Aircraft inspection dynamic boundary generation and sharing method and system based on master-slave cooperation

Abstract

The invention discloses an aircraft inspection dynamic boundary generation and sharing method and system based on master-slave cooperation, and relates to the technical field of unmanned vehicle cooperative inspection and aviation ground safety detection.The method comprises the steps that an unmanned vehicle is selected from a vehicle team to serve as a boundary generation vehicle, and other vehicles serve as boundary consumption vehicles; the boundary generation vehicle runs around the aircraft along a preset path, point cloud data are collected, filtering processing and target extraction are carried out on the point cloud data, and aircraft point cloud is obtained; projecting the aircraft point cloud to a horizontal plane, calculating a convex hull boundary by adopting a Graham scanning algorithm, and performing safety expansion processing on the convex hull boundary to generate a safety boundary; and packaging the security boundary into a boundary data packet, and broadcasting the boundary data packet to the boundary consumer vehicle. According to the invention, the overall operation efficiency and economical efficiency of multi-vehicle cooperative inspection are improved.

Description

Technical Field

[0001] This invention relates to the field of unmanned vehicle collaborative inspection and aviation ground safety inspection technology, and in particular to a method and system for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration. Background Technology

[0002] With the increasing demand for operation and maintenance management of large civil aircraft and airport hangar areas, automated inspection technology based on unmanned platforms (such as unmanned vehicles and drones) has gradually become an important part of aircraft safety inspection. Traditional inspections usually rely on manual or unmanned equipment with fixed flight paths for close-range observation. Their inspection boundaries are mostly statically preset and cannot be adjusted in a timely manner according to dynamic changes in aircraft attitude, parking position, or apron environment. In recent years, although technologies such as multi-sensor fusion positioning, laser point cloud reconstruction, and autonomous navigation have been applied in relevant scenarios, they still have significant shortcomings in aircraft inspection scenarios. For example, inspection boundaries are mostly estimated independently by a single machine, lacking a master-slave collaborative generation mechanism, resulting in lagging boundary information updates; boundary models often use coarse-grained bounding boxes or fixed-radius buffers, which cannot take into account both fine geometric contours and safety distance requirements; and the lack of a highly consistent boundary sharing mechanism among inspection equipment makes multi-vehicle collaborative inspections prone to planning conflicts, duplicate coverage, or safety distance imbalances due to inconsistent boundary interpretations. Therefore, there is an urgent need to develop an inspection boundary generation method that combines high-precision point cloud processing, dynamic boundary modeling, and inter-vehicle sharing mechanisms, so as to enable multi-vehicle collaboration in completing high-safety-level aircraft inspection operations.

[0003] CN120602923A discloses a method and apparatus for sharing urban inspection data based on unmanned aerial vehicles (UAVs). This method samples multi-dimensional UAV inspection data, constructs feature vectors, generates feature values ​​using a hash algorithm, and then encrypts the inspection data before sharing, thus solving the problem of low security in UAV inspection data sharing. While this method has certain advantages in data encapsulation, secure transmission, and key management, its focus is on the secure sharing of inspection data, without addressing the boundary generation logic for inspection targets. Furthermore, it lacks mechanisms for constructing target contours using point clouds, dynamically expanding safe zones, or a master-slave boundary distribution mechanism for vehicle-to-vehicle collaboration. Therefore, it cannot be applied to the high-precision boundary model generation and sharing requirements of aircraft inspection tasks.

[0004] CN118640894B proposes a method and system for automatically generating lightweight environmental maps for robotic building inspections. This method extracts geometric and non-geometric information from building BIM models, utilizes the Alpha Shapes algorithm to extract internal and external boundaries, and further optimizes the accuracy of the environmental map construction. While this method possesses high-quality boundary extraction capabilities in built environment scenarios, it relies on static building BIM models and is not suitable for real-world airport environments with dynamically changing aircraft parking conditions, complex structures, and constantly changing unobstructed environments. Furthermore, this method only focuses on reconstructing environmental element boundaries and does not address multi-device collaboration, real-time boundary updates, or cross-vehicle sharing mechanisms, thus failing to meet the real-time boundary generation and distribution requirements based on point clouds in aircraft inspection scenarios. Summary of the Invention

[0005] The purpose of this section is to outline some aspects of the embodiments of the present invention and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section, as well as in the abstract and title of the present application, to avoid obscuring the purpose of this section, the abstract and title of the invention. Such simplifications or omissions shall not be used to limit the scope of the present invention.

[0006] Given that existing technologies still suffer from problems such as boundary generation relying on static models, lack of master-slave collaborative mechanisms, insufficient real-time boundary sharing, and low accuracy in constructing security boundaries in aircraft inspection scenarios, this invention is proposed.

[0007] Therefore, the problem to be solved by this invention is how to realize the dynamic boundary generation and sharing of aircraft inspection based on master-slave collaboration.

[0008] To solve the above-mentioned technical problems, the present invention provides the following technical solution: In a first aspect, embodiments of the present invention provide a method for generating and sharing dynamic boundaries for aircraft inspections based on master-slave collaboration, which includes selecting one unmanned vehicle from the fleet as the boundary generation vehicle and the remaining vehicles as boundary consumption vehicles. The boundary generation vehicle travels around the aircraft along a predetermined path and collects point cloud data. The point cloud data is then filtered and the target is extracted to obtain the aircraft point cloud. The aircraft point cloud is projected onto a horizontal plane, the Graham scan algorithm is used to calculate the convex hull boundary, and the convex hull boundary is subjected to safe expansion processing to generate a safe boundary. The security boundary is encapsulated into a boundary data packet and broadcast to the boundary consumption vehicle, wherein the boundary consumption vehicle integrates the security boundary into the local cost map for real-time path planning.

