A method and system for safety inspection in open-pit mines based on digital twin models

Abstract

This invention relates to the field of digital twin model technology, and provides a method and system for safety inspection of open-pit mines based on digital twin models. The method includes accessing historically archived geographic information data of the mining area and real-time surface images and laser point cloud data of the mining area transmitted by a UAV platform. After fusion processing and dynamic updates, a three-dimensional digital twin model is constructed. Based on the information contained in the three-dimensional digital twin model, and combined with preset inspection task coverage and target point requirements, a three-dimensional inspection path that meets safety constraints and optimizes task efficiency is planned and generated through spatial analysis and obstacle avoidance calculations. Based on the waypoint coordinates, flight altitude, and speed parameters in the three-dimensional inspection path, flight control commands are calculated and generated, and sent to the UAV inspection platform via a communication unit, driving it to perform autonomous flight inspection tasks along the planned path. This invention improves the automation and intelligence level of closed-loop management in mines.

Description

Technical Field

[0001] This invention relates to the field of digital twin model technology, and in particular to a method and system for safety inspection of open-pit mines based on digital twin models. Background Technology

[0002] Safety inspections are essential for ensuring safe production in open-pit mines, spoil heaps, and tailings ponds. Mining companies are required to conduct regular safety inspections of their mining areas according to relevant regulations, and routinely carry out anti-violation (anti-three violations) operations, including illegal command, illegal operation, and violations of labor discipline. Currently, the most commonly used inspection method is still traditional manual ground inspection, which has many problems. It is not only time-consuming and labor-intensive, but also poses safety hazards to safety management personnel during inspections, and has blind spots. As intelligent equipment with autonomous flight capabilities, drones have advantages such as high flexibility, low flight costs, and high operational coverage. They can execute tasks through radio remote control or preset programs, and their multi-sensor integrated systems can adapt to complex working conditions. In mining areas with rugged terrain and complex environments, drone inspections, through multimodal data acquisition technology, can significantly improve the efficiency and accuracy of slope disease identification, while avoiding the risks of close-range manual operations. Equipped with high-definition imaging equipment and utilizing LiDAR and visual fusion algorithms, it can generate high-precision 3D models of real-world scenes with a single click. It can capture defect features at millimeter-level resolution and transmit data in real time, constructing a full-element 3D scene model. Its high efficiency, safety, economy, and high precision make it a core technology for mines, regular inspections, and dynamic monitoring and early warning, providing crucial support for open-pit mine inspections. Currently, open-pit mine inspections are transitioning from purely manual inspections to drone-assisted inspections. In the industry context, with the advancement of smart mine construction, using drones for aerial photography and 3D modeling has become standard practice for leading mining companies. However, existing technical solutions are mostly in the offline processing, single-function, or semi-automatic stage, still exhibiting significant limitations in fully automatic closed-loop monitoring, deep fusion and recognition of multi-modal data, and real-time dynamic early warning for violations.

[0003] In recent years, artificial intelligence technology has made groundbreaking progress, with deep learning algorithms in the field of machine vision demonstrating powerful feature extraction and pattern recognition capabilities. Compared with traditional machine learning methods, deep learning models based on convolutional neural networks exhibit higher accuracy and faster processing speeds in tasks such as image classification, object detection, and semantic segmentation, and show stronger generalization performance, especially in complex scenarios. Various deep neural network architectures have been successfully applied in the field of inspection, providing efficient solutions for processing the massive amounts of image data generated by drone inspections. By building intelligent analysis platforms, they have replaced traditional manual analysis methods, significantly improving the automation level and inspection efficiency of mines.

[0004] Traditional manual ground inspection: Safety management personnel, carrying rangefinders, cameras, and other tools, periodically walk or drive into the mining area, spoil heap, and tailings dam for on-site inspections. They visually observe slope cracks, measure drainage ditch siltation, and manually record violations, which are then compiled into the safety management system. High safety risks: Personnel must enter rugged terrain or active work areas, facing threats of landslides, falls, and vehicle injuries. Blind spots exist: Limited by ground-level perspective, it is difficult to detect minute deformations and hidden dangers at the top of steep slopes or within complex terrain. Poor timeliness: Data collection and analysis rely entirely on manual labor, making real-time early warning of dynamic hazards impossible. Conventional UAV aerial surveying and modeling system: Utilizes rotary-wing or fixed-wing UAVs equipped with visible light cameras to take pictures along a preset flight path. Subsequent offline 3D reconstruction is performed on a workstation using oblique photogrammetry software, such as ContextCapture, generating a 3D reality model of the mine for reserve calculation and terrain analysis. Low automation: Relies on manual on-site launch and battery replacement, making unattended routine inspections impossible. The data modality is limited: it mainly relies on visible light imagery, making it difficult to obtain accurate surface point clouds at night, in foggy conditions, or in areas with vegetated vegetation. It also lacks sensitivity to deep slope deformation. Furthermore, it lacks intelligent recognition: the model serves only as a visual base map; violations such as not wearing safety helmets, vehicle violations, and minor defects like eaves and cracks still require manual annotation on the model.

[0005] Fixed sensor-based slope monitoring systems deploy GNSS displacement monitoring stations, guy wire displacement meters, or fixed long-range laser scanners at key locations on mine slopes. Physical deformation data is transmitted back to the monitoring center via wired or wireless networks for threshold alarms. However, these systems have several limitations: limited coverage (being point-based monitoring that only covers localized areas with deployed sensors and cannot detect sudden hazards in newly excavated faces or non-critical areas); difficult deployment and maintenance (frequent blasting operations in mines easily damage fixed equipment, and the need to relocate equipment as mining progresses leads to high costs); and limited functionality (only monitoring physical displacement and unable to identify violations during operations or visual hazards such as blockages in drainage systems).

[0006] Existing technologies suffer from a disconnect between automated closed-loop systems and multimodal intelligent analysis. Specifically, there is a lack of a comprehensive solution that integrates automated takeoff and landing systems, real-time terrain-following flight based on real-time point clouds, laser and vision-based Multi-SLAM fusion modeling, and deep learning-based real-time automatic analysis of reserves, hidden dangers, and violations. Existing solutions often exhibit a disconnect between data acquisition, model building, and AI early warning, making it difficult to meet the comprehensive, all-weather, and unmanned safety closed-loop management requirements of open-pit mines. Summary of the Invention

[0007] To address the aforementioned technical problems, this invention provides a method for safety inspection of open-pit mines based on a digital twin model, comprising the following steps: establishing a three-dimensional digital twin model that is synchronized in real time with the geometry and attributes of the physical mine by integrating historical geographic archives with real-time imagery and point cloud data collected by UAVs; utilizing the terrain, facilities, and dynamic change information contained in the three-dimensional digital twin model, combined with the task objectives, performing spatial analysis and obstacle avoidance calculations to automatically generate the globally optimal inspection path; converting the planned optimal path into flight control commands executable by the UAV to drive the UAV to perform the inspection; and analyzing the data collected by the autonomous flight inspection mission to extract key information.

[0008] Optionally, historical archived geographic information data of the mining area and real-time surface images and laser point cloud data of the mining area transmitted by the UAV platform are integrated, processed and dynamically updated to construct a three-dimensional digital twin model that maintains the geometric shape and attribute information of the physical mining area. Based on the topography, facility distribution and dynamic change information contained in the three-dimensional digital twin model, combined with the preset inspection task coverage and target point requirements, a three-dimensional inspection path that meets the safety constraints and the optimal task efficiency is planned and generated through spatial analysis and obstacle avoidance calculation.

[0009] Optionally, the process of planning and generating a 3D inspection path that satisfies both safety constraints and optimal task efficiency includes the following steps: High-precision digital elevation data of the mining area surface, 3D contour data of permanent buildings and fixed facilities, and real-time data on spoil heap dumping progress, mining face advancement position, and temporary stockpile space occupancy are extracted from the 3D digital twin model. This data is then uniformly converted into voxel-based constraint elements containing spatial coordinates and obstacle attributes. When writing these voxel-based constraint elements into the spatial constraint set, the constraint elements are configured hierarchically based on the task status identifier issued by the task planning system: when the task status identifier is a regular safety inspection mode, the boundary data of the blasting warning zone is written into the set as a mandatory horizontal avoidance constraint, and the UAV is not allowed to enter the warning zone airspace during path search; when the task status identifier is a special clearing mode for blasting operations, the voxel-based avoidance constraint corresponding to the blasting warning zone is dynamically revoked, leaving only the physical obstacle voxel constraint, allowing path planning to include the airspace inside the warning zone into the passable area to support UAVs in performing full-field scanning and personnel clearing tasks within the warning line; thus forming a spatial constraint set for path search that is dynamically configured according to the task status. Based on the preset inspection task coverage, the required waypoint sequence positions are marked in the spatial coordinate system of the three-dimensional digital twin model. At the same time, according to the task target point requirements, the three-dimensional coordinates and orientation information of each target point are extracted. The waypoint sequence and target points are jointly constructed into a set of necessary nodes for path planning. Combined with the formed set of spatial constraints, a search for passable areas based on spatial connectivity is performed within the mining area represented by the three-dimensional digital twin model. Several candidate spatial polyline segment sequences that can connect all necessary nodes and completely avoid obstacles occupying space are selected. The obtained candidate spatial polyline segment sequences are mapped back to the three-dimensional digital twin model one by one. Using the accurate terrain profile data and facility distribution data provided by the three-dimensional digital twin model, the total spatial length of each candidate sequence is calculated. At the same time, the proximity between each polyline segment and the mine bench slope, as well as the intersection with overhead lines and transportation roads, are analyzed. The candidate sequence with the shortest total spatial length and which maintains a set safe distance from all fixed facilities and dynamic obstacles in the model is selected. The candidate sequence is then transformed into a three-dimensional inspection path composed of continuous waypoint coordinates and flight speeds between adjacent waypoints.

[0010] Optionally, the process of calculating the total spatial length of each candidate sequence includes the following steps: For each selected candidate spatial polyline segment sequence, the topographic profile data of the area traversed by the candidate spatial polyline segment sequence is extracted from the three-dimensional digital twin model; at the same time, the vertical profile data corresponding to each turning point of the polyline segment is extracted, and the straight line segment of the polyline segment in three-dimensional space is divided into several continuous spatial micro-segments according to the topographic undulation characteristics, and the two endpoints of each micro-segment are located inside the same topographic unit. Based on the spatial micro-segments obtained from segmentation, the precise three-dimensional coordinates of the start and end points of each micro-segment are extracted from the three-dimensional digital twin model. Using the spatial geometric properties of the mine bench surface and slope surface provided by the model, the projected length of each micro-segment on the horizontal plane is calculated. At the same time, based on the slope characteristics of the terrain crossed by the spatial micro-segment, the horizontal projected length is converted into the actual slope distance of the spatial micro-segment in three-dimensional space. For spatial micro-segments that cross the edge of the bench, their projected lengths on the upper and lower bench planes are calculated separately and summed to obtain the spatial length of the spatial micro-segment. The calculated spatial lengths of each spatial micro-segment are summed up. At the same time, the spatial location data of the overhead lines and transportation road facilities distributed along the candidate polyline segments are extracted from the three-dimensional digital twin model. For spatial micro-segments that intersect with overhead lines, the corresponding vertical detour distance is added based on the line sag data provided by the three-dimensional digital twin model. For spatial micro-segments that cross transportation roads, the corresponding horizontal avoidance distance is added based on the road width. The summed result is combined with the detour avoidance distance to obtain the total spatial length of the candidate sequence.

