A mine scene unmanned aerial vehicle and robot dog cooperative operation system and method

Abstract

This invention discloses a collaborative operation system and method for a drone and a robotic dog in a mining scenario. The system includes a drone, a robotic dog, and a wireless communication module. The drone and the robotic dog transmit data through the wireless communication module. The drone includes a laser terrain scanner, and the robotic dog is equipped with a high-definition camera and an infrared sensor. The laser terrain scanner is used to perform a global terrain scan of the mining blasting area to obtain laser point cloud data. The high-definition camera is used to identify obstacle elevation values, and the external sensors are used to collect road slope data. This invention utilizes a bidirectional BFS search algorithm combined with adjacent grid selection weights and priorities to generate multiple planned paths. The drone sends the planned paths to the robotic dog in advance, and the robotic dog only needs to select the optimal path based on the actual road conditions. The lightweight MobileNet model improves the robotic dog's computation speed and time, and the selection of the optimal path based on actual road conditions requires low hardware computing power.

Description

Technical Field

[0001] This application relates to the field of intelligent mining operations, and in particular to a collaborative operation system and method of drones and robot dogs in mining scenarios. Background Technology

[0002] With the development of intelligent mining, the collaborative operation of drones and robot dogs has been widely used in mining blasting and other scenarios, becoming an important way to replace manual labor in high-risk and tedious operations, effectively avoiding the safety risks of manual labor in complex mining environments and improving the convenience of operations.

[0003] In existing collaborative operation technologies, drones mainly undertake tasks such as global terrain scanning and aperture planning, while robot dogs are responsible for specific operations such as inspection, verification, cleaning, and transportation near the aperture points. The path planning of robot dogs mostly relies on the road condition data collected by their own cameras to be completed in real time, and often uses complex algorithms such as deep learning and edge computing. Such algorithms have extremely high requirements for the hardware computing power of robot dogs. If the hardware configuration is insufficient, problems such as slow computing speed and delayed path planning are likely to occur, affecting the efficiency of operation.

[0004] At the same time, due to the large number of robot dogs used, if each robot dog is equipped with high-performance hardware to meet the algorithm's operation requirements, or if each robot dog independently and dynamically adjusts its path, it will significantly increase the costs of equipment procurement, operation and maintenance, and technology implementation. This will result in excessively high costs for intelligent mining operations, making it difficult to promote and apply on a large scale and fully leverage the advantages of collaborative operations between drones and robot dogs.

[0005] There is an urgent need for a collaborative operation solution that can address the difficulty in simultaneously satisfying the aforementioned cost and efficiency requirements of drones and robotic dogs. Summary of the Invention

[0006] The purpose of this invention is to propose a collaborative operation system and method for drones and robotic dogs in mining scenarios to solve the technical problems mentioned in the background section above.

[0007] The first aspect of this invention proposes a collaborative operation system of drones and robotic dogs in a mining scenario, comprising: The system includes a drone, a robot dog, and a wireless communication module. The drone and the robot dog transmit data through the wireless communication module. The drone includes a laser terrain scanner, and the robot dog is equipped with a high-definition camera and an infrared sensor. The laser terrain scanner is used to perform a global terrain scan of the mining blasting area to obtain laser point cloud data. The high-definition camera is used to identify the elevation values ​​of obstacles, and the external sensor is used to collect road slope data. The UAV is configured to convert laser point cloud data into a mine-independent coordinate system elevation model (DEM), calculate the actual resistance line of each borehole point based on the elevation model (DEM), and select points whose actual resistance lines meet a first threshold as potential borehole points. The UAV is configured to mark the reachability attribute of each grid cell according to the slope and terrain undulation of each grid cell in the elevation model DEM, and then generate a list of hole coordinates based on the reachability attribute and potential hole points. The list of hole coordinates includes hole number, mine-specific coordinates, hole depth and hole diameter. The drone is configured to start from the robot dog's initial position and end at a single potential aperture point, call a bidirectional search (BFS) algorithm to calculate a basic path, and generate multiple candidate paths by adjusting the filtering priority of adjacent grids in the BFS algorithm. The overlap rate between the candidate paths and the basic path is less than a second threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The drone then selects candidate paths based on the robot dog's obstacle-crossing ability and combines the selected candidate paths with the basic path to form a planned path. The robot dog is configured to receive the planned path from the drone, input the obstacle elevation values ​​and road slope into a lightweight MobileNet model, and output the optimal path in the planned path. The robot dog includes a lightweight MobileNet model. The robot dog is configured to receive the list of hole coordinates from the drone and move to the location of each hole after the blasting is completed according to the optimal path. It uses a high-definition camera and infrared sensor to verify the accuracy of the hole position, hole depth, hole diameter and the cleanliness inside the hole, generate a hole inspection report and feed it back to the drone and big data platform. The drone is configured to perform aerial scanning of the blasting area, collect post-blast terrain images and rock block size distribution data to identify potential hazard areas, and synchronize the coordinates of the potential hazard areas to the robot dog. The robot dog is configured to investigate the potential hazard areas and clear any remaining obstacles in the potential hazard areas.

[0008] In a specific embodiment, the UAV is configured to convert laser point cloud data into a mine-independent coordinate system elevation model (DEM), and calculate the actual resistance line for each potential borehole point based on the elevation model DEM, specifically as follows: The laser point cloud data is preprocessed, including noise reduction and smoothing. The original scanning coordinates of the preprocessed laser point cloud data are converted into three-dimensional coordinates in the mine-independent coordinate system using a coordinate transformation formula. The three-dimensional coordinates are rasterized, and the average elevation and average coordinate of all point clouds in each grid are calculated as the elevation value of each grid and the mine-independent coordinates to generate a complete elevation model DEM. The elevation model DEM includes mine-independent coordinates, grid elevation values ​​and grid slope values. Based on the generated elevation model DEM, and combined with the pre-set hole spacing, grid elevation value and grid slope value, potential hole placement points are determined, and the shortest vertical distance from the potential hole placement point to the corresponding blasting free surface is calculated. This distance is the actual resistance line length of the potential hole placement point.

[0009] In a specific embodiment, the UAV is configured to mark the reachability attribute of each grid cell in the elevation model DEM based on the slope and terrain undulation, and then generate a list of aperture coordinates based on the reachability attribute and the actual resistance line, specifically: Extract the core parameters of all grids in the elevation model DEM. The core parameters include the mine-independent coordinates, elevation value, slope value and terrain undulation degree of each grid. The terrain undulation degree is calculated by taking the current grid as the center, taking 3×3 neighboring grids, and calculating the difference between the highest and lowest elevation values ​​in the neighborhood. Based on the reachability attribute determination threshold, a raster reachability attribute distribution map is generated, and the mine-independent coordinates of all reachable rasters are marked. The reachability attribute determination threshold includes the slope threshold and the terrain undulation threshold. Extract the mine-independent coordinates of potential borehole points, associate them with the mine-independent coordinates of all reachable grids. If the mine-independent coordinates of a borehole point fall within the mine-independent coordinates of a reachable grid, retain the potential borehole point; otherwise, remove the potential borehole point. For the retained potential borehole points, recheck the slope and terrain undulation of the grid in which they are located to form a preliminary list of borehole points. The initial borehole point list is supplemented with parameters to ensure that the list includes the borehole number, mine-specific coordinates, borehole depth, and borehole diameter required by the system. All the improved borehole point data are integrated and sorted in order of borehole number to form a borehole coordinate list.

