An unmanned aerial vehicle airworthiness passage planning method based on linear scanning laser and AI

Abstract

This invention discloses a method for planning airworthiness channels for unmanned aerial vehicles (UAVs) based on linear scanning lasers and AI, relating to the field of UAV navigation technology. The method includes: acquiring a three-dimensional flight space region above the target road and setting a baseline; controlling the UAV to fly along the baseline, simultaneously acquiring the 3D coordinates and image data of obstacle vertices using an onboard linear scanning laser; intelligently extracting the 2D contours of obstacles using an AI model, and adaptively expanding the contours based on their type; mapping the contours to 3D space and rotating them around the perpendicular lines of the vertices to construct the 3D obstacle region; and removing all obstacle regions from the flight space using difference operations to obtain a precise UAV airworthiness channel. This invention integrates linear scanning and AI vision to achieve efficient and lightweight environmental perception and 3D modeling, overcoming the problems of high computational load and cost associated with traditional point cloud methods, and improving the real-time performance, safety, and accuracy of path planning.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) navigation and path planning technology, specifically to a UAV airworthiness channel planning method based on linear scanning laser and AI, for enabling safe and efficient flight of UAVs in complex environments such as above urban roads. Background Technology

[0002] With the rapid iteration of drone technology, its applications in logistics, municipal inspection, geographic surveying, and emergency rescue are becoming increasingly widespread. However, when drones fly at low altitudes, especially above urban roads, they face complex obstacle environments, such as trees (including branching structures), streetlights, utility poles, and billboards (including supporting structures) on both sides of the road. These obstacles directly threaten the flight safety of drones, thus necessitating precise and efficient airworthiness channel planning solutions.

[0003] Existing UAV path planning technologies mostly rely on lidar or visual sensors for environmental perception. The core of these technologies is to construct a 3D environment model by collecting dense point cloud data, and then plan the flight path. However, this type of method has the following significant drawbacks:

[0004] High workload and computational cost: The process of collecting dense point cloud data is time-consuming, and subsequent data processing relies on high-performance computing equipment, which makes it difficult to meet the needs of real-time planning for UAVs.

[0005] Poor equipment compatibility: High-precision lidar equipment is large and heavy, which not only increases the load of UAVs, affecting their maneuverability and endurance, but also has high equipment procurement costs, which is not conducive to large-scale application.

[0006] Insufficient accuracy in contour extraction: Traditional point cloud processing methods lack intelligent judgment in the extraction of obstacle contours, and easily overlook detailed structures such as tree branches and billboard support protrusions, resulting in safety hazards in the planned passage and failing to guarantee the flight safety of drones.

[0007] To address the shortcomings of the existing technologies, this invention proposes a lightweight, efficient, and high-precision method for planning airworthiness channels for unmanned aerial vehicles (UAVs). By combining linear scanning lasers with AI models, it solves the core problem of UAV airworthiness channel planning in complex environments. Summary of the Invention

[0008] To solve the above-mentioned technical problems, the present invention adopts the following solution:

[0009] A method for planning airworthiness corridors for unmanned aerial vehicles (UAVs) based on linear scanning lasers and AI includes the following steps:

[0010] S1. Obtain the three-dimensional flight space area of ​​the drone above the target road, wherein the three-dimensional flight space area of ​​the drone is the preset flight space range of the drone above the target road;

[0011] S2. Set a baseline within the three-dimensional space area of ​​the UAV flight and control the UAV to fly along the baseline;

[0012] S3. During the flight of the UAV along the baseline, the UAV collects the three-dimensional coordinate data of the vertices of obstacles and images containing obstacles in the three-dimensional space area of ​​the UAV's flight using linear scanning lasers installed on both sides of the fuselage;

[0013] Unlike traditional point cloud scanning technology, which requires the collection of massive amounts of dense point cloud data, this step only collects the 3D coordinate data of obstacle vertices and images containing obstacles, without the need for redundant point cloud information. This significantly reduces the amount of data collected and the storage pressure, laying a data foundation for lightweight UAV operations.

