk level judgment unit 300 includes a production risk factor acquisition module 301 and a production risk factor analysis module 302. The production risk factor acquisition module 301 collects texts related to open-pit mine production safety risk factors through an octopus collector, and performs co-word analysis on the collected texts related to open-pit mine production safety risk factors through Citespace software to obtain production risk factors in the mining area production operation process. The production risk factor analysis module 302 is electrically connected to the production risk factor acquisition module 301. The production risk factor analysis module 302 performs a correlation analysis between mining area production safety and mining area production risk factors on each production risk factor through SPSS data analysis software, and retains the first eight production risk factors based on the correlation analysis results between mining area production safety and mining area production risk factors, thereby realizing the screening of production risk factors and improving the accuracy of subsequent risk analysis; The mining area risk level judgment unit 300 also includes a risk assessment system generation module 303 and a risk level judgment module 304. The risk assessment system generation module 303 is electrically connected to the production risk factor analysis module 302. The risk assessment system generation module 303 constructs a risk assessment system based on the screened production risk factors. The risk level judgment module 304 is electrically connected to the risk assessment system generation module 303. The risk level judgment module 304 is based on the production risk factors in the risk assessment system. The Delphi method is used to judge the importance of the production risk factors in pairs, and the judgment results are further calculated by the SuperDecisions3.0 software to obtain the weight value of the corresponding production risk factor, thereby collecting data on the corresponding production risk factors and analyzing the acquired data. The data is standardized by the Max-Min standardization method so that the distribution range of the collected data is between 0 and 1. In combination with the weight values of the corresponding production risk factors obtained by calculation, the standardized processing values of each production risk factor are multiplied by the weight values of the corresponding production risk factors, and each product result is accumulated to solve the risk value of production safety in the mining area. The risk level is judged based on the solution result. Specifically, when 0 < solution result ≤ 0.2, it is judged that the risk level is very low; when 0.2 < solution result ≤ 0.4, it is judged that the risk level is low; when 0.4 < solution result ≤ 0.6, it is judged that the risk level is good; when 0.6 < solution result ≤ 0.8, it is judged that the risk level is high; when 0.8 < solution result ≤ 1, it is judged that the risk level is very high; The mining area monitoring alarm unit 400 includes an alarm warning notification module 401. The alarm warning notification module 401 is provided with alarm warning levels one to five from low to high corresponding to the risk level of the mining area, and based on the risk level of the mining area production operation process obtained by the risk level judgment module 304, an alarm warning signal of a corresponding level is issued to the corresponding mining work area, and the corresponding level of alarm warning signal notification is performed on the display screen set in the corresponding mining work area. It also includes based on the position information of the corresponding target frame of each category of target in the acquired image, for targets with spatial dependence, the spatial position relationship between the corresponding targets is calculated by the overlap method, and when it is detected that the spatial position relationship of the corresponding target does not meet the safety production requirements of the mining area, an alarm is generated. An alarm signal is generated, otherwise no alarm signal is generated. Taking the tires and vehicle limiting devices of the mining vehicle during the maintenance process of the mining vehicle in the mining equipment maintenance area as an example, when the overlap between the target frame corresponding to the tires of the mining vehicle during the vehicle maintenance process and the target frame corresponding to the vehicle limiting device is not greater than 0, it is considered that the vehicle limiting device is not correctly placed. When the overlap between the target frame corresponding to the tires of the mining vehicle during the vehicle maintenance process and the target frame corresponding to the vehicle limiting device is greater than 0, a coordinate system is established with the center position of the target frame corresponding to the tires of the mining vehicle as the coordinate origin. When it is detected that the target frame corresponding to the vehicle limiting device is located above the center of the target frame corresponding to the tires of the mining vehicle, it is also considered that the vehicle limiting device is not correctly placed. Otherwise, it is considered that the vehicle limiting device is correctly placed. The mining area monitoring alarm unit 400 also includes an alarm detection notification module 402, which is electrically connected to the alarm warning notification module 401. The alarm detection notification module 402 receives the alarm signal generated by the alarm warning notification module 401 through a signal receiver, and sends an alarm notification through a speaker at the same time, and controls the camera pan to save the current camera pan position information and zoom factor information, and then controls the camera pan to focus and advance, and continuously obtains and tracks the position of the alarm target to ensure that when there are multiple target alarms in the picture, each alarm target is tracked in sequence without repetition, and the camera pan after reinforcement learning ensures that the scale of the alarm target in the picture is appropriate, and ends the tracking after the camera pan alarm video acquisition is completed and returns to the position before tracking, and saves and uploads the alarm video.