[0010] In a specific embodiment, the drone uses the robot dog's starting position as the starting point and a single hole point as the ending point, and calls a bidirectional search (BFS) algorithm to calculate the basic path, specifically: The drone uses the bidirectional search (BFS) algorithm, taking the grid corresponding to the robot dog's starting position as the starting grid and the grid corresponding to a single hole point as the ending grid. It sets up an initial start queue and an end queue, and stores the start grid and the end grid respectively. Calculate the Euclidean distance between each adjacent reachable grid and the starting grid and the ending grid in the mine-independent coordinate system. The two queues search for adjacent reachable grids according to the Euclidean distance. The bidirectional search stops when a reachable grid is searched by both the starting queue and the ending queue. The search path on the starting side and the search path on the ending side are then concatenated to obtain the basic path.

[0011] In a specific embodiment, multiple candidate paths are generated by adjusting the filtering priority of adjacent grids in the bidirectional search (BFS) algorithm. The overlap rate between the candidate paths and the base path is less than a second threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The candidate paths are then filtered based on the obstacle-crossing ability constraints of the robot dog. The filtered candidate paths and the base path are combined to form a planned path, specifically: Based on the bidirectional search (BFS) algorithm, the bidirectional search process is repeatedly executed by adjusting the filtering priority of adjacent grid cells to generate multiple differentiated candidate paths. Each adjustment focuses on only one priority indicator to ensure path differentiation. Specifically: Set the slope value index weight to the highest, select the adjacent reachable grid with the lowest slope value, call the bidirectional search (BFS) algorithm to complete the search, and generate the first alternative path; Set the elevation value index weight to the highest, select the adjacent reachable grid with the lowest elevation value, call the bidirectional search (BFS) algorithm to complete the search, and generate a second alternative path. Set the weight of terrain undulation to the highest, select the adjacent reachable grid with the lowest terrain undulation, call the bidirectional search (BFS) algorithm to complete the search, and generate a third alternative path; The overlap rate of each generated alternative path with the base path is calculated one by one. Alternative paths with an overlap rate less than the second threshold are selected to form a set of alternative paths. The system receives obstacle-crossing capability parameters from the robot dog and retains valid candidate paths whose grids all meet the obstacle capability parameters, including maximum elevation and maximum slope. The drone combines the basic path with the effective alternative paths to form the planned path.

[0012] In a specific embodiment, the robot dog is configured to receive the planned path from the drone, input the maximum obstacle elevation value and the maximum road gradient into the lightweight MobileNet model, and output the optimal path from the planned path, specifically: The robot dog receives the planned path transmitted by the drone through the wireless communication module, stores the planned path in its own control module, and starts the lightweight MobileNet model to complete the model initialization. The robot dog moves to the starting point of the path at the current hole point, activates its onboard high-definition camera and infrared sensor, and simultaneously collects on-site environmental data of the planned path. The high-definition camera collects images of obstacles on the path in real time, extracts the elevation values ​​of obstacles and the maximum road slope through image recognition algorithms, and transmits them to the lightweight MobileNet model simultaneously. The lightweight MobileNet model transmits the optimal path to the robot dog control module. Based on the robot dog's forward distance or time, the lightweight MobileNet model receives real-time obstacle elevation values ​​and maximum road gradients again, and outputs the optimal path.

[0013] In a specific embodiment, the robot dog receives the list of hole coordinates from the drone and moves to the location of each hole after blasting according to the optimal path. Using a high-definition camera and infrared sensor, it verifies the accuracy of the hole position, hole depth, hole diameter, and cleanliness inside the hole, generates a hole inspection report, and feeds it back to the drone and the big data platform. Specifically: The robot dog retrieves the optimal path output by the lightweight MobileNet model, starts the navigation module, and moves step by step along the optimal path. It uses a high-definition camera to identify obstacles around the path and an infrared sensor to detect road conditions and slope in real time, dynamically adjusting the moving speed. When the robot dog moves to the target hole point, it stops moving. The robot dog uses a high-definition camera and infrared sensor to aim at the orifice of the target hole, captures a clear image of the orifice, extracts the orifice diameter elevation value through an image recognition algorithm, retrieves the standard orifice diameter of the hole at that point from the hole coordinate list, and calculates the deviation rate between the measured orifice diameter and the standard orifice diameter to determine whether the orifice diameter is qualified. The robot dog emits an infrared detection signal to detect the vertical distance from the hole opening to the bottom of the hole to detect the actual hole depth. It retrieves the standard hole depth of the hole point in the hole layout coordinate list and calculates the deviation rate between the actual hole depth and the standard hole depth to determine whether the hole depth is qualified. The robot dog controls a high-definition camera to extend into the hole through the opening and take pictures of the hole to identify whether there are residual stones, mud and water inside the hole, so as to determine whether the cleanliness of the hole is up to standard. After all the holes have been inspected, the robot dog control module integrates all the single-hole inspection records and generates a complete inspection report in a standardized format.

[0014] In a specific embodiment, the drone performs an aerial scan of the blasting area, collecting post-blast terrain images and rock block size distribution data to identify potential hazard areas. The drone then synchronizes the coordinates of these potential hazard areas to the robot dog. Specifically: The drone follows a preset scanning path and uses a high-definition camera to photograph the blasting area, collect topographic images of the blasting area, and identify at least one of the image features in the images, namely, topographic collapse, slope collapse, and damage around the blast holes. Laser topographic scanners scan the rocks scattered after blasting, extract rock block size distribution data, and calculate the elevation values ​​of the length, width and height of each rock block based on the rock block size distribution data to classify the rock block size grades. The UAV's onboard system automatically identifies potential hazard areas based on the collected post-blast terrain images and rock mass grades.

[0015] In a specific embodiment, the robot dog inspects the potential hazard area and removes any remaining obstacles there, specifically as follows: The robot dog receives the hazard areas after the explosion transmitted by the drone, retrieves its own navigation module, compares the coordinates of the hazard areas with its current position coordinates, plans the optimal movement path from the current position to each hazard area, and simultaneously activates the high-definition camera, infrared sensor and self-cleaning device. If a potential hazard area is identified, the robot dog uses its mechanical claw to grab the remaining rocks and transport them to a pre-designated waste disposal area. After the cleanup is completed, a high-definition camera takes an image of the cleaned area to confirm that no obstacles remain. Record the elevation, location, and destination of each obstacle to be cleared, forming an obstacle clearing record, which is then simultaneously fed back to the drone and big data platform.

[0016] A second aspect of this invention proposes a method for collaborative operation of a drone and a robotic dog in a mining scenario, comprising: A drone equipped with a laser terrain scanner performs a global terrain scan of the mining blasting area to obtain laser point cloud data; The drone converts laser point cloud data into a mine-independent coordinate system elevation model (DEM), and calculates the actual resistance line of each potential borehole point based on the elevation model (DEM). The UAV marks the reachability attribute of each grid cell in the elevation model DEM based on the slope and terrain undulation. Then, based on the reachability attribute and the actual resistance line, it generates a hole coordinate list, which includes hole number, mine-specific coordinates, hole depth and hole diameter. The drone starts from the robot dog's starting position and ends at a single point of observation. It then uses a bidirectional search (BFS) algorithm to calculate the basic path. By adjusting the filtering priority of adjacent grids in the BFS algorithm, multiple alternative paths are generated. The overlap rate between the alternative paths and the basic path is less than a first threshold. The filtering priority includes slope, elevation, and terrain undulation. The drone then selects alternative paths based on the robot dog's obstacle-crossing ability and combines the selected alternative paths with the basic path to form the planned path. The robot dog receives the planned path from the drone, uses its onboard high-definition camera to identify the elevation values ​​of obstacles, and uses its onboard infrared sensor to collect road slope data. The robot dog includes a lightweight MobileNet model, which selects the optimal path from the planned path based on the obstacle elevation values ​​and road slope. The robot dog receives the list of hole coordinates from the drone and moves to the location of each hole after the blasting is completed according to the optimal path. Through high-definition cameras and infrared sensors, it verifies the accuracy of the hole position, hole depth, hole diameter and the cleanliness inside the hole, generates a hole inspection report and feeds it back to the drone and big data platform. The drone scans the blasting area from the air, collecting post-blast terrain images and rock block size distribution data to identify potential hazard areas. The drone then synchronizes the coordinates of the potential hazard areas to the robot dog, which inspects the hazard areas and clears any remaining obstacles.

