A method and system for visualizing coal mine information

Abstract

The present invention relates to the field of information visualization. In particular, it relates to a method and system for visualizing coal mine information, which includes the following steps: obtaining a full-scale mining area monitoring video; performing cache delay optimization and worker visual calibration, and extracting multiple worker position boundary boxes; identifying key reference points in the mining area based on the full-scale mining area monitoring video, and performing three-dimensional structural modeling to construct a three-dimensional structural model of the mining area; performing real-time dynamic tracking based on multiple worker position boundary boxes, and performing instant mobile mapping of the three-dimensional structural model of the mining area to construct a real-time mining area twin model; obtaining multi-dimensional environmental monitoring parameters of the mining area, and performing toxic gas concentration fluctuation fitting and hazardous gas area positioning, and marking hazardous gas areas. The present invention realizes real-time and dynamic visualization of coal mine information, and improves the operational safety of the mining area.

Description

Technical Field

[0001] The present invention relates to the field of information visualization, and in particular to a method and system for visualizing coal mine information. Background Art

[0002] With the continuous development of information technology and digitalization, the coal mining industry is undergoing a gradual digital transformation in production, management, and safety. As a high-risk, high-intensity operating environment, coal mines place production safety and operational efficiency at the core of the industry's priorities. Traditional coal mine information management relies heavily on manual records, static drawings, or two-dimensional diagrams. Information transmission and processing often suffer from issues such as poor timeliness and unclear visualization. This makes it difficult to quickly and accurately obtain real-time mine information in complex mining environments, increasing safety hazards and operational risks.

[0003] As coal mining operations increase in depth and complexity, traditional mine management models and information monitoring methods are no longer adapting to new demands. Complex geological conditions, frequent environmental changes, unstable equipment operation, and widely distributed personnel within mines place higher demands on real-time monitoring and visualization of coal mine information. Especially in the event of sudden disasters and accidents, obtaining comprehensive and accurate mine information in the shortest possible time for rapid decision-making and emergency response becomes crucial for coal mine safety management.

[0004] Traditional coal mine information monitoring systems rely on two-dimensional floor plans or static data displays, which are unable to effectively display the spatial layout of the mine, the operating status of complex equipment, or real-time operational status. This approach often suffers from limitations such as delayed information updates, difficulty visualizing spatial relationships, and difficulty conducting interactive three-dimensional spatial analysis, making it difficult to provide effective decision support. With the maturity of the Internet of Things, big data, and cloud computing technologies, real-time data collection and analysis have become crucial for coal mine safety management. However, data collection alone is not sufficient to improve decision-making efficiency. Transforming this real-time data into intuitive and easy-to-understand three-dimensional visualizations has become a new challenge. Summary of the Invention

[0005] In order to solve the above technical problems, the present invention proposes a method and system for visualizing coal mine information to solve at least one of the above technical problems.

[0006] To achieve the above object, the present invention provides a method for visualizing coal mine information, comprising the following steps:

[0007] Step S1: Obtain omnidirectional mining area monitoring video; perform cache delay optimization and worker visual calibration, and extract multiple worker position bounding boxes;

[0008] Step S2: identifying key reference points in the mining area based on the all-round mining area monitoring video, and performing three-dimensional structural modeling to construct a three-dimensional structural model of the mining area;

[0009] Step S3: Real-time dynamic tracking is performed based on the bounding boxes of multiple worker positions, and real-time mobile mapping of the mining area 3D structure model is performed to build a real-time mining area twin model;

[0010] Step S4: Obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting and dangerous gas area positioning, and mark the dangerous gas area;

[0011] Step S5: Inferring dangerous terrain based on the real-time mining twin model, and then highlighting and visualizing the dangerous gas area to obtain a warning map of the dangerous area;

[0012] Step S6: Based on the warning map of the dangerous area, real-time early warning decisions are made and dynamic holographic visualization is performed to build a holographic visualization model of the mining area information.

[0013] This invention uses video surveillance and computer vision technology to accurately locate each worker's location, helping managers understand the distribution of workers within the mine in real time. It reduces video stream latency, ensuring data processing speed and real-time response, and safeguarding subsequent dynamic modeling and real-time feedback. Accurate worker location calibration enables real-time monitoring of workers' dynamic behavior, reducing the likelihood of workers entering dangerous areas and improving mine safety. By identifying key reference points within the mine, the accuracy of 3D modeling is ensured. The 3D model provides mine managers with a comprehensive understanding of the mine's spatial layout. The 3D structural model helps visualize the complex mine environment, enhancing the spatial awareness of workers and managers. It provides a solid foundation for subsequent worker location calibration and dangerous area warnings. Through dynamic tracking, the system updates each worker's position changes in real time, ensuring mine managers have an accurate understanding of worker distribution. The 3D model of the mine adjusts in real time based on worker movement, creating a dynamic twin model of the mine, providing real-time data support for management decisions. Through real-time dynamic mapping and tracking, managers can promptly detect abnormal worker behavior and prevent workers from entering dangerous areas. By acquiring multi-dimensional environmental data in real time, the system monitors changes in the mining environment, ensuring managers are immediately aware of gas conditions. Fluctuations in toxic gas concentrations are fitted to effectively predict potential changes in hazardous gas concentrations, allowing for timely identification of hazardous areas and improving predictive safety management. By combining gas concentrations with terrain features, a comprehensive warning map of hazardous areas is generated, comprehensively highlighting risk points within the mining area. Highlighting and visualizing hazardous areas visually alert managers and workers to hazardous locations, facilitating timely response. Visual warning maps help decision-makers and workers quickly locate hazardous areas, improving response efficiency and safety assurance. Real-time identification of hazardous situations and decision-making recommendations reduce delays in human decision-making and ensure safe operations within the mining area. Dynamic holographic visualization technology allows managers to view real-time information within a virtual environment, intuitively monitoring mine operations and significantly improving decision-making efficiency. Workers can use augmented reality devices to visualize their safety status and routes, enhancing worker safety.

[0014] In this specification, a coal mine information visualization system is provided, which is used to execute the above-mentioned coal mine information visualization method, including:

[0015] The worker vision calibration module is used to obtain omnidirectional mining area monitoring videos; it also performs cache latency optimization and worker vision calibration to extract multiple worker position bounding boxes;

[0016] The 3D structure module is used to identify key reference points in the mining area based on the omnidirectional mining area monitoring video, and to perform 3D structure modeling to construct a 3D structural model of the mining area;

[0017] The twin model module is used to perform real-time dynamic tracking based on the bounding boxes of multiple worker positions and to perform instant mobile mapping of the 3D structural model of the mine area to build a real-time twin model of the mine area;

[0018] The toxic gas sensing module is used to obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting, locate hazardous gas areas, and mark hazardous gas areas;

[0019] The dangerous area rendering module is used to infer dangerous terrain based on the real-time mining twin model, and then highlight and visualize the dangerous gas areas to obtain a warning map of the dangerous area;

[0020] The real-time warning module is used to make real-time warning decisions and perform dynamic holographic visualization based on the warning map of dangerous areas, and to build a holographic visualization model of mining area information.

[0021] This invention accurately identifies worker locations, ensuring monitoring of each worker. In emergencies, their specific locations can be quickly determined, reducing the risk of workplace accidents. By optimizing video processing buffering and latency, worker location bounding boxes can be updated in real time, improving system response speed, reducing latency, and ensuring smooth dynamic tracking. 3D modeling visually displays the spatial structure of the mine, including complex underground passages and mine facilities, helping managers understand the mine's layout. The 3D model of the mine can serve as a foundation for future design and planning, improving resource management efficiency and optimizing the layout of mine facilities. Real-time modeling of the mine's 3D structure allows managers to identify structural issues (such as tunnel collapse) and implement timely maintenance or repairs. The twin model updates worker locations in real time and reflects them in the 3D model, providing mine managers with accurate real-time data and ensuring they are always aware of personnel movements within the mine. Through real-time dynamic tracking, the twin model can help predict potential dangerous situations, enabling proactive prevention and emergency response plans. Real-time tracking of worker locations helps managers optimize work schedules and avoid congestion in high-risk areas. Real-time gas concentration analysis enables early warning and measures to prevent toxic gas leaks from causing disasters such as fires or explosions. Accurately fitting toxic gas concentration fluctuations helps locate risky areas, providing a basis for personnel evacuation and evacuation. 3D rendering allows all hazardous areas to be visually displayed within the mine twin model, significantly improving mine managers' ability to identify potential risks. Real-time marking of hazardous areas within the mine facilitates emergency management, allowing workers to quickly receive instructions and take swift emergency response measures. Rendered hazardous area maps enable workers to quickly identify and avoid hazardous areas, improving mine safety. Fusion analysis of real-time data allows mines to rapidly respond to potential safety incidents, issuing early warnings and reducing the likelihood of accidents. Dynamic holographic visualization provides mine managers with a clearer view of hazardous areas and the overall mine safety status, enabling quick and accurate decision-making. The holographic visualization model of the mine allows simulation of various hazards within the mine, further enhancing safety and control capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0022] Figure 1 This is a schematic flow chart of the steps of a method for visualizing coal mine information according to the present invention;

[0023] Figure 2 Detailed implementation flow chart of step S1;

[0024] Figure 3 Detailed implementation flow chart of step S2;

[0025] Figure 4Schematic diagram of the detailed implementation steps of step S3. DETAILED DESCRIPTION

[0026] It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0027] This application provides a method and system for visualizing coal mine information. The implementation entities of the method and system include, but are not limited to, mechanical equipment, data processing platforms, cloud server nodes, and network upload devices that are equipped with the system, which can be considered general computing nodes of this application. The data processing platform includes, but is not limited to, at least one of an audio and image management system, an information management system, and a cloud data management system.

