Mine accident intelligent detection collaborative management early warning method
By combining multi-source collaborative inspection with UAV intelligent perception, and integrating multimodal data fusion and intelligent interception network sensors, the shortcomings of multi-source information fusion and dynamic response in mine safety early warning have been solved. This has enabled real-time monitoring and efficient emergency response of mine accidents, and improved the real-time and systematic nature of mine safety management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- GENERAL PROSPECTING INSTITUTE OF CHINA NATIONAL ADMINISTRATION OF COAL GEOLOGY
- Filing Date
- 2025-12-30
- Publication Date
- 2026-05-12
AI Technical Summary
Existing mine safety early warning technologies lack the ability to integrate and dynamically respond to multi-source information, especially in the real-time monitoring and emergency response to accidents such as slope instability and rockfalls. They are insufficient to achieve comprehensive and high-precision risk identification and location, and lack effective on-site visualization guidance and collaborative emergency response mechanisms after the early warning is issued.
By combining multi-source collaborative inspection task planning with UAV intelligent perception, multi-modal data fusion analysis and intelligent interception network sensor network, real-time monitoring and dynamic response to mine accidents can be achieved. UAVs can be used for precise aerial verification and light warning zone delineation to build a three-dimensional collaborative emergency response system.
It significantly improves the timeliness and accuracy of mine safety inspections, enables real-time monitoring and precise location of sudden risks, enhances the efficiency and reliability of on-site safety control, and ensures closed-loop management of the entire process from early warning to response.
Smart Images
Figure CN122014347A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the technical field of mine accident early warning methods, specifically a mine accident intelligent detection, collaborative management, and early warning method. Background Technology
[0002] Existing mine safety early warning technologies mostly focus on single monitoring methods or local risk identification, making it difficult to achieve multi-source information synergy and dynamic closed-loop response.
[0003] In the prior art, Chinese patent CN117036082A discloses a smart mine management system and method. This system collects mine environmental information and personnel physiological information, extracts features based on DS evidence theory, and uses a dual neural network model to collaboratively calculate accident probabilities to achieve graded early warning. Although this technology introduces data fusion and model collaboration mechanisms, its early warning process still mainly relies on a fixed-deployment ground sensor network, lacking real-time three-dimensional perception and rapid response capabilities for sudden and localized risks, especially in the monitoring and emergency response to geological disasters such as slope instability and rockfalls.
[0004] In actual mining operations, accidents such as slope instability, landslides, and rockfalls are characterized by their suddenness, wide impact, and high risk. Ground sensors alone are insufficient for comprehensive and high-precision risk identification and location. Furthermore, existing systems lack effective on-site visualization guidance and collaborative emergency response mechanisms after warnings are issued. Especially at night or in low-visibility conditions, personnel evacuation and on-site control are inefficient, making it difficult to achieve closed-loop management from warning to response.
[0005] Therefore, there is a need for a method for intelligent detection, collaborative management, and early warning of mine accidents that can integrate aerial and ground monitoring resources, achieve multimodal data fusion analysis, and possess intelligent triggering and dynamic response capabilities, in order to improve the real-time performance, accuracy, and systematic nature of mine safety management. Summary of the Invention
[0006] The purpose of this invention is to provide a method for intelligent detection, collaborative management and early warning of mine accidents, so as to solve the problems mentioned in the background art.
[0007] To achieve the above objectives, the present invention provides the following technical solution: a method for intelligent detection, collaborative management, and early warning of mine accidents, specifically comprising the following steps: Step 1: Multi-source collaborative inspection task planning and triggering; Step 2: Simultaneous acquisition of intelligent sensing and multimodal data by UAV; Step 3: Receive all data from this inspection, and the system will automatically link data from the same time period and region for multi-source data fusion analysis and risk assessment. Step 4: Tiered early warning issuance and collaborative management response; Step 5: The net sensing layer is triggered; Step Six: Precise Aerial Verification and Change Detection; Step 7: Dynamic risk assessment and delineation of light warning zones; Step 8: Collaborative emergency response and closed-loop management.
