Unmanned aerial vehicle aerial photography path planning method for dynamic modeling

Abstract

The invention relates to the field of building surveying and mapping, and provides an unmanned aerial vehicle aerial photography path planning method for dynamic modeling, which comprises a target data acquisition step, a modeling structure analysis step, a surveying and mapping point income calculation step, an aerial photography path planning step and a path interference compensation step. By checking and avoiding interference obstacles of the unmanned aerial vehicle, the beneficial effect of improving the stability of the modeling aerial photography system of the unmanned aerial vehicle is achieved. Through accurate structural analysis, scientific income calculation and optimized path planning, scientificity, high efficiency and adaptability of unmanned aerial vehicle aerial photography path planning are comprehensively improved, and high-quality data support is provided for dynamic modeling.

Description

Technical Field

[0001] The invention relates to the field of architectural surveying and mapping, and in particular to a method for planning an unmanned aerial vehicle aerial photography path for dynamic modeling. Background Art

[0002] Traditional UAV aerial photography path planning methods often rely on static, preset flight paths, which lack sufficient consideration of the dynamic characteristics of the modeling target and environmental changes. For example, some methods may only set one or more fixed flight paths for image acquisition based on the approximate location and range of the modeling target. This method is difficult to adapt to the complex shape and structural characteristics of the modeling target, and may lead to the omission or repeated acquisition of key information, thus affecting the accuracy and efficiency of modeling.

[0003] In addition, the existing UAV aerial photography path planning technology lacks flexibility and pertinence when facing modeling targets with different structures. For modeling targets with regular three-dimensional structures, such as buildings with regular shapes, traditional planning methods may not be able to effectively optimize the aerial photography path to fully utilize the symmetry and structural repeatability of these targets to reduce unnecessary flights and mapping work. For special-shaped three-dimensional structures, such as artistic buildings with complex curves and broken line turns, existing methods may find it difficult to accurately capture these shape changes, resulting in poor modeling results. In addition, existing technologies lack comprehensive consideration of environmental factors. UAVs may encounter various obstacles during aerial photography, such as trees and telephone poles, which may pose a threat to the flight path and safety of UAVs. Traditional path planning methods often lack real-time detection and avoidance strategies for obstacles on the aerial photography route, which may cause UAVs to collide during flight or need to frequently adjust the flight path, thereby reducing aerial photography efficiency and modeling quality. In addition, existing UAV aerial photography path planning methods are also insufficient in comprehensive benefit evaluation. During the aerial photography process, different surveying points may have different sampling information and surveying costs, such as pixel clarity, texture richness, feature prominence, UAV flight distance, flight tortuosity, and environmental information. Traditional planning methods often do not fully consider the impact of these factors on comprehensive benefits, making it difficult to optimize the aerial photography path and reduce surveying costs while ensuring modeling quality.

[0004] Therefore, in order to solve the problems existing in the above-mentioned prior art, the present invention proposes a UAV aerial photography path planning method for dynamic modeling to improve aerial photography efficiency, modeling quality and system stability. Summary of the invention

[0005] In view of the shortcomings of the prior art, the object of the present invention is to provide a UAV aerial photography path planning method for dynamic modeling.

[0006] To achieve the above object, the present invention provides the following technical solutions: A UAV aerial photography path planning method for dynamic modeling includes the following steps: A target data acquisition step, in which an image of the modeling target captured by a camera mounted on the drone is obtained as a model analysis image, and the size, shape and structural features of the modeling target are obtained according to the model analysis image as basic model information; A modeling structure analysis step, determining the structure type of the modeling target through a structure analysis model according to the basic model information, and obtaining basic surveying and mapping points of the modeling target according to the structure type combined with the basic model information analysis; The step of calculating the benefit of the surveying and mapping points is to obtain the sampling information and the surveying and mapping cost of the basic surveying and mapping points through the image analysis according to the model analysis, and to obtain the comprehensive benefit value of the basic surveying and mapping points through the comprehensive benefit model according to the sampling information and the surveying and mapping cost of the basic surveying and mapping points; The aerial photography path planning step is to divide the basic surveying points into regions and form several key surveying areas through the surveying point planning strategy according to the structural characteristics, prioritize the basic surveying points in each key surveying area according to the comprehensive benefit value, and select the basic surveying points according to the priority ranking results to complete the drone aerial photography path planning.

