A method for unmanned aerial vehicle intelligent inspection of a road construction site

Abstract

The application discloses a kind of road construction site unmanned plane intelligent inspection method, belong to unmanned plane inspection control technical field.This method collects the real-time position information, attitude information and wind field distribution data of unmanned plane in current inspection area, generates inspection state data set;According to wind field distribution data, reconstruct three-dimensional wind disturbance potential field through space interpolation, extract disturbance gradient feature, calculate the thrust variation of each rotor and dynamically correct the preplanned inspection route, generate the corrected inspection track;Control unmanned plane to fly along the corrected inspection track, while collecting deformation monitoring data on the surface of road, generate deformation spatial distribution map;Identify abnormal deformation area and generate fine inspection task plan, send to unmanned plane flight control system execution.This scheme can reduce the influence of wind disturbance on track, and carry out targeted fine inspection on abnormal deformation area.

Description

Technical Field

[0001] This invention relates to the field of unmanned aerial vehicle (UAV) inspection and control technology, specifically to an intelligent UAV inspection method for road construction sites. Background Technology

[0002] Drone inspections at road construction sites typically involve operators pre-setting flight paths. The drones, equipped with visible light cameras or laser scanners, fly along fixed waypoints using satellite positioning and inertial navigation to collect data on the construction area. For monitoring road surface deformation, common methods include deploying ground monitoring points within the construction area or performing a uniform scan of the entire area, followed by manual interpretation or offline data processing to identify deformation. During scanning, the drone follows a fixed flight path and fixed scanning parameters, making it impossible to adjust the focus of observation based on the location and extent of abnormal areas during flight. As the scope of road construction expands and the site environment becomes more complex, these fixed flight paths and uniform scanning methods are increasingly insufficient for on-site inspection needs. Tower cranes, barriers, material stockpiles, transport vehicles, and construction machinery at construction sites alter local airflow, resulting in significant differences in wind fields within a small space, which change over time. Pre-planned flight paths do not incorporate real-time wind field distribution information, making it easy for drones to deviate from their intended flight path and exhibit significant attitude fluctuations when traversing areas of strong winds or eddies. Using a single wind speed value or overall average wind compensation is insufficient to reflect the spatial gradient of the wind field in the horizontal and vertical directions. This makes it impossible to adjust thrust for different rotors in response to local wind disturbances, resulting in shifts in the scanned point cloud position, insufficient overlap between adjacent scan zones, or missed scans. While some UAV platforms are equipped with wind speed sensors, the wind field data is only used for flight safety alerts or overall hovering compensation; the spatial inhomogeneity of the wind field is not fully quantified, and rotor thrust adjustments lack a basis corresponding to local wind disturbances. Regarding deformation monitoring, point cloud data needs to be exported and manually compared with the initial surface model. After uniform scanning of the entire area, abnormal areas must be manually identified. Abnormal areas cannot be automatically identified during inspections, nor can detailed local inspection tasks be generated in real time. The data density and observation angles acquired in key areas are insufficient, often requiring additional re-measurement or supplementary ground measurements. Summary of the Invention

[0003] To address the problem that existing UAV inspections at road construction sites cannot automatically and precisely inspect areas with deviations from flight paths and abnormal deformations under wind disturbance conditions, this paper proposes an intelligent UAV inspection method for road construction sites. This method reconstructs a three-dimensional wind disturbance potential field based on real-time wind field distribution data and dynamically corrects the inspection flight path. At the same time, it automatically identifies abnormal deformation areas based on the spatial distribution of deformation and generates a detailed inspection task plan, adapting to the inspection needs of complex wind fields and dispersed abnormal deformation areas.

[0004] To achieve the above objectives, the present invention provides the following technical solution: The present invention provides an intelligent inspection method for road construction sites using unmanned aerial vehicles (UAVs), comprising: collecting real-time location information, attitude information, and wind field distribution data of the UAV in the current inspection area to generate an inspection status dataset; reconstructing a three-dimensional wind disturbance potential field through spatial interpolation based on the wind field distribution data in the inspection status dataset, and extracting the disturbance gradient features of the three-dimensional wind disturbance potential field, wherein the disturbance gradient features include the rate of change of disturbance intensity in the horizontal direction, the rate of change of disturbance intensity in the vertical direction, and the disturbance vorticity distribution; calculating the thrust change of each rotor based on the disturbance gradient features, and dynamically correcting the pre-planned inspection route using the thrust change to generate a corrected inspection trajectory; controlling the UAV to fly along the corrected inspection trajectory, while simultaneously collecting deformation monitoring data of the road surface to generate a deformation spatial distribution map; identifying abnormal deformation areas based on the deformation spatial distribution map, and generating a detailed inspection task plan for the abnormal deformation areas; and sending the detailed inspection task plan to the UAV flight control system for execution. This method reconstructs a three-dimensional wind disturbance potential field using wind field distribution data and extracts disturbance gradient features, dynamically corrects the pre-planned inspection route, enables UAVs to maintain stable flight in complex wind field environments at road construction sites, and improves the accuracy of road deformation monitoring and the efficiency of abnormal area identification.

[0005] As a preferred embodiment of the present invention, the step of collecting real-time location information, attitude information, and wind field distribution data of the UAV in the current inspection area to generate an inspection status dataset includes: arranging multiple wind speed sensors at symmetrical positions on the UAV body; synchronously collecting the wind field distribution data using the multiple wind speed sensors at a preset sampling frequency; aligning the real-time location information, attitude information, and wind field distribution data according to the sampling time using timestamps; and storing the aligned data in a circular buffer to form the inspection status dataset. This solution can ensure the temporal consistency of multi-source data and improve the real-time performance of data acquisition.

[0006] As a preferred embodiment of the present invention, the reconstruction of the three-dimensional wind disturbance potential field through spatial interpolation includes: dividing the current inspection area into multiple three-dimensional voxel grids; for each three-dimensional voxel grid, collecting wind field distribution data located within and adjacent to the grid, and calculating the wind disturbance vector at the center point of the grid using an inverse distance weighted interpolation algorithm; and aggregating the wind disturbance vectors of all three-dimensional voxel grids to form the three-dimensional wind disturbance potential field. This scheme can construct a continuous three-dimensional wind disturbance potential field, providing a reliable data foundation for disturbance gradient feature extraction.

[0007] As a preferred embodiment of the present invention, the extraction of the disturbance gradient features of the three-dimensional wind disturbance potential field includes: selecting multiple spatial sampling points in the three-dimensional wind disturbance potential field; for each spatial sampling point, calculating the wind disturbance vector difference between the point and its horizontally adjacent sampling points to obtain the horizontal disturbance intensity change rate; calculating the wind disturbance vector difference between the point and its vertically adjacent sampling points to obtain the vertical disturbance intensity change rate; calculating the wind disturbance vector rotation in the neighborhood of the point to obtain the disturbance vorticity distribution; and combining the horizontal disturbance intensity change rate, the vertical disturbance intensity change rate, and the disturbance vorticity distribution to form the disturbance gradient features. This scheme can comprehensively reflect the spatial variation and rotation characteristics of the wind disturbance potential field, improving the targeting of subsequent thrust corrections.

