Unmanned aerial vehicle surveying and mapping method for multi-angle photography

Abstract

The invention discloses a multi-angle photographing unmanned aerial vehicle surveying and mapping method, and relates to the technical field of unmanned aerial vehicle surveying and mapping, through forming a photographing angle score set Vgl, illumination suitability quantitative evaluation in a full attitude domain is realized, and an attitude combination with the highest score in the score set Vgl is selected for image acquisition, so that the imaging quality is improved, and the imaging efficiency is improved. A data basis of optimal illumination conditions is provided for modeling, after image acquisition is completed, a point cloud set Pcd is constructed through an image set Img, an illumination imperfection index Lck is calculated, and the illumination shielding influence in the model is quantitatively evaluated; if the illumination incompleteness index Lck exceeds the illumination shielding threshold value LckThr, updating the attitude angle candidate space and driving an iterative optimization process; the problems of large image shadow interference, exposure imbalance and three-dimensional model defects caused by factors such as natural illumination dynamic change, uncontrollable attitude and surface feature structure shielding in an existing surveying and mapping unmanned aerial vehicle method are effectively solved.

Description

Technical Field

[0001] The present invention relates to the technical field of UAV mapping, and specifically to a UAV mapping method for multi-angle photography. Background Art

[0002] With the in-depth application of remote sensing technology, three-dimensional modeling, and spatial data analysis in multiple industries, the broad field of surveying and mapping science and engineering has witnessed rapid development. In this broad field, aerial and UAV mapping, as an efficient way to obtain geographical information currently, has gradually replaced traditional ground measurement methods and become one of the important means of modern mapping.

[0003] In the invention with Chinese Patent Application No. CN202411523934.3, a mapping UAV and a mapping method for reducing mapping errors are disclosed. The present invention uses the radar device and camera of the UAV to monitor the surrounding environment in real time, determines the position of obstacles through data fusion technology and image recognition algorithms, establishes an obstacle avoidance decision model according to the target mapping task, introduces the simulated annealing algorithm to optimize the model and formulate an obstacle avoidance strategy, and finally feeds back the obstacle avoidance strategy to the control center. The control center conducts route planning and controls the attitude adjustment speed of the UAV in real time to obtain the final route. Through the fusion of multiple sensors and optimization algorithms, the present invention can detect obstacles more accurately, formulate reasonable obstacle avoidance strategies, improve mapping accuracy, reduce mapping errors, and ensure the flight safety of the UAV and the accuracy of mapping data.

[0004] It can be seen therefrom that although solutions to the error problem caused by obstacles in the mapping process are proposed, mainly by jointly detecting obstacles through a radar device and image recognition, and introducing the simulated annealing algorithm for obstacle avoidance strategy optimization, thereby improving the route safety and obstacle avoidance efficiency. However, the external environmental information it relies on is mainly the position and distribution of static obstacles in space, ignoring an important variable - the profound impact of natural light conditions on the image imaging quality and modeling integrity.

[0005] When this technology is applied to multi-angle photography mapping, due to the lack of consideration of factors such as the solar altitude angle, incident direction, and dynamic light changes, a large number of shadow occlusion areas may appear on the surface of the same building or ground object during flights at different times. Especially under the low-angle light conditions at dawn and dusk, long strip-shaped shadows are extremely likely to form, resulting in the lack of a large number of target structure details in the collected images, and ultimately affecting the modeling texture matching and facade reconstruction accuracy. In addition, the original scheme lacks adjustment strategies for high-reflection areas, such as glass curtain walls and water surfaces. In these scenarios, it is easy to cause overexposure, artifacts, or even image failure due to strong reflections, and in severe cases, it will lead to local defects in the mapping results or the interruption of model reconstruction, unable to meet the requirements of high-precision three-dimensional modeling tasks. Summary of the Invention

[0006] In view of the deficiencies of the prior art, the present invention provides a method for UAV mapping with multi-angle photography, which solves the problems mentioned in the background art.

[0007] To achieve the above objectives, the present invention is realized through the following technical solutions: A method for UAV mapping with multi-angle photography, including the following steps:

[0008] S1. Collect the solar incident parameters of the current UAV operation area, and combine them to generate an illumination environment data set Lenv;

[0009] S2. Extract the edge morphological features from the mapping target ground object image, construct the main structure direction vector Vec, and combine it with the solar incident vector Sun to calculate the angle Spd between the structure direction and the incident light;

[0010] S3. Collect the heading angle Yaw and attitude inclination angle Ptc of the current UAV, calculate the heading correction angle Alp between the solar incident direction, construct the illumination effective projection score Ill, and output the illumination adaptability score Gls under the current angle combination;

[0011] S4. Calculate multiple illumination adaptability scores Gls based on different combinations of the heading angle Yaw and attitude inclination angle Ptc, form a shooting angle score set Vgl, and select the attitude combination with the highest score in the shooting angle score set Vgl to send down for image acquisition;

[0012] S5. After performing image acquisition and shooting, obtain the image set Img, construct the three-dimensional model point cloud set Pcd, and calculate the illumination deficiency index Lck;

[0013] S6. When the illumination deficiency index Lck exceeds the set threshold, construct a feedback optimization parameter set Fbk and use it to update the attitude angle candidate space.