[0008] Preferably, in step one, the collaborative inspection task planning generates daily and periodic inspection plans based on safety management procedures. When ground micro-seismic sensors detect abnormal vibrations, slope radar detects small displacements exceeding the threshold, gas concentration monitoring point data is abnormal, or a hidden danger is reported manually, the data center control system automatically obtains meteorological information and adjusts the special inspection based on the meteorological information.
[0009] Preferably, in step two, the UAV flies automatically along a preset route and uses an onboard millimeter-wave radar and vision system to achieve obstacle avoidance at night. The multimodal data synchronous acquisition in step two includes: high-definition night vision video stream, infrared thermal image data and lidar point cloud.
[0010] Preferably, in step three, the information receiving process involves the UAV detecting an abnormal temperature in the area using infrared sensors, then retrieving historical data from fixed temperature sensors near that area to observe temperature trends. The UAV imagery reveals new cracks on the slope, which are compared with the displacement increment from the lidar point cloud and slope monitoring radar data to verify the severity and activity of the cracks. Visible and infrared images are analyzed in depth, and a recognition model identifies deficiencies in protective facilities. A point cloud change detection model automatically compares the current and previous scanned point clouds to generate a high-precision differential model, identifying areas where settlement, landslides, and material accumulation changes exceed limits. A comprehensive risk assessment model combines the identified hazard characteristics with environmental and production activity data, inputting them into the risk assessment model to output the risk level and probability.
[0011] Preferably, after the risk assessment is completed in step three, an analysis report is generated, and the early warning rule engine is activated according to preset rules. In step four, when a high warning occurs, a strong reminder is triggered simultaneously on the command center's large screen, in the relevant area broadcast, and on the responsible person's mobile APP. When a low warning occurs, the warning information is automatically associated with the details of the hidden danger, the on-site video stream, and the emergency response plan, and maintenance and review work orders are automatically generated and dispatched to the mobile terminals of the responsible personnel.
[0012] Preferably, in step five, an intelligent interception net is deployed below the potential instability area of the slope. Multi-mode sensor nodes are arranged in a matrix on the interception net. The sensors integrate stress, vibration, tilt angle sensing and wireless transmission modules. The sensor network is completed and the signal is connected to the early warning center. The unique ID and precise coordinates of each sensor have been marked in the digital twin model.
[0013] Preferably, in step five, during normal operation, the sensors periodically report status data to form a background baseline. When the interception net is impacted by landslides and falling rocks, a group of sensors simultaneously trigger an over-threshold alarm, but communication is still maintained. Based on the spatiotemporal relationship of the sensor signal sequence, the early warning center uses triangulation or shock wave propagation algorithms to automatically delineate the suspected impact source area on the map. When some sensor signals suddenly and completely fail, the early warning center directly identifies the physical area where the sensor that lost the signal is located as a direct danger area. In both cases, the system automatically generates the highest priority UAV emergency verification task.
[0014] Preferably, in step six, when the drone receives an emergency alarm from the intelligent interception network, the drone no longer follows the regular flight path, but flies directly to the alarm area and lowers its altitude to conduct close reconnaissance. It takes high-definition photos, performs thermal imaging scans, and rapid LiDAR scans of the suspected impact source area and the directly dangerous area from multiple angles. The images are transmitted back to the command center in real time, and AI assists in analyzing the scale, boundaries, stability, and threat level of the landslide.
[0015] Preferably, in step seven, the early warning center integrates the interception network data and the drone verification data to dynamically generate three layers of areas on the 3D map: a red high-risk area, a yellow warning area, and a green safe assembly area. The system automatically dispatches a dedicated warning drone to the site airspace. Based on the received regional geographical coordinates, the drone uses a strong light projector to project clear red and yellow danger zone boundaries onto the ground from the air, and projects a green assembly marker in the green area.
[0016] Preferably, in step eight, the system automatically locks the personnel location cards in the red and yellow areas, sends a forced evacuation alarm to their handheld terminals, and can automatically control the power outage of the work equipment in the area. The command center directs the rescue personnel to the green safe assembly area on the three-dimensional situation map and plans a safe reconnaissance or disposal route. The drone continuously monitors the site, tracks the personnel evacuation and slope stability. Every step of the disposal process is recorded in the system. After the event, all data is packaged into a case study for optimizing the alarm threshold of the interception network, the AI recognition model, and the emergency response plan.