[0008] As a preferred embodiment of the present invention, the step of calculating the thrust variation of each rotor based on the disturbance gradient characteristics includes: calculating the additional thrust correction values ​​of the UAV along the flight path in the horizontal plane in the lateral and longitudinal directions according to the horizontal disturbance intensity change rate; calculating the additional thrust correction value of the UAV in the vertical direction according to the vertical disturbance intensity change rate; calculating the yaw moment correction value of the UAV according to the disturbance vorticity distribution; and mapping the additional thrust correction value and the yaw moment correction value to the thrust variation of each rotor according to the arrangement position of each rotor. This solution can convert wind disturbance gradient characteristics into thrust compensation amounts for each rotor, improving the attitude stability and flight path maintenance capability of the UAV in windy environments.

[0009] As a preferred embodiment of the present invention, the step of dynamically correcting the pre-planned inspection route using the thrust change to generate a corrected inspection track includes: discretizing the pre-planned inspection route into multiple waypoints; for each waypoint, adjusting the desired position offset of the waypoint according to the thrust change to obtain a corrected waypoint; and reconnecting all the corrected waypoints in the order of the original route to form the corrected inspection track. This solution can improve the safety of inspection flights and the quality of data acquisition by locally adjusting the route according to wind disturbance conditions while preserving the original inspection mission coverage.

[0010] As a preferred embodiment of the present invention, controlling the UAV to fly along the corrected inspection trajectory includes: acquiring angular velocity and acceleration data output by the UAV's inertial measurement unit in real time; comparing the angular velocity and acceleration data with the desired position on the corrected inspection trajectory to calculate the current flight error; and adjusting the rotational speed of each rotor using a proportional-integral-derivative (PID) control method based on the current flight error, so that the UAV tracks the corrected inspection trajectory. This solution can achieve accurate tracking of the corrected trajectory and reduce flight error.

[0011] As a preferred embodiment of the present invention, the step of collecting deformation monitoring data of the road surface and generating a deformation spatial distribution map includes: acquiring a three-dimensional point cloud of the road surface using an airborne laser scanner and recording the position and attitude information of the UAV corresponding to each scanning point; converting the three-dimensional point cloud to a world coordinate system to obtain a road surface coordinate sequence; overlaying the road surface coordinate sequence with pre-stored initial road surface reference data to calculate the deformation displacement at each location; and filling the deformation displacement into a two-dimensional grid according to its spatial location to obtain the deformation spatial distribution map. This solution can obtain a high-precision spatial distribution of road surface deformation, providing a quantitative basis for the identification of abnormal areas.

[0012] As a preferred embodiment of the present invention, identifying abnormal deformation regions based on the deformation spatial distribution map includes: marking spatial location points with deformation displacements greater than a set limit as candidate deformation points in the deformation spatial distribution map; performing neighborhood connectivity analysis on the candidate deformation points, merging interconnected candidate deformation points into a connected region; calculating the area and average deformation displacement of each connected region, and identifying connected regions with areas exceeding a preset area limit or average deformation displacement exceeding a preset deformation limit as the abnormal deformation regions. This solution can automatically identify abnormal deformation regions of engineering significance, reducing missed detections and false detections.

[0013] As a preferred embodiment of the present invention, sending the detailed inspection task plan to the UAV flight control system for execution includes: the detailed inspection task plan comprising a sequence of hovering observation points and a sequence of scanning parameters; the UAV flight control system parsing the sequence of hovering observation points to obtain the three-dimensional coordinates and arrival order of each hovering observation point; the UAV sequentially flying to the hovering observation points, and adjusting the gimbal pointing and the resolution of the airborne laser scanner at each hovering observation point according to the corresponding scanning parameters to perform the inspection. This solution enables detailed supplementary inspections of abnormal deformation areas, improving the resolution and targeting of data acquisition in abnormal areas.

[0014] The technical effects and advantages provided by the present invention in the above technical solution are as follows:

[0015] A three-dimensional wind disturbance potential field is reconstructed through spatial interpolation. The horizontal and vertical disturbance intensity change rates and disturbance vorticity distribution are extracted. The thrust change of each rotor is calculated from the disturbance gradient characteristics, and the pre-planned inspection route is dynamically corrected. This method converts discrete wind field distribution data into wind disturbance vectors and gradient information in three-dimensional space, which can describe the non-uniform changes of the wind field in the horizontal and vertical directions and the rotational effect of vortices on the UAV body. The horizontal disturbance intensity change rate is used to calculate the additional thrust correction values ​​in the lateral and longitudinal directions along the track in the horizontal plane, the vertical disturbance intensity change rate is used to calculate the additional thrust correction value in the vertical direction, and the disturbance vorticity distribution is used to calculate the yaw moment correction value. Then, the additional thrust correction values ​​and yaw moment correction values ​​are mapped to the thrust change of each rotor according to the rotor arrangement position. Unlike methods that rely solely on a single wind speed value or overall average wind compensation, this method uses rotor thrust variation to correspond to local wind disturbance gradients. This suppresses attitude disturbances in roll, pitch, and yaw directions when the UAV traverses areas of strong winds, eddies, and abrupt airflow changes at road construction sites, reducing the deviation between the actual flight trajectory and the corrected inspection track. With reduced track deviation, the point cloud positions obtained by the onboard laser scanner maintain better consistency with the pre-planned scan zones, improving overlap between adjacent scan zones, reducing missed scans and duplicate acquisitions, and enhancing the spatial continuity and availability of deformation data compared to fixed-route inspection methods. Based on road surface deformation monitoring data, a spatial distribution map of deformation is generated. Neighborhood connectivity analysis is performed on candidate points where deformation displacement exceeds set limits. Abnormal deformation areas are identified based on the area of ​​connected regions and average deformation displacement. A detailed inspection task plan, including a sequence of hovering observation points and scanning parameters, is then generated and sent to the UAV flight control system for execution. This method integrates anomaly area identification and refined task generation into a single inspection process, eliminating the need for pre-setting fixed monitoring points on the road surface and for manual interpretation and flight path replanning after a full-area scan. The UAV adjusts its hovering observation points, gimbal pointing, and laser scanning resolution based on the distribution, connectivity, and deformation degree of the anomaly areas, enabling locally intensified observation of these areas, while maintaining the same scanning density in non-anomaly areas. Compared to uniform full-area scanning and subsequent manual re-measurement, this local task generation method reduces the ineffective scanning of areas without significant deformation, allowing anomaly areas to obtain spatial sampling density and observation angles that match their deformation characteristics. Attached Figure Description

[0016] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0017] Figure 1 This is a flowchart of a method for intelligent inspection of road construction sites using drones.

[0018] Figure 2 This is a schematic diagram of the process for collecting wind field distribution data and aligning timestamps using drones;

[0019] Figure 3 This is a flowchart of UAV inspection trajectory correction based on perturbation gradient and vorticity. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] See Figure 1 This invention provides an intelligent inspection method for road construction sites using unmanned aerial vehicles (UAVs), comprising: collecting real-time location information, attitude information, and wind field distribution data of the UAV in the current inspection area to generate an inspection status dataset; reconstructing a three-dimensional wind disturbance potential field through spatial interpolation based on the wind field distribution data in the inspection status dataset, and extracting the disturbance gradient features of the three-dimensional wind disturbance potential field, including the rate of change of disturbance intensity in the horizontal direction, the rate of change of disturbance intensity in the vertical direction, and the distribution of disturbance vorticity; calculating the thrust change of each rotor based on the disturbance gradient features, and dynamically correcting the pre-planned inspection route using the thrust change to generate a corrected inspection trajectory; controlling the UAV to fly along the corrected inspection trajectory, while simultaneously collecting deformation monitoring data of the road surface to generate a deformation spatial distribution map; identifying abnormal deformation areas based on the deformation spatial distribution map, and generating a detailed inspection task plan for the abnormal deformation areas; and sending the detailed inspection task plan to the UAV flight control system for execution.

