Unmanned aerial vehicle autonomous mapping and rescue path visualization generation method for post-disaster surface mine

Abstract

The invention discloses an unmanned aerial vehicle autonomous mapping and rescue path visualization generation method for a post-disaster surface mine. The method comprises the following steps: controlling an unmanned aerial vehicle to autonomously explore in a GNSS-signal-free environment and obtain environment sensing data; a three-dimensional point cloud map is constructed based on the data, and a globally consistent three-dimensional dynamic reference navigation map is obtained through registration with a prior CAD drawing; geological risk areas in the map are identified and semantic annotation is carried out; converting the three-dimensional map into a two-dimensional grid map with a risk weight through a multivariable risk assessment model based on a labeling result; planning a rescue path by adopting an improved A algorithm; and finally, fusing the risk grid map, the path and the point cloud map, and generating and outputting a visual navigation result containing the risk thermodynamic diagram and the path identifier. According to the method, rapid, autonomous and safety evaluation of the post-disaster surface mine is realized, visual and reliable visual navigation support is provided for rescue workers, and the post-disaster emergency response efficiency and safety are remarkably improved.

Description

Technical Field

[0001] This invention belongs to the field of intelligent unmanned systems and mine emergency rescue technology, and in particular relates to a method for autonomous mapping and rescue path visualization generation by unmanned aerial vehicles (UAVs) for post-disaster open-pit mines. Background Technology

[0002] Open-pit mines are prone to secondary disasters such as bench collapses and spoil heap landslides triggered by earthquakes, heavy rainfall, or blasting, seriously threatening the safety of workers. Traditional manual surveys are inefficient and risky, unable to quickly cover large areas of the post-disaster environment, and lack the ability to accurately identify key geological features such as bench cracks and loose rock masses. While UAV technology offers the advantage of high-altitude scanning, existing solutions have significant drawbacks: 3D mapping methods (such as LIO-SAM) only achieve environmental reconstruction and cannot identify the unique bench structure risks of open-pit mines; GPS interruption and drastic environmental changes after a disaster cause SLAM drift, resulting in a high failure rate for accurate relocation across different time periods; path planning algorithms primarily serve the UAV's own obstacle avoidance during flight, without considering the safety of ground rescue personnel in high-risk areas such as areas with rubble accumulation and slopes greater than 30°; the system relies on ground stations or cloud processing, cannot operate independently in mining areas without network access, and lacks localized visualization output, hindering on-site decision-making.

[0003] Existing technologies do not incorporate optimized exploration strategies for the stepped structures of open-pit mines. For example, traditional frontier exploration fails to differentiate between the priority of step edges and transport roads, resulting in insufficient coverage of high-risk areas; conventional landmark matching becomes ineffective due to post-disaster environmental changes, failing to support pre- and post-disaster comparative analysis. Furthermore, path planning does not quantify geological risks as a cost function; areas with crack density greater than 50% or gravel coverage greater than 70% are still included in the travel path, increasing secondary risks for rescue personnel. In addition, rescue personnel need to travel along existing transport roads, but existing algorithms do not utilize this prior information, resulting in circuitous and unsafe planned paths. Summary of the Invention

[0004] To address the aforementioned technical problems, this invention proposes a method for autonomous mapping and rescue path visualization generation using unmanned aerial vehicles (UAVs) in post-disaster open-pit mines, thereby resolving the issues present in the existing technologies.

[0005] To achieve the above objectives, this invention provides a method for autonomous mapping and rescue path visualization generation using unmanned aerial vehicles (UAVs) in post-disaster open-pit mines, comprising: The system controls a drone to autonomously explore in an environment without global navigation satellite system signals, acquiring environmental perception data; and constructs a three-dimensional point cloud map based on the environmental perception data. The three-dimensional point cloud map is registered to obtain a globally consistent three-dimensional dynamic reference navigation map; geological risk areas are identified and semantically labeled in the three-dimensional dynamic reference navigation map; Based on the semantic annotation results, the three-dimensional dynamic reference navigation map is converted into a two-dimensional raster map with risk weights; Rescue routes are planned based on two-dimensional grid maps with risk weights; By integrating the risk weight grid map, the rescue route, and the 3D point cloud map, a visual navigation result containing a risk heat map and route markers is generated and output.