[0028] See also Figures 1 to 4 The present invention provides a method for visualizing coal mine information, which comprises the following steps:

[0029] Step S1: Obtain omnidirectional mining area monitoring video; perform cache delay optimization and worker visual calibration, and extract multiple worker position bounding boxes;

[0030] Step S2: identifying key reference points in the mining area based on the all-round mining area monitoring video, and performing three-dimensional structural modeling to construct a three-dimensional structural model of the mining area;

[0031] Step S3: Real-time dynamic tracking is performed based on the bounding boxes of multiple worker positions, and real-time mobile mapping of the mining area 3D structure model is performed to build a real-time mining area twin model;

[0032] Step S4: Obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting and dangerous gas area positioning, and mark the dangerous gas area;

[0033] Step S5: Inferring dangerous terrain based on the real-time mining twin model, and then highlighting and visualizing the dangerous gas area to obtain a warning map of the dangerous area;

[0034] Step S6: Based on the warning map of the dangerous area, real-time early warning decisions are made and dynamic holographic visualization is performed to build a holographic visualization model of the mining area information.

[0035] This invention uses video surveillance and computer vision technology to accurately locate each worker's location, helping managers understand the distribution of workers within the mine in real time. It reduces video stream latency, ensuring data processing speed and real-time response, and safeguarding subsequent dynamic modeling and real-time feedback. Accurate worker location calibration enables real-time monitoring of workers' dynamic behavior, reducing the likelihood of workers entering dangerous areas and improving mine safety. By identifying key reference points within the mine, the accuracy of 3D modeling is ensured. The 3D model provides mine managers with a comprehensive understanding of the mine's spatial layout. The 3D structural model helps visualize the complex mine environment, enhancing the spatial awareness of workers and managers. It provides a solid foundation for subsequent worker location calibration and dangerous area warnings. Through dynamic tracking, the system updates each worker's position changes in real time, ensuring mine managers have an accurate understanding of worker distribution. The 3D model of the mine adjusts in real time based on worker movement, creating a dynamic twin model of the mine, providing real-time data support for management decisions. Through real-time dynamic mapping and tracking, managers can promptly detect abnormal worker behavior and prevent workers from entering dangerous areas. By acquiring multi-dimensional environmental data in real time, the system monitors changes in the mining environment, ensuring managers are immediately aware of gas conditions. Fluctuations in toxic gas concentrations are fitted to effectively predict potential changes in hazardous gas concentrations, allowing for timely identification of hazardous areas and improving predictive safety management. By combining gas concentrations with terrain features, a comprehensive warning map of hazardous areas is generated, comprehensively highlighting risk points within the mining area. Highlighting and visualizing hazardous areas visually alert managers and workers to hazardous locations, facilitating timely response. Visual warning maps help decision-makers and workers quickly locate hazardous areas, improving response efficiency and safety assurance. Real-time identification of hazardous situations and decision-making recommendations reduce delays in human decision-making and ensure safe operations within the mining area. Dynamic holographic visualization technology allows managers to view real-time information within a virtual environment, intuitively monitoring mine operations and significantly improving decision-making efficiency. Workers can use augmented reality devices to visualize their safety status and routes, enhancing worker safety.

[0036] In the embodiment of the present invention, see Figure 1 , is a schematic flow chart of the steps of a method for visualizing coal mine information according to the present invention. In this example, the steps of the method for visualizing coal mine information include:

[0037] Step S1: Obtain omnidirectional mining area monitoring video; perform cache delay optimization and worker visual calibration, and extract multiple worker position bounding boxes;

[0038] In this example, multiple high-resolution cameras are deployed throughout the mine to ensure coverage of key areas. These cameras should have night vision and earthquake resistance to operate in a variety of environmental conditions. Assume that 10 cameras are deployed throughout the mine, distributed across key work areas, entrances and exits, and hazardous areas. This ensures that each key area is monitored by at least one camera. Configure the video capture system and set acquisition parameters (such as resolution, frame rate, and video encoding format). Typically, the video resolution is 1080p and the frame rate is 30 fps to ensure that workers' movements are captured. The H.264 codec is used for video compression to reduce storage space while ensuring video quality. Each camera has a storage capacity of 1 TB, ensuring at least one week of monitoring video can be stored. The captured video data is transmitted to the central control system in real time using a stable network protocol (such as RTSP or RTMP). This ensures the real-time and integrity of the video data. The network bandwidth is set to 10 Mbps to support simultaneous transmission of video streams from multiple cameras, ensuring that data from each camera can be seamlessly connected to the control center. During video capture and transmission, monitor the system's latency and analyze the latency of the video stream transmission. Use network monitoring tools (such as Wireshark) to assess network latency and packet loss. If latency exceeds 200 milliseconds during certain periods, optimization is required to ensure real-time monitoring. Design and implement an effective caching strategy, using a ring buffer or queue structure to store recent video data. This strategy can help reduce transmission latency and improve video streaming stability. Set the buffer size to 30 seconds of video data to prevent data loss and ensure data flow continuity even under poor network conditions. After implementing the caching strategy, continuously monitor network latency to ensure optimization results. Record latency changes before and after optimization to evaluate the effectiveness of the caching strategy. After optimization, the monitored latency dropped to 150 milliseconds, meeting real-time monitoring requirements and ensuring timely reflection of worker activities. Prepare a calibration scene to ensure that the camera accurately captures the worker's location. Calibration markers (such as colored dots or squares) can be set to identify the worker's location in the video. Use the calibration tool to place multiple, distinctly colored markers on the ground in key areas of the mine to facilitate algorithm recognition. Select and apply an appropriate worker detection algorithm (such as YOLO, SSD, or other real-time object detection algorithms) to analyze the video stream frame by frame and extract the worker's bounding boxes. Ensure that the algorithm works stably under various lighting conditions. Use the YOLOv5 model, set the input resolution to 640x640, and ensure that workers are accurately detected and bounding boxes are generated in each frame. Record the bounding box information of the detected worker locations, including the bounding box coordinates and recognition confidence of each worker. Ensure that all data can be stored in a database for subsequent analysis and visualization.The recording format is "timestamp, worker ID, bounding box coordinates (x1, y1, x2, y2), confidence level" to ensure data integrity and traceability.

[0039] Step S2: identifying key reference points in the mining area based on the all-round mining area monitoring video, and performing three-dimensional structural modeling to construct a three-dimensional structural model of the mining area;

[0040] In this example, key frames are extracted from a 360-degree mining area monitoring video for subsequent reference point identification. The extraction frequency is set to one frame per second to ensure that dynamic changes within the mining area are captured. Assuming a 10-minute monitoring video, a total of 600 frames are extracted. Each extracted frame undergoes preprocessing, including noise removal and contrast enhancement, to improve the accuracy of subsequent feature extraction. Key reference points are identified in each frame using a feature point detection algorithm (such as SIFT, ORB, or SURF). These feature points are typically objects with significant textures or shapes within the mining area, such as ore piles, equipment, and signage. The ORB algorithm is used to detect 500 feature points in each frame, with a threshold of 10 to ensure that only the most significant feature points are retained for subsequent analysis. The coordinates and descriptors of each feature point are recorded. Feature points in each frame are matched, and identical feature points across frames are associated using the FLANN (Fast Library for Approximate Nearest Neighbors) or BFMatcher algorithms. A matching threshold (e.g., 0.75) is set to select high-quality matching pairs. If a feature point A in one frame matches in the next frame, and the matching degree exceeds a threshold, the feature point pair is recorded. Ultimately, approximately 350 stable matching feature point pairs are retained to ensure the reliability of the feature points. Before 3D modeling, the camera must be calibrated to obtain its intrinsic and extrinsic parameters. Multiple calibration images are captured using the checkerboard calibration method, and calibration is performed by calculating the camera matrix. Ten checkerboard images are captured at different angles, and the Zhang Zhengyou calibration method is used to calculate the camera's intrinsic parameter matrix and distortion coefficients to ensure the accuracy of 3D reconstruction. Based on the feature matching results, a 3D point cloud is generated using triangulation. The camera's intrinsic and extrinsic parameters are used to convert the pixel coordinates of each feature point pair into 3D world coordinates. If the pixel coordinates of feature point A are (x1, y1) and (x2, y2), its 3D coordinates are calculated as (X, Y, Z) using the camera parameters. This generates a point cloud dataset containing thousands of 3D points. The generated point cloud data is processed, including downsampling, denoising, and point cloud registration. A Voxel Grid filter was used to downsample the point cloud to reduce data size and improve computational efficiency. The point cloud data was reduced from 100,000 points to 20,000 points. Subsequently, the ICP (Iterative Closest Point) algorithm was used to align point clouds from different viewpoints to ensure the accuracy and integrity of the point cloud data. Based on the processed point cloud data, a 3D model was constructed using 3D modeling software such as Meshlab or Blender. By generating a mesh and applying texture mapping, the point cloud data was converted into a visual 3D model.The optimized point cloud data was then converted into a 5,000-triangle mesh model, to which texture mapping was added to ensure the visualisation accurately reflected the appearance of the mining area.