[0014] Preferably, the S1 includes S11 and S12;

[0015] S11. Through the current geographical location and timestamp information of the UAV operation, obtain the longitude and latitude (Lon, Lat) and time Utc, calculate the solar azimuth angle Azm and solar altitude angle Alt at the current moment, then construct the unit solar incident direction vector Dir to represent the direction of the sun's rays in the three-dimensional space coordinate system under the current space-time conditions, and through the illumination intensity sensing sensor mounted on the top of the UAV fuselage, collect the surface irradiance Irr at the current position in real time;

[0016] The unit solar incident direction vector Dir is obtained through the following construction formula:

[0017] Dir = [cos(Alt) * cos(Azm), cos(Alt) * sin(Azm), sin(Alt)], where Dir represents the solar incident direction vector, which is a three-dimensional unit vector specifically indicating the solar irradiation direction. cos represents the cosine function and sin represents the sine function;

[0018] Among them, the solar azimuth angle Azm is obtained through the following calculation formula:

[0019] ;

[0020] In the formula, SPAAzm represents the solar azimuth angle calculation sub-function based on the NRELSPA model, Lon represents longitude, Lat represents latitude, and Utc represents time;

[0021] The solar altitude angle Alt is obtained through the following calculation formula:

[0022] ;

[0023] In the formula, SPAAlt represents the solar altitude angle calculation sub-function based on the NRELSPA model;

[0024] S12. Package and integrate the obtained unit solar incident direction vector Dir, solar altitude angle Alt, solar azimuth angle Azm, and surface irradiance intensity Irr to construct an illumination environment parameter set Lenv, which is used to describe the solar irradiation direction characteristics and radiation intensity state under the current natural illumination conditions;

[0025] The specific form of the illumination environment parameter set Lenv is the illumination environment parameter set Lenv = {Dir, Alt, Azm, Irr}.

[0026] Preferably, the S2 includes S21;

[0027] S21. Based on the image structure recognition algorithm, perform feature recognition processing on the target ground object in the surveying and mapping image, extract the contour edge information of the target ground object, and combine the image geometric parameters and camera pose information to restore the starting and ending points of the edge line segment from pixel coordinates to three-dimensional space coordinates, which are respectively marked as the starting point coordinate Pts and the ending point coordinate Pte. After integrating all edge line segments, obtain an edge feature set EdgeSet. By performing normalization processing and direction averaging on the edge feature set EdgeSet, construct a main structure direction vector Vec representing the main orientation of the target ground object and the normal direction of the main facade, which is used to represent the main orientation characteristics of the target ground object in the current perspective or the normal direction of its main facade;

[0028] The specific form of the edge feature set EdgeSet is the edge feature set EdgeSet = {E(1), E(2), ……, E(i)|i ∈ N}, where N represents the total number of edge segments. Among them, E(i) represents the edge feature unit of the i-th edge segment, specifically E(i) = {Pts, Pte}, Pts represents the spatial coordinates (x(1), y(1), z(1)) of the starting point of the edge segment, and Pte represents the spatial coordinates (x(2), y(2), z(2)) of the ending point of the edge segment;

[0029] The main structure direction vector Vec is obtained through the following calculation formula:

[0030] ;

[0031] In the formula, (x(i,1), y(i,1), z(i,1)) represents the spatial coordinates of the starting point of the i-th edge segment, and (x(i,2), y(i,2), z(i,2)) represents the spatial coordinates of the ending point of the i-th edge segment.

[0032] Preferably, S2 includes S22;

[0033] S22. Perform direction matching based on the obtained main structure direction vector Vec and the unit solar incident direction vector Dir, and calculate the included angle value Spd between the main structure direction vector Vec and the unit solar incident direction vector Dir through the three-dimensional space vector included angle formula;

[0034] The included angle value Spd is obtained through the following calculation formula:

[0035] ;

[0036] In the formula, arccos represents the inverse cosine function, represents the vector dot product operation.

[0037] Preferably, S3 includes S31;

[0038] S31. Collect the heading angle Yaw and the attitude inclination angle Ptc of the current UAV, and combine with the unit solar incident direction vector Dir to calculate the heading offset angle Alp between the camera view direction of the current UAV and the solar irradiation direction, which reflects the difference between the current heading configuration and the main direction of illumination. Based on the obtained heading offset angle Alp, further construct the illumination effective projection score Ill with the attitude inclination angle Ptc to quantify the projectability effectiveness of the solar incident light on the surface of the target ground object under the current body attitude;

[0039] The heading offset angle Alp is obtained through the following calculation formula:

[0040] ;

[0041] Wherein, Cam represents a three-dimensional unit line-of-sight vector, which is specifically obtained by converting the yaw angle and the pitch angle Ptc of the UAV camera.

[0042] The specific conversion formula of the three-dimensional unit line-of-sight vector Cam is as follows:

[0043] ;

[0044] Wherein, cos represents the cosine function, and sin represents the sine function.

[0045] The effective light projection score Ill is obtained by constructing a formula.

[0046] Preferably, S3 includes S32;

[0047] S32: Using the effective light projection score Ill and the included angle value Spd as the evaluation basic data, calculating the light adaptability score Gls under the current attitude angle combination through constructing a light adaptability evaluation function.

[0048] The light adaptability score Gls is obtained by the following calculation formula:

[0049] ;

[0050] Wherein, γ represents a penalty coefficient, which is specifically used to suppress the scoring deviation when the attitude direction is inconsistent with the ground object direction, and the specific value is set by the user.

[0051] Preferably, S4 includes S41;

[0052] S41: Based on the unit solar incident direction vector Dir and the main structure direction vector Vec, constructing a candidate shooting attitude parameter set PoseSet based on the combinations of multiple yaw angles and pitch angles Ptc within the controllable angle range of the UAV, calculating the light adaptability score Gls for each combined attitude in the candidate shooting attitude parameter set PoseSet, and then integrating the light adaptability scores Gls of each combined attitude in the candidate shooting attitude parameter set PoseSet to form a score set Vgl.

[0053] Among them, the specific form of the candidate shooting pose parameter set PoseSet is the candidate shooting pose parameter set PoseSet = {Pose(1), Pose(2), ……, Pose(j)|j ∈ M}, where M represents the total number of combined poses, and Pose(j) represents the j-th group of combined poses, specifically Pose(j) = {Yaw(j), Ptc(j)}, and Yaw(j) and Ptc(j) respectively represent the yaw angle Yaw and the attitude pitch angle Ptc of the j-th group of combined poses;

[0054] The specific form of the scoring set Vgl is the scoring set Vgl = {Gls(1), Gls(1), ……, Gls(j)|j ∈ M}, where Gls(j) represents the illumination adaptability score Gls obtained by calculating the j-th group of combined poses.