[0009] As a preferred embodiment of the master-slave collaborative aircraft inspection dynamic boundary generation and sharing method described in this invention, the boundary data packet includes the vertex coordinate sequence of the security boundary S, the current timestamp, the confidence level C, and the boundary generation vehicle identifier; the specific formula for the confidence level C is as follows: ; in, For the effective number of point clouds, The preset upper limit threshold for the number of valid point clouds. The weighting coefficients for the point cloud quantity factor. The weighting coefficients for the time decay factor. These are the weighting coefficients for the distance factor. The duration since the last structural update of the security boundary. The time decay coefficient, The average distance of the scan. The maximum distance threshold for effective scanning of aircraft by lidar.

[0010] As a preferred embodiment of the master-slave collaborative aircraft inspection dynamic boundary generation and sharing method described in this invention, it further includes: The boundary consumption vehicle adjusts its safety strategy based on confidence level C; When the confidence level C is higher than the confidence level threshold, the safety boundary is directly applied. When the confidence level C is lower than the confidence level threshold, the safety boundary is expanded further before use.

[0011] As a preferred embodiment of the master-slave collaborative aircraft inspection dynamic boundary generation and sharing method described in this invention, the method for generating the safety boundary S is as follows: Set the vehicle safety radius as δ, and perform Minkowski sum operation on the convex hull boundary H and the disk B with radius δ centered at the origin to obtain the safety boundary S; Traverse all vertices of the convex hull boundary H Calculate vertices interior angle of ; Calculate the vertex interior angle of Unit vector in the direction of the bisector Along the unit vector The direction will be the vertex Offset the vertex outward by a distance to obtain the new, expanded vertex. ; Connect all the new vertices after expansion. This forms a safe boundary after smooth expansion.

[0012] As a preferred embodiment of the master-slave collaborative aircraft inspection dynamic boundary generation and sharing method described in this invention, wherein: the expanded new vertex The specific formula is as follows: ; in, As vertex interior angle of The unit vector in the direction of the angle bisector. For all vertices of the convex hull boundary H, For the vehicle's safe radius, As vertex inside angle.

[0013] As a preferred embodiment of the master-slave collaborative aircraft inspection dynamic boundary generation and sharing method described in this invention, the method for calculating the convex hull boundary is as follows: The three-dimensional coordinates of the aircraft point cloud are vertically projected onto the horizontal plane to obtain a two-dimensional projection point set. Traverse the set of two-dimensional projection points and find the point with the smallest y-coordinate as the base point. and press the remaining points relative to the base point. The points are sorted by their polar angles to obtain a point sequence. Initialize the stack structure and set the base point Push the first point in the aforementioned point sequence onto the stack in sequence; Traverse the remaining points in the sorted point sequence, for the current point Get the top of the stack and the apex of the stack The cross product of vectors is calculated using the following formula: ; When the cross product of vectors is less than or equal to zero, then the stack vertex If the point is concave, pop it from the stack; Repeat this process until the cross product of the vectors is greater than zero or only one point remains in the stack, then move the current point... Push onto the stack; After traversing all points, the points stored in the stack are the set of vertices of the convex hull boundary H. These vertices are then connected in polar angle order to form the convex hull boundary H.

[0014] As a preferred embodiment of the aircraft inspection dynamic boundary generation and sharing method based on master-slave collaboration described in this invention, the method for acquiring the aircraft point cloud is as follows: Based on point cloud data, a statistical filtering algorithm is used to remove noise points; The RANSAC algorithm is used to fit the ground plane to the filtered point cloud data, extract the plane equation parameters, mark the point cloud data with a distance less than the ground threshold from the plane equation as ground points and remove them to obtain non-ground point clouds; Euclidean clustering is performed on the non-terrestrial point cloud to calculate the point cloud data. With point cloud data The Euclidean distance between them; When the Euclidean distance is less than a preset distance threshold, the point cloud data will be... With point cloud data Points are grouped into the same point cluster, the number of points in each cluster is counted, and the cluster with the largest number of points is selected as the aircraft point cloud.

[0015] Secondly, embodiments of the present invention provide a dynamic boundary generation and sharing system for aircraft inspection based on master-slave collaboration, comprising: a boundary generation vehicle and a boundary consumption vehicle; the boundary generation vehicle includes a computing processing terminal, a navigation module, a communication module, and a power supply module; The computing terminal includes an industrial control module and an edge computing module; the navigation module includes a lidar, a vision module, and an RTK module; and the communication module includes a 4G module, a 5G module, a WiFi module, and a Bluetooth module.

[0016] Thirdly, embodiments of the present invention provide a computer device, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement any step of the above-described method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration.

[0017] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program thereon, wherein: when the computer program is executed by a processor, it implements any step of the above-described method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration.