[0041] Step S3: Real-time dynamic tracking is performed based on the bounding boxes of multiple worker positions, and real-time mobile mapping of the mining area 3D structure model is performed to build a real-time mining area twin model;

[0042] In this embodiment, the worker position bounding box data extracted from the previous step is updated in real time. Each bounding box should contain the worker's location information (such as coordinates, width, and height) as well as the recognition confidence level. The monitoring system receives the video stream and bounding box data in real time to ensure data accuracy and timeliness. Assume that the monitoring system updates the worker's location every second and records the bounding box of worker A as (220, 340, 50, 70) in real time, indicating that its center point coordinates are (220, 340), the bounding box width is 50 pixels, and the height is 70 pixels. A dynamic tracking algorithm (such as the Kalman filter or the SORT (Simple Online and Realtime Tracking) algorithm) is used to track the worker's bounding box in real time. These algorithms can handle changes in worker position and improve tracking accuracy through prediction and update steps. If the worker moves from position (220, 340) to (225, 345) in the monitoring video, the Kalman filter can predict the next position based on the worker's velocity and acceleration, ensuring tracking continuity. Tracking results are recorded in real time in a database, including the worker's ID, timestamp, current location, and bounding box information. Ensure that all data is readily accessible for subsequent analysis and visualization. Ensure that a 3D structural model of the mine site can be integrated with the worker's real-time location data. The model should include information about the mine's topography, equipment, and major access routes. Convert the model to a format suitable for real-time updates (such as those used in virtual reality or augmented reality). The coordinate system of the mine site model must be consistent with the coordinate system of the worker's location, ensuring that the worker's location can be directly mapped to the 3D model. Based on the real-time tracked worker location, apply a motion mapping algorithm to update the worker's location to the 3D model. Linear interpolation or other interpolation methods can be used to smooth the worker's movement path to ensure smooth model updates. If worker A moves from (220, 340) to (225, 345) in 1 second, update their position on the model from (220, 340, 0) to (225, 345, 0) to ensure accurate representation on the ground. Render the updated worker locations into the 3D structural model of the mine in real time, ensuring that managers can intuitively see the real-time distribution of workers within the mine. Use a graphics rendering engine (such as Unity or UnrealEngine) for visualization. Integrate the workers' real-time location information with the 3D model of the mine to form a real-time mine twin model. Ensure data synchronization and consistency to reflect the actual situation in the mine in real time. Set an update frequency of once per second to ensure that each update accurately reflects the location and status of workers. In the real-time mine twin model, provide global and local views to facilitate comprehensive monitoring by managers. The global view shows the situation of the entire mine, while the local view highlights the specific area where the workers are located.The layout of the entire mine area is displayed in the global view, and the area where worker A is located is highlighted in the local view to facilitate quick identification of his location.

[0043] Step S4: Obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting and dangerous gas area positioning, and mark the dangerous gas area;

[0044] In this example, various sensors are deployed within the mining area, including oxygen sensors, combustible gas sensors (such as methane and carbon monoxide), toxic gas sensors (such as hydrogen sulfide), temperature and humidity sensors, and air pressure sensors. Sensor coverage is ensured in key areas of the mining area to comprehensively monitor environmental conditions. Assuming 12 gas sensors, 6 oxygen sensors, and other environmental parameter sensors are deployed within the mining area, ensure that each sensor accurately measures and achieves the specified sensitivity level (for example, a gas sensor with a detection sensitivity of 1 ppm). A data acquisition system is configured, with a set data acquisition frequency (e.g., once per second) and real-time transmission of collected data to the central monitoring system. A stable communication protocol (such as LoRa or Zigbee) is used for data transmission to ensure real-time data integrity. Each sensor's acquisition cycle is set to 1 second to ensure data stability under varying environmental conditions (such as high humidity and high temperature), and a data caching mechanism is implemented to prevent data loss. Real-time toxic gas concentration data is extracted from the database and cleaned and preprocessed. Outliers and noise are removed to ensure data accuracy and reliability. Use statistical methods (such as the Z-score) to identify and eliminate data points with abnormal concentration fluctuations to ensure the accuracy of subsequent fluctuation fitting. Select an appropriate mathematical model to fit the fluctuations of toxic gas concentrations. Common models include linear regression, polynomial fitting, and time series analysis methods (such as the ARIMA model). Choose the model that best suits the current data characteristics. If the extracted methane concentration data exhibits nonlinear fluctuations over a period of time, select a quadratic polynomial fitting model for concentration fluctuation analysis. Apply the selected fitting model to model the toxic gas concentration data and generate a concentration fluctuation curve. Evaluate the model's fit by calculating goodness-of-fit metrics (such as the R² value), and adjust model parameters to optimize the fitting results. Based on the results of the toxic gas concentration fluctuation fitting, set thresholds for hazardous gas concentrations. Generally, if a gas concentration exceeds a safety threshold (for example, a methane concentration exceeding 1.5%), the area is marked as hazardous. The safety threshold for carbon monoxide is set at 50 ppm; if the monitored concentration exceeds this value, the area is considered high-risk. Continuously monitor data from each sensor and determine in real time whether any hazardous gas concentration exceeds the set threshold. If toxic gas concentrations in a particular area exceed a threshold, the area is immediately marked as hazardous. For example, if a carbon monoxide concentration of 70 ppm is detected within a monitoring area, the area is marked as a "hazardous gas zone" and the relevant parameters are recorded. Identified hazardous gas areas are marked to ensure they are clearly displayed on the mine's visualization platform. Different colors or symbols are used to represent different types of hazardous areas (e.g., red for high-risk areas, yellow for medium-risk areas).

[0045] Step S5: Inferring dangerous terrain based on the real-time mining twin model, and then highlighting and visualizing the dangerous gas area to obtain a warning map of the dangerous area;

[0046] In this example, a three-dimensional model of the mining area and environmental monitoring data are used to collect terrain information, including slope, undulation, and potential landslide areas. These terrain features are extracted using graphics processing software to ensure data accuracy. A digital elevation model (DEM) is used to generate elevation data for the mining area, and the average slope of each area is calculated. Areas with slopes exceeding 30° are considered potentially hazardous terrain. Criteria for identifying hazardous terrain are set based on the mining area's operational requirements and safety standards. Areas with slopes exceeding a set threshold (e.g., 30°) or with significant geological fractures or fissures are considered hazardous terrain. If an area has a calculated slope of 35° and cracks are present on the surface, the area is marked as "high-risk terrain." Algorithms (such as rule-based inference systems or machine learning models) are used to analyze the collected terrain data and infer potential hazardous terrain areas. This algorithm can comprehensively consider multiple terrain features to make a comprehensive assessment. If an area meets both the slope and cracks hazard criteria, it is marked as "hazardous terrain," and its specific location and type are recorded. Integrate hazardous terrain areas with previously identified hazardous gas area data to form a comprehensive hazardous area dataset. Ensure that all data is easily accessible for subsequent visualization. If areas where carbon monoxide concentrations exceed safety thresholds overlap with high-slope areas, mark them as "comprehensive hazardous areas." Select appropriate visualization rendering technology to highlight hazardous areas with different colors or symbols within the 3D mine model. Graphics rendering engines such as Unity or Unreal Engine can be used for visualization. Highlight all hazardous terrain areas in red, and areas with excessive gas concentrations in yellow, ensuring clear distinction between areas of different risk levels. Configure the visualization system to update and display hazardous area information in real time. When the monitoring system detects a change in hazardous area status, the 3D model immediately reflects this change. If gas concentrations in a particular area return to safe levels, the area's color in the visualization model changes from red to green, providing real-time feedback to management. Based on the integrated hazardous area data, design and generate a hazardous area warning map. The map should include the overall layout of the mine, hazardous terrain, and areas with excessive gas concentrations, and ensure clear and readable information. The map scale is set to 1:1000 to ensure the location and size of each hazardous area are accurately identified. Dangerous area markings are integrated into the warning map, and necessary legends and explanations are included to help users understand the risk level of each area. The legend indicates that red represents high-risk terrain and yellow represents areas with excessive gas levels. Users can use this legend to quickly identify the safety status of the current mining area.