[0006] Optionally, the process of controlling a drone to conduct autonomous exploration includes: Prior information about the open-pit mine structure is obtained. Based on this prior information and a real-time constructed 3D occupancy grid map, a semantically enhanced frontier exploration algorithm is used to control the autonomous exploration of a UAV. Specifically, the semantically enhanced frontier exploration algorithm identifies frontier points belonging to unexplored areas in the 3D occupancy grid map. These frontier points are then filtered using the prior information. For each filtered frontier point, its distance to the UAV's current position and semantic saliency are calculated. A comprehensive evaluation is then performed based on the distance and semantic saliency, and the optimal exploration direction is selected according to the comprehensive evaluation result.

[0007] Optionally, the process of obtaining a globally consistent 3D dynamic reference navigation map includes: The environmental perception data is processed using a laser-vision-inertial tightly coupled synchronous positioning and mapping algorithm to generate the 3D point cloud map in real time. The 3D point cloud map is then spatially aligned with the pre-stored mine structure CAD drawings using an iterative nearest point algorithm to obtain a globally consistent 3D dynamic reference navigation map.

[0008] Optionally, the process of identifying and semantically annotating geological risk areas in a 3D dynamic reference navigation map includes: A lightweight 3D target detection network deployed on the UAV's onboard edge computing unit is used to process the 3D dynamic reference navigation map in real time, identify step cracks, loose rock and gravel accumulation areas, and output their category and 3D location information as semantic annotations.

[0009] Optionally, the process of converting the three-dimensional dynamic reference navigation map into a two-dimensional raster map with risk weights based on the semantic annotation results includes: The semantically annotated 3D dynamic reference navigation map is projected onto a 2D plane and divided into grids to obtain a 2D grid map. Based on the semantic annotation information corresponding to each grid, a multivariate risk assessment model is used to dynamically calculate the risk weight of each grid. Based on the risk weight values ​​of all grids, a 2D grid map with risk weights is generated.

[0010] Optionally, the multivariate risk assessment model calculates the risk weight of the grid cell according to the following formula: ; in, For grid Risk weights, For grid The slope value, For grid Crack density, For grid The ratio of gravel coverage area, For grid Light intensity, These are adjustable weighting coefficients.

[0011] Optionally, the process of planning rescue routes based on a two-dimensional raster map with risk weights includes: Adopting improved A The algorithm performs path planning; it treats each passable grid in the 2D raster map as a path node; based on the improved A... The algorithm's heuristic function calculates the total cost of a node based on the actual cost from the starting point to the node, and searches for and generates the rescue path based on the principle of minimizing the total cost; where the improved A The heuristic function of the algorithm is: ; in, For heuristic functions, Let n be the risk weight. Let n be the slope of node n. Let n be the crack density at node n. This represents the risk sensitivity coefficient.

[0012] Optionally, the process of obtaining the visualized navigation result includes: generating a color-coded risk heat map using a color mapping function based on the risk weight values ​​of each grid in a two-dimensional raster map with risk weights; smoothing and visualizing the rescue path, and overlaying and fusing it with the three-dimensional point cloud map and the risk heat map to obtain the visualized navigation result.

[0013] Optionally, it also includes: extracting key structural feature points and their descriptors from the annotated 3D dynamic reference navigation map to construct a relocation anchor point library; performing accurate relocation based on the relocation anchor point library; and updating and maintaining the global consistency of the 3D dynamic reference navigation map.