[0007] As a further improvement of the present invention, the structural analysis model obtains the structural type of the modeling target according to the shape and structural feature analysis of the modeling target, and the structural type includes conventional three-dimensional structure and special-shaped three-dimensional structure; when the structural analysis model determines that the similarity between the shape of the modeling target and the preset geometric parameters is greater than the preset threshold, and the number of symmetrical structures and structural repeating units in the structural features of the modeling target is greater than the preset threshold, the output structural type is a conventional three-dimensional structure, otherwise the output structural type is a special-shaped three-dimensional structure; when the structural type is a conventional three-dimensional structure, basic surveying and mapping points are selected from the geometric center plane, connection points and key feature points of each geometric shape of the modeling target according to the shape and size of the modeling target. When the structural type is a special-shaped three-dimensional structure, basic surveying and mapping points are set at special nodes according to the shape information of the modeling target, including densely selecting basic surveying and mapping points at positions where the curvature of the curve changes dramatically, and setting basic surveying and mapping points at the turning points of the broken line according to the model analysis image analysis.

[0008] As a further improvement of the present invention, the surveying and mapping benefit calculation strategy includes extracting the pixel clarity, texture richness and feature significance of the area where the basic surveying and mapping points are located in the model analysis image according to the structure type as sampling information, and analyzing the UAV flight distance, the flight tortuosity of the UAV at key points and the environmental information according to the size and shape of the modeling target to obtain the surveying and mapping cost, and calculating the sampling information of each surveying and mapping point and the surveying and mapping cost through a comprehensive benefit model to obtain the comprehensive benefit value of each basic surveying and mapping point.

[0009] As a further improvement of the present invention, the comprehensive income model is configured with: ; Among them, R i is the comprehensive benefit value of the i-th basic surveying and mapping point, C i is the pixel definition of the i-th basic mapping point, T i is the texture richness of the i-th basic mapping point, S i is the feature significance of the i-th basic mapping point, indicating the prominence of the feature of the mapping point; D i is the flight distance from the UAV to the i-th basic mapping point, W i is the flight tortuosity of the UAV at the i-th basic mapping point, indicating the flight difficulty of the UAV at this point; E i is the environmental impact factor of the i-th basic surveying and mapping point.

[0010] As a further improvement of the present invention, the aerial photography path planning step also includes dividing the key surveying and mapping areas according to the spatial coordinates of the basic surveying and mapping points through distribution density and the structural characteristics of the modeling target, constructing a comprehensive benefit matrix for the basic surveying and mapping points in each key surveying and mapping area based on the comprehensive benefit value, confirming the weight of each basic surveying and mapping point by analyzing the comprehensive benefit matrix, and sorting the basic surveying and mapping points from high to low according to the weight.

[0011] As a further improvement of the present invention, the path interference compensation step also includes: extracting the theoretical target image of the corresponding position of the theoretical target model according to the basic surveying and mapping points of the sampling image, extracting the image features of the sampling image and the theoretical target image respectively through edge detection and pixel coordinate calibration, the image features including the contour perimeter, area, shape descriptor, texture features and key point positions of the modeling target in the sampling image, statistically matching the spatial distribution of the key point position deviations in the two images to obtain a spatial distribution value, calculating the deviation coefficient through the interference deviation model according to the image feature difference between the sampling image and the theoretical target image, and setting a first-level deviation threshold, a second-level deviation threshold and a spatial distribution threshold, when the deviation coefficient is higher than the first-level deviation threshold and lower than the second-level deviation threshold, and the spatial distribution value is lower than the spatial distribution threshold, outputting that the basic surveying and mapping point has a microscopic deviation, and when the deviation coefficient is higher than the second-level deviation threshold, and the spatial distribution value is higher than the spatial distribution threshold, outputting that the basic surveying and mapping point has a global deviation.