[0047] Step S6: Based on the warning map of the dangerous area, real-time early warning decisions are made and dynamic holographic visualization is performed to build a holographic visualization model of the mining area information.

[0048] In this embodiment, a real-time monitoring system is deployed within the mining area to continuously monitor changes in the status of hazardous areas. The system should be able to obtain the latest data from sensors and video surveillance and compare this information with the hazardous area warning map. The data collection frequency is set to once per second to ensure that the system receives the latest data from gas sensors and video surveillance in real time, updating the status of hazardous areas in real time. A real-time early warning mechanism is designed. When a worker enters a hazardous area or the concentration of hazardous gas exceeds a threshold, the system should immediately issue an alarm and notify relevant personnel. Conditional triggering can be used to implement these warnings. If a worker enters an area marked as "high risk," the system will immediately trigger an alarm, record the entry time, worker ID, and hazardous area information, and send it to the management terminal. The early warning information is displayed on the monitoring system interface, including the specific location of the hazardous area, the identity of the affected workers, and specific safety recommendations (such as evacuation or taking protective measures). On the monitoring interface, the hazardous area is marked with a red box, and an alert message is displayed next to it, such as "Worker A is in the hazardous area, please evacuate immediately." Appropriate holographic visualization technology (such as augmented reality (AR) or virtual reality (VR)) is selected to display real-time data and hazardous area information in the mining area. Ensure the technology supports dynamic updates and multi-user interaction. Use AR technology to overlay a 3D mine model with real-time monitoring data, ensuring managers can visually visualize each worker and their location relative to hazardous areas. Integrate real-time monitoring data (such as worker locations, hazardous gas concentrations, and hazardous area information) into the holographic visualization model. Ensure the model updates in real time to reflect the mine's status based on the latest data changes. If gas concentrations in a particular area return to safe levels, the system should change the color of that area in the holographic model from red to green, dynamically reflecting the change in safety status. Verify the constructed holographic visualization model to ensure it accurately reflects the actual conditions at the mine. Compare the model with field data to verify the consistency between worker locations and hazardous area markings. If the error between the worker locations in the holographic model and the real-time monitoring data is within 5 meters, the model is considered verified, ensuring the model's credibility. Collect feedback from managers on the holographic visualization model, analyze user experience issues, and optimize the system. Ensure the model's ease of use and practicality to better support mine safety management. If managers report that certain features are difficult to use, adjust the interface and optimize functionality based on this feedback to ensure users can quickly access the information they need. Regularly update the base map and data of the holographic visualization model to ensure that the model always reflects the latest status of the mine site. Establish a maintenance plan to ensure the long-term stable operation of the system. Update the 3D model of the mine site monthly to reflect new equipment layouts and environmental changes, ensuring that managers always have the most up-to-date information.

[0049] In this embodiment, refer to Figure 2, is a flowchart of the detailed implementation steps of step S1. In this embodiment, the detailed implementation steps of step S1 include:

[0050] Step S11: Acquire a full range of mining area monitoring videos based on multiple high-definition cameras in the mining area;

[0051] Step S12: Optimizing the cache delay of the all-round mining area monitoring video to construct a delay-optimized monitoring video;

[0052] Step S13: Perform deep visual inspection on the latency optimization monitoring video and mark each worker node;

[0053] Step S14: Calculate the real-time position of the worker nodes to generate the real-time position coordinates of each worker;

[0054] Step S15: calibrate the image bounding box of the time-delay optimized monitoring video according to the real-time position coordinates of each worker, and extract multiple worker position bounding boxes.

[0055] In this embodiment, multiple high-definition cameras are strategically placed within the mining operation area to ensure coverage of the entire operation area. Cameras should feature high resolution (e.g., 1080p or higher) to capture clear video images. For a 1,000-square-meter operation area, at least six cameras may be required, positioned in different corners and strategic locations to ensure effective monitoring of every worker activity area. A video acquisition server is configured to receive real-time video streams from each camera. Ensure the system has sufficient bandwidth and storage capacity to process high-definition video data. Set the system's network bandwidth to at least 10 Mbps to support the concurrent transmission of multiple video streams. Use an efficient encoding format (e.g., H.264) to optimize video stream transmission efficiency and storage space. To optimize video caching latency, design an efficient video caching mechanism. This mechanism should dynamically adjust caching strategies based on network conditions and video stream priority to ensure real-time and smooth video streaming. Prioritize video streams from key worker areas over those from other areas to ensure that video streams in critical areas are not interrupted when bandwidth is insufficient. Analyze the cache latency of the video stream, identify the main factors affecting latency (such as network fluctuations and storage speed), and apply optimization algorithms to adjust them. Utilize adaptive bitrate streaming (ABR) technology to dynamically adjust the resolution and frame rate of the video stream based on network conditions, ensuring video stream consistency even when bandwidth fluctuates. After video cache optimization, generate a latency-optimized monitoring video to ensure a smooth visual experience during real-time monitoring. Record latency data before and after optimization for subsequent analysis. The optimized video stream latency should be controlled within 100 milliseconds to ensure that workers' real-time movements are promptly reflected in the monitoring system. Use deep learning techniques (such as convolutional neural networks (CNNs)) to train a visual model for worker detection. This model should be able to accurately identify and label workers in the video. Use a labeled dataset containing multiple worker activity scenes for training to ensure that the model can recognize different worker postures and movements. Input the latency-optimized monitoring video into the deep vision detection model for real-time processing. The model should be able to identify each worker and label them within the video frame. When the model outputs the labeling results, it draws a bounding box around each worker and labels the worker's identity information (such as work number or name). The detected worker node information (such as location coordinates and labeling time) is recorded in the database for subsequent query and analysis. This ensures data integrity and real-time performance. If a worker's coordinates in a video frame are (300, 450), they are recorded as "work number, timestamp, coordinates." By analyzing the worker labels in the video frame, the real-time coordinates of each worker are calculated. Taking into account the video resolution, the coordinates are accurate and consistent with the actual scene.If the video resolution is 1920x1080 and the worker's coordinates in the image are (300, 450), these coordinates are converted to the actual work area coordinates and corrected based on the camera's installation height and angle. Based on each worker's real-time location coordinates, a corresponding bounding box is drawn in the monitoring video. Multiple worker location bounding boxes are extracted to facilitate subsequent data analysis and visualization. If a worker's bounding box coordinates are between (290, 440) and (310, 460), this box is drawn in the video and its location information is recorded. The extracted multiple worker location bounding boxes are integrated to generate a comprehensive worker monitoring report, including each worker's real-time coordinates and status information, which is then visualized in the monitoring system. A graphical user interface (GUI) displays each worker's real-time location and work status, forming a dynamic monitoring platform that facilitates safety management and scheduling.

[0056] In this embodiment, the specific steps of optimizing the cache delay of the omnidirectional mining area monitoring video and constructing the delay-optimized monitoring video are as follows:

[0057] Performing frame-by-frame video histogram brightness distribution recognition on the omnidirectional mining area monitoring video to obtain brightness distribution characteristics of each frame;

[0058] Calculating the global average brightness value of the omnidirectional mining area monitoring video;

[0059] Performing brightness deviation position detection on the brightness distribution characteristics of each frame according to the global average brightness value, and marking the brightness deviation area of ​​each frame;

[0060] Perform local brightness optimization on the brightness deviation area of ​​each frame to construct a brightness optimization monitoring video;

[0061] Identify abnormal noise points in brightness optimization monitoring videos and mark multiple video noise points;

[0062] Calculating the pixel value of the video noise;

[0063] Adaptively filter and denoise the pixel values ​​to obtain a filtered and optimized monitoring video;

[0064] Calculate the decoding delay of the filter optimization monitoring video and identify the video decoding delay parameters;

[0065] Calculate the maximum bit rate of the current network status;

[0066] Perform dynamic rate control optimization based on the maximum bit rate and video decoding delay parameters to generate a dynamically optimized bit rate;

[0067] Based on the dynamic optimization bit rate, cache delay optimization is performed on the filtered optimization monitoring video to construct a delay optimized monitoring video.