[0017] It has the following effects: 1. A drone equipped with a laser terrain scanner performs a global terrain scan of the mine blasting area to acquire laser point cloud data. After preprocessing, this data is converted into a Depth Model (DEM) in a mine-independent coordinate system. Based on this DEM, the actual resistance line of each borehole point is calculated, and points whose actual resistance lines meet a first threshold are selected as potential borehole points. This method replaces manual labor in performing complex terrain scanning and borehole selection, significantly improving the accuracy of borehole location, avoiding the safety risks of manual measurement in the high-risk environment of the mine, and significantly improving the efficiency of early-stage borehole planning.

[0018] 2. The UAV extracts core parameters such as slope and terrain undulation from each grid cell in the elevation model (DEM), marks each grid cell with reachability attributes, and then associates them with the locations of potential borehole points. Borehole points located in inaccessible grid cells are removed, and parameters such as borehole number, coordinates, depth, and diameter are added to generate a standardized borehole coordinate list. This ensures that the borehole points are compatible with the UAV's navigation capabilities, reduces invalid borehole placement and subsequent redundancy, ensures that the borehole placement plan fits the actual mine terrain, and improves the feasibility of borehole placement operations.

[0019] 3. The drone uses the grid corresponding to the robot dog's starting position as the starting point and the grid corresponding to a single potential aperture point as the ending point. It then uses a bidirectional BFS algorithm to calculate the Euclidean distance between each adjacent reachable grid and the starting and ending grids, optimizes the search direction, and concatenates the paths from the starting and ending sides to obtain the basic path. Compared to traditional path planning algorithms, the bidirectional BFS algorithm, combined with Euclidean distance optimization, significantly accelerates path planning, ensures optimal basic path distance and slope adaptation, and guarantees smooth passage for the robot dog.

[0020] 4. The drone generates multiple candidate paths with an overlap rate of less than a second threshold with the basic path by adjusting the grid selection priority (slope value, elevation value, terrain undulation) in the bidirectional BFS algorithm. These candidate paths are then combined with the robot dog's obstacle-crossing capability parameters (maximum elevation, maximum slope) to further select effective alternative paths. This enriches the robot dog's path selection, adapts to the complex and varied road conditions in mines, avoids the problem of no single path being accessible, and improves the flexibility and reliability of path planning.

[0021] 5. The robot dog is equipped with a lightweight MobileNet model. After receiving the planned path transmitted from the drone, it uses a high-definition camera and infrared sensors to collect data such as the elevation of obstacles and the slope of the road. After inputting this data into the model, it quickly outputs the optimal path from the planned path and can dynamically adjust it according to real-time road conditions. This significantly reduces the robot dog's reliance on hardware computing power, improves the speed of optimal path selection, avoids operational delays caused by slow computation, and adapts to the complex and dynamic road conditions in mines.

[0022] 6. The robot dog moves to the locations of each hole after blasting according to the optimal path. Using high-definition cameras and infrared sensors, it verifies the accuracy of the hole location, depth, diameter, and cleanliness of the hole, generating a standardized inspection report and providing feedback. The drone scans the blasting area to identify potential hazards, and the robot dog receives the coordinates to check and clear any remaining obstacles. This automates the entire process of hole inspection, hazard identification, and cleanup, reducing manual labor, lowering mining operation costs, mitigating the risks of high-risk manual operations, and ensuring the safe and orderly progress of subsequent operations. Attached Figure Description

[0023] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings: Figure 1 This is a block diagram of a collaborative operation system between drones and robotic dogs in a mining scenario; Figure 2 This is a flowchart of a collaborative operation method between drones and robot dogs in a mining scenario. Detailed Implementation

[0024] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and not intended to limit it. Furthermore, it should be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings.

[0025] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0026] like Figure 1As shown, the first aspect of this invention proposes a collaborative operation system 100 for drones and robotic dogs in a mining scenario, including a drone, a robotic dog, and a wireless communication module. The drone and the robotic dog are each equipped with a wireless communication module, which employs 5G+Mesh dual-module wireless communication to transmit data. The drone includes a laser terrain scanner and a high-definition camera. The laser terrain scanner is used to perform a global terrain scan of the mining blasting area to obtain laser point cloud data, and the drone's camera is used to identify the flight path. The robotic dog is equipped with a high-definition camera and an infrared sensor. The high-definition camera is used to identify obstacle elevation values, and the external sensor is used to collect road slope data. Drones can perform a comprehensive, blind-spot-free scan of the blasting area in a mine, avoiding the terrain blind spots and safety risks associated with manual scanning. The laser topographic scanner collects high-precision, detailed laser point cloud data, which can accurately reproduce the terrain features of the area. This provides a reliable and comprehensive data source for subsequent elevation model construction, actual resistance line calculation, and borehole design, ensuring the accuracy and scientific nature of subsequent operations.

[0027] The UAV is configured to convert laser point cloud data into a mine-independent coordinate system elevation model (DEM), calculate the actual resistance line of each borehole point based on the elevation model (DEM), and select points whose actual resistance lines meet a first threshold as potential borehole points. The laser point cloud data is converted into a DEM with an independent coordinate system for the mine, which accurately matches the coordinate standards of the mine site and ensures smooth connection of subsequent operations such as hole layout and navigation. Based on the DEM, the actual resistance line of each potential hole layout point can be accurately calculated to ensure that the hole layout design meets the requirements of the mine blasting process and avoids poor blasting effect, excessively large rock blocks or safety hazards caused by unreasonable resistance lines, providing a core basis for the selection of potential hole layout points.

[0028] The UAV is configured to mark the reachability attribute of each grid cell according to the slope and terrain undulation of each grid cell in the elevation model DEM, and then generate a list of hole coordinates based on the reachability attribute and potential hole points. The list of hole coordinates includes hole number, mine-specific coordinates, hole depth and hole diameter. By marking the reachability attribute of the grid, areas that the robot dog can safely pass through can be screened in advance, avoiding setting the hole points in inaccessible areas and reducing the waste of subsequent work. The generated hole coordinate list contains complete parameters such as hole number, coordinates, hole depth, and hole diameter, which standardizes the hole layout. This provides a clear and specific comparison basis for the robot dog's subsequent hole inspection work, and also makes it easier for staff to control the hole layout quality and improve the operability of the hole layout design.