[0055] Preferably, S4 includes S42;

[0056] S42. Screen out the pose combination Pose(j) with the largest score value from the constructed scoring set Vgl, and then extract the yaw angle Yaw(j) and the attitude pitch angle Ptc(j) of the j-th combined pose from the pose combination Pose(j). The pose combination Pose(j) has the best illumination matching and shooting potential under the current illumination conditions. Generate a pose control instruction with the pose combination Pose(j), where the instruction contains the target yaw angle Yaw(j) and the attitude pitch angle Ptc(j), and send it to the execution module through the UAV flight control system to drive the UAV to adjust to the state of the pose combination Pose(j) and then execute the image acquisition task.

[0057] Preferably, S5 includes S51;

[0058] S51. Based on the image set Img captured by executing the image acquisition task, extract the shaded pixel regions with brightness lower than the global mean for each image in the image set Img, and mark the pixel positions of the shaded pixel regions in the image plane. Convert the pixel position coordinates of all pixels in each image into three-dimensional space coordinates through pixel backprojection to form a point cloud set Pcd of the three-dimensional model of each image. Extract the total number Ntot of image pixels of each image based on the point cloud set Pcd, and then extract the total number Nshd of shaded pixels marked as shaded pixel regions. Calculate the proportion of the total number Nshd of shaded pixels in the total number Ntot of image pixels to obtain the illumination defect index Lck.

[0059] Preferably, S6 includes S61;

[0060] S61. Based on a preset light occlusion threshold LckThr, when the light incompleteness index Lck exceeds the light occlusion threshold LckThr, it indicates that there is shadow occlusion in the current image acquisition pose combination of the UAV, which affects the image imaging quality and modeling integrity, and triggers a feedback optimization mechanism;

[0061] The feedback optimization mechanism extracts the pose combination Pose(j), included angle value Spd, light adaptability score Gls, and light incompleteness index Lck of the current UAV shot, combines and constructs a feedback optimization parameter set Fbk, and uses the feedback optimization parameter set Fbk as the input basis for the optimization strategy to update the pose angle candidate space;

[0062] Updating the pose angle candidate space includes expanding the value range of the pose tilt angle Ptc in the pose candidate space and finely adjusting and correcting the heading angle Yaw according to the direction deviation characteristics of the included angle value Spd.

[0063] The present invention provides a UAV mapping method for multi-angle photography, which has the following beneficial effects

[0064] (1) By forming a shooting angle score set Vgl, a quantitative evaluation of light adaptability in the full pose domain is realized. By selecting the pose combination with the highest score in the score set Vgl for image acquisition, not only the imaging quality is improved, but also a data basis with the optimal lighting conditions is provided for modeling. Further, after the image acquisition is completed, a point cloud set Pcd is constructed through the image set Img, and the light incompleteness index Lck is calculated to quantitatively evaluate the influence of light occlusion in the model; if the light incompleteness index Lck exceeds the light occlusion threshold LckThr, the pose angle candidate space is updated and an iterative optimization process is driven, effectively solving the problems of large image shadow interference, exposure imbalance, and three-dimensional model defects caused by factors such as natural light dynamic changes, uncontrollable poses, and ground object structure occlusion in the existing mapping UAV methods.

[0065] (2) Calculate the heading offset angle Alp between the current body orientation and the unit solar incident direction vector Dir, and then construct the effective illumination projection score Ill that reflects the illumination projection efficiency to comprehensively characterize the projection effectiveness of the illumination on the surface of the target ground object in the current posture. On this basis, input the effective illumination projection score Ill and the included angle value Spd calculated from the main structure direction vector Vec and the unit solar incident direction vector Dir into the illumination adaptability evaluation function to generate the illumination adaptability score Gls, which clearly indicates whether the current posture angle combination meets the illumination condition requirements for image acquisition. This process not only combines the ground object structure features with the flight posture, but also introduces natural illumination factors into the posture angle control logic, significantly improving the adaptive adjustment ability of the UAV's body orientation configuration and the discrimination accuracy of image acquisition timing selection when performing mapping tasks under complex illumination conditions.

[0066] (3) By taking the unit solar incident direction vector Dir and the main structure direction vector Vec as the input reference, traverse all combinations of the heading angle Yaw and the attitude pitch angle Ptc within the current controllable angle range of the UAV to construct the candidate shooting posture parameter set PoseSet, and calculate the illumination adaptability score Gls for each posture combination in turn, thereby forming the score set Vgl, realizing the global scoring modeling of the responsiveness of multi-angle postures to the current illumination environment. Further, by selecting the posture combination Pose(j) with the maximum score value in the score set Vgl as the optimal configuration of the current mapping task in the illumination adaptation dimension, and generating posture control instructions through the flight control system to adjust the UAV flight posture in real time, so that the image acquisition task always carries out around the optimal illumination angle and the target ground object structure direction. This method not only avoids the problems of shadow occlusion and uneven exposure caused by relying on fixed flight paths or one-way shooting in the traditional path, but also has high flexibility and real-time performance in actual flight operations, can dynamically adapt to the change of the sun angle, and ensure that each image acquisition process has the maximum imaging advantage under the illumination geometric conditions. Description of the Drawings

[0067] Figure 1 Schematic diagram of the steps of a UAV mapping method for multi-angle photography according to the present invention

[0068] Figure 2 Schematic diagram of the posture combination scoring and optimal screening mechanism. Detailed Embodiment

[0069] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0070] Embodiment 1

[0071] The present invention provides a method for unmanned aerial vehicle mapping with multi-angle photography. Please refer to Figure 1 , which includes the following steps:

[0072] S1. Collect the solar incident parameters of the current unmanned aerial vehicle operation area and combine them to generate an illumination environment data set Lenv;

[0073] S2. Extract the edge morphological features from the surveyed target ground object image, construct a main structure direction vector Vec, and combine it with the solar incident vector Sun to calculate the angle Spd between the structure direction and the incident light;

[0074] S3. Collect the heading angle Yaw and attitude inclination angle Ptc of the current unmanned aerial vehicle, calculate the heading correction angle Alp between the solar incident direction, construct an illumination effective projection score Ill, and output the illumination adaptability score Gls under the current angle combination;

[0075] S4. Calculate multiple illumination adaptability scores Gls based on different combinations of the heading angle Yaw and the attitude inclination angle Ptc to form a shooting angle score set Vgl, and select the attitude combination with the highest score in the shooting angle score set Vgl to send down for image acquisition;

[0076] S5. After performing image acquisition and shooting, obtain an image set Img, construct a three-dimensional model point cloud set Pcd, and calculate the illumination defect index Lck;

[0077] S6. When the illumination defect index Lck exceeds the set threshold, construct a feedback optimization parameter set Fbk and use it to update the attitude angle candidate space.

[0078] In this embodiment, by generating the light environment data set Lenv, an accurate description basis for the current natural light environment is established; subsequently, combined with the edge structure recognition algorithm, the main structure direction vector Vec of the target ground object is extracted, and the included angle Spd between the structure direction and the light direction is jointly calculated with the solar incident vector Dir, realizing the evaluation of the matching degree between the geometric structure of the photographed object and the light; at the same time, the heading angle Yaw and the attitude inclination angle Ptc of the UAV are collected, the heading correction angle Alp between the camera orientation and the light direction is constructed, and the light effective projection score Ill is constructed based on its matching with the structure characteristics, and the light adaptation score Gls is output, and then the shooting angle score set Vgl is formed, realizing the quantitative evaluation of the light adaptation in the full attitude domain. By selecting the attitude combination with the highest score in the score set Vgl for image acquisition, not only the imaging quality is improved, but also the optimal data basis of the light condition is provided for modeling. Further, after the image acquisition is completed, the point cloud set Pcd is constructed through the image set Img, and the light deficiency index Lck is calculated to quantitatively evaluate the influence of light occlusion in the model; if the light deficiency index Lck exceeds the light occlusion threshold LckThr, the optimal attitude combination, the included angle value Spd, the light adaptation score Gls and the light deficiency index Lck are used to form the feedback optimization parameter set Fbk, update the attitude angle candidate space and drive the iterative optimization process, effectively solving the problems of large image shadow interference, exposure imbalance and three-dimensional model defects caused by factors such as natural light dynamic changes, uncontrollable attitude and ground object structure occlusion in the existing mapping UAV methods, realizing the intelligent acquisition optimization of UAV images driven by light, and having wide practicability and engineering adaptability.

[0079] Embodiment 2

[0080] Specifically: S1 includes S11 and S12;

[0081] S11. Through the current geographical location and timestamp information of the UAV operation, the longitude and latitude (Lon, Lat) and the time Utc are obtained, the solar azimuth angle Azm and the solar altitude angle Alt at the current moment are calculated, and then the unit solar incident direction vector Dir is constructed to represent the direction of the sun's rays in the three-dimensional space coordinate system under the current spatio-temporal conditions, and the surface irradiance Irr at the current position is collected in real time through the light intensity sensing sensor mounted on the top of the UAV fuselage;

[0082] The unit solar incident direction vector Dir is obtained through the following construction formula:

[0083] Dir = [cos(Alt) * cos(Azm), cos(Alt) * sin(Azm), sin(Alt)] represents the solar incident direction vector, which is a three-dimensional unit vector and specifically represents the solar irradiation direction. cos represents the cosine function and sin represents the sine function;

[0084] Among them, the solar azimuth Azm is obtained through the following calculation formula:

[0085] ;

[0086] In the formula, SPAAzm represents the solar azimuth calculation sub-function based on the NRELSPA model, which is specifically used to calculate the projection direction of the sun on the horizontal plane under the current geographical location and time conditions. Lon represents longitude, Lat represents latitude, and Utc represents time;

[0087] The solar altitude angle Alt is obtained through the following calculation formula:

[0088] ;

[0089] In the formula, SPAAlt represents the solar altitude angle calculation sub-function based on the NRELSPA model, which is specifically used to calculate the elevation angle of the sun's rays relative to the horizon;

[0090] S12. Package and integrate the obtained unit solar incident direction vector Dir, solar altitude angle Alt, solar azimuth Azm, and surface irradiation intensity Irr to construct an illumination environment parameter set Lenv, which is used to describe the solar irradiation direction characteristics and radiation intensity state under the current natural illumination conditions. The illumination environment parameter set Lenv is used as standard input data for subsequent steps such as ground object structure direction analysis, attitude adjustment angle derivation, and image acquisition condition judgment in multi-angle mapping processes related to illumination;

[0091] The specific form of the illumination environment parameter set Lenv is the illumination environment parameter set Lenv = {Dir, Alt, Azm, Irr}.

[0092] In this embodiment, an input foundation for dynamic adaptation to the natural environment is established. Specifically, the solar azimuth angle Azm and solar altitude angle Alt are calculated using the current geographical location and time information of the UAV, and a unit solar incident direction vector Dir is further constructed to achieve a three-dimensional spatial representation of the solar irradiation direction under the current spatio-temporal conditions. At the same time, combined with the surface irradiance intensity Irr obtained in real-time by the light intensity sensing sensor mounted on the top of the UAV, the system effectively integrates key physical parameters related to light to form a standardized input structure light environment parameter set Lenv, providing a unified, accurate, and quantifiable light model input for the entire multi-angle photogrammetry process. This enables core calculation processes such as the analysis of the main structure direction of the target ground object, the adjustment of the attitude angle, and the light matching score in subsequent steps to have data consistency and spatio-temporal synchronization, significantly improving the adaptability of the system under non-ideal natural conditions and the real-time decision-making accuracy of attitude control, and laying a data benchmark guarantee at the light level for the subsequent modeling reliability and angle optimization.

