UAV swarm scanning method for autonomous detection of hidden cracks in highways

Abstract

This invention relates to the field of intelligent detection technology, specifically to a UAV swarm scanning method for autonomous detection of hidden cracks in highways. The method includes the following steps: based on the scanning task blocks, identifying abrupt changes in trajectory curvature and abrupt changes in attitude angles; extracting flight direction and attitude angle change curves; analyzing spatial pointing consistency; filtering blocks deviating from the flight path; analyzing image coverage density and flight direction overlap; filtering and matching abnormal blocks; evaluating elevation change trends and deviation degrees; and outputting a dynamic replanning instruction set for the UAV swarm. This invention introduces a dynamic flight trajectory and attitude angle recognition mechanism to ensure high accuracy in scanning tasks even in complex environments. By analyzing changes in flight direction vectors and attitude angles, it dynamically adjusts the flight paths between task blocks, ensuring uniform image coverage, improving scanning accuracy and completeness, and enabling real-time evaluation of swarm scanning collaborative performance, significantly improving task execution efficiency and accuracy.

Description

Technical Field

[0001] This invention relates to the field of intelligent detection technology, and in particular to a drone swarm scanning method for autonomous detection of hidden cracks in highways. Background Technology

[0002] The field of intelligent detection technology involves technologies related to the automated perception, identification, and judgment of the state of physical objects. It mainly includes core aspects such as sensor information acquisition, target feature extraction, state recognition and interpretation, and multi-source data collaborative processing. It is geared towards scenarios such as transportation infrastructure, industrial equipment, and public safety. By deploying mobile or fixed detection carriers and combining visible light imaging, infrared imaging, laser ranging, attitude positioning, and path planning, it systematically detects and analyzes the surface morphological structure changes and potential defects of the detected objects, forming an overall technical system based on automated perception and aimed at intelligent judgment.

[0003] The traditional UAV swarm scanning method for autonomous detection of invisible cracks in highways involves multiple UAVs flying collaboratively along a preset route to acquire continuous images and location information of the road surface using onboard visible light cameras, infrared imaging devices, and positioning components. The scanning area is then divided into segments according to the route, with each UAV acquiring images of its corresponding road segment. After scanning, the collected image data is summarized for manual or semi-automatic analysis of surface and near-surface cracks. This method relies on fixed altitude, fixed shooting angle, and a pre-set scanning sequence to complete the detection of invisible cracks in highways.

[0004] Existing UAV swarm scanning methods rely on fixed flight altitudes, shooting angles, and preset scanning sequences, limiting their adaptability to complex terrain and unexpected anomalies. In practice, fixed altitudes and angles can easily lead to uneven image coverage or omissions in some areas, especially under the influence of undulating ground or complex obstacles, causing deviations in flight trajectories and imaging attitudes, affecting image continuity and quality. Furthermore, the lack of a real-time dynamic scheduling mechanism between scanning task blocks can easily lead to low flight task execution efficiency, failing to adjust task paths according to real-time flight status, thus affecting detection accuracy and execution speed. These limitations make it difficult for existing methods to achieve optimal performance in large-scale highway hidden crack detection, and they may even fail to complete tasks efficiently in certain specific environments. Summary of the Invention

[0005] The purpose of this invention is to overcome the shortcomings of existing technologies and propose a drone swarm scanning method for autonomous detection of hidden cracks in highways.

[0006] To achieve the above objectives, the present invention adopts the following technical solution: a drone swarm scanning method for autonomous detection of hidden cracks in highways, comprising the following steps:

[0007] S1: Based on the scanning task blocks divided by the highway segment, obtain the flight trajectory and imaging attitude angle sequence of the UAV within the block, identify the trajectory curvature change point and attitude angle jump interval in the spatial coordinate system, and generate a flight anomaly status identification dataset by combining the block number and geofence boundary.

[0008] S2: Based on the flight anomaly state recognition dataset, extract the flight direction vector and imaging attitude angle change curve of the scan task block, analyze the spatial pointing consistency of the two at the block boundary, filter blocks with continuous boundaries and flight trajectories deviating from the preset route, and form the flight behavior anomaly area recognition result.

[0009] S3: Based on the results of the abnormal flight behavior region identification, extract the image coverage density matrix and heading overlap rate features within the block, analyze the image coverage uniformity and heading overlap stability, screen and match abnormal blocks, and obtain the cluster scanning collaborative performance evaluation data table.

[0010] S4: Based on the cluster scanning collaborative efficiency evaluation data table, analyze the road surface elevation change trend corresponding to the block, evaluate the degree of deviation from the original digital elevation model, mark the scanning risk level of the block according to the deviation range, and output the UAV scanning task intervention scheduling table.

[0011] As a further aspect of the present invention, the flight anomaly state identification dataset includes flight trajectory curvature change block numbers, imaging attitude angle change points, block geofence coordinates, and attitude synchronization markers on the time axis. The flight behavior anomaly area identification results include flight trajectory deviation block markers, block boundary continuity units, and block intersection pointing consistency blocks. The cluster scanning collaborative performance evaluation data table includes sparse continuous coverage blocks, heading overlap fluctuation areas, coverage void overlap areas, and collaborative failure block numbers. The UAV scanning task intervention scheduling table includes risk level labels, image missing ratio values, elevation adaptation deviation indicators, and scan integrity offset levels.

[0012] As a further aspect of the present invention, the steps for obtaining the flight anomaly identification dataset are as follows:

[0013] S111: Based on the scanning task blocks divided by the highway segment, extract the flight trajectory and imaging attitude angle change curve of the UAV within the block, perform spatial projection alignment on the two types of data within the same block, and obtain the flight trajectory and attitude synchronization trend sequence.

[0014] S112: Based on the flight trajectory and attitude synchronization trend sequence, identify the curvature change point and the attitude angle change curve jump point in the flight trajectory, perform time window matching on the two types of points, extract the time period when the curvature change exceeds the set threshold and the attitude angle jump occurs synchronously, and generate a set of high-frequency abnormal flight time periods.