[0068] In this embodiment, the all-round mining area monitoring video is processed frame by frame to extract each frame of image. Video processing tools (such as OpenCV) can be used to extract frames to ensure that each frame can be analyzed separately. Set the video frame extraction frequency to 30 frames per second. If the video length is 10 minutes, a total of 1,800 frames of image are extracted. A brightness histogram is calculated for each frame of image to obtain the brightness distribution characteristics of each frame. The brightness histogram can reflect the number of pixels with different brightness values ​​in the image. Usually, the image is converted into a grayscale image for analysis. If the histogram of a frame shows that the brightness value is between 0 and 255, the calculation result shows that the number of pixels with a brightness value of 100 is 500 and the number of pixels with a brightness value of 150 is 300, then these features are recorded for subsequent analysis. Based on the brightness distribution characteristics of all frames, the global average brightness value is calculated. The global average brightness value can be obtained by adding the brightness value of each frame and dividing it by the total number of frames. If the total brightness value obtained from 1800 frames is 150,000, the global average brightness value is calculated as: global average brightness = 150,000 / 1800 = 83.33. The brightness distribution characteristics of each frame are compared with the global average brightness value to identify areas of brightness deviation. A threshold (such as ±10) can be set to determine whether there is a significant brightness deviation. If the average brightness of a frame is 75 and the deviation from the global average brightness of 83.33 exceeds 10, the frame is marked as a brightness deviation frame. To mark the brightness deviation area in each frame, image processing techniques (such as contour detection) can be used to identify the specific deviation area and draw a bounding box on the image. If the deviation area is a rectangular area in the upper left corner, a red border is drawn within the area to clearly mark the brightness deviation.

[0069] For areas of brightness deviation in each frame, a local brightness optimization algorithm is applied. Technologies such as histogram equalization or adaptive histogram equalization (CLAHE) can be used to optimize brightness. If the brightness of a deviation area is too low, CLAHE can be used to increase the brightness of that area to bring it closer to the global average brightness. The processed frames are reassembled to generate a brightness-optimized monitoring video. This ensures a more uniform brightness distribution in the optimized video, improving visualization. After optimization, the overall brightness distribution of the video is more balanced, facilitating subsequent analysis and monitoring. Image processing techniques (such as Gaussian filtering) are used to detect noise in the brightness-optimized monitoring video to identify abnormal noise points. Noise points typically appear as points with a significant brightness difference from the surrounding area. A brightness difference threshold is set. If the brightness difference between a pixel and the average of the surrounding pixels exceeds the threshold, the pixel is marked as a noise point. Detected noise points are marked and their pixel values ​​are recorded. This ensures that the problem area in the video can be quickly located. If the coordinates of a noise point are (500, 300) and its pixel value is 255, the "noise point coordinates, pixel value" are recorded.

[0070] Based on the pixel values ​​of identified noise points, an adaptive filtering algorithm (such as a median filter) is applied to denoise the image. This algorithm effectively removes isolated noise points while preserving edge information. If multiple noise points are detected within a certain area, a median filter is used to adjust the pixel values ​​in that area to more closely approximate the median value of the surrounding pixels. After denoising, the processed frames are reassembled to generate a filtered, optimized surveillance video. This ensures effective noise removal and improves video quality. After filtering, the video is visually clearer, facilitating subsequent monitoring and analysis. Decoding delay is calculated on the filtered, optimized surveillance video, measuring the time delay between the video signal input and the display output. This can be accurately measured by setting a timestamp. If the measured delay during decoding is 200 milliseconds, this parameter is recorded for subsequent analysis. The calculated decoding delay parameter is recorded in a database and associated with the video processing results for subsequent performance evaluation. The recording format is "video frame number, decoding delay time" to ensure data integrity. By monitoring network status in real time, the highest bitrate in the current environment is calculated. This can be performed using a network bandwidth test tool to ensure accurate network performance assessment. If the current network bandwidth test result is 10 Mbps, the maximum bit rate can be calculated to accommodate the transmission of the video stream. The maximum bit rate is recorded in the database for subsequent use.

[0071] Ensure that dynamic bitrate control accurately considers current network conditions. A dynamic bitrate control model is established based on the current network's maximum bitrate and decoding delay parameters. This model should be able to adjust the video stream's bitrate based on real-time network conditions to ensure smooth playback. An adaptive bitrate control policy is set to automatically reduce video quality to 720p when network bandwidth falls below 5 Mbps. The model calculates a dynamically optimized bitrate and applies it to video streaming, ensuring good video quality and smooth playback under varying network conditions. If network conditions are good, the bitrate is set to 5 Mbps; if conditions are poor, it is reduced to 2 Mbps. Based on the dynamically optimized bitrate, video cache latency is optimized to ensure an effective balance between video quality and smoothness during streaming. A larger cache is set when network conditions are good; a smaller cache is used to improve response speed when network conditions are unstable. All optimization steps are integrated to generate a final latency-optimized monitoring video, ensuring good visual quality and smooth playback. The resulting video adapts to varying network conditions, providing a stable monitoring experience.

[0072] In this embodiment, refer to Figure 3 , is a flowchart of the detailed implementation steps of step S2. In this embodiment, the detailed implementation steps of step S2 include:

[0073] Step S21: identifying key reference points in the mining area based on the time-delay optimized monitoring video to obtain multiple mining area feature points;

[0074] Step S22: performing a three-dimensional structural analysis of the mining area on multiple mining area feature points to generate a three-dimensional structural feature of the mining area;

[0075] Step S23: performing spatial topological analysis of the feature points based on the multiple mining area feature points to obtain a spatial topological relationship of the feature points;

[0076] Step S24: Perform multi-level three-dimensional modeling on the three-dimensional structural features of the mining area based on the spatial topological relationship of the feature points to construct a three-dimensional structural model of the mining area.

[0077] In this example, key reference points in a mining area are identified. These reference points are typically distinctive objects within the mining area, such as ore piles, equipment, and signage. Feature extraction is performed on each frame using the SIFT (Scale-Invariant Feature Transform) or ORB (Oriented FAST and Rotated BRIEF) algorithms to ensure accurate identification of key reference points. Suppose 10 key points are detected in a video, each located in a different mining area. Using a feature point matching algorithm (such as FLANN or BFMatcher), the identified feature points are matched to ensure consistent identification of the same feature across frames. The matching results are filtered, retaining key points with high matching scores and removing noise and false matches. A matching threshold of 0.75 is set; if a feature point's matching score across frames falls below this threshold, it is removed. Ultimately, 5 to 8 stable key reference points are retained. The coordinate information of the identified mining area feature points is recorded in a database, including both the image coordinates and the corresponding real-world coordinates, for subsequent analysis and modeling. Select an appropriate 3D reconstruction algorithm (such as structured light or stereo vision) to perform 3D structural analysis on the identified feature points. This algorithm should be able to infer 3D structure from 2D image data. Using stereo vision, depth information of feature points is calculated from multiple images. Combined with known camera parameters, a 3D point cloud is generated. The generated 3D point cloud is processed, including downsampling, noise removal, and point cloud registration. This ensures the accuracy and usability of the point cloud data, providing high-quality basic data for subsequent 3D modeling. Downsampling the point cloud using a Voxel Grid filter reduces the number of points to 50% of the original size, and point cloud registration is performed using the Iterative Closest Point (ICP) algorithm. 3D structural features of the mining area are generated from the point cloud data, typically including the overall shape of the mining area, the relative positions of feature points, and the main structures within the mining area. Topological relationships between feature points, including adjacency and connectivity, are defined. This step ensures a clear description of the relative positions and relationships of each feature point in 3D space. If a direct connection exists between feature point A and feature point B, it is recorded as "feature point A - feature point B," and the connection weight is set based on the spatial distance. Apply a topological analysis algorithm (such as Delaunay triangulation or Voronoi diagram) to perform topological analysis on feature points and generate a topological structure diagram between the feature points. Using the Delaunay triangulation method, connect the feature points into a triangular network to form a topological relationship diagram, ensuring that each feature point reflects its position in space through its connection relationship. Design a multi-level 3D modeling framework based on the spatial topological relationships of the feature points. This framework should be able to combine feature points and topological relationships to generate a clearly layered 3D model. Set the hierarchical relationship as "base layer (ground), first layer (equipment), second layer (ore pile)" to ensure a clear layered model structure.A multi-level 3D structural model is generated based on topological relationships and feature points. Using 3D modeling software (such as Blender or SketchUp), the feature points and topological relationships are combined to form a complete 3D model. The feature points and their connections are converted into 3D objects, and corresponding colors and materials are assigned to better reflect the actual mining area. The generated 3D structural model is verified to ensure that it accurately reflects the spatial layout and characteristics of the actual mining area. The model's accuracy and reliability are assessed by comparing it with field survey data. If the deviation between the feature points in the model and the field measurement data is less than 5%, the model is considered verified and appropriate optimization adjustments are made.