[0093] Embodiment 3

[0094] Specifically: S2 includes S21;

[0095] S21. Based on the image structure recognition algorithm, perform feature recognition processing on the target ground object in the surveying and mapping image, extract the contour edge information of the target ground object, and combine the image geometric parameters and camera pose information to restore the start and end points of the edge line segment from pixel coordinates to three-dimensional space coordinates, respectively marked as the start point coordinate Pts and the end point coordinate Pte. After integrating all edge line segments, obtain the edge feature set EdgeSet. By performing normalization processing and direction averaging on the edge feature set EdgeSet, construct a main structure direction vector Vec representing the main orientation of the target ground object and the normal direction of the main facade, which is used to represent the main orientation characteristics of the target ground object or the normal direction of its main facade under the current perspective. The main structure direction vector Vec is the core comparison benchmark for subsequent solar light direction deviation analysis;

[0096] The specific form of the edge feature set EdgeSet is the edge feature set EdgeSet = {E(1), E(2), ……, E(i)|i ∈ N}, where N represents the total number of edge line segments. Among them, E(i) represents the edge feature unit of the i-th edge line segment, specifically E(i) = {Pts, Pte}, Pts represents the spatial coordinates (x(1), y(1), z(1)) of the start point of the edge line segment, and Pte represents the spatial coordinates (x(2), y(2), z(2)) of the end point of the edge line segment;

[0097] The main structure direction vector Vec is obtained through the following calculation formula:

[0098] ;

[0099] In the formula, (x(i, 1), y(i, 1), z(i, 1)) represents the spatial coordinates of the starting point of the i-th edge segment, and (x(i, 2), y(i, 2), z(i, 2)) represents the spatial coordinates of the ending point of the i-th edge segment; represents the direction vector of the line segment (ending point minus starting point), represents normalizing the magnitude of the direction vector to obtain a unit vector;

[0100] The main structure direction vector Vec is a unit spatial direction vector used to represent the main structure orientation or the normal direction of the main facade of the target ground object in the current perspective in the surveyed image. It is specifically obtained by calculating the direction vector from the starting point coordinates Pts and the ending point coordinates Pte of each edge segment in the edge feature set EdgeSet extracted from the image and then performing normalization processing. The main structure direction vector Vec can accurately reflect the most significant structural orientation characteristics of the target ground object in three-dimensional space and is an important geometric reference parameter for performing solar irradiation direction deviation analysis, attitude adjustment, and illumination matching evaluation. It is used as a key input vector in the subsequent calculation of the included angle value Spd.

[0101] S2 includes S22;

[0102] S22: Perform direction matching based on the obtained main structure direction vector Vec and the unit solar incident direction vector Dir, and calculate the included angle value Spd between the main structure direction vector Vec and the unit solar incident direction vector Dir through the three-dimensional space vector included angle formula, which is used to represent the deviation degree between the solar irradiation direction and the main structure direction of the ground object under the current illumination conditions. The included angle value Spd is an important indicator reflecting the illumination adaptability: when the value of the included angle value Spd is small, it indicates that the sunlight irradiation direction approaches the main structure plane of the target ground object, and higher illumination sufficiency and less shadow occlusion can be obtained; conversely, when the included angle value Spd is large, there is an obvious backlight or backlighting effect, and the image imaging quality will be affected, and optimization processing needs to be carried out through attitude adjustment or shooting strategy delay.

[0103] The included angle value Spd is obtained through the following calculation formula:

[0104] ;

[0105] In the formula, arccos represents the inverse cosine function, which is used to restore the cosine value to the included angle value, represents the vector dot product operation, which specifically represents calculating the direction similarity between two vectors.

[0106] In this embodiment, by matching the spatial relationship between the main structure direction vector Vec and the unit solar incident direction vector Dir, a quantization index included angle value Spd based on the deviation degree of the incident angle of illumination is established. The main structure direction vector Vec is calculated through the spatial starting point coordinates Pts and the ending point coordinates Pte of each edge segment in the edge feature set EdgeSet, and has geometric representativeness that combines structural feature expression and illumination coupling. It not only realizes the spatial relationship modeling between the solar irradiation direction and the main structure of the ground object, but also can dynamically determine whether there are adverse factors such as backlight, backlighting or side light under the current illumination conditions, and has a pre-intervention value for the image brightness uniformity, shadow occlusion degree and imaging geometric accuracy. The setting of this step effectively improves the geometric accuracy of the recognition of the ground object structure direction during the multi-angle photography process, and enhances the rationality and initiative of the shooting attitude adjustment under the scenes of unstable illumination or complex structure orientations, constituting an important criterion source for the subsequent attitude optimization control.

[0107] Embodiment 4

[0108] Specifically: S3 includes S31;

[0109] S31. Collect the heading angle Yaw and the attitude inclination angle Ptc of the current unmanned aerial vehicle (UAV), and combine with the unit solar incident direction vector Dir to calculate the heading offset angle Alp between the camera viewing direction of the current UAV and the solar irradiation direction, which reflects the difference between the current heading configuration and the main direction of illumination. Based on the obtained heading offset angle Alp, an illumination effective projection score Ill is constructed with the attitude inclination angle Ptc to quantify the projectability effectiveness of the solar incident light on the surface of the target ground object under the current body attitude.

[0110] The heading offset angle Alp is obtained through the following calculation formula:

[0111] ;

[0112] In the formula, Cam represents a three-dimensional unit line-of-sight vector, which is specifically obtained by converting the heading angle Yaw and the attitude pitch angle Ptc of the UAV camera.