[0014] S4. Input the three-dimensional coordinate data of the obstacle vertex and the image containing the obstacle into the image information parsing AI model, and extract the two-dimensional contour of the obstacle by the image information parsing AI model;

[0015] This step extracts only the two-dimensional outline of the obstacle, rather than the complete three-dimensional point cloud data, further simplifying the data volume and reducing the computational load of subsequent data processing. Combined with the three-dimensional coordinate data of the vertices, it can meet the modeling requirements, achieving a dual lightweighting of data acquisition and processing.

[0016] S5. Based on the three-dimensional coordinate data of the obstacle vertex, the two-dimensional outline of the obstacle extracted in step S4 is mapped to the three-dimensional space region of the UAV flight described in step S1, and then the two-dimensional outline of the obstacle is rotated around the perpendicular line passing through its corresponding vertex to form a three-dimensional region of the obstacle.

[0017] This step cleverly utilizes the symmetrical features of common obstacles such as trees and streetlights in urban road environments, combining physical form with data processing logic. It quickly constructs the three-dimensional region of the obstacle through "two-dimensional contour + vertex vertical line rotation", eliminating the need for complex three-dimensional modeling algorithms. This simplifies the modeling process while ensuring the accuracy of the three-dimensional region, achieving efficient and precise obstacle modeling.

[0018] S6. Perform a difference operation between the three-dimensional flight space region of the UAV and the three-dimensional obstacle region to obtain the airworthiness channel of the UAV.

[0019] Preferably, the three-dimensional space region for the UAV flight is a cuboid structure; the baseline is a straight line connecting the center point of the front surface and the center point of the rear surface of the cuboid structure; the top surface of the cuboid structure is 8 to 15 meters away from the target road, the bottom surface of the cuboid structure is parallel to the target road and 3 to 5 meters away from the target road, the left and right sides of the cuboid structure are 2 to 4 meters away from the baseline, and the front and rear directions of the cuboid structure extend along the baseline.

[0020] Preferably, in step S3, when the linear scanning laser has a lightweight structure, the weight of the linear scanning laser is ≤ 5% of the total weight of the UAV, the scanning frequency of the linear scanning laser is ≥ 10Hz, and the scanning angle of the linear scanning laser covers ≥ 80% of the front contour of the obstacle, the program for acquiring the three-dimensional coordinate data of the obstacle vertex and the image containing the obstacle is started.

[0021] By combining the lightweight design of the linear scanning laser, only vertex coordinates and two-dimensional contour-related data are collected, avoiding the increased load problem caused by the large amount of data in traditional point cloud scanning equipment. This further enhances the maneuverability and endurance of the drone, achieving coordinated lightweighting of hardware adaptation and data optimization.

[0022] Preferably, in step S4, the image information parsing AI model further includes expanding the two-dimensional contour of the obstacle: expanding the extracted two-dimensional contour of the obstacle outward according to a preset ratio to generate an extended two-dimensional contour of the obstacle; in step S5, based on the three-dimensional coordinate data of the vertices of the extended two-dimensional contour of the obstacle, mapping the extended two-dimensional contour of the obstacle extracted in step S4 to the three-dimensional space region of the UAV flight described in step S1, and rotating the extended two-dimensional contour of the obstacle around the perpendicular line passing through its corresponding vertex to form an extended three-dimensional region of the obstacle; in step S6, the difference operation is: performing a difference operation between the three-dimensional space region of the UAV flight and the three-dimensional region of the obstacle and / or the extended three-dimensional region of the obstacle to obtain the airworthiness channel of the UAV.

[0023] For obstacles with regular or irregular contours, the ingenious modeling logic of "two-dimensional contour + vertex vertical line rotation" is continued. For the extended two-dimensional contour of irregular obstacles, the symmetry feature is also used to complete the construction of the three-dimensional region. This maintains the consistency of the modeling method and adapts to different types of obstacles through extended processing, thus taking into account both the versatility and safety of modeling.

[0024] Preferably, the rule for selecting whether to use a two-dimensional obstacle profile or an extended two-dimensional obstacle profile is as follows: when the obstacle is an object with a regular profile, the two-dimensional obstacle profile is used; when the obstacle is an object with an irregular profile, the extended two-dimensional obstacle profile is used.