[0022] In specific implementation, please refer to Figure 2 A wind speed sensor is positioned beneath each of the four arms of the drone's fuselage. The four sensors are symmetrically distributed about the center of the drone's fuselage, with equal horizontal and vertical distances from the center. They are also aligned in the same direction, with their air inlets facing forward. Each wind speed sensor is connected to the drone's flight control system's data acquisition unit via a serial peripheral interface. The data acquisition unit provides a unified clock signal to all four sensors.

[0023] Wind field distribution data is collected synchronously using four wind speed sensors at a preset sampling frequency of 50 Hz. Each wind speed sensor outputs wind speed components in three orthogonal directions at each sampling moment. The data acquisition unit reads the output values ​​of the four wind speed sensors at each sampling moment and combines the output values ​​of the four wind speed sensors to form wind field distribution data. Each sampling point in the wind field distribution data includes the sampling time and the wind speed components of each of the four wind speed sensors.

[0024] The real-time location information, attitude information, and wind field distribution data are timestamped according to the sampling time. Real-time location information is output by the satellite positioning module at a sampling frequency of 10 Hz, and attitude information is output by the inertial measurement unit at a sampling frequency of 200 Hz. Using the sampling time of the real-time location information as the reference time, for each reference time, two adjacent sampling points that are time-sequentially before and after the reference time are selected from the wind field distribution data. The aligned wind field distribution data is then calculated using a linear interpolation formula:

[0025]

[0026] in, Indicates at the reference time The wind vector in the aligned wind field distribution data; This indicates the location of the wind field distribution data at the reference time. Previous adjacent sampling time Wind vector at the location; This indicates the location of the wind field distribution data at the reference time. Subsequent adjacent sampling times Wind vector at the location; This indicates the location of the wind field distribution data at the reference time. Previous adjacent sampling times; This indicates the location of the wind field distribution data at the reference time. The subsequent adjacent sampling times; The sampling time of the real-time location information is used as the reference time for timestamp alignment. Attitude information is also timestamp aligned to the reference time. For two adjacent sampling points of attitude information before and after the reference time, the aligned attitude information is calculated using the same linear interpolation method as the wind field distribution data.

[0027] In practice, the aligned data is stored in a circular buffer to form an inspection status dataset. The circular buffer is a fixed-length buffer allocated in the UAV flight control system's memory, containing 6000 storage units. Each storage unit stores the real-time position information, attitude information, and aligned wind field distribution data corresponding to a reference time. The circular buffer is managed using write pointers and read pointers. The write pointer points to the next storage unit to be written to, and the read pointer points to the earliest written storage unit that has not yet been read. When the write pointer reaches the end of the buffer, it jumps to the beginning of the buffer, overwriting the earliest written storage unit. After each data alignment for a reference time is completed, the aligned real-time position information, attitude information, and aligned wind field distribution data are written to the storage unit pointed to by the write pointer in the circular buffer according to the timestamp order. The write pointer moves forward one storage unit, forming the inspection status dataset.

[0028] In practice, the current inspection area is divided into a two-dimensional grid along a first horizontal direction and a second horizontal direction in the horizontal plane. The first horizontal direction is east-west, and the second horizontal direction is north-south. Vertically, it is divided into multiple height layers. Each three-dimensional voxel grid has a horizontal side length of 2 meters and a vertical side length of 1 meter. The origin is taken as the southwest corner of the current inspection area, and the first horizontal direction is... axial direction, with the second horizontal direction as The axial direction, with the vertical upward direction as... Establish a voxel grid coordinate system along the axis. Divide the current inspection area into... A horizontal row, Horizontal columns and A vertical layer, The integer obtained by dividing the total length of the current inspection area along the first horizontal direction by 2 meters and rounding it up. The integer obtained by dividing the total length of the current inspection area along the second horizontal direction by 2 meters and rounding it up. The integer obtained by dividing the total vertical height of the current inspection area by 1 meter and rounding it up.

[0029] For each 3D voxel grid, wind field distribution data are collected within and adjacent to that 3D voxel grid. For the... The three-dimensional voxel mesh, with the first The center point of each three-dimensional voxel grid is used as the center of a sphere, forming a spherical neighborhood with a radius of 3 meters. Wind field distribution data sampling points whose spatial coordinates fall within this spherical neighborhood are used to calculate the... The nearest sampling points of the wind disturbance vector at the center point of the three-dimensional voxel mesh. Let this be used to calculate the... The total number of neighboring sampling points of the wind disturbance vector at the center point of the three-dimensional voxel grid is .like If the value is less than 4, the radius of the spherical neighboring region is increased by 1 meter and neighboring sampling points are collected again until... Not less than 4.

[0030] Calculate the first using the inverse distance weighted interpolation algorithm The wind disturbance vector at the center point of a 3D voxel grid is calculated using the following formula:

[0031]

[0032] in, Indicates the first Wind disturbance vector at the center point of a three-dimensional voxel grid; Indicates the first Spatial coordinates of the center point of a three-dimensional voxel grid; Indicates the use of calculating the first The index of the neighboring sampling points of the wind disturbance vector at the center point of a three-dimensional voxel grid. The value range is 1 to ; Indicates the use of calculating the first The total number of neighboring sampling points of the wind disturbance vector at the center point of a three-dimensional voxel grid; Indicates the first Wind velocity vectors of neighboring sampling points; Indicates the first The center point of the three-dimensional voxel mesh and the first Spatial distance between neighboring sampling points; This represents the power exponent in the inverse distance weighted interpolation algorithm. The value of is 2. When the power exponent in the inverse distance weighted interpolation algorithm is 2, the weight decreases with the square of the spatial distance, thus maintaining local interpolation stability when the wind field data is sparse. This indicates a smoothing term to prevent division by zero. The value is 0.001 meters, when the first... The nearest sampling point and the first Spatial distance between the center points of the three-dimensional voxel grid When it is zero, the smoothing term This avoids the denominator being zero while also not significantly affecting the interpolation result.

[0033] Perform the above calculations on all three-dimensional voxel grids within the current inspection area to obtain the wind disturbance vector at the center point of each three-dimensional voxel grid. Store the wind disturbance vectors at the center points of all three-dimensional voxel grids according to the spatial arrangement order of the three-dimensional voxel grids to form a three-dimensional wind disturbance potential field.

[0034] Multiple spatial sampling points are selected in the three-dimensional wind disturbance potential field. The center point of each three-dimensional voxel grid is used as a spatial sampling point, and the total number of spatial sampling points is the same as the total number of three-dimensional voxel grids.

