Unmanned aerial vehicle autonomous inspection orthoimage generation method

Abstract

The invention discloses a method for generating an autonomous inspection orthoimage of an unmanned aerial vehicle, and relates to the technical field of unmanned aerial vehicle surveying and mapping and autonomous navigation. Global optimization of an air route is realized through a path optimization strategy fused by a genetic algorithm in combination with grid characteristics of an inspection area and image parameter constraints, and candidate waypoints are used as nodes; redundant waypoints are screened through leg smoothness factors, route complexity is reduced, a genetic algorithm takes flight height and waypoint spacing as constraints, a fitness function containing flight distance, turning times and overlapping rate is constructed, a global optimal path is found through population iteration, multi-target requirements are optimized and balanced, and it is ensured that the route meets the image acquisition precision requirement and also meets the requirement of image acquisition. The flight distance can be shortened, the turning frequency is reduced, the cruising ability of the unmanned aerial vehicle is adapted, efficient propelling of the inspection task is guaranteed, meanwhile, waypoint coordinates output through simulation directly adapt to a flight control system, and it is guaranteed that actual flight parameters are consistent with planning parameters.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of unmanned aerial vehicle surveying and mapping and autonomous navigation, in particular to a method for generating orthographic images by autonomous inspection of unmanned aerial vehicles. BACKGROUND

[0002] The unmanned aerial vehicle orthographic image technology has become an indispensable efficient data acquisition means in the fields of power inspection, geological disaster monitoring, smart city modeling and basic surveying and mapping by collecting vertical photographic images with uniform scale and no distortion from the air and through splicing processing. Compared with traditional manual surveying and mapping, this technology has the advantages of high efficiency, low cost, uniform data precision, and the ability to operate in complex and dangerous areas. With the continuous development of unmanned aerial vehicle platforms, high-precision positioning technology and camera sensor technology, the industry application is evolving towards full coverage, high precision and full automation. To achieve high-quality orthographic image results, the first prerequisite is to generate an efficient flight route that can adapt to complex operating environments and meet the geometric quality requirements of subsequent splicing. Therefore, the intelligent generation method of the unmanned aerial vehicle autonomous inspection route is the core key technology to improve the efficiency and quality of the entire operation process, and is also a hot spot of current research and application.

[0003] The existing technology still has significant defects in structure and method when dealing with complex practical application scenarios. First, in terms of terrain adaptability, most existing solutions use fixed flight altitudes or uniform overlap rate parameters, without fully considering the elevation differences and topography undulations within the inspection area. This leads to inconsistent ground resolution of images in areas with large undulations when flying at a fixed altitude, resulting in a large difference in image detail between flat areas and hilly areas. A fixed overlap rate setting can cause data redundancy in low-lying areas and insufficient coverage in high-altitude areas, leading to subsequent image splicing failures, which severely restricts the generation accuracy and reliability of orthographic images. Second, in terms of intelligence and automation, existing methods generally lack deep real-time fusion capabilities with three-dimensional geographic information, and cannot dynamically adjust the three-dimensional coordinates of the waypoints and the attitude of the aircraft according to the real terrain. When facing obstacles with complex three-dimensional structures such as high-voltage transmission lines, wind turbines and high-rise buildings, existing obstacle avoidance algorithms mostly rely on a pre-set two-dimensional plane obstacle library, which not only makes it difficult to autonomously generate a safe and reliable three-dimensional avoidance path, but also requires manual secondary confirmation and intervention, greatly increasing the operation risk and time cost, and failing to meet the real-time requirements of large-scale automated inspection.

[0004] In summary, the prior art lacks deep integration with three-dimensional geographic information at the structural level of route planning, resulting in the existing unmanned aerial vehicle orthographic image route generation method being difficult to meet the increasingly complex industry application requirements in terms of efficiency, accuracy and safety. Therefore, there is an urgent need for a fully automatic and adaptive route generation method that can deeply integrate real-time terrain analysis, dynamic parameter adjustment and intelligent three-dimensional obstacle avoidance decision-making to break through the bottleneck of existing technology. SUMMARY

[0005] The purpose of the present application is to overcome the shortcomings of the prior art and provide a method for generating orthographic images for autonomous inspection of unmanned aerial vehicles. The method uses a path optimization strategy based on genetic algorithms to achieve global optimization of the route by combining grid features and image parameter constraints in the inspection area. The method uses candidate waypoints as nodes and filters redundant waypoints using a flight segment smoothness factor to reduce the complexity of the route. The genetic algorithm uses flight height and waypoint spacing as constraints to construct a fitness function that includes flight distance, number of turns and overlap rate. The global optimal path is found by iterating through the population, optimizing and balancing multiple objectives to ensure that the route meets the image acquisition accuracy requirements while reducing flight distance and the number of turns to adapt to the endurance of the unmanned aerial vehicle and ensure efficient progress of the inspection mission. At the same time, the simulated output of the waypoint coordinates directly adapts to the flight control system to ensure that the actual flight parameters are consistent with the planned parameters, thereby avoiding image quality problems and flight safety risks caused by defects in route design from the source.

