Unmanned aerial vehicle path planning method and system based on laser scanning

Abstract

The invention relates to the technical field of unmanned aerial vehicle path planning, in particular to an unmanned aerial vehicle path planning method and system based on laser scanning. The method comprises the following steps: scanning a roadway environment in real time by using a laser scanner to obtain original radial section point cloud data; analyzing a structural stability index and a dangerous rock potential energy threat value based on the original radial section point cloud data to determine a risk attribution area of the current flight path of the unmanned aerial vehicle; judging whether a high-risk detour area exists or not according to the risk attribution area, and if yes, dynamically planning a risk detour path so as to adjust a single pre-scanning voyage and generate an intelligent safety execution route; and instructing the unmanned aerial vehicle to fly to the terminal point along the intelligent safety execution route, and taking the terminal point as the starting point of the next round of path planning until the iterative exploration of the target underground mine space is completed. According to the method, the structural stability and potential energy threat of the surrounding rock are quantified in real time to realize safe path planning for actively avoiding surrounding rock instability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle path planning, and particularly relates to an unmanned aerial vehicle path planning method and system based on laser scanning. BACKGROUND

[0002] In long-term mining old roadway, complex geological structure of mining area or experienced ground pressure activity area, the structural stability of the environment is not uniform. The roof and two sides of the roadway are usually internally developed with a large number of joints and through cracks due to the comprehensive influence of stress release, weathering and underground water erosion and other factors. These cracks will gradually expand, eventually leading to the separation of part of the rock mass from the parent rock (main rock mass), forming a "dangerous rock mass" or "suspended body" suspended above the roadway by only a few connection points. These dangerous rock masses are one of the most dangerous major safety hazards in underground mines, and their state is extremely unstable, and they will fall at any time due to a small disturbance (such as equipment vibration, wind flow impact or even the downwash airflow generated by the unmanned aerial vehicle rotor), causing serious safety accidents. The existing unmanned aerial vehicle path planning method usually regards the scanned environment as a geometric structure stable and homogeneous whole, and its core safety logic mainly focuses on "obstacle avoidance" in physical distance, and lacks real-time perception and evaluation ability of the structural stability of the surrounding rock (rock mass around the roadway), and cannot avoid safety accidents caused by rock mass falling. SUMMARY

[0003] Based on this, the present application provides an unmanned aerial vehicle path planning method and system based on laser scanning to solve at least one of the above technical problems.

[0004] To achieve the above purpose, an unmanned aerial vehicle path planning method based on laser scanning comprises the following steps: Step S1: setting a single pre-scanning range, instructing the unmanned aerial vehicle to fly along the range, and using a laser scanner to scan the roadway environment in real time to obtain original radial section point cloud data; Step S2: separating the free point cluster set based on the original radial section point cloud data; analyzing the structural stability index and dangerous rock potential energy threat value of the free point cluster set to determine the risk attribution area of the current unmanned aerial vehicle flight path; Step S3: determining whether there is a high-risk detour area according to the risk attribution area, if there is, dynamically planning a risk detour path to adjust the single pre-scanning range, and generating an intelligent safe execution route; Step S4: instructing the unmanned aerial vehicle to fly along the intelligent safe execution route to its end point, and taking the end point as the starting point of the next round of path planning until the iterative exploration of the target underground mine space is completed.

[0005] The application also provides a UAV path planning system based on laser scanning, which executes the UAV path planning method based on laser scanning as described above, and comprises: A voyage scanning and sensing module is configured to set a single pre-scanning voyage, instruct the UAV to fly along the voyage, and use a laser scanner to scan the tunnel environment in real time to obtain original radial profile point cloud data. A real-time risk assessment module is configured to separate free point clusters from the original radial profile point cloud data, analyze the structural stability index and dangerous rock potential energy threat value of the free point clusters to determine the risk attribution area of the current UAV flight path. An intelligent flight path generation module is configured to determine whether there is a high-risk detour area according to the risk attribution area, and if there is, dynamically plan a risk detour path to adjust the single pre-scanning voyage and generate an intelligent safe execution flight path. A navigation control and iteration module is configured to instruct the UAV to fly along the intelligent safe execution flight path to its endpoint, and use the endpoint as the starting point of the next round of path planning until the iterative exploration of the target underground mine space is completed.

[0006] The application has the following advantages: On the one hand, the application separates and quantitatively evaluates free point clusters by deeply analyzing the original radial profile point cloud data, realizes the leap of the UAV from traditional geometric obstacle avoidance to the core ability of surrounding rock structure stability perception. The prior art only regards all rock surfaces as homogeneous obstacle boundaries, while the application separates potential "dangerous rock bodies" (i.e. free point clusters) and further calculates their structural stability index and dangerous rock potential energy threat value, which can accurately identify and quantify weak areas that pose a potential falling threat to flight safety. Furthermore, this risk assessment mechanism based on structural stability enables the UAV to distinguish between "stable rock walls" and "hanging dangerous rocks", fundamentally solving the major safety defect that the traditional path planning method cannot foresee and avoid rock falling risks, greatly improving the survival ability and operation safety of the UAV in complex geological environments.

[0007] On the other hand, the application dynamically generates an intelligent safe execution route based on the risk assessment results, realizing the path planning mode from passive obstacle avoidance to active risk avoidance. When a high-risk bypass area is identified, the application does not simply mark it as a no-fly zone, but constructs a risk potential field according to the structural stability index and the dangerous rock potential energy threat value, and dynamically plans an optimal risk bypass path considering the safety distance and bypass cost. This dynamic adjustment and re-planning mechanism ensures that the unmanned aerial vehicle can actively move away from the danger source and avoid triggering the dangerous rock to fall due to its own flight disturbance (such as wash airflow). At the same time, the intelligent route not only ensures flight safety, but also can be optimized for performing multi-angle and high-density supplementary scanning tasks on the identified dangerous rock mass, thereby obtaining its accurate geometric and attitude information under the premise of safety, providing unprecedented high-value data support for subsequent manual intervention and disaster control.

[0008] On the other hand, the unmanned aerial vehicle will repeat the cycle of "scanning-evaluation-planning-flight" every time it completes a section of intelligent safe execution route, so it can gradually and safely penetrate into unknown and potentially dangerous areas. This iterative exploration mode ensures that the unmanned aerial vehicle always operates within its latest perceived safe boundary, effectively avoiding the problem of entering unknown dangerous areas due to one-time long-distance planning, significantly enhancing the robustness and success rate of the entire exploration task, and ultimately realizing comprehensive and safe autonomous exploration of the target underground mine space. BRIEF DESCRIPTION OF DRAWINGS

[0009] Fig. 1 The figure is a schematic diagram of the step flow of the unmanned aerial vehicle path planning method based on laser scanning of the application; Fig. 2 The figure is a schematic diagram of the module of the unmanned aerial vehicle path planning system based on laser scanning of the application; The implementation of the object of the application, functional characteristics and advantages will be further described with reference to the embodiments and the accompanying drawings. DETAILED DESCRIPTION

[0010] To achieve the above-mentioned object, please refer to Figs. 1-2 The application provides a kind of unmanned aerial vehicle path planning method based on laser scanning, comprising the following steps: Step S1: set single pre-scanning voyage, instruct unmanned aerial vehicle to fly along the voyage, and obtain original radial profile point cloud data by using laser scanner to scan tunnel environment in real time; In the embodiment of the application, when the unmanned aerial vehicle starts a surveying task, first, the sonar is used for forward-looking detection to determine an initial flight voyage that is safe and has no obvious obstacles. When the unmanned aerial vehicle flies along this voyage, the laser scanner carried by the unmanned aerial vehicle continuously scans the surrounding environment in 360 degrees, and collects high-density three-dimensional point cloud data.