[0018] Compared with existing technologies, the advantages of this invention are as follows: By establishing a master-slave collaborative architecture, one unmanned vehicle in the fleet is designated as the boundary generation vehicle, and the remaining vehicles are designated as boundary consumption vehicles, thus realizing a collaborative operation mode with a clear division of labor. This avoids the waste of computing power caused by repeated scanning by multiple vehicles and improves the overall fleet operation efficiency. The boundary generation vehicle collects point cloud data and performs filtering and target extraction while driving around the aircraft. Noise points are removed using statistical filtering algorithms, ground points are filtered out using the RANSAC algorithm, and aircraft point clouds are extracted using Euclidean clustering, ensuring the accuracy and reliability of the data. Accurate extraction of the aircraft contour is achieved by projecting the aircraft point cloud onto a horizontal plane and using the Graham scan algorithm to calculate the convex hull boundary. This algorithm has a time complexity of O(nlogn) and high computational efficiency. The Minkowski principle is combined to refine the convex hull edge... The boundary undergoes safety expansion processing, adaptively offsetting according to the direction of the angle bisector of the vertex's interior angle. The generated safety boundary ensures safe vehicle distances while avoiding excessive conservatism that could reduce the drivable area, achieving a balance between safety and traffic efficiency. By encapsulating the safety boundary into a data packet containing vertex coordinate sequences, timestamps, confidence levels, and vehicle identifiers, and broadcasting it to the boundary consumer vehicle based on the DDS protocol, the boundary consumer vehicle can dynamically adjust its safety strategy based on the confidence level. When the confidence level is high, the boundary is directly used; when the confidence level is low, additional expansion is performed. This achieves real-time and reliable data sharing, allowing the boundary consumer vehicle to integrate the shared boundary into its local cost map for path planning without performing complex boundary calculations itself. This significantly reduces the sensor configuration requirements and computational burden of a single vehicle, improving the overall operational efficiency and economy of multi-vehicle collaborative inspections. Attached Figure Description

[0019] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. Wherein: Figure 1 The flowchart shows a method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration. Figure 2 A three-dimensional schematic diagram of boundary generation for a master-slave collaborative method for dynamic boundary generation and sharing in aircraft inspection. Figure 3 This is a planar schematic diagram of boundary generation based on a master-slave collaborative method for dynamic boundary generation and sharing in aircraft inspection. Figure 4 This is a schematic diagram illustrating the dynamic boundary generation and sharing method for aircraft inspection based on master-slave collaboration. Figure 5This is a system architecture diagram of a master-slave collaborative aircraft inspection dynamic boundary generation and sharing system. Detailed Implementation

[0020] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments.

[0021] Based on the embodiments of this invention, all other embodiments obtained by those skilled in the art without inventive effort should fall within the scope of protection of this invention.

[0022] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0023] As mentioned in the background section, existing aircraft inspection processes still commonly suffer from problems such as reliance on static models for inspection boundary generation, insufficient boundary accuracy, lagging real-time updates, and a lack of consistent boundary sharing in multi-vehicle collaboration. To address these issues, this invention provides a method for dynamic boundary generation and sharing in aircraft inspection based on master-slave collaboration.

[0024] Example 1 Reference Figures 1-4 , Figure 1 This is a flowchart illustrating a method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration, according to an embodiment of the present invention. Figure 1 As shown, a method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration includes: S1: Select one autonomous vehicle from the fleet as the boundary generation vehicle, and the remaining vehicles as boundary consumption vehicles; It should be noted that the boundary generation vehicle is equipped with a lidar and a GNSS / IMU integrated navigation module.

[0025] S2: The boundary generation vehicle travels around the aircraft along a predetermined path and collects point cloud data. The point cloud data is then filtered and the target is extracted to obtain the aircraft point cloud. Specifically, such as Figure 2 As shown, point cloud data in the radar coordinate system The boundary generation vehicle establishes a global coordinate system, using latitude and longitude provided by GNSS and attitude angles provided by IMU to generate point cloud data in the lidar coordinate system. By rotation matrix Translation vector Convert to point cloud data in global coordinate system The conversion formula is: ; in, These are the three-dimensional coordinate components of the point cloud in the lidar coordinate system. This is the translation vector used to determine the real-time position of the vehicle based on the boundary.

[0026] It should be noted that the rotation matrix The translation vector is calculated from the pitch, roll, and yaw angles provided by the IMU. The three-dimensional position coordinates provided by GNSS are used for determination; both the boundary generation vehicle and the boundary consumption vehicle use a global coordinate system for data processing and communication to ensure that all point cloud data and boundary data are expressed in a unified coordinate system.

[0027] Furthermore, based on point cloud data, a statistical filtering algorithm is used to remove noise points; the number of neighboring points of each point is calculated, and when the number of neighboring points is less than a set threshold, this point is marked as a noise point and removed.

[0028] Furthermore, the RANSAC algorithm is used to fit the ground plane to the filtered point cloud data, extracting the plane equation parameters. Point cloud data with a distance less than the ground threshold from the plane equation are marked as ground points and then removed to obtain non-ground point clouds. Euclidean clustering is then performed on the non-ground point clouds to calculate the point cloud data. With point cloud data The Euclidean distance between them is given by the following formula: ; in, For point cloud data With point cloud data The Euclidean distance between them For point cloud data The three-dimensional coordinates For point cloud data The three-dimensional coordinates.