[0025] Preferably, the method further includes the following steps:

[0026] S7. Collect real-time flight data of the UAV, including the UAV's current three-dimensional coordinates and flight trajectory data;

[0027] S8. Perform spatial matching between the real-time flight data and the UAV airworthiness channel obtained in step S6 to determine whether the UAV is within the airworthiness channel:

[0028] If the current three-dimensional coordinates or flight trajectory of the UAV fall within the spatial range of the UAV's airworthiness channel, the match is considered successful and the UAV is in the airworthiness zone.

[0029] If the current three-dimensional coordinates or flight trajectory of the drone exceed the spatial range of the drone's airworthiness channel, the matching is deemed to have failed and the drone has exceeded the airworthiness zone.

[0030] Preferably, when the matching fails and the drone is determined to be outside the airworthiness zone, the method further includes: generating an alarm signal and, based on the latest dynamic airworthiness channel data, planning the shortest recovery path for the drone to return to the airworthiness zone.

[0031] The beneficial effects of this invention are as follows:

[0032] 1. Lightweight and high efficiency: The linear scanning laser replaces the traditional dense point cloud scanning and combines it with AI models for intelligent contour extraction, which greatly reduces the burden of data acquisition and processing, and realizes lightweight and high-frequency environmental perception to meet the real-time operation needs of UAVs.

[0033] 2. High modeling accuracy: By fusing the two-dimensional contours extracted by AI with the three-dimensional coordinates of laser ranging, and rotating around the vertical line of the vertex to construct a three-dimensional obstacle area, high-precision modeling of complex obstacles (such as tree branches and billboard supports) is achieved, improving the safety of channel planning.

[0034] 3. Adaptive safety boundary: An obstacle contour expansion mechanism is introduced, which can dynamically generate a safety margin according to the obstacle type and automatically expand the obstacle avoidance area for irregular obstacles, effectively coping with environmental uncertainties.

[0035] 4. Real-time monitoring and autonomous correction: By matching the real-time flight data of the UAV with the planned channel, continuous monitoring of the airworthiness status is achieved, and alarms and automatic recovery path planning are triggered when the UAV exceeds the safe area, thus constructing a closed-loop safety protection system.

[0036] 5. High engineering practicality: Specific limitations were made on key hardware (laser weight, scanning frequency) and triggering conditions to ensure the feasibility and reliability of the technical solution on a real UAV platform, providing an implementable solution for urban low-altitude UAV applications. Attached Figure Description

[0037] Figure 1 This is an overall flowchart of the method of the present invention;

[0038] Figure 2 A schematic diagram showing the positional relationship between the three-dimensional space region of the drone's flight and the target road;

[0039] Figure 3 This is a two-dimensional outline diagram of a street lamp;

[0040] Figure 4 This is a three-dimensional schematic diagram of the streetlight area;

[0041] Figure 5 An extended two-dimensional outline diagram of a tree;

[0042] Figure 6 A schematic diagram of the expanded 3D region for trees;

[0043] Figure 7 This is a schematic diagram of the airworthiness corridor for unmanned aerial vehicles (UAVs).

[0044] Attached label: 1-Target road, 2-UAV flight 3D space area, 3-UAV, 4-Baseline, 5-Linear scanning laser, 6-Street lamp 2D outline, 7-Street lamp 3D area, 8-Tree 2D outline, 9-Tree extended 2D outline, 10-Tree extended 3D area, 11-UAV airworthiness channel. Detailed Implementation

[0045] The present invention will be further described in detail below with reference to the embodiments and accompanying drawings, but the embodiments of the present invention are not limited thereto. Example

[0046] Figure 1 This is an overall flowchart of the method of the present invention. Figure 1 As shown, a method for UAV airworthiness channel planning based on linear scanning laser and AI includes the following steps:

[0047] A method for planning airworthiness corridors for unmanned aerial vehicles (UAVs) based on linear scanning lasers and AI includes the following steps:

[0048] S1. Obtain the three-dimensional flight space region 2 of the drone above the target road 1. The three-dimensional flight space region 2 of the drone is the preset flight space range of the drone 3 above the target road 1.