[0014] Compared with the prior art, the present invention has the following advantages and technical effects: This invention achieves efficient autonomous perception and safety rescue decision support for post-disaster open-pit mine environments by integrating prior information guidance, multi-sensor tightly coupled mapping, 3D semantic risk identification, quantitative risk assessment and improved path planning, and local visualization rendering. Its semantic enhancement significantly improves the reconnaissance coverage efficiency of high-risk areas; laser-vision-inertial SLAM and CAD registration and relocation anchor point library ensure the consistency and cross-time reliability of the global map; and it quantifies geological features based on a multivariate risk model and incorporates A... The algorithm's heuristic function generates rescue paths that can proactively avoid high-risk areas; finally, through real-time fusion of airborne edge computing, it generates a visualized result containing risk heatmaps and paths, providing intuitive and reliable navigation data for on-site rescue, and improving the autonomy, safety, and timeliness of post-disaster emergency response. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings: Figure 1 This is a schematic diagram of the method structure according to an embodiment of the present invention; Figure 2 This is a three-dimensional visualization output diagram of an embodiment of the present invention. Detailed Implementation

[0016] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0017] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0018] Example 1 like Figure 1 As shown, this embodiment provides a method for autonomous mapping and rescue path visualization generation using unmanned aerial vehicles (UAVs) in post-disaster open-pit mines, including: Step 1: System Deployment and Initialization. At the post-disaster open-pit mine site, the UAV system, equipped with LiDAR, a visual camera, and an inertial measurement unit (IMU) sensor, completes system initialization via onboard edge computing equipment. The system loads prior information about the mine structure (including CAD drawings of bench layout, transportation road directions, and the locations of key structures) as the semantic guidance basis for environmental exploration. Simultaneously, a pre-trained lightweight semantic segmentation model is loaded, optimized for typical geological features of open-pit mines.

[0019] Based on prior information about the open-pit mine structure, a semantically enhanced frontier exploration algorithm is used to control the autonomous exploration behavior of a UAV in an environment without Global Navigation Satellite System (GNSS) signals. The semantically enhanced frontier exploration algorithm identifies the "frontier" points of unexplored areas based on a three-dimensional occupied grid map.

[0020] The prior information on the open-pit mine structure includes CAD drawings of the bench layout, transportation road direction, and the location of key structures, which are used to guide drones to prioritize the exploration of high-risk areas.

[0021] The implementation process of the semantically enhanced frontier exploration algorithm includes: acquiring a 3D occupied grid map, identifying frontier points in unexplored areas, calculating the exploration priority score of each frontier point, selecting the frontier point with the highest exploration priority as the exploration target of the UAV, then guiding the UAV to fly towards that frontier point, and repeating the above process until all unexplored areas are covered.

[0022] The scoring function for the exploration direction of the frontier is defined as follows: ; in, Prioritize exploration of cutting-edge points. Let Euclidean distance be the distance from the front edge to the current position. The semantic saliency weights of the frontier points (output by the semantic segmentation network, reflecting the intensity of risk features such as cracks and loose rock masses). and These are adjustable weighting coefficients.

[0023] Step 2, Autonomous Exploration and Mapping. A 3D point cloud map is constructed in real time using a laser-vision-inertial tightly coupled simultaneous localization and mapping (SLAM) algorithm. Spatial registration is then performed with pre-stored mine structure CAD drawings using an iterative nearest point (ICP) algorithm to generate a globally consistent, high-precision dynamic reference navigation map. Step 3, Precise Relocalization. The system extracts geometric feature points (such as corner points and planes) from the point cloud map, obtains local descriptors through the SuperPoint feature extractor, and combines geometric stability scores to select persistent landmarks. Landmark selection criteria include geometric stability, texture uniqueness, and environmental matching to ensure that selected landmarks remain stable in the post-disaster environment for a long time. The system builds a landmark anchor point library to store landmark locations and descriptor indexes. When performing cross-time period relocalization tasks, the system achieves pose estimation through descriptor matching and geometric consistency verification.