[0029] Preferably, when the Euclidean distance is less than a preset distance threshold, the point cloud data is... With point cloud data Points are grouped into clusters of the same type of point, and the number of points in each cluster is counted. The cluster with the largest number of points is selected as the aircraft point cloud. The specific formula is as follows: ; in, Let k be the k-th point cluster.

[0030] It should be noted that the threshold is set based on point cloud density and aircraft structural characteristics to ensure that the main body of the aircraft can be effectively separated from other auxiliary objects; the preset distance threshold ranges from 0.5 meters to 2 meters.

[0031] S3: Project the aircraft point cloud onto the horizontal plane, use the Graham scan algorithm to calculate the convex hull boundary, and perform safe expansion processing on the convex hull boundary to generate a safe boundary; Specifically, such as Figure 3 As shown, the three-dimensional coordinates of the aircraft point cloud are vertically projected onto the horizontal plane to obtain a two-dimensional projection point set; the two-dimensional projection point set is then traversed to find the point with the smallest y-coordinate as the base point. and the remaining points relative to the base point Sort the points by their polar angles to obtain a point sequence; initialize the stack structure and set the base points... Push the first point in the sorted point sequence onto the stack; iterate through the remaining points in the sorted point sequence, and for the current point... Get the top of the stack and the apex of the stack The cross product of vectors is calculated using the following formula: ; Furthermore, when the cross product of vectors is less than or equal to zero, the stack vertex... If the point is concave, pop it from the stack; repeat this process until the cross product of the vectors is greater than zero or only one point remains in the stack, then pop the current point. Push the points onto the stack; after traversing all points, the points stored in the stack are the set of vertices of the convex hull boundary H, and these vertices are connected in polar angle order to form the convex hull boundary H.

[0032] It should be noted that if multiple minimum y-coordinates exist, the point with the smallest x-coordinate is selected as the base point. .

[0033] Furthermore, setting the vehicle's safety radius as δ, we perform a Minkowski sum operation on the convex hull boundary H and a disk B centered at the origin with a radius of δ to obtain the safety boundary S. The specific formula is as follows: ; in, For safety boundaries, Let H be the vehicle's safety radius, and H be the convex hull boundary. A circular structural element with radius δ. Let H be any point on the convex hull boundary H. Let be any vector within the circular structuring element B.

[0034] It should be noted that the vehicle's safe radius δ is determined based on the vehicle's dimensions and safety margin, and its value ranges from 1 meter to 5 meters; interior angle Through vertex Its two adjacent vertices and The included angle between the two sides formed is calculated.

[0035] Specifically, traverse all vertices of the convex hull boundary H. Calculate vertices interior angle of ; Calculate vertices interior angle of Unit vector in the direction of the bisector Along the unit vector Direction will be the vertex Offset the vertex outward by a distance to obtain the new, expanded vertex. The specific formula is as follows: ; in, As vertex interior angle of The unit vector in the direction of the angle bisector. For all vertices of the convex hull boundary H, For the vehicle's safe radius, As vertex inside angle.

[0036] Furthermore, connect all the newly expanded vertices. This forms a safe boundary after smooth expansion.

[0037] Preferably, when the interior angle is close to 180 degrees, the offset distance approaches the vehicle's safe radius δ; when the interior angle is 90 degrees, the safe boundary S forms a smooth outward expansion boundary curve.

[0038] S4: Encapsulate the security boundary into a boundary data packet and broadcast it to the boundary consumer vehicle, which integrates the security boundary into the local cost map for real-time path planning.

[0039] It should be noted that the boundary data packet includes the vertex coordinate sequence of the security boundary S, the current timestamp t, the confidence level C, and the boundary generation vehicle identifier; Specifically, the formula for the confidence level C is as follows: ; in, For the effective number of point clouds, This is a preset upper limit threshold for the number of valid point clouds, with a value range of [5000, 20000]. The weighting coefficients for the point cloud quantity factor satisfy 0 < <1, The weighting coefficients for the time decay factor satisfy 0 < <1, The weighting coefficients of the distance factor satisfy 0 < <1, The duration since the last structural update of the security boundary. This is the time decay coefficient, with a value range of [0.01, 0.5], used to control the rate at which the confidence level decays over time. The average distance of the scan. The maximum distance threshold for effective scanning of aircraft by lidar is [30, 100] meters.

[0040] It should be noted that structural updates refer to updates where the number or sequence of vertices of the safe boundary polygon changes.

[0041] Furthermore, the safety strategy for the boundary consumption vehicle is adjusted based on the confidence level C; when the confidence level C is higher than the confidence level threshold, the safety boundary is used directly; when the confidence level C is lower than the confidence level threshold, the safety boundary is expanded further before use.

[0042] Furthermore, such as Figure 4 As shown, the boundary generation vehicle broadcasts boundary data packets to the boundary consumption vehicle via the DDS data distribution service protocol. After receiving the boundary data packets, the boundary consumption vehicle extracts the current timestamp, calculates the time difference between the current timestamp and the local current time, and determines that the boundary data packet is valid when the time difference is less than the preset time threshold. The vehicle then transforms the vertex coordinates of the security boundary S to the local cost map coordinate system, marks the area enclosed by the security boundary S as a restricted area in the local cost map, and performs path planning based on the updated local cost map to generate a safe path by avoiding the restricted area.