[0035] For each spatial sampling point, the wind disturbance vector difference between that spatial sampling point and its horizontally adjacent sampling points is calculated to obtain the rate of change of horizontal disturbance intensity. For the ... For each spatial sampling point, two horizontally adjacent spatial sampling points are selected along the positive and negative directions of the first horizontal direction, respectively. Similarly, two horizontally adjacent spatial sampling points are selected along the positive and negative directions of the second horizontal direction, respectively. Calculate the... The magnitude of the wind disturbance vector difference between the nth spatial sampling point and the adjacent spatial sampling point along the positive direction of the first horizontal direction is used to calculate the nth... The magnitude of the wind disturbance vector difference between the nth spatial sampling point and the adjacent spatial sampling point along the negative direction of the first horizontal direction is used to calculate the nth... The magnitude of the wind disturbance vector difference between the nth spatial sampling point and the adjacent spatial sampling point along the positive direction of the second horizontal direction is used to calculate the nth... The magnitude of the wind disturbance vector difference between the first spatial sampling point and its adjacent spatial sampling point along the negative direction of the second horizontal direction. The average of the magnitudes of the four wind disturbance vector differences is then calculated, and the average value is divided by the horizontal side length of the three-dimensional voxel mesh, which is 2 meters, to obtain the magnitude of the first... The rate of change of horizontal disturbance intensity at each spatial sampling point.

[0036] Calculate the wind disturbance vector difference between this spatial sampling point and its vertically adjacent sampling points to obtain the rate of change of vertical disturbance intensity. For the ... For each spatial sampling point, two spatial sampling points perpendicularly adjacent to it are selected along the positive and negative vertical directions, respectively. Calculate the... The magnitude of the wind disturbance vector difference between the nth spatial sampling point and its adjacent spatial sampling point along the positive vertical direction is used to calculate the nth... The magnitude of the wind disturbance vector difference between the nth spatial sampling point and its adjacent spatial sampling point along the negative vertical direction is calculated. The average of these two wind disturbance vector differences is then divided by the vertical side length of the 3D voxel mesh, which is 1 meter, to obtain the magnitude of the nth... The rate of change of vertical disturbance intensity at each spatial sampling point.

[0037] Calculate the wind disturbance vector rotation in the neighborhood of the spatial sampling point to obtain the disturbance vorticity distribution. For the th The spatial sampling point, with the first Centered on the first spatial sampling point, select six adjacent spatial sampling points along the positive direction of the first horizontal direction, the negative direction of the first horizontal direction, the positive direction of the second horizontal direction, the negative direction of the second horizontal direction, the positive direction of the vertical direction, and the negative direction of the vertical direction. Record the first... The wind disturbance vector at each spatial sampling point is The wind disturbance vector of adjacent spatial sampling points along the positive direction of the first horizontal direction is The wind disturbance vector of the adjacent spatial sampling points along the negative direction of the first horizontal direction is The wind disturbance vector of adjacent spatial sampling points along the positive direction of the second horizontal direction is The wind disturbance vector of the adjacent spatial sampling points along the negative direction of the second horizontal direction is The wind disturbance vector of adjacent spatial sampling points along the positive vertical direction is The wind disturbance vector of adjacent spatial sampling points along the negative vertical direction is The calculation method for the wind disturbance vector rotation is as follows: subtract the component of the wind disturbance vector change between adjacent spatial sampling points along the first horizontal direction in the second horizontal direction from the component of the wind disturbance vector change between adjacent spatial sampling points along the second horizontal direction in the first horizontal direction, to obtain the first component. The vorticity component of each spatial sampling point in the vertical direction; subtracting the component of the wind disturbance vector change between adjacent spatial sampling points along the second horizontal direction in the vertical direction from the component of the wind disturbance vector change between adjacent spatial sampling points along the vertical direction in the second horizontal direction, yields the vorticity component of the first spatial sampling point in the vertical direction. The vorticity component of the first spatial sampling point in the first horizontal direction; subtracting the component of the wind disturbance vector change between adjacent spatial sampling points in the first horizontal direction from the component of the wind disturbance vector change between adjacent spatial sampling points in the first horizontal direction in the vertical direction, yields the vorticity component of the first spatial sampling point in the first horizontal direction. The vorticity components of the three spatial sampling points in the second horizontal direction. These three vorticity components together constitute the first... The perturbation eddy current distribution at each spatial sampling point.

[0038] The first The horizontal disturbance intensity change rate, vertical disturbance intensity change rate, and disturbance vorticity distribution of each spatial sampling point are combined to form the... The perturbation gradient characteristics of each spatial sampling point are obtained by performing the above calculation on all spatial sampling points, thus obtaining the perturbation gradient characteristics of the entire current inspection area.

[0039] In specific implementation, please refer to Figure 3 The horizontal disturbance intensity change rate in the disturbance gradient features is obtained. Based on the tangent direction of the pre-planned inspection route at the waypoint, a track coordinate system is established. The longitudinal axis of the track coordinate system coincides with the tangent direction of the pre-planned inspection route at the waypoint, and the transverse axis of the track coordinate system is perpendicular to the longitudinal axis in the horizontal plane and points to the right of the track. The horizontal disturbance intensity change rate is projected from the voxel grid coordinate system to the track coordinate system to obtain the disturbance intensity change rate components along the longitudinal direction and the transverse direction of the track.

[0040] Based on the rate of change of horizontal disturbance intensity, the additional thrust correction values ​​for the UAV along its trajectory in the horizontal plane are calculated for both the lateral and longitudinal directions. The same formula is used to calculate both the lateral and longitudinal additional thrust correction values:

[0041]

[0042] in, This represents the additional thrust correction value in the lateral or longitudinal direction of the flight path, expressed in Newtons. It is used when calculating the lateral additional thrust correction value. This represents the lateral thrust correction value for the flight path. When calculating the longitudinal thrust correction value for the flight path... This indicates the longitudinal thrust correction value for the flight path; The total mass of the drone is expressed in kilograms. The total mass of the drone is calculated in real time by the drone's flight control system based on the baseline mass measured before takeoff and the mass consumed by the battery during flight. This represents the horizontal aerodynamic response coefficient, which is dimensionless and ranges from 0.8 to 1.2, with a default value of 1.0. The horizontal aerodynamic response coefficient is determined based on the calibration relationship between the change in horizontal thrust and the rate of change in horizontal wind disturbance, as measured by the UAV in wind tunnel tests. This represents the equivalent wind disturbance acceleration, calculated by multiplying the projected component of the rate of change of horizontal disturbance intensity by the characteristic length of the UAV, in meters per second squared. The value is obtained by projecting the rate of change of horizontal disturbance intensity onto the track coordinate system. For the longitudinal component of the track, the longitudinal projection component is substituted into... The longitudinal thrust correction value of the trajectory is obtained; for the lateral component of the trajectory, the lateral projection component of the trajectory is substituted into... The lateral thrust correction value of the trajectory is obtained.

[0043] The additional thrust correction value in the vertical direction of the UAV is calculated based on the rate of change of vertical disturbance intensity. This is obtained by multiplying the rate of change of vertical disturbance intensity by the total mass of the UAV, and then multiplying by the vertical aerodynamic response coefficient. The vertical aerodynamic response coefficient is a dimensionless coefficient, ranging from 0.9 to 1.1, with a default value of 1.0. The vertical aerodynamic response coefficient is determined based on the calibration relationship between the change in vertical thrust and the rate of change of vertical wind disturbance measured in wind tunnel tests. The direction of the additional thrust correction value in the vertical direction is opposite to the direction of the rate of change of vertical disturbance intensity, used to counteract the wind disturbance effect in the vertical direction.