[0006] To solve the above technical problems, the present application provides the following technical solution: a method for generating orthographic images for autonomous inspection of unmanned aerial vehicles, the specific steps of which are as follows: S100, obtaining boundary information and digital elevation model data of a target area, dividing the area into uniform grids according to a predetermined resolution, and taking each grid center point as a candidate waypoint; S200, dynamically calculating the flight height, number of route segments, waypoint spacing and overlap rate based on the camera focal length, sensor size, predetermined ground resolution and overlap rate requirements; ; S300, generating parallel routes along the long direction of the area, performing elevation interpolation on each waypoint based on the digital elevation model to dynamically adjust the flight height, and comparing it with the obstacle database to perform obstacle avoidance operations such as lateral offset and height increase for waypoints within a safety threshold; S400, merging adjacent waypoints and reducing the number of turns using a genetic algorithm, performing collision detection and overlap rate verification in a simulation environment, and outputting the final waypoint coordinates; ​​​S500, the unmanned aerial vehicle flies along the route and automatically triggers shooting at the waypoint, and transmits the image data in real time, when detecting image missing and sudden obstacles, re-divide the grid according to 1 / 2 of the grid resolution in S100 for the area, generate local reflight route combining the parameter calculation logic of S200 and execute reflight, finally complete the orthographic image data acquisition of the whole area.

[0007] Further, in S100, the specific steps of obtaining digital elevation model data and grid division are: Import the boundary polygon of the target area through the geographic information system interface and manually draw it by the user; Read the point cloud data in the boundary polygon area by digital elevation model and real-time scanning of airborne laser radar; According to the preset ground resolution target and camera parameter division resolution, and combining the terrain complexity reflected by the digital elevation model, the target area is divided into uniform grid, and each grid is given a unique identifier and the latitude and longitude elevation coordinates of the center point.

[0008] Further, in S200, for the flight height The calculation is based on the requirement of preset ground resolution GSD, camera focal length And the pixel size of the camera sensor, through the proportional relationship of lens focal length and pixel size, combined with the preset ground resolution, the required flight height , wherein, the ground resolution GSD is a customized parameter input by the user, which represents the actual size of the ground represented by each pixel in the image, represent the actual focal length of the camera lens, represent the pixel size of the camera sensor, that is, the physical size of a single photosensitive unit on the sensor, is the sensor area.

[0009] Further, in S200, the waypoint spacing includes the heading spacing And the lateral spacing , which is determined by the heading overlap rate And the lateral overlap rate , and the heading overlap rate , the lateral overlap rate , according to the camera field of view angle and the flight height The ground coverage width And length of a single image are calculated, then the heading spacing , used to ensure that the adjacent shooting points meet the preset image overlap ratio in the flight direction, the lateral spacing This is used to ensure that the images between adjacent parallel flight paths meet the preset overlap requirements in the lateral direction.

[0010] Furthermore, in S200, regarding the number of routes... The width of the target area and the lateral spacing Joint decision, that is ,in, This represents the total number of routes required. This represents the actual width of the inspection area along its shorter side. This represents the lateral distance between two adjacent flight paths. The function is called the floor function, which rounds the calculation result up. This calculation ensures that the generated parallel flight paths can completely cover the entire target area, and the number of flight paths is the minimum value that satisfies the lateral overlap condition. This optimizes the flight path, reduces redundant flight, and improves inspection efficiency while ensuring image quality.

[0011] Furthermore, in step S300, based on the digital elevation model data acquired in step S100, elevation interpolation calculations are performed on each generated initial waypoint to determine the actual terrain elevation of its location. ; The elevation interpolation uses bilinear interpolation, which employs a weighted average of grid point data with known elevations around the waypoint to obtain the interpolation result, thereby establishing three-dimensional waypoint coordinates that correspond to the actual terrain. Flight altitude calculated from S200 And the terrain elevation obtained by interpolation Dynamically correct the absolute flight altitude of the waypoint The correction is as follows: ,in, The safety redundancy height is preset according to the operation specifications to ensure that there is a sufficient safety gap between the drone and the ground vegetation. The principle of this operation is to enable the drone to strictly follow the changes in terrain and fly in a terrain-following manner, thereby maintaining a constant relative altitude and ground resolution throughout the inspection area, and avoiding inconsistent image resolution or blind spots caused by changes in terrain.

[0012] Furthermore, after the S300 completes the elevation adjustment, it performs three-dimensional spatial conflict detection between each three-dimensional waypoint and a preset obstacle database. The obstacle database includes static obstacles and dynamic obstacles. The coordinates and height information of the static obstacles are pre-imported through a geographic information system, and the information of the dynamic obstacles is updated in real time through the UAV's onboard millimeter-wave radar. A safety threshold is set for each type of obstacle. When the three-dimensional Euclidean distance between a waypoint and an obstacle is detected to be less than the safety threshold, obstacle avoidance is automatically triggered. The obstacle avoidance operation adopts a horizontal offset strategy in the horizontal plane, that is, a new waypoint is generated in a direction perpendicular to the original flight path, and the horizontal distance between the new waypoint and the obstacle is greater than the safety threshold. When the horizontal offset cannot be implemented due to airspace restrictions, a vertical altitude boost strategy is activated, that is, the absolute flight altitude of the waypoint is raised to the level of the obstacle and a safety redundancy is added. This forms an obstacle-crossing flight path.