[0015] S113: For the set of high-frequency abnormal flight time periods, associate the corresponding block number with the geofence boundary information, extract the location of the block where the abnormal flight occurred, and generate a flight abnormality status identification dataset.

[0016] As a further aspect of the present invention, the steps for obtaining the abnormal flight behavior region identification results are specifically as follows:

[0017] S211: Based on the flight anomaly identification dataset, extract the flight direction vector and imaging attitude angle change curve of the scan task block, extract the spatial projection line of the two at the block boundary, identify the number and distribution density of boundary points where the vector angle on the boundary line is less than the preset tolerance, and obtain the block boundary pointing consistency map.

[0018] S212: Based on the block boundary orientation consistency map, filter the boundary areas where the boundary point density is higher than the benchmark density, compare the block boundaries of the overall digital map of the highway, identify continuous and dense boundary zones belonging to the same scanning task, and obtain the block trajectory deviation zone of the scanning task.

[0019] S213: Invoke the scanning task block trajectory deviation zone, perform integrated analysis on the block boundary pointing consistency, flight trajectory curvature dispersion, image coverage density balance and heading overlap delay, identify abnormal response intensity values, match the block number according to the intensity value layer, and form the flight behavior abnormal area identification result.

[0020] As a further aspect of the present invention, the steps for obtaining the cluster scanning collaborative performance evaluation data table are as follows:

[0021] S311: Based on the flight behavior abnormal area identification results, extract the image coverage density matrix and heading overlap rate features of the numbered blocks in the layer, perform spatial grid alignment on the data in the blocks, identify the coverage density fluctuation value and heading overlap rate offset, and obtain the local scan abnormal response feature set.

[0022] S312: Based on the local scan anomaly response feature set, the coverage uniformity and heading overlap stability within the block are jointly determined using the following formula:

[0023] ;

[0024] Identify coverage and overlap coupling indices, filter blocks in the layer whose coupling indices are lower than a preset cooperation threshold, and establish a spatial distribution map of scan cooperation failure.

[0025] in, Represents the coverage and overlap coupling index. Representing the The coverage uniformity error of each block Representing the The heading stability error of each block The area representing the block. The total number of blocks;

[0026] S313: Call the aforementioned scan collaboration failure spatial distribution map, aggregate the blocks in the coupling index layer that are lower than the collaboration identification benchmark, mark the block codes and geographical coordinates corresponding to the continuous failure areas, and obtain the cluster scan collaboration performance evaluation data table.

[0027] As a further aspect of the present invention, the step of obtaining the UAV scanning task intervention scheduling table specifically includes:

[0028] S411: Based on the location number of the cluster scanning collaborative efficiency evaluation data table, extract the road surface elevation change curve of the specified number block, perform spatial grid unified processing, identify the elevation change gradient per unit distance, and obtain the block elevation anomaly gradient set.

[0029] S412: Based on the block elevation anomaly gradient set, retrieve the reference elevation curve in the original digital elevation model, compare the current elevation gradient sequence with the reference curve, identify the block elevation adaptation deviation level, extract and mark blocks whose deviation level exceeds the tolerance limit, and obtain the block set of elevation mismatch sudden increase blocks.

[0030] S413: Based on the set of blocks with sudden increases in elevation mismatch, bind the deviation level value of each block to the location number in the digital highway map, sort them according to risk level, and output the UAV scanning task intervention scheduling table.

[0031] As a further aspect of the present invention, the reference elevation curve refers to a standard elevation sequence extracted from the original digital elevation model that is consistent with the spatial range of the location number, and interpolated and aligned according to the same sampling interval as the elevation change gradient per unit distance to form a reference elevation curve.

[0032] The block elevation adaptation deviation level refers to the point-by-point calculation of the difference between the elevation change gradient per unit distance and the corresponding gradient of the reference elevation curve, and the division into multiple discrete levels based on the magnitude of the difference, where each level corresponds to a fixed numerical range.

[0033] As a further aspect of the present invention, the method further includes step S5:

[0034] S5: Call the UAV scanning task intervention scheduling table, identify the corresponding number of the scanning task block in the highway functional zoning map, extract the list of backup flight paths, compare the path availability with the terrain adaptation priority, filter the block numbers that need to be reassigned for flight tasks, and output the UAV cluster dynamic replanning instruction set.

[0035] The UAV swarm dynamic replanning instruction set includes reassignment of target block numbers, flight altitude adjustment parameters, heading overlap correction items, and path switching trigger types.

[0036] As a further aspect of the present invention, the step of obtaining the dynamic replanning instruction set of the UAV swarm is specifically as follows:

[0037] S511: Call the UAV scanning task intervention scheduling table, extract the number of the scanning task block in the highway functional zoning map, spatially map the block risk level value with the geofence boundary, identify the block information corresponding to the terrain adaptation level, and generate a scanning task risk distribution map.

[0038] S512: Based on the risk distribution map of the scanning task, extract the backup flight path number and path adaptation level, match the block risk level with the path adaptation level, identify the block number with insufficient path coverage, and obtain the scanning task path adaptation mismatch list.

[0039] S513: Based on the scan task path adaptation disconnect list, extract the key block numbers that need to be used for the backup path according to the level number in the terrain adaptation priority, output the adjustment control parameters linked with the original flight path in sequence, and output the UAV swarm dynamic replanning instruction set.