[0029] The drone is configured to start from the robot dog's initial position and end at a single potential aperture point, call a bidirectional search (BFS) algorithm to calculate a basic path, and generate multiple candidate paths by adjusting the filtering priority of adjacent grids in the BFS algorithm. The overlap rate between the candidate paths and the basic path is less than a second threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The drone then selects candidate paths based on the robot dog's obstacle-crossing ability and combines the selected candidate paths with the basic path to form a planned path. The bidirectional search (BFS) algorithm is highly efficient, quickly calculating the optimal basic path for the robot dog to reach the hole point, thus improving path planning efficiency. By adjusting the filtering priority, multiple differentiated alternative paths are generated, and the overlap rate with the basic path is controlled to avoid path redundancy. At the same time, combined with the robot dog's obstacle-crossing ability, all alternative paths are ensured to be feasible. The combined planned path balances efficiency and flexibility, avoids operation stagnation caused by a single path being blocked, improves the operation fault tolerance rate, and provides support for the robot dog's subsequent real-time route selection.

[0030] The robot dog is configured to receive the planned path from the drone, input the obstacle elevation values ​​and road slope into a lightweight MobileNet model, and output the optimal path in the planned path. The robot dog includes a lightweight MobileNet model. High-definition cameras and infrared sensors work together to capture the size of obstacles and the slope of the road in real time and accurately, ensuring that the robot dog has a comprehensive understanding of the road conditions. The lightweight MobileNet model balances computing speed and accuracy, requires no complex hardware support, and can quickly select the optimal route from the planned path based on real-time road conditions. It can dynamically avoid obstacles and adjust its movement status, reducing the risk of the robot dog getting stuck, malfunctioning, or trapped, ensuring the safety and efficiency of the robot dog's movement, and adapting to the complex and ever-changing road conditions in the mine.

[0031] The robot dog is configured to receive the list of hole coordinates from the drone and move to the location of each hole after the blasting is completed according to the optimal path. It uses a high-definition camera and infrared sensor to verify the accuracy of the hole position, hole depth, hole diameter and the cleanliness inside the hole, generate a hole inspection report and feed it back to the drone and big data platform. The robot dog moves along the optimal path to ensure efficient arrival at each borehole point, improving the efficiency of borehole inspection. Dual sensors work together to verify the accuracy of borehole location, depth, diameter, and cleanliness, avoiding errors from manual inspection and safety risks associated with high-altitude and underground operations. It generates standardized borehole inspection reports and provides real-time feedback, enabling synchronous sharing of inspection data. This allows staff to monitor the quality of blasted boreholes in real time, providing a basis for subsequent operations, while also automating, standardizing, and making the borehole inspection process traceable.

[0032] The drone is configured to perform aerial scanning of the blasting area, collect post-blast terrain images and rock block size distribution data to identify potential hazard areas, and synchronize the coordinates of the potential hazard areas to the robot dog. The robot dog is configured to investigate the potential hazard areas and clear any remaining obstacles in the potential hazard areas.

[0033] The drone's aerial scanning has a wide coverage and high efficiency, and can quickly collect post-blast terrain images and rock block size distribution data, accurately identify potential hazard areas such as terrain collapse, slope collapse, and damage around blast holes, and promptly detect potential safety risks. The coordinates of the hazard areas are synchronized to the robot dog in real time, realizing seamless connection between hazard investigation and subsequent cleanup operations, preventing the hazard from expanding, providing accurate positioning basis for the robot dog's precise investigation and cleanup, and improving the safety and continuity of post-blasting operations in the mine.

[0034] In a specific embodiment, the UAV is configured to convert laser point cloud data into a mine-independent coordinate system elevation model (DEM), and calculate the actual resistance line for each potential borehole point based on the elevation model DEM, specifically as follows: The robot dog accurately connects to the hazard areas reported by the drone, efficiently completing hazard identification and clearing of residual obstacles, replacing manual cleaning, reducing labor intensity and safety risks; it records cleaning information in detail and provides real-time feedback, forming a complete cleaning record, enabling traceability of hazard cleaning operations; clearing hazard areas can eliminate safety hazards, prevent residual obstacles from hindering the progress of subsequent operations, and ensure the safe and orderly conduct of subsequent operations in the mine.

[0035] The laser point cloud data is preprocessed, including noise reduction and smoothing. The original scanning coordinates of the preprocessed laser point cloud data are converted into three-dimensional coordinates in the mine-independent coordinate system using a coordinate transformation formula. Original coordinates to mine-independent 3D coordinates: Obtain the parameters of the mine-independent coordinate system (origin, coordinate axis direction, projection method) and the external parameters of the UAV laser scanner. Substitute the original scanning coordinates of the preprocessed laser point cloud data into the preset coordinate transformation formula to complete the coordinate mapping. Finally, output the (X,Y,Z) 3D coordinates in the mine-independent coordinate system to ensure consistency with the actual coordinates on the mine site.

[0036] The preset coordinate transformation formula (adapted to mining scenarios, combined with scanner extrinsic parameters) is a rigid transformation formula for three-dimensional space, specifically: , Where (x, y, z) are the original scan coordinates, and (X, Y, Z) are the mine-specific coordinates. , , ) represents the translation parameter, while a1-a3, b1-b3, and c1-c3 are rotation parameters, determined by the scanner's extrinsic parameters and coordinate system parameter calibration.

[0037] The three-dimensional coordinates are rasterized, and the average elevation and average coordinate of all point clouds in each grid are calculated as the elevation value of each grid and the mine-independent coordinates to generate a complete elevation model DEM. The elevation model DEM includes mine-independent coordinates, grid elevation values ​​and grid slope values. Rasterization: This refers to dividing the three-dimensional coordinates of laser point clouds in the mine's independent coordinate system into uniform grids with a preset precision. Taking each grid as a unit, the average elevation and average coordinates of all point clouds within the grid are calculated and used as the elevation value and mine-independent coordinates of that grid, respectively. This achieves the normalization of point cloud data and provides a foundation for the subsequent generation of the elevation model (DEM).

[0038] Based on the generated elevation model DEM, and combined with the pre-set hole spacing, grid elevation value and grid slope value, potential hole placement points are determined, and the shortest vertical distance from the potential hole placement point to the corresponding blasting free surface is calculated. This distance is the actual resistance line length of the potential hole placement point.

[0039] Hole spacing refers to the horizontal distance between two adjacent hole placement points preset in the mine blasting design, used to control the hole density; grid elevation value refers to the average elevation of all laser point clouds in each grid after gridding, representing the terrain height corresponding to the grid; grid slope value refers to the degree of terrain inclination of the grid calculated based on the grid elevation value, representing whether the terrain is gentle or not.

[0040] Location of boreholes in the mine: Located within the blasting area of ​​the mine, the elevation of boreholes in open-pit mines is within the range of 1200-1500m, and in underground mines it is within the range of -500 to -200m; and within the accessible grid in the DEM with a slope ≤25°, meeting the preset hole spacing of 2-4m in open-pit mines and 1.5-3m underground, with gentle terrain, suitable for borehole laying operations and robot dog passage, which are the specific points preset for laying blasting holes.

[0041] Location of the blasting free face in the mine: It is located at the point of abrupt change in terrain elevation within the blasting area of ​​the mine, that is, the area where the grid elevation drops sharply in the DEM and the slope is ≥45°, such as the slope of an open mine. It is the free face where the rock is first broken and thrown outward during blasting, and serves as the reference surface for calculating the actual resistance line.