[0024] A lightweight 3D target detection network is used to identify areas such as step cracks, loose rock mass and gravel accumulation in real time. The identification results are then semantically annotated on the dynamic reference navigation map, and key structural features are extracted to construct a relocation anchor point library.

[0025] The lightweight 3D target detection network is obtained by performing channel pruning, knowledge distillation and quantization on the standard 3D detection network. It is deployed on the airborne edge computing unit of the UAV to realize real-time identification and 3D positioning of areas such as step cracks, loose rock and gravel accumulation.

[0026] The implementation process of the lightweight 3D target detection network includes: channel pruning of the standard 3D detection network to reduce the computational load; knowledge transfer from the large model to the small model through knowledge distillation; quantization of the network to improve inference speed; and deployment on an airborne edge computing unit to achieve real-time identification and 3D localization of step cracks, loose rock mass, and gravel accumulation areas.

[0027] The construction process of the relocation anchor point library includes: extracting geometric feature points from the point cloud map; obtaining local descriptors through the SuperPoint feature extractor; filtering persistent landmarks by combining geometric stability scores; and storing landmark locations and descriptor indexes.

[0028] Step 4, Risk Semantic Mapping. Based on a multivariate risk assessment model, the risk weight of each grid is dynamically calculated, and the dynamic reference navigation map is converted into a two-dimensional raster map with risk weights. The formula for calculating the risk weight of each grid is as follows: ; in, For grid Risk weights, For grid The slope value, For grid Crack density, For grid The ratio of gravel coverage area, For grid Light intensity, These are adjustable weighting coefficients.

[0029] The implementation process of the multivariate risk assessment model includes: constructing a risk weight calculation model based on geological risk characteristics (such as slope, crack density, gravel coverage ratio, and light intensity); determining the weight coefficients of each risk factor through experiments and data analysis; dynamically calculating the risk weight of each grid and generating a risk weight raster map.

[0030] The process of generating a risk-weighted raster map includes: projecting a 3D point cloud map into a 2D raster map; calculating the slope value, crack density, gravel coverage ratio, and light intensity of each grid; calculating the risk weight of each grid according to preset weight coefficients; and generating a 2D raster map with risk weights.

[0031] The formula for calculating the area ratio covered by gravel is: in, For grid The point cloud volume, This represents the minimum point cloud volume. This represents the maximum point cloud volume.

[0032] Step 5, Rescue Route Planning. The user specifies the start and end points via a local input device (such as a mobile phone). The system, based on the aforementioned risk weight grid map, employs an improved A / B algorithm. The algorithm plans the rescue route, wherein the improvement A The heuristic function of an algorithm is defined as: in, For heuristic functions, Let n be the risk weight. Let n be the slope of node n. Let n be the crack density at node n. This is the risk sensitivity coefficient. The improved A... The algorithm's heuristic function design takes into account the safety of ground rescue personnel, incorporates risk factors into the heuristic function, and prioritizes paths with low risk weights and favorable passage conditions.

[0033] Improvement A The algorithm implementation process includes: taking the risk weight grid map as input, and taking the starting point and the ending point as input; calculating the total cost of each node, including the actual path cost and the heuristic function value; selecting nodes for expansion in order of increasing total cost until the ending point is reached; backtracking the path and generating a recommended rescue path.

[0034] Improvement A The total cost function of the algorithm is defined as: ; in, For nodes The total cost, The actual path cost from the starting point to node n. This is a heuristic function.

[0035] Step 6, Visualization Output. The risk weight raster map, recommended rescue route, and 3D point cloud map are fused and rendered to generate a visualized navigation result including a risk heatmap and route markers, which is then output to a local display device. The 3D visualization output effect is as follows: Figure 2As shown, this module serves as an offline visualization terminal, integrating the overall environmental view with risk assessment to generate a 3D rescue navigation map. The system constructs a quantitative heatmap based on risk weights, setting low, medium, and high thresholds to intuitively map geological hazards such as areas with dense cracks. Simultaneously, it utilizes the B-spline algorithm to perform continuous smooth fitting on discrete path nodes and overlays high-contrast direction vectors, ultimately outputting an intelligent rescue path that combines risk warning accuracy with kinematic feasibility.