[0040] Compared with the prior art, the advantages and positive effects of the present invention are as follows:

[0041] This invention effectively solves the adaptation problem of existing technologies in complex terrain and sudden situations by introducing a dynamic flight trajectory and attitude angle recognition mechanism. It can identify flight trajectory anomalies in real time, ensuring high accuracy of scanning tasks even in complex terrain or changing environments. By analyzing the flight direction vector and attitude angle change curves, the flight paths between task blocks are dynamically adjusted to ensure more uniform image coverage, improving scanning accuracy and completeness. Simultaneously, it can evaluate the collaborative efficiency of swarm scanning in real time and assess scanning risks based on ground elevation change trends, dynamically adjusting flight tasks. This avoids the inefficiencies and path mismatch problems of traditional methods, significantly improving the execution efficiency and accuracy of UAV swarm scanning tasks. Attached Figure Description

[0042] Figure 1 This is a schematic diagram of the main steps of the present invention;

[0043] Figure 2 This is a flowchart illustrating the acquisition of the flight anomaly state identification dataset in this invention.

[0044] Figure 3 This is a flowchart illustrating the process of obtaining the results of identifying abnormal flight behavior regions in this invention.

[0045] Figure 4 This is a flowchart illustrating the process of obtaining the cluster scanning collaborative performance evaluation data table in this invention.

[0046] Figure 5 This is a flowchart illustrating the process of obtaining the UAV scanning task intervention scheduling table in this invention.

[0047] Figure 6 This is a flowchart illustrating the process of obtaining the dynamic replanning instruction set for UAV swarms in this invention. Detailed Implementation

[0048] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0049] In the description of this invention, it should be understood that the terms "length," "width," "upper," "lower," "front," "rear," "left," "right," "vertical," "horizontal," "top," "bottom," "inner," and "outer," etc., indicating orientation or positional relationships, are based on the orientation or positional relationships shown in the accompanying drawings and are only for the convenience of describing the invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation, and therefore should not be construed as a limitation of the invention. Furthermore, in the description of this invention, "a plurality of" means two or more, unless otherwise explicitly specified.

[0050] Example 1: Please refer to Figure 1 This invention provides a technical solution for an autonomous detection method of hidden cracks in highways using a drone swarm scanning system, comprising the following steps:

[0051] S1: Based on the scanning task blocks divided by the highway segment, obtain the flight trajectory and imaging attitude angle sequence of the UAV within the block, identify the trajectory curvature change point and attitude angle jump interval in the spatial coordinate system, and generate a flight anomaly status identification dataset by combining the block number and geofence boundary.

[0052] S2: Based on the flight anomaly identification dataset, extract the flight direction vector and imaging attitude angle change curve of the scan task block, analyze the spatial pointing consistency of the two at the block boundary, filter blocks with continuous boundaries and flight trajectories deviating from the preset route, and form the flight behavior anomaly area identification result.

[0053] S3: Based on the results of abnormal flight behavior region identification, extract the image coverage density matrix and heading overlap rate features within the block, analyze the image coverage uniformity and heading overlap stability, screen and match abnormal blocks, and obtain the cluster scanning collaborative performance evaluation data table.

[0054] S4: Based on the cluster scanning collaborative efficiency evaluation data table, analyze the road surface elevation change trend corresponding to the block, evaluate the degree of deviation from the original digital elevation model, mark the scanning risk level of the block according to the deviation range, and output the UAV scanning task intervention scheduling table.

[0055] S5: Call the UAV scanning task intervention scheduling table, identify the corresponding number of the scanning task block in the highway functional zoning map, extract the list of alternative flight paths, compare the path availability with the terrain adaptation priority, filter the block numbers that need to be reassigned for flight tasks, and output the UAV cluster dynamic replanning instruction set.

[0056] The flight anomaly identification dataset includes the block number of flight trajectory curvature change, the location of imaging attitude angle change, the block geofence coordinates, and the attitude synchronization marker on the time axis. The flight behavior anomaly area identification results include the flight trajectory deviation block marker, the block boundary continuity unit, and the block intersection pointing consistency block. The swarm scanning collaborative performance evaluation data table includes the sparse continuous coverage block, the heading overlap fluctuation area, the coverage hole overlap area, and the collaborative failure block number. The UAV scanning task intervention scheduling table includes the risk level label, the image missing ratio value, the elevation adaptation deviation index, and the scan integrity offset level. The UAV swarm dynamic replanning instruction set includes the reassignment target block number, flight altitude adjustment parameters, heading overlap correction item, and path switching trigger type.

[0057] Please see Figure 2 The specific steps for obtaining the flight anomaly identification dataset are as follows:

[0058] S111: Based on the scanning task blocks divided by the highway segment, extract the flight trajectory and imaging attitude angle change curve of the UAV within the block, perform spatial projection alignment on the two types of data within the same block, and obtain the flight trajectory and attitude synchronization trend sequence.

[0059] Based on the scanning task blocks divided by the highway section, the time-series position coordinate data recorded by the airborne high-precision positioning unit, covering longitude, latitude, and ellipsoidal elevation information, is retrieved. This data is then converted into local Cartesian coordinates with the highway centerline as the reference axis using the Gauss-Kruger projection method, constructing a flight trajectory point set containing three-dimensional spatial coordinates. Simultaneously, the raw inertial navigation data recorded by the inertial measurement unit is retrieved, and the pitch, roll, and yaw angle data strictly corresponding to the position coordinate timestamps are analyzed using the Kalman filter algorithm to construct a three-dimensional imaging attitude angle sequence. Subsequently, B-spline curve fitting is performed on the extracted flight trajectory point set, calculating the tangent direction vector and radius of curvature of each sampling point on the trajectory line, and then the three-dimensional imaging attitude angles are mapped to the waterline. The planar and vertical planes are decomposed and projected. A cubic spline interpolation algorithm is used to resample the sampling frequency of the flight trajectory data and the imaging attitude angle data to 10Hz to ensure the consistency of the time reference. On this basis, the two types of data within the same block are spatially projected and aligned to establish a unified spatial reference system. The rate of change of tangent curvature of the flight trajectory per unit time and the rate of change of angular velocity of the imaging attitude angle within the corresponding time window are calculated respectively. Cross-correlation analysis is performed on the two normalized rate of change curves, and the Pearson correlation coefficients of the two sets of sequences are calculated at zero delay and different time delays. The maximum value of the correlation coefficient is selected as the index to characterize the degree of coordination between the two changes, and the synchronization trend sequence of flight trajectory and attitude is obtained.