[0011] In one implementation manner of the embodiment of the present application, assuming that the UAV is located at the entrance of the tunnel, the wideband sonar detector carried by the UAV detects a 120° sector in front of the UAV, the maximum detection distance is 50 meters, and a group of initial acoustic echo data is obtained. Analyzing the echo data, it is identified that the farthest passable distance is 42 meters, a safety front-out coefficient of 80% is set, and the length of the single pre-scanning voyage is calculated to be 42 meters 80% = 33.6 meters, according to the geometric center of the echo profile, a central axis pointing to the depth of the tunnel and having a length of 33.6 meters is generated as the single pre-scanning voyage. The UAV is instructed to fly at a uniform speed along the 33.6-meter voyage, and in this process, the dual-axis rotating laser scanner carried by the UAV continuously scans the plane orthogonal to the flight direction at a frequency of 20 Hz. The scanner generates a radial profile containing about 1800 laser points every rotation, which constitutes the original radial profile point cloud data.

[0012] Step S2: separating free point clusters from the original radial profile point cloud data; analyzing the structural stability index and the dangerous rock potential energy threat value of the free point clusters to determine the risk attribution region of the current UAV flight path; In one implementation manner of the embodiment of the present application, the obtained original radial profile point cloud data is voxelized, and the grid size is set to 10cmx10cmx10cm. For each voxel, the position eccentricity between the geometric center of all laser pulse points inside and the geometric center of the corresponding voxel is calculated, and the standard deviation of the echo intensity values of all points is obtained to obtain the intensity fluctuation value. The geometric aggregation threshold is set to 5 cm, and the material uniformity threshold is set to 15. When the position eccentricity of a voxel is less than 5 cm and the intensity fluctuation value is less than 15, the voxel is marked as a high-confidence voxel. The centroid set of the pulse points in all high-confidence voxels constitutes the structural confidence point cloud. The region growing algorithm is performed on the structural confidence point cloud to identify and extract the largest connected point cloud set in volume, which is determined as the main rock mass. Then, the structural confidence point cloud and the main rock mass are subjected to Boolean difference set operation to obtain the initial overhanging body. The Euclidean distance clustering analysis is performed on the initial overhanging body, and the clustering distance threshold is set to 50 cm. After clustering, the small clusters (noise) with a point number less than 50 are filtered out, and the remaining point cloud clusters are the free point clusters. Assuming that three free point clusters are identified, they are marked as C1, C2 and C3 respectively. Taking C1 as an example, assuming that its structural stability index is 1.4 and the dangerous rock potential energy threat value is 15.68, and the high-risk screening threshold is set to 1.5, since the structural stability index of C1 is 1.4 which is less than 1.5 and the dangerous rock potential energy threat value 15.68 is high, the region where C1 is located and the surrounding space are marked as a high-risk bypass region, and other regions are normal flight regions.

[0013] Step S3: Determine whether there is a high-risk bypass area according to the risk attribution area, if there is, dynamically plan a risk bypass path to adjust the single pre-scanning voyage and generate an intelligent safe execution route; In the embodiment of the present application, once a high-risk area is detected, the original straight pre-scanning voyage will be abandoned, for example, for the high-risk free point cluster C1 identified in the previous step, a spatial influence radius of 10 meters is set with the centroid as the center. Within this radius, a Gaussian potential field function is constructed, the peak intensity A is proportional to the dangerous rock potential energy threat value ( ), and the decay width sigma is proportional to the inverse of the structural stability index ( ). The potential field data is superimposed on the three-dimensional space grid map to form the flight risk potential field data. In the flight risk potential field data, the starting point (current position of the UAV) and the end point (original 33.6 meters) of the original "single pre-scanning voyage" are taken as the starting point and target point of the path search. The path cost function is set: Cost=w1 path length+w2 cumulative total value of risk potential field crossed+w3 estimated energy consumption, wherein the weights w1, w2 and w3 are adjustable. The fast expanding random tree algorithm is used to search in the risk potential field, and the algorithm will tend to avoid areas with high potential field values when expanding (because it will significantly increase the cost). When the search reaches the maximum number of iterations (such as 5000 times) or finds a path with a cost lower than the preset threshold, the current lowest cost feasible path is output as the risk bypass path, which gracefully bypasses the high-risk area where C1 is located and finally reaches the original target point. The new path is the intelligent safe execution route.

[0014] Step S4: instruct the UAV to fly along the intelligent safe execution route to its end point, and take the end point as the starting point of the next round of path planning until the iterative exploration of the target underground mine space is completed.

[0015] In the embodiment of the present application, the generated intelligent safe execution route (represented as a series of three-dimensional space waypoints) is sent to the flight control system of the UAV, and the flight control system drives the UAV to fly accurately along the safe path. When the UAV flies to the end point of the intelligent safe execution route (a new position 33.6 meters deep) and hovers stably, it marks the completion of this exploration cycle, and immediately takes the current position of the UAV as the starting point of the next round of path planning. Subsequently, the whole process returns to step S1: the UAV performs sonar forward detection from the new position again, sets a new "single pre-scanning voyage", and repeats the complete closed loop of "scanning-evaluating-planning-flying". This iterative process continues, advancing safely into the tunnel a little at a time. When the sonar detects the end of the tunnel or the total exploration mileage reaches the preset value, the task is completed.

[0016] Preferably, step S1 comprises the following steps: Step S11: Before the UAV enters the target underground mine space, a fan-shaped acoustic detection is performed on the front area by using the wideband sonar detector carried by the UAV, and initial acoustic echo data containing main contour reflection information is obtained; Step S12: Based on the farthest echo boundary identified in the initial acoustic echo data, a safe front distance is calculated and set, and a single pre-scanning range is obtained; Step S13: The UAV is instructed to fly along the central axis of the single pre-scanning range to the end point of the single pre-scanning range, and in this process, a dual-axis rotating laser scanner arranged orthogonally to the flight direction is used to perform real-time radial scanning on the sides of the roadway and the roof and floor of the target underground mine space, and original radial profile point cloud data is obtained.

[0017] In the embodiment of the present application, at the starting point of each iteration exploration, the UAV first performs a forward long-distance rough detection, which aims to quickly evaluate the accessibility and general contour of the front space, and provides a basis for setting a safe flight scanning distance. In order to reduce cost and power consumption, acoustic detection method is used in this stage.