[0042] In a specific embodiment, the UAV is configured to mark the reachability attribute of each grid cell in the elevation model DEM based on the slope and terrain undulation, and then generate a list of aperture coordinates based on the reachability attribute and the actual resistance line, specifically: Extract the core parameters of all rasters in the elevation model (DEM). These core parameters include the mine-specific coordinates, elevation value, slope value, and terrain undulation degree for each raster. The terrain undulation degree is calculated by taking a 3×3 neighboring raster as the center and calculating the difference between the highest and lowest elevation values ​​within that neighboring raster. Extracting these core parameters also obtains all key data related to raster location and terrain, providing a complete data source for raster reachability labeling and borehole point selection. This ensures subsequent operations are supported by data and avoids blind judgments. Raster Mine Independent Coordinates: After rasterization, the average coordinates of all laser point clouds within each raster correspond to the mine independent coordinate system (X,Y,Z), representing the specific location of the raster in the mine; Raster Elevation Value: The average elevation of all laser point clouds within each raster corresponds to the Z-axis of the mine independent coordinate system, representing the raster terrain height; Generating Raster Reachability Distribution Map: Clearly distinguishing between reachable and inaccessible rasters, marking the coordinates of reachable rasters, defining the range for subsequent borehole point screening, and improving screening efficiency; Raster slope value: The degree of terrain inclination calculated based on the raster elevation value, representing whether the raster terrain is gentle or not, in degrees; Raster terrain undulation: Taking the current raster as the center, take 3×3 neighboring raster cells, calculate the difference between the highest and lowest elevation values ​​in the neighboring cells, representing the degree of terrain undulation around the raster.

[0043] Based on the reachability attribute determination threshold, a raster reachability attribute distribution map is generated, and the mine-independent coordinates of all reachable rasters are marked. The reachability attribute determination threshold includes a slope threshold and a terrain undulation threshold; the slope threshold is preferably ≤25°, and the terrain undulation threshold is preferably ≤0.8m. Extract the mine-independent coordinates of potential borehole points and associate them with the mine-independent coordinates of all reachable grids. If the mine-independent coordinates of a borehole point fall within the mine-independent coordinates of a reachable grid, retain the potential borehole point; otherwise, remove the potential borehole point. For the retained potential borehole points, re-check the slope and terrain undulation of the grid where they are located to form a preliminary list of borehole points. Screen potential borehole points: remove borehole points located in inaccessible grids, and re-check the slope and undulation to ensure that the retained borehole points are suitable for operation, forming a qualified preliminary list. The initial borehole location list is supplemented with parameters to ensure it includes the borehole location number, mine-specific coordinates, borehole depth, and borehole diameter required by the system. All the improved borehole location data is then integrated and sorted by borehole location number to form a borehole coordinate list. The core information of the borehole locations is then refined and sorted by borehole location number to form a standardized borehole coordinate list, facilitating accurate borehole location identification by the robot dog and efficient subsequent borehole inspection operations.

[0044] In a specific embodiment, the drone uses the robot dog's starting position as the starting point and a single hole point as the ending point, and calls a bidirectional search (BFS) algorithm to calculate the basic path, specifically: The drone uses the bidirectional search (BFS) algorithm, taking the grid corresponding to the robot dog's starting position as the starting grid and the grid corresponding to a single hole point as the ending grid. It sets up an initial start queue and an end queue, and stores the start grid and the end grid respectively. Calculate the Euclidean distance between each adjacent reachable grid and the starting grid and the ending grid in the mine-independent coordinate system. The two queues search for adjacent reachable grids according to the Euclidean distance. The bidirectional search stops when a reachable grid is searched by both the starting queue and the ending queue. The search path on the starting side and the search path on the ending side are then concatenated to obtain the basic path.

[0045] The formula for calculating the Euclidean distance between adjacent grid cells is as follows: , Where (X1,Y1,Z1) are the mine-independent coordinates of the current grid, (X2,Y2,Z2) are the mine-independent coordinates of the adjacent reachable grids, and d is the Euclidean distance between the two grids, which is used to help determine the search direction and prioritize searching the adjacent grids that are closer to each other to improve search efficiency.

[0046] In a specific embodiment, multiple candidate paths are generated by adjusting the filtering priority of adjacent grids in the bidirectional search (BFS) algorithm. The overlap rate between the candidate paths and the base path is less than a second threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The candidate paths are then filtered based on the obstacle-crossing ability constraints of the robot dog. The filtered candidate paths and the base path are combined to form a planned path, specifically: Based on the bidirectional search (BFS) algorithm, the bidirectional search process is repeatedly executed by adjusting the filtering priority of adjacent grid cells to generate multiple differentiated candidate paths. Each adjustment focuses on only one priority indicator to ensure path differentiation. Specifically: Set the slope value index weight to the highest, select the adjacent reachable grid with the lowest slope value, call the bidirectional search (BFS) algorithm to complete the search, and generate the first alternative path; Set the elevation value index weight to the highest, select the adjacent reachable grid with the lowest elevation value, call the bidirectional search (BFS) algorithm to complete the search, and generate a second alternative path. Set the weight of terrain undulation to the highest, select the adjacent reachable grid with the lowest terrain undulation, call the bidirectional search (BFS) algorithm to complete the search, and generate a third alternative path; By adjusting the single priority focus, the three alternative paths are made significantly different, avoiding path homogenization. This provides the robot dog with multiple travel plans to adapt to different terrain requirements. At the same time, relying on the efficiency of the bidirectional BFS algorithm, the path search is completed quickly, taking into account both path diversity and planning efficiency, and adapting to the complex and ever-changing terrain conditions of the mine.

[0047] The overlap rate of each generated alternative path with the base path is calculated one by one. Alternative paths with an overlap rate less than the second threshold are selected to form a set of alternative paths. The second threshold is 0.05-0.1. Redundant paths with high overlap with the basic path are eliminated to ensure that the alternative paths have real differentiation and backup value. This avoids the problem that the robot dog has no alternative when a single path is blocked. At the same time, it reduces the number of paths, reduces redundancy in subsequent screening, improves the rationality of path planning, and reduces the computational pressure on the robot dog's subsequent route selection.

[0048] Receive the obstacle crossing capability parameters of the robot dog, and retain the valid alternative paths in the alternative paths in which all grids meet the obstacle capability parameters. The obstacle capability parameters include the maximum elevation value and the maximum slope value. Path selection is based on the robot dog's actual obstacle-crossing capabilities, eliminating paths exceeding the robot dog's limits to avoid unreachable paths, robot dog jamming, or malfunctions. This ensures that alternative paths match the robot dog's hardware performance, reducing the risk of operational failures and guaranteeing path feasibility and safety. It retains the efficiency of the basic path while providing the flexibility of alternative paths, offering ample support for the robot dog to select routes based on real-time road conditions. This avoids the limitations of a single path, improves the continuity and fault tolerance of mining operations, and ensures the orderly progress of operations.

[0049] The drone combines the basic path with the effective alternative paths to form the planned path.

[0050] In a specific embodiment, the robot dog is configured to receive the planned path from the drone, input the maximum obstacle elevation value and the maximum road gradient into the lightweight MobileNet model, and output the optimal path from the planned path, specifically: The robot dog receives the planned path transmitted by the drone through the wireless communication module, stores the planned path in its own control module, and starts the lightweight MobileNet model to complete the model initialization. The robot dog moves to the starting point of the path at the current hole point, activates its onboard high-definition camera and infrared sensor, and simultaneously collects on-site environmental data of the planned path. The high-definition camera collects images of obstacles on the path in real time, extracts the elevation values ​​of obstacles and the maximum road slope through image recognition algorithms, and transmits them to the lightweight MobileNet model simultaneously. By using dual sensors to collect data collaboratively, the system can accurately obtain real-time road condition information, providing a real and accurate data source for the model to output the optimal path. This avoids route selection errors due to missing road condition information, ensures the robot dog's passage safety, and reduces the probability of path obstruction.