[0011] Optionally, the process of combining the summation result with the detour distance includes the following steps: From the calculated spatial length of each spatial micro-segment, extract the spatial length data of all spatial micro-segments that are not marked as intersecting with overhead lines or crossing transport roads. Accumulate the spatial lengths of the spatial micro-segments according to their order in the candidate spatial polyline segment sequence to obtain the basic path length data. Simultaneously, extract the vertical detour distance data corresponding to all spatial micro-segments marked as intersecting with overhead lines, and the horizontal avoidance distance data corresponding to all spatial micro-segments marked as crossing transport roads. Sort the detour and avoidance distance data according to the position of their respective micro-segments in the sequence to form a set of corrected distances to be merged. Starting with the obtained basic path length data, the first set of detour or avoidance distance data is sequentially taken from the set of corrected distances to be merged and merged with the starting value to obtain the first corrected path length. Then, the next set of distance data is sequentially taken from the set of corrected distances and merged with the previous corrected path length to obtain the second corrected path length. This process continues until all distance data in the set of corrected distances to be merged has been merged. Each merging operation adds the detour or avoidance distance to the path length of the interval where the spatial micro-segment is located, based on the actual position of the micro-segment corresponding to the distance data in the sequence. The final path length data is used as the initial total spatial length of the candidate spatial polyline segment sequence. Mine ground subsidence monitoring point data at the start and end points of the candidate sequence, as well as boundary data of the blasting warning zones along the route, are extracted from the 3D digital twin model. Based on the cumulative ground subsidence at the start and end points, the absolute elevation coordinates of the corresponding waypoints are corrected by subtracting the subsidence from the original Z-axis coordinates to obtain corrected waypoint coordinates reflecting the current actual surface elevation. Using the corrected 3D waypoint coordinates as a basis, the spatial length of the flight segment between the start point and adjacent waypoints, and between the end point and adjacent waypoints, are recalculated. The recalculated segment lengths replace the original values ​​of the corresponding intervals in the initial total spatial length, completing the path length correction caused by ground subsidence. Based on the safety distance requirements of the blasting warning zone, the micro-segments crossing the edge of the warning zone are finely adjusted horizontally. The corrected length data is then determined as the total spatial length of the candidate sequence.

[0012] Optionally, the process of adding the detour or avoidance distance to the path length of the interval containing the spatial micro-segment includes the following steps: From the formed set of distances to be merged and corrected, the first set of detour or avoidance distance data is taken out in sequence. At the same time, the position index of the spatial micro-segment in the candidate spatial polyline segment sequence is extracted from the example data. Based on the position index, the start and end positions of the interval corresponding to the spatial micro-segment are located from the obtained basic path length data. The path length data within the interval is separated from the basic path length data and used as the interval length of the micro-segment to be processed. The first set of detour or avoidance distance data is merged with the micro-segment interval length to be processed. For micro-segments that intersect with overhead lines, the vertical detour distance and the spatial micro-segment interval length are spatially geometrically superimposed along the line direction in three-dimensional space to form a new micro-segment interval length that includes the vertical detour path. For spatial micro-segments that cross transportation roads, the horizontal avoidance distance and the micro-segment interval length are spatially geometrically superimposed along the road vertical direction in the horizontal plane to form a new micro-segment interval length that includes the horizontal avoidance path. The resulting new micro-segment interval length is the actual flight path length of the spatial micro-segment after detour or avoidance processing. The obtained new micro-segment interval length is replaced with the original interval position of the spatial micro-segment in the basic path length data, while keeping the interval lengths of all other micro-segments before and after the spatial micro-segment unchanged, and the data is recombined to form the updated basic path length data. The currently processed detour or avoidance distance data is removed from the set of corrected distances to be merged, resulting in a reduced set of corrected distances. The updated basic path length data is used as the starting value for the next merging operation to process the next set of detour or avoidance distance data.

[0013] Optionally, the process of forming a new micro-segment interval length that includes the vertical bypass path includes the following steps: The spatial coordinate data of the axis of the overhead line intersecting the spatial micro-segment is extracted from the three-dimensional digital twin model to determine the spatial intersection point of the spatial micro-segment and the axis; at the same time, the three-dimensional coordinates of the start and end points of the micro-segment are extracted from the obtained interval length data of the micro-segment to be processed, and the straight path of the micro-segment between the original start and end points is divided into the pre-intersection segment and the post-intersection segment with the spatial intersection point as the boundary. Based on the extracted vertical detour distance data, the sag height data of the overhead line at the intersection and the safety clearance requirements on both sides of the line are extracted from the three-dimensional digital twin model. A detour height point that meets the safety clearance requirements is set vertically upward along the line axis at the spatial intersection position. The three-dimensional coordinates of the detour height point are spatially correlated with the end point of the section before the intersection and the starting point of the section after the intersection to construct a spatial zigzag line segment from the end point of the section before the intersection through the detour height point to the starting point of the section after the intersection. The three-dimensional spatial length of the constructed spatial zigzag segment is merged with the spatial lengths of the segment before and after the intersection. The new spatial length obtained after merging is used to replace the total straight-line length from the starting point to the ending point in the original spatial micro-segment interval length, forming a new micro-segment interval length that includes the vertical bypass path.

[0014] Optionally, the process of forming a new micro-segment interval length that includes the horizontal avoidance path includes the following steps: The coordinate data of the centerline of the transportation road crossed by the spatial micro-segment and the coordinate data of the road boundary line are extracted from the three-dimensional digital twin model to determine the intersection of the horizontal projection of the micro-segment and the centerline of the road. At the same time, the projection coordinates of the start and end points of the micro-segment on the horizontal plane are extracted from the length data of the micro-segment to be processed. The horizontal projection line between the original start and end points of the micro-segment is divided into the approach segment and the departure segment with the intersection of the horizontal projection as the boundary. Based on the extracted horizontal avoidance distance data, the safety distance requirement data outside the road boundary line is extracted from the three-dimensional digital twin model. At the intersection of the horizontal projections, a avoidance point that meets the safety requirements is set by translating to one side along the vertical direction of the road centerline. The horizontal coordinates of the avoidance point are spatially associated with the end point of the approach section and the start point of the departure section in the horizontal plane to construct a horizontal polygonal line segment from the end point of the approach section through the avoidance point to the start point of the departure section. The length of the constructed horizontal plane polyline segment is merged with the actual slant distance lengths of the approach and departure segments on their respective terrain profiles. The new spatial length obtained after merging is used to replace the total slant distance length from the start point to the end point in the original spatial micro-segment interval length, forming a new micro-segment interval length that includes the horizontal avoidance path.

[0015] Optionally, when generating inspection tasks, the task planning system automatically sets the task status identifier based on the type of inspection instruction currently issued: for routine tasks such as daily slope inspections, facility inspections, and anti-illegal inspections, the task status identifier is set to the routine safety inspection mode; for tasks involving clearing and confirming the warning zone before blasting operations, the task status identifier is set to the blasting operation special clearing mode; the task status identifier is transmitted to the spatial constraint condition set construction stage of the 3D digital twin model generation module before path planning, serving as the input parameter for the voxel constraint hierarchical configuration; based on the waypoint coordinates, flight altitude, and speed parameters in the 3D inspection path, the system calculates and generates flight control instructions including attitude adjustment and heading control, which are sent to the UAV inspection platform through the communication unit to drive it to execute autonomous flight inspection tasks along the planned path; the system analyzes the data collected by the autonomous flight inspection tasks, extracts key information, and uses it for dynamic management of reserves, identification and judgment of major hidden dangers, slope stability monitoring, management of blasting operations, anti-illegal inspections, inspection of drainage systems, dynamic monitoring of spoil heaps, and emergency early warning.

[0016] The open-pit mine safety inspection system based on a digital twin model provided by this invention includes: The drone inspection platform is used to carry recording equipment to carry out flight inspection tasks. It can realize functions such as human-machine interaction, remote visual monitoring, voice intercom, dynamic recognition and data storage. The unmanned aerial vehicle (UAV) system is used to control the flight of the UAV, collect environmental data in the mining area, and transmit data to the ground control station. It includes an airport, sensors, data transmission, and a ground control station. The ground control station is used to send inspection commands and receive and display inspection data. The mission planning system is used to execute functions such as UAV trajectory, inspection time, inspection frequency and sensor type according to inspection instructions, and to generate flight control instructions based on the three-dimensional inspection path to drive the UAV platform to fly autonomously along the planned path. The 3D digital twin model generation module is used to construct and update a 3D digital twin model synchronized with the physical mining area based on historical and real-time mining area geographic information data; and to plan the optimal 3D inspection path for the UAV platform based on the 3D digital twin model and inspection task requirements. The intelligent recognition module is used to analyze real-time acquired image data, identify inspection targets of preset categories, and generate recognition results containing target location and category information; it also analyzes data collected by autonomous flight inspection missions, extracts key information, and uses it for dynamic management of reserves, identification and judgment of major hidden dangers, slope stability monitoring, management of blasting operations, prevention of illegal construction, inspection of drainage systems, dynamic monitoring of spoil heaps, and emergency early warning.

[0017] This invention enables routine, closed-loop inspections of mines around the clock and without human intervention. Compared to traditional manual inspections, which require on-site personnel to launch drones and manually change batteries, and are limited by terrain, this invention achieves full automation. The invention integrates an AC / DC charging system and a remote control link through an intelligent airport, supporting drone takeoff from Airport A and landing at Airport B, as well as centimeter-level autonomous takeoff and landing. The hardware layer integrates an intelligent airport with automatic charging, environmental monitoring, and an RTK base station, coupled with a remote data transmission link. A high-precision digital twin model with millimeter-level resolution is constructed, eliminating monitoring blind spots. Compared to traditional single-vision modal or fixed-point sensors, this solves the problem of missing point clouds under vegetation cover and complex terrain. The lidar point frequency reaches 1,920,000 points / second (three echoes), with an elevation accuracy of less than 5cm and hovering accuracy of ±0.1 meters (vertical / horizontal). A multi-SLAM fusion algorithm combining laser and vision, combined with real-time terrain-following flight technology, is used to construct a global terrain DEM. This invention enables real-time AI-powered intelligent identification and early warning of violations and minor defects, achieving automated judgment by integrating data collection and analysis, compared to manual annotation on visual base maps. It utilizes deep learning algorithms to make real-time judgments on features such as canopy structures, cracks, lack of safety helmets, and vehicle violations. The application layer deploys a deep neural network model specific to mining scenarios, integrating target detection and semantic segmentation algorithms to improve the precision of dynamic reserve management and blasting effect analysis. Compared to traditional manual surveying or estimation, it provides accurate volume calculation based on 3D point clouds. Based on dynamic scanning and point cloud volume calculation, it can quickly calculate the amount of soil received at spoil heaps and the reserves in mining areas. The dynamic reserve management algorithm in the data application module uses multi-period point cloud comparison for difference calculation, enhancing safety monitoring capabilities in extreme environments (nighttime, heavy fog), overcoming the limitations of traditional visible light inspections that cannot operate at night or in low visibility conditions. The thermal imaging camera sensitivity supports detection from -40℃ to 150℃; the UAV payload integrates telephoto, wide-angle, and thermal imaging multimodal sensors.

[0018] This invention completely changes the traditional manual inspection model of open-pit mines by constructing an integrated, fully automated inspection system that combines sensing, transmission, computing, and alarm functions. By integrating a drone-based intelligent airport, multimodal fusion perception, and a deep learning analysis platform, it not only solves the problem of monitoring blind spots in dangerous areas such as steep slopes, spoil heaps, and tailings ponds, but also achieves a leap from physical displacement monitoring to full-dimensional identification of operational behavior and environmental hazards through millimeter-level digital twin models and real-time AI algorithms. This significantly reduces the operational risks for safety management personnel, improves the automation and intelligence level of closed-loop management in mines, and provides a scientific and accurate data foundation for safe production in smart mines.

[0019] Other features and advantages of the invention will be set forth in the following description, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures particularly pointed out in the written description and the accompanying drawings.

[0020] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0021] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used in conjunction with embodiments of the invention to explain the invention and do not constitute a limitation thereof. In the drawings: Figure 1 This is a flowchart of the open-pit mine safety inspection method based on a digital twin model in Embodiment 1 of the present invention; Figure 2 This is a schematic diagram of the open-pit mine safety inspection method based on a digital twin model in Embodiment 1 of the present invention. Figure 3 This is a process diagram of constructing a three-dimensional digital twin model that maintains the geometric shape and attribute information of the physical mining area in Embodiment 2 of the present invention; Figure 4 This is a process diagram for planning and generating a three-dimensional inspection path that satisfies both safety constraints and optimal task efficiency in Embodiment 9 of the present invention. Figure 5 This is an engineering diagram illustrating the analysis of data collected during autonomous flight inspection missions in Embodiment 15 of the present invention. Figure 6 This is a block diagram of the open-pit mine safety inspection system based on a digital twin model in Embodiment 16 of the present invention. Detailed Implementation

[0022] The preferred embodiments of the present invention will be described below with reference to the accompanying drawings. It should be understood that the preferred embodiments described herein are for illustration and explanation only and are not intended to limit the present invention.