[0044] Based on the disturbance vorticity distribution, the yaw moment correction value of the UAV is calculated. The vertical vorticity component is extracted from the disturbance vorticity distribution. The vertical vorticity component is multiplied by the moment of inertia of the UAV about its vertical axis, and then multiplied by the yaw moment correction coefficient to obtain the yaw moment correction value. The yaw moment correction coefficient is a dimensionless coefficient, ranging from 0.1 to 0.3, with a default value of 0.2. The yaw moment correction coefficient is determined based on the calibration relationship between the change in yaw moment measured in a wind tunnel rotation test of the UAV about its vertical axis and the vertical vorticity component. The direction of the yaw moment correction value is opposite to the direction of the vertical vorticity component, used to counteract the yaw moment caused by wind disturbance vorticity.

[0045] The additional thrust correction and yaw moment correction values ​​are mapped to the thrust variation of each rotor according to their arrangement. For a quadcopter UAV, the four rotors are arranged symmetrically about the center of symmetry of the UAV body at the ends of the first, second, third, and fourth rotor arms, respectively. Let the longitudinal additional thrust correction value be denoted as... The lateral thrust correction value is The vertical additional thrust correction value is The yaw moment correction value is The change in first rotor thrust is calculated as follows: The change in first rotor thrust equals the vertical additional thrust correction value divided by four, minus the lateral additional thrust correction value multiplied by the first rotor lateral lever arm coefficient, minus the longitudinal additional thrust correction value multiplied by the first rotor longitudinal lever arm coefficient, plus the yaw moment correction value multiplied by the first rotor anti-torque distribution coefficient. The first rotor lateral lever arm coefficient is obtained by dividing the lateral distance from the first rotor center to the UAV body symmetry center by the sum of the lateral distances from the four rotor centers to the UAV body symmetry center, and the first rotor lateral lever arm coefficient is 0.25. The first rotor longitudinal lever arm coefficient is obtained by dividing the longitudinal distance from the first rotor center to the UAV body symmetry center by the sum of the longitudinal distances from the four rotor centers to the UAV body symmetry center, and the first rotor longitudinal lever arm coefficient is 0.25. The first rotor anti-torque distribution coefficient is determined by the proportion of the first rotor's anti-torque characteristic to the total anti-torque characteristic of the four rotors, and the first rotor anti-torque distribution coefficient is 0.25. The thrust changes of the second, third, and fourth rotors are calculated in the same way. The signs of the lateral lever arm coefficient and the longitudinal lever arm coefficient are determined according to the positions of the second, third, and fourth rotors in the airframe coordinate system. The anti-torque distribution coefficients of the second, third, and fourth rotors are all 0.25.

[0046] The pre-planned inspection route is discretized into multiple waypoints. Based on the arc length between adjacent original waypoints in the pre-planned inspection route, discrete waypoints are generated on the original route at 1-meter intervals. Each discrete waypoint contains the original three-dimensional coordinates on the pre-planned inspection route and the heading angle at that discrete waypoint.

[0047] For each waypoint, the desired position offset of that waypoint is adjusted according to the thrust change to obtain the corrected waypoint. For the... The nth discrete waypoint is obtained. The thrust variations of the first, second, third, and fourth rotors at discrete waypoints are calculated. These variations are then added together to obtain the total thrust correction. The total thrust correction is decomposed along the longitudinal, lateral, and vertical directions of the flight path to obtain a three-dimensional vector of total thrust correction. This three-dimensional vector is divided by the total mass of the UAV and multiplied by the position offset correction factor to obtain the desired three-dimensional vector of position offset. The position offset correction factor is a dimensionless coefficient, ranging from 0.1 to 0.5, with a default value of 0.25. The position offset correction factor is determined based on the maximum allowable horizontal and vertical deviations from the pre-planned inspection route during UAV operation. When the maximum allowable horizontal deviation is 0.5 meters and the maximum allowable vertical deviation is 0.3 meters, the position offset correction factor is set to 0.25 to ensure that the maximum desired position offset does not exceed the allowable deviation distance. The desired position offset three-dimensional vector is then multiplied by the first rotor thrust variation. Adding the original three-dimensional coordinates of the discrete waypoints together yields the first... The corrected three-dimensional coordinates of the waypoints.

[0048] All corrected waypoints are reconnected in the original order to form a corrected inspection track. A cubic spline interpolation algorithm is then used to interpolate all corrected waypoints, generating a continuous and smooth corrected inspection track. The corrected inspection track retains the original order of all corrected waypoints, with the track curve passing through each corrected waypoint. The track between adjacent corrected waypoints is determined by the cubic spline interpolation function.

[0049] In specific implementations, the angular velocity and acceleration data output by the inertial measurement unit of the unmanned aerial vehicle are acquired in real time. The inertial measurement unit in the flight control system of the unmanned aerial vehicle comprises three orthogonally mounted gyroscopes and three orthogonally mounted accelerometers. The inertial measurement unit outputs three-axis angular velocity data and three-axis acceleration data at a sampling frequency of 200 Hz. The flight control system of the unmanned aerial vehicle reads the angular velocity data and the acceleration data output by the inertial measurement unit in each sampling period. The angular velocity data includes the angular velocity component around the first horizontal axis of the body coordinate system, the angular velocity component around the second horizontal axis of the body coordinate system, and the angular velocity component around the vertical axis of the body coordinate system. The acceleration data includes the acceleration component along the first horizontal axis of the body coordinate system, the acceleration component along the second horizontal axis of the body coordinate system, and the acceleration component along the vertical axis of the body coordinate system.

[0050] The flight control system of the unmanned aerial vehicle updates the attitude quaternion of the unmanned aerial vehicle by using the angular velocity data output by the inertial measurement unit. Taking the th sampling period as an example, let the attitude quaternion of the unmanned aerial vehicle at the initial moment of the th sampling period be , the angular velocity vector output by the inertial measurement unit in the th sampling period be , then the attitude quaternion of the unmanned aerial vehicle at the end moment of the th sampling period is obtained through quaternion differential update. The updated attitude quaternion of the unmanned aerial vehicle is used to calculate a rotation matrix, the acceleration data output by the inertial measurement unit is converted from the body coordinate system to the world coordinate system, so as to obtain the acceleration vector in the world coordinate system. The gravitational acceleration vector is subtracted from the acceleration vector in the world coordinate system, so as to obtain the motion acceleration vector in the world coordinate system. The motion acceleration vector in the world coordinate system is integrated, so as to obtain the velocity vector of the unmanned aerial vehicle in the world coordinate system. The velocity vector in the world coordinate system is integrated, so as to obtain the position coordinates of the unmanned aerial vehicle in the world coordinate system. Meanwhile, the real-time position information output by the satellite positioning module at a sampling frequency of 10 Hz is used to correct the position coordinates in the world coordinate system obtained by integration, so as to obtain the current real-time coordinates of the unmanned aerial vehicle in the world coordinate system.

[0051] The angular velocity and acceleration data are compared with the desired position on the corrected inspection track to calculate the current flight error. The waypoint closest to the UAV's current real-time coordinates in the world coordinate system is extracted from the corrected inspection track and used as the desired position. The three-dimensional coordinate difference between the UAV's current real-time coordinates in the world coordinate system and the desired position is calculated to obtain the position error. The position error includes components along the first horizontal axis, the second horizontal axis, and the vertical axis of the world coordinate system. Simultaneously, based on the tangent direction of the corrected inspection track at the desired position, the desired velocity vector is determined. The magnitude of the desired velocity vector is the cruise speed of the pre-planned inspection route. The velocity error is obtained by subtracting the UAV's velocity vector in the world coordinate system from the desired velocity vector. The velocity error includes components along the first horizontal axis, the second horizontal axis, and the vertical axis of the world coordinate system. Based on the desired velocity vector and the curvature of the corrected inspection track at the desired position, the desired attitude angles are determined. These desired attitude angles include the desired pitch angle, desired roll angle, and desired yaw angle. Subtracting the desired attitude angles from the current attitude angles of the UAV yields the attitude error. The attitude error includes pitch angle error, roll angle error, and yaw angle error.