[0060] S112: Based on the flight trajectory and attitude synchronization trend sequence, identify the curvature change points and attitude angle change curve jump points in the flight trajectory, perform time window matching on the two types of points, extract the time period when the curvature change exceeds the set threshold and the attitude angle jump occurs synchronously, and generate a set of high-frequency abnormal flight time periods.

[0061] Based on the synchronization trend sequence of flight trajectory and attitude, a first-order difference operation is performed on the preprocessed flight trajectory curvature sequence to obtain the curvature change gradient sequence. A curvature change gradient judgment threshold of 0.05 per meter is set. The sequence is traversed to filter out time points where the absolute value of the gradient exceeds this threshold as curvature abrupt change points. Simultaneously, the second derivative is calculated on the imaging attitude angle sequence to obtain angular acceleration. An angular acceleration judgment threshold of 15 degrees per square second is set. Time points exceeding this threshold are filtered out as attitude angle jump points. Time window matching is performed on the two types of points, with a time matching tolerance window of 0.5 seconds. If an attitude angle jump point exists within 0.5 seconds before and after each curvature abrupt change point, the two are merged and the moment is marked as a joint anomaly. If no such point exists, no merging is performed. If the time interval between two adjacent anomaly intervals of the joint anomaly point and its two preceding and following 2-second intervals is less than 1 second, interval fusion is performed to form a continuous time period. The time period in which the curvature abrupt change exceeds the set threshold and the attitude angle jump occurs synchronously is extracted. The number of anomalous sampling points in each time period is counted. Time periods with more than 10 sampling points are retained to generate a high-frequency anomalous flight time period set.

[0062] S113: For high-frequency abnormal flight time periods, associate the corresponding block number with geofence boundary information, extract the location of the block where abnormal flight occurred, and generate a flight anomaly status identification dataset.

[0063] For a set of high-frequency abnormal flight time periods, the system iterates through each start and end time in the time period set, retrieves all spatial coordinate points corresponding to that time period from the original flight trajectory database, calls a vector polygon database storing highway scanning task block division information, and uses a ray casting algorithm to determine the polygon range of each spatial coordinate point to which the block belongs, thus determining the specific block number where the anomaly occurred. At the same time, it reads the geofence boundary coordinate sequence corresponding to the block, constructs a boundary vector model, and calculates the shortest Euclidean distance between each point in the abnormal trajectory point set and the geofence boundary line segment. If the calculated distance is less than the preset safety buffer distance (set to 2 meters), the anomaly is marked as having a risk of crossing the boundary. The block number, the start and end time of the abnormal time period, the corresponding trajectory coordinate set, and the risk of crossing the boundary are combined into a structured data entry. The location of the block where the abnormal flight occurred is extracted, and all entries are written into a relational database or time series database in chronological order to generate a flight anomaly status identification dataset.

[0064] Please see Figure 3 The specific steps for obtaining the results of abnormal flight behavior area identification are as follows:

[0065] S211: Based on the flight anomaly identification dataset, extract the flight direction vector and imaging attitude angle change curve of the scan task block, extract the spatial projection line of the two at the block boundary, identify the number and distribution density of boundary points where the vector angle on the boundary line is less than the preset tolerance, and obtain the block boundary pointing consistency map.

[0066] Based on the flight anomaly identification dataset, the block numbers marked as anomalies are read from the dataset. For two adjacent blocks (defined as preceding block A and following block B), the tangent vectors of the UAV's terminal flight trajectory when it flies out of block A and the tangent vectors of its front flight trajectory when it flies into block B are extracted respectively. Simultaneously, the instantaneous normal vectors of the corresponding imaging attitude angles at the boundaries are extracted. The spatial projection lines of these two vectors at the block boundary are extracted. The three-dimensional flight vectors are projected onto a two-dimensional horizontal plane. The dot product of the terminal vector of block A and the front vector of block B is calculated, and the inverse cosine function is used... The angle between two vectors on the boundary line is calculated. The preset tolerance for the vector angle is set to 10 degrees. The boundary positions in all abnormal records are traversed, and the number of boundary points with an angle less than 10 degrees is counted. Then, the kernel density estimation method is used, and the search radius is set to 5 meters to calculate the distribution of boundary points per unit length on the boundary line. The number and distribution density of boundary points with vector angles less than the preset tolerance are identified. The density values ​​are mapped to gray values ​​from 0 to 255, and a visual grayscale image reflecting the smoothness of the connection between blocks is drawn to obtain the block boundary orientation consistency map.

[0067] S212: Based on the block boundary orientation consistency map, filter the boundary areas where the density of boundary points is higher than the baseline density, compare the block boundaries of the overall digital map of the highway, identify continuous and dense boundary zones belonging to the same scanning task, and obtain the deviation of the scanning task block trajectory from the zoning.

[0068] Based on the block boundary alignment consistency map, a baseline density value of 3 boundary points per 10 meters is set. The density data in the map is binarized and thresholded to extract the connected components of high-density areas. The geometric center coordinates and spatial coverage of these areas are obtained. The block boundaries of the overall digital road map are compared with those of the road. The extracted high-density area polygons are spatially superimposed with the scanning task block vector boundaries in the road digital map. The area ratio of the overlapping parts is calculated, and discrete noise areas with an overlap rate of less than 50% are removed. Continuous and dense boundary zones belonging to the same scanning task are identified. Based on the spatial distribution characteristics of the dense zones, the block areas where they are located are marked as potential flight trajectory deviation zones. The relative position coordinates of these areas within the block (i.e., from the starting mileage marker to the ending mileage marker) are recorded to obtain the trajectory deviation zone of the scanning task block.