[0017] Compared with the prior art, the beneficial effects of the present invention are: 1. This application achieves automation and intelligence in mine safety inspections through multi-source collaborative inspection task planning and intelligent triggering mechanisms. The system can generate daily and periodic inspection plans based on safety management procedures, and automatically trigger special inspection tasks when anomalies are detected by equipment such as micro-seismic sensors, slope radar, and gas monitoring, or when potential hazards are manually reported. Simultaneously, the system can dynamically adjust inspection strategies based on real-time meteorological information, significantly improving the targeting and timeliness of inspections, ensuring that potential hazards are detected and responded to promptly.
[0018] 2. This application constructs a ground-based perception layer by deploying an intelligent interception network and a multi-mode sensor network, enabling real-time monitoring and precise location of sudden events such as rockfalls and landslides on slopes. The interception network sensors periodically report status data to form a baseline and trigger alarms when impacted. The system uses signal sequences for spatiotemporal analysis to automatically delineate dangerous areas and generate UAV emergency verification tasks, forming a three-dimensional monitoring and emergency response system that significantly improves the speed and accuracy of emergency response.
[0019] 3. This application achieves three-dimensional and visual identification of on-site risk areas through dynamic risk assessment and light warning zone delineation. The system integrates ground and aerial data to dynamically generate red, yellow, and green risk zones on a 3D map, and deploys dedicated warning drones to project corresponding colored light spots and assembly markers onto the ground. Especially at night or in low-visibility conditions, this method can intuitively and clearly guide personnel evacuation and assembly, greatly improving the efficiency and reliability of on-site safety control. Attached Figure Description
[0020] Figure 1 This is a schematic diagram of the overall process of the present invention. Detailed Implementation
[0021] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0022] Example: Figure 1 As shown, the present invention provides a technical solution for a method for intelligent detection, collaborative management, and early warning of mine accidents, which specifically includes the following steps: Step 1: Multi-source collaborative inspection task planning and triggering. In Step 1, the collaborative inspection task planning generates daily and periodic inspection plans based on the safety management procedures. Ground micro-seismic sensors detect abnormal vibrations, slope radar detects small displacements exceeding the threshold, gas concentration monitoring point data is abnormal, and manual reports of potential hazards are received. The data center control system automatically obtains meteorological information and adjusts the special inspection based on the meteorological information. Step 2: Intelligent perception and simultaneous collection of multimodal data by UAV. In step 2, the UAV flies automatically along a preset route and uses onboard millimeter-wave radar and vision system to achieve obstacle avoidance at night. The simultaneous collection of multimodal data in step 2 includes: high-definition night vision video stream, infrared thermal image data and lidar point cloud. Step 3: Receive all data from this inspection. Simultaneously, the system automatically associates data from data sources within the same time period and area, performing multi-source data fusion analysis and risk assessment. In Step 3, the information reception process involves the UAV infrared sensor detecting abnormal temperatures in the area, then retrieving historical data from nearby fixed temperature sensors to examine temperature trends. The UAV imagery reveals new cracks on the slope, which are compared with displacement increments from the lidar point cloud and slope monitoring radar data to verify the severity and activity of the cracks. In-depth analysis of visible light and infrared images is performed, and a recognition model identifies deficiencies in protective facilities. A point cloud change detection model automatically compares the current and previous scanned point clouds to generate a high-precision differential model, identifying areas where settlement, landslides, and material accumulation changes exceed limits. A comprehensive risk assessment model inputs various identified hazard characteristics, combined with environmental and production activity data, into the risk assessment model, outputting risk levels and probabilities. Step 4: Tiered early warning issuance and collaborative management response. After the risk assessment in Step 3 is completed, an analysis report is generated. The early warning rule engine is activated