[0023] The terminology used in the embodiments of this application is for the purpose of describing particular embodiments only and is not intended to limit the embodiments of this application. The singular forms "a," "say," and "this" used in the embodiments of this application are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the terms used herein refer to and / or include any or all possible combinations of one or more associated listed items.

[0024] The inventive concept of this invention lies in constructing a data-driven, virtual-real synchronized, closed-loop feedback autonomous inspection system for open-pit mines using unmanned aerial vehicles (UAVs). The specific conceptual logic is as follows: Breaking the limitations of traditional static maps, by integrating historical geographical archives with real-time UAV-collected imagery and point cloud data, a 3D digital twin model is established that maintains real-time geometric and attribute synchronization with the physical mine, providing an accurate virtual copy for inspection operations. Abandoning the inefficient manual route planning mode, the constructed digital twin model, utilizing its contained precise terrain, facility, and dynamic change information, combined with task objectives, performs spatial analysis and obstacle avoidance calculations to automatically generate the globally optimal inspection path. The planned optimal inspection path is transformed into flight control commands executable by the UAV, driving the UAV to perform inspections in the real physical space. Simultaneously, newly collected data is fed back to update the twin model, forming a closed-loop system where virtual planning guides real flight, and real data feeds back into the virtual model for continuous evolution. The 3D digital twin model, as the core decision-making layer for path planning, directly participates in task generation; path planning is automatically completed based on the 3D digital twin model containing dynamic change information, avoiding the lag and inaccuracy of manual planning.

[0025] This invention employs technologies such as UAV airports, machine vision, digital imaging, laser point clouds, and laser-vision Multi-SLAM fusion algorithms to form a complete chain architecture encompassing UAV airports, monitoring and perception, intelligent models, and software systems. It mainly includes the following aspects: UAV platform and control system composition: including a flight platform, sensor system, data transmission system, and ground control station; reasonable task planning is formulated based on the characteristics and monitoring needs of different open-pit mines, including flight paths, inspection time, inspection rate, and monitoring accuracy. Monitoring and perception: Open-pit mines typically need to complete UAV inspections, panoramic imaging, and laser point cloud monitoring of open-pit mining areas, spoil heaps, and tailings ponds to provide data support for production safety management. Data analysis and application: Images, videos, and LiDAR data are collected through sensors mounted on the UAVs, and the collected data is analyzed to extract key information for dynamic reserve management, identification and judgment of major hidden dangers, slope stability monitoring, blasting operation management, anti-illegal construction, drainage system inspection, soil volume, stability monitoring, dynamic monitoring of spoil heaps, and emergency early warning.

[0026] Example 1: As Figure 1 As shown, this embodiment of the invention provides a safety inspection method for open-pit mines based on a digital twin model, comprising the following steps: S100: Accesses historical archived geographic information data of the mining area and real-time surface images and laser point cloud data of the mining area transmitted back by the UAV platform. After fusion processing and dynamic updates, a three-dimensional digital twin model that keeps the geometric shape and attribute information of the physical mining area synchronized is constructed. S200: Based on the terrain, facility distribution and dynamic change information contained in the three-dimensional digital twin model, combined with the preset inspection task coverage and target point requirements, a three-dimensional inspection path that meets safety constraints and optimal task efficiency is planned and generated through spatial analysis and obstacle avoidance calculation. S300: Based on waypoint coordinates, flight altitude, and speed parameters in the three-dimensional inspection path, it calculates and generates flight control commands including attitude adjustment and heading control, and sends them to the UAV inspection platform through the communication unit to drive it to perform autonomous flight inspection tasks along the planned path; it analyzes the data collected by the autonomous flight inspection tasks, extracts key information, and uses it for dynamic management of reserves, identification and judgment of major hidden dangers, slope stability monitoring, management of blasting operations, prevention of illegal construction, inspection of drainage systems, dynamic monitoring of spoil heaps, and emergency early warning.

[0027] The process of calculating and generating flight control commands based on waypoint coordinates, flight altitude, and speed parameters in the 3D inspection path is as follows: First, the coordinates of a series of discrete waypoints in the 3D inspection path are transformed into the UAV body coordinate system. Combined with the flight altitude and speed parameters set at each waypoint, the target spatial position and target motion state that the UAV needs to reach in the current flight segment are calculated. Then, the UAV's current real-time attitude data, including position, attitude angles, and angular velocity, are compared with the target attitude data. The desired attitude angles required to achieve state adjustment are calculated through the control algorithm, including pitch angle, roll angle, yaw angle, and throttle control. Finally, the desired attitude angles and control quantities are converted into standard control commands that the UAV flight control system can recognize. These commands are then sent to the UAV inspection platform through the communication unit to drive it to adjust its attitude and heading, ensuring that the UAV can smoothly and accurately perform autonomous flight inspection tasks along the planned path.

[0028] The working principle and beneficial effects of the above technical solution are as follows: (Refer to Appendix for details.) Figure 2 This embodiment constructs a three-dimensional digital twin model that is synchronously updated with the physical mining area by fusing historical geographic data with real-time UAV imagery and laser point clouds, achieving dynamic and accurate mapping of the mining area's geometric shape and attribute information. Based on the terrain, facilities, and change information in the model, combined with the preset inspection range and target locations, a three-dimensional inspection path that meets safety constraints and is optimal in efficiency is generated through spatial analysis and obstacle avoidance calculation. Flight control commands are calculated based on the waypoint coordinates, altitude, and speed parameters in the path, driving the UAV to autonomously execute the inspection task along the planned path, ultimately realizing the automation, accuracy, and real-time nature of mining area safety inspection, improving the efficiency of hazard identification and operational safety.

[0029] This embodiment utilizes drones combined with digital twin and AI recognition technologies for mine safety monitoring. By integrating LiDAR, multimodal sensors, and deep learning algorithms, it achieves automated inspection and early warning of mine slopes, spoil heaps, tailings ponds, and operational activities. The drone-based automated inspection system features functions such as establishing a real-world mine model, automatic model updates, positioning and navigation, remote control, automatic take-off and landing, automatic inspection, and rapid early warning. It can achieve automated, unmanned, and precise inspection of large-scale mining areas, providing data support for mine slope monitoring, drilling and blasting operations, inspection of major hidden dangers, emergency drills and handling, and anti-illegal operations. It also provides innovative ideas for the transformation from traditional manual inspection modes to an intelligent inspection system based on drones and AI.

[0030] Example 2: As Figure 3 As shown in Example 1, the process of constructing a three-dimensional digital twin model that maintains geometric shape and attribute information synchronized with the physical mining area, provided by this embodiment of the invention, specifically includes the following steps: S101: Access historical archived mining area geographic information data as a spatial reference, and simultaneously access mining area surface images and laser point cloud data transmitted in real time from the UAV platform. Use the laser and vision Multi-SLAM fusion algorithm to jointly solve the two types of real-time data to generate dense three-dimensional point cloud data of the mining area with accurate geographic coordinates. S102: Based on the generated dense 3D point cloud data of the mining area, combined with the terrain trajectory and relative height data recorded by the UAV during real-time terrain-following flight, a high-precision digital elevation model covering the entire mining area is constructed through grid interpolation processing. S103: The high-precision digital elevation model is integrated with historically archived geographic information data of the mining area, and the local areas that have changed are dynamically updated based on the data transmitted back in real time by the UAV platform, ultimately forming a three-dimensional digital twin model that is synchronized with the physical mining area in terms of geometric shape and attribute information.

[0031] The working principle and beneficial effects of the above technical solution are as follows: This embodiment achieves high-precision real-time reconstruction and dynamic synchronization of the three-dimensional scene of the mining area. By fusing laser point clouds, image data, and historical geographic information, and combining Multi-SLAM algorithm with terrain-following flight trajectory, a dense three-dimensional point cloud with accurate geographic coordinates is constructed, and a high-precision digital elevation model is further generated. This ensures that the three-dimensional digital twin model is consistent with the physical mining area in terms of geometry. At the same time, dynamic updates of local areas are achieved through real-time data injection, enabling the model to reflect real-time changes in the surface, facilities, and terrain of the mining area. It supports the fusion of multi-source heterogeneous data and unified spatial reference mapping. Historical archived geographic information data is used as a spatial reference, and spatiotemporal alignment and fusion processing are performed with image, point cloud, and terrain-following flight trajectory data collected in real time by UAVs. This solves the problem of differences in coordinate systems, resolution, and temporal sequence of multi-source data, forming a three-dimensional expression with both global consistency and local detail accuracy, providing a reliable data foundation for spatial analysis and path planning. To enhance the model's adaptability to complex terrain and dynamic environments, the constructed digital elevation model, based on relative altitude data recorded during terrain-following flight and a gridded interpolation method, accurately depicts the undulating terrain, slope morphology, and subtle surface features of the mining area. Combined with real-time data transmission for dynamic correction of changing areas, the model possesses the ability to track dynamic processes such as blasting operations, spoil heap accumulation, and slope displacement, providing high-fidelity 3D scene support for applications such as slope stability monitoring and dynamic reserve management. This embodiment provides a computable 3D environmental foundation for autonomous inspection and intelligent analysis. The resulting digital twin model not only contains geometric and attribute information but also possesses precise spatial relationships and topological structures. This allows the model to be directly used for quantitative analysis such as spatial obstacle avoidance calculations for inspection paths, automatic extraction of slope deformation, and identification and location of hazard areas, achieving a technological leap from static scene reproduction to dynamic intelligent perception.

[0032] Example 3: Based on Example 2, the process of jointly solving two types of real-time data using a laser and vision Multi-SLAM fusion algorithm provided in this embodiment of the invention specifically includes the following steps: S1011: Extract repeatable visual feature points from real-time transmitted surface images of the mining area, and simultaneously extract edge points and planar points representing geometric structures from synchronized laser point cloud data. Associate and map the visual feature points with the edge points and planar points in the laser point cloud through pre-calibrated sensor relative pose parameters to form a joint feature set that integrates visual texture information and laser geometric structure. S1012: Based on the joint feature set of adjacent acquisition times, the relative motion parameters of the UAV platform between each acquisition time are iteratively calculated by calculating the projection matching error of feature points in space. Based on the relative motion parameters, the laser point cloud data of consecutive times are spatially registered and stitched to generate a local dense point cloud sub-map reflecting the current flight area. S1013: Globally match the generated local dense point cloud sub-map with the known geographic landmarks and terrain features contained in the historical archived mining area geographic information data. By adjusting the spatial position and orientation of the sub-map, it achieves the greatest possible geometric consistency with the historical geographic benchmark, and outputs dense 3D point cloud data of the mining area that covers the current flight area and has accurate geographic coordinates.

[0033] The working principle and beneficial effects of the above technical solution are as follows: This embodiment realizes complementary feature fusion and robust pose estimation of multimodal sensor data. By simultaneously extracting the geometric structural features of visual feature points in the image and edge points and planar points in the laser point cloud, and performing correlation mapping based on pre-calibrated sensor parameters, a joint feature set with both texture information and geometric structure is formed. The joint features significantly enhance the ability to identify low-texture areas, dynamic lighting changes, and repetitive structures, thereby improving the robustness and accuracy of relative motion parameter calculation between adjacent time points. It supports high-precision local stitching of continuous point clouds and dynamic scene consistency reconstruction. Based on the relative motion parameters calculated by the joint feature set, the continuously acquired laser point clouds are spatially registered and stitched to generate locally dense point cloud sub-images. By iteratively optimizing the feature projection matching error, the cumulative drift is effectively reduced, ensuring the consistency of the internal geometric structure of the sub-image, and providing accurate local three-dimensional representation of the dynamically changing surface of the mining area. Establishing a globally consistent spatial alignment between real-time data and historical geographic benchmarks involves globally matching local point cloud sub-maps with known landmarks and terrain features in historical archived geographic information. By optimizing the spatial position and orientation of the sub-maps, precise alignment between real-time sensing data and historical geographic benchmarks is achieved. Absolute geographic coordinates are assigned to the point cloud data to eliminate accumulated errors in local reconstruction, ensuring reliable geometric accuracy and spatial traceability of the generated dense 3D point cloud in the global coordinate system. This results in a real-time, dense, and computable 3D point cloud data base suitable for large-scale open-pit mining areas. The final output is a dense 3D point cloud with accurate geographic coordinates, incorporating detailed information from real-time visual and laser data while inheriting the global consistency of historical geographic data. This data base can be directly used for quantitative analysis such as digital elevation model generation, change detection, and volume measurement, providing a high-precision, dynamically updatable 3D spatial foundation for twin model construction and inspection path planning.