[0049] S2. Set a baseline 4 within the three-dimensional flight space region 2 of the UAV, and control the UAV 3 to fly along the baseline 4;

[0050] S3. During the flight of UAV 3 along the baseline 4, the UAV 3 collects the three-dimensional coordinate data of the obstacle vertices and images containing the obstacles in the three-dimensional space region 2 of the UAV flight through the linear scanning lasers 5 installed on both sides of the fuselage;

[0051] In this embodiment, the linear scanning laser only collects core data and corresponding images such as the three-dimensional coordinates of street lamp vertices (e.g., the top of the lamp post, the edge of the lampshade), and key tree vertices (e.g., the top of the trunk, the branching vertices of the main branches). Compared with the hundreds of thousands or even millions of point cloud data collected by traditional point cloud scanning, the amount of data is reduced by more than 90%, and the storage occupation is greatly reduced, effectively alleviating the data storage burden of the UAV.

[0052] S4. Input the three-dimensional coordinate data of the obstacle vertex and the image containing the obstacle into the image information parsing AI model, and extract the two-dimensional contour of the obstacle by the image information parsing AI model;

[0053] The image information parsing AI model extracts only the columnar 2D outlines of streetlights and the approximate fan-shaped 2D outlines of trees from the collected images, eliminating the need to process redundant point cloud data. This improves data processing efficiency by more than 80%, and when combined with vertex 3D coordinate data, it can meet the subsequent modeling needs, achieving lightweight data processing.

[0054] S5. Based on the three-dimensional coordinate data of the obstacle vertex, the two-dimensional outline of the obstacle extracted in step S4 is mapped to the three-dimensional space region 2 of the UAV flight described in step S1, and then the two-dimensional outline of the obstacle is rotated around the perpendicular line passing through its corresponding vertex to form a three-dimensional region of the obstacle.

[0055] For regularly symmetrical obstacles like streetlights, their two-dimensional outline is rectangular. Rotating it 360° around the vertical line from the top of the pole precisely forms a cylindrical three-dimensional area for the streetlight. For approximately symmetrical obstacles like trees, their extended two-dimensional outline is fan-shaped. Rotating it 360° around the vertical line from the top of the trunk forms an approximately spherical three-dimensional area covering the entire extension of the branches. This design cleverly utilizes the physical symmetry of obstacles to quickly convert two-dimensional data into a three-dimensional model without the need for complex 3D reconstruction algorithms, significantly improving modeling efficiency and accuracy.

[0056] S6. Perform a difference operation between the three-dimensional space region 2 of the UAV flight and the three-dimensional region of the obstacle to obtain the UAV airworthiness channel 11.

[0057] Preferably, the three-dimensional space region 2 for UAV flight is a cuboid structure; the baseline 4 is a straight line connecting the center point of the front surface and the center point of the rear surface of the cuboid structure; the top surface of the cuboid structure is 8 to 15 meters away from the target road 1, the bottom surface of the cuboid structure is parallel to the target road 1 and is 3 to 5 meters away from the target road 1, the left and right sides of the cuboid structure are 2 to 4 meters away from the baseline 4 respectively, and the front and rear directions of the cuboid structure extend along the baseline 4.

[0058] Preferably, in step S3, when the linear scanning laser 5 has a lightweight structure, the weight of the linear scanning laser 5 is ≤ 5% of the total weight of the UAV 3, the scanning frequency of the linear scanning laser 5 is ≥ 10Hz, and the scanning angle of the linear scanning laser 5 covers ≥ 80% of the front contour of the obstacle, the obstacle vertex three-dimensional coordinate data and the image acquisition program containing the obstacle are started.

[0059] In this embodiment, the laser weighs only 0.3kg. Combined with the lightweight data acquisition mode of "vertex coordinates + two-dimensional contour", it avoids the problem of excessive drone load caused by the large amount of data and bulky equipment of traditional point cloud scanning equipment, ensuring the drone's flight maneuverability and endurance, and achieving dual lightweight adaptation of hardware and data.