according to preset rules. When a high-level early warning occurs in Step 4, a strong reminder is triggered simultaneously on the command center's large screen, in relevant areas, and on the responsible person's mobile APP. When a low-level early warning occurs, the early warning information is automatically associated with hazard details, on-site video streams, and emergency response plans. Repair and review work orders are automatically generated and dispatched to the mobile terminals of the responsible personnel. Step 5: The interception net sensing layer is triggered. An intelligent interception net is deployed below the potential instability area of the slope. Multi-mode sensor nodes are arranged in a matrix on the interception net. The sensors integrate stress, vibration, tilt angle sensing and wireless transmission modules. The sensor network is completed and the signal is connected to the early warning center. The digital twin model has marked the unique ID and precise coordinates of each sensor. Step 5 is monitored normally during operation. The sensors periodically report status data to form a background baseline. When the interception net is impacted by landslide rocks, a group of sensors simultaneously triggers an over-threshold alarm, but communication is still maintained. The early warning center automatically delineates the suspected impact source area on the map based on the spatiotemporal relationship of the sensor signal sequence, using triangulation or shock wave propagation algorithms. When some sensor signals are suddenly completely interrupted, the early warning center directly determines the physical area where the sensor that lost the signal is located as a direct danger area. In both cases, the system automatically generates the highest priority UAV emergency verification task.
[0023] Step Six: Precise Aerial Verification and Change Detection. In Step Six, the drone receives an emergency alarm from the intelligent interception network. Instead of following the regular flight path, the drone flies directly to the alarm area and lowers its altitude to conduct close-range reconnaissance. It takes high-definition photos, thermal imaging scans, and rapid LiDAR scans of the suspected impact source area and the directly dangerous area from multiple angles. The images are transmitted back to the command center in real time, and AI assists in analyzing the scale, boundaries, stability, and threat level of the landslide. Step 7: Dynamic Risk Assessment and Light Warning Zone Delineation. In Step 7, the warning center integrates interception network data and drone verification data to dynamically generate three layers of areas on a 3D map: a red high-risk zone, a yellow warning zone, and a green safe assembly zone. The system automatically dispatches dedicated warning drones to the site airspace. Based on the received regional geographic coordinates, the drones project clear red and yellow danger zone boundaries onto the ground using high-intensity projection lights, and project green assembly markers into the green zone. Step 8: Collaborative Emergency Response and Closed-Loop Management. In Step 8, the system automatically locks the personnel location cards in the red and yellow areas, sends a forced evacuation alarm to their handheld terminals, and can automatically control the power outage of the operating equipment in the area. On the 3D situation map, the command center directs rescue personnel to the green safe assembly area and plans safe reconnaissance or disposal routes. Drones continuously monitor the site, track the personnel evacuation and slope stability. Every step of the disposal process is recorded in the system. After the event, all data is packaged into a case study for optimizing the alarm threshold of the interception network, the AI recognition model, and the emergency response plan.
[0024] When using this solution: Step 1: Collaborative Inspection Task Planning. Based on safety management procedures, daily and periodic inspection plans are generated. Ground micro-seismic sensors detect abnormal vibrations, slope radar detects small displacements exceeding thresholds, gas concentration monitoring point data is abnormal, and manual reports of potential hazards are received. The data center control system automatically obtains meteorological information and adjusts special inspections based on the meteorological information. The early warning center automatically generates collaborative inspection task packages based on trigger conditions. The drone sub-tasks include: designated routes covering all slope areas, key inspections of cracks, seepage, and falling rocks, equipped with infrared thermal imagers, lidar, and high-resolution night vision cameras, specifying the visible light video, infrared temperature matrix, and 3D point cloud data to be transmitted. Through the collaborative management platform, the tasks are automatically dispatched to the drone hangar, robot control station, and corresponding monitoring positions.