[0034] Example 4: Based on Example 3, the process for forming a joint feature set that integrates visual texture information and laser geometry provided in this embodiment of the invention specifically includes the following steps: S10111: Identify and extract repeatable visual feature points from the real-time surface images of the mining area transmitted back by the UAV platform, and record the coordinates of each visual feature point in the image pixel coordinate system; identify and extract geometric edge points representing the edges of mine steps, slope ridges and ore pile outlines, as well as geometric plane points representing flat areas and slopes from the synchronized laser point cloud data of the mining area, and record the three-dimensional spatial coordinates of each edge point and plane point in the lidar coordinate system; S10112: Based on the pre-calibrated relative pose parameters between the visual sensor and the lidar, the three-dimensional spatial coordinates of the extracted edge points and planar points are transformed from the lidar coordinate system to the visual sensor coordinate system; and using the intrinsic parameter model of the visual sensor, the transformed three-dimensional spatial coordinates are mapped to the image pixel plane to calculate the pixel projection coordinates corresponding to each edge point and planar point. S10113: Perform spatial proximity matching between the obtained pixel projection coordinates of each edge point and plane point and the pixel coordinates of the extracted visual feature points, and select edge points or plane points whose pixel distance to the visual feature points is within a set threshold; associate and store the three-dimensional spatial coordinates of the edge points or plane points with the pixel coordinates and texture description information of the corresponding visual feature points to form a joint feature set that simultaneously contains the geometric structure and visual texture information of the mine surface.

[0035] The working principle and beneficial effects of the above technical solution are as follows: This embodiment achieves unified spatial representation and precise association of heterogeneous sensor data. By transforming geometric feature points in the lidar coordinate system to the visual sensor coordinate system and further projecting them onto the image pixel plane, a strict geometric correspondence between the lidar point cloud and the image pixels is established. This eliminates spatial inconsistencies caused by differences in installation position, viewing angle, and coordinate system between sensors, providing a precise mathematical basis for multimodal data fusion. A highly discriminative multimodal joint feature description is constructed by associating and storing the three-dimensional spatial coordinates of geometric feature points with the pixel coordinates and texture description information of visual feature points. This forms a composite feature that simultaneously contains the three-dimensional geometric structure of the mine surface, such as step edges and slope ridges, and two-dimensional visual texture information. The composite feature possesses both geometric invariance and texture discriminability, significantly enhancing the robustness of identifying typical features in the mining area, such as ore piles, slopes, and roads, especially in scenarios where single sensors are prone to failure, such as changes in lighting, dust interference, or low-texture areas. To improve the reliability and efficiency of feature matching, feature matching is performed on the pixel plane based on the spatial proximity criterion. This transforms the feature association problem between 3D point clouds and 2D images into distance calculation in 2D space, significantly reducing matching complexity. By setting a distance threshold to filter reliable association pairs, false matches are effectively eliminated, ensuring the accuracy and consistency of the joint feature set and providing high-quality input for motion estimation and point cloud registration. Stable feature constraints that can be used for high-precision pose estimation are formed. The final generated joint feature set provides both 3D geometric position and 2D texture information, constituting multi-dimensional constraints. In motion parameter calculation, features can be jointly optimized by minimizing reprojection error or geometric alignment error, thereby significantly improving the accuracy and stability of UAV pose estimation and laying a reliable feature foundation for real-time 3D reconstruction of dynamic scenes in mining areas.

[0036] Example 5: Based on Example 3, the process of iteratively calculating the relative motion parameters of the UAV platform between each data acquisition moment provided in this embodiment of the invention specifically includes the following steps: S10121: Select the same-name joint feature unit that contains both visual feature points and associated edge points or planar points from the joint feature set of the previous acquisition time and the joint feature set of the current acquisition time. Record the pixel coordinates of the visual feature points in the images of the two time times in each set of the same-name joint feature unit, as well as the three-dimensional spatial coordinates of the associated edge points or planar points in the lidar coordinate system at the current time. S10122: Based on the pre-set initial motion parameters of the UAV platform, the three-dimensional spatial coordinates of the edge points or planar points in the joint feature unit at the current moment are projected onto the image plane at the previous moment to obtain the corresponding projected pixel coordinates; the projected pixel coordinates are compared with the actual pixel coordinates of the same visual feature points at the previous moment, the distance deviation between the two on the pixel plane is calculated, and the distance deviations of all the same joint feature units are accumulated to obtain the total projection matching error under the current motion parameters; S10123: Adjust the motion parameters of the UAV platform and repeat the calculation process of the total projection matching error until the total projection matching error reaches the minimum value within the set threshold range. Output the corresponding motion parameters as the relative motion parameters of the UAV platform from the previous acquisition time to the current acquisition time for spatial registration and stitching of laser point cloud data.

[0037] The working principle and beneficial effects of the above technical solution are as follows: This embodiment achieves cross-modal data correspondence by associating visual features with edge points or planar points of LiDAR as joint feature units; it constructs a motion parameter optimization model by utilizing the mapping relationship between pixel coordinates and three-dimensional spatial coordinates; it obtains high-precision relative motion estimation by iteratively adjusting motion parameters to minimize projection matching error; and finally achieves spatial consistency registration and seamless stitching of LiDAR point cloud data, improving the positioning and mapping accuracy of multi-source sensor fusion.

[0038] Example 6: Based on Example 5, the process of comparing the projected pixel coordinates with the actual pixel coordinates of the corresponding visual feature points at the previous time step provided by this embodiment of the invention specifically includes the following steps: S101221: From each selected group of joint feature units with the same name, extract the actual pixel coordinates of the visual feature points in the image at the previous time step. At the same time, extract the projected pixel coordinates of the edge points or plane points associated with the visual feature points in the joint feature units with the same name after they are projected onto the image plane at the previous time step through the current motion parameters. Combine the actual pixel coordinates and the projected pixel coordinates to form a pair of pixel coordinates for the deviation to be calculated. S101222: For each pair of pixel coordinates, in the pixel plane of the image at the previous moment, determine the row and column positions of the actual pixel coordinates and the row and column positions of the projected pixel coordinates respectively. Based on the physical scale of the image pixels, calculate the row direction offset of the projected pixel coordinate row position relative to the actual pixel coordinate row position, and the column direction offset of the projected pixel coordinate column position relative to the actual pixel coordinate column position. Combine the row direction offset and the column direction offset geometrically to obtain the straight-line distance of the pixel coordinate pair on the pixel plane. Use the straight-line distance as the pixel deviation value of the joint feature unit under the current motion parameters. S101223: The pixel deviation values ​​of all the calculated joint feature units are summed up. The sum is the total projection matching error of the UAV platform from the previous acquisition time to the current acquisition time under the current set motion parameters. The total projection matching error is used to determine whether the optimal motion parameters have been reached.

[0039] The working principle and beneficial effects of the above technical solution are as follows: This embodiment extracts the actual pixel coordinates and projected pixel coordinates in the same-name joint feature unit to form a pixel coordinate pair for calculating the deviation; in the image pixel plane, the offsets in the row direction and column direction are calculated respectively, and geometrically synthesized into a straight-line distance to obtain the pixel deviation value of each joint feature unit; all pixel deviation values ​​are accumulated to generate the total projection matching error; a direct correlation between motion parameters and pixel-level errors is established, providing a quantifiable objective function for iterative optimization, thereby ensuring that the solution of relative motion parameters can converge to the solution that optimizes the consistency of multimodal feature projection, and finally achieving high-precision spatiotemporal alignment between laser point cloud and visual data.

[0040] Example 7: Based on Example 2, the process of constructing a high-precision digital elevation model covering the entire mining area provided by this embodiment of the invention specifically includes the following steps: S1021: Extract the absolute position coordinates of the UAV at each trajectory point from the terrain-following trajectory data recorded during the real-time terrain-following flight of the UAV, and extract the vertical distance between the UAV at each trajectory point and the surface of the mining area directly below from the synchronously recorded relative altitude data; perform difference calculation between the absolute position coordinates of the UAV and the corresponding vertical distance to calculate the absolute elevation value of the surface point of the mining area directly below each trajectory point, and at the same time use the planar coordinates of each trajectory point as the planar position of the surface point to form a set of discrete surface sampling points distributed along the flight trajectory. S1022: From the generated dense 3D point cloud data of the mining area with accurate geographic coordinates, all point cloud data points are selected, and their planar coordinates are spatially overlaid with the planar coordinates of the obtained discrete surface sampling points. Point cloud data points that are too close to the plane of the sampling points and whose elevation difference exceeds the normal range of the mine steps or slopes are marked as flying points or noise points and removed. The remaining point cloud data points constitute the effective point cloud set of the mining area surface. S1023: Using the obtained discrete surface sampling points as the elevation control framework, and combining them with the obtained effective point cloud set of the mining area, the entire mining area is divided into grids according to the preset grid spacing. Within each grid cell, based on the elevation of the discrete surface sampling points and the effective point cloud elevation in the grid and its neighborhood, the elevation value of the grid node is calculated using an interpolation method that takes into account the steep changes in the mine terrain. Finally, a high-precision digital elevation model that covers the entire mining area and matches the shape of the mine steps, slopes and pits is generated.

[0041] The working principle and beneficial effects of the above technical solution are as follows: This embodiment calculates discrete elevation sampling points on the ground along the flight path by integrating terrain-following flight trajectory data and synchronous relative height measurement; using dense three-dimensional point clouds with accurate geographic coordinates, effective point clouds are obtained by eliminating flying points and noise points through spatial overlay analysis and elevation difference thresholding; using discrete elevation sampling points as the control framework, combined with effective point cloud data, an interpolation method that takes into account the steep changes in mine terrain is used to calculate elevation values ​​on regular grid nodes; the complementarity and verification of multi-source elevation data are realized, effectively suppressing data anomalies, and the generated digital elevation model of the entire mining area can accurately reflect the complex mine terrain features such as steps, slopes and pits, providing a reliable spatial terrain benchmark for mine planning, monitoring and engineering quantity calculation.

[0042] Example 8: Based on Example 7, the process of calculating the elevation values ​​of grid nodes using an interpolation method that takes into account the abrupt changes in mine terrain, provided in this embodiment of the invention, specifically includes the following steps: S10231: Based on the divided regular grid, for each grid node whose elevation value is to be calculated, select all valid point cloud data points within the set neighborhood of the grid node from the obtained set of valid point clouds of the mining area surface; at the same time, select all discrete surface sampling points within the same neighborhood from the obtained set of discrete surface sampling points, and combine the selected valid point clouds and discrete sampling points to form the set of points to be used by the grid node. S10232: Based on the characteristics of the drone's terrain-following flight trajectory reflected by discrete surface sampling points, the planar coordinates of each point in the set of points to be used are compared with the planar projection of the terrain-following trajectory line. Points located near steep terrain feature lines such as the edge line of the mine bench, the top line and bottom line of the slope, and the boundary line of the mining pit are identified. These points are marked as steep terrain feature points and given higher retention weights. The remaining points located on the bench plane or stable slope are marked as gentle terrain feature points and given regular weights. S10233: Based on the assigned conventional weights, points within the same terrain feature unit as the grid node are selected from the set of points to be used according to their weights. Points belonging to the same step plane or the same slope surface as the grid node are selected first. The local terrain trend at the grid node is fitted using the elevation values ​​of these selected points. The final elevation value of the grid node is calculated based on the local terrain trend to ensure that the elevation of the grid node accurately represents the vertical abrupt change at the edge of the mine step and the continuous inclination of the slope surface.

[0043] The working principle and beneficial effects of the above technical solution are as follows: This embodiment constructs a set of points to be used for grid nodes by fusing effective point clouds and discrete sampling points; identifies abrupt terrain change feature points based on the terrain-following trajectory and assigns differentiated weights; and selects points within the same terrain unit based on the weights to fit the local terrain trend, thereby calculating the elevation of the grid nodes. It realizes targeted modeling of abrupt terrain changes such as the edge of the bench, the top line of the slope, the bottom line of the slope, and the boundary line of the mining pit, accurately expressing the vertical change characteristics while preserving the continuity of the terrain, ensuring that the generated digital elevation model can accurately depict the complex geomorphic structure of the mine, and providing a reliable terrain data foundation for slope stability analysis, bench planning, and reserve calculation.