[0019] Working principle: The mining area data acquisition module 101 collects relevant data of the mining area to be monitored, and the cruise monitoring point setting module 102 determines the cruise monitoring point of the camera pan / tilt table, thereby determining the cruise path of the camera pan / tilt table to monitor the mining area operation process. The monitoring video tracking module 103 performs reinforcement learning on the inspection process of the camera pan / tilt table. The mining area image acquisition module 104 collects video images of the operation process in the mining operation area during the actual cruise process. The image detection model construction module 201 constructs an image target detection model, and the sample data set construction module 202 constructs an image data set. The model training and verification module 203 trains the image target detection model, and the detection result acquisition module 204 performs model training on the image target detection model. 04 detects the collected image and obtains the detection result, obtains the position information of the target box corresponding to each category of target in the collected image, obtains the relevant production risk factors in the production operation process of the mining area through the production risk factor acquisition module 301, screens the production risk factors through the production risk factor analysis module 302, builds a risk assessment system through the risk assessment system generation module 303, determines the risk level through the risk level judgment module 304, performs alarm detection based on the risk level of the production operation process of the mining area through the alarm warning notification module 401, and also includes processing the image detection result, and performing alarm detection based on the processing result, and performing alarm notification and alarm recording operations based on the alarm detection result through the alarm detection notification module 402.
[0020] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.
[0021] Although embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.
Claims
1. The mining area production safety monitoring system based on target detection is characterized by: The invention comprises a mining area inspection image acquisition unit (100), wherein the mining area inspection image acquisition unit (100) collects relevant data of the mining area to be monitored, and formulates the operation trajectory of the staff in each mining operation area based on the relevant information of the mining area operation, thereby determining the monitoring trajectory of the camera pan / tilt platform, and setting the cruise monitoring point of the camera pan / tilt platform on the monitoring trajectory of the camera pan / tilt platform, and in the actual mining area operation process, by performing reinforcement learning on the camera pan / tilt platform, the camera pan / tilt platform can update the position of the actual monitoring point according to the actual operation trajectory of the staff, and collect relevant video images of the operation process; A mining area image recognition and analysis unit (200), the mining area image recognition and analysis unit (200) constructing an image target detection model and a sample data set for training the image target detection model, performing model training and precision verification operations on the image target detection model through the image data set, and when the precision of the image target detection model does not meet the use requirements, performing model optimization on the image target detection model to ensure the accuracy of the image target detection model, and then detecting the collected image through the image target detection model that meets the use requirements, and obtaining the detection result; A mining area risk level judgment unit (300), wherein the mining area risk level judgment unit (300) obtains relevant production risk factors in the mining area production operation process by performing a co-word analysis on text related to open-pit mine production safety risk factors, and screens the obtained production risk factors by performing a correlation analysis on mining area production safety and mining area production risk factors, and then constructs a risk assessment system based on the screened production risk factors, thereby evaluating the risk level of the mining area production operation process; A mining area monitoring and alarm unit (400), wherein the mining area monitoring and alarm unit (400) issues an alarm warning signal of a corresponding level based on the risk level of the mining area production operation process, and simultaneously performs an alarm warning signal notification of a corresponding level, and also includes processing the image detection result, performing alarm detection based on the processing result and in combination with the mining area production safety requirements, and then performing an alarm notification and an alarm video recording operation based on the alarm detection result.