[0025] Step Two: The drone automatically flies along a preset route, utilizing onboard millimeter-wave radar and a vision system for obstacle avoidance at night. Multimodal data collection in Step Two includes: high-definition night vision video streams, infrared thermal imaging data, and lidar point clouds. The high-definition night vision video stream is used for real-time AI analysis of slope morphology, the presence of unauthorized intrusions, and abnormal vehicle parking. Infrared thermal imaging data is used to monitor abnormal temperature rises in power distribution facilities and the mining environment to prevent fires. LiDAR point clouds are used to scan mining slopes and spoil heaps, comparing them with historical point cloud models to calculate changes in earthwork volume and potential slip surfaces. The drone is equipped with an edge computing module, running a lightweight AI model to perform real-time analysis of the video stream. Once a suspected hazard is detected, its location is immediately marked, and the drone flies closer for focused scanning. Simultaneously, high-priority data is transmitted back via the 5G network. After the drone completes its flight path, all raw data and preprocessed results are transmitted back to the data analysis center. The drone then returns to base and begins automatic charging and data export.
[0026] Step 3: Receive all data from this inspection. Simultaneously, the system automatically links to data sources from the same time period and region for multi-source data fusion analysis and risk assessment. The information reception process involves the UAV infrared sensor detecting abnormal temperatures in the area, then retrieving historical data from nearby fixed temperature sensors to examine temperature trends. The UAV imagery reveals new cracks on the slope, which are compared with displacement increments from the lidar point cloud and slope monitoring radar data to verify the severity and activity of the cracks. In-depth analysis of visible light and infrared images is performed, and a recognition model identifies deficiencies in protective facilities. A point cloud change detection model automatically compares the current and previous scans of the point cloud, generating a high-precision differential model to identify areas exceeding limits for settlement, landslides, and material accumulation changes. A comprehensive risk assessment model inputs the characteristics of various identified hazards, combined with environmental and production activity data, into the risk assessment model, outputting risk levels and probabilities, and generating a structured risk analysis report. This report includes: a list of hazards, the geographical location of each hazard, image evidence, related data, risk level, and the type of accident that may develop from it. Step Four: The risk analysis report has been generated. The early warning rule engine is activated according to the preset rules, and tiered early warnings are issued and collaborative management responses are implemented. After the risk assessment in Step Three is completed, an analysis report is generated, and the early warning rule engine is activated according to the preset rules. When a high-level early warning occurs in Step Four, a strong reminder is triggered simultaneously on the command center's large screen, in relevant areas, and on the responsible person's mobile APP. When a low-level early warning occurs, the early warning information is automatically associated with hazard details, on-site video streams, and emergency response plans, and maintenance and review work orders are automatically generated and dispatched to the mobile terminals of the responsible personnel. Based on the location and type of the early warning, the command center can notify nearby inspection personnel or vehicles to verify the situation with one click through the system and plan a safe route. If the early warning involves personnel safety, the system can automatically link up to prevent personnel with location cards in the area from entering, or control nearby equipment to shut down. The personnel handling the situation can report the on-site situation and upload photos after the handling through their mobile terminals. The system tracks work order status until the loop is closed. All early warning and handling information is visualized and annotated in the 3D digital twin model of the mine, providing a three-dimensional perspective for decision-making. A large amount of closed-loop data of early warning, handling and result is accumulated over a certain period. The closed-loop data is used to retrain the AI recognition model and optimize the recognition accuracy. Based on the false alarms and missed alarms of the early warning, the threshold and logic of the early warning rules are adjusted, and a new version of the AI algorithm model and early warning rules are released and updated to the entire system.
[0027] Step 5: The barrier sensing layer is triggered. An intelligent barrier net is deployed below the potential instability area of the slope. Multi-mode sensor nodes are arranged in a matrix on the barrier net. The sensors integrate stress, vibration, tilt angle sensing and wireless transmission modules. The sensor network is completed and the signal is connected to the early warning center. The unique ID and precise coordinates of each sensor have been marked in the digital twin model. Step 5 is monitored normally during operation. The sensors periodically report status data to form a background baseline. Example 1: When the interception net is impacted by landslides and falling rocks, a group of sensors are affected by a sudden increase in stress and violent vibration, triggering an over-threshold alarm in sequence or simultaneously, but still maintaining communication. The early warning center automatically delineates the suspected impact source area on the map based on the spatiotemporal relationship of the sensor signal sequence, using triangulation or shock wave propagation algorithms. Example 2: When some sensors are buried or destroyed, and the signal is suddenly and completely interrupted, the early warning center directly identifies the physical area where the signal-lost sensor is located as a direct danger area. Regardless of scenario one or two, the system automatically generates the highest priority drone emergency inspection task.