[0069] S213: Call the scan task block trajectory deviation zone, perform integrated analysis on the block boundary pointing consistency, flight trajectory curvature dispersion, image coverage density balance and heading overlap delay, identify abnormal response intensity values, match the block number according to the intensity value layer, and form the flight behavior abnormal area identification result;

[0070] The formula for calculating the curvature dispersion of the flight trajectory is:

[0071] ;

[0072] in, This represents the dispersion of the flight trajectory curvature. Represents the total number of trajectory points. Representing adjacent trajectory points The change in heading angle between them Represents the standard perspective. This represents the straight-line distance between points on the trajectory. Representing adjacent trajectory points The time interval between;

[0073] Call the scan task block trajectory deviation zone, obtain the target trajectory segment to be analyzed from the deviation zone, and set the standard angle. The value is 5 degrees (0.087 radians). This angle represents the permissible heading deviation during normal straight-line cruise. The total number of trajectory points within the trajectory segment is extracted. For 50 points, calculate adjacent trajectory points. and The straight-line distance between The average distance was measured to be 2 meters, and the time interval between adjacent trajectory points was extracted. The average time is 0.2 seconds. The change in heading angle between adjacent trajectory points is calculated point by point. Assuming the change is 0.1 radians at the first point, 0.15 radians at the second point, and so on, the formula is as follows: In this formula, Representing the The point and the first The change in heading angle between points For standard perspective normalization factor, It represents the spatial displacement distance. For time increments;

[0074] This formula introduces a standard angle. The course change was dimensionless, eliminating the differences in angle sensitivity among different tasks, while a velocity term was introduced. As dynamic weights, the minute angular jitter during high-speed flight affects the dispersion. The contribution increases nonlinearly, thus capturing unstable states at high speeds more sensitively;

[0075] Substitute data for calculation: Calculate the weighted deviation term for a single point, taking point 1 as an example. Speed ​​item The product is 11.5;

[0076] Taking point 2 as an example, Speed ​​item The product is 17.2;

[0077] Assuming the average product of the remaining 48 points is 5, then the sum is... Final calculation ;

[0078] Calculated The value 5.37 represents the curvature dispersion of the trajectory segment. This value is compared with the preset dispersion threshold of 3.0. Since 5.37 is greater than 3.0, the region is determined to be abnormal. Combining the normalized scores of image coverage density uniformity and heading overlap delay, a comprehensive abnormal response intensity value is calculated by weighted summation. If the intensity value exceeds 0.7 (out of 1.0), the abnormal response intensity value is identified. Based on the intensity value layer, the block number is matched to form the flight behavior abnormal region identification result.

[0079] Please see Figure 4 The specific steps for obtaining the cluster scanning collaborative performance evaluation data table are as follows:

[0080] S311: Based on the results of abnormal flight behavior region identification, extract the image coverage density matrix and heading overlap rate features of numbered blocks in the layer, perform spatial grid alignment on the data in the block, identify the coverage density fluctuation value and heading overlap rate offset, and obtain the local scan abnormal response feature set.

[0081] Based on the results of abnormal flight behavior area identification, for the identified abnormal blocks, the center point coordinates and ground coverage polygons of all acquired images within them are obtained. The data within the blocks are spatially grid-aligned to establish a regular grid with a resolution of 1m × 1m. All image coverage areas are traversed, and the number of images falling into each grid is counted to construct an image coverage density matrix. The value of each element in the matrix represents the cumulative number of times the location was photographed. The standard deviation of all non-zero elements in the matrix is ​​calculated as the coverage density fluctuation value. The ratio of the ground overlap length of adjacent images to the total ground length of the images is calculated along the flight heading to obtain the flight heading overlap rate sequence. The average difference between this sequence and the preset overlap rate standard (e.g., 80%) is calculated. The coverage density fluctuation value and the flight heading overlap rate offset are identified. The coverage density fluctuation value (e.g., 1.5) and the overlap rate offset (e.g., -15%) are combined to obtain the local scanning abnormal response feature set.

[0082] S312: Based on the local scan anomaly response feature set, the coverage uniformity and forward overlap stability within the block are jointly determined using the following formula:

[0083] ;

[0084] Identify coverage and overlap coupling indices, filter blocks in the layer whose coupling indices are lower than a preset cooperation threshold, and establish a spatial distribution map of scan cooperation failure.

[0085] in, Represents the coverage and overlap coupling index. Representing the The coverage uniformity error of each block Representing the The heading stability error of each block The area representing the block. The total number of blocks;

[0086] Based on the local scan anomaly response feature set, select the quantity The continuous blocks are treated as independent analysis units, and the analysis unit is targeted at the first... For each block, cover density fluctuation values ​​are extracted from the feature set, and the fluctuation values ​​are converted into a dimensionless cover uniformity index using an inverse normalization mapping algorithm. The mapping range is set to 0 to 1, with a value of 0.2 representing a state of large fluctuations and low uniformity. At the same time, the absolute value of the heading overlap rate offset is extracted and converted into a heading stability index. A value of 0.1 represents a state with large overlap deviation and low stability, defining the geometric area of ​​a single block. The area is 1000 square meters. Assuming the index values ​​of the remaining 9 blocks within the analysis unit are the same as those of the first block... All blocks are completely identical, using the formula: ;

[0087] Representing the The coverage uniformity performance score of each block Representing the The heading stability performance score of each block Represents the projected area of ​​a single block. Representing the total number of blocks involved in the calculation, this formula performs a root mean square operation on the sum of the squares of various performance indicators and incorporates an area parameter. Spatial dimension normalization is performed to calculate the density parameter characterizing the degree of scan quality coordination per unit area. Substituting the set values, the calculation is performed. First, the sum of squares of the single-block performance index is calculated, i.e., the square of 0.2 is added to the square of 0.1, resulting in 0.05. Then, the calculation... The cumulative performance of each block is calculated by multiplying 10 by 0.05, resulting in 0.5. The denominator parameter is then calculated by multiplying the total number of blocks (10) by the area of ​​a single block (1000), resulting in 10000. A division operation is performed to obtain an intermediate quotient of 0.00005. Finally, the square root of the quotient is taken to obtain the final coverage and overlap coupling index. The coefficient is approximately 0.007. The coverage and overlap coupling index is identified, and a preset coordination threshold of 0.01 is set. This threshold is determined based on the unit area efficiency density benchmark required to meet the minimum scan reconstruction quality. The calculated result is compared with the threshold. The value of 0.007 is lower than the preset coordination threshold of 0.01, indicating that the comprehensive scan coordination efficiency per unit area of ​​the region is insufficient and there is a risk of data acquisition quality failure. Blocks in the layer with a coupling index lower than the preset coordination threshold are selected. The selected blocks are marked with attributes in the vector map layer to establish a spatial distribution map of scan coordination failure.