[0044] Example 9: As Figure 4 As shown, based on Example 1, the process of planning and generating a three-dimensional inspection path that satisfies both safety constraints and optimal task efficiency provided by this embodiment of the invention specifically includes the following steps: S201: Extract high-precision digital elevation data of the mining area surface, 3D contour data of permanent buildings and fixed facilities, and real-time data on spoil heap dumping progress, mining face advancement position, and temporary stockpile space occupancy from the 3D digital twin model. All data is uniformly converted into voxelized constraint elements containing spatial coordinates and obstacle attributes. Before writing the voxelized constraint elements into the spatial constraint condition set, the task status identifier issued by the task planning system is first read. If the task status identifier is a regular safety inspection mode, the voxel units corresponding to the blasting warning zone boundary data are marked as forced horizontal avoidance. The path search algorithm treats such voxels as insurmountable rigid obstacles. If the task status is identified as a special clearing mode for blasting operations, the avoidance attribute of the voxel unit corresponding to the blasting warning zone is dynamically revoked, restoring it to a passable state, enabling the path search algorithm to plan a clearing scan route in the airspace within the warning zone. The aforementioned task status identifier is automatically parsed and set by the task planning system according to the instruction type field after receiving the task instruction from the ground control station, and is transmitted through the system's internal communication interface before path planning begins, forming a set of spatial constraints for path search that is dynamically configured according to the task status. S202: Based on the preset inspection task coverage, mark the required waypoint sequence positions in the spatial coordinate system of the three-dimensional digital twin model. At the same time, according to the task target point requirements, extract the three-dimensional coordinates and orientation information of each target point, and construct the waypoint sequence and target points together as a set of necessary nodes for path planning. Combined with the formed set of spatial constraints, perform a search for passable areas based on spatial connectivity within the mining area represented by the three-dimensional digital twin model, and select several candidate spatial polyline segment sequences that can connect all necessary nodes and completely avoid obstacles occupying space. S203: Map the obtained candidate spatial polyline segment sequences back into the 3D digital twin model one by one. Utilize the accurate terrain profile data and facility distribution data provided by the 3D digital twin model to calculate the total spatial length of each candidate sequence. Simultaneously analyze the proximity of each polyline segment to the mine bench slope and its intersection with overhead lines and transportation roads. Select the candidate sequence with the shortest total spatial length that maintains a set safe distance from all fixed facilities and dynamic obstacles in the model. Transform the candidate sequence into a 3D inspection path composed of continuous waypoint coordinates and flight speeds between adjacent waypoints.

[0045] The working principle and beneficial effects of the above technical solution are as follows: This embodiment constructs voxelized spatial constraints by integrating high-precision terrain, fixed facilities and dynamic obstacle data; combined with the necessary nodes defined in the inspection task, it searches and generates candidate polyline segment sequences that connect all nodes and completely avoid obstacles in three-dimensional space; it uses the accurate data of the three-dimensional digital twin model to evaluate the spatial length of each candidate sequence and the safe distance from facilities such as steps, slopes and overhead lines, and finally selects the shortest sequence that meets the safety requirements to convert it into an executable flight path; it realizes the unified constraint expression and efficient path search of multi-source heterogeneous spatial data, and optimizes the spatial efficiency of the inspection path while ensuring that the UAV avoids obstacles throughout the process and maintains a safe distance, thereby generating a three-dimensional optimal inspection path that is both safe and meets the task coverage requirements.

[0046] Example 10: Based on Example 9, the process for calculating the total spatial length of each candidate sequence provided in this embodiment of the invention specifically includes the following steps: S2031: For each selected candidate spatial polyline segment sequence, extract the topographic profile data of the area traversed by the candidate spatial polyline segment sequence from the three-dimensional digital twin model, including the elevation values ​​of each point along the ground projection trajectory of the polyline segment, the position of the step edge, and the points where the slope inclination changes; at the same time, extract the vertical profile data corresponding to each turning point of the polyline segment, and divide the straight segment of the polyline segment in three-dimensional space into several continuous spatial micro-segments according to the topographic undulation characteristics, with the two endpoints of each micro-segment located inside the same topographic unit; S2032: Based on the spatial micro-segments obtained by segmentation, the precise three-dimensional coordinates of the start and end points of each micro-segment are extracted from the three-dimensional digital twin model. Using the spatial geometric properties of the mine bench surface and slope surface provided by the model, the projected length of each micro-segment on the horizontal plane is calculated. At the same time, based on the slope characteristics of the terrain crossed by the spatial micro-segment, the horizontal projected length is converted into the actual slope distance of the micro-segment in three-dimensional space. For spatial micro-segments that cross the edge of the bench, the projected lengths on the upper and lower bench planes are calculated respectively and summed to obtain the spatial length of the spatial micro-segment. S2033: The calculated spatial lengths of each spatial micro-segment are summed up. At the same time, the spatial location data of the overhead lines and transportation road facilities distributed along the candidate polyline segment are extracted from the three-dimensional digital twin model. For spatial micro-segments that intersect with overhead lines, the corresponding vertical detour distance is added based on the line sag data provided by the three-dimensional digital twin model. For spatial micro-segments that cross transportation roads, the corresponding horizontal avoidance distance is added based on the road width. The summed result is combined with the detour avoidance distance to obtain the total spatial length of the candidate sequence.

[0047] The working principle and beneficial effects of the above technical solution are as follows: This embodiment extracts the topographic profile data along the candidate broken line segment and divides it into spatial micro-segments located within the same topographic unit according to the topographic undulation characteristics; based on the precise coordinates and geometric attributes provided by the digital twin model, the actual slant distance length of each micro-segment in three-dimensional space is calculated; after accumulating the lengths of all micro-segments, the overhead line sag and road width data are further integrated to increase the necessary detour and avoidance distances; it realizes the accurate measurement of the three-dimensional spatial path length under complex mining terrain, not only considering the real impact of topographic undulation on path length, but also dynamically incorporating facility avoidance requirements, so that the length calculation results can accurately reflect the safe operation path that the inspection drone needs to perform in actual flight, providing a spatial length basis that conforms to actual operating conditions for path optimization.

[0048] Example 11: Based on Example 10, the process of merging the cumulative summation result with the detour avoidance distance provided in this embodiment of the invention specifically includes the following steps: S20331: From the calculated spatial length of each spatial micro-segment, extract the spatial length data of all spatial micro-segments that are not marked as intersecting with overhead lines or crossing transport roads. Accumulate the spatial lengths of the spatial micro-segments according to their order in the candidate spatial polyline segment sequence to obtain the basic path length data. At the same time, extract the vertical detour distance data corresponding to all spatial micro-segments marked as intersecting with overhead lines, and the horizontal avoidance distance data corresponding to all spatial micro-segments marked as crossing transport roads. Sort the detour and avoidance distance data according to the position of their respective micro-segments in the sequence to form a set of corrected distances to be merged. S20332: Using the obtained basic path length data as the starting value, the first set of detour or avoidance distance data is sequentially taken from the set of corrected distances to be merged and merged with the starting value to obtain the first corrected path length; then the next set of distance data is sequentially taken from the set of corrected distances and merged with the previous corrected path length to obtain the second corrected path length; and so on, until all distance data in the set of corrected distances to be merged has been merged. Each merging operation adds the detour or avoidance distance to the path length of the interval where the spatial micro-segment is located based on the actual position of the micro-segment corresponding to the set of distance data in the sequence. S20333: The final path length data is used as the initial total spatial length of the candidate spatial polyline segment sequence. Mine ground subsidence monitoring point data at the start and end points of the candidate sequence, as well as boundary data of the blasting warning zones along the route, are extracted from the 3D digital twin model. Based on the cumulative ground subsidence at the start and end points, the absolute elevation coordinates (Z-axis coordinates) of the corresponding waypoints are corrected: the subsidence amount is subtracted from the original Z-axis coordinates to obtain corrected waypoint coordinates reflecting the current actual surface elevation, ensuring that the UAV maintains the planned relative flight altitude above the subsided surface. Based on the corrected waypoint 3D coordinates, the spatial length of the flight segment between the starting point and the adjacent intermediate waypoint (and the spatial length of the flight segment between the ending point and the adjacent intermediate waypoint) is recalculated. The spatial length of the affected flight segment is recalculated using the 3D spatial slant distance formula. The original value of the corresponding interval in the initial total spatial length is replaced with the recalculated flight segment length to complete the path length correction caused by ground subsidence. According to the safety distance requirements of the blasting warning zone, the micro-segments passing through the edge of the warning zone are finely adjusted in the horizontal direction. The corrected length data is determined as the total spatial length of the candidate sequence.

[0049] The working principle and beneficial effects of the above technical solution are as follows: This embodiment combines the basic path length with the sequentially arranged detour and avoidance distances to form the initial total spatial length; then, by combining the ground settlement monitoring data of the starting point and the ending point and the safety requirements of the blasting warning zone along the route, the initial length is finely adjusted in the vertical and horizontal directions; the path length calculation is dynamically corrected from basic geometric measurement to actual operational constraints, which not only ensures the accurate reflection of the length of avoidance operations of fixed facilities such as overhead lines and transportation roads, but also further integrates dynamic geological and safety control factors of the mine, so that the final total spatial length reflects the real flight path requirements under multi-dimensional constraints such as terrain undulation, facility avoidance, ground settlement and safety warning, thereby providing a length evaluation benchmark that is closer to the actual operation scenario for path optimization.

[0050] Example 12: Based on Example 11, the process of adding the detour or avoidance distance to the path length of the spatial micro-segment provided in this embodiment of the invention specifically includes the following steps: S203321: From the formed set of distances to be merged and corrected, take out the first set of detour or avoidance distance data in sequence, and extract the position index of the spatial micro-segment in the candidate spatial polyline segment sequence from the example data; based on the position index, locate the start and end positions of the interval corresponding to the spatial micro-segment from the obtained basic path length data, and separate the path length data within the interval from the basic path length data as the interval length of the micro-segment to be processed; S203322: The first set of detour or avoidance distance data is merged with the micro-segment interval length to be processed. For micro-segments that intersect with overhead lines, the vertical detour distance and the spatial micro-segment interval length are spatially geometrically superimposed along the line direction in three-dimensional space to form a new micro-segment interval length containing the vertical detour path. For spatial micro-segments that cross transportation roads, the horizontal avoidance distance and the micro-segment interval length are spatially geometrically superimposed along the road vertical direction in the horizontal plane to form a new micro-segment interval length containing the horizontal avoidance path. The resulting new micro-segment interval length is the actual flight path length of the spatial micro-segment after detour or avoidance processing. S203323: Replace the obtained new micro-segment interval length with the original interval position of the spatial micro-segment in the basic path length data, while keeping the interval lengths of all other micro-segments before and after the spatial micro-segment unchanged, and recombine to form the updated basic path length data; remove the currently processed detour or avoidance distance data from the set of corrected distances to be merged, obtain the reduced corrected distance set, and use the updated basic path length data as the starting value for the next merging operation to process the next set of detour or avoidance distance data.

[0051] The working principle and beneficial effects of the above technical solution are as follows: This embodiment generates a new micro-segment interval length containing the detour or avoidance path by spatially and geometrically superimposing the detour or avoidance distance with the corresponding micro-segment interval length, and replaces the corresponding position in the basic path length with the new micro-segment interval length, while keeping the other intervals unchanged; it realizes the accurate correction of the path length in local intervals, ensuring that the vertical detour at the intersection of overhead lines and the horizontal avoidance at the crossing of roads are geometrically accurate in three-dimensional space, so that the updated path length data can reflect the additional flight distance caused by facility avoidance in actual flight segment by segment, providing a locally accurate and globally consistent length correction basis for the overall path length calculation.

[0052] Example 13: Based on Example 12, the process for forming a new micro-segment interval length including a vertical bypass path provided by this embodiment of the invention specifically includes the following steps: S2033221: Extract the axial spatial coordinate data of the overhead lines intersecting the spatial micro-segment from the three-dimensional digital twin model to determine the spatial intersection point of the spatial micro-segment and the axis; at the same time, extract the three-dimensional coordinates of the start and end points of the micro-segment from the obtained interval length data of the micro-segment to be processed, and divide the straight path of the micro-segment between the original start and end points into the pre-intersection segment and the post-intersection segment with the spatial intersection point as the boundary. S2033222: Based on the extracted vertical detour distance data, extract the sag height data of the overhead line at the intersection and the safety clearance requirement data on both sides of the line from the three-dimensional digital twin model. Set a detour height point that meets the safety clearance requirement at the spatial intersection position along the vertical direction of the line axis. Spatially associate the three-dimensional coordinates of the detour height point with the end point of the section before the intersection and the starting point of the section after the intersection to construct a spatial zigzag line segment from the end point of the section before the intersection through the detour height point to the starting point of the section after the intersection. S2033223: The three-dimensional spatial length of the constructed spatial zigzag segment is merged with the spatial lengths of the segment before and after the intersection. The new spatial length obtained after merging is used to replace the total straight-line length from the starting point to the ending point in the original spatial micro-segment interval length, forming a new micro-segment interval length that includes the vertical bypass path.