[0018] In one implementation manner of the embodiment of the present application, the UAV carries a remote ultrasonic sensor array composed of three transmitting / receiving units, which can detect a fan-shaped area of -60° to +60° in the front horizontal direction. The sensor array transmits a series of coded ultrasonic pulses to -60°, -40°,..., +40°, +60° in turn with a step angle of 20°. After the detection in each direction is completed, the receiving unit captures a series of acoustic signals reflected by the roadway wall, obstacles or the far boundary. After amplification and filtering, these signals are recorded as initial acoustic echo data containing flight time and signal amplitude. Specifically, for the detection in the 0° (forward) direction, assuming that the received initial acoustic echo data is a time-domain waveform in which a significant peak appears at t1=0.25s and a weaker peak appears at t2=0.28s, which represents two main reflection surfaces in the forward direction.

[0019] In one implementation manner of the embodiment of the present application, the processing unit analyzes the initial acoustic echo data in seven directions one by one. For each direction, a constant threshold detection method is used to identify the valid peak value in the echo signal, and the flight time of the last valid peak value is recorded This is considered as the time corresponding to the farthest detection distance in this direction.

[0020] In another implementation manner of the embodiment of the present application, assuming that among the seven detection directions, the flight time of the last valid peak value is calculated is 0.25 seconds, the speed of sound in air is known to be approximately 340 m / s, then the distance to the farthest echo boundary is calculated as follows: The calculation is as follows: Note that the result is divided by 2 because the sound wave travels a distance of one way there and one way back.

[0021] The final safe forward distance is then calculated based on this farthest echo boundary distance, and a pre-defined safety factor is introduced The safety factor is set to reserve enough reaction and braking distance. Its mathematical expression is as follows: ; Where: is the safe forward distance; is the identified farthest echo boundary distance; is a dimensionless coefficient between 0 and 1, which is dynamically adjusted according to the current flight speed of the UAV, and its empirical model can be set as where is the UAV speed (m / s), and assuming that the UAV is currently in a hovering state , then .

[0022] In an implementation manner of the embodiment of the application, the UAV flight control system receives a single pre-scanning flight instruction with a length of 38 meters, and the flight control system generates a straight line track with a length of 38 meters as a central axis, taking the current position of the UAV as the starting point and the front direction. The UAV flies along this axis at a constant speed of 1.0 m / s. Specifically, the rotating axis of the laser scanner (for example, Velodyne VLP-16) carried on the UAV maintains a strict orthogonal relationship with the flight direction (central axis) of the UAV, and when the UAV flies forward, the 16-line laser beams of the scanner rotate 360° horizontally at a frequency of 20 Hz, forming a “scanning disc surface” that moves forward continuously and is perpendicular to the flight path. This scanning disc surface cuts the tunnel space in real time, and the laser beams scan the right side, the roof, the left side and the floor of the tunnel in turn, and then return to the right side to complete a radial scan. Each laser emitter returns one or more echo points in each scan, and the three-dimensional coordinates (based on the scanner's own coordinate system) and echo intensity values of the echo points are combined with the real-time pose information (position and attitude) provided by the inertial measurement unit (IMU) of the UAV, and are converted to the global coordinate system through coordinate transformation. All these point cloud sets after coordinate transformation jointly constitute the original radial profile point cloud data of this flight.

[0023] Preferably, before the isolated free point cluster set is separated from the original radial profile point cloud data in step S2, the method further comprises:​​ The original radial profile point cloud data is subjected to three-dimensional space rasterization processing to obtain three-dimensional scanning point cloud data; The three-dimensional scanning point cloud data is subjected to spatial voxel grid division, and then the laser pulse points in the three-dimensional scanning point cloud data are classified into corresponding spatial voxel grids to obtain voxelized pulse data; Based on the voxelized pulse data, the Euclidean distance between the three-dimensional coordinate geometric center of all laser pulse points in each voxel and the corresponding voxel geometric center is calculated to obtain a position eccentricity; Meanwhile, the standard deviation of the echo intensity values of all laser pulse points in each voxel in the voxelized pulse data is calculated to obtain an intensity fluctuation value; It is judged whether the position eccentricity is less than a preset geometric aggregation threshold value and whether the intensity fluctuation value is less than a preset material uniformity threshold value, and if so, the voxel is marked as a high-confidence voxel; Based on the high-confidence voxel, the centroid points of the pulse points in the voxel are analyzed, and the set of all centroid points constitutes a structure confidence point cloud.

[0024] In the embodiment of the application, each collected data frame is traversed, and the point cloud in the local coordinate system of the frame is transformed into the global coordinate system (usually taking the task starting point as the origin) by using the pose information (a 4x4 homogeneous transformation matrix) therein. All the transformed point clouds are combined together to form three-dimensional scanning point cloud data covering the entire single pre-scanning flight area.

[0025] In one implementation manner of the embodiment of the application, the edge length of the voxel is set to 10 cm, a three-dimensional voxel grid covering the entire range of the three-dimensional scanning point cloud data is created, each laser pulse point in the three-dimensional scanning point cloud data is traversed, the voxel index (i, j, k) to which the laser pulse point belongs is calculated according to its three-dimensional coordinates (x, y, z), and the point is stored in the voxel corresponding to the index. All non-empty voxels and the set of laser pulse points contained therein constitute voxelized pulse data.

[0026] In the embodiment of the application, the three-dimensional coordinate geometric center of all laser pulse points in each voxel is calculated, and the geometric center of the voxel itself is calculated at the same time, and then the Euclidean distance between the two center points is calculated, which is defined as the position eccentricity. For a voxel V containing n points, the position eccentricity of the voxel V is The calculation process of the position eccentricity is as follows: the three-dimensional coordinate geometric center of all n laser pulse points in the voxel is calculated The geometric center of the voxel V itself is obtained ​​​The coordinates of which are determined directly by the voxel index and the edge length. The Euclidean distance between the two center points is calculated, i.e. the position eccentricity : ; wherein: is the position eccentricity, in meters; is the geometric center coordinate of the point cloud within the voxel; is the geometric center coordinate of the voxel itself. Specifically, for each non-empty voxel in the voxelized pulse data, the standard deviation of the echo intensity values of all laser pulse points within the voxel is calculated, which is defined as the intensity fluctuation value. Then the position eccentricity is compared with a preset geometric concentration threshold, and the intensity fluctuation value is compared with a preset material uniformity threshold. It should be noted that the geometric concentration threshold is set to 5 cm in the present case, and the value is set according to the ranging accuracy of the laser scanner and the general flatness of the roadway surface; the material uniformity threshold is set to 15 in the present case, which is set based on the prior knowledge of the statistical distribution of the echo intensity of typical rock, coal wall and other materials.