[0078] In this embodiment, refer to Figure 4 , is a flowchart of the detailed implementation steps of step S3. In this embodiment, the detailed implementation steps of step S3 include:

[0079] Step S31: calculating the spatial separation of worker working areas based on multiple worker position bounding boxes;

[0080] Step S32: dynamically deconstruct the spatial intervals of the workers' work areas to construct a dynamic work distribution map of the workers;

[0081] Step S33: marking the real-time positions of workers on the three-dimensional structure model of the mining area according to the dynamic work distribution map of workers, and obtaining a real-time worker distribution mining area structure model;

[0082] Step S34: performing real-time dynamic tracking on the marked bounding boxes of multiple workers to obtain the real-time movement trajectory of each worker;

[0083] Step S35: Based on the real-time movement trajectory of each worker, the real-time worker distribution mining area structure model is instantly mapped to construct a real-time mining area twin model.

[0084] In this embodiment, the position bounding box data of multiple workers are obtained from the monitoring system. These bounding boxes should contain the position coordinates of each worker and the size information of the bounding box. The bounding boxes are usually calibrated by deep visual recognition technology. Assume that the bounding boxes of 5 workers are detected, namely A (200, 300, 50, 70), B (250, 320, 50, 70), etc., and the coordinates and size of the bounding box of each worker are recorded. According to the coordinates of the worker's bounding box, the spatial interval between each worker's working area is calculated. The Euclidean distance formula can be used to calculate the distance between two workers to determine whether their working areas overlap. If the center point of worker A is (225, 335) and the center point of worker B is (275, 355), the spatial interval is calculated as: Pixels are analyzed using a dynamic distribution deconstruction algorithm. Based on the spatial interval data of worker work areas, worker work patterns within the mine are analyzed. Clustering algorithms (such as K-means or DBSCAN) can be used to identify worker work distribution. Set the clustering algorithm parameters. For example, we choose the K-means algorithm and set the number of clusters to 3 to identify the primary work areas within the mine. Based on the clustering results, generate a dynamic worker work distribution map. This map should visually display each worker's work area within the mine and its dynamic changes. The worker clustering results show that workers are distributed in three primary areas, each marked with a different color. The generated work distribution map clearly illustrates hotspots of worker activity. Save the generated dynamic worker work distribution map and record information such as the number of workers and activity frequency in each area for subsequent analysis and decision-making. The record format is "area number, number of workers, activity frequency" to ensure information integrity. Obtain the constructed 3D structural model of the mine area and ensure that it can be integrated with the dynamic worker work distribution map. The 3D model should include information such as the mine's topography, equipment, and major routes. Feature points in the mine model include ore piles, equipment locations, and main access roads, ensuring they can be mapped to the dynamic positions of workers. Based on the workers' real-time locations and the dynamic work distribution map, the 3D structural model of the mine is updated, marking the workers' specific locations within the model. Real-time marking is achieved by mapping the workers' coordinates to the 3D model's coordinate system. If Worker A's real-time location is (220, 340, 0) meters, the worker's location marker is updated in the 3D model to ensure that their work status within the mine is reflected in real time. The updated real-time worker distribution mine structural model is verified to ensure the accuracy of the worker's location markers. The accuracy of the markers is assessed by comparing them with real-time data from the monitoring system. If the monitoring system confirms that the error between Worker A's actual position and the marked position in the model is less than 5%, the update is recorded as successful and the updated model is saved. The bounding boxes of multiple workers are dynamically tracked in real time to record each worker's movement within the mine. Trajectory tracking can be achieved using techniques such as Kalman filtering or optical flow. If a worker moves from position (220, 340) to (230, 350) in the monitoring video, this dynamic trajectory is recorded. Real-time movement trajectory data is stored in a database to ensure that each worker's activity history can be tracked. Trajectory data should include information such as timestamps and location coordinates. The recording format is "worker number, timestamp, current location" to facilitate subsequent analysis and visualization. Visualize the worker's movement trajectory to vividly reflect the worker's movement path within the mine area. This can be achieved by drawing trajectory lines on the three-dimensional model of the mine area. Based on the movement trajectory of worker A, a line segment connecting his various positions is drawn on the model to intuitively display his work path. Based on each worker's real-time movement trajectory, the real-time worker distribution mine structure model is instantly mapped. Ensure that the model can reflect the dynamic changes of workers in real time.If worker A moves 10 meters during the monitoring period, their position is updated in the model to ensure that their activity status is reflected in real time. The real-time worker distribution is combined with the three-dimensional structural model of the mine to construct a real-time mine twin model. This model should be able to reflect the distribution and activity of workers at different time periods. The generated mine twin model can display the real-time location and movement trajectory of workers and their relationship with the mine environment. The generated real-time mine twin model is verified to ensure that it accurately reflects the actual situation in the mine. The accuracy of the model is evaluated by comparing it with field data. If the error between the real-time model and the actual worker location is less than 5%, it is marked as verified successfully and corresponding optimization is performed.

[0085] In this embodiment, step S4 includes the following steps:

[0086] Acquire multi-dimensional environmental monitoring parameters of mining areas based on multiple sensors;

[0087] Calculating oxygen content changes based on the multi-dimensional environmental monitoring parameters to obtain a real-time oxygen content curve;

[0088] Performing toxic gas detection on the multi-dimensional environmental monitoring parameters and extracting toxic gas parameters;

[0089] Identify the gas type of toxic gas parameters and perform gas concentration fluctuation fitting to generate concentration change diagrams of multiple types of toxic gases;

[0090] Air safety risk inference is carried out based on the concentration change graphs of multiple types of toxic gases and the real-time oxygen content curves, and dangerous gas areas are located and marked in the real-time mining area twin model.

[0091] In this embodiment, a variety of sensors are strategically deployed within the mining area, including oxygen sensors, toxic gas sensors (such as carbon monoxide, methane, and hydrogen sulfide), temperature and humidity sensors, and air pressure sensors. This ensures that sensors cover key areas within the mining area to monitor environmental changes. Assuming 10 oxygen sensors and 5 toxic gas sensors are deployed, these sensors should possess high precision and rapid response capabilities. The oxygen sensor measurement range is 0-30% (volume ratio) with an accuracy of ±0.1%. A data acquisition system is configured to regularly read sensor data. The data acquisition frequency is set to once per second to ensure timely monitoring of environmental changes. The system sends data from each sensor to a central control system in the format of "timestamp, sensor type, and measurement value," ensuring the use of an efficient communication protocol (such as LoRa or Zigbee). Based on the real-time collected oxygen concentration data, changes in oxygen content are calculated and a real-time oxygen content curve is generated. Data smoothing techniques (such as moving average) can be used to reduce the impact of transient fluctuations. If the oxygen concentration data collected over a 5-minute period is 20.5%, 20.4%, 20.6%, 20.3%, and 20.5%, calculate a moving average and generate curve data points. Plot the calculated oxygen content data into a graph to display the changing trend of oxygen concentration. Ensure that the graph clearly reflects the real-time changes in oxygen concentration. If the graph shows that the oxygen concentration has fluctuated between 20.3% and 20.6% over the past 10 minutes, record this fluctuation and set an alarm threshold (e.g., below 20.2%) to facilitate subsequent risk warnings. Perform real-time detection of toxic gases based on multi-dimensional environmental monitoring parameters. Use toxic gas sensors to monitor the concentrations of gases such as carbon monoxide, methane, and hydrogen sulfide. Set the carbon monoxide sensor's monitoring range to 0-1000 ppm with an accuracy of ±5 ppm. If the real-time carbon monoxide concentration reaches 150 ppm, record this value. Extract and store the collected toxic gas parameters to ensure rapid access and analysis. Record the type and concentration of the toxic gas for subsequent analysis. Use a gas type recognition algorithm (such as a machine learning-based classifier) ​​to analyze real-time monitored gas parameters and identify the gas type. Using a trained model, if the real-time monitored gas concentration matches known characteristics, it is identified as carbon monoxide, methane, etc. Perform concentration fluctuation fitting on the extracted toxic gas parameters, using methods such as linear regression or polynomial fitting, to generate a graph showing the concentration changes of various toxic gases. If the methane concentration data monitored over the past 60 minutes is 30 ppm, 35 ppm, 32 ppm, and 40 ppm, a methane concentration curve is generated through fitting. The fitting results are plotted as a concentration change graph to display the concentration fluctuations of different toxic gases. Ensure that the graph clearly reflects the trend of gas concentration changes. If the graph shows a gradual increase in methane concentration over the past hour, attention should be paid to the risk of exceeding safety thresholds.An air safety risk assessment model was established based on oxygen content curves and toxic gas concentration graphs. This model should consider the interplay between oxygen and toxic gas concentrations to infer air safety risks. Based on the risk assessment results, potentially hazardous gas areas within the mine are located and marked. If the risk score exceeds a set threshold (e.g., 0.7), the area is marked as hazardous. If the oxygen concentration in a particular area is below 20.2% and the methane concentration exceeds 40 ppm, the area is marked as "high risk." The results of locating hazardous gas areas are applied to a real-time mine twin model to ensure that the model reflects air safety risks within the mine in real time. Hazardous gas areas are highlighted in red in the 3D mine model, and a report is generated to facilitate decision-making by management.