[0036] The risk weighted raster map, recommended rescue routes, and 3D point cloud map are fused and rendered to generate a visual navigation result containing a risk heatmap and route markers, which is then output to a local display device. The color mapping of the risk heatmap is achieved using the following formula: ; in, For grid The risk weight (calculated from formula 3 in step 4). These represent the intensity values ​​(range 0-255) of the red, green, and blue channels, respectively, mapped to RGB color values. This mapping formula quantifies the risk value into an intuitive visual code: green... Yellow indicates low-risk areas, medium-risk areas indicate medium-risk areas, and red indicates low-risk areas. This indicates a high-risk area.

[0037] The process of fusion rendering includes: projecting the risk weight raster map onto the background of the 3D point cloud map; generating a risk heat map based on the risk weight calculation results; converting the coordinates of the recommended rescue route into continuous trajectories and overlaying them on the heat map; and generating a visual navigation result that includes the risk heat map and route markers.

[0038] The recommended rescue path is rendered using B-spline curve smoothing, and the path point coordinates are... The trajectory can be fitted to a continuous trajectory using the following formula: in, Let be the coordinates of any point on the path. For p-order B-spline basis functions, Here are the coordinates of the control points, and n is the number of path points. The fitted path is indicated by white arrows. Overlaid on the risk heatmap, arrows indicate the direction of the travel path. Finally, the system outputs the fused, visualized navigation results to the local display device in both image and trajectory file formats. Image format: Generate PNG file with a resolution of 1920×1080, including 3D point cloud background, risk heat map and path overlay; Track file: Generates path coordinate data in JSON format, including timestamps and risk weight values, for easy subsequent analysis.

[0039] The visual navigation results include an overlay of a 3D point cloud map, a risk heat map, and a recommended route. The risk heat map uses color coding to indicate the risk level, with red indicating high-risk areas and green indicating low-risk areas. The recommended route is overlaid on the map as a white arrow.

[0040] The visual navigation results are output to the local display device in image format and trajectory file format. The image format is a high-resolution PNG file, and the trajectory file is structured data containing path coordinates and risk weight values.

[0041] This invention eliminates the need for ground stations or cloud processing in post-disaster open-pit mine environments. It can operate independently on an UAV-borne edge computing unit, supporting real-time generation and display of visualized navigation results even in network-free environments. Through a closed-loop process encompassing "autonomous mapping, precise relocation, risk identification, path planning, and local visualization," it achieves intelligent environmental perception and rescue decision-making in post-disaster open-pit mines, providing scientific and efficient technical support for mine emergency rescue.

[0042] This invention addresses the unique challenges of post-disaster scenarios in open-pit mines by proposing an integrated approach: it identifies step risks through structure-guided autonomous mapping, achieves precise cross-time relocation based on persistent landmarks, quantifies geological features as a risk cost function, generates priority rescue routes along transportation roads, and supports local visualization output. This method overcomes the limitations of existing technologies, achieving a closed loop from environmental perception to decision support, significantly improving the efficiency and safety of post-disaster emergency response, and providing a scientific basis for mine rescue.

[0043] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for autonomous mapping and rescue route visualization generation using unmanned aerial vehicles (UAVs) in post-disaster open-pit mines, characterized in that, Includes the following steps: Controlling drones to autonomously explore in environments without global navigation satellite system signals, and obtaining environmental perception data; A 3D point cloud map is constructed based on the environmental perception data; The three-dimensional point cloud map is registered to obtain a globally consistent three-dimensional dynamic reference navigation map; Identify and semantically annotate geological risk areas in a 3D dynamic reference navigation map; Based on the semantic annotation results, the three-dimensional dynamic reference navigation map is converted into a two-dimensional raster map with risk weights; Rescue routes are planned based on two-dimensional grid maps with risk weights; By integrating the risk weight grid map, the rescue route, and the 3D point cloud map, a visual navigation result containing a risk heat map and route markers is generated and output.

2. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 1, characterized in that, The process of controlling a drone to conduct autonomous exploration includes: Prior information about the open-pit mine structure is obtained. Based on this prior information and a real-time constructed 3D occupancy grid map, a semantically enhanced frontier exploration algorithm is used to control the autonomous exploration of a UAV. Specifically, the semantically enhanced frontier exploration algorithm identifies frontier points belonging to unexplored areas in the 3D occupancy grid map. These frontier points are then filtered using the prior information. For each filtered frontier point, its distance to the UAV's current position and semantic saliency are calculated. A comprehensive evaluation is then performed based on the distance and semantic saliency, and the optimal exploration direction is selected according to the comprehensive evaluation result.

3. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 1, characterized in that, The process of obtaining a globally consistent 3D dynamic reference navigation map includes: The environmental perception data is processed using a laser-vision-inertial tightly coupled synchronous positioning and mapping algorithm to generate the 3D point cloud map in real time. The 3D point cloud map is then spatially aligned with the pre-stored mine structure CAD drawings using an iterative nearest point algorithm to obtain a globally consistent 3D dynamic reference navigation map.

4. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 1, characterized in that, The process of identifying and semantically annotating geological risk areas in a 3D dynamic reference navigation map includes: A lightweight 3D target detection network deployed on the UAV's onboard edge computing unit is used to process the 3D dynamic reference navigation map in real time, identify step cracks, loose rock and gravel accumulation areas, and output their category and 3D location information as semantic annotations.

5. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 1, characterized in that, Based on the semantic annotation results, the process of converting the 3D dynamic reference navigation map into a 2D raster map with risk weights includes: The semantically annotated 3D dynamic reference navigation map is projected onto a 2D plane and divided into grids to obtain a 2D grid map. Based on the semantic annotation information corresponding to each grid, a multivariate risk assessment model is used to dynamically calculate the risk weight of each grid. Based on the risk weight values ​​of all grids, a 2D grid map with risk weights is generated.

6. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 5, characterized in that, The multivariate risk assessment model calculates the risk weight of the grid cell using the following formula: ； in, For grid Risk weights, For grid The slope value, For grid Crack density, For grid The ratio of gravel coverage area, For grid Light intensity, These are adjustable weighting coefficients.

7. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 5, characterized in that, The process of planning rescue routes based on a two-dimensional raster map with risk weights includes: Adopting improved A The algorithm performs path planning; it treats each passable grid in the 2D raster map as a path node; based on the improved A... The algorithm's heuristic function calculates the total cost of a node based on the actual cost from the starting point to the node, and searches for and generates the rescue path based on the principle of minimizing the total cost; where the improved A The heuristic function of the algorithm is: ； in, For heuristic functions, Let n be the risk weight. Let n be the slope of node n. Let n be the crack density at node n. This represents the risk sensitivity coefficient.

8. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 1, characterized in that, The process of obtaining the visualized navigation result includes: generating a color-coded risk heat map using a color mapping function based on the risk weight values ​​of each grid in the two-dimensional raster map with risk weights; smoothing and visualizing the rescue path, and overlaying and fusing it with the three-dimensional point cloud map and the risk heat map to obtain the visualized navigation result.

9. The method for autonomous mapping and rescue route visualization generation by unmanned aerial vehicles (UAVs) in post-disaster open-pit mines according to claim 1, characterized in that, It also includes: extracting key structural feature points and their descriptors from the annotated 3D dynamic reference navigation map to construct a relocation anchor point library; performing accurate relocation based on the relocation anchor point library; and updating and maintaining the global consistency of the 3D dynamic reference navigation map.

Images