[0013] Furthermore, in S400, the specific process of merging adjacent waypoints and reducing the number of turns using the genetic algorithm includes: The initial waypoints of the initial route are initially screened, using the coordinates of the candidate waypoints in S100 as nodes and the straight-line distance between adjacent waypoints as the basic path cost. Simultaneously, a segment smoothness factor is introduced as a heuristic function to delete redundant waypoints and merge consecutive waypoints. The segment smoothness factor... ,in, This is the segment smoothness factor, with a value ranging from 0 to 1. The angle between two adjacent flight segments is the angle between the line connecting the end of the previous flight segment and the current waypoint, and between the current waypoint and the start of the next flight segment. When it approaches 1, it indicates that the angle of the flight segment is... Approaching 0 degrees, the flight path is smooth; when When it approaches 0, it indicates that the angle of the flight segment is... Approaching 90 degrees, the flight path shows a clear turn; For parallel routes generated in S300, if the straight-line distance between two adjacent waypoints is less than 0.5 meters, they are identified as redundant waypoints and deleted. If the straight-line distance between three consecutive waypoints is less than 0.5 meters, they are deleted. When the average value is greater than 0.95, intermediate waypoints are merged, and only the first and last waypoints are retained. This filtering process can reduce the number of waypoints in the initial route by 15% to 20%, thereby reducing the route complexity while ensuring coverage accuracy. The selected flight routes are globally optimized, using the flight altitude and waypoint spacing calculated in S200 as constraints, with the optimization objectives being to minimize the total flight distance, the number of turns, and the image overlap rate. The number of flight routes and the order of waypoints are used as gene encodings to construct a length of [missing information]. Chromosomes, among which For the total number of waypoints after filtering, perform iterative optimization, including: Population size: set to 50, that is, 50 candidate route schemes are maintained in each iteration, and each scheme corresponds to a complete route; Fitness function: Constructing a fitness function based on the requirements of the inspection task. wherein, is the fitness value, used to evaluate the quality of the candidate flight path, is the total flight distance of the flight path, i.e. the sum of the straight-line distances between all adjacent waypoints, is the number of turns, is the overlap rate compliance rate, including the heading overlap rate , the lateral overlap rate , and the proportion of the number of image pairs in the abnormal area to the total number of image pairs, , , respectively, the weight coefficients of the total flight distance, the number of turns, and the overlap rate compliance rate, and satisfy ; Selection operation: the probability of an individual being selected is proportional to the fitness value, and for individuals with a fitness value , 5% of the individuals are directly entered into the next generation; The crossover operation adopts the two-point crossover method, and the crossover probability is set to 0.7, that is, two parent chromosomes, i.e. two different flight paths, are selected, the gene fragments of the waypoints between the crossover points are exchanged, and the child chromosomes are generated; Mutation operation: random mutation method is used, the mutation probability is set to 0.05, the waypoint gene in the chromosome is selected, and according to the coordinates of the adjacent nodes in S100, the waypoint is replaced by one of the eight adjacent grid centers, and the flight path is locally adjusted; Iteration number: set to 50 to 100 times, and the iteration termination condition is that the fitness value change rate of the individual in the last 10 iterations is <1%, and the final waypoint coordinates are output.

[0014] Further, in S500, the logic for detecting image loss and sudden obstacles is: By analyzing the real-time returned image metadata and low-resolution thumbnails, when 3 consecutive waypoints do not receive valid images, it is determined that the image is missing; When the airborne obstacle avoidance radar detects a moving object with a relative speed > 5m / s, it is determined that a sudden obstacle has occurred; The execution logic of the local reflight flight path is: When image loss and sudden obstacles are identified, the current main flight path task is suspended, and the boundary coordinates of the region are recorded. Using the boundary as a new target area, the local reflight process is triggered, which uses half of the initial grid resolution in S100 to redivide the abnormal area to generate reflight candidate waypoints; Based on the reflight candidate waypoints, the flight parameters are recalculated, including the flight height, the number of flight paths, and the waypoint spacing, to generate a local reflight flight path; After the UAV finishes the local supplementary flight path, it automatically returns to the main flight path breakpoint to continue the task.

[0015] Compared with the prior art, the method for generating orthographic images by the UAV autonomous inspection has the following beneficial effects: Firstly, the path optimization strategy of the genetic algorithm is combined with the grid features of the inspection area and the image parameter constraints to realize global optimization of the flight path, candidate waypoints are taken as nodes, redundant waypoints are filtered through a flight segment smoothness factor to reduce the complexity of the flight path, the genetic algorithm takes the flight height and the waypoint spacing as constraints, constructs a fitness function containing the flight distance, the number of turns and the overlap rate, finds the global optimal path through population iteration, optimizes and balances the multi-objective requirements, ensures that the flight path meets the image acquisition accuracy requirements, shortens the flight distance and reduces the number of turns, adapts to the endurance of the UAV, and guarantees efficient progress of the inspection task, at the same time, the waypoint coordinates simulated and output are directly adapted to the flight control system to ensure that the actual flight parameters are consistent with the planned parameters, thereby avoiding image quality problems and flight safety risks caused by defects in the flight path design. Additional advantages, objects, and features of the application will be set forth in part in the description which follows, and in part will become apparent to those skilled in the art upon examination of the following specification, or can be learned by practice of the application. BRIEF DESCRIPTION OF DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiment or prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and for those skilled in the art, other drawings can also be obtained without creative labor on the basis of these drawings.

[0017] Figure 1 An operation flowchart of the method for generating orthographic images by the UAV autonomous inspection; Figure 2 A step framework diagram of the method for generating orthographic images by the UAV autonomous inspection; Figure 3 An execution logic diagram when image missing and sudden obstacles are monitored. DETAILED DESCRIPTION

[0018] In order to further illustrate the technical means and effects adopted by the present application to achieve the predetermined application purpose, the following will combine the drawings and the preferred embodiments to specifically describe the specific embodiments, structures, features and effects of the present application.