[0027] In the embodiment of the present application, after the judgment of all non-empty voxels is completed, only the data of all voxels marked as high-confidence voxels is retained, the centroid points (i.e. the average value of three-dimensional coordinates) of all pulse points within them are calculated, and the set of all these centroid points constitutes a new point cloud data set that is purified and down-sampled, which is the structure-confidence point cloud.

[0028] Preferably, the step S2 of separating the free point cluster set from the original radial profile point cloud data includes: Performing a region growing operation on the structure-confidence point cloud to identify and extract the largest connected point cloud set in volume, and determining this set as the main rock mass; Performing a point cloud Boolean difference set operation on the structure-confidence point cloud and the main rock mass to separate all points not belonging to the main rock mass, and obtaining the initial overhang; Performing a Euclidean distance clustering analysis on the initial overhang to segment the dispersed point clouds on the initial overhang into multiple independent point cloud clusters, and obtaining the point cloud cluster set; Calculating the number of points in each cluster in the point cloud cluster set, and filtering out small clusters with a number of points less than a preset noise point number threshold, and the remaining cluster set is the free point cluster set.

[0029] In the embodiment of the present application, the region growing algorithm based on normal vector and distance constraint is adopted, and the point with the lowest Z coordinate in the structure confidence point cloud is selected as the initial seed point, because this point is most likely to belong to the stable roadway floor. Then, the k-d tree structure is used for efficient neighborhood search, and the adjacent points with a distance less than 25 cm and a normal vector angle less than 10 degrees from the point cloud in the current region are added to the region. This process is iterated until no new point can be added, thereby forming a connected point cloud set. The process is repeated for all unvisited points until all points are classified into a certain set.

[0030] In an implementation manner of the embodiment of the present application, it is assumed that the total number of the original structure confidence point cloud is 509,410, and the main rock mass determined in the previous step contains 485,120 points, and then the point cloud Boolean difference set operation is performed, that is, all points existing in the main rock mass set are removed from the total set of the structure confidence point cloud. Specifically, the remaining point cloud set obtained after the difference set operation is the initial overhanging body, which contains 24,290 points.

[0031] In the embodiment of the present application, the initial overhanging body contains multiple potential dangerous rock bodies separated in space, and this step effectively separates these spatially independent entities through the clustering algorithm. It should be noted that the Euclidean distance clustering is adopted because the spatial proximity is the most basic basis for judging whether the point cloud belongs to the same physical entity. The Euclidean distance clustering analysis is performed on the initial overhanging body containing 24,290 points, and the key parameter clustering tolerance is set to 50 cm. The threshold is set according to the fact that it is significantly greater than the average point spacing of the structure confidence point cloud (about 10 cm), but less than the safety spacing between the independent dangerous rock bodies, thereby effectively distinguishing different overhanging rock blocks. Specifically, after the clustering analysis is completed, the initial overhanging body is divided into four independent point cloud clusters, which are named cluster 1, cluster 2, cluster 3 and cluster 4, and the number of points is 15,340, 8,300, 380 and 270 respectively. The set of the four point cloud clusters, that is, the point cloud cluster set.

[0032] In one implementation form of the embodiment of the present application, a noise point threshold value of 500 points is set, which is based on engineering experience. An entity composed of less than 500 centroid points (at a 10 cm voxel resolution) is too small in physical volume and is generally not considered a significant safety hazard that requires path avoidance. Specifically, the point cloud cluster set obtained in the previous step is traversed to check the number of points in each cluster. Since the number of points in cluster 1 (15,340 points) and cluster 2 (8,300 points) is greater than 500, they are retained. Cluster 3 (380 points) and cluster 4 (270 points) are determined to be noise and filtered out because their point numbers are less than 500. Therefore, the final remaining cluster set is the free point cluster set containing only cluster 1 and cluster 2.

[0033] Preferably, the step S2 of analyzing the structural stability index and the dangerous rock potential energy threat value of the free point cluster set to determine the risk attribution region of the current UAV flight path comprises: searching for a point set in the free point cluster set and the surface points in the main rock mass whose spatial distance is less than a preset connection distance threshold value, and defining the point set as a connection point set; analyzing the hanging index and the fracture density value in the free point cluster set based on the connection point set; weighting calculation according to the hanging index and the fracture density value to evaluate the structural stability of the connection region and obtain the structural stability index; screening a high-risk free point cluster in the free point cluster set based on a preset high-risk screening threshold value and the structural stability index; calculating the local curvature or surface variation degree of each point in the high-risk free point cluster, and identifying a point set with a curvature or surface variation degree greater than a preset convexity threshold value as a convex shape to segment an isolated convex cluster; evaluating the dangerous rock potential energy threat value according to the isolated convex cluster; determining the risk attribution region of the current UAV flight path based on the structural stability index and the dangerous rock potential energy threat value; wherein the risk attribution region comprises a normal flight region and a high-risk detour region.

[0034] In the embodiment of the present application, for each free point cluster identified in the previous step (taking cluster 1 containing 15,340 points as an example), a k-d tree data structure is used to quickly search for the nearest neighbor point of each point in cluster 1 in the main rock mass. The preset connection distance threshold value is set to 15 cm, which is slightly larger than the voxel resolution of the point cloud (10 cm) and can effectively capture physically adjacent or connected point pairs through a slight crack. Therefore, when it is detected that the Euclidean distance between a point in cluster 1 and its nearest neighbor point in the main rock mass is less than 15 cm, the pair of points is simultaneously stored in a temporary set. After traversing all points in cluster 1, the temporary set is determined as the connection point set. It is assumed that the final connection point set contains 850 points.

[0035] In an implementation form of the embodiment of the application, the calculation of the hanging index takes the volume of the entire free point cluster and the connection surface area formed by the connection point set as inputs, while the calculation of the fracture density value focuses on the point cloud normal vector distribution characteristics of the connection point set in the neighborhood of one side of the host rock.

[0036] In an implementation form of the embodiment of the application, assuming that the hanging index of cluster 1 is 1.5 (the larger the value, the less stable) and the fracture density value is 0.4 (the larger the value, the less stable) obtained through detailed calculation, the following weighted reciprocal sum model is used to calculate the structure stability index : ; wherein, is the structure stability index, the smaller the value, the less stable the structure; is the hanging index; is the fracture density value; and : the weights of the hanging index and the fracture density value respectively, the sum of which is 1, which are empirically set to 0.6 and 0.4 respectively here, to indicate that the hanging form occupies a more important position in stability evaluation.

[0037] In an implementation form of the embodiment of the application, the structure stability index of each free point cluster calculated in the previous step is compared with a preset high-risk screening threshold, which is set to 1.5, which is a critical value determined based on mine safety regulations and ground pressure disaster statistical data, and a structure is considered to have a higher instability possibility if it is higher than this value.