[0051] Maximum elevation N1 ≤ robot dog preset maximum elevation X1, maximum slope M ≤ robot dog preset maximum slope X2, output basic path when N1 ≥ 4 / 1X1, M ≥ 1 / 4X2; Maximum elevation N1 ≤ machine dog preset maximum elevation X1, maximum slope M ≤ machine dog preset maximum slope X2, output the first alternative path when N1 ≤ 1 / 4X1; When the maximum elevation value N1 is less than or equal to the robot dog's preset maximum elevation X1, and the maximum slope value M is less than or equal to the robot dog's preset maximum slope X2, and M is less than or equal to M1 / 4X2, the second alternative path is output. If the maximum elevation value N1 ≤ the robot dog's preset maximum elevation X1, the maximum slope value M ≤ the robot dog's preset maximum slope X2, and the maximum elevation value N1 - minimum elevation value N2 ≤ 1 / 10X1, the third alternative path will be output first.

[0052] Adapted to mining scenarios and the lightweight requirements of robotic dogs, a lightweight convolutional neural network algorithm is selected, with YOLOv5s being the preferred choice. This algorithm balances recognition speed and accuracy, quickly extracting features from path obstacle images captured by high-definition cameras, accurately identifying obstacle contours and elevation values, and simultaneously calculating and extracting obstacle elevation values ​​and maximum road gradients. It requires no complex hardware support, meeting the needs of robotic dogs for real-time data acquisition and rapid data transmission, and works in conjunction with the lightweight MobileNet model to improve operational efficiency.

[0053] The lightweight MobileNet model transmits the optimal path to the robot dog control module. Based on the robot dog's forward distance or time, the lightweight MobileNet model receives real-time obstacle elevation values ​​and maximum road gradients again, and outputs the optimal path.

[0054] Real-time dynamic adjustment of the optimal path can promptly avoid the risks of passage caused by sudden obstacles and changes in road conditions, ensuring that the robot dog always moves along the safest and most efficient path. At the same time, it relies on a lightweight model to quickly complete calculations, avoiding sluggish calculations and further improving operational efficiency and safety.

[0055] In a specific embodiment, the robot dog receives the list of hole coordinates from the drone and moves to the location of each hole after blasting according to the optimal path. Using a high-definition camera and infrared sensor, it verifies the accuracy of the hole position, hole depth, hole diameter, and cleanliness inside the hole, generates a hole inspection report, and feeds it back to the drone and the big data platform. Specifically: The robot dog retrieves the optimal path output by the lightweight MobileNet model, starts the navigation module, and moves step by step along the optimal path. It uses a high-definition camera to identify obstacles around the path and an infrared sensor to detect road conditions and slope in real time, dynamically adjusting the moving speed. When the robot dog moves to the target hole point, it stops moving. The robot dog uses a high-definition camera and infrared sensor to aim at the orifice of the target hole, captures a clear image of the orifice, extracts the orifice diameter elevation value through an image recognition algorithm, retrieves the standard orifice diameter of the hole at that point from the hole coordinate list, and calculates the deviation rate between the measured orifice diameter and the standard orifice diameter to determine whether the orifice diameter is qualified. The robot dog emits an infrared detection signal to detect the vertical distance from the hole opening to the bottom of the hole to detect the actual hole depth. It retrieves the standard hole depth of the hole point in the hole layout coordinate list and calculates the deviation rate between the actual hole depth and the standard hole depth to determine whether the hole depth is qualified. The robot dog controls a high-definition camera to extend into the hole through the opening and take pictures of the hole to identify whether there are residual stones, mud and water inside the hole, so as to determine whether the cleanliness of the hole is up to standard. After all the holes have been inspected, the robot dog control module integrates all the single-hole inspection records and generates a complete inspection report in a standardized format.

[0056] In a specific embodiment, the drone performs an aerial scan of the blasting area, collecting post-blast terrain images and rock block size distribution data to identify potential hazard areas. The drone then synchronizes the coordinates of these potential hazard areas to the robot dog. Specifically: The drone follows a preset scanning path and uses a high-definition camera to photograph the blasting area, collect topographic images of the blasting area, and identify at least one of the image features in the images, namely, topographic collapse, slope collapse, and damage around the blast holes. Laser topographic scanners scan the rocks scattered after blasting, extract rock block size distribution data, and calculate the elevation values ​​of the length, width and height of each rock block based on the rock block size distribution data to classify the rock block size grades. The UAV's onboard system automatically identifies potential hazard areas based on the collected post-blast terrain images and rock mass grades.

[0057] In a specific embodiment, the robot dog inspects the potential hazard area and removes any remaining obstacles there, specifically as follows: The robot dog receives the hazard areas after the explosion transmitted by the drone, retrieves its own navigation module, compares the coordinates of the hazard areas with its current position coordinates, plans the optimal movement path from the current position to each hazard area, and simultaneously activates the high-definition camera, infrared sensor and self-cleaning device. If a potential hazard area is identified, the robot dog uses its mechanical claw to grab the remaining rocks and transport them to a pre-designated waste disposal area. After the cleanup is completed, a high-definition camera takes an image of the cleaned area to confirm that no obstacles remain. Record the elevation, location, and destination of each obstacle to be cleared, forming an obstacle clearing record, which is then simultaneously fed back to the drone and big data platform.

[0058] A second aspect of this invention proposes a method for collaborative operation of a drone and a robotic dog in a mining scenario, comprising: S201: A drone equipped with a laser terrain scanner to perform a global terrain scan of the mining blasting area to obtain laser point cloud data; S202: The UAV converts laser point cloud data into a mine-independent coordinate system elevation model (DEM), and calculates the actual resistance line of each potential borehole point based on the elevation model (DEM). S203: The UAV marks the reachability attribute of each grid cell in the elevation model DEM according to the slope and terrain undulation. Then, based on the reachability attribute and the actual resistance line, it generates a hole coordinate list, which includes hole number, mine-specific coordinates, hole depth and hole diameter. S204: The UAV starts from the robot dog's starting position and ends at a single hole point. It calls the bidirectional search (BFS) algorithm to calculate the basic path. By adjusting the filtering priority of adjacent grids in the BFS algorithm, multiple candidate paths are generated. The overlap rate between the candidate paths and the basic path is less than a first threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The UAV selects candidate paths based on the robot dog's obstacle-crossing ability and combines the selected candidate paths with the basic path to form a planned path. S205: The robot dog receives the planned path from the drone, uses its onboard high-definition camera to identify obstacle elevation values, and uses its onboard infrared sensor to collect road slope data; the robot dog includes a lightweight MobileNet model, which selects the optimal path from the planned path based on obstacle elevation values ​​and road slope. S206: The robot dog receives the list of hole coordinates from the drone and moves to the location of each hole after the blasting is completed according to the optimal path. Through a high-definition camera and infrared sensor, it verifies the accuracy of the hole position, hole depth, hole diameter and the cleanliness inside the hole, generates a hole inspection report and feeds it back to the drone and big data platform. S207: The drone scans the blasting area from the air, collecting post-blast terrain images and rock block size distribution data to identify potential hazard areas. The drone synchronizes the coordinates of the potential hazard areas to the robot dog, which then investigates the potential hazard areas and clears any remaining obstacles.