[0053] The working principle and beneficial effects of the above technical solution are as follows: This embodiment extracts the axis coordinates and safety clearance data of the overhead line, sets a detour high point that meets safety requirements at the intersection, constructs a spatial zigzag line segment containing the high point, and merges its length with the lengths of the segments before and after the intersection to replace the original straight segment length; it realizes accurate three-dimensional avoidance path modeling of the overhead line intersection area, ensuring that the flight path strictly follows the line sag and safety clearance constraints in the vertical direction, so that the length of the micro-segment interval can accurately reflect the actual spatial distance of the vertical detour trajectory required by the UAV to safely cross the overhead line, thereby providing a local three-dimensional geometric correction that complies with safety regulations for path length calculation.

[0054] Example 14: Based on Example 13, the process for forming a new micro-segment interval length including a horizontal avoidance path provided by this embodiment of the invention specifically includes the following steps: S2033224: Extract the centerline coordinates and road boundary coordinates of the transportation road traversed by the spatial micro-segment from the 3D digital twin model to determine the intersection of the horizontal projection of the micro-segment with the road centerline; at the same time, extract the projection coordinates of the start and end points of the micro-segment on the horizontal plane from the interval length data of the micro-segment to be processed, and divide the horizontal projection line between the original start and end points of the micro-segment into approaching and departing segments with the intersection of the horizontal projections as the boundary. S2033225: Based on the extracted horizontal avoidance distance data, extract the safety distance requirement data outside the road boundary line from the three-dimensional digital twin model, and set a avoidance point that meets the safety requirements by translating it to one side along the vertical direction of the road centerline at the intersection of the horizontal projections. Then, spatially associate the horizontal coordinates of the avoidance point with the end point of the approach section and the start point of the departure section in the horizontal plane to construct a horizontal polygonal line segment from the end point of the approach section through the avoidance point to the start point of the departure section. S2033226: The length of the constructed horizontal plane polyline segment is merged with the actual slant distance lengths of the approach and departure segments on their respective terrain profiles. The new spatial length obtained after merging is used to replace the total slant distance length from the start point to the end point in the original spatial micro-segment interval length, forming a new micro-segment interval length that includes the horizontal avoidance path.

[0055] The working principle and beneficial effects of the above technical solution are as follows: This embodiment extracts the centerline, boundary line and safety distance data of the transportation road, sets a lateral avoidance point that meets the safety requirements at the intersection of the horizontal projection, constructs a horizontal plane polyline segment containing the avoidance point, and merges its length with the actual slant distance length of the approach segment and departure segment on the terrain profile to replace the original micro-segment's full slant distance length; it realizes accurate modeling of the horizontal plane avoidance path in the road crossing area, ensuring that the flight path strictly follows the safety distance constraint outside the road boundary in the horizontal direction. At the same time, it combines the horizontal avoidance trajectory with the actual terrain undulation, so that the length of the micro-segment interval can accurately reflect the actual slant distance length of the horizontal bypass trajectory required by the UAV to safely cross the transportation road in three-dimensional space, thereby providing a local three-dimensional space correction that complies with safety regulations for path length calculation.

[0056] Example 15: As Figure 5 As shown in Example 1, the process of analyzing data collected by autonomous flight inspection missions provided in this embodiment of the invention specifically includes the following steps: S301: From the surface images of the mining area collected by the autonomous flight inspection mission, based on the geographic coordinates and timestamps at the time of acquisition, the image data is spatially registered with the synchronously acquired laser point cloud data to generate mining area real-scene fusion data with three-dimensional spatial coordinates and spectral information; from the mining area real-scene fusion data, by setting the geometric morphological feature parameters of the edges of mine steps, slopes, transportation roads and spoil heaps, the initial contours and location information of key features of the mine are extracted to form a basic feature layer of the mining area containing the spatial distribution of steps, slopes, roads and spoil heaps; S302: Perform overlay analysis on the historical element data of the corresponding area in the constructed 3D digital twin model of the mining area to identify areas in the element layer where the geometry has changed, including the displacement of the step outline, local bulges or depressions on the slope surface, the expansion range of the spoil heap boundary, and the relocation of the transport road; at the same time, based on the visual characteristics of personnel safety helmets, transport vehicle trajectories, and excavation equipment operating postures from real-time acquired surface images of the mining area, identify the spatial position and movement status of workers, transport vehicles, and mining equipment in the mining area's basic element layer, forming dynamic monitoring information of the mining area that includes the spatial coordinates and changes of the changed areas, as well as the position and status of personnel and equipment; S303: Spatially correlate and compare the obtained dynamic monitoring information of the mining area with the preset mine safety thresholds. For areas where the bench displacement exceeds the design range, the local deformation of the slope exceeds the stability limit, or the boundary of the spoil heap intrudes into the warning area, mark their spatial coordinates and changes as areas of reserve change or potential hazard areas. For situations where personnel enter non-working areas, transport vehicles deviate from the prescribed roads, or the safety distance between mining equipment and the slope is insufficient, mark their spatial coordinates and behavioral characteristics as records of illegal operations or safety risk events. Classify and organize the marked areas of reserve change, potential hazard areas, records of illegal operations, and safety risk events to form a key information set for dynamic reserve management, major hazard identification, slope monitoring, blasting operation management, anti-illegal operations, drainage system inspection, spoil heap monitoring, and emergency early warning.

[0057] The working principle and beneficial effects of the above technical solution are as follows: This embodiment generates real-world fusion data of the mining area by fusing real-time acquired images with laser point clouds, extracts key features of the mine to form a basic feature layer, and then performs overlay analysis with historical 3D models to identify geometric changes in surface features and the spatial position and movement of personnel and equipment; furthermore, it compares the monitoring information with preset safety thresholds spatially to automatically label and classify areas of excessive changes and violations; it realizes automated and quantitative monitoring of changes in surface features and production activities in the mining area, and can systematically identify changes in reserves, potential hazards, violations, and safety risk events, providing integrated key information support for dynamic management of mine reserves, judgment of major hazards, slope monitoring, blasting operation management, anti-violation of three violations, inspection of drainage systems, monitoring of spoil heaps, and emergency early warning, thereby improving the refinement and intelligence of mine safety supervision and production management.

[0058] The specific application scenarios in this embodiment include: Dynamic storage management, based on dynamic scanning and point cloud volume calculation, quickly calculates changes in storage capacity, optimizes transportation distance and real-time cost accounting, including blasting costs, daily costs, weekly costs, and monthly costs, reducing traditional metering and weighing processes and saving transportation costs.

[0059] The identification of major hidden dangers and the early warning of hidden disasters are based on dynamic scanning and AI analysis. The mining area is dynamically scanned before the start of work every day to plan the mining area and identify goafs or karst caves that threaten the safety of open-pit mining, and early warning is issued.

[0060] Identifying major hidden dangers involves identifying the presence of equipment, materials, or processes that are prohibited by national regulations. AI analysis is used to identify such prohibited equipment, materials, or processes, preventing obsolete equipment and processes from entering the mining area. Prohibited equipment, materials, or processes include: pot blasting, bottom caving, undercutting, single-wall mining without stratification, secondary crushing of large ore blocks using blasting, medium-deep hole drilling equipment without pressure stabilization devices, manual loading and unloading of ore during centralized shoveling operations, and dry drilling operations without dust collection devices.

[0061] Identifying major hidden dangers involves determining whether there has been any failure to strictly adhere to a top-down mining sequence, such as step-by-step or layer-by-layer mining. This is done based on dynamic scanning and point cloud volume calculation.

[0062] The identification of major hidden dangers is based on dynamic scanning and point cloud volume calculation. The number and width of the working platform are calculated, and it is determined whether the working slope angle is greater than the design working slope angle, or whether the final slope step height exceeds the design height.

[0063] Identification of major hidden dangers, AI analysis and early warning or electronic fences, and judgment of mining or destruction of ore or rock pillars or ore bodies that are required to be preserved by design.

[0064] The identification of major hidden dangers is based on dynamic scanning and AI analysis and early warning, such as the appearance of transverse and longitudinal radial cracks on the slope; upward bulging or protrusion at the leading edge and toe of the slope; and rapid expansion of cracks at the trailing edge. This is superior to traditional slope radar, which can only monitor slopes from one dimension or the radar illumination surface.

[0065] The identification of major hidden dangers is based on dynamic scanning and point cloud computing to determine whether the road slope exceeds the design standard, match the optimal transportation route, and avoid transportation accidents caused by excessively steep road slopes.

[0066] The identification of major hidden dangers is based on dynamic scanning and AI analysis and early warning. It is determined whether the open-pit mine has constructed flood control or drainage ditches and flood discharge facilities as designed, and whether there are any damages or leaks in their quantity and operation.

[0067] The identification of major hidden dangers is based on dynamic scanning and AI analysis and early warning. There are densely populated areas within twice the total height of the spoil heap; whether interception and drainage facilities have been built around the hillside spoil heap as designed.

[0068] The identification of major hidden dangers is based on dynamic scanning and point cloud computing to determine whether the open-pit mine has set up safety platforms and cleaning platforms as designed, and whether the number or width of the steps meets the specified width.

[0069] The identification of major hidden dangers is based on dynamic scanning and AI analysis and early warning, to determine whether there is unauthorized backfilling operations at the waste dump.

[0070] The identification of major hidden dangers is based on dynamic scanning and AI analysis and early warning, including situations where open-pit mines fail to stop operations and evacuate on-site personnel in a timely manner during extreme weather.

[0071] Slope monitoring is crucial because open-pit mine slopes are prone to landslides and collapses. Manual inspections are difficult to detect even minor cracks and pose a high risk. Traditional sensors are costly to deploy and have limited coverage.

[0072] This invention utilizes a drone equipped with lidar and a high-resolution camera to collect point cloud data through terrain-following flight, constructing a millimeter-level real-scene 3D model; it automatically compares digital twin models from different time periods to identify whether the slope has transverse / longitudinal cracks, slope toe bulges, or canopy structures, thereby achieving quantitative monitoring of slope stability and disaster early warning.

[0073] Based on dynamic scanning and point cloud computing, it was determined that: ① Umbrella rock and turquoise; ② The slope showed transverse and longitudinal radial cracks; ③ The slope foot showed an upward bulge, and the cracks at the rear edge expanded rapidly.

[0074] For blasting operations, pre-blast site clearance confirmation is difficult, and post-blast manual measurement of large block ratio and flyrock range is not only inefficient but also carries the risk of unexploded detonators. Before blasting, the task planning system switches the task status to a blasting-specific clearance mode. The path planning module dynamically removes the voxel-based avoidance constraints of the blasting warning zone based on this mode, incorporating the airspace within the warning zone into the passable area, and replans a scanning route covering the entire area within the warning line. The drone enters the warning zone airspace along the planned clearance route and performs a full-field scan of personnel and machinery within the warning line using AI target detection, ensuring that a confirmation signal is transmitted back to the ground control station after clearance is completed. After the clearance task is completed, the task planning system restores the task status to the regular safety inspection mode, and the voxel-based avoidance constraints of the blasting warning zone are reactivated. After blasting, a 3D model of the blast pile is instantly generated using laser point clouds, automatically analyzing the large block ratio, flyrock range, and blast pile shape to assist in optimizing subsequent blasting parameter design and achieve accurate calculation of blasting costs and effects.

[0075] AI-powered intelligent identification and on-site supervision of illegal operations in mining areas: Mining areas are large and work sites are scattered, making it impossible for supervisors to monitor every work site in real time. This makes it difficult to eradicate illegal commands, unauthorized operations, violations of labor discipline, and other illegal activities. Drones automatically patrol along predetermined routes, utilizing deep learning algorithms integrated into the intelligent analysis platform to identify in real time whether workers are wearing safety helmets, whether vehicles are illegally cutting in line, and whether anyone has accidentally entered blasting warning zones or special control areas. Once a violation is detected, the system immediately captures evidence and remotely pushes warning information to the management terminal.