[0038] In an implementation form of the embodiment of the application, for each point in the high-risk free point cluster, a local quadratic surface is fitted based on its neighborhood points (for example, all points within a radius of 0.2 meters), and the principal curvatures of the point are calculated. The larger one of the two principal curvatures is taken as the local curvature value of the point, which can effectively reflect the convexity degree of the position of the point. The preset convexity threshold is set to 0.8, which is used to distinguish the sharp, convex edges or corner points on the rock mass from the relatively flat surface. The set of all points with a local curvature greater than 0.8 is identified as a convex form, and a small-range Euclidean distance clustering is performed on these points, and the largest connected part is segmented out to form an isolated convex cluster.

[0039] In the embodiment of the present application, as long as at least one high-risk isolated point cluster (such as cluster 1 in this example) is identified in the scene, a high-risk bypass area is defined according to its spatial position and size, the axis-aligned bounding box of the high-risk isolated point cluster 1 is calculated, and a safety buffer of 5 meters is extended in each direction of the bounding box, and the enlarged three-dimensional space is defined as the high-risk bypass area. Therefore, in the entire three-dimensional space covered by the current single pre-scanning voyage, all other areas except the explicitly marked high-risk bypass area are determined as normal flight areas.

[0040] Preferably, the isolated protruding cluster is evaluated to obtain the dangerous rock potential energy threat value, including: A minimum convex envelope of all outer layer points in the isolated protruding cluster is constructed to obtain a dangerous rock convex envelope; The number of all non-empty voxel units containing at least one point cloud data in the envelope range of the isolated protruding cluster is counted, and the number is multiplied by the volume of a single voxel unit to obtain a solid filling product; The solid filling product is compared with the total volume of the dangerous rock convex envelope to calculate the ratio of the development degree of internal cavities or cracks as the structure looseness; The dangerous rock potential energy threat value is determined based on the volume of the dangerous rock convex envelope and the structure looseness, and according to the height of the centroid point in the isolated protruding cluster relative to the roadway floor.

[0041] In the embodiment of the present application, a three-dimensional QuickHull algorithm is used to process the isolated protruding cluster containing 1,200 points segmented from the high-risk isolated point cluster in the previous step. The algorithm can efficiently find a minimum convex polyhedron containing all points. After the algorithm is executed, a closed three-dimensional mesh model composed of 98 triangular facets is generated. By calculating the volume enclosed by the closed mesh, it is obtained that the total volume is 1.5 cubic meters. The geometric body and the volume are jointly defined as the dangerous rock convex envelope.

[0042] In an implementation manner of the embodiment of the present application, in the geometric range of the dangerous rock convex envelope, all voxel units with an edge length of 10 centimeters are counted, and non-empty voxel units containing at least one isolated protruding cluster point cloud data inside are counted. Assuming that the total number of counted non-empty voxel units is 11,500, the volume of a single voxel unit is 0.001 cubic meters, and the two are multiplied to obtain a solid filling product. It should be noted that all voxels in the envelope range (i.e., the dangerous rock convex envelope) of the isolated protruding cluster are traversed, and it is checked whether each voxel contains at least one point cloud data from the cluster. If yes, the voxel is marked as a non-empty voxel unit. The total number of non-empty voxel units is multiplied by the volume of a single voxel unit, so as to obtain a volume value that can more truly reflect the solid part of the rock mass.

[0043] Specifically, the average value of three-dimensional coordinates of all points in the isolated convex cluster is calculated to obtain a centroid point, and then the vertical distance from the centroid point to the pre-identified roadway floor plane is calculated to obtain the height of the centroid point 8.5 meters. In an implementation form of the embodiment, the dangerous rock potential energy threat value is calculated using the following formula : ; wherein: is the dangerous rock potential energy threat value, a comprehensive risk indicator; is the total volume of the dangerous rock convex hull, representing the maximum mass scale thereof; is a correction coefficient taking into account the internal fragmentation degree, the higher the porosity, the greater the coefficient, meaning that it will break into more small pieces in the falling process, affecting a wider range; is the centroid point height, representing the initial potential energy.

[0044] Preferably, the analysis of the hanging index and the fracture density value in the free point cluster based on the connection point set comprises: a local fitting surface is constructed based on the connection point set, and then the surface area of the local fitting surface is calculated as the connection area; the minimum circumscribed geometric body of each free point cluster in the free point cluster is analyzed, and the overhanging volume is calculated; the overhanging volume is divided by the connection area to obtain the hanging index; the neighborhood point cloud of the connection point set on one side of the main rock mass is extracted to obtain the connection area parent rock point cloud; the normal vector of each point in the connection area parent rock point cloud is calculated, and then the dispersion degree of the vector direction in the normal vector is counted and quantified as a scalar value to obtain the fracture density value.

[0045] In the embodiment, the ball pivoting algorithm is used to process the connection point set containing 850 points obtained in the previous step. The algorithm constructs a triangular mesh surface model by simulating a virtual small ball rolling on the surface of the point cloud. The radius of the virtual small ball is set to 20 centimeters, which is slightly larger than the average distance of the point cloud, sufficient to cross small cavities but not to smooth important geometric features too much. Finally, a local fitting surface composed of 1,580 small triangles is generated. By accumulating the areas of all triangles, the total surface area of the surface is calculated to be 0.8 square meters, which is determined as the connection area.

[0046] It should be noted that the minimum circumscribed geometric body (especially the minimum circumscribed bounding box with variable direction) can more compactly and accurately wrap the irregularly shaped point cloud than the axis-aligned bounding box, thereby avoiding overestimation of the volume.

[0047] In one implementation of the embodiment of the application, for the cluster 1 (containing 15,340 points) in the cluster set of free points, the principal component analysis (PCA) method is applied, first, three principal directions (eigenvectors) of the point cloud distribution are calculated, then the point cloud is rotated to align with these principal directions, and the maximum and minimum coordinate ranges thereof in the new coordinate system are calculated, thereby constructing the minimum circumscribed bounding box of the point cloud. Specifically, the length, width and height of the minimum circumscribed bounding box are calculated to be 2.0 meters, 1.0 meter and 0.6 meter respectively, and thus the volume thereof is 2.0 1.0 0.6=1.2 cubic meters, which is determined as the overhanging volume.

[0048] Specifically, the following mathematical formula is used to calculate the hanging index : ; wherein: is the hanging index (dimensionless), the greater the value, the greater the volume supported per unit of connecting area, and the more unstable the structure; is the overhanging volume calculated in the previous step (unit: cubic meters); is the connecting area calculated in the first step (unit: square meters); the obtained value is substituted into the formula: =1.2 cubic meters / 0.8 square meters=1.5, and thus the hanging index of the free point cluster is determined to be 1.5.

[0049] In one implementation of the embodiment of the application, from the connecting point set containing 850 points, all points belonging to the host rock (assuming 420 points) are screened out, and then a radius search is performed within the host rock point cloud with the 420 points as the center, and the search radius is set to 30 centimeters. The radius aims to include sufficient environmental information near the connecting points, while avoiding the introduction of too far and irrelevant point cloud data. All host rock points within the search radius are merged and duplicate points are removed, and finally a point cloud subset containing 2,500 points is obtained, which is defined as the connecting area host rock point cloud.