[0113] The three-dimensional unit line-of-sight vector Cam is specifically converted through the following calculation formula:

[0114] ;

[0115] In the formula, cos represents the cosine function, and sin represents the sine function.

[0116] The illumination effective projection score Ill is obtained through Build formula acquisition, where the effective light projection score Ill is a score value in the range of [−1, 1]. Usually, the positive value range [0, 1] is actually concerned. The closer the score is to 1, the more suitable the camera pose is for the sun direction, which is beneficial to light projection.

[0117] S3 includes S32;

[0118] S32. Take the effective light projection score Ill and the included angle value Spd as the evaluation basic data, and calculate the light adaptability score Gls under the current pose angle combination through constructing a light adaptability evaluation function, which is used to jointly evaluate whether the current body orientation is suitable for performing the image acquisition task;

[0119] The light adaptability score Gls is obtained through the following calculation formula:

[0120] ;

[0121] In the formula, γ represents the penalty coefficient, which is specifically used to suppress the scoring deviation when the attitude direction is inconsistent with the ground object direction, and the specific value is set by the user;

[0122] The light adaptability score Gls comprehensively considers the geometric consistency between the current UAV pose angle and the sun incident direction, the irradiation coordination between the target ground object structure plane and the light direction, and the spatial synchronization between the camera orientation and the target illumination direction, and is used to quantify the imaging adaptability degree of the current shooting pose under the light condition. The higher the light adaptability score Gls value, the more suitable the current pose is for performing the mapping image acquisition task.

[0123] In this embodiment, the quantitative evaluation of the adaptability between the current UAV body pose and the sun incident direction is realized, effectively solving the problem that the image light projection efficiency decreases due to the unreasonable configuration of the heading angle Yaw and the attitude tilt angle Ptc. Specifically, first collect the heading angle Yaw and the attitude tilt angle Ptc of the current UAV, calculate the heading offset angle Alp between the current body orientation and the unit sun incident direction vector Dir, and then construct the effective light projection score Ill reflecting the light projection efficiency, comprehensively depicting the projection effectiveness of the light on the target ground object surface under the current pose. On this basis, the effective light projection score Ill and the included angle value Spd calculated by the main structure direction vector Vec and the unit sun incident direction vector Dir are jointly input into the light adaptability evaluation function to generate the light adaptability score Gls, clearly indicating whether the current pose angle combination meets the light condition requirements for image acquisition. This process not only combines the ground object structure characteristics with the flight pose, but also introduces natural light factors into the attitude angle control logic, significantly improving the UAV's adaptive adjustment ability for the body orientation configuration and the discrimination accuracy of the image acquisition timing selection when performing mapping tasks under complex light conditions.

[0124] Example 5

[0125] Please refer to Figure 1 and Figure 2 , specifically: S4 includes S41;

[0126] S41. Based on the unit solar incident direction vector Dir and the main structure direction vector Vec, a candidate shooting pose parameter set PoseSet is constructed based on combinations of multiple heading angles Yaw and attitude pitch angles Ptc within the controllable angle range of the UAV. For each combined pose in the candidate shooting pose parameter set PoseSet, a lighting adaptability score Gls is calculated. Then, the lighting adaptability scores Gls of each combined pose in the candidate shooting pose parameter set PoseSet are integrated to form a score set Vgl, which is used to map all feasible pose combinations and provide a unified shooting preference basis under lighting conditions;

[0127] Among them, the specific form of the candidate shooting pose parameter set PoseSet is candidate shooting pose parameter set PoseSet = {Pose(1), Pose(2), ……, Pose(j)|j ∈ M}, where M represents the total number of combined poses, and Pose(j) represents the jth combined pose, specifically Pose(j) = {Yaw(j), Ptc(j)}, and Yaw(j) and Ptc(j) respectively represent the heading angle Yaw and attitude pitch angle Ptc of the jth combined pose;

[0128] The specific form of the score set Vgl is score set Vgl = {Gls(1), Gls(1), ……, Gls(j)|j ∈ M}, and Gls(j) represents the lighting adaptability score Gls obtained by calculating the jth combined pose.

[0129] S4 includes S42;

[0130] S42. Select the pose combination Pose(j) with the maximum score value from the constructed score set Vgl, and then extract the heading angle Yaw(j) and attitude pitch angle Ptc(j) of the jth combined pose from the pose combination Pose(j). The pose combination Pose(j) has the best lighting matching and shooting potential under the current lighting conditions. Generate a pose control instruction for the pose combination Pose(j), and the instruction contains the target heading angle Yaw(j) and attitude pitch angle Ptc(j), and send it to the execution module through the UAV flight control system to drive the UAV to adjust to the state of the pose combination Pose(j) and then execute the image acquisition task, ensuring that in the natural light dynamic change environment, the mapping image acquisition can always preferentially match the best lighting conditions and structural angles, improving the geometric accuracy and visual quality of the image.

[0131] Example illustration of attitude combination scoring and optimization:

[0132] Assume that the unit solar incident direction vector Dir and the main structure direction vector Vec have been obtained in the current lighting environment. The system traverses the range of the heading angle Yaw and the attitude pitch angle Ptc allowed for the UAV to form the following candidate shooting attitude parameter set PoseSet: Table 1: Candidate shooting attitude parameter set PoseSet

[0133] Pose(j) of pose combination Yaw(j) of heading angle Ptc(j) of attitude pitch angle Ill(j) of effective light projection score Spd(j) of included angle value Gls(j) of light adaptability score Pose(1) 45° 10° 0.72 12° 0.66 Pose(2) 60° 15° 0.81 8° 0.77 Pose(3) 90° 12° 0.65 20° 0.50 Pose(4) 30° 5° 0.60 25° 0.38

[0134] Combined with the above scoring data, a scoring set Vgl is formed:

[0135] Scoring set Vgl = {Gls(1)=0.66, Gls(2)=0.77, Gls(3)=0.50, Gls(4)=0.38};

[0136] It can be seen from the scoring set Vgl that the attitude combination with the highest score is Pose(2).