[0092] In this embodiment, step S5 includes the following steps:

[0093] Perform terrain morphology calculation on the real-time mining area twin model to generate terrain morphology parameters;

[0094] Perform terrain semantic analysis based on terrain morphological parameters to obtain terrain semantic features;

[0095] Infer dangerous terrain based on the semantic features of terrain appearance and obtain dangerous terrain areas;

[0096] Based on dangerous terrain areas and dangerous gas areas, the real-time mining area twin model is highlighted and visualized to obtain a warning map of the dangerous area.

[0097] In this embodiment, data from a real-time twin model of the mining area is acquired. This model typically includes 3D structural information about the mining area, including features such as topography, equipment locations, and ore piles. This ensures that the model reflects the latest status of the mining area. The model may also include elevation data for the mining area, ensuring that the 3D coordinates of each point accurately describe the topographic changes within the mining area. Based on the 3D model of the mining area, an algorithm is used to calculate terrain morphological parameters, including slope, undulation, concavity, and elevation change. This calculation can be performed using digital elevation model (DEM) technology. For example, if the elevation range of a certain area in the mining area is 100 meters, the slope variation can be calculated to yield a slope of 15° and an undulation of 0.3 meters. All calculated parameters are recorded for subsequent analysis. The calculated terrain morphological parameters are stored in a database to ensure quick access to this data for subsequent analysis. The record format is "area number, slope, undulation, elevation change," etc. For example, if an area has a slope of 10° and an undulation of 0.5 meters, it would be recorded as "area A, slope 10°, undulation 0.5 meters." Select an appropriate semantic analysis method to analyze terrain morphological parameters to extract semantic features of the terrain. Machine learning or deep learning models (such as convolutional neural networks) can be used for feature extraction. Use the trained model to analyze terrain elevation data and identify different terrain features (such as ridges, valleys, and flat areas). Organize the analysis results to extract semantic features of the terrain, including terrain type, terrain variation patterns, and corresponding risk characteristics. Identify an area as steep slope and label it as "high-risk terrain." If an area's features display as "steep slope," record it as "Area B, Type: Steep Slope, Risk Level: High." Based on the extracted semantic features of the terrain, establish a dangerous terrain inference model. This model should comprehensively consider slope, terrain type, and its corresponding risk level to infer potentially dangerous terrain areas. Set a threshold rule: if the slope is greater than 30° and the terrain type is "steep slope," label it as "high-risk terrain." Apply the inference model to analyze the mining area's terrain and identify dangerous terrain areas. Using computer algorithms, automatically label areas with potential risks. If a slope of 35° and a "steep slope" type is detected in a certain area, the area will be marked as a "hazardous terrain area". Integrate the previously identified hazardous gas areas with the hazardous terrain areas to form a comprehensive hazardous area dataset. Ensure that both gas risks and terrain risks are reflected. If a hazardous gas area intersects with a hazardous terrain area, it will be marked as a "high-risk area". Based on the comprehensive hazardous area data, the real-time mine twin model is highlighted and visually rendered. Use three-dimensional visualization tools to ensure that hazardous areas can be intuitively presented to management personnel. Highlight all hazardous terrain and hazardous gas areas in red to ensure that they are easy to identify on the map. Generate the final hazardous area warning map and display it in real time in the mine monitoring system.Ensure managers have timely access to hazardous area information so they can take appropriate safety measures. Create dynamic dashboards that update the status of hazardous areas in real time, ensuring that warning maps always reflect the latest safety conditions in the mine.

[0098] In this embodiment, the specific steps of step S6 are:

[0099] Perform real-time worker behavior trajectory prediction on the real-time mine twin model to obtain the predicted behavior trajectory of each worker;

[0100] Based on the danger zone warning map, the predicted trajectory of each worker's behavior is judged to determine the area of ​​intrusion, marking the workers who enter the potential danger zone;

[0101] Make real-time early warning decisions based on potential workers entering dangerous areas and build a dangerous area intrusion prediction strategy;

[0102] Perform dynamic holographic visualization of dangerous area intrusion prediction strategies and real-time mining area twin models, and build a holographic visualization model of mining area information.

[0103] In this example, historical worker movement trajectory data is extracted from the real-time mine twin model. This data should include information such as the worker's location coordinates, timestamp, and movement speed to ensure that it reflects the worker's behavior patterns. The location data of each worker over the past hour is collected. Assume that worker A's movement trajectory is (220, 340), (225, 345), and (230, 350). This data is recorded for subsequent analysis. A worker behavior prediction model is constructed using machine learning or deep learning techniques. Time series prediction algorithms such as LSTM (Long Short-Term Memory) can be used to capture the temporal dependencies of worker movement. The model is trained to predict future movement trajectories by inputting past worker movement trajectory data. Assume that the model training predicts that worker A's coordinates at the next moment will be (235, 355). Based on the output of the prediction model, a predicted behavior trajectory for each worker is generated. These trajectories should reflect the worker's likely movement path over the next period of time. If worker A is predicted to move to (235, 355), (240, 360), and (245, 365) within the next 5 minutes, these predicted locations are recorded to form a complete behavioral trajectory. A previously generated danger zone warning map is obtained. This map should clearly identify dangerous terrain and hazardous gas areas within the mine. This ensures data timeliness and accuracy. Suppose the map shows that area B has a slope of 35° and excessive methane concentration, marking it as a "high-risk area." Each worker's predicted trajectory is collided with the hazardous area to determine whether the worker has entered the hazardous area. Geometric methods can be used to detect whether a trajectory point intersects the hazardous area. If worker A's predicted trajectory point (240, 360) intersects hazardous area B, the point is marked as a potential hazardous area entry. Detected potential hazardous area entries are recorded in the database for subsequent warning decisions. Ensure that the record includes worker ID, predicted location, hazardous area information, and more. The record format is "worker ID, predicted location, hazardous area, timestamp" to ensure data integrity and traceability. Develop a real-time early warning decision-making strategy based on potential entries into dangerous areas. Ensure that the strategy can quickly respond to workers entering dangerous areas and initiate appropriate safety measures. Set an alarm to trigger and notify relevant management personnel if a worker enters a dangerous area within a certain period of time. Configure a real-time monitoring system to continuously monitor the relationship between workers' behavioral trajectories and dangerous areas. Once a worker is detected entering a dangerous area, immediately execute the early warning strategy. If worker A enters the dangerous area during the predicted trajectory, an alarm is automatically sent to the mine management center, and the information is highlighted in the monitoring system. Record the execution results of the early warning decision in the database for subsequent analysis and optimization. Ensure that all early warning events are traceable. Select appropriate holographic visualization technology to display real-time information and dangerous areas in the mine area. Augmented reality (AR) or virtual reality (VR) technology can be used to achieve more intuitive visualization effects.Using AR technology, a 3D model of the mine site is overlaid with real-time data, ensuring managers can intuitively visualize worker locations and hazardous areas. A dynamic holographic visualization model is constructed based on the real-time mine site twin model, predicted worker behavior trajectories, and hazardous area information. This model should be able to update in real time to display the mine's safety status. The model uses different colors to identify hazardous areas, and dynamic lines to illustrate worker movement trajectories, ensuring real-time and accurate information.

[0104] In this embodiment, a coal mine information visualization system is provided, which is used to execute the above-mentioned coal mine information visualization method, including:

[0105] The worker vision calibration module is used to obtain omnidirectional mining area monitoring videos; it also performs cache latency optimization and worker vision calibration to extract multiple worker position bounding boxes;

[0106] The 3D structure module is used to identify key reference points in the mining area based on the omnidirectional mining area monitoring video, and to perform 3D structure modeling to construct a 3D structural model of the mining area;

[0107] The twin model module is used to perform real-time dynamic tracking based on the bounding boxes of multiple worker positions and to perform instant mobile mapping of the 3D structural model of the mine area to build a real-time twin model of the mine area;

[0108] The toxic gas sensing module is used to obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting, locate hazardous gas areas, and mark hazardous gas areas;

[0109] The dangerous area rendering module is used to infer dangerous terrain based on the real-time mining twin model, and then highlight and visualize the dangerous gas areas to obtain a warning map of the dangerous area;

[0110] The real-time warning module is used to make real-time warning decisions and perform dynamic holographic visualization based on the warning map of dangerous areas, and to build a holographic visualization model of mining area information.