[0043] It should be noted that the DDS data distribution service protocol supports publish-subscribe communication mode; after the border consumer vehicle subscribes to the border data topic, it automatically receives border data packets; the preset time threshold ranges from 0.5 seconds to 3 seconds; the value of the restricted area is set to 255, indicating that it is not allowed to pass.

[0044] Specifically, the boundary generation vehicle periodically updates the security boundary S and broadcasts new boundary data packets. The update period ranges from 0.1 seconds to 1 second. If the boundary consumption vehicle does not receive a valid boundary data packet within 3 consecutive update periods, the boundary generation vehicle is deemed to have failed, triggering a re-election mechanism to select a new boundary generation vehicle from the boundary consumption vehicles.

[0045] In summary, this invention establishes a master-slave collaborative architecture, designating one unmanned vehicle in the fleet as the boundary generation vehicle and the remaining vehicles as boundary consumption vehicles. This achieves a clearly defined collaborative operation mode, avoiding the waste of computing power caused by multiple vehicles repeatedly scanning and improving the overall fleet operation efficiency. The boundary generation vehicle collects point cloud data while circling the aircraft, performs filtering and target extraction, and uses statistical filtering algorithms to remove noise points, RANSAC algorithms to filter ground points, and Euclidean clustering to extract aircraft point clouds, ensuring data accuracy and reliability. By projecting the aircraft point cloud onto a horizontal plane and using the Graham scan algorithm to calculate the convex hull boundary, accurate extraction of the aircraft contour is achieved. This algorithm has a time complexity of O(nlogn) and high computational efficiency. Furthermore, the Minkowski principle is used to securely define the convex hull boundary. The expansion process adaptively offsets the boundary based on the angle bisector of the vertex's interior angle. This generates a safe boundary that ensures safe vehicle distances while avoiding excessive conservatism that could reduce the drivable area, achieving a balance between safety and traffic efficiency. By encapsulating the safe boundary into a data packet containing vertex coordinate sequences, timestamps, confidence levels, and vehicle identifiers, and broadcasting it to the boundary consumer vehicle based on the DDS protocol, the boundary consumer vehicle can dynamically adjust its safety strategy based on the confidence level. When the confidence level is high, the boundary is directly used; when the confidence level is low, additional expansion is performed. This achieves real-time and reliable data sharing, allowing the boundary consumer vehicle to integrate the shared boundary into its local cost map for path planning without performing complex boundary calculations. This significantly reduces the sensor configuration requirements and computational burden for a single vehicle, improving the overall operational efficiency and economy of multi-vehicle collaborative inspections.

[0046] Example 2 Based on the teachings of the above embodiments, other aspects of the present invention also propose a dynamic boundary generation and sharing system for aircraft inspection based on master-slave collaboration, including a boundary generation vehicle and a boundary consumption vehicle.

[0047] Specifically, such as Figure 5 As shown, the boundary generation vehicle includes a computing processing terminal, a navigation module, a communication module, and a power supply module.

[0048] Furthermore, the computing processing terminal includes an industrial control module and an edge computing module. The industrial control module is responsible for scheduling, task distribution, and data processing of each module. The edge computing module analyzes and inspects image recognition faults in real time and analyzes sensor data to provide beyond-line-of-sight perception information for autonomous vehicles. The navigation module includes a lidar, a vision module, and an RTK module. The lidar generates a 3D point cloud containing depth information, accurately measures the contour, size, and distance of objects, and collects point cloud data. The vision module perceives objects, spatial structures, light, and color in the environment through image data and generates a map or model of the environment in the computer, performing obstacle avoidance and visual inspection. The RTK module achieves centimeter-level or millimeter-level high-precision positioning and works with the lidar to convert the point cloud to a unified global coordinate system in real time. Furthermore, the communication module includes a 4G module, a 5G module, a WiFi module, and a Bluetooth module. The 4G and 5G modules enable network communication, establish a link with the backend system, and broadcast border data packets to the border consumption vehicle. The WiFi module enables access to WiFi hotspots, builds a local area network to achieve short-range high-speed communication, and can also serve as a mobile hotspot base station. The Bluetooth module enables interconnection with other devices, allowing portable devices to access and be controlled. Preferably, the communication module broadcasts boundary data packets using a DDS-based communication network protocol; the boundary consumer vehicle verifies the timeliness of the boundary data packets through the communication module and integrates the security boundary S as a restricted area into the local cost map.

[0049] Example 3 This embodiment provides a calibrable multi-step histogram peak reading circuit and method for use in lidar sensors. To verify the beneficial effects of the present invention, scientific demonstration is carried out through economic benefit calculations and simulation experiments.

[0050] Specifically, the experiment was conducted in the actual apron environment of Nanjing Lukou International Airport. Two AGV-M200 autonomous inspection vehicles were used, both equipped with Velodyne VLP-32C 32-line lidar with a scanning frequency of 10Hz, a horizontal field of view of 360 degrees, a vertical field of view of +15° to -25°, a ranging range of 0.3 meters to 200 meters, a ranging accuracy of ±3 centimeters, and an angular resolution of 0.2 degrees. The navigation and positioning system used Qianxun Location's RTK-GNSS module, with a positioning accuracy of ±2 centimeters in the horizontal direction and ±5 centimeters in the vertical direction, and a data update rate of 10Hz. The in-vehicle computing platform is equipped with an Intel Core i7-11800H processor with a main frequency of 2.3GHz, 8 cores and 16 threads, 16GB of DDR4 memory and a 512GB solid-state drive, running the Ubuntu 20.04 operating system and ROS Noetic middleware; the communication module adopts Huawei 5G industrial module, operating on the n78 (3.5GHz) frequency band, with an uplink peak rate of 180Mbps, a downlink peak rate of 1.2Gbps, and an average end-to-end latency of 18 milliseconds.