[0088] S313: Call the spatial distribution map of scanning collaborative failure, aggregate the blocks in the coupling index layer that are lower than the collaborative identification benchmark, mark the block codes and geographical coordinates corresponding to the continuous failure areas, and obtain the cluster scanning collaborative performance evaluation data table.

[0089] The algorithm calls upon the spatial distribution map of collaborative scanning failures and uses an eight-neighbor connected component labeling algorithm to traverse the raster cells within the layer. It identifies sets of pixels with the same failure attributes and continuous spatial locations. Low-index blocks with shared boundaries or vertex connections are merged into an independent failed connected domain object. For each aggregated connected domain, it traverses all the original block cells contained within it and extracts the unique code (ID) of each block stored in the database to form a block code list belonging to the connected domain. At the same time, it calculates the geometric attributes of the aggregated region, including the coordinates of the four vertices of the circumscribed rectangle and the latitude and longitude coordinates of the geometric center point calculated based on the polygon moment method, to determine the spatial core location of the failed region. Furthermore, it calculates the arithmetic mean of the coupling index of all blocks within the connected domain as a quantitative indicator to measure the overall failure degree of the region. It constructs structured data entries containing unique identifiers of connected domains, block code sets, center location coordinates, and average coupling indexes. The generated entries are written sequentially to the storage medium to obtain the cluster scanning collaborative performance evaluation data table.

[0090] Please see Figure 5 The specific steps for obtaining the UAV scanning mission intervention schedule are as follows:

[0091] S411: Based on the location number of the cluster scanning collaborative efficiency evaluation data table, extract the road surface elevation change curve of the specified number block, perform spatial grid unified processing, identify the elevation change gradient per unit distance, and obtain the block elevation anomaly gradient set.

[0092] Based on the location number in the cluster scanning collaborative efficiency evaluation data table, the high-precision airborne LiDAR point cloud database is accessed to retrieve all ground echo point data within the spatial range corresponding to that location number. A cloth simulation filtering algorithm is used to remove non-ground interference points such as vegetation, vehicles, and pedestrians, retaining only the true road surface point cloud. A sampling step size of 0.5 meters is set along the centerline of the highway, and the road surface point cloud is longitudinally sliced ​​and Kriging interpolated to generate a continuous and smooth road surface elevation sampling sequence (e.g., 100.00 meters, 100.50 meters, 101.20 meters...). The sequence is then subjected to spatial grid unification processing, and the elevation sequence is mapped to a linear reference strictly aligned with the UAV flight trajectory timestamp. The elevation change gradient per unit distance is identified, and the ratio of the elevation difference to the horizontal distance between adjacent sampling points is calculated using the first-order finite difference method. For example, if the slope value between the previous and next points is calculated to be 0.5, the entire elevation sequence is traversed to obtain the gradient value set point by point, thus obtaining the block elevation anomaly gradient set.

[0093] S412: Based on the block elevation anomaly gradient set, retrieve the reference elevation curve in the original digital elevation model, compare the current elevation gradient sequence with the reference curve, identify the block elevation adaptation deviation level, extract and mark blocks whose deviation level exceeds the tolerance limit, and obtain the block set of elevation mismatch bursts.

[0094] A reference elevation curve is a standard elevation sequence extracted from the original digital elevation model that is consistent with the spatial range of the location number, and then interpolated and aligned according to the sampling interval that is the same as the elevation change gradient per unit distance to form a reference elevation curve.

[0095] Block elevation adaptation deviation level refers to the point-by-point calculation of the difference between the elevation change gradient per unit distance and the corresponding gradient of the reference elevation curve, and the division into multiple discrete levels based on the magnitude of the difference, with each level corresponding to a fixed numerical range.

[0096] Based on the block elevation anomaly gradient set, theoretical design elevation data that is completely consistent with the current block location coordinates are retrieved from the pre-built highway building information model (BIM) or standard design elevation database. A reference gradient sequence is generated according to the same sampling step size and interpolation method as the measured data. The current elevation gradient sequence is compared with the reference curve. The absolute value of the difference between the measured elevation gradient and the theoretical reference gradient is calculated point by point. For example, if the measured gradient is 0.05 and the reference gradient is 0.02, the difference is calculated to be 0.03. The block elevation adaptation deviation level is identified. Based on the calculated difference magnitude, the block is quantitatively graded according to the preset grading standard. The specific grading logic is shown in Table 1.

[0097] Table 1: Classification of Block Elevation Adaptation Deviation Levels

[0098] ;

[0099] Referring to Table 1, if the average gradient difference calculated for a certain block is 0.06, it is classified as a level 3 risk according to Table 1. Blocks whose deviation level exceeds the tolerance limit (set to level 2) are extracted and marked. The rating results of all blocks are traversed, and the unique identification codes of blocks classified as level 3 and level 4 are selected to form a high-risk object set, thus obtaining the set of blocks with sudden increase in elevation mismatch.