[0028] Step Six: The drone's workflow revolves around precise aerial verification and change detection, and is divided into two modes: routine automated patrol and emergency verification, forming an all-weather, precise monitoring system for the mining area's slopes.
[0029] Routine automated comparative inspections are planned tasks and automatically initiated at fixed times each day. Drones, equipped with dual-light cameras, travel along a pre-set high-precision flight path to capture gridded images of the entire mining area, obtaining visible light and infrared images. After the data is transmitted back, the system's change detection AI automatically compares the day's images with previous images and a baseline 3D model. The system can intelligently determine the nature of the changes: if it identifies construction activities that conform to the plan, such as bench excavation or road extension, it is automatically classified as a construction change, and the 3D model and archived images are updated simultaneously; if it identifies features such as new cracks, local bulging, seepage and slippery areas, or abnormal loading, it is immediately marked as abnormal. After the process is completed, the system automatically generates a daily slope change report, archives normal changes, and automatically upgrades abnormal changes to a medium-risk warning, triggering a manual review process.
[0030] Emergency precision verification, as an event-driven task, is directly triggered by the critical alarm caused by the interruption of the interception network signal in step five. Upon receiving the alarm, the drone immediately switches to target-oriented mode and flies directly to the alarm area for close-range reconnaissance. In the target area, the drone descends to a lower altitude and performs multi-angle high-definition photography, thermal imaging scanning, and rapid lidar scanning to collect data on the suspected impact source area and the directly hazardous area. All images and data are transmitted back to the command center in real time and analyzed with AI assistance to quickly assess the landslide's size, boundaries, stability, and potential threats to facilities or personnel below. This allows for the immediate acquisition of key on-site information, accurately confirming the scope, severity, and impact of the accident or hazard, and providing direct decision-making support for the next stage.
[0031] Step Seven: Once the emergency verification drone transmits high-definition on-site imagery, thermal imaging, and lidar data, the process is immediately initiated. The early warning center will fuse and analyze the impact data from the ground-based "intelligent interception network" and the precise reconnaissance data from the aerial drones, dynamically delineating three layers of risk control zones on a digital 3D map of the mining area: Red high-risk zone: This zone covers the landslide itself and the unstable area above it that may experience secondary landslides. Entry of any personnel or equipment into this zone is strictly prohibited. Yellow warning zone: The area surrounding the red zone is designated as the area that may be affected by flying rocks, debris flows, or secondary disasters. Entry is prohibited for non-essential rescue personnel. Green safe assembly zone: A safe space designated for personnel assembly and command deployment in a safe area that is upwind or lateral and has stable and open terrain.
[0032] Once the area delineation command is issued, the system automatically dispatches dedicated warning drones deployed nearby to fly over the site. These drones are equipped with high-intensity projectors or high-brightness LED arrays. Based on the received geographical coordinates of the red, yellow, and green zones, they project conspicuous light field boundaries directly onto the ground: red light spots cover high-risk areas, yellow light halos delineate warning areas, and clear green arrows are projected in the green safety areas to guide personnel evacuation.
[0033] This measure creates a three-dimensional, visually visible on-site safety control zone. The lighting effect is particularly significant at night or in low visibility conditions. It allows all personnel on-site to instantly understand the risk distribution and safety situation visually, without the need for complex equipment, ensuring that commands are communicated and executed unambiguously and efficiently. Step 8: Once the light warning zone is successfully established on site and the warning information is synchronized to all personnel terminals and the command screen, the system immediately enters the collaborative emergency response state.
[0034] First, the system automatically executes safety linkage operations: based on the designated red and yellow electronic fence zones, it automatically scans and locks onto all personnel and vehicles wearing location cards within the area, sending a forced evacuation alarm with vibration and a strong audible alert to their terminals. If necessary, the system can remotely and automatically cut off the power to the engineering equipment in the area, forcibly stopping operations to eliminate risks.