[0012] As a further improvement of the present invention, the interference deviation model is configured as follows: ; Where D is the deviation coefficient, which is used to determine the degree of deviation between the sampled image and the theoretical target image; n represents the number of image pairs at the modeled target in the image, which is usually 1. When the basic mapping point needs to be sampled continuously, n is the number of corresponding sampled image pairs and theoretical target image pairs; and are the contour perimeters of the i-th sampled image and the theoretical target image, respectively, max( , ) means normalization by the maximum value; and are the area components of the i-th sample image and the theoretical target image in the j direction, respectively. The area component represents the length of the modeled target after mapping in the j direction. m represents the number of area components. and Represent the norm of the area vector of the i-th sampled image and the theoretical target image respectively; and denote the shape descriptors of the i-th sampled image and the theoretical target image, respectively, and d H Represents the Hausdorff distance between shape descriptors, which represents the distance measure between shape descriptors; and Respectively represent the texture feature values ​​of the i-th sampled image and the theoretical target image, is the domain of texture features; and Respectively represent the position vector of the kth key point of the i-th sampled image and the theoretical target image, p is the number of key points, and represents the norm of the set of all key point position vectors of the i-th sampled image and the theoretical target image, represents the sum of the distances from all key points to the centroid, α, β, γ, δ and ε represent the weight factors of each parameter, where the value of ε is greater than α, β, γ, δ, and α+β+γ+δ+ε=1.

[0013] As a further improvement of the present invention, the microscopic path adjustment strategy includes calculating the basic surveying and mapping point offset according to the contour perimeter, area, shape descriptor, texture feature and key point position deviation value of the modeling target in the sampling image and the theoretical target image, adjusting the basic surveying and mapping point position coordinates according to the basic surveying and mapping point offset, and re-acquiring the modeling target sampling image of the updated basic surveying and mapping point position, and the basic surveying and mapping point offset calculation configuration includes: ; in, is a comprehensive correction vector, which indicates the offset of the coordinates of the basic surveying and mapping points. (v PAx 、v PAy ) is the first correction vector based on the deviation of contour perimeter and area, where , , k P is the contour perimeter deviation correction coefficient, k A is the area deviation correction factor, is the contour perimeter deviation, is the area deviation, P0 is the reference contour perimeter of the modeled target in the theoretical target image, A0 is the reference area of ​​the modeled target in the theoretical target image, and θ represents the correction direction angle calculated based on the deviation of the contour perimeter and area. Jx 、v Jy ) is the second correction vector based on the deviation of the shape descriptor, , , where ΔH i is the i-th element in the difference vector matrix, ω i is the weight coefficient of the i-th element in the x direction, ω ＇ i is the weight coefficient of the i-th element in the y direction, (v Xx 、v Xy ) is the third correction vector based on the deviation of texture features, , , where k X is the texture deviation correction coefficient, is the correction direction angle determined according to the texture deviation distribution, (Δx, Δy) is the key point position deviation; As a further improvement of the present invention, the global path adjustment strategy includes calculating the benefit adjustment factor of the basic surveying and mapping point at the corresponding position of the sampling image according to the deviation of the contour perimeter, area, shape descriptor, texture feature and key point position, the benefit adjustment factor including texture richness adjustment factor, feature significance adjustment factor and flight distance adjustment factor, and recalculating the comprehensive benefit adjustment value of the basic surveying and mapping point through a comprehensive benefit recalculation model according to the comprehensive benefit adjustment factor, replacing the comprehensive benefit value of the basic surveying and mapping point with the recalculated comprehensive benefit adjustment value as the final comprehensive benefit value, re-prioritizing the basic surveying and mapping points in each key surveying and mapping area according to their comprehensive benefit values, and reselecting the basic surveying and mapping points for the second aerial photography path planning according to the priority ranking result.