[0050] In the embodiment of the application, for each point in the connecting area host rock point cloud, a local plane is fitted using its 20 nearest neighbor points, and the normal vector of the plane is taken as the normal vector of the point After calculating the normal vectors of all 2,500 points, the average vector of all normal vectors is calculated Then, the included angle between each normal vector and the average vector is calculated Finally, all the included angles standard deviation of the angle and take it as the fissure density value Therefore, assuming that the final calculated standard deviation of the angle is 0.4 radian, the fissure density value of the connection region is determined as 0.4, and the greater the value, the more discrete the normal vector distribution, the more uneven the rock surface, and the more developed the potential fissure.

[0051] Preferably, in step S3, it is determined whether there is a high-risk bypass region according to the risk attribution region, and if there is, a risk bypass path is dynamically planned to adjust the single pre-scan flight range, including: If the risk attribution region is a normal flight region, the UAV continues to perform the single pre-scan flight; If the risk attribution region is a high-risk bypass region, a spatial influence radius is set according to the high-risk bypass region; A Gaussian potential field function of risk intensity decay with distance is constructed based on the dangerous rock potential threat value and the structure stability index within the spatial influence radius with the centroid of the high-risk bypass region as the center, and the high-risk bypass region is expanded outwardly by a safety distance to obtain flight risk potential field data; wherein the dangerous rock potential threat value is mapped to the peak intensity of the Gaussian potential field function, and the inverse of the structure stability index is taken as a parameter affecting the decay rate thereof; A risk bypass path is dynamically planned according to the flight risk potential field data; The single pre-scan flight range is dynamically adjusted and re-planned through the risk bypass path to generate an intelligent safe execution route.

[0052] In the embodiment of the present application, all identified free point clusters are analyzed, and the calculated structure stability index is higher than the preset high-risk screening threshold (for example, 1.5), and it is determined that the risk attribution region of the current exploration area is all normal flight region. Therefore, the flight control system will not start the path re-planning program, but directly execute the original single pre-scan flight range set by the sonar detection in step S1, that is, instructing the UAV to fly along the predetermined 33.8-meter central axis in a straight line to the end point.

[0053] In another implementation manner of the embodiment of the present application, it is assumed that the identified free point cluster 1 in step S2 is determined as a high-risk free point cluster, thereby determining the existence of a high-risk bypass region. Specifically, the minimum circumscribed bounding box of the high-risk free point cluster 1 is calculated to obtain a diagonal length of 2.3 meters, and then a 5-meter safety buffer distance is added on this basis to set a spatial influence radius Therefore, the finally determined spatial influence radius is 2.3 meters + 5 meters = 7.3 meters, which means that the subsequent risk potential field will take effect within this radius range.

[0054] In the embodiment of the present application, the centroid coordinates of the high-risk free point cluster 1 are calculated as Then a three-dimensional Gaussian potential field function is constructed The mathematical expression is: ; Wherein: The risk intensity value of any point in space ; The peak intensity of the Gaussian potential field, which is mapped from the dangerous rock potential energy threat value , , It is a preset weight coefficient, which is used to scale the potential energy threat value to the appropriate risk cost order of magnitude, which is set to 10 here; σ is the standard deviation of the Gaussian function, which determines the decay rate (influence range) of the potential field, which is determined by the reciprocal of the structural stability index SSI, , It is a preset scale coefficient, which is used to adjust the size of the influence range, which is set to 2 here.

[0055] In the embodiment of the application, a rapid extended random tree star algorithm is used to search for a path in the flight risk potential field data, with the starting point of the original single-time pre-scanning flight path (the current position of the unmanned aerial vehicle) as the starting point and the endpoint (33.8 meters away) as the target point.

[0056] In one implementation manner of the embodiment of the application, the entry point and the exit point where the original single-time pre-scanning flight path (38-meter straight line) intersects with the flight risk potential field data are determined, the straight line segment between the entry point and the exit point in the original flight path is deleted, and then the risk bypass path composed of a series of waypoints obtained by the planning in the previous step is inserted into the deleted position.

[0057] In another implementation manner of the embodiment of the application, in order to ensure the smoothness of the flight of the unmanned aerial vehicle, the B-spline curve is applied to the spliced waypoint sequence for smoothing processing, and a continuous and curvature-varying flight path is generated. The new, safe and feasible path is directly encapsulated and named as an intelligent safe execution flight path, which is then issued to the flight control system of the unmanned aerial vehicle for execution.

[0058] Preferably, the dynamic planning of the risk bypass path according to the flight risk potential field data comprises: Determining the starting point and the target point of the bypass path in the flight risk potential field data based on the single-time pre-scanning flight path; Setting a path cost function, which includes three components of path length, cumulative total value of risk potential field crossed and flight energy consumption; Performing a rapid extended random tree search on the flight risk potential field data by using the path cost function to explore a feasible path connecting the starting point and the target point; When the convergence condition is met or the maximum number of iterations is reached, the current lowest-cost feasible path is output as the risk-avoiding path.

[0059] In an implementation form of the embodiment of the application, the target point of the avoiding path is set as the end point coordinate of the original single pre-scan voyage, which is calculated by translating the starting point coordinate along the positive direction vector of the UAV nose direction by 33.8 meters, and the two points jointly define the solution boundary of the path planning problem.

[0060] In an implementation form of the embodiment of the application, the set path cost function has a specific mathematical expression as follows: ; Wherein: is a candidate path from the starting point to the target point; is the total path length of the path , which is directly calculated by integrating the path; is the total value of the risk potential field accumulated by the path , which is obtained by accumulating the risk intensity values of each grid passed by the path; is the estimated flight energy consumption of the path , and its calculation model is , wherein is the curvature of the path, is the total elevation change, is the energy consumption coefficient, which respectively represents the energy consumption related to distance, turning and climbing; , , are the weight coefficients of the three components respectively, which are set according to the task priority, and their sum is 1. In this case, in order to give priority to safety, they are set to 0.2, 0.6 and 0.2 respectively.

[0061] In an implementation form of the embodiment of the application, the algorithm starts from the starting point, randomly generates a sampling point in the three-dimensional space, then finds the node in the tree closest to the sampling point, and expands a fixed step (for example, 1 meter) in the direction of the sampling point. If the newly generated node and its connected path do not collide with any region with a risk value exceeding a preset safety threshold (for example, 80% of the peak intensity), the new node is added to the tree, and the nodes in its neighborhood are checked to try to reconnect through the new node to optimize the path cost. This process is repeated until a branch of the tree reaches the vicinity of the target point.

[0062] In the embodiment of the present application, this step defines the termination condition of the search algorithm, and outputs the final planning result when the condition is met. Specifically, two parallel termination conditions are set: one is the convergence condition, that is, in the next 100 iterations, the improvement of the optimal path cost value found is less than 0.1%, which indicates that the algorithm has basically converged; the second is the maximum number of iterations, that is, the total number of iterations of the algorithm reaches the preset upper limit of 5000 times, to ensure that the planning process is completed within a limited time.