[0137] In this embodiment, taking the unit solar incident direction vector Dir and the main structure direction vector Vec as the input reference, traversing all combinations of the heading angle Yaw and the attitude pitch angle Ptc within the current controllable angle range of the UAV, constructing the candidate shooting attitude parameter set PoseSet, and calculating the lighting adaptability score Gls of each attitude combination in turn, so as to form the scoring set Vgl, realizing the global scoring modeling of the responsiveness of multi-angle attitudes to the current lighting environment. Further, by selecting the attitude combination Pose(j) with the largest scoring value in the scoring set Vgl as the optimal configuration for the current mapping task in the lighting adaptation dimension, and generating attitude control commands through the flight control system to adjust the UAV flight attitude in real time, the image acquisition task is always carried out around the optimal lighting angle and the target ground object structure direction. This method not only avoids the problems of shadow occlusion and uneven exposure caused by relying on fixed flight paths or one-way shooting in the traditional path, but also has high flexibility and real-time performance in actual flight operations, can dynamically adapt to the change of the sun angle, and ensures that the maximum imaging advantage under the lighting geometry conditions is available in each image acquisition process.

[0138] Example 6

[0139] Specifically: The S5 includes S51;

[0140] S51. Based on the image set Img captured by performing the image acquisition task, for each image in the image set Img, extract the shaded pixel region with brightness lower than the global mean, mark the pixel positions of the shaded pixel region in the image plane, convert the pixel position coordinates of all pixels in each image into three-dimensional space coordinates through pixel back-projection to form a point cloud set Pcd of the three-dimensional model of each image, extract the total number Ntot of image pixels of each image based on the point cloud set Pcd, then extract the total number Nshd of shaded pixels marked as the shaded pixel region, and obtain the illumination defect index Lck by calculating the ratio of the total number Nshd of shaded pixels to the total number Ntot of image pixels.

[0141] The above S6 includes S61;

[0142] S61. Based on the preset illumination occlusion threshold LckThr, when the illumination defect index Lck exceeds the illumination occlusion threshold LckThr, it indicates that there is shadow occlusion in the current image acquisition attitude combination of the UAV, which affects the image imaging quality and modeling integrity, and triggers a feedback optimization mechanism;

[0143] The feedback optimization mechanism extracts the attitude combination Pose(j), included angle value Spd, illumination adaptability score Gls, and illumination defect index Lck of the current UAV capture, combines and constructs a feedback optimization parameter set Fbk, and uses the feedback optimization parameter set Fbk as the input basis of the optimization strategy to update the attitude angle candidate space;

[0144] The update of the attitude angle candidate space includes expanding the value range of the pitch angle Ptc in the attitude candidate space and fine-tuning and correcting the yaw angle Yaw according to the direction deviation characteristics of the included angle value Spd.

[0145] In this embodiment, after obtaining the image set Img, the shadow pixel region is extracted based on the image brightness feature, the 3D point cloud set Pcd of the image is constructed by using the pixel back-projection method, and the total number Nshd of shadow pixels and the total number Ntot of image pixels are counted. Furthermore, the illumination deficiency index Lck is calculated to quantitatively characterize the degree of image shadow interference under the current pose combination. On this basis, with the illumination occlusion threshold LckThr as the criterion, when the illumination deficiency index Lck exceeds the threshold, the feedback optimization mechanism is triggered, and the feedback optimization parameter set Fbk is constructed by combining the pose combination Pose(j), the included angle value Spd, the illumination adaptability score Gls, and the illumination deficiency index Lck corresponding to the current image, so as to drive the dynamic adjustment of the pose candidate space, including expanding the value range of the pose tilt angle Ptc and fine-tuning and correcting the heading angle Yaw based on the direction deviation feature. This mechanism not only establishes a closed-loop evaluation system for the modeling quality after image acquisition and the illumination adaptability of the image, but also realizes the intelligent optimization and update of the pose parameters, provides a more illumination-responsive angle scheduling guarantee for the subsequent image acquisition and modeling process, solves the problem that the image imperfection cannot be perceived under the traditional static path, and has the comprehensive advantages of high adaptability, high stability, and strong scene perception ability.

[0146] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A UAV mapping method for multi-angle photography, characterized in that: It includes the following steps: S1. Collect the solar incident parameters of the current UAV operation area, and combine them to generate the illumination environment dataset Lenv; S2. Extract the edge morphological features from the surveyed target ground object image, construct the main structure direction vector Vec, and combine it with the solar incident vector Sun to calculate the angle Spd between the structure direction and the incident light; S3. Collect the heading angle Yaw and attitude inclination angle Ptc of the current UAV, calculate the heading correction angle Alp between the UAV and the solar incident direction, construct the illumination effective projection score Ill, and output the illumination adaptability score Gls under the current angle combination; S4. Calculate multiple illumination adaptability scores Gls based on different combinations of the heading angle Yaw and the attitude inclination angle Ptc, form the shooting angle score set Vgl, and select the attitude combination with the highest score in the shooting angle score set Vgl to send and execute image acquisition; S5. After performing image acquisition and shooting, obtain the image set Img, construct the 3D model point cloud set Pcd, and calculate the illumination deficiency index Lck; S6. When the illumination deficiency index Lck exceeds the set threshold, construct the feedback optimization parameter set Fbk and use it to update the attitude angle candidate space.