[0014] As a further improvement of the present invention, the comprehensive income recalculation model is configured with: ; Among them, among them, is the comprehensive benefit adjustment value, g1 is the texture richness adjustment factor, which is used to indicate the impact of the deviation value on the texture richness in the comprehensive benefit value. , , are the adjustment coefficients for the contour perimeter and area, respectively, and g2 is the feature significance adjustment factor , , is the adjustment factor, which is used to indicate the impact of the deviation value on the characteristic significance in the comprehensive benefit value, and g3 is the flight distance adjustment factor , which is used to indicate the influence of the deviation value on the flight distance in the comprehensive benefit value. g4 is the flight tortuosity adjustment factor, which is used to indicate the influence of the deviation value on the flight tortuosity in the comprehensive benefit value.

[0015] The beneficial effects of the present invention are as follows: the present invention divides key surveying and mapping areas according to the spatial coordinates, distribution density and structural characteristics of the modeling target of the basic surveying and mapping points. In each area, a comprehensive benefit matrix is ​​constructed based on the comprehensive benefit value, and the weights of each basic surveying and mapping point are determined and sorted by methods such as principal component analysis. This method makes aerial photography path planning more scientific and reasonable, gives priority to high-value surveying and mapping points, improves aerial photography efficiency, and reduces unnecessary flight paths. In addition, the present invention introduces interference compensation and flight route adjustment strategies, and achieves the beneficial effect of improving the stability of the drone modeling aerial photography system by checking and avoiding obstacles that interfere with the drone. Through precise structural analysis, scientific benefit calculation and optimized path planning, the scientificity, efficiency and adaptability of drone aerial photography path planning are comprehensively improved, providing high-quality data support for dynamic modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 The present invention is a flow chart of a method for planning a path for aerial photography of an unmanned aerial vehicle for dynamic modeling.

[0017] Figure 2 It is a flow chart of the surveying and mapping revenue calculation strategy of the present invention.

[0018] Figure 3 It is a flow chart of the path interference compensation steps of the present invention. DETAILED DESCRIPTION

[0019] The present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. The same parts are represented by the same reference numerals. It should be noted that the words "front", "rear", "left", "right", "upper" and "lower" used in the following description refer to directions in the accompanying drawings, and the words "bottom surface" and "top surface", "inner" and "outer" refer to directions toward or away from the geometric center of a specific component, respectively.

[0020] What are the shortcomings of the current existing technology? Therefore, the present invention designs this unmanned aerial photography path planning method for dynamic modeling, such as Figure 1 As shown, including: The target data acquisition step is to acquire an image of the modeling target captured by the camera carried by the drone as a model analysis image, and obtain the size, shape and structural features of the modeling target as basic model information based on the model analysis image analysis; the high-precision camera carried by the drone is used to collect images of the modeling target in multiple directions, and the collected images cover all angles of the modeling target to ensure that the model information acquisition can more comprehensively display the size, shape and structural features of the modeling target. The collected image is preprocessed, and a filtering algorithm such as Gaussian filtering is used to remove sensor noise and environmental noise; Gaussian filtering performs convolution operations on each pixel in the image by setting a suitable kernel size and standard deviation, smoothing out noise points while retaining the edge and detail information of the image. The target data acquisition step also includes, after acquiring the model analysis image, using a visual algorithm to analyze the actual size of the modeling target based on the model analysis image to obtain the size of the modeling target, and using a contour extraction algorithm to obtain the contour information of the modeling target based on the model analysis image, and using a shape descriptor to quantify the contour information to obtain the shape of the modeling target; and by constructing a deep learning target detection algorithm to identify and locate the model analysis image to obtain predefined structural features. After image preprocessing, the size data of the modeling target is extracted by analyzing the image according to the preprocessed model through a computer vision algorithm, and the actual size of the modeling target is estimated through a reference object of known proportion in the image; the contour information of the modeling target is obtained by a contour extraction algorithm, and then the shape is quantitatively described using a shape descriptor, which is a basic shape unit preset in a database and used to determine the contour information of the modeling target.