[0051] Furthermore, during the experiment, AGV-01 was initially designated as the boundary generation vehicle, and AGV-02 as the boundary consumption vehicle. AGV-01 started from the southeast corner of the parking position and traveled clockwise around the A320 aircraft along a preset trajectory at an average speed of 0.8 m / s. The preset trajectory maintained a safe distance of 12 to 15 meters from the outer contour of the aircraft fuselage, with a total trajectory length of approximately 185 meters. During the trajectory, the lidar continuously scanned and collected point cloud data, with each frame containing approximately 118,000 to 132,000 three-dimensional coordinate points. The point cloud data was first subjected to statistical filtering, with a neighbor threshold of 8 and a distance threshold of 0.15 meters to remove outlier noise points. After filtering, each frame of point cloud data retained approximately [missing information]. The ground plane was fitted with 95,000 to 108,000 points. The RANSAC algorithm was used for ground plane fitting, with an interior point distance threshold of 0.08 meters and a maximum of 500 iterations. After obtaining the ground plane equation, ground point clouds within 0.08 meters of the plane were removed, resulting in approximately 38,000 to 52,000 non-ground point clouds. Euclidean clustering analysis was performed on the non-ground point clouds, with an inter-cluster distance threshold of 0.8 meters and a minimum cluster size of 200 points. The largest extracted point cluster was the main aircraft point cloud. The effective number of aircraft point clouds, N, fluctuated between 8,200 and 14,800 points, with an average of 11,350 points. After projecting the aircraft point clouds onto the horizontal plane, the Graham scan algorithm was used to calculate the convex hull boundary. The algorithm first finds the base point with the smallest y-coordinate, sorts the remaining points by polar angle, and uses a stack structure to check the cross product of vectors point by point to complete the convex hull calculation. The extracted convex hull boundary H contains 12 to 16 vertices.

[0052] Furthermore, a safety expansion process is applied to the convex hull boundary H. The vehicle safety radius δ is set to 1.2 meters. All vertices of the convex hull are traversed, and the interior angles and angle bisectors of each vertex are calculated. An adaptive offset is then performed to generate the safety boundary S. The confidence level C is calculated based on the number of effective point clouds, the time decay factor, and the average scanning distance. The weighting coefficients are set to 0.35, 0.40, and 0.25, respectively. The time decay factor is set to 0.08. The upper limit of the number of effective point clouds is set to 15,000 points, and the maximum scanning distance D_max is set to 80 meters. During the entire detour, the confidence level C is maintained between 0.72 and 0.89. The boundary generation vehicle encapsulates the vertex coordinate sequence of the safety boundary S, the current timestamp, the confidence level C, and the vehicle identifier into a boundary data packet. The data packet size is approximately 800 to 1200 bytes, and it is broadcast to the boundary consumer vehicle AGV-02 via the DDS protocol at a period of 100 milliseconds. After receiving the boundary data packet, the boundary consumption vehicle extracts the timestamp and compares it with the local time. If the time difference is less than 50 milliseconds, the data is deemed valid. The safety boundary S is then transformed to the local cost map coordinate system. The area enclosed by boundary S is marked as a restricted zone in an 80m × 80m grid map with a resolution of 0.1m, and its cost is set to 255. Based on the updated cost map, the boundary consumption vehicle performs real-time path planning, using an improved A* algorithm to generate the optimal path from its current location to the inspection target point, maintaining a distance of more than 1.5 meters from the restricted zone boundary throughout the process to ensure inspection safety. To verify the advantages of this invention, a comparative experiment was conducted simultaneously. Two unmanned vehicles independently performed point cloud acquisition, boundary calculation, and path planning in the same scenario, recording various performance indicators. The entire experiment consisted of three sets of repeated tests, each lasting approximately 15 minutes. Key performance parameters recorded included single-vehicle computation latency, boundary consistency error, total system power consumption, communication bandwidth usage, average CPU utilization, peak memory usage, boundary update frequency, boundary generation delay, path planning success rate, and system response time.

[0053] Specifically, as shown in Table 1, from the perspective of boundary consistency error index, in the existing technical solution, the two unmanned vehicles independently perform boundary calculations. Due to the slight differences in point cloud acquisition perspective, filtering parameters, and the randomness of algorithm calculation, there is a significant deviation in the boundary generated by the two vehicles. The boundary consistency errors in the three tests reached 18.7 cm, 22.3 cm, and 19.5 cm, respectively, with an average error of 20.17 cm. This inconsistency in boundary understanding can cause serious problems in multi-vehicle collaborative inspection scenarios: when the two vehicles have a deviation of nearly 20 cm in judging the safety boundary of the same aircraft, it may lead to one vehicle believing that a certain area is safe to pass through while the other vehicle determines it to be a restricted area, thereby causing problems such as path planning conflicts, duplicate coverage, or imbalance of safety distance, which seriously affects the efficiency and safety of collaborative operations. In contrast, the present invention adopts a master-slave collaborative architecture, in which the boundary generation vehicle uniformly calculates the safety boundary and broadcasts it to the boundary consumption vehicle, ensuring that all vehicles in the fleet make decisions based on the exact same boundary information. The boundary consistency errors in the three tests were only 0.8 cm, 1.2 cm and 0.6 cm, with an average error of only 0.87 cm, which is 95.7% lower than the prior art.