2. The method for UAV mapping with multi-angle photography according to claim 1, characterized in that: The S1 includes S11 and S12; S11. Through the current geographical location and timestamp information of the UAV operation, obtain the longitude and latitude (Lon, Lat) and time Utc, calculate the solar azimuth angle Azm and solar altitude angle Alt at the current moment, then construct the unit solar incident direction vector Dir, and use the light intensity sensing sensor mounted on the top of the UAV fuselage to collect the surface irradiance Irr at the current position in real time; The unit solar incident direction vector Dir is obtained through the following construction formula: Dir = [cos(Alt) * cos(Azm), cos(Alt) * sin(Azm), sin(Alt)], where Dir represents the solar incident direction vector, which is a three-dimensional unit vector, specifically representing the solar irradiation direction, cos represents the cosine function, and sin represents the sine function; S12. Package and integrate the obtained unit solar incident direction vector Dir, solar altitude angle Alt, solar azimuth angle Azm, and surface irradiance Irr to construct the illumination environment parameter set Lenv = {Dir, Alt, Azm, Irr}.

3. The method for UAV mapping with multi-angle photography according to claim 2, wherein: The S2 includes S21; S21. Based on the image structure recognition algorithm, perform feature recognition processing on the target ground object in the surveyed image, extract the contour edge information of the target ground object, restore the start and end points of the edge line segment from pixel coordinates to three-dimensional space coordinates, and mark them as the start point coordinate Pts and the end point coordinate Pte respectively. After integrating all edge line segments, obtain the edge feature set EdgeSet, and construct the main structure direction vector Vec representing the main orientation and the main facade normal direction of the target ground object through normalization processing and direction averaging of the edge feature set EdgeSet; The specific form of the edge feature set EdgeSet is the edge feature set EdgeSet = {E(1), E(2), ……, E(i)|i ∈ N}, where N represents the total number of edge segments. Among them, E(i) represents the edge feature unit of the i-th edge segment, specifically E(i) = {Pts, Pte}, Pts represents the spatial coordinates (x(1), y(1), z(1)) of the starting point of the edge segment, and Pte represents the spatial coordinates (x(2), y(2), z(2)) of the ending point of the edge segment.

4. The UAV mapping method for multi-angle photography according to claim 3, wherein: The above-mentioned S2 includes S22; S22. Perform direction matching based on the obtained main structure direction vector Vec and the unit solar incident direction vector Dir, and calculate the included angle value Spd between the main structure direction vector Vec and the unit solar incident direction vector Dir through the three-dimensional space vector included angle formula.

5. A method for UAV mapping with multi-angle photography according to claim 4, characterized in that: The above-mentioned S3 includes S31; S31. Collect the heading angle Yaw and the attitude inclination angle Ptc of the current unmanned aerial vehicle (UAV), and combine with the unit solar incident direction vector Dir to calculate the heading offset angle Alp between the camera viewing direction of the current UAV and the solar irradiation direction. Based on the obtained heading offset angle Alp, further construct the effective light projection score Ill with the attitude inclination angle Ptc. The effective light projection score Ill is obtained by constructing a formula.

6. The method for unmanned aerial vehicle mapping with multi-angle photography according to claim 5, characterized in that: The above-mentioned S3 includes S32; S32. Use the effective light projection score Ill and the included angle value Spd as the evaluation basic data, and calculate the light adaptability score Gls under the current attitude angle combination by constructing a light adaptability evaluation function. The light adaptability score Gls is obtained through the following calculation formula: ； In the formula, γ represents the penalty coefficient, which is specifically used to suppress the scoring deviation when the attitude direction is inconsistent with the ground object direction, and the specific value is set by the user.

7. The method for UAV mapping with multi-angle photography according to claim 6, characterized in that: The above-mentioned S4 includes S41; S41. Based on the unit solar incident direction vector Dir and the main structure direction vector Vec, construct a candidate shooting attitude parameter set PoseSet based on the combinations of multiple heading angles Yaw and attitude pitch angles Ptc within the controllable angle range of the UAV. Calculate the light adaptability score Gls for each combined attitude in the candidate shooting attitude parameter set PoseSet, and then integrate the light adaptability scores Gls of each combined attitude in the candidate shooting attitude parameter set PoseSet to form a score set Vgl.

8. The method for UAV mapping with multi-angle photography according to claim 7, characterized in that: The above-mentioned S4 includes S42; S42. Screen out the attitude combination Pose(j) with the maximum score value from the constructed score set Vgl, then extract the heading angle Yaw(j) and the attitude pitch angle Ptc(j) of the j-th combined attitude from the attitude combination Pose(j), generate an attitude control instruction for the attitude combination Pose(j), and send it to the execution module through the UAV flight control system to drive the UAV to adjust to the state of the attitude combination Pose(j) and then execute the image acquisition task.

9. The method for UAV mapping with multi-angle photography according to claim 8, wherein: The above-mentioned S5 includes S51; S51. Based on the image set Img captured by performing the image acquisition task, for each image in the image set Img, extract the shaded pixel region with brightness lower than the global mean, and mark the pixel positions of the shaded pixel region in the image plane. Convert the pixel position coordinates of all pixels in each image into three-dimensional space coordinates through pixel back-projection to form a point cloud set Pcd of the three-dimensional model of each image. Based on the point cloud set Pcd, extract the total number Ntot of image pixels of each image, and then extract the total number Nshd of shaded pixels marked as the shaded pixel region. By calculating the ratio of the total number Nshd of shaded pixels to the total number Ntot of image pixels, obtain the illumination defect degree index Lck.

10. A method for UAV mapping with multi-angle photography according to claim 9, characterized in that: The above S6 includes S61; S61. Based on the preset illumination shielding threshold LckThr, when the illumination defect degree index Lck exceeds the illumination shielding threshold LckThr, it indicates that there is shadow occlusion in the current image acquisition pose combination of the UAV, which affects the image imaging quality and modeling integrity, and triggers the feedback optimization mechanism; The feedback optimization mechanism extracts the pose combination Pose(j), the included angle value Spd, the illumination adaptability score Gls, and the illumination defect degree index Lck of the current UAV shot, combines and constructs a feedback optimization parameter set Fbk, and uses the feedback optimization parameter set Fbk as the input basis of the optimization strategy to update the pose angle candidate space.

Images