[0019] Embodiment one: the embodiment provides a working principle of a method for generating orthographic images by unmanned aerial vehicle autonomous inspection, aiming to solve the problems of the existing technology in the aspects of terrain adaptability, obstacle avoidance safety and inspection efficiency of unmanned aerial vehicle inspection route planning, such as Figure 2 As shown in the figure, the method starts from target area basic data acquisition and grid division, dynamically calculates flight key parameters in combination with camera parameters and terrain data, optimizes the route based on real terrain and obstacle information and performs obstacle avoidance operation, then globally optimizes the route by genetic algorithm to reduce redundant flight points and turning times, finally monitors image data in real time during actual flight and generates local reflight route for abnormal situations, realizes high-precision, high-efficiency and high-safety orthographic image data collection in the whole area, and is suitable for power inspection, geological disaster monitoring, smart city modeling and basic surveying and mapping and other complex scenes.

[0020] First, enter the target area basic data acquisition and grid division stage (S100), in the acquisition target area boundary information link, adopt two complementary ways to guarantee the accuracy and flexibility of boundary data, on the one hand, through the geographic information system (GIS) interface imports the boundary polygon of target area, this way is suitable for the scene that has clear geographic boundary data, can directly obtain the area boundary confirmed by professional surveying and mapping, avoid the error that manual drawing can produce;The other hand, allow user to draw boundary polygon manually, this way is suitable for no preset GIS data or need to adjust the situation of temporary inspection area, user can determine the target area range according to actual inspection demand, through the point selection coordinate or drag drawing mode, the combination of two kinds of ways not only meets the accuracy requirement of standardization operation, also takes into account the flexibility of field operation;Acquire digital elevation model (DEM) data for reflecting the terrain undulation characteristics of target area, provide basis for height adjustment of flight point, and the point cloud data scanned by airborne laser radar in real time can further supplement the fine terrain and ground object information in the region, improve the fineness of terrain data, specifically: after determining the target area boundary, call the digital elevation model database, extract the DEM data in the boundary polygon range, the data is stored in the form of regular grid, each grid unit contains the corresponding elevation value, at the same time, the unmanned aerial vehicle scans the boundary polygon region in real time through the airborne laser radar equipment before taking off or at the initial stage of flight, obtains the three-dimensional point cloud data in the region, the point cloud data contains a large number of three-dimensional coordinates (longitude, latitude, elevation) of discrete points, which can accurately capture the detailed terrain information of DEM data that may be missed, such as ground vegetation and small protrusions, combining DEM data with airborne laser radar point cloud data can form a more comprehensive and accurate target area terrain data set, according to the target area terrain data set, the continuous target area is converted into discrete and standardized candidate flight point carriers, laying a foundation for subsequent flight point generation and parameter calculation, before dividing the grid, determine the initial preset ground resolution, adjust the initial preset ground resolution combined with camera parameters (such as sensor size, focal length), after determining the final grid resolution, divide the target area into uniform grids according to the resolution, each grid is as a basic image acquisition unit, for the convenience of subsequent flight point management and parameter association, give each grid a unique identifier (such as the code generated based on the row and column number of the grid in the region), at the same time, calculate the longitude and latitude and elevation coordinates of the center point of each grid, the center point is the candidate flight point, the longitude and latitude coordinates of the candidate flight point are calculated through the position of the grid in the target area coordinate system, and the elevation coordinate is obtained by fusing and interpolating DEM data and airborne laser radar point cloud data, which ensures that the elevation of each candidate flight point can truly reflect the terrain height of its position.

[0021] Then, enter the flight key parameter dynamic calculation phase (S200), based on the target area information obtained in S100, camera hardware parameters and preset image quality requirements (overlap rate), accurately calculate the unmanned aerial vehicle flight height , flight line number , flight point spacing and , wherein, for the flight height , through the geometric relationship between the camera optical parameters and the preset ground resolution (GSD), determine the required flight height of the unmanned aerial vehicle to ensure that the captured image can meet the preset ground resolution requirement, the ground resolution GSD is a customized parameter input by the user according to the inspection task requirement, representing the actual ground size corresponding to each pixel in the image, the camera focal length is the actual optical focal length of the camera lens, which is a fixed parameter of the camera and directly affects the imaging range and magnification of the lens; the pixel size of the camera sensor is the physical size of a single photosensitive unit on the sensor, which is also a fixed parameter of the camera, and its size determines the sensor's ability to capture light signals and image resolution. When calculating the flight height , when the unmanned aerial vehicle is at a height , the field of view of the camera lens will form a coverage area on the ground, and the ground size corresponding to each pixel in the image (i.e. GSD) is proportional to the flight height , inversely proportional to the camera focal length , and proportional to the pixel size , at the same time, considering the influence of the camera sensor area on the imaging range (the larger the sensor area, the wider the ground coverage range under the same focal length and height), the sensor area is included in the calculation to correct the flight height, and the final flight height calculation formula is: The flight height calculated by the formula can ensure that each image captured by the camera during the flight of the unmanned aerial vehicle meets the user's preset ground resolution requirement, providing a unified accuracy basis for subsequent image stitching and orthographic image generation; the calculation of flight point spacing is the key to ensuring that adjacent images meet the overlap rate requirement, which includes heading spacing (flight point interval along the flight direction) and lateral spacing (flight point interval perpendicular to the flight direction), which requires the combination of heading overlap rate and lateral overlap rate , wherein, the heading overlap rate is the overlap ratio of adjacent two images in the flight direction, and the lateral overlap rate This refers to the overlap ratio between images of two adjacent parallel flight paths. The overlap ratio is set to ensure that sufficient matching feature points can be found during subsequent image stitching, avoiding stitching gaps or misalignments due to insufficient overlap. In this embodiment, the flight path overlap ratio is set. Lateral overlap rate Based on the camera's field of view and flight altitude Calculate the ground coverage width of a single image and length Then the heading spacing Calculated using this formula This ensures that images taken at adjacent waypoints have sufficient overlap in the flight direction, providing ample matching features for image stitching, and lateral spacing. This spacing ensures seamless lateral image transition between adjacent parallel flight paths, avoiding blind spots; regarding the number of flight paths... The width of the target area Lateral spacing The width of the target area is determined jointly. This refers to the actual width of the inspection area along its shorter side. The shorter side is chosen as the basis for width calculation because when generating parallel flight paths along the longer side of the area, the coverage area along the shorter side directly determines the required number of flight paths. This minimizes the number of turns between flight paths, improving flight efficiency. Therefore, the number of flight paths... ,in This is an up-rounding function. Its function is to round the calculation result up to an integer. Under the premise of satisfying the lateral overlap rate condition, it minimizes the number of flight paths, thereby reducing redundant flights and improving inspection efficiency.