[0052] Based on the current flight error, the rotational speed of each rotor is adjusted using proportional-integral-derivative (PID) control to enable the UAV to track the corrected inspection trajectory. The vertical thrust adjustment is calculated based on the vertical position error, using the following formula:

[0053]

[0054] in, This indicates the vertical thrust adjustment, measured in Newtons. The vertical thrust adjustment is added to the baseline vertical thrust required for the drone to hover. The position error component along the vertical axis of the world coordinate system is expressed in meters. The position error component along the vertical axis of the world coordinate system is equal to the vertical component of the UAV's current real-time coordinates in the world coordinate system minus the vertical component of the desired position. This represents the position error components along the vertical axis of the world coordinate system from the start of the current sampling period to the current time. The integral is expressed in meters per second. This represents the position error component along the vertical axis of the world coordinate system with respect to time. The derivative, expressed in meters per second; This represents the vertical proportional gain, expressed in Newtons per meter. The vertical proportional gain is set to 1.5 Newtons per meter. The vertical proportional gain is determined based on the debugging conditions that the vertical position step response of the UAV during hovering has no overshoot and the adjustment time is less than 0.5 seconds. This represents the vertical integral gain, expressed in Newtons per meter-second. The vertical integral gain is set to 0.2 Newtons per meter-second. The vertical integral gain is determined based on the debugging condition that the steady-state error of the vertical position of the UAV during hovering is less than 0.02 meters. This represents the vertical differential gain, measured in Newton-seconds per meter. The vertical differential gain is set to 0.5 Newton-seconds per meter. The vertical differential gain is determined based on the adjustment conditions where the vertical position response damping ratio of the UAV during hovering is between 0.7 and 1.0.

[0055] In practical implementation, the horizontal position error component, horizontal velocity error component, and attitude error are used, employing the same proportional-integral-derivative (PID) control structure as the vertical direction, to calculate the roll moment adjustment, pitch moment adjustment, and yaw moment adjustment. The roll moment adjustment is obtained by multiplying the position error component along the second horizontal axis of the world coordinate system, the velocity error component along the second horizontal axis of the world coordinate system, and the roll angle error by their respective proportional gain, integral gain, and derivative gain, and then summing them. The pitch moment adjustment is obtained by multiplying the position error component along the first horizontal axis of the world coordinate system, the velocity error component along the first horizontal axis of the world coordinate system, and the pitch angle error by their respective proportional gain, integral gain, and derivative gain, and then summing them. The yaw moment adjustment is obtained by multiplying the yaw angle error by its respective proportional gain, integral gain, and derivative gain, and then summing them.

[0056] The vertical thrust adjustment is added to the baseline vertical thrust required for the UAV to hover, yielding the total vertical thrust. The total vertical thrust, roll moment adjustment, pitch moment adjustment, and yaw moment adjustment are then mapped to the rotational speed adjustments of the four rotors according to the rotor arrangement of the quadcopter UAV. The four rotors are a first rotor, a second rotor, a third rotor, and a fourth rotor, which are symmetrically arranged about the UAV's fuselage center of symmetry. The calculation method for the rotational speed adjustment of the four rotors is as follows: The total vertical thrust is evenly distributed across the four rotors to obtain the baseline thrust component for each rotor. The roll moment adjustment is distributed according to the lever arm of the four rotors along the second horizontal axis of the airframe coordinate system, increasing the thrust of the rotor corresponding to the positive direction of the second horizontal axis and decreasing the thrust of the rotor corresponding to the negative direction of the second horizontal axis. The pitch moment adjustment is distributed according to the lever arm of the four rotors along the first horizontal axis of the airframe coordinate system, increasing the thrust of the rotor corresponding to the positive direction of the first horizontal axis and decreasing the thrust of the rotor corresponding to the negative direction of the first horizontal axis. The yaw moment adjustment is distributed according to the anti-torque characteristics of the four rotors, increasing the rotational speed of the rotor generating positive anti-torque and decreasing the rotational speed of the rotor generating negative anti-torque. The baseline thrust component of each rotor is added to the allocated values ​​for the roll moment, pitch moment, and yaw moment to obtain the target thrust for each rotor. Based on the quadratic relationship between the target thrust and rotor speed of each rotor, the target speed of each rotor is calculated from the target thrust. Subtracting the current speed from the target speed of each rotor yields the speed adjustment amount for each rotor. The UAV flight control system adjusts the output of the corresponding electronic speed controller according to the speed adjustment amount for each rotor, ensuring that the speeds of the first, second, third, and fourth rotors reach the target speeds, thus enabling the UAV to track the corrected inspection trajectory.

[0057] In practice, an airborne laser scanner acquires a 3D point cloud of the road surface and records the UAV's position and attitude information for each scan point. The airborne laser scanner is mounted beneath the UAV, with its scanning plane at a 45-degree angle to the UAV's transverse axis. The scanner acquires the 3D point cloud at a scanning frequency of 10 Hz. In each scanning cycle, the scanner emits a laser pulse, which is reflected from the road surface and received by the scanner's receiver. The slant distance between the scanner and the scanned point on the road surface is calculated based on the time difference between the emission and reception times of the laser pulse. The scanner also records the emission angle of the laser pulse, including both the vertical and horizontal emission angles within the scanner's coordinate system. The UAV's position and attitude information for each scan point is provided by the UAV flight control system. The UAV's position information is the 3D coordinates in the world coordinate system output by the satellite positioning module, and the UAV's attitude information is the attitude quaternion output by the inertial measurement unit. At the beginning of each scanning cycle, the UAV flight control system records the UAV's position and attitude information. It then timestamps the UAV's position and attitude information with the scan point data within the current scanning cycle to form a scan point data record. The scan point data record includes the scanning time, slant range, vertical launch angle, horizontal launch angle, the UAV's three-dimensional coordinates in the world coordinate system, and the UAV's attitude quaternion.

[0058] Transform the 3D point cloud into the world coordinate system to obtain the road surface coordinate sequence. (The rest of the text appears to be a series of numbers and symbols, possibly related to a road surface coordinate sequence.) Taking the first scan point as an example, the... The slant distance in the data record of each scan point is denoted as The vertical launch angle is denoted as The horizontal launch angle is denoted as , No. The three-dimensional coordinates of the UAV corresponding to each scan point in the world coordinate system are plotted. , No. The quaternion of the UAV attitude corresponding to each scan point is denoted as . Calculate the first [missing information] based on the vertical and horizontal emission angles. Local coordinates of the i-th scan point in the coordinate system of the airborne laser scanner: The first coordinate component of each scan point in the airborne laser scanner coordinate system is equal to the slant distance multiplied by the cosine of the vertical emission angle, and then multiplied by the cosine of the horizontal emission angle. The second coordinate component of each scan point in the airborne laser scanner coordinate system is equal to the slant distance multiplied by the cosine of the vertical emission angle, and then multiplied by the sine of the horizontal emission angle. The third coordinate component of each scan point in the airborne laser scanner coordinate system is equal to the slant distance multiplied by the sine of the vertical emission angle. According to the... The quaternion of the UAV attitude corresponding to the scanning point, the rotation matrix from the computer-mounted laser scanner coordinate system to the world coordinate system, and then the quaternion of the scanning point. Multiplying the local coordinates of the i-th scan point in the coordinate system of the airborne laser scanner with the rotation matrix yields the i-th scan point. The offset vector of the nth scan point relative to the UAV. The three-dimensional coordinates of the UAV corresponding to the first scan point in the world coordinate system, plus the first... The offset vector of the i-th scan point relative to the UAV is obtained. The three-dimensional coordinates of each scan point in the world coordinate system are obtained. The above transformation is performed on each scan point in the three-dimensional point cloud to obtain the road surface coordinate sequence. Each element in the road surface coordinate sequence contains the first horizontal coordinate, the second horizontal coordinate, and the vertical coordinate of a scan point in the world coordinate system.