[0059] A drone equipped with a laser terrain scanner completes a global scan of the blasting area to acquire laser point cloud data, providing high-precision basic data for subsequent operations and avoiding blind spots and safety risks associated with manual scanning. The point cloud data is converted into a mine-independent coordinate system (DEM) to calculate the actual resistance lines of potential borehole points, ensuring that the borehole design aligns with the blasting process and improving the rationality of the borehole layout. Based on the DEM grid slope and terrain undulations, the drone marks reachability attributes and generates a complete list of borehole coordinates with parameters, avoiding the placement of borehole points in inaccessible areas and providing clear evidence for borehole verification. A bidirectional search (BFS) algorithm is used to plan a basic path, adjusting the grid selection priority to generate differentiated alternative paths. This is combined with the robot dog's obstacle-crossing capabilities to form a planned path, improving path feasibility and fault tolerance. The robot dog receives the planned path, collects road condition data through a high-definition camera and infrared sensors, and selects the optimal path using a lightweight MobileNet model, achieving dynamic obstacle avoidance and ensuring safe and efficient movement. The robot dog arrives at the borehole point along the optimal path, and dual sensors verify borehole position, depth, and other indicators, generating a borehole verification report to replace manual verification and improve accuracy and safety. Drones scan the post-blasting area to identify potential hazards and synchronize their coordinates, while robotic dogs inspect and clear these hazards, recording and providing feedback to achieve closed-loop hazard management. This method, through the collaboration of drones and robotic dogs, automates the entire process of borehole planning, path navigation, borehole inspection, and hazard clearing, improving mining operation efficiency, safety, and standardization.

Claims

1. A collaborative operation system of drones and robot dogs in a mining scenario, characterized in that, include: The system includes a drone, a robot dog, and a wireless communication module. The drone and the robot dog transmit data through the wireless communication module. The drone includes a laser terrain scanner, and the robot dog is equipped with a high-definition camera and an infrared sensor. The laser terrain scanner is used to perform a global terrain scan of the mining blasting area to obtain laser point cloud data. The high-definition camera is used to identify the elevation values ​​of obstacles, and the external sensor is used to collect road slope data. The UAV is configured to convert laser point cloud data into a mine-independent coordinate system elevation model (DEM), calculate the actual resistance line of each borehole point based on the elevation model (DEM), and select points whose actual resistance lines meet a first threshold as potential borehole points. The UAV is configured to mark the reachability attribute of each grid cell according to the slope and terrain undulation of each grid cell in the elevation model DEM, and then generate a list of hole coordinates based on the reachability attribute and potential hole points. The list of hole coordinates includes hole number, mine-specific coordinates, hole depth and hole diameter. The drone is configured to start from the robot dog's initial position and end at a single potential aperture point, call a bidirectional search (BFS) algorithm to calculate a basic path, and generate multiple candidate paths by adjusting the filtering priority of adjacent grids in the BFS algorithm. The overlap rate between the candidate paths and the basic path is less than a second threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The drone then selects candidate paths based on the robot dog's obstacle-crossing ability and combines the selected candidate paths with the basic path to form a planned path. The robot dog is configured to receive the planned path from the drone, input the obstacle elevation values ​​and road slope into a lightweight MobileNet model, and output the optimal path in the planned path. The robot dog includes a lightweight MobileNet model. The robot dog is configured to receive the list of hole coordinates from the drone and move to the location of each hole after the blasting is completed according to the optimal path. It uses a high-definition camera and infrared sensor to verify the accuracy of the hole position, hole depth, hole diameter and the cleanliness inside the hole, generate a hole inspection report and feed it back to the drone and big data platform. The drone is configured to perform aerial scanning of the blasting area, collect post-blast terrain images and rock block size distribution data to identify potential hazard areas, and synchronize the coordinates of the potential hazard areas to the robot dog. The robot dog is configured to investigate the potential hazard areas and clear any remaining obstacles in the potential hazard areas.

2. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 1, characterized in that, The UAV is configured to convert laser point cloud data into a DEM (Demographic Elevation Model) in a mine-independent coordinate system, and to calculate the actual resistance line for each potential borehole point based on the DEM. Specifically: The laser point cloud data is preprocessed, including noise reduction and smoothing. The original scanning coordinates of the preprocessed laser point cloud data are converted into three-dimensional coordinates in the mine-independent coordinate system using a coordinate transformation formula. The three-dimensional coordinates are rasterized, and the average elevation and average coordinate of all point clouds in each grid are calculated as the elevation value of each grid and the mine-independent coordinates to generate a complete elevation model DEM. The elevation model DEM includes mine-independent coordinates, grid elevation values ​​and grid slope values. Based on the generated elevation model DEM, and combined with the pre-set hole spacing, grid elevation value and grid slope value, potential hole placement points are determined, and the shortest vertical distance from the potential hole placement point to the corresponding blasting free surface is calculated. This distance is the actual resistance line length of the potential hole placement point.

3. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 1, characterized in that, The UAV is configured to mark the reachability attribute for each grid cell in the elevation model DEM based on the slope and terrain undulation, and then generate a list of aperture coordinates based on the reachability attribute and the actual resistance line, specifically: Extract the core parameters of all grids in the elevation model DEM. The core parameters include the mine-independent coordinates, elevation value, slope value and terrain undulation degree of each grid. The terrain undulation degree is calculated by taking the current grid as the center, taking 3×3 neighboring grids, and calculating the difference between the highest and lowest elevation values ​​in the neighborhood. Based on the reachability attribute determination threshold, a raster reachability attribute distribution map is generated, and the mine-independent coordinates of all reachable rasters are marked. The reachability attribute determination threshold includes the slope threshold and the terrain undulation threshold. Extract the mine-independent coordinates of potential borehole points, associate them with the mine-independent coordinates of all reachable grids. If the mine-independent coordinates of a borehole point fall within the mine-independent coordinates of a reachable grid, retain the potential borehole point; otherwise, remove the potential borehole point. For the retained potential borehole points, recheck the slope and terrain undulation of the grid in which they are located to form a preliminary list of borehole points. The initial borehole point list is supplemented with parameters to ensure that the list includes the borehole number, mine-specific coordinates, borehole depth, and borehole diameter required by the system. All the improved borehole point data are integrated and sorted in order of borehole number to form a borehole coordinate list.

4. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 1, characterized in that, The drone starts at the robot dog's initial position and ends at a single hole point, using a bidirectional search (BFS) algorithm to calculate the basic path, specifically: The drone uses the bidirectional search (BFS) algorithm, taking the grid corresponding to the robot dog's starting position as the starting grid and the grid corresponding to a single hole point as the ending grid. It sets up an initial start queue and an end queue, and stores the start grid and the end grid respectively. Calculate the Euclidean distance between each adjacent reachable grid and the starting grid and the ending grid in the mine-independent coordinate system. The two queues search for adjacent reachable grids according to the Euclidean distance. The bidirectional search stops when a reachable grid is searched by both the starting queue and the ending queue. The search path on the starting side and the search path on the ending side are then concatenated to obtain the basic path.

5. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 4, characterized in that, By adjusting the filtering priority of adjacent grids in the bidirectional search (BFS) algorithm, multiple candidate paths are generated. The overlap rate between the candidate paths and the base path is less than a second threshold. The filtering priority includes slope value, elevation value, and terrain undulation degree. The candidate paths are then filtered based on the robot dog's obstacle-crossing ability. The filtered candidate paths and the base path are combined to form the planned path, specifically: Based on the bidirectional search (BFS) algorithm, the bidirectional search process is repeatedly executed by adjusting the filtering priority of adjacent grid cells to generate multiple differentiated candidate paths. Each adjustment focuses on only one priority indicator to ensure path differentiation. Specifically: Set the slope value index weight to the highest, select the adjacent reachable grid with the lowest slope value, call the bidirectional search (BFS) algorithm to complete the search, and generate the first alternative path; Set the elevation value index weight to the highest, select the adjacent reachable grid with the lowest elevation value, call the bidirectional search (BFS) algorithm to complete the search, and generate a second alternative path. Set the weight of terrain undulation to the highest, select the adjacent reachable grid with the lowest terrain undulation, call the bidirectional search (BFS) algorithm to complete the search, and generate a third alternative path; The overlap rate of each generated alternative path with the base path is calculated one by one. Alternative paths with an overlap rate less than the second threshold are selected to form a set of alternative paths. Receive the obstacle crossing capability parameters of the robot dog, and retain the valid alternative paths in the alternative paths in which all grids meet the obstacle capability parameters. The obstacle capability parameters include the maximum elevation value and the maximum slope value. The drone combines the basic path with the effective alternative paths to form the planned path.

6. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 1, characterized in that, The robot dog is configured to receive the planned path from the drone, inputting the maximum obstacle elevation and maximum road gradient into a lightweight MobileNet model, and outputting the optimal path from the planned path, specifically: The robot dog receives the planned path transmitted by the drone through the wireless communication module, stores the planned path in its own control module, and starts the lightweight MobileNet model to complete the model initialization. The robot dog moves to the starting point of the path at the current hole point, activates its onboard high-definition camera and infrared sensor, and simultaneously collects on-site environmental data of the planned path. The high-definition camera collects images of obstacles on the path in real time, extracts the elevation values ​​of obstacles and the maximum road slope through image recognition algorithms, and transmits them to the lightweight MobileNet model simultaneously. The lightweight MobileNet model transmits the optimal path to the robot dog control module. Based on the robot dog's forward distance or time, the lightweight MobileNet model receives real-time obstacle elevation values ​​and maximum road gradients again, and outputs the optimal path.

7. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 1, characterized in that, The robot dog receives the list of hole coordinates from the drone and moves to the location of each hole after blasting according to the optimal path. Using a high-definition camera and infrared sensors, it verifies the accuracy of the hole positions, depth, diameter, and cleanliness of the holes, generates a hole inspection report, and feeds it back to the drone and the big data platform. Specifically: The robot dog retrieves the optimal path output by the lightweight MobileNet model, starts the navigation module, and moves step by step along the optimal path. It uses a high-definition camera to identify obstacles around the path and an infrared sensor to detect road conditions and slope in real time, dynamically adjusting the moving speed. When the robot dog moves to the target hole point, it stops moving. The robot dog uses a high-definition camera and infrared sensor to aim at the orifice of the target hole, captures a clear image of the orifice, extracts the orifice diameter elevation value through an image recognition algorithm, retrieves the standard orifice diameter of the hole at that point from the hole coordinate list, and calculates the deviation rate between the measured orifice diameter and the standard orifice diameter to determine whether the orifice diameter is qualified. The robot dog emits an infrared detection signal to detect the vertical distance from the hole opening to the bottom of the hole to detect the actual hole depth. It retrieves the standard hole depth of the hole point in the hole layout coordinate list and calculates the deviation rate between the actual hole depth and the standard hole depth to determine whether the hole depth is qualified. The robot dog controls a high-definition camera to extend into the hole through the opening and take pictures of the hole to identify whether there are residual stones, mud and water inside the hole, so as to determine whether the cleanliness of the hole is up to standard. After all the holes have been inspected, the robot dog control module integrates all the single-hole inspection records and generates a complete inspection report in a standardized format.

8. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 1, characterized in that, The drone performs an aerial scan of the blasting area, collecting post-blast terrain images and rock block size distribution data to identify potential hazard areas. The drone then synchronizes the coordinates of these hazard areas to the robot dog. Specifically: The drone follows a preset scanning path and uses a high-definition camera to photograph the blasting area, collect topographic images of the blasting area, and identify at least one of the image features in the images, namely, topographic collapse, slope collapse, and damage around the blast holes. Laser topographic scanners scan the rocks scattered after blasting, extract rock block size distribution data, and calculate the elevation values ​​of the length, width and height of each rock block based on the rock block size distribution data to classify the rock block size grades. The UAV's onboard system automatically identifies potential hazard areas based on the collected post-blast terrain images and rock mass grades.

9. The collaborative operation system of drones and robot dogs in a mining scenario according to claim 8, characterized in that, The robot dog inspects the potential hazard area and removes any remaining obstacles, specifically: The robot dog receives the hazard areas after the explosion transmitted by the drone, retrieves its own navigation module, compares the coordinates of the hazard areas with its current position coordinates, plans the optimal movement path from the current position to each hazard area, and simultaneously activates the high-definition camera, infrared sensor and self-cleaning device. If a potential hazard area is identified, the robot dog uses its mechanical claw to grab the remaining rocks and transport them to a pre-designated waste disposal area. After the cleanup is completed, a high-definition camera takes an image of the cleaned area to confirm that no obstacles remain. Record the elevation, location, and destination of each obstacle to be cleared, forming an obstacle clearing record, which is then simultaneously fed back to the drone and big data platform.

10. A method for collaborative operation of drones and robot dogs in a mining scenario, characterized in that, include: A drone equipped with a laser terrain scanner performs a global terrain scan of the mining blasting area to obtain laser point cloud data; The drone converts laser point cloud data into a mine-independent coordinate system elevation model (DEM), and calculates the actual resistance line of each potential borehole point based on the elevation model (DEM). The UAV marks the reachability attribute of each grid cell in the elevation model DEM based on the slope and terrain undulation. Then, based on the reachability attribute and the actual resistance line, it generates a hole coordinate list, which includes hole number, mine-specific coordinates, hole depth and hole diameter. The drone starts from the robot dog's starting position and ends at a single point of observation. It then uses a bidirectional search (BFS) algorithm to calculate the basic path. By adjusting the filtering priority of adjacent grids in the BFS algorithm, multiple alternative paths are generated. The overlap rate between the alternative paths and the basic path is less than a first threshold. The filtering priority includes slope, elevation, and terrain undulation. The drone then selects alternative paths based on the robot dog's obstacle-crossing ability and combines the selected alternative paths with the basic path to form the planned path. The robot dog receives the planned path from the drone, uses its onboard high-definition camera to identify the elevation values ​​of obstacles, and uses its onboard infrared sensor to collect road slope data. The robot dog includes a lightweight MobileNet model, which selects the optimal path from the planned path based on the obstacle elevation values ​​and road slope. The robot dog receives the list of hole coordinates from the drone and moves to the location of each hole after the blasting is completed according to the optimal path. Through high-definition cameras and infrared sensors, it verifies the accuracy of the hole position, hole depth, hole diameter and the cleanliness inside the hole, generates a hole inspection report and feeds it back to the drone and big data platform. The drone scans the blasting area from the air, collecting post-blast terrain images and rock block size distribution data to identify potential hazard areas. The drone then synchronizes the coordinates of the potential hazard areas to the robot dog, which inspects the hazard areas and clears any remaining obstacles.

Images