[0076] During extreme weather and flood season, drainage systems are prone to blockages, and spoil heaps and tailings ponds are susceptible to landslides. Manual inspections are inefficient and extremely dangerous in severe weather. The system is linked to weather warnings and automatically dispatches drones for specialized inspections after rain. Infrared and high-definition cameras are used to identify siltation, blockages, or leaks in drainage ditches, and to monitor the height of spoil heaps and tailings ponds, as well as slope subsidence.

[0077] Inspection of spoil heaps: 1) The amount of spoil heaped and the elements of the heap; 2) The stability of the spoil heap, whether there are cracks, subsidence, bulges, collapses, or landslides on the slope; 3) Whether the toe of the slope has been eroded, hollowed out, or overloaded; 4) Whether there is any random work, stay, or passage at the bottom or front of the slope; 5) Whether the top intercepting ditch is unobstructed and whether it has collapsed or blocked; 6) Whether the bottom drainage blind ditch and seepage well are functioning properly; 7) Whether there is any water accumulation, seepage, or muddy water flowing out after rain.

[0078] Identification of major hidden dangers: There are activities such as mining, excavation, and blasting in the tailings pond area or on the tailings dam that endanger the safety of the tailings pond.

[0079] Identification of major hidden dangers includes: severe piping, soil deformation, etc. in the dam body; through cracks, collapse, and sliding signs in the dam body; large-area longitudinal cracks in the dam body, with a large area of ​​seepage water escaping at high levels or large-area swamp formation.

[0080] The system identifies major hidden dangers, automatically calculates the slope of the dam body, and determines whether the average external slope ratio of the dam body or the external slope ratio of the sub-dam is steeper than the design slope ratio.

[0081] The identification of major hidden dangers involves automatically identifying and measuring the height of the dam and the area of ​​the tailings dam. If the height of the dam exceeds the design total dam height, or if the tailings dam exceeds the design capacity for storing tailings, it is considered a major hidden danger.

[0082] The system automatically identifies major hidden dangers, including the length of the dry beach and the elevation of the water level in tailings dams. This is indicated if the flood control height and dry beach length of a wet tailings dam are less than the design values, or if the flood control height and flood control width of a dry tailings dam are less than the design values.

[0083] The system can identify major hidden dangers, automatically identify particles of materials discharged from tailings ponds, and identify tailings and waste materials entering the pond that are not designed into the system.

[0084] The system can identify major hidden dangers, automatically identify the water flow around the tailings dam, and identify wastewater entering the dam that is not designed for it.

[0085] Emergency warnings, information releases, and urgent evacuation and site clearance are required in the following situations: 1. When a yellow rainstorm warning is issued by the weather forecast; 2. When significant changes in geological structures are discovered, or when unexplored underground goaf areas, abandoned tunnels, or karst passages are found; 3. When accelerated changes are observed in the surface or internal displacement monitoring of the mining slope or spoil heap; 4. When there are obvious cracks on the mining slope, or when there are loose rocks, dangerous rocks, or overhanging structures on the slope; 5. When non-uniform settlement occurs in the spoil heap, or when the foundation heaves upward; 6. When large areas of the mining transport roads are icy, or when visibility is less than 150 meters in rainy or foggy weather.

[0086] Example 16: As Figure 6 As shown, based on Examples 1-15, the open-pit mine safety inspection system based on a digital twin model provided in this embodiment of the invention includes: The drone inspection platform is used to carry recording equipment to carry out flight inspection tasks. It can realize functions such as human-machine interaction, remote visual monitoring, voice intercom, dynamic recognition and data storage. The unmanned aerial vehicle (UAV) system is used to control the flight of the UAV, collect environmental data in the mining area, and transmit data to the ground control station. It includes an airport, sensors, data transmission, and a ground control station. The ground control station is used to send inspection commands and receive and display inspection data. The mission planning system is used to execute functions such as UAV trajectory, inspection time, inspection frequency and sensor type according to inspection instructions, and to generate flight control instructions based on the three-dimensional inspection path to drive the UAV platform to fly autonomously along the planned path. The 3D digital twin model generation module is used to construct and update a 3D digital twin model synchronized with the physical mining area based on historical and real-time mining area geographic information data; and to plan the optimal 3D inspection path for the UAV platform based on the 3D digital twin model and inspection task requirements. The intelligent recognition module is used to analyze real-time acquired image data, identify inspection targets of preset categories, and generate recognition results containing target location and category information; it also analyzes data collected by autonomous flight inspection missions, extracts key information, and uses it for dynamic management of reserves, identification and judgment of major hidden dangers, slope stability monitoring, management of blasting operations, prevention of illegal construction, inspection of drainage systems, dynamic monitoring of spoil heaps, and emergency early warning.

[0087] The working principle and beneficial effects of the above technical solution are as follows: The data application end includes dynamic reserve management, major hidden danger identification, slope monitoring, blasting operations, anti-illegal construction, drainage system inspection, spoil heap inspection, and emergency early warning. This embodiment achieves intelligent safety inspection with a closed-loop process by integrating a UAV inspection platform, a UAV system, a mission planning system, a 3D digital twin model generation module, and an intelligent recognition module. The UAV inspection platform has human-computer interaction, remote visualization, voice intercom, dynamic recognition, and data storage functions, realizing visualized monitoring and real-time interaction of the inspection process, improving on-site response capabilities and operational convenience. The UAV system realizes flight control, environmental data acquisition, and real-time feedback through airports, sensors, data transmission units, and ground control stations, ensuring the continuity and reliability of data acquisition and providing a data source for model updates and identification analysis. The mission planning system automatically generates flight control commands that meet safety constraints and are optimally efficient based on inspection instructions and the 3D digital twin model, enabling the UAV to fly autonomously along the planned path, reducing manual intervention and improving the consistency and repeatability of inspection operations. The 3D digital twin model generation module integrates historical and real-time geographic information data to construct and dynamically update a 3D model synchronized with the physical mining area. This provides a high-precision spatial benchmark for path planning and situation analysis, supporting visualized monitoring and change tracking of the mining area's status. The intelligent recognition module automatically analyzes real-time acquired image data, identifies pre-defined inspection targets, and generates structured results containing location and category information, enabling early detection and rapid location of anomalies.

[0088] In summary, this embodiment achieves automation, precision, and intelligence in mine area inspections through data fusion, model-driven approaches, and intelligent analysis, thereby improving the ability to identify safety risks, the efficiency of inspection operations, and the level of decision support, forming a complete technical closed loop from data collection, processing, analysis to feedback.

[0089] In this embodiment, the aircraft adopts an industrial-grade quadcopter drone system with a diagonal of 498.5 mm. The fuselage is a single-piece molded structure made of carbon fiber. It utilizes a high-performance, high-efficiency power system, coupled with a folding carbon fiber propeller design, offering advantages such as high structural strength and light weight. The power battery is housed inside the aircraft fuselage, which helps maintain battery warmth. The drone can be equipped with an infrared / visible light gimbal pod, built-in WIFI / 4G and long-range high-speed image / data transmission links, hovering accuracy: vertical: ±0.1 meters; horizontal: ±0.1 meters. It features built-in high-precision RTK dual-antenna positioning / orientation, forward-looking and downward-looking active obstacle avoidance, manual remote control and automatic flight path functions, takeoff from airport A and landing at airport B, supports remote control and equipment management, and fully automatic takeoff and landing operations.

[0090] Sensors: ① The drone payload uses image sensors: Wide-angle camera: 4 / 3 CMOS, 20 million effective pixels, DFOV: 84°, equivalent focal length: 24mm, aperture: f / 2.8 to f / 11, focus point: 1m to infinity; Medium telephoto camera: 1 / 1.3-inch CMOS, 8 million effective pixels, DFOV: 35°, equivalent focal length: 70mm, aperture: f / 2.8, focus point: 3m to infinity; Telephoto camera: 1 / 1.5-inch CMOS, 48 million effective pixels, DFOV: 15°, equivalent focal length: 168mm, aperture: f / 2.8, focus point: 3m to infinity. ② Mapping camera range: 80m@10% reflectivity; 200m@54% reflectivity; 300m@90% reflectivity; accuracy: elevation <5cm@100m; LiDAR parameters: point frequency 1,920,000 points / second (triple echo); Inertial navigation system parameters: heading accuracy 0.038°; attitude accuracy 0.008°; IMU data frequency 200Hz. Thermal imaging camera: -40℃ to 150℃, sensitivity ≤50mk@F1.0, ambient temperature 25℃.

[0091] The airport primarily facilitates the autonomous takeoff and landing of drones. Upon receiving a flight mission, the automated takeoff and landing platform opens its doors, allowing the drone to take off and enter autonomous cruise mode according to a pre-planned flight path from the ground station. After completing its cruise mission, the drone autonomously lands on the platform. The automated takeoff and landing platform features a metal frame roof and a double-sided opening design. It consists of a drone protection canopy, equipment bay, waterproof platform, and external equipment. Takeoff and landing markers are installed inside the platform. Internally, it is equipped with an AC / DC conversion system for powering the platform's motors and charging the drones; a remote data transmission link and a local fiber optic / 4G transmission system for communication and control between the drones, the platform, and the remote monitoring center; and Wi-Fi, data transmission, and image transmission equipment for communication with the drones, enabling and disabling charging, and other short-range communication. Cameras, meteorological equipment, and an RTK base station are installed both internally and externally. Fixed cameras operate continuously 24 / 7, protecting the platform while monitoring the surrounding area. Meteorological equipment is mainly used for drone takeoffs and landings and for collecting local meteorological data, such as wind direction, wind speed, rainfall, and temperature. The RTK base station is used for drone differential positioning services, enabling centimeter-level takeoffs and landings. All external devices are connected to the airport's internal communication network via routers, aggregating data to the remote control center.

[0092] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of equivalents of this invention, this invention is also intended to include these modifications and variations.

Claims

1. A safety inspection method for open-pit mines based on a digital twin model, characterized in that, The process includes the following steps: By integrating historical geographic archives with real-time imagery and point cloud data collected by UAVs, a three-dimensional digital twin model that is synchronized in real time with the geometry and attributes of the physical mine is established; Using the terrain, facilities, and dynamic change information contained in the three-dimensional digital twin model, spatial analysis and obstacle avoidance calculations are performed in combination with the mission objectives to automatically generate the globally optimal inspection path; The planned optimal path is converted into flight control commands that can be executed by the UAV to drive the UAV to perform the inspection. Analyze the data collected during autonomous flight inspection missions and extract key information.

2. The open-pit mine safety inspection method based on a digital twin model as described in claim 1, characterized in that, By integrating historical archived geographic information data of the mining area with real-time surface images and laser point cloud data transmitted from the UAV platform, and through fusion processing and dynamic updates, a three-dimensional digital twin model that maintains geometric shape and attribute information synchronization with the physical mining area is constructed. Based on the topography, facility distribution and dynamic change information contained in the three-dimensional digital twin model, and combined with the preset inspection task coverage and target point requirements, a three-dimensional inspection path that meets safety constraints and optimizes task efficiency is planned and generated through spatial analysis and obstacle avoidance calculation.