[0060] Preferably, in step S4, the image information parsing AI model further includes the expansion processing of the two-dimensional contour of the obstacle: the extracted two-dimensional contour of the obstacle is expanded outward according to a preset ratio to generate an extended two-dimensional contour of the obstacle; in step S5, based on the three-dimensional coordinate data of the vertices of the extended two-dimensional contour of the obstacle, the extended two-dimensional contour of the obstacle extracted in step S4 is mapped to the three-dimensional space region 2 of the UAV flight described in step S1, and the extended two-dimensional contour of the obstacle is rotated around the perpendicular line passing through its corresponding vertex to form an extended three-dimensional region of the obstacle; in step S6, the difference operation is: the three-dimensional space region 2 of the UAV flight is performed with the three-dimensional region of the obstacle and / or the extended three-dimensional region of the obstacle to obtain the UAV airworthiness channel 11.

[0061] After the extended two-dimensional outline of the tree is rotated around the vertical line of the vertex, the resulting extended three-dimensional region can completely cover the irregular extensions such as branching branches. This not only simplifies the modeling by utilizing the approximate symmetry of the tree, but also compensates for the modeling errors of the irregular structure through the extension process, demonstrating a clever combination of physical form and data processing.

[0062] Preferably, the rule for selecting whether to use a two-dimensional obstacle profile or an extended two-dimensional obstacle profile is as follows: when the obstacle is an object with a regular profile, the two-dimensional obstacle profile is used; when the obstacle is an object with an irregular profile, the extended two-dimensional obstacle profile is used.

[0063] Preferably, the method further includes the following steps:

[0064] S7. Collect real-time flight data of UAV 3, including the current three-dimensional coordinates and flight trajectory data of UAV 3;

[0065] S8. Perform spatial matching between the real-time flight data and the UAV airworthiness channel 11 obtained in step S6 to determine whether the UAV 3 is within the airworthiness channel:

[0066] If the current three-dimensional coordinates or flight trajectory of UAV 3 fall within the spatial range of the UAV airworthiness channel 11, the match is determined to be successful and UAV 3 is in the airworthiness zone.

[0067] If the current three-dimensional coordinates or flight trajectory of UAV 3 exceed the spatial range of the UAV airworthiness channel 11, the matching is deemed to have failed and UAV 3 has exceeded the airworthiness zone.

[0068] Preferably, when the matching fails and the UAV 3 is determined to be outside the airworthiness zone, the method further includes: generating an alarm signal and, based on the latest dynamic airworthiness channel data, planning the shortest recovery path for the UAV 3 to return to the airworthiness zone.

[0069] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings.

[0070] Target scenario: Planning of a drone logistics delivery channel (1000 meters long) on ​​a section of a main urban road.

[0071] Step S1: Preset flight space

[0072] like Figure 2 As shown, the three-dimensional space region 2 for UAV flight is defined as a cuboid structure. A coordinate system is established with the center point of the starting end of the target road 1 as the origin. The top surface of the cuboid is set to be 10 meters away from the plane of the target road 1, the bottom surface of the cuboid is set to be 4 meters away from the plane of the target road 1, the left and right sides of the cuboid are each 3 meters away from the baseline 4, and the cuboid extends 1000 meters along the baseline in the front-back direction, forming an aerial flight corridor with a cross-section of 6 meters (width) × 6 meters (height).

[0073] Step S2: Fly along the baseline

[0074] Baseline 4 is set as a straight line connecting the center point of the front surface of the cuboid (coordinates: X=0, Y=0, Z=7) and the center point of the rear surface (coordinates: X=1000, Y=0, Z=7). The UAV 3 is controlled to fly at a constant speed of 5 m / s along the baseline 4.

[0075] Step S3: Environmental Data Acquisition

[0076] The linear scanning lasers mounted on both sides of the drone weigh 0.3 kg (5% of the drone's total weight of 6 kg) and have a scanning frequency of 15 Hz. When the laser scanning angle covers ≥80% of the frontal outline of obstacles (such as streetlights and trees) on both sides of the road, it simultaneously acquires the 3D coordinate data of the obstacle vertices and high-resolution images containing the obstacles.