[0063] In an implementation manner of the embodiment of the present application, it is assumed that a feasible path connecting the starting point and the target point is found at the 3800th iteration, and in the next 100 iterations, the cost function value is reduced from 125.8 to 125.7, and the improvement rate is less than 0.1%, which meets the convergence condition. Therefore, the search process is immediately terminated, and the feasible path recorded in the current tree with a cost value of 125.7 is output as the final planning result, and the path is determined as the risk bypass path.

[0064] Especially important is that when the current square region is determined as a high-risk bypass region, the path planning process specifically further includes: extracting a high-risk isolated point cluster in the high-risk bypass region, and locking the geometric center point of the cluster as the target of the sensor aiming, defined as a target to be surveyed; instructing the unmanned aerial vehicle to immediately stop advancing and adjust its own position to maintain a preset safe observation distance from the target to be surveyed, to enter a survey standby state; based on the intelligent safety, performing a flight path planning to cover multiple key observation viewing angles of the survey scanning waypoints; based on the survey standby state, setting a gimbal pitch angle and yaw angle instruction for the survey scanning waypoints, to ensure that the laser scanner always focuses on the target to be surveyed throughout the flight, to obtain a multi-viewpoint column; dynamically adjusting the scanning line density of the laser scanner, and locally encrypting the multi-viewpoint column to obtain enhanced survey data.

[0065] In an implementation manner of the embodiment of the present application, the high-risk isolated point cluster triggering the generation of the high-risk bypass region (i.e. cluster 1 identified in the previous step) is directly extracted from the definition data of the high-risk bypass region, the three-dimensional coordinates of the 15,340 points contained in cluster 1 are traversed, the geometric center point coordinates of the point cluster are calculated by averaging the X, Y and Z coordinates of all points, and the coordinates are assumed to be (15.2, 8.5, 9.1). The coordinate point is locked and officially defined as the target to be surveyed in the task.

[0066] In an implementation form of the embodiment of the application, the preset safe observation distance is set to 8 meters, which is determined by comprehensively considering the optimal working range of the laser scanner and sufficient reaction space reserved for emergency situations, a direction vector between the current position of the UAV and the target to be surveyed is calculated, and the UAV is moved to a position point 8 meters away from the target in the opposite direction of the vector. After the UAV hovers stably at this position, its state is updated to the survey standby state.

[0067] In the embodiment of the application, 7 survey scanning waypoints are planned on a safe hemispherical surface tangent to the original intelligent safe execution route, with the target to be surveyed as the spherical center and the safe observation distance of 8 meters as the radius. These waypoints are distributed in a circular arc shape and cover the key observation angles from 45 degrees left front to 45 degrees right front of the target to be surveyed. Specifically, for each of the 7 survey scanning waypoints generated in the previous step , a direction vector from the waypoint to the target to be surveyed is calculated.

[0068] In an implementation form of the embodiment of the application, the direction vector is used to calculate the yaw angle and the pitch angle of the gimbal that the UAV needs to set at the waypoint: ; ; wherein: is a two-parameter arctangent function for accurately calculating the yaw angle, is the length of the vector .

[0069] In the embodiment of the application, before the UAV starts to perform the flight task of the multi-view waypoint list, an instruction is sent to the controller of the laser scanner to dynamically lower the rotation scanning frequency from 20 Hz for navigation to 5 Hz for fine scanning. It should be noted that the reduction of the scanning frequency means that the distribution of the laser scanning beam on the target surface will become more intensive in unit time, thereby significantly improving the density of the point cloud. Therefore, after the UAV flies along the multi-view waypoint list and continuously scans the target, the point cloud data obtained by the UAV is about 4 times denser than the conventional scanning data on the surface of the target to be surveyed. This high-density point cloud set after local encryption is defined as enhanced survey data, The application further provides a UAV path planning system based on laser scanning, which performs the UAV path planning method based on laser scanning as described above. The UAV path planning system based on laser scanning comprises: ​The voyage scanning and sensing module S101 is configured to set a single pre-scanning voyage, instruct the UAV to fly along the voyage, and use a laser scanner to scan the tunnel environment in real time to obtain original radial profile point cloud data; The real-time risk assessment module S102 is configured to separate free point clusters from the original radial profile point cloud data, analyze the structural stability index and dangerous rock potential energy threat value of the free point clusters, and determine a risk attribution area of the current UAV flight path. The intelligent route generation module S103 is configured to determine whether there is a high-risk detour area according to the risk attribution area, dynamically plan a risk detour path if there is a high-risk detour area, adjust the single pre-scanning voyage, and generate an intelligent safe execution route. The navigation control and iteration module S104 is configured to instruct the UAV to fly along the intelligent safe execution route to its endpoint, and use the endpoint as a starting point for the next round of path planning, until the iterative exploration of the target underground mine space is completed.

[0070] Therefore, from any point of view, the embodiments should be regarded as exemplary and non-limiting, the scope of the present application being defined by the appended claims rather than the above description, and all variations falling within the meaning and scope of the equivalent elements of the application file are intended to be included in the present application.