[0035] The command center uses a 3D situation map that integrates real-time early warning information for visualized command and dispatch. Commanders can clearly see the distribution of personnel, directly direct personnel in danger zones to evacuate to green safe assembly areas along planned safe routes, and plan safe travel and operation routes for the upcoming emergency reconnaissance teams.
[0036] The entire response process was monitored and recorded. On-site drones continuously monitored the situation, transmitting real-time footage of personnel evacuation progress, slope stability changes, and rescue operations. Every key step, instruction, and on-site status from alarm triggering to completion of the response was automatically recorded and time-stamped by the system.
[0037] After the hazard was averted and the warning was lifted, the system did not cease operation but entered a crucial closed-loop learning phase. All the multimodal data from this incident, from the initial sensor triggering to the final safe evacuation of personnel (including sensor waveforms, drone imagery, laser point clouds, communication logs, command records, etc.), will be automatically packaged into a complete digital case library. This case will be used for in-depth analysis and debriefing to optimize the alarm sensitivity and algorithms of the intelligent interception network, train the change detection AI model, and iteratively upgrade the entire emergency response plan, thus enabling the system to continuously evolve.
[0038] After the emergency was completely over, the system returned to normal monitoring mode, and a complete digital archive of the incident was preserved.
[0039] Through multi-source collaborative inspection task planning and intelligent triggering mechanisms, the system achieves automation and intelligence in mine safety inspections. The system can generate daily and periodic inspection plans based on safety management procedures and automatically trigger specific inspection tasks when anomalies are detected by equipment such as microseismic sensors, slope radar, and gas monitoring, or when potential hazards are manually reported. Simultaneously, the system can dynamically adjust inspection strategies based on real-time meteorological information, significantly improving the targeting and timeliness of inspections, ensuring that hazards are detected and responded to promptly. The system can automatically correlate various monitoring data from the same time period and area, utilizing point cloud change detection, image recognition, and a comprehensive risk assessment model to conduct in-depth analysis and risk level determination of hazards such as cracks, settlement, and landslides. After risk assessment, the system automatically generates a structured report and initiates tiered early warnings according to preset rules, achieving full automation from hazard discovery to risk assessment.
[0040] Through dynamic risk assessment and the delineation of light-based warning zones, a three-dimensional and visual identification of on-site risk areas is achieved. The system integrates ground and aerial data to dynamically generate red, yellow, and green risk zones on a 3D map, and deploys dedicated warning drones to project corresponding colored light spots and assembly markers onto the ground. Especially at night or in low-visibility conditions, this method provides intuitive and clear guidance for personnel evacuation and assembly, greatly improving the efficiency and reliability of on-site safety control.
[0041] After the incident, all data was packaged into a case library to optimize sensor alarm thresholds, train AI recognition models, iterate emergency plans, promote continuous learning and upgrading of the system, and continuously improve the level of intelligence in mine safety management.
[0042] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention. No reference numerals in the claims should be construed as limiting the scope of the claims.
Claims
1. A method for intelligent detection, collaborative management, and early warning of mine accidents, characterized in that, Specifically, the following steps are included: Step 1: Multi-source collaborative inspection task planning and triggering; Step 2: Simultaneous acquisition of intelligent sensing and multimodal data by UAV; Step 3: Receive all data from this inspection, and the system will automatically link data from the same time period and region for multi-source data fusion analysis and risk assessment. Step 4: Tiered early warning issuance and collaborative management response; Step 5: The net sensing layer is triggered; Step Six: Precise Aerial Verification and Change Detection; Step 7: Dynamic risk assessment and delineation of light warning zones; Step 8: Coordinated emergency response and closed-loop management.
2. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: In step one, the collaborative inspection task planning generates daily and periodic inspection plans based on safety management procedures. Ground micro-seismic sensors detect abnormal vibrations, slope radar detects small displacements exceeding thresholds, gas concentration monitoring point data is abnormal, and manual reports of potential hazards are received. The data center control system automatically obtains meteorological information and adjusts the special inspections accordingly.
3. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: In step two, the UAV flies automatically along a preset route and uses onboard millimeter-wave radar and vision system to achieve obstacle avoidance at night. The multimodal data collection in step two includes: high-definition night vision video stream, infrared thermal image data and lidar point cloud.
4. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 3, characterized in that: In step three, the information reception process involves the UAV detecting abnormal temperatures in the area using infrared sensors. It then retrieves historical data from nearby fixed temperature sensors to examine temperature trends. UAV imagery reveals new cracks on the slope, which are compared with displacement increments from the lidar point cloud and slope monitoring radar data to verify the severity and activity of the cracks. In-depth analysis of visible light and infrared images is performed, and a recognition model identifies deficiencies in protective facilities. A point cloud change detection model automatically compares the current and previous point clouds, generating a high-precision differential model to identify areas exceeding limits for settlement, landslides, and material accumulation changes. A comprehensive risk assessment model integrates the identified hazard characteristics with environmental and production activity data, inputting them into the risk assessment model to output risk levels and probabilities.
5. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: After the risk assessment is completed in step three, an analysis report is generated. The early warning rule engine is activated according to the preset rules. In step four, when a high warning occurs, a strong reminder is triggered simultaneously on the command center's large screen, in the relevant area broadcast, and on the responsible person's mobile APP. When a low warning occurs, the warning information is automatically associated with the hazard details, on-site video stream, and emergency response plan, and maintenance and review work orders are automatically generated and dispatched to the mobile terminals of the responsible personnel.
6. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 5, characterized in that: Below the potential instability zone of the slope, an intelligent interception net is deployed. Multi-mode sensor nodes are arranged in a matrix on the interception net. The sensors integrate stress, vibration, tilt angle sensing and wireless transmission modules. The sensor network is completed and the signal is connected to the early warning center. The unique ID and precise coordinates of each sensor have been marked in the digital twin model.
7. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: Step 5: During operation, normal monitoring is performed. The sensors periodically report status data to form a background baseline. When the interception net is impacted by landslides and falling rocks, a group of sensors simultaneously trigger an over-threshold alarm, but communication is still maintained. Based on the spatiotemporal relationship of the sensor signal sequence, the early warning center uses triangulation or shock wave propagation algorithms to automatically delineate the suspected impact source area on the map. When some sensor signals suddenly and completely fail, the early warning center directly identifies the physical area where the sensor that lost the signal is located as a direct danger area. In both cases, the system automatically generates the highest priority UAV emergency verification task.
8. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: In step six, the drone receives an emergency alarm from the intelligent interception network. Instead of following its regular flight path, the drone flies directly to the alarm area and lowers its altitude to conduct close-range reconnaissance. It takes high-definition photos, performs thermal imaging scans, and uses lidar to quickly scan the suspected impact source area and the directly dangerous area from multiple angles. The images are transmitted back to the command center in real time, and AI assists in analyzing the scale, boundaries, stability, and threat level of the landslide.
9. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: In step seven, the early warning center integrates the interception network data and drone verification data to dynamically generate three layers of areas on a 3D map: a red high-risk area, a yellow warning area, and a green safe assembly area. The system automatically dispatches a dedicated warning drone to the site airspace. Based on the received regional geographic coordinates, the drone projects clear red and yellow danger zone boundaries onto the ground using a high-intensity projection light, and projects a green assembly marker in the green area.
10. The intelligent detection, collaborative management, and early warning method for mine accidents according to claim 1, characterized in that: In step eight, the system automatically locks onto the personnel location cards in the red and yellow areas, sends a forced evacuation alarm to their handheld terminals, and can automatically control the power outage of the work equipment in the area. On the three-dimensional situation map, the command center directs the rescue personnel to the green safe assembly area and plans safe reconnaissance or disposal routes. Drones continuously monitor the site, track the personnel evacuation and slope stability. Every step of the disposal process is recorded in the system. After the event, all data is packaged into a case study for optimizing the alarm threshold of the interception network, the AI recognition model, and the emergency response plan.