[0022] Next, the flight path generation, waypoint elevation adjustment, and obstacle avoidance phase (S300) begins. The theoretical parameters calculated in S200 are combined with real terrain and obstacle information to generate a 3D flight path adapted to the actual environment. This ensures the UAV can fly along terrain undulations while avoiding static and dynamic obstacles within the area, guaranteeing flight safety and image accuracy. Based on the digital elevation model (DEM) data of the target area and the planar coordinates (latitude and longitude) of candidate waypoints acquired in S100, elevation interpolation is performed on each initial waypoint to determine its actual terrain elevation. , the elevation value of the waypoint is calculated by weighted average method using the data of four known elevation grid points around the waypoint. In this embodiment, for the candidate waypoint P, the nearest four grid points P1(x1, y1, z1), P2(x2, y1, z2), P3(x1, y2, z3), P4(x2, y2, z4) (where x and y are plane coordinates, and z is the elevation) are found in the DEM data. According to the x-coordinate distance between the waypoint P and P1 and P2, the elevation approximation value z12 of the P point in the y1 direction is calculated. Then, according to the y-coordinate distance between the waypoint P and P1 and P3, the elevation approximation value z13 of the P point in the x1 direction is calculated. Finally, the y-coordinate distance between the waypoint P and y1 and y2 is used to perform weighted average on z12 and z13, to obtain the final elevation value of the waypoint P , the flight height calculated in S200 , dynamically correct the absolute flight height of the waypoint The absolute flight height refers to the flight height of the unmanned aerial vehicle relative to the sea level. Its correction formula is: , wherein is the safety redundancy height preset according to the operation specification. Through the correction operation, the unmanned aerial vehicle can strictly follow the terrain changes to perform the terrain-following flight, maintain the constant relative height and ground resolution in the entire inspection area, avoid the inconsistent image resolution or blind area caused by the terrain changes, and complete the elevation adjustment of the waypoint. After the elevation adjustment of the waypoint, obstacle conflict detection and obstacle avoidance operation are performed on each three-dimensional waypoint to cope with the static and dynamic obstacles in the region. First, an obstacle database is constructed, which includes two types of information: the coordinate and height information of the static obstacle (such as high-voltage transmission tower, high-rise building, mountain cliff, etc.), which is pre-imported by the geographic information system (GIS) before flight, to ensure that the unmanned aerial vehicle has mastered the distribution of the fixed obstacles in the region before taking off; and the information of the dynamic obstacle (such as other aircraft, birds, temporary construction equipment, etc.), which is captured by the millimeter wave radar on the unmanned aerial vehicle in real time during the flight to provide data support for dynamic obstacle avoidance. Safety threshold values are set for each type of obstacle. The obstacle avoidance operation adopts the strategy of “horizontal offset first, height increase second”: first, try to perform horizontal offset in the horizontal plane, that is, generate a new avoidance waypoint in the direction perpendicular to the original flight path. The plane coordinates of the new waypoint need to satisfy that the horizontal distance from the obstacle is greater than the safety threshold value, and the new waypoint is still within the target inspection area and does not affect the image overlap rate of the adjacent waypoints. The advantage of the horizontal offset strategy is that it can maintain the relative flight height of the unmanned aerial vehicle unchanged, ensure the image resolution, and avoid other risks caused by height changes. When the horizontal offset cannot be implemented due to airspace restrictions (such as the edge of the target area, adjacent no-fly zone), the height increase strategy in the vertical direction is enabled. At this time, the absolute flight height of the waypoint Raise it to a height above the top of the obstacle, and add a safety margin of safety. This process creates an obstacle-crossing flight path, ensuring that the drone can safely fly over obstacles. After implementing the altitude-boosting strategy, the image overlap rate between the waypoint and surrounding waypoints is re-verified. If the overlap rate fails to meet the standard due to altitude boosting, the altitude of the surrounding waypoints needs to be adaptively adjusted to ensure that the overall image quality of the flight path is not affected.