[0059] The road surface coordinate sequence is overlaid with pre-stored initial road surface reference data to calculate the deformation displacement at each location. The pre-stored initial road surface reference data is a road surface elevation model obtained before road construction through ground laser scanning or total station measurements. The road surface elevation model is stored in a two-dimensional regular grid format, with each grid cell having a side length of 0.05 meters. Each grid cell records a reference vertical coordinate. For the first... Each scan point is marked with its first horizontal coordinate in the world coordinate system. The second horizontal coordinate is marked Vertical coordinates are marked Find the horizontal coordinates in the pre-stored initial road surface reference data. The nearest grid cell, if its horizontal coordinate is... If the distance between them is less than half the side length of the grid, i.e., 0.025 meters, then the reference vertical coordinate in that grid is used as the reference vertical coordinate of the corresponding position in the initial road surface reference data, denoted as . If no grid cell meets the distance requirement, bilinear interpolation is used, utilizing the bounding grid in the initial road surface reference data. The reference vertical coordinates at the corresponding position are calculated using the reference vertical coordinates of the four nearest grid cells. The formula for calculating the deformation displacement at each scanning point is:

[0060]

[0061] in, Indicates the first The deformation displacement at the nth scan point, in meters. A deformation displacement less than zero indicates that the nth scan point is the nth scan point. At the nth scanning point, the road surface subsides relative to the initial road surface; a deformation displacement greater than zero indicates that the nth scanning point... The road surface at each scanning point bulges relative to the initial road surface; Indicates the first The vertical coordinates of each scan point in the world coordinate system, in meters; This indicates that in the initial surface reference data of the road, the horizontal coordinate is... The corresponding reference vertical coordinates, in meters; Indicates the first The first horizontal coordinate of each scan point in the world coordinate system, in meters; Indicates the first The second horizontal coordinate of each scan point in the world coordinate system, in meters.

[0062] The deformation displacement is filled into a two-dimensional grid according to its spatial location to obtain a deformation spatial distribution map. The two-dimensional grid of the deformation spatial distribution map has a side length of 0.1 meters and covers the road surface area of ​​the current inspection area. For each scan point in the road surface coordinate sequence, the first horizontal coordinate of the scan point in the world coordinate system is divided by 0.1 meters and rounded down to obtain the grid column number, and the second horizontal coordinate of the scan point in the world coordinate system is divided by 0.1 meters and rounded down to obtain the grid row number. The deformation displacement at the scan point is added to the corresponding grid. When multiple scan points fall into the same grid of the deformation spatial distribution map, the deformation displacement at the multiple scan points is averaged to obtain the deformation displacement of that grid. For grids that do not fall into any scan point, nearest neighbor filling is performed using the deformation displacement of neighboring grids to obtain a continuous deformation spatial distribution map.

[0063] In the deformation spatial distribution map, spatial locations where the deformation displacement exceeds a set limit are marked as candidate deformation points. The set limit is 0.01 meters, determined based on the allowable vertical displacement deviation of the road surface during road construction. Vertical displacement deviations exceeding 0.01 meters will be deemed unqualified according to the road construction quality acceptance standards. Each grid in the deformation spatial distribution map is traversed; if the deformation displacement of a grid is greater than 0.01 meters, the center point of that grid is marked as a candidate deformation point, and the candidate deformation point retains the deformation displacement of that grid.

[0064] Neighborhood connectivity analysis is performed on candidate deformation points, merging interconnected candidate deformation points into a single connected region. An eight-neighborhood connectivity criterion is used, where the neighborhood of each candidate deformation point includes the eight raster center points adjacent to it in the raster row and column directions. Connectivity labels are created for all candidate deformation points, with each point initially assigned an independent label. All candidate deformation points are traversed, checking if other candidate deformation points exist in their eight neighborhoods. If other candidate deformation points exist and their connectivity labels differ from the current candidate deformation point's label, the two labels are merged into a single label. This process is repeated until no more connectivity labels need merging; at this point, candidate deformation points with the same connectivity label form a connected region.

[0065] Calculate the area and average deformation displacement of each connected region. Connected regions whose area exceeds a preset area limit or whose average deformation displacement exceeds a preset deformation limit are designated as abnormal deformation regions. The area of ​​a connected region is calculated as follows: count the number of candidate deformation points within the connected region, multiply the number of candidate deformation points by the area of ​​a single grid cell in the deformation spatial distribution map (0.01 square meters) to obtain the area of ​​the connected region. The average deformation displacement of a connected region is calculated as follows: add the deformation displacements of all candidate deformation points within the connected region, and then divide by the number of candidate deformation points within the connected region to obtain the average deformation displacement of the connected region. The preset area limit is 0.5 square meters, which is based on the statistical analysis of the contact area of ​​a single construction machinery tire. Local deformations smaller than 0.5 square meters generally do not affect the overall quality of road construction. The preset deformation limit is 0.015 meters, which is determined based on the severe deformation threshold in the construction quality acceptance standards. When the area of ​​a connected region exceeds 0.5 square meters or the average deformation displacement of a connected region exceeds 0.015 meters, that connected region is designated as an abnormal deformation region.

[0066] In implementation, a detailed inspection task plan is generated for the abnormal deformation areas. For each abnormal deformation area, the boundary polygon of the abnormal deformation area is calculated, and the vertices of the boundary polygon are used as hovering observation points for detailed inspection. The sequence of hovering observation points is arranged according to the boundary order of the abnormal deformation area. For abnormal deformation areas with an area greater than 1 square meter, internal hovering observation points are evenly distributed at 0.5-meter intervals within the abnormal deformation area and added to the hovering observation point sequence. Each hovering observation point in the sequence contains three-dimensional coordinates. The first and second horizontal coordinates of the hovering observation point are taken from the horizontal coordinates of the grid center of the candidate deformation point, and the third coordinate is taken from a height of 3 meters above the abnormal deformation area. Each scanning parameter in the scanning parameter sequence corresponds to a hovering observation point. The scanning parameters include the gimbal pointing angle and the resolution of the airborne laser scanner. For hovering observation points located at the boundary of the abnormal deformation area, the gimbal pointing angle is set to face inwards from the abnormal deformation area with a pitch angle of 60 degrees, and the resolution of the airborne laser scanner is set to 0.02 meters. For hovering observation points located within the abnormal deformation region, the gimbal pointing angle is set vertically downwards, and the resolution of the airborne laser scanner is set to 0.01 meters. Each scanning parameter in the scanning parameter sequence corresponds one-to-one with each hovering observation point in the hovering observation point sequence, forming a detailed inspection task plan. The detailed inspection task plan includes the hovering observation point sequence and the scanning parameter sequence.