2. The mining area production safety monitoring system based on target detection according to claim 1 is characterized in that: The mining area inspection image acquisition unit (100) comprises a mining area data acquisition module (101) and a cruise monitoring point setting module (102). The mining area data acquisition module (101) collects relevant data of the mining area to be monitored, including geographical information, environmental information, operator information, daily working area information of the operators, operating equipment information and operating cycle arrangement information of the mining area, and divides the mining area into a mining working area and a mining equipment maintenance area, and divides the mining working area into different and adjacent operating areas based on the operating cycle arrangement information. The cruise monitoring point setting module (102) is electrically connected to the mining area data acquisition module (101). The cruise monitoring point setting module (102) obtains the operating trajectory information of the mining area operators in the mining operating area within the corresponding operating cycle based on the operating cycle arrangement information corresponding to each mining operating area and further based on the working area information of the mining area operators in the mining operating area every day within a preset operating cycle, and then determines the cruise monitoring point of the camera pan / tilt head based on the operating trajectory information, thereby realizing the monitoring of the mining area operating process.
3. The mining area production safety monitoring system based on target detection according to claim 2 is characterized in that: The mining area inspection image acquisition unit (100) further comprises a monitoring video tracking module (103) and a mining area image acquisition module (104). The monitoring video tracking module (103) is electrically connected to the cruise monitoring point setting module (102). The monitoring video tracking module (103) performs reinforcement learning on the inspection process of the camera pan / tilt, so that in the actual mining area operation process, through the automatic adjustment of the camera pan / tilt, the image of the mining area operation process subsequently acquired is located at a suitable position in the picture. The specific operation is to use GYM to simulate and model the scene of the mining area operation area camera pan / tilt monitoring operation process, adopt a modeling method with a target frame as the state space, and design the state space, action space and reward function. The mining area image acquisition module (104) is electrically connected to the monitoring video tracking module (103). The mining area image acquisition module (104) uses the camera pan / tilt to collect video images of the operation process in the mining operation area during the actual cruise.
4. The mining area production safety monitoring system based on target detection according to claim 1 is characterized in that: The mining area image recognition and analysis unit (200) comprises an image detection model construction module (201) and a sample data set construction module (202). The image detection model construction module (201) constructs an image target detection model based on a YOLOv5 network, and the image target detection model specifically comprises four parts: an input end, a feature extraction network, a neck network, and an output end. The sample data set construction module (202) obtains video images of the operation process in the mining operation area collected by a historical camera pan-tilt from the mining area management personnel, and extracts a frame of the video every 2 seconds to obtain a sample image, and annotates the obtained sample image through LabelImg. In the process of labeling the image through LabelImg, a corresponding txt file is automatically generated for each image, and the txt file specifically contains the target category in the target frame of the sample image and the coordinate information of the target frame, and constructs a sample data set based on the annotated sample image.
5. The mining area production safety monitoring system based on target detection according to claim 4 is characterized in that: The mining area image recognition and analysis unit (200) further comprises a model training and verification module (203) and a detection result acquisition module (204). The model training and verification module (203) is electrically connected to the image detection model construction module (201) and the sample data set construction module (202) respectively. The model training and verification module (203) is used to perform model training on the image target detection model through the image data set, and to perform accuracy verification on the trained image target detection model through the model accuracy algorithm. When the calculation result is greater than 0.95, it indicates that the accuracy of the image target detection model meets the requirement. When the calculation result is not greater than 0.95, it indicates that the accuracy of the image target detection model does not meet the requirement and the image target detection model needs to be optimized and adjusted. Specifically, the GIOU Loss loss function is selected as the optimization function of the image target detection model. Through the GIOU The Loss function detects the overlap between the detection frame and the target frame, thereby optimizing the accuracy of the detection frame. The detection result acquisition module (204) is electrically connected to the model training verification module (203). The detection result acquisition module (204) detects the collected image using an image target detection model that meets the use requirements, obtains the detection result, and obtains the position information of the target frame corresponding to each category of target in the collected image.