[0023] Next, the route global optimization and simulation verification stage (S400) is entered. The initial route generated in S300 is optimized using a genetic algorithm to remove redundant waypoints and reduce the number of turns. Specifically, the initial waypoints are initially screened; when the straight-line distance between two adjacent waypoints is less than 0.5 meters, they are identified as redundant and deleted. Using the candidate waypoint coordinates from S100 as nodes and the straight-line distance between adjacent waypoints as the basic path cost, a segment smoothness factor is introduced as a heuristic function to delete redundant waypoints and merge consecutive waypoints. ,in, This is the segment smoothness factor, with a value ranging from 0 to 1. The angle between two adjacent flight segments is the angle between the line connecting the end of the previous flight segment and the current waypoint, and between the current waypoint and the start of the next flight segment. When it approaches 1, it indicates that the angle of the flight segment is... Approaching 0 degrees, the flight path is smooth; when When it approaches 0, it indicates that the angle of the flight segment is... Approaching 90 degrees, the flight path exhibits a significant turning point. For parallel routes generated in S300, if the straight-line distance between two adjacent waypoints is less than 0.5 meters, they are identified as redundant waypoints and deleted. When the angle between adjacent segments formed by three consecutive waypoints corresponds to... When the average value is >0.95, intermediate waypoints are merged, retaining only the first and last waypoints. This screening process reduces the number of waypoints in the initial route by 15% to 20%, reducing route complexity while maintaining coverage accuracy. The screened routes are then globally optimized, using the flight altitude and waypoint spacing calculated in S200 as constraints, with the optimization objectives being minimizing total flight distance, minimizing the number of turns, and achieving image overlap. The number of routes and waypoint order are used as gene encodings to construct a length of [missing information]. Chromosomes, among which For the total number of waypoints after filtering, perform iterative optimization, including: Population size: set to 50, that is, 50 candidate route schemes are maintained in each iteration, and each scheme corresponds to a complete route; Fitness function: Constructing a fitness function based on the requirements of the inspection task. ,in, fitness value, for evaluating the quality of candidate flight path, total flight distance, for the sum of straight-line distances between all adjacent waypoints, turning number, overlapping rate, including heading overlapping rate , lateral overlapping rate , the proportion of image pairs in the image overlapping rate , , total flight distance, turning number, and overlapping rate, respectively, and satisfy ; Selection operation: the probability of individual being selected is proportional to the fitness value, and for individuals with fitness value , 5% of them are directly entered into the next generation; The crossover operation adopts two-point crossover method, and the crossover probability is set to 0.7, that is, two parent chromosomes, that is, two different flight paths, are selected, and the order of the waypoint gene fragments between the crossover points is exchanged to generate child chromosomes; Mutation operation: random mutation method is adopted, mutation probability is set to 0.05, waypoint gene in chromosome is selected, according to the coordinates of adjacent nodes in S100 grid, the waypoint is replaced by one of the eight adjacent grid centers, and local adjustment of flight path is carried out; Iteration number: set to 50 to 100 times, and the iteration termination condition is that the change rate of individual fitness value in continuous 10 iterations is <1%, and the final waypoint coordinates are output, which are directly imported into the unmanned aerial vehicle flight control system to ensure that the actual flight parameters are consistent with the planning parameters.

[0024] Finally, enter the image acquisition and local reflight phase (S500), execute the UAV flight along the route and automatic shooting operation, after the UAV flight control system imports the final waypoint coordinates output by S400, plan the flight path according to the waypoint sequence, in the flight process, the UAV positions itself in real time through the onboard GPS and inertial navigation system, when it reaches the preset waypoint, automatically triggers the camera shooting function, the shooting frequency matches the waypoint spacing (ensures that the images taken by adjacent waypoints meet the overlap rate requirement), at the same time, the camera returns the image data (including raw images, image metadata such as shooting time, coordinates, exposure parameters, etc.) taken to the ground control center in real time, the ground control center stores and preliminarily analyzes the returned data in real time, by analyzing the real-time returned image metadata and low-resolution thumbnails, the ground control center sets the continuous waypoint image receiving threshold to 3, that is, when 3 consecutive waypoints do not receive valid images (including not receiving image data, missing image metadata, image thumbnail blurred and cannot be identified, etc.), it is determined that the image is missing, at the same time, through the real-time scanning of the three-dimensional space around the UAV by the UAV onboard obstacle avoidance radar, when a moving object with a relative speed > 5m / s is detected, it is determined that a sudden obstacle has occurred, when image missing or sudden obstacle is detected, the system executes the generation and execution logic of the local reflight route, the specific steps are as follows: Task pause and area recording: the system immediately sends instructions to pause the current main route task of the UAV, and the UAV hovers at a safe height; at the same time, the ground control center records the boundary coordinates of the area where the image is missing or the sudden obstacle occurs, the boundary range is centered on the abnormal waypoint and extends 2 times the initial grid resolution in S100, ensuring that the entire area where the problem may exist is covered; Local grid redivision: taking the recorded abnormal area boundary as the new target area, redivide the grid according to 1 / 2 of the initial grid resolution in S100, redivide the grid with 1 / 2 resolution in order to increase the waypoint density of the reflight area, ensure that the reflight image has higher accuracy and more detailed coverage, and make up for the possible coverage blind area or insufficient accuracy of the original route; Re-calculation of reflight parameters: combine the parameter calculation logic of S200 to re-calculate the flight height, route number and waypoint spacing of the reflight route; Reflight route generation and execution: based on the redivided reflight candidate waypoints and calculated reflight parameters, generate a local reflight route, after receiving the reflight route instructions, the UAV flies from the hovering position to the starting waypoint of the reflight area, executes the shooting task according to the reflight route, returns the reflight image data in real time, and the ground control center verifies the reflight image in real time to ensure that there is no missing and the overlap rate meets the standard; Return to the main route: after completing the reflight task, the UAV automatically flies to the breakpoint of the main route (i.e. the waypoint at the time of suspension), continues to execute the unfinished main route task until the image acquisition of the entire target area is completed.

[0025] In summary, this embodiment achieves the entire process of generating autonomous inspection orthophotos by UAVs through five closely connected stages from S100 to S500. The entire process deeply integrates three-dimensional geographic information, dynamic parameter calculation, intelligent obstacle avoidance decision-making, and global path optimization technology to ensure that the generated orthophotos meet the requirements of uniform scale, no distortion, and high stitching quality.