[0071] The above description is merely one specific implementation of the present application, enabling those skilled in the art to understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments shown herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for path planning of a UAV based on laser scanning, characterized in that, The method comprises the following steps: Step S1: set a single pre-scanning voyage, instruct the UAV to fly along the voyage, and use the laser scanner to scan the tunnel environment in real time to obtain original radial profile point cloud data; Step S2: separate free point clusters from the original radial profile point cloud data; analyze the structural stability index and dangerous rock potential energy threat value of the free point clusters to determine the risk attribution area of the current UAV flight path; Step S3: determine whether there is a high-risk detour area according to the risk attribution area, if there is, dynamically plan a risk detour path to adjust the single pre-scanning voyage, and generate an intelligent safe execution route; Step S4: instruct the UAV to fly along the intelligent safe execution route to its endpoint, and take the endpoint as the starting point of the next round of path planning until the iterative exploration of the target underground mine space is completed. 2.The laser scanning based UAV path planning method of claim 1, wherein, Step S1 comprises the following steps: Step S11: before the UAV enters the target underground mine space, use the wideband sonar detector carried by the UAV to perform fan-shaped acoustic detection on the front area, and obtain initial acoustic echo data containing main contour reflection information; Step S12: based on the farthest echo boundary identified in the initial acoustic echo data, calculate and set a safe front distance to obtain a single pre-scanning voyage; Step S13: instruct the UAV to fly along the central axis of the single pre-scanning voyage to the endpoint of the single pre-scanning voyage, and in the process, use the double-axis rotating laser scanner deployed orthogonally to the flight direction to perform real-time radial scanning on the two sides of the tunnel and the roof and floor of the target underground mine space, to obtain original radial profile point cloud data. 3.The laser scanning based UAV path planning method of claim 1, wherein, Before separating the free point clusters from the original radial profile point cloud data in step S2, the following steps are further included: perform three-dimensional spatial gridding processing on the original radial profile point cloud data to obtain three-dimensional scanning point cloud data; perform spatial voxel grid division on the three-dimensional scanning point cloud data, and then classify the laser pulse points in the three-dimensional scanning point cloud data into corresponding spatial voxel grids to obtain voxelized pulse data; based on the voxelized pulse data, calculate the Euclidean distance between the three-dimensional coordinate geometric center of all laser pulse points in each voxel and the corresponding voxel geometric center to obtain the position eccentricity; at the same time, calculate the standard deviation of the echo intensity values of all laser pulse points in each voxel in the voxelized pulse data to obtain the intensity fluctuation value; determine whether the position eccentricity is less than a preset geometric aggregation threshold value and whether the intensity fluctuation value is less than a preset material uniformity threshold value, if yes, mark the voxel as a high-confidence voxel; based on the high-confidence voxel, analyze the centroid points of the pulse points in the voxel, and use the set of all centroid points to form a structure confidence point cloud. 4.The laser scanning based UAV path planning method of claim 3, wherein, In step S2, separating the free point clusters from the original radial profile point cloud data comprises: perform region growing operation on the structure confidence point cloud to identify and extract the largest connected point cloud set, and determine the set as the main rock mass; perform point cloud Boolean difference set operation on the structure confidence point cloud and the main rock mass to separate all points not belonging to the main rock mass to obtain an initial overhanging body; Performing a Euclidean distance clustering analysis on the initial overhangs to segment the scattered point cloud on the initial overhangs into multiple independent point cloud clusters, to obtain a point cloud cluster set; Calculate the number of points in each cluster in the point cloud cluster set, and filter out the small clusters with a number of points less than a preset noise point number threshold, and the remaining cluster set is the set of free point clusters. 5.The laser scanning based UAV path planning method of claim 4, wherein, The structure stability index and the dangerous rock potential energy threat value of the free point cluster set in step S2 are analyzed to determine the risk attribution area of the current UAV flight path, including: Search for a point set in the free point cluster set and the surface points in the main rock mass whose spatial distance is less than a preset connection distance threshold, and define this point set as a connection point set; Based on the connection point set, analyze the hanging index and the fracture density value in the free point cluster set; According to the hanging index and the fracture density value, weighted calculation is performed to evaluate the structural stability of the connection area, and a structure stability index is obtained; Based on the preset high-risk screening threshold and the structure stability index, screen the high-risk free point cluster in the free point cluster set; Calculate the local curvature or surface variation degree of each point in the high-risk free point cluster, and identify the point set with a curvature or surface variation degree greater than a preset convexity threshold as a convex shape to segment out an isolated convex cluster; According to the isolated convex cluster, the dangerous rock potential energy threat value is evaluated; Based on the structure stability index and the dangerous rock potential energy threat value, the risk attribution area of the current UAV flight path is determined; wherein the risk attribution area includes a normal flight area and a high-risk bypass area. 6.The laser scanning based UAV path planning method of claim 5, wherein, According to the isolated convex cluster, the dangerous rock potential energy threat value is evaluated, including: Construct a minimum convex envelope of all outer points in the isolated convex cluster to obtain a dangerous rock convex body; Count the number of non-empty voxel units containing at least one point cloud data within the envelope range of the isolated convex cluster, and multiply the number by the volume of a single voxel unit to obtain a solid filling product; Compare the solid filling product with the total volume of the dangerous rock convex body, calculate the ratio of the degree of internal cavity or fracture development as the structure loosening degree; Based on the volume of the dangerous rock convex body and the structure loosening degree, and according to the height of the centroid point in the isolated convex cluster relative to the roadway floor, the dangerous rock potential energy threat value is determined. 7.The laser scanning based UAV path planning method of claim 5, wherein, Based on the connection point set, the hanging index and the fracture density value in the free point cluster set are analyzed, including: Based on the connection point set, a local fitting surface is constructed, and then the surface area of the local fitting surface is calculated as the connection area; Analyze the minimum circumscribed geometric body of each free point cluster in the free point cluster set, and calculate the overhanging volume; Divide the overhanging volume by the connection area to obtain the hanging index; Extract the neighborhood point cloud of the connection point set on one side of the main rock mass to obtain the connection area parent rock point cloud; Calculate the normal vector of each point in the connection area parent rock point cloud, and then count the dispersion degree of each vector direction in the normal vector and quantify it as a scalar value to obtain a fracture density value. 8.The laser scanning based UAV path planning method of claim 1, wherein, In step S3, whether there is a high-risk bypass area is determined according to the risk attribution area, and if there is, a risk bypass path is dynamically planned to adjust the single pre-scanning flight range, including: If the risk attribution area is a normal flight area, the UAV continues to perform the single pre-scanning flight; If the risk attribution area is a high-risk bypass area, a spatial influence radius is set according to the high-risk bypass area; A Gaussian potential field function with risk intensity decaying with distance is constructed based on the dangerous rock potential threat value and the structure stability index within the spatial influence radius centered on the centroid of the high-risk bypass area, and the high-risk bypass area is expanded outward by a safety distance to obtain flight risk potential field data; wherein the dangerous rock potential threat value is mapped to the peak intensity of the Gaussian potential field function, and the inverse of the structure stability index is taken as a parameter affecting its decay rate; A risk bypass path is dynamically planned according to the flight risk potential field data; The single pre-scanning flight range is dynamically adjusted and re-planned through the risk bypass path to generate an intelligent safe execution flight path. 9.The laser scanning based UAV path planning method of claim 8, wherein, The risk bypass path is dynamically planned according to the flight risk potential field data, including: The starting point and target point of the bypass path are determined in the flight risk potential field data based on the single pre-scanning flight range; A path cost function is set, which includes three components: path length, cumulative total value of risk potential field crossed, and flight energy consumption; A rapid expansion random tree search is performed on the flight risk potential field data using the path cost function to explore feasible paths connecting the starting point and the target point; When the convergence condition is met or the maximum number of iterations is reached, the current feasible path with the lowest cost is output as the risk bypass path.

10. A laser scanning based unmanned aerial vehicle path planning system, characterized in that, The laser scanning-based UAV path planning system for executing the method according to claim 1 comprises: A flight range scanning and perception module for setting a single pre-scanning flight range, instructing the UAV to fly along the flight range, and using a laser scanner to scan the tunnel environment in real time to obtain original radial profile point cloud data; A real-time risk assessment module for separating out a cluster of free points based on the original radial profile point cloud data; analyzing the structure stability index and the dangerous rock potential threat value of the cluster of free points to determine the risk attribution area of the current UAV flight path; An intelligent flight path generation module for determining whether there is a high-risk bypass area according to the risk attribution area, and if there is, dynamically planning a risk bypass path to adjust the single pre-scanning flight range and generate an intelligent safe execution flight path; A navigation control and iteration module for instructing the UAV to fly along the intelligent safe execution flight path to its endpoint, and taking the endpoint as the starting point of the next round of path planning until the iterative exploration of the target underground mine space is completed.

Images