[0021] The modeling structure analysis step determines the structural type of the modeling target through the structural analysis model according to the basic model information, and obtains the basic surveying and mapping points of the modeling target according to the structural type combined with the basic model information analysis; by constructing a structural analysis model based on deep learning, a convolutional neural network is used to train a large amount of image data of different types of structures, so that the model can more accurately identify the structural type of the modeling target, and according to the determined structural type and basic information, different judgment criteria are used to analyze the basic surveying and mapping points.

[0022] The structural analysis model obtains the structural type of the modeling target according to the shape and structural characteristics of the modeling target, and the structural type includes a conventional three-dimensional structure and a special-shaped three-dimensional structure; when the structural analysis model determines that the similarity between the shape of the modeling target and the preset geometric parameters is greater than a preset threshold, and the number of symmetric structures and structural repeating units in the structural characteristics of the modeling target is greater than a preset threshold, the output structural type is a conventional three-dimensional structure, otherwise the output structural type is a special-shaped three-dimensional structure. In which, the actual shape parameter set of the modeling target is assumed to be , for example, for a rectangular parallelepiped shape, Set to long, Set to width, Set to High. The default geometry parameter set is .

[0023] At this time, the shape similarity S is configured as: The closer S is to 1, the better the modeling target shape is. The shape is similar to the preset geometric shape. By analyzing the structural features of the modeling target, the spatial geometric transformation algorithm is used to detect whether there are symmetrical parts in the target structure, and the number of symmetrical structures is recorded. The structural repetitive units in the modeling target are identified by the pattern recognition algorithm: the modeling target is feature extracted to obtain its structural feature vector set, and then the similar structural feature vectors are clustered into one category by the clustering algorithm. At this time, each category is a structural repetitive unit, and the classification standard is adjusted by setting the similarity threshold.

[0024] After completing the classification of structural features, basic surveying points for modeling targets are selected according to the structural feature types.

[0025] When the structure type is a conventional three-dimensional structure, basic surveying points are selected from the geometric center plane, connection points and key feature points of each geometric shape of the modeling target according to the shape and size of the modeling target. For conventional three-dimensional structures, conventional three-dimensional structures are mostly composed of common geometric shapes based on the shape characteristics of the modeling target. Taking the rectangular building in the model analysis image as an example, the length, width and height are determined according to its size information, and basic surveying points are set at the center of each face, the midpoint of the edge and the vertex. This is because these parts can effectively reflect the overall shape and size of the cuboid, and in subsequent modeling, these points can be used to accurately construct the framework of the model. This strategy takes into account the shape and size factors of the modeling target.

[0026] Using the symmetry of the modeling target, select points at corresponding positions on the symmetry plane as basic surveying points. For example, for a cylindrical building, the structural features of the corresponding points on both sides of the symmetry axis are consistent. By selecting several points on one side, the corresponding points on the other side can be determined based on the symmetry relationship. This not only reduces the workload of surveying and mapping, but also ensures the symmetry of the model. For conventional three-dimensional structures composed of repeating units, analyze the distribution pattern of repeating units in the model analysis image. Select representative points from each repeating unit, such as each workshop unit in an industrial plant, and select points with obvious features such as the entrance and center. These points can represent the structural characteristics of the repeating unit, so selecting this point as the basic surveying point can improve modeling efficiency.