[0100] S413: Based on the set of blocks with sudden increases in elevation mismatch, bind the deviation level value of each block to the location number in the digital highway map, sort them according to risk level, and output the UAV scanning task intervention scheduling table;

[0101] Based on the set of blocks with sudden increases in elevation mismatch, an associated data structure containing "unique block identifier - deviation level value - center mileage marker" is constructed. The list of this structure is prioritized using a quicksort algorithm, with the sorting rule set to place severely mismatched blocks with a deviation level of 4 at the front of the list. For blocks with the same deviation level, they are sorted in ascending order according to the value of their center mileage markers to ensure the orderliness of the scheduling response. For the sorted list, a scheduling instruction containing specific location information and risk level is generated for each item. The sorted list is then encapsulated into a standardized data exchange format (such as JSON or XML) and the UAV scanning mission intervention scheduling table is output.

[0102] Please see Figure 6 The specific steps for obtaining the dynamic replanning instruction set for drone swarms are as follows:

[0103] S511: Call the UAV scanning mission intervention schedule table, extract the number of the scanning mission block in the highway functional zoning map, spatially map the block risk level value with the geofence boundary, identify the block information corresponding to the terrain adaptation level, and generate a scanning mission risk distribution map.

[0104] The system invokes the UAV scanning mission intervention schedule table, reads the highway function attribute fields (such as "extra-large bridge section", "long tunnel exit", "sharp bend section") corresponding to high-risk blocks recorded in the schedule table, spatially maps the block risk level values ​​to the geofence boundaries, and in the geographic information processing module, attaches the risk level values ​​(such as 3 and 4) in the schedule table as attribute data to the vector polygon objects of the corresponding blocks. Based on the risk level values, the rendering rules are set to fill level 4 risk areas in red and level 3 risk areas in orange, realizing spatial visualization of risk levels. At the same time, the system reads the geofence coordinate sequence associated with the block, verifies the fence height and range restrictions, identifies the block information corresponding to the terrain adaptation level, and generates a scanning mission risk distribution map.

[0105] S512: Based on the risk distribution map of the scanning mission, extract the backup flight path number and path adaptation level, match the block risk level with the path adaptation level, identify the block number with insufficient path coverage, and obtain the scanning mission path adaptation mismatch list.

[0106] Based on the risk distribution map of the scanning mission, the system accesses the UAV flight plan database and retrieves all planned backup flight route resources (including high-altitude return routes and lateral detour routes) within the airspace. It reads the pre-set adaptation level attributes of each backup path (for example, backup path A is capable of adapting to a level 4 risk environment, while backup path B is only suitable for a level 2 risk environment). The system matches the risk level of the block with the path adaptation level and uses spatial topology analysis to check whether the block location marked as high risk (e.g., level 4) in the map is covered by backup paths with an adaptation level greater than or equal to 4. If a level 4 risk block is detected to be currently only covered by a backup path with an adaptation level of 2, or not covered by any backup path, it is determined that there is a path guarantee gap in the area. The block number with insufficient path coverage is identified, and a list of scanning mission path adaptation mismatches is obtained.

[0107] S513: Based on the scan mission path adaptation disconnect list, according to the level number in the terrain adaptation priority, extract the key block number that needs to be used for the backup path, output the adjustment control parameters linked with the original flight path in sequence, and output the UAV swarm dynamic replanning instruction set.

[0108] Based on the list of disjointed paths in the scanning mission, for each disjointed block recorded in the list, the associated terrain adaptation priority parameters are queried (e.g., bridge sections have a higher adaptation priority than straight road sections). The processing order of replanning is determined according to the priority, and the path adjustment parameters of high-priority blocks are calculated first. The spatial coordinates (longitude, latitude, and elevation) of the optimal transition point from the currently executed main flight path to the target backup path are calculated, and the heading angle adjustment, the rate of change of altitude during climb or descent, and the flight speed correction value are calculated when switching to the backup path. The adjustment control parameters that are smoothly linked with the original flight path are output in sequence. For example, an instruction code containing "at kilometer marker K10+500, turn the heading 30 degrees to the right, climb to 150 meters at 2 meters per second, and switch to backup path B" is generated, and the dynamic replanning instruction set of the UAV swarm is output.

[0109] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments that can be applied to other fields. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A drone swarm scanning method for autonomous detection of hidden cracks in highways, characterized in that, Includes the following steps: S1: Based on the scanning task blocks divided by the highway segment, obtain the flight trajectory and imaging attitude angle sequence of the UAV within the block, identify the trajectory curvature change point and attitude angle jump interval in the spatial coordinate system, and generate a flight anomaly status identification dataset by combining the block number and geofence boundary. S2: Based on the flight anomaly state recognition dataset, extract the flight direction vector and imaging attitude angle change curve of the scan task block, analyze the spatial pointing consistency of the two at the block boundary, filter blocks with continuous boundaries and flight trajectories deviating from the preset route, and form the flight behavior anomaly area recognition result. The specific steps for obtaining the results of the abnormal flight behavior region identification are as follows: S211: Based on the flight anomaly identification dataset, extract the flight direction vector and imaging attitude angle change curve of the scan task block, extract the spatial projection line of the two at the block boundary, identify the number and distribution density of boundary points where the vector angle on the boundary line is less than the preset tolerance, and obtain the block boundary pointing consistency map. S212: Based on the block boundary orientation consistency map, filter the boundary areas where the boundary point density is higher than the benchmark density, compare the block boundaries of the overall digital map of the highway, identify continuous and dense boundary zones belonging to the same scanning task, and obtain the block trajectory deviation zone of the scanning task. S213: Call the scanning task block trajectory deviation zone, perform integrated analysis on the block boundary pointing consistency, flight trajectory curvature dispersion, image coverage density balance and heading overlap delay, identify abnormal response intensity values, match the block number according to the intensity value layer, and form the flight behavior abnormal area identification result; S3: Based on the results of the abnormal flight behavior region identification, extract the image coverage density matrix and heading overlap rate features within the block, analyze the image coverage uniformity and heading overlap stability, screen and match abnormal blocks, and obtain the cluster scanning collaborative performance evaluation data table. The specific steps for obtaining the cluster scanning collaborative performance evaluation data table are as follows: S311: Based on the flight behavior abnormal area identification results, extract the image coverage density matrix and heading overlap rate features of the numbered blocks in the layer, perform spatial grid alignment on the data in the blocks, identify the coverage density fluctuation value and heading overlap rate offset, and obtain the local scan abnormal response feature set. S312: Based on the local scan anomaly response feature set, the coverage uniformity and heading overlap stability within the block are jointly determined using the following formula: ； Identify coverage and overlap coupling indices, filter blocks in the layer whose coupling indices are lower than a preset cooperation threshold, and establish a spatial distribution map of scan cooperation failure. in, Represents the coverage and overlap coupling index. This represents the coverage uniformity error of the k-th block. The denominator represents the heading stability error of the k-th block, E represents the area of ​​the block, and m is the total number of blocks. S313: Call the aforementioned scan collaboration failure spatial distribution map, aggregate the blocks in the coupling index layer that are lower than the collaboration identification benchmark, mark the block codes and geographical coordinates corresponding to the continuous failure areas, and obtain the cluster scan collaboration performance evaluation data table. S4: Based on the cluster scanning collaborative efficiency evaluation data table, analyze the road surface elevation change trend corresponding to the block, evaluate the degree of deviation from the original digital elevation model, mark the scanning risk level of the block according to the deviation range, and output the UAV scanning task intervention scheduling table.