3. The open-pit mine safety inspection method based on a digital twin model as described in claim 2, characterized in that, The process of planning and generating a 3D inspection path that satisfies both safety constraints and optimal task efficiency includes the following steps: High-precision digital elevation data of the mining area surface, 3D contour data of permanent buildings and fixed facilities, and real-time data on the progress of spoil heaps, the advancing position of mining faces, and the space occupancy of temporary stockpiles are extracted from the 3D digital twin model. The data is uniformly converted into voxel-based constraint elements containing spatial coordinates and obstacle attributes. When writing the voxel-based constraint elements into the spatial constraint condition set, the constraint elements are configured hierarchically according to the task status identifier issued by the task planning system: when the task status identifier is the regular safety inspection mode, the boundary data of the blasting warning zone is written into the set as a mandatory horizontal avoidance constraint, and the UAV must not enter the warning zone airspace during the path search process; when the task status identifier is the blasting operation special clearing mode, the voxel-based avoidance constraint corresponding to the blasting warning zone is dynamically canceled, and only the physical obstacle voxel constraint is retained, allowing the path planning to include the airspace inside the warning zone into the passable area, so as to support the UAV to perform full-field scanning and personnel clearing tasks within the warning line; Form a set of spatial constraints for path search that are dynamically configured according to the task state; Based on the preset inspection task coverage, the required waypoint sequence positions are marked in the spatial coordinate system of the three-dimensional digital twin model. At the same time, according to the task target point requirements, the three-dimensional coordinates and orientation information of each target point are extracted. The waypoint sequence and target points are jointly constructed into a set of necessary nodes for path planning. Combined with the formed set of spatial constraints, a search for passable areas based on spatial connectivity is performed within the mining area represented by the three-dimensional digital twin model. Several candidate spatial polyline segment sequences that can connect all necessary nodes and completely avoid obstacles occupying space are selected. The obtained candidate spatial polyline segment sequences are mapped back to the three-dimensional digital twin model one by one. Using the accurate terrain profile data and facility distribution data provided by the three-dimensional digital twin model, the total spatial length of each candidate sequence is calculated. At the same time, the proximity between each polyline segment and the mine bench slope, as well as the intersection with overhead lines and transportation roads, are analyzed. The candidate sequence with the shortest total spatial length and which maintains a set safe distance from all fixed facilities and dynamic obstacles in the model is selected. The candidate sequence is then transformed into a three-dimensional inspection path composed of continuous waypoint coordinates and flight speeds between adjacent waypoints.

4. The open-pit mine safety inspection method based on a digital twin model as described in claim 3, characterized in that, The process of calculating the total spatial length of each candidate sequence includes the following steps: For each selected candidate spatial polyline segment sequence, the topographic profile data of the area traversed by the candidate spatial polyline segment sequence is extracted from the three-dimensional digital twin model; at the same time, the vertical profile data corresponding to the point is extracted at each turning point of the polyline segment, and the straight line segment of the polyline segment in three-dimensional space is divided into several continuous spatial micro-segments according to the topographic undulation characteristics, and the two endpoints of each micro-segment are located inside the same topographic unit. Based on the spatial micro-segments obtained from segmentation, the precise three-dimensional coordinates of the start and end points of each micro-segment are extracted from the three-dimensional digital twin model. Using the spatial geometric properties of the mine bench surface and slope surface provided by the model, the projected length of each micro-segment on the horizontal plane is calculated. At the same time, based on the slope characteristics of the terrain crossed by the spatial micro-segment, the horizontal projected length is converted into the actual slope distance of the spatial micro-segment in three-dimensional space. For spatial micro-segments that cross the edge of the bench, their projected lengths on the upper and lower bench planes are calculated separately and summed to obtain the spatial length of the spatial micro-segment. The calculated spatial lengths of each spatial micro-segment are summed up. At the same time, the spatial location data of the overhead lines and transportation road facilities distributed along the candidate polyline segments are extracted from the three-dimensional digital twin model. For spatial micro-segments that intersect with overhead lines, the corresponding vertical detour distance is added based on the line sag data provided by the three-dimensional digital twin model. For spatial micro-segments that cross transportation roads, the corresponding horizontal avoidance distance is added based on the road width. The summed result is combined with the detour avoidance distance to obtain the total spatial length of the candidate sequence.

5. The open-pit mine safety inspection method based on a digital twin model as described in claim 4, characterized in that, The process of combining the summation result with the detour distance includes the following steps: From the calculated spatial length of each spatial micro-segment, extract the spatial length data of all spatial micro-segments that are not marked as intersecting with overhead lines or crossing transport roads. Accumulate the spatial lengths of the spatial micro-segments according to their order in the candidate spatial polyline segment sequence to obtain the basic path length data. Simultaneously, extract the vertical detour distance data corresponding to all spatial micro-segments marked as intersecting with overhead lines, and the horizontal avoidance distance data corresponding to all spatial micro-segments marked as crossing transport roads. Sort the detour and avoidance distance data according to the position of their respective micro-segments in the sequence to form a set of corrected distances to be merged. Using the obtained basic path length data as the starting value, the first set of detour or avoidance distance data is taken out sequentially from the set of corrected distances to be merged, and merged with the starting value to obtain the path length after the first correction. Then, take the next set of distance data from the corrected distance set in order and merge it with the path length after the previous correction to obtain the path length after the second correction. This process continues until all distance data in the corrected distance set to be merged has been merged. Each merging operation adds the detour or avoidance distance to the path length of the interval where the spatial micro-segment is located, based on the actual position of the micro-segment corresponding to the distance data in the sequence. The final path length data is used as the initial total spatial length of the candidate spatial polyline segment sequence. Mine ground subsidence monitoring point data at the start and end points of the candidate sequence, as well as boundary data of the blasting warning zones along the route, are extracted from the 3D digital twin model. Based on the cumulative ground subsidence at the start and end points, the absolute elevation coordinates of the corresponding waypoints are corrected by subtracting the subsidence from the original Z-axis coordinates to obtain corrected waypoint coordinates reflecting the current actual surface elevation. Using the corrected 3D waypoint coordinates as a basis, the spatial length of the flight segment between the start point and adjacent waypoints, and between the end point and adjacent waypoints, are recalculated. The recalculated segment lengths replace the original values ​​of the corresponding intervals in the initial total spatial length, completing the path length correction caused by ground subsidence. Based on the safety distance requirements of the blasting warning zone, the micro-segments crossing the edge of the warning zone are finely adjusted horizontally. The corrected length data is then determined as the total spatial length of the candidate sequence.

6. The open-pit mine safety inspection method based on a digital twin model as described in claim 5, characterized in that, The process of adding the detour or avoidance distance to the path length of the interval containing the spatial micro-segment includes the following steps: From the formed set of distances to be merged and corrected, the first set of detour or avoidance distance data is taken out in sequence. At the same time, the position index of the spatial micro-segment in the candidate spatial polyline segment sequence is extracted from the example data. Based on the position index, the start and end positions of the interval corresponding to the spatial micro-segment are located from the obtained basic path length data. The path length data within the interval is separated from the basic path length data and used as the interval length of the micro-segment to be processed. The first set of detour or avoidance distance data is merged with the micro-segment interval length to be processed. For micro-segments that intersect with overhead lines, the vertical detour distance and the spatial micro-segment interval length are spatially geometrically superimposed along the line direction in three-dimensional space to form a new micro-segment interval length that includes the vertical detour path. For spatial micro-segments that cross transportation roads, the horizontal avoidance distance and the micro-segment interval length are spatially geometrically superimposed along the road vertical direction in the horizontal plane to form a new micro-segment interval length that includes the horizontal avoidance path. The resulting new micro-segment interval length is the actual flight path length of the spatial micro-segment after detour or avoidance processing. The obtained new micro-segment interval length is replaced with the original interval position of the spatial micro-segment in the basic path length data, while keeping the interval lengths of all other micro-segments before and after the spatial micro-segment unchanged, and the data is recombined to form the updated basic path length data. The currently processed detour or avoidance distance data is removed from the set of corrected distances to be merged, resulting in a reduced set of corrected distances. The updated basic path length data is used as the starting value for the next merging operation to process the next set of detour or avoidance distance data.

7. The open-pit mine safety inspection method based on a digital twin model as described in claim 6, characterized in that, The process of forming a new micro-segment interval length that includes a vertical bypass path includes the following steps: The spatial coordinate data of the axis of the overhead line intersecting the spatial micro-segment is extracted from the three-dimensional digital twin model to determine the spatial intersection point of the spatial micro-segment and the axis; at the same time, the three-dimensional coordinates of the start and end points of the micro-segment are extracted from the obtained interval length data of the micro-segment to be processed, and the straight path of the micro-segment between the original start and end points is divided into the pre-intersection segment and the post-intersection segment with the spatial intersection point as the boundary. Based on the extracted vertical detour distance data, the sag height data of the overhead line at the intersection and the safety clearance requirements on both sides of the line are extracted from the three-dimensional digital twin model. A detour height point that meets the safety clearance requirements is set vertically upward along the line axis at the spatial intersection position. The three-dimensional coordinates of the detour height point are spatially correlated with the end point of the section before the intersection and the starting point of the section after the intersection to construct a spatial zigzag line segment from the end point of the section before the intersection through the detour height point to the starting point of the section after the intersection. The three-dimensional spatial length of the constructed spatial zigzag segment is merged with the spatial lengths of the segment before and after the intersection. The new spatial length obtained after merging is used to replace the total straight-line length from the starting point to the ending point in the original spatial micro-segment interval length, forming a new micro-segment interval length that includes the vertical bypass path.

8. The open-pit mine safety inspection method based on a digital twin model as described in claim 6, characterized in that, The process of forming a new micro-segment interval length that includes the horizontal avoidance path includes the following steps: The coordinate data of the centerline of the transportation road crossed by the spatial micro-segment and the coordinate data of the road boundary line are extracted from the three-dimensional digital twin model to determine the intersection of the horizontal projection of the micro-segment and the centerline of the road. At the same time, the projection coordinates of the start and end points of the micro-segment on the horizontal plane are extracted from the length data of the micro-segment to be processed. The horizontal projection line between the original start and end points of the micro-segment is divided into the approach segment and the departure segment with the intersection of the horizontal projection as the boundary. Based on the extracted horizontal avoidance distance data, the safety distance requirement data outside the road boundary line is extracted from the three-dimensional digital twin model. At the intersection of the horizontal projections, a avoidance point that meets the safety requirements is set by translating to one side along the vertical direction of the road centerline. The horizontal coordinates of the avoidance point are spatially associated with the end point of the approach section and the start point of the departure section in the horizontal plane to construct a horizontal polygonal line segment from the end point of the approach section through the avoidance point to the start point of the departure section. The length of the constructed horizontal plane polyline segment is merged with the actual slant distance lengths of the approach and departure segments on their respective terrain profiles. The new spatial length obtained after merging is used to replace the total slant distance length from the start point to the end point in the original spatial micro-segment interval length, forming a new micro-segment interval length that includes the horizontal avoidance path.

9. The open-pit mine safety inspection method based on a digital twin model as described in claim 1, characterized in that, When generating inspection tasks, the task planning system automatically sets the task status identifier based on the type of inspection instruction currently issued: for routine tasks such as daily slope inspection, facility inspection, and anti-illegal inspection, the task status identifier is set to the routine safety inspection mode; for tasks such as clearing and confirming the warning zone before blasting operations, the task status identifier is set to the special clearing mode for blasting operations; the task status identifier is transmitted to the spatial constraint condition set construction stage of the 3D digital twin model generation module before path planning, as an input parameter for the hierarchical configuration of voxel constraints; based on the waypoint coordinates, flight altitude, and speed parameters in the 3D inspection path, the system calculates and generates flight control commands including attitude adjustment and heading control, and sends them to the UAV inspection platform through the communication unit, driving it to perform autonomous flight inspection tasks along the planned path; Data collected from autonomous flight inspection missions is analyzed to extract key information for dynamic management of reserves, identification and judgment of major hidden dangers, slope stability monitoring, management of blasting operations, prevention of illegal construction, inspection of drainage systems, dynamic monitoring of spoil heaps, and emergency early warning.

10. A safety inspection system for open-pit mines based on a digital twin model, characterized in that, Include The drone inspection platform is used to carry recording equipment to carry out flight inspection tasks. It can realize human-computer interaction, remote visual, voice intercom, dynamic recognition and data storage functions. The unmanned aerial vehicle (UAV) system is used to control the flight of the UAV, collect environmental data in the mining area, and transmit data to the ground control station. It includes an airport, sensors, data transmission, and a ground control station. The ground control station is used to send inspection commands and receive and display inspection data. The mission planning system is used to execute functions such as UAV trajectory, inspection time, inspection frequency and sensor type according to inspection instructions, and to generate flight control instructions based on the three-dimensional inspection path to drive the UAV platform to fly autonomously along the planned path. The 3D digital twin model generation module is used to construct and update a 3D digital twin model synchronized with the physical mining area based on historical and real-time mining area geographic information data; and to plan the optimal 3D inspection path for the UAV platform based on the 3D digital twin model and inspection task requirements. The intelligent recognition module is used to analyze the real-time acquired image data, identify the inspection targets of the preset categories, and generate recognition results containing the target location and category information; Data collected from autonomous flight inspection missions is analyzed to extract key information for dynamic management of reserves, identification and judgment of major hidden dangers, slope stability monitoring, management of blasting operations, prevention of illegal construction, inspection of drainage systems, dynamic monitoring of spoil heaps, and emergency early warning.

Images