[0067] In practice, the detailed inspection task plan is sent to the UAV flight control system for execution. This plan is transmitted via the UAV's data link. After transmission, the flight control system parses the hovering observation point sequence, obtaining the 3D coordinates and arrival order of each point. The system then reads the 3D coordinates of each hovering observation point sequentially. The UAV flies to each observation point in turn, adjusting the gimbal pointing and the resolution of the onboard laser scanner at each point according to the corresponding scanning parameters to perform the inspection. Based on the 3D coordinates of the current hovering observation point, the flight control system controls the UAV to fly from its current position to above the point and maintain hovering. Finally, based on the gimbal pointing angle in the scanning parameters for the current hovering observation point, the flight control system controls the gimbal to rotate, aligning the scanning direction of the onboard laser scanner with the gimbal pointing angle in the scanning parameters. The UAV flight control system adjusts the onboard laser scanner's resolution based on the scanning parameters corresponding to the current hovering observation point, ensuring that the onboard laser scanner performs a routine scan at the hovering observation point according to the resolution specified in the scanning parameters. After completing the scan of the current hovering observation point, the UAV flight control system controls the UAV to fly to the next hovering observation point, until all hovering observation points in the sequence have been scanned.

[0068] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any changes or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application.

Claims

1. A method for intelligent inspection of road construction sites using unmanned aerial vehicles (UAVs), characterized in that, The method includes: Collect real-time location, attitude, and wind field distribution data of the UAV in the current inspection area to generate an inspection status dataset. Based on the wind field distribution data in the inspection status dataset, a three-dimensional wind disturbance potential field is reconstructed by spatial interpolation, and the disturbance gradient features of the three-dimensional wind disturbance potential field are extracted. The disturbance gradient features include the rate of change of disturbance intensity in the horizontal direction, the rate of change of disturbance intensity in the vertical direction, and the disturbance vorticity distribution. Based on the disturbance gradient characteristics, the thrust change of each rotor is calculated, and the thrust change is used to dynamically correct the pre-planned inspection route to generate the corrected inspection track. Control the drone to fly along the corrected inspection trajectory, while collecting deformation monitoring data of the road surface and generating a deformation spatial distribution map; Based on the deformation spatial distribution map, abnormal deformation areas are identified, and a detailed inspection task plan is generated for the abnormal deformation areas. The detailed inspection task plan is sent to the UAV flight control system for execution.

2. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The process involves collecting real-time location information, attitude information, and wind field distribution data of the UAV in the current inspection area to generate an inspection status dataset, including: Multiple wind speed sensors are placed symmetrically on the drone body; The wind field distribution data is collected synchronously using the multiple wind speed sensors at a preset sampling frequency; The real-time location information, the attitude information, and the wind field distribution data are timestamped according to the sampling time. The aligned data is stored in a circular cache to form the inspection status dataset.

3. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The reconstruction of the three-dimensional wind disturbance potential field through spatial interpolation includes: Divide the current inspection area into multiple three-dimensional voxel grids; For each three-dimensional voxel grid, wind field distribution data located inside and near the grid are collected, and the wind disturbance vector at the center point of the grid is calculated using an inverse distance weighted interpolation algorithm. The wind disturbance vectors of all three-dimensional voxel meshes are collected to form the three-dimensional wind disturbance potential field.

4. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The extraction of the perturbation gradient features of the three-dimensional wind disturbance potential field includes: Multiple spatial sampling points are selected in the three-dimensional wind disturbance potential field; For each spatial sampling point, the wind disturbance vector difference between that point and its horizontally adjacent sampling points is calculated to obtain the rate of change of the horizontal disturbance intensity. Calculate the wind disturbance vector difference between this point and the adjacent sampling points in the vertical direction to obtain the rate of change of the vertical disturbance intensity. Calculate the wind disturbance vector rotation in the neighborhood of this point to obtain the disturbance vorticity distribution; The disturbance intensity variation rate in the horizontal direction, the disturbance intensity variation rate in the vertical direction, and the disturbance vorticity distribution are combined to form the disturbance gradient feature.

5. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The calculation of the thrust change of each rotor based on the disturbance gradient characteristics includes: Based on the rate of change of the horizontal disturbance intensity, calculate the additional thrust correction values ​​of the UAV along the horizontal and longitudinal directions of the flight path in the horizontal plane; Based on the rate of change of the vertical disturbance intensity, calculate the additional thrust correction value of the UAV in the vertical direction; Based on the disturbance vortex distribution, calculate the yaw moment correction value of the UAV; The additional thrust correction value and the yaw moment correction value are mapped to the thrust change of each rotor according to the arrangement position of each rotor.

6. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The process of dynamically correcting the pre-planned inspection route using the thrust change to generate a corrected inspection track includes: The pre-planned inspection route is discretized into multiple waypoints; For each waypoint, the desired position offset of the waypoint is adjusted according to the thrust change to obtain the corrected waypoint; All the corrected waypoints are reconnected in the order of the original route to form the corrected inspection track.

7. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The control of the UAV to fly along the corrected inspection trajectory includes: Real-time acquisition of angular velocity and acceleration data output by the inertial measurement unit of the UAV; The angular velocity and acceleration data are compared with the expected position on the corrected inspection track to calculate the current flight error; Based on the current flight error, the rotational speed of each rotor is adjusted using proportional-integral-derivative control to enable the UAV to track the corrected inspection trajectory.

8. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, The process of collecting deformation monitoring data from the road surface and generating a spatial distribution map of the deformation includes: The airborne laser scanner acquires a 3D point cloud of the road surface and records the position and attitude information of the UAV corresponding to each scan point. The three-dimensional point cloud is transformed into the world coordinate system to obtain the road surface coordinate sequence; The road surface coordinate sequence is overlaid with the pre-stored initial road surface reference data to calculate the deformation displacement at each location; The deformation displacement is filled into a two-dimensional grid according to its spatial location to obtain the deformation spatial distribution map.

9. The intelligent inspection method for road construction sites using unmanned aerial vehicles (UAVs) according to claim 1, characterized in that, The step of identifying abnormal deformation regions based on the deformation spatial distribution map includes: In the deformation spatial distribution map, spatial locations where the deformation displacement is greater than a set limit are marked as candidate deformation points; Perform neighborhood connectivity analysis on the candidate deformation points, and merge the interconnected candidate deformation points into a connected region; Calculate the area and average deformation displacement of each connected region, and designate the connected regions whose area exceeds a preset area limit or whose average deformation displacement exceeds a preset deformation limit as the abnormal deformation regions.

10. The intelligent unmanned aerial vehicle (UAV) inspection method for road construction sites according to claim 1, characterized in that, Sending the detailed inspection task plan to the UAV flight control system for execution includes: The detailed inspection task plan includes a sequence of hovering observation points and a sequence of scanning parameters; After transmission, the UAV flight control system analyzes the hovering observation point sequence to obtain the three-dimensional coordinates and arrival order of each hovering observation point; The drone flies sequentially to the hovering observation points, and at each hovering observation point, it adjusts the gimbal pointing and the resolution of the airborne laser scanner according to the corresponding scanning parameters to perform inspection.

Images