[0026] Example 2: Figure 1 As shown, this embodiment provides a method for generating orthophotos of autonomous UAV inspections. The workflow for generating orthophotos of autonomous UAV inspections includes the following steps: (1) Obtaining regional information and dividing the grid Obtain the boundary information of the target area.

[0027] Obtain the digital elevation model (DEM) data for this area.

[0028] The area is divided into a uniform grid based on the preset ground resolution.

[0029] The center point of each grid is used as a candidate waypoint.

[0030] (2) Calculate flight parameters Based on camera focal length, sensor size, preset ground resolution, and overlap requirements, the following parameters are dynamically calculated: flight altitude, number of flight paths, and waypoint spacing (including heading spacing and lateral spacing).

[0031] (3) Generate flight path and adjust elevation Parallel routes are generated along the longer side of the region.

[0032] Elevation interpolation is performed at each waypoint to obtain the actual terrain height.

[0033] The flight altitude is dynamically adjusted according to the terrain to maintain a constant relative altitude.

[0034] (4) Obstacle detection and obstacle avoidance Perform 3D collision detection on each waypoint and obstacle database.

[0035] If the distance between the waypoint and the obstacle is less than the safety threshold, an obstacle avoidance operation will be performed.

[0036] Prioritize lateral offset.

[0037] If you cannot deviate, increase your flight altitude to overcome the obstacle.

[0038] (5) Route optimization and verification Genetic algorithms are used to optimize waypoints, remove redundant waypoints, and merge consecutive waypoints to reduce turns.

[0039] Optimize waypoint order and route structure, collision detection and overlap rate checking in simulation environment.

[0040] Output final waypoint coordinates.

[0041] (6) Execute flight and abnormality handling As shown in Figure 3 , the UAV flies along the planned route, automatically takes pictures at the waypoints, and real-time transmits image data.

[0042] If image loss or sudden obstacles are detected.

[0043] Pause the current task.

[0044] Re-divide the abnormal area at 1 / 2 of the original grid resolution.

[0045] Generate a local re-flight route and execute it.

[0046] Return to the main route to continue the task after completion.

[0047] The above is only the preferred embodiment of the present application, and is not intended to limit the present application in any form. Although the present application has been disclosed as above with the preferred embodiment, it is not intended to limit the present application. Any person skilled in the art can make some changes or modifications to the above disclosed technical content without departing from the scope of the technical solution of the present application, and any equivalent embodiments with equivalent changes and modifications are still within the scope of the present application.

Claims

1. A method for generating autonomous inspection orthophotos by unmanned aerial vehicles (UAVs), characterized in that, The specific steps of this method are as follows: S100: Obtain the boundary information and digital elevation model data of the target area, and divide the area into uniform grids according to a preset resolution, with the center point of each grid as a candidate waypoint; S200 dynamically calculates flight altitude based on camera focal length, sensor size, preset ground resolution, and overlap rate requirements. Number of routes waypoint spacing and ; S300 generates parallel flight routes along the long side of the region, performs elevation interpolation on each waypoint based on the digital elevation model to dynamically adjust the flight altitude, and compares it with the obstacle database. For waypoints with a distance less than the safety threshold, it performs obstacle avoidance operations such as lateral offset and altitude increase. S400 uses a genetic algorithm to merge adjacent waypoints and reduce the number of turns. Collision detection and overlap rate verification are performed in a simulation environment, and the final waypoint coordinates are output. The S500 drone flies along the flight path and automatically triggers shooting at waypoints, transmitting image data back in real time. When image loss or sudden obstacles are detected, the area is re-divided into grids at half the grid resolution in S100. Combined with the parameter calculation logic of S200, a local supplementary flight path is generated and the supplementary flight is executed, ultimately completing the acquisition of orthophoto data for the entire area.

2. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In step S100, the specific steps for obtaining digital elevation model data and grid division are as follows: The boundary polygons of the target area are imported via a geographic information system interface and drawn manually by the user. The point cloud data within the boundary polygon area is scanned in real time using a digital elevation model and airborne lidar. The resolution is determined based on the preset ground resolution target and camera parameters, and the resolution is adjusted in combination with the terrain undulation complexity reflected by the digital elevation model. The target area is divided into uniform grids, and each grid is assigned a unique identifier and the latitude, longitude and elevation coordinates of its center point.

3. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In S200, regarding flight altitude The calculation is based on the preset ground resolution (GSD) requirements and camera focal length. and the pixel size of the camera sensor By using the ratio of lens focal length to pixel size, combined with a preset ground resolution, the required flight altitude can be obtained. Wherein, the ground resolution GSD is a customized parameter input by the user, representing the actual ground size represented by each pixel in the image. Represents the actual focal length of the camera lens. Represents the pixel size of a camera sensor, that is, the physical size of a single photosensitive unit on the sensor. The sensor area is [area].

4. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In S200, the waypoint spacing includes the heading spacing. and lateral spacing Its calculation is based on the heading overlap rate. and lateral overlap rate Confirmed, and heading overlap rate Lateral overlap rate Based on the camera's field of view and the flight altitude Calculate the ground coverage width of a single image. and length Then the heading spacing This is used to ensure that adjacent shooting points meet a preset image overlap ratio in the flight direction, and the lateral spacing... This is used to ensure that the images between adjacent parallel flight paths meet the preset overlap requirements in the lateral direction.

5. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In S200, regarding the number of routes The width of the target area and the lateral spacing Joint decision, that is ,in, This represents the total number of routes required. This represents the actual width of the inspection area along its shorter side. This represents the lateral distance between two adjacent flight paths. This is a rounding function used to round the calculated result up.

6. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In step S300, based on the digital elevation model data obtained in step S100, elevation interpolation calculations are performed on each generated initial waypoint to determine the actual terrain elevation of its location. ; The elevation interpolation uses bilinear interpolation, which employs a weighted average of grid point data with known elevations around the waypoint to obtain the interpolation result, thereby establishing three-dimensional waypoint coordinates that correspond to the actual terrain. Flight altitude calculated from S200 The terrain elevation obtained by interpolation Dynamically correct the absolute flight altitude of the waypoint The correction is as follows: ,in, This is a safety redundancy height preset according to the operating specifications.

7. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, After the S300 completes the elevation adjustment, it performs three-dimensional spatial conflict detection between each three-dimensional waypoint and the preset obstacle database. The obstacle database contains static obstacles and dynamic obstacles. The coordinates and height information of the static obstacles are pre-imported through a geographic information system, and the information of the dynamic obstacles is updated in real time through the UAV's onboard millimeter-wave radar. A safety threshold is set for each type of obstacle. When the three-dimensional Euclidean distance between a waypoint and an obstacle is detected to be less than the safety threshold, obstacle avoidance is automatically triggered. The obstacle avoidance operation adopts a horizontal offset strategy in the horizontal plane, that is, a new waypoint is generated in a direction perpendicular to the original flight path, and the horizontal distance between the new waypoint and the obstacle is greater than the safety threshold. When the horizontal offset cannot be implemented due to airspace restrictions, a vertical altitude boost strategy is activated, that is, the absolute flight altitude of the waypoint is raised to the level of the obstacle and a safety redundancy is added. This forms an obstacle-crossing flight path.

8. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In S400, the specific process of merging adjacent waypoints and reducing the number of turns using the genetic algorithm includes: The waypoints of the initial route are initially screened, using the coordinates of the candidate waypoints in S100 as nodes and the straight-line distance between adjacent waypoints as the basic path cost. Simultaneously, a segment smoothness factor is introduced as a heuristic function to delete redundant waypoints and merge consecutive waypoints. The segment smoothness factor... ,in, This is the segment smoothness factor, with a value ranging from 0 to 1. It is the angle between two adjacent segments, that is, the angle between the line connecting the end of the previous segment and the current waypoint, and the line connecting the current waypoint and the start of the next segment; For parallel routes generated in S300, if the straight-line distance between two adjacent waypoints is less than 0.5 meters, they are identified as redundant waypoints and deleted. If the straight-line distance between three consecutive waypoints is less than 0.5 meters, they are deleted. When all values ​​are greater than 0.95, intermediate waypoints will be merged, and only the first and last waypoints will be retained; The selected flight routes are globally optimized, using the flight altitude and waypoint spacing calculated in S200 as constraints, with the optimization objectives being to minimize the total flight distance, the number of turns, and the image overlap rate. The number of flight routes and the order of waypoints are used as gene encodings to construct a length of [missing information]. Chromosomes, among which For the total number of waypoints after filtering, perform iterative optimization, including: Population size: set to 50, that is, 50 candidate route schemes are maintained in each iteration, and each scheme corresponds to a complete route; Fitness function: Constructing a fitness function based on the requirements of the inspection task. ,in, The fitness value is used to evaluate the merits of candidate routes. This is the total flight distance of the route, which is the sum of the straight-line distances between all adjacent waypoints. For the number of turns, The overlap rate compliance rate includes heading overlap rate. Lateral overlap rate The proportion of image pairs within the total number of image pairs. , , These are the weighting coefficients for the total flight distance, number of turns, and overlap rate compliance rate of the flight route, respectively, and they satisfy the following conditions: ; Selection operation: The probability of an individual being selected is directly proportional to its fitness value. Of the individuals, 5% are retained and directly enter the next generation; The crossover operation uses a two-point crossover method with a crossover probability of 0.

7. That is, two parent chromosomes are selected, which are two different routes, and the gene segments of the waypoint sequence between the crossover points are exchanged to generate offspring chromosomes. Mutation operation: A random mutation method is used with a mutation probability of 0.

05. The waypoint gene in the chromosome is selected, and the waypoint is replaced with one of the candidate waypoints of the center of the eight adjacent grids in S100 according to the coordinates of the adjacent nodes of the grid, so as to make local adjustments to the flight path. Number of iterations: set to 50 to 100 times. The iteration termination condition is that the rate of change of the fitness value of an individual is <1% in 10 consecutive iterations. The final waypoint coordinates are output.

9. The method for generating autonomous inspection orthophotos by a UAV according to claim 1, characterized in that, In S500, the logic for detecting image gaps and sudden obstacles is as follows: By analyzing the real-time transmitted image metadata and low-resolution thumbnails, when no valid images are received for three consecutive waypoints, it is determined that the images are missing. When the airborne obstacle avoidance radar detects a moving object with a relative speed greater than 5 m / s, it is determined to be a sudden obstacle. The execution logic for the partial supplementary flight route is as follows: When image loss or sudden obstacles are identified, the current main flight path is paused and the boundary coordinates of the area are recorded. The boundary is used as the new target area to trigger a local re-flight process. This process uses half of the initial grid resolution in S100 to re-grid the abnormal area and generate candidate waypoints for re-flight. Based on the candidate waypoints for re-flight, the flight parameters, including flight altitude, number of routes, and distance between waypoints, are recalculated to generate local re-flight routes; After completing the aforementioned partial supplementary flight route, the drone automatically returns to the interruption point of the main flight route to continue its mission.

Images