[0077] This step only collects key coordinates and corresponding images of the top vertex of the street lamp and the branching vertex of the tree trunk. Compared with the traditional point cloud scanning that collects dense point cloud data of the entire street lamp and the entire tree, the amount of data is greatly reduced, the storage cost is reduced, and there is no need for high-performance equipment to process massive point clouds, which meets the lightweight requirements of drone real-time operations.

[0078] Step S4: AI Contour Extraction

[0079] The collected 3D coordinate data of obstacle vertices and high-resolution images containing obstacles are input into the image information analysis AI model.

[0080] like Figure 3 As shown, for streetlights with regular contours, the image information parsing AI model directly extracts the two-dimensional contour of the streetlight.

[0081] like Figure 5 As shown, for trees with irregular outlines and branching branches, the image information parsing AI model extracts the tree's two-dimensional outline 8 and then expands it outward by 8% to generate an extended two-dimensional outline 9, in order to reserve a safety margin.

[0082] This step extracts only the two-dimensional contour, avoiding the processing of redundant point cloud data in traditional point cloud scanning, significantly reducing the amount of data processing. Combined with the three-dimensional coordinates of the vertices, subsequent modeling can be completed, further enhancing the lightweight advantage.

[0083] Step S5: Construction of 3D Obstacle Region

[0084] The two-dimensional outline 6 of the street lamp and the extended two-dimensional outline 9 of the tree are respectively mapped to the three-dimensional space region 2 of the UAV flight described in step S1.

[0085] like Figure 4 As shown, the two-dimensional outline 6 of the street lamp is rotated 360° around the vertical line (parallel to the Z-axis) of its vertex to form the three-dimensional area 7 of the street lamp.

[0086] like Figure 6 As shown, the tree's extended two-dimensional outline 9 is also rotated 360° around its vertex perpendicular line to form the tree's extended three-dimensional region 10.

[0087] Streetlights, as regular symmetrical objects, can be precisely transformed into cylindrical three-dimensional regions by rotating their two-dimensional outlines. Trees, as approximately symmetrical objects, can cover the entire extension range of branches by rotating their extended two-dimensional outlines. This modeling method cleverly utilizes the physical symmetry of obstacles to quickly associate two-dimensional data with three-dimensional shapes. It can complete accurate modeling without complex algorithms, balancing efficiency and accuracy.

[0088] Step S6: Airworthiness Path Generation

[0089] like Figure 7 As shown, the three-dimensional space region 2 of the UAV flight is subjected to a difference operation with all the identified obstacle three-dimensional regions (including the street light three-dimensional region 7 and the tree extended three-dimensional region 10) to finally obtain a three-dimensional safe space without obstacles, namely the UAV airworthiness channel 11.

[0090] Steps S7-S8: Real-time monitoring and security decision making

[0091] During the drone's flight, its real-time flight data (such as current 3D coordinates: X=150, Y=1, Z=8) is continuously collected. This coordinate is then matched with the spatial range of the drone's airworthiness channel 11: if the coordinate is within the drone's airworthiness channel 11, the drone is determined to be in the airworthiness zone and continues normal flight.

[0092] If the drone's position deviates into an obstacle area due to factors such as airflow (e.g., coordinates change to X=100, Y=3, Z=8, landing in a streetlight area), it is determined to be outside the safe flight zone. The system immediately generates an audible and visual alarm signal and, based on the latest channel data, plans the shortest recovery path back to the safe flight zone, guiding the drone to automatically fly back to safe coordinates (e.g., X=100, Y=2, Z=8).