[0027] When the structure type is a special-shaped three-dimensional structure, basic mapping points are set at special nodes according to the shape information of the modeling target, including densely selecting basic mapping points at locations where the curvature of the curve changes dramatically, and setting basic mapping points at the turning points of the broken line according to the model analysis image analysis. For special-shaped three-dimensional structures, the shape of the special-shaped three-dimensional structure is irregular. According to the complex shape of the target in the model analysis image, basic mapping points are densely set at locations where the curvature of the curve changes greatly, the turning points of the broken line, and the surface mutations. For example, for artistic buildings with peculiar shapes, the curved and turning parts of the surface can accurately capture the shape changes by increasing the number of mapping points. This strategy is based on the shape and structural characteristics of the modeling target to ensure that sufficient information can be obtained to accurately depict the complex shape; according to the structural characteristics of the modeling target, for the unique structures in the special-shaped three-dimensional structure, such as the special modeling parts of large sculptures, cantilever structures in buildings, etc., mapping points are set comprehensively and meticulously. By analyzing the model analysis image, the boundaries and internal structures of these parts are clarified, and the mapping points are evenly distributed to fully obtain their structural information. This takes into account the unique structural characteristics of the special-shaped structure and its importance in modeling; considering that there may be occlusions and hidden parts in the special-shaped three-dimensional structure, the size and shape information of the modeling target is used, combined with the results of the model analysis image observed from different angles, to estimate the position and structure of the occlusions and hidden parts. Mapping points are set at the junction of the visible part and the possible occlusion and hidden part, as well as the key hidden parts judged based on experience, so as to restore the complete structure to the greatest extent in subsequent modeling.

[0028] The step of calculating the benefit of the surveying and mapping points is to obtain the sampling information and the surveying and mapping cost of the basic surveying and mapping points through the image analysis according to the model analysis, and to obtain the comprehensive benefit value of the basic surveying and mapping points through the comprehensive benefit model according to the sampling information and the surveying and mapping cost of the basic surveying and mapping points; The surveying and mapping revenue calculation strategy includes, for example, Figure 2 As shown, according to the structure type, the pixel clarity, texture richness and feature significance of the area where the basic surveying and mapping points are located are extracted from the model analysis image as sampling information, and the UAV flight distance, the flight tortuosity of the UAV at key points and the environmental information are analyzed according to the size and shape of the modeling target to obtain the surveying and mapping cost. The sampling information of each surveying and mapping point and the surveying and mapping cost are calculated by the comprehensive benefit model to obtain the comprehensive benefit value of each basic surveying and mapping point.

[0029] When the structure type is a conventional three-dimensional structure, considering that conventional three-dimensional structures have regular shapes and relatively stable characteristics, when calculating the sampling information, the focus is on analyzing the integrity and clarity of the images of the basic surveying and mapping points at different shooting angles. The pixel information of the area where the basic surveying and mapping points are located is extracted from the model analysis image. If the pixels in the area are evenly distributed, the edges are clear, and the contrast with the surrounding area is moderate, a higher sampling information score is given. For example, for the basic surveying and mapping points of a rectangular building, when the shooting angle can fully present the characteristics of the surface where the point is located, and the image is not blurred or deformed, the sampling information score is high. At the same time, combined with the size information of the modeling target, for the basic surveying and mapping points that are far away but can be fully included in the model analysis image, appropriate score adjustments are made according to their proportion in the image. If the proportion of the basic surveying and mapping points in the image is too small, it will affect the accuracy of its information and reduce the sampling information score accordingly. Since the position and shape of the conventional three-dimensional structure are relatively fixed, the surveying and mapping cost mainly considers the flight distance and flight difficulty of the drone to reach the basic surveying and mapping points. The flight distance is calculated based on the spatial coordinates of the basic surveying and mapping points and the initial position of the drone. If the basic mapping point is located near the straight line of the UAV's flight direction and there are no obstacles around, the flight difficulty is low and the mapping cost is reduced accordingly; on the contrary, if the UAV needs to make a large turn, cross a complex environment or avoid a no-fly zone to reach the basic mapping point, the mapping cost will increase. For example, when the basic mapping point is located on a conventional building in an open area, the flight cost is low; if it is located on a conventional building in an area with tall buildings and there are many flight restrictions around it, the flight cost will increase.