[0054] Table 1. Comparison between the present invention and prior art Parameter name Existing Technology - Experiment 1 Existing technology - Experiment 2 Existing technology - Test 3 This invention - Experiment 1 This invention - Experiment 2 This invention - Experiment 3 Single-vehicle boundary calculation delay 142ms 138ms 145ms 143ms 141ms 146ms Boundary consistency error 18.7cm 22.3cm 19.5cm 0.8cm 1.2cm 0.6cm Total system power consumption 286W 291W 283W 178W 182W 175W Communication bandwidth usage 12.3KB / s 13.1KB / s 11.8KB / s 94.5KB / s 97.2KB / s 92.8KB / s CPU average utilization 67.2% 69.5% 66.8% 42.3% 43.8% 41.7% Peak memory usage 1428MB 1452MB 1435MB 896MB 912MB 883MB Boundary update frequency 9.2Hz 9.5Hz 9.3Hz 9.8Hz 10.0Hz 9.7Hz Boundary generation delay 148ms 152ms 146ms 145ms 149ms 143ms Path planning success rate 94.3% 95.1% 93.8% 99.2% 99.5% 99.3% System response time 265ms 272ms 258ms 198ms 203ms 195ms Furthermore, analyzing the total system power consumption, existing solutions require each autonomous vehicle to be equipped with a high-performance LiDAR and computing unit, independently executing a complete set of complex algorithms, including point cloud filtering, ground segmentation, target clustering, convex hull calculation, and safety expansion. The total system power consumption measured in three tests was 286 watts, 291 watts, and 283 watts, with an average power consumption of 286.67 watts. This high power consumption not only shortens the autonomous vehicle's range and increases charging frequency and maintenance costs, but also limits the expansion of the fleet size. The solution of this invention, through a role-based division of labor mechanism, concentrates the computationally intensive boundary generation task on the boundary generation vehicle, while the boundary consumption vehicle only needs to receive, verify, and fuse boundary data, without performing complex point cloud processing and geometric calculations, significantly reducing the computational load. The total system power consumption in the three tests was 178 watts, 182 watts, and 175 watts, with an average power consumption of 178.33 watts, a reduction of 37.8% compared to existing technologies.

[0055] Furthermore, considering the two computational resource indicators of average CPU utilization and peak memory usage, existing solutions exhibit average CPU utilization between 66.8% and 69.5%, averaging 67.83%, and peak memory usage between 1428MB and 1452MB, averaging 1438.33MB. Such high resource usage means the computing platform operates under high load for extended periods, increasing hardware overheating and failure risks, and limiting the system's ability to perform other tasks simultaneously (such as image recognition and path planning optimization). In the solution of this invention, the computational burden on the boundary consumer vehicle is significantly reduced, with average CPU utilization decreasing to 41.7% to 43.8%, averaging 42.60%, a 37.2% reduction compared to existing technologies; peak memory usage decreases to 883MB to 912MB, averaging 897MB, a 37.6% reduction compared to existing technologies.

[0056] Preferably, from the perspective of communication bandwidth usage, existing technical solutions require less inter-vehicle communication because each vehicle independently calculates its boundary. The communication bandwidth usage in the three tests was 12.3KB / s, 13.1KB / s, and 11.8KB / s, with an average of 12.40KB / s. The solution of this invention requires the boundary generating vehicle to periodically broadcast boundary data packets to the boundary consuming vehicle, containing information such as vertex coordinate sequences, timestamps, and confidence levels. The communication bandwidth usage in the three tests was 94.5KB / s, 97.2KB / s, and 92.8KB / s, with an average of 94.83KB / s, approximately 7.6 times that of existing technologies. Although the communication bandwidth usage has increased, the absolute value is still very small (less than 100KB / s), occupying only 0.0005% of bandwidth resources in a 5G communication environment (uplink speed 180Mbps), and will not constitute a communication bottleneck. In terms of boundary update frequency and system response time, the boundary update frequency of existing technologies ranges from 9.2Hz to 9.5Hz, with an average of 9.33Hz; the boundary update frequency of the present invention ranges from 9.7Hz to 10.0Hz, with an average of 9.83Hz, showing a slight improvement. More importantly, regarding system response time (the total delay from environmental change to generating a usable boundary and completing path planning), existing technologies require 258ms to 272ms, with an average of 265ms; the present invention requires only 195ms to 203ms, with an average of 198.67ms, representing a 25.0% improvement in response speed.

[0057] Specifically, regarding the key safety indicator of path planning success rate, the existing technical solutions achieved success rates of 94.3%, 95.1%, and 93.8% in three trials, with an average of 94.40%, resulting in a planning failure rate of approximately 5.6%. The main reasons for planning failures were conflicts caused by inconsistent boundaries and algorithm timeouts due to limited computing resources per vehicle. The path planning success rates of the proposed solution were 99.2%, 99.5%, and 99.3%, with an average of 99.33%, and a failure rate of only 0.67%, representing a reduction of 88.0% compared to existing technologies.