[0093] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Based on the technical essence of the present invention, any simple modifications, equivalent substitutions, and improvements made to the above embodiments within the spirit and principles of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A method for planning airworthiness corridors for unmanned aerial vehicles (UAVs) based on linear scanning laser and AI, characterized in that, Includes the following steps: S1. Obtain the three-dimensional space area (2) of the UAV flight above the target road (1), wherein the three-dimensional space area (2) of the UAV flight is the preset flight space range of the UAV (3) above the target road (1); S2. Set a baseline (4) within the three-dimensional flight space region (2) of the UAV, and control the UAV (3) to fly along the baseline (4); S3. During the flight of the UAV (3) along the baseline (4), the UAV (3) collects the three-dimensional coordinate data of the obstacle vertices and the images containing the obstacles in the three-dimensional space region (2) of the UAV flight by using the linear scanning lasers (5) installed on both sides of the fuselage; S4. Input the 3D coordinate data of the obstacle vertices and the image containing the obstacle into the image information parsing AI model, and have the image information parsing AI model extract the 2D outline of the obstacle; the image information parsing AI model also includes the expansion processing of the 2D outline of the obstacle: the extracted 2D outline of the obstacle is expanded outward according to a preset ratio to generate an extended 2D outline of the obstacle; the rule for selecting to use the 2D outline of the obstacle or the extended 2D outline of the obstacle is: when the obstacle is an object with a regular outline, the 2D outline of the obstacle is used; when the obstacle is an object with an irregular outline, the extended 2D outline of the obstacle is used. S5. Based on the three-dimensional coordinate data of the obstacle vertex, the two-dimensional outline of the obstacle extracted in step S4 is mapped to the three-dimensional space region (2) of the UAV flight described in step S1, and then the two-dimensional outline of the obstacle is rotated around the perpendicular line passing through its corresponding vertex to form the three-dimensional region of the obstacle. Based on the three-dimensional coordinate data of the vertices of the extended two-dimensional contour of the obstacle, the extended two-dimensional contour of the obstacle extracted in step S4 is mapped to the three-dimensional space region of the UAV flight described in step S1. The extended two-dimensional contour of the obstacle is rotated around the perpendicular line passing through its corresponding vertex to form an extended three-dimensional region of the obstacle. S6. Perform a difference operation between the three-dimensional space region (2) of the UAV flight and the three-dimensional region of the obstacle and / or the extended three-dimensional region of the obstacle to obtain the airworthiness channel (11) of the UAV.

2. The method for UAV airworthiness channel planning based on linear scanning laser and AI according to claim 1, characterized in that, The three-dimensional space region (2) of the UAV flight is a cuboid structure; the baseline (4) is a straight line connecting the center point of the front surface and the center point of the rear surface of the cuboid structure; the top surface of the cuboid structure is 8 to 15 meters away from the target road (1), the bottom surface of the cuboid structure is parallel to the target road (1) and is 3 to 5 meters away from the target road (1), the left and right sides of the cuboid structure are 2 to 4 meters away from the baseline (4) respectively, and the front and back directions of the cuboid structure extend along the baseline (4).

3. The method for UAV airworthiness channel planning based on linear scanning laser and AI according to claim 1, characterized in that, In step S3, the linear scanning laser (5) is a lightweight structure. The weight of the linear scanning laser (5) is less than 5% of the total weight of the UAV (3). The scanning frequency of the linear scanning laser (5) is greater than or equal to 10Hz. When the scanning angle of the linear scanning laser (5) covers ≥80% of the front contour of the obstacle, the three-dimensional coordinate data of the obstacle vertex and the image acquisition program containing the obstacle are started.

4. The method for UAV airworthiness channel planning based on linear scanning laser and AI according to claim 1, characterized in that, It also includes the following steps: S7. Collect real-time flight data of the UAV (3), wherein the real-time flight data includes the current three-dimensional coordinates and flight trajectory data of the UAV (3); S8. Perform spatial matching between the real-time flight data and the UAV airworthiness channel (11) obtained in step S6 to determine whether the UAV (3) is within the airworthiness channel: If the current three-dimensional coordinates or flight trajectory of the UAV (3) fall within the spatial range of the UAV airworthiness channel (11), the match is determined to be successful and the UAV (3) is in the airworthiness zone. If the current three-dimensional coordinates or flight trajectory of the UAV (3) exceed the spatial range of the UAV airworthiness channel (11), the matching is determined to be unsuccessful and the UAV (3) is out of the airworthiness area.

5. The method for UAV airworthiness channel planning based on linear scanning laser and AI according to claim 4, characterized in that, When the matching fails and the drone (3) is determined to be outside the airworthiness zone, the method further includes: generating an alarm signal and planning the shortest recovery path for the drone (3) to return to the airworthiness zone based on the latest dynamic airworthiness channel data.

Images