[0111] This invention accurately identifies worker locations, ensuring monitoring of each worker. In emergencies, their specific locations can be quickly determined, reducing the risk of workplace accidents. By optimizing video processing buffering and latency, worker location bounding boxes can be updated in real time, improving system response speed, reducing latency, and ensuring smooth dynamic tracking. 3D modeling visually displays the spatial structure of the mine, including complex underground passages and mine facilities, helping managers understand the mine's layout. The 3D model of the mine can serve as a foundation for future design and planning, improving resource management efficiency and optimizing the layout of mine facilities. Real-time modeling of the mine's 3D structure allows managers to identify structural issues (such as tunnel collapse) and implement timely maintenance or repairs. The twin model updates worker locations in real time and reflects them in the 3D model, providing mine managers with accurate real-time data and ensuring they are always aware of personnel movements within the mine. Through real-time dynamic tracking, the twin model can help predict potential dangerous situations, enabling proactive prevention and emergency response plans. Real-time tracking of worker locations helps managers optimize work schedules and avoid congestion in high-risk areas. Real-time gas concentration analysis enables early warning and measures to prevent toxic gas leaks from causing disasters such as fires or explosions. Accurately fitting toxic gas concentration fluctuations helps locate risky areas, providing a basis for personnel evacuation and evacuation. 3D rendering allows all hazardous areas to be visually displayed within the mine twin model, significantly improving mine managers' ability to identify potential risks. Real-time marking of hazardous areas within the mine facilitates emergency management, allowing workers to quickly receive instructions and take swift emergency response measures. Rendered hazardous area maps enable workers to quickly identify and avoid hazardous areas, improving mine safety. Fusion analysis of real-time data allows mines to rapidly respond to potential safety incidents, issuing early warnings and reducing the likelihood of accidents. Dynamic holographic visualization provides mine managers with a clearer view of hazardous areas and the overall mine safety status, enabling quick and accurate decision-making. The holographic visualization model of the mine allows simulation of various hazards within the mine, further enhancing safety and control capabilities.

[0112] The present invention is therefore intended to be illustrative and non-restrictive in all respects, with the scope of the invention being defined by the appended claims rather than the foregoing description, and all changes that come within the meaning and range of equivalents of the application documents are intended to be embraced therein.

[0113] The foregoing description is intended only to provide specific embodiments of the present invention, which will enable those skilled in the art to understand and implement the present invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention is not intended to be limited to the embodiments shown herein, but is to be construed in the widest possible manner consistent with the principles and novel features disclosed herein.

Claims

1. A method for visualizing coal mine information, characterized in that: The following steps are involved: Step S1: Obtain all-round mining area monitoring video; It also performs cache latency optimization and worker visual calibration to extract multiple worker position bounding boxes; Step S2: identifying key reference points in the mining area based on the all-round mining area monitoring video, and performing three-dimensional structural modeling to construct a three-dimensional structural model of the mining area; Step S3: Real-time dynamic tracking is performed based on the bounding boxes of multiple worker positions, and real-time mobile mapping of the mining area 3D structure model is performed to build a real-time mining area twin model; Step S4: Obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting and dangerous gas area positioning, and mark the dangerous gas area; Step S5: Inferring dangerous terrain based on the real-time mining twin model, and then highlighting and visualizing the dangerous gas area to obtain a warning map of the dangerous area; Step S6: Based on the warning map of the dangerous area, a real-time early warning decision is made and dynamic holographic visualization is performed to construct a holographic visualization model of the mining area information; Among them, the specific steps of step S1 are: Obtain all-round mining area monitoring videos based on multiple high-definition cameras in the mining area; Optimizing cache latency of the all-around mining area monitoring video to construct a latency-optimized monitoring video; Perform deep visual inspection on the latency optimization monitoring video and mark each worker node; Calculate the real-time position of the worker nodes to generate the real-time position coordinates of each worker; Perform image bounding box calibration on the latency-optimized monitoring video based on the real-time location coordinates of each worker, extracting multiple worker location bounding boxes. The specific steps of optimizing the cache delay of the all-round mining area monitoring video and constructing the delay-optimized monitoring video are as follows: Performing frame-by-frame video histogram brightness distribution recognition on the omnidirectional mining area monitoring video to obtain brightness distribution characteristics of each frame; Calculating the global average brightness value of the omnidirectional mining area monitoring video; Performing brightness deviation position detection on the brightness distribution characteristics of each frame according to the global average brightness value, and marking the brightness deviation area of ​​each frame; Perform local brightness optimization on the brightness deviation area of ​​each frame to construct a brightness optimization monitoring video; Identify abnormal noise points in brightness optimization monitoring videos and mark multiple video noise points; Calculating the pixel value of the video noise; Adaptively filter and denoise the pixel values ​​to obtain a filtered and optimized monitoring video; Calculate the decoding delay of the filter optimization monitoring video and identify the video decoding delay parameters; Calculate the maximum bit rate of the current network status; Perform dynamic rate control optimization based on the maximum bit rate and video decoding delay parameters to generate a dynamically optimized bit rate; Optimize the cache delay of the filtered optimized monitoring video based on the dynamic optimized bit rate to construct a delay optimized monitoring video; The specific steps of step S5 are: Perform terrain morphology calculation on the real-time mining area twin model to generate terrain morphology parameters; Perform terrain semantic analysis based on terrain morphological parameters to obtain terrain semantic features; Infer dangerous terrain based on the semantic features of terrain appearance and obtain dangerous terrain areas; Highlight and visualize the real-time mining twin model based on dangerous terrain and hazardous gas areas to obtain a warning map of dangerous areas; The specific steps of step S6 are: Perform real-time worker behavior trajectory prediction on the real-time mine twin model to obtain the predicted behavior trajectory of each worker; Based on the danger zone warning map, the predicted trajectory of each worker's behavior is judged to determine the area of ​​intrusion, marking the workers who enter the potential danger zone; Make real-time early warning decisions based on potential workers entering dangerous areas and build a dangerous area intrusion prediction strategy; Perform dynamic holographic visualization of dangerous area intrusion prediction strategies and real-time mining area twin models, and build a holographic visualization model of mining area information.

2. The method for visualizing coal mine information according to claim 1, characterized in that: The specific steps of step S2 are: Identify key reference points in the mining area using time-delay optimized monitoring videos to obtain multiple mining area feature points; Conduct three-dimensional structural analysis of multiple mining area feature points to generate three-dimensional structural features of the mining area; Perform spatial topological analysis of feature points based on multiple mining area feature points to obtain the spatial topological relationship of feature points; Based on the spatial topological relationship of characteristic points, multi-level 3D modeling is performed on the 3D structural features of the mining area to construct a 3D structural model of the mining area.

3. The method for visualizing coal mine information according to claim 1, characterized in that: The specific steps of step S3 are: Calculating the spatial separation of worker working areas based on multiple worker position bounding boxes; Dynamically deconstruct the spatial distribution of workers' work areas and construct a dynamic work distribution map of workers; Mark the real-time positions of workers on the three-dimensional structure model of the mining area according to the dynamic work distribution map of workers, and obtain the real-time worker distribution mining area structure model; Perform real-time dynamic tracking of multiple workers’ marked bounding boxes to obtain the real-time movement trajectory of each worker; Based on the real-time movement trajectory of each worker, the real-time worker distribution mining area structure model is instantly mapped to build a real-time mining area twin model.

4. The method for visualizing coal mine information according to claim 1, characterized in that: The specific steps of step S4 are: Acquire multi-dimensional environmental monitoring parameters of mining areas based on multiple sensors; Calculating oxygen content changes based on the multi-dimensional environmental monitoring parameters to obtain a real-time oxygen content curve; Performing toxic gas detection on the multi-dimensional environmental monitoring parameters and extracting toxic gas parameters; Identify the gas type of toxic gas parameters and perform gas concentration fluctuation fitting to generate concentration change diagrams of multiple types of toxic gases; Air safety risk inference is carried out based on the concentration change graphs of multiple types of toxic gases and the real-time oxygen content curves, and dangerous gas areas are located and marked in the real-time mining area twin model.

5. A coal mine information visualization system, characterized in that: The method for visualizing coal mine information according to claim 1 comprises: The worker vision calibration module is used to obtain omnidirectional mining area monitoring videos; it also performs cache latency optimization and worker vision calibration to extract multiple worker position bounding boxes; The 3D structure module is used to identify key reference points in the mining area based on the omnidirectional mining area monitoring video, and to perform 3D structure modeling to construct a 3D structural model of the mining area; The twin model module is used to perform real-time dynamic tracking based on the bounding boxes of multiple worker positions and to perform instant mobile mapping of the 3D structural model of the mine area to build a real-time twin model of the mine area; The toxic gas sensing module is used to obtain multi-dimensional environmental monitoring parameters of the mining area, perform toxic gas concentration fluctuation fitting, locate hazardous gas areas, and mark hazardous gas areas; The dangerous area rendering module is used to infer dangerous terrain based on the real-time mining twin model, and then highlight and visualize the dangerous gas areas to obtain a warning map of the dangerous area; The real-time warning module is used to make real-time warning decisions and perform dynamic holographic visualization based on the warning map of dangerous areas, and to build a holographic visualization model of mining area information.

Images