[0058] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration, characterized in that: include, Select one autonomous vehicle from the fleet as the boundary generation vehicle, and the remaining vehicles as boundary consumption vehicles; The boundary generation vehicle travels around the aircraft along a predetermined path and collects point cloud data. The point cloud data is then filtered and the target is extracted to obtain the aircraft point cloud. The aircraft point cloud is projected onto a horizontal plane, the Graham scan algorithm is used to calculate the convex hull boundary, and the convex hull boundary is subjected to safe expansion processing to generate a safe boundary. The security boundary is encapsulated into a boundary data packet and broadcast to the boundary consumption vehicle, wherein the boundary consumption vehicle integrates the security boundary into the local cost map for real-time path planning.

2. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 1, characterized in that: The boundary data packet includes the vertex coordinate sequence of the security boundary S, the current timestamp, the confidence level C, and the boundary generation vehicle identifier; the specific formula for the confidence level C is as follows: ； in, For the effective number of point clouds, The preset upper limit threshold for the number of valid point clouds. The weighting coefficients for the point cloud quantity factor. The weighting coefficients for the time decay factor. These are the weighting coefficients for the distance factor. The duration since the last structural update of the security boundary. The time decay coefficient, The average distance of the scan. The maximum distance threshold for effective scanning of aircraft by lidar.

3. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 2, characterized in that: It also includes, The boundary consumption vehicle adjusts its safety strategy based on confidence level C; When the confidence level C is higher than the confidence level threshold, the safety boundary is directly applied. When the confidence level C is lower than the confidence level threshold, the safety boundary is expanded further before use.

4. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 1, characterized in that: The method for generating the security boundary S is as follows: Set the vehicle safety radius as δ, and perform Minkowski sum operation on the convex hull boundary H and the disk B with radius δ centered at the origin to obtain the safety boundary S; Traverse all vertices of the convex hull boundary H Calculate vertices interior angle of ; Calculate the vertex interior angle of Unit vector in the direction of the bisector Along the unit vector The direction will be the vertex Offset the vertex outward by a distance to obtain the new, expanded vertex. ; Connect all the new vertices after expansion. This forms a safe boundary after smooth expansion.

5. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 4, characterized in that: The expanded new vertex The specific formula is as follows: ； in, As vertex interior angle of The unit vector in the direction of the angle bisector. For all vertices of the convex hull boundary H, For the vehicle's safe radius, As vertex inside angle.

6. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 4, characterized in that: The method for calculating the convex hull boundary is as follows: The three-dimensional coordinates of the aircraft point cloud are vertically projected onto the horizontal plane to obtain a two-dimensional projection point set. Traverse the set of two-dimensional projection points and find the point with the smallest y-coordinate as the base point. and press the remaining points relative to the base point. The points are sorted by their polar angles to obtain a point sequence. Initialize the stack structure and set the base point Push the first point in the aforementioned point sequence onto the stack in sequence; Traverse the remaining points in the sorted point sequence, for the current point Get the top of the stack and the apex of the stack The cross product of vectors is calculated using the following formula: ； When the cross product of vectors is less than or equal to zero, then the stack vertex If the point is concave, pop it from the stack; Repeat this process until the cross product of the vectors is greater than zero or only one point remains in the stack, then move the current point... Push onto the stack; After traversing all points, the points stored in the stack are the set of vertices of the convex hull boundary H. These vertices are then connected in polar angle order to form the convex hull boundary H.

7. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 6, characterized in that: The method for obtaining the aircraft point cloud is as follows: Based on point cloud data, a statistical filtering algorithm is used to remove noise points; The RANSAC algorithm is used to fit the ground plane to the filtered point cloud data, extract the plane equation parameters, mark the point cloud data with a distance less than the ground threshold from the plane equation as ground points and remove them to obtain non-ground point clouds; Euclidean clustering is performed on the non-terrestrial point cloud to calculate the point cloud data. With point cloud data The Euclidean distance between them; When the Euclidean distance is less than a preset distance threshold, the point cloud data will be... With point cloud data Points are grouped into the same point cluster, the number of points in each cluster is counted, and the cluster with the largest number of points is selected as the aircraft point cloud.

8. A system for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration, based on the method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in any one of claims 1 to 7, characterized in that: It includes a boundary generation vehicle and a boundary consumption vehicle; the boundary generation vehicle includes a computing processing terminal, a navigation module, a communication module, and a power module; The computing terminal includes an industrial control module and an edge computing module; the navigation module includes a lidar, a vision module, and an RTK module; and the communication module includes a 4G module, a 5G module, a WiFi module, and a Bluetooth module.

9. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 8, characterized in that: The industrial control module is responsible for scheduling, task distribution and data processing of each module; the edge computing module analyzes the inspection images in real time to identify faults and analyzes sensor data to provide beyond-line-of-sight perception information for autonomous vehicles.

10. The method for generating and sharing dynamic boundaries for aircraft inspection based on master-slave collaboration as described in claim 8, characterized in that: The communication module broadcasts the boundary data packets using a DDS-based communication network protocol; the boundary consumer vehicle verifies the timeliness of the boundary data packets through the communication module and integrates the security boundary S as a restricted area into the local cost map.

Images