2. The UAV swarm scanning method for autonomous detection of hidden cracks in highways according to claim 1, characterized in that, The flight anomaly identification dataset includes flight trajectory curvature change block numbers, imaging attitude angle change points, block geofence coordinates, and attitude synchronization markers on the time axis. The flight behavior anomaly area identification results include flight trajectory deviation block markers, block boundary continuity units, and block intersection pointing consistency blocks. The cluster scanning collaborative performance evaluation data table includes sparse continuous coverage blocks, heading overlap fluctuation areas, coverage void overlap areas, and collaborative failure block numbers. The UAV scanning task intervention scheduling table includes risk level labels, image missing ratio values, elevation adaptation deviation indicators, and scan integrity offset levels.

3. The UAV swarm scanning method for autonomous detection of hidden cracks in highways according to claim 1, characterized in that, The specific steps for obtaining the flight anomaly identification dataset are as follows: S111: Based on the scanning task blocks divided by the highway segment, extract the flight trajectory and imaging attitude angle change curve of the UAV within the block, perform spatial projection alignment on the two types of data within the same block, and obtain the flight trajectory and attitude synchronization trend sequence. S112: Based on the flight trajectory and attitude synchronization trend sequence, identify the curvature change point and the attitude angle change curve jump point in the flight trajectory, perform time window matching on the two types of points, extract the time period when the curvature change exceeds the set threshold and the attitude angle jump occurs synchronously, and generate a set of high-frequency abnormal flight time periods. S113: For the set of high-frequency abnormal flight time periods, associate the corresponding block number with the geofence boundary information, extract the location of the block where the abnormal flight occurred, and generate flight abnormality status identification data.

4. The UAV swarm scanning method for autonomous detection of hidden cracks in highways according to claim 1, characterized in that, The specific steps for obtaining the UAV scanning task intervention schedule table are as follows: S411: Based on the location number of the cluster scanning collaborative efficiency evaluation data table, extract the road surface elevation change curve of the specified number block, perform spatial grid unified processing, identify the elevation change gradient per unit distance, and obtain the block elevation anomaly gradient set. S412: Based on the block elevation anomaly gradient set, retrieve the reference elevation curve in the original digital elevation model, compare the current elevation gradient sequence with the reference curve, identify the block elevation adaptation deviation level, extract and mark blocks whose deviation level exceeds the tolerance limit, and obtain the block set of elevation mismatch sudden increase blocks. S413: Based on the set of blocks with sudden increases in elevation mismatch, bind the deviation level value of each block to the location number in the digital highway map, sort them according to risk level, and output the UAV scanning task intervention scheduling table.

5. The UAV swarm scanning method for autonomous detection of hidden cracks in highways according to claim 4, characterized in that, The reference elevation curve refers to the standard elevation sequence extracted from the original digital elevation model that is consistent with the spatial range of the location number, and interpolated and aligned according to the sampling interval that is the same as the elevation change gradient per unit distance to form the reference elevation curve. The block elevation adaptation deviation level refers to the point-by-point calculation of the difference between the elevation change gradient per unit distance and the corresponding gradient of the reference elevation curve, and the division into multiple discrete levels based on the magnitude of the difference, where each level corresponds to a fixed numerical range.

6. The UAV swarm scanning method for autonomous detection of hidden cracks in highways according to claim 1, characterized in that, The method also includes step S5: S5: Call the UAV scanning task intervention scheduling table, identify the corresponding number of the scanning task block in the highway functional zoning map, extract the list of backup flight paths, compare the path availability with the terrain adaptation priority, filter the block numbers that need to be reassigned for flight tasks, and output the UAV cluster dynamic replanning instruction set. The UAV swarm dynamic replanning instruction set includes reassignment of target block numbers, flight altitude adjustment parameters, heading overlap correction items, and path switching trigger types.

7. The UAV swarm scanning method for autonomous detection of hidden cracks in highways according to claim 6, characterized in that, The specific steps for obtaining the dynamic replanning instruction set of the UAV swarm are as follows: S511: Call the UAV scanning task intervention scheduling table, extract the number of the scanning task block in the highway functional zoning map, spatially map the block risk level value with the geofence boundary, identify the block information corresponding to the terrain adaptation level, and generate a scanning task risk distribution map. S512: Based on the risk distribution map of the scanning task, extract the backup flight path number and path adaptation level, match the block risk level with the path adaptation level, identify the block number with insufficient path coverage, and obtain the scanning task path adaptation mismatch list. S513: Based on the scan task path adaptation disconnect list, extract the key block numbers that need to be used for the backup path according to the level number in the terrain adaptation priority, output the adjustment control parameters linked with the original flight path in sequence, and output the UAV swarm dynamic replanning instruction set.

Images