6. The mining area production safety monitoring system based on target detection according to claim 1 is characterized in that: The mining area risk level judgment unit (300) comprises a production risk factor acquisition module (301) and a production risk factor analysis module (302). The production risk factor acquisition module (301) collects text related to open-pit mine production safety risk factors through an Octopus collector, and performs co-word analysis on the collected text related to open-pit mine production safety risk factors through Citespace software to obtain production risk factors in the mining area production operation process. The production risk factor analysis module (302) is electrically connected to the production risk factor acquisition module (301). The production risk factor analysis module (302) performs a correlation analysis between mining area production safety and mining area production risk factors on each production risk factor through SPSS data analysis software, and retains the first eight production risk factors based on the correlation analysis results between mining area production safety and mining area production risk factors, thereby realizing the screening of production risk factors and improving the accuracy of subsequent risk analysis.
7. The target detection-based mining production safety monitoring system according to claim 6 is characterized in that: The mining area risk level judgment unit (300) further comprises a risk assessment system generation module (303) and a risk level judgment module (304), wherein the risk assessment system generation module (303) is electrically connected to the production risk factor analysis module (302), wherein the risk assessment system generation module (303) constructs a risk assessment system based on the screened production risk factors, wherein the risk level judgment module (304) is electrically connected to the risk assessment system generation module (303), wherein the risk level judgment module (304) performs a pairwise importance judgment on the production risk factors based on the production risk factors in the risk assessment system through the Delphi method, and further The first step is to calculate the evaluation results through SuperDecisions3.0 software to obtain the weight value of the corresponding production risk factor, and then collect data on the corresponding production risk factors and standardize the acquired data through the Max-Min standardization method so that the distribution range of the collected data is between 0 and 1. Combined with the calculated weight value of the corresponding production risk factor, the standardized processing value of each production risk factor is multiplied by the weight value of the corresponding production risk factor, and each product result is accumulated to achieve the solution of the risk value of mine production safety, and make risk level judgment based on the solution result.
8. The mining area production safety monitoring system based on target detection according to claim 7 is characterized in that: The mining area monitoring alarm unit (400) includes an alarm warning notification module (401), wherein the alarm warning notification module (401) is provided with alarm warning levels one to five from low to high corresponding to the risk level of the mining area, and based on the risk level of the mining area production operation process obtained by the risk level judgment module (304), an alarm warning signal of a corresponding level is issued to the corresponding mining work area, and the alarm warning signal notification of the corresponding level is performed on a display screen set in the corresponding mining work area, and also includes based on the position information of the corresponding target frame of each category of target in the acquired image, for targets with spatial dependence, using an overlap method to calculate the spatial position relationship between the corresponding targets, and when it is detected that the spatial position relationship of the corresponding targets does not meet the safety production requirements of the mining area, an alarm signal is generated, otherwise no alarm signal is generated.
9. The target detection-based mining production safety monitoring system according to claim 8 is characterized in that: The mining area monitoring alarm unit (400) further comprises an alarm detection notification module (402), wherein the alarm detection notification module (402) is electrically connected to the alarm warning notification module (401), wherein the alarm detection notification module (402) receives the alarm signal generated by the alarm warning notification module (401) through a signal receiver, and simultaneously issues an alarm notification through a loudspeaker, and controls the camera pan to save the current position information and zoom factor information of the camera pan, and then controls the camera pan to focus and advance, and continuously obtains and tracks the position of the alarm target, ensuring that when there are multiple target alarms in the picture, each alarm target is tracked in sequence without duplication, and the camera pan after reinforcement learning ensures that the scale of the alarm target in the picture is appropriate, and after the camera pan alarm video is collected, the tracking is terminated and the position before tracking is returned, and the alarm video is saved and uploaded.