[0030] When the structure type is a special-shaped three-dimensional structure, due to the complex shape and rich details of the special-shaped three-dimensional structure, when calculating the sampling information, in addition to considering the integrity and clarity of the image, it is also necessary to focus on the information volume of the unique structural features contained in the basic mapping points. The feature information such as texture and curvature around the basic mapping points is extracted from the model analysis image. If the point is located in the key feature part of the special-shaped structure, such as the extreme point of the complex surface, the center position of the unique decoration, etc., and these features are clearly distinguishable in the image, a higher sampling information score is given. At the same time, considering the information complementarity between different basic mapping points, if a basic mapping point can provide key information different from other points, which helps to build a more comprehensive model, its sampling information score is improved. For example, in a special-shaped building with a unique sculptural shape, the texture and shape of the sculpture surface are rich in variations. The basic mapping points located in the area with obvious texture changes will have their sampling information scores improved due to the inclusion of unique texture information. The irregularity of the special-shaped three-dimensional structure increases the difficulty and risk of the UAV flight, and the mapping cost calculation needs to consider more factors comprehensively. In addition to the flight distance, the energy and time consumed by the complex posture adjustments of the drone in order to obtain the best shooting angle when approaching the basic mapping point must also be considered. For example, when the basic mapping point is located in a recessed part of an irregular building or under a protruding overhang, the drone needs to adjust its posture to avoid collision and find a suitable shooting angle, which will significantly increase the mapping cost. In addition, due to the occlusion problem of the irregular structure, the drone needs to adjust its position multiple times to obtain complete information, and each adjustment increases the mapping cost. At the same time, combined with the previously collected model analysis images and structural feature data, if it is found that the environmental interference in a certain area is large, such as strong winds, light reflections, etc., it will also increase the mapping cost of the basic mapping points in that area.

[0031] The comprehensive income model configuration is: ; in, is the comprehensive benefit adjustment value, g1 is the texture richness adjustment factor, which is used to indicate the impact of the deviation value on the texture richness in the comprehensive benefit value. , , are the adjustment coefficients for the contour perimeter and area, respectively, and g2 is the feature significance adjustment factor , , is the adjustment factor, which is used to indicate the impact of the deviation value on the characteristic significance in the comprehensive benefit value, and g3 is the flight distance adjustment factor , which is used to indicate the influence of the deviation value on the flight distance in the comprehensive benefit value. g4 is the flight tortuosity adjustment factor, which is used to indicate the influence of the deviation value on the flight tortuosity in the comprehensive benefit value.

[0032] Ci is the pixel definition of the area where the ith basic mapping point is located, and is the definition quantization value of the image of the mapping point calculated by the image processing algorithm; T i is the texture richness of the area where the i-th basic mapping point is located, and is the quantitative value of the texture complexity of the image of the mapping point obtained by the texture analysis algorithm; S i is the feature significance of the area where the i-th basic mapping point is located, and is the quantitative value of the feature prominence of the mapping point obtained based on the feature extraction algorithm; D i is the flight distance from the UAV to the i-th basic mapping point, is the actual flight distance calculated based on the coordinates of the UAV and the mapping point; W i is the flight tortuosity of the UAV at the key point of the i-th basic mapping point, and is the quantitative value of the tortuosity obtained by analyzing the curvature of the UAV flight path and other information; E i is the environmental information influencing factor of the ith basic mapping point, which is a quantitative value obtained by comprehensively considering the influence of environmental factors such as wind speed and light on mapping; ω1, ω2, ω3 are the weight coefficients of pixel clarity, texture richness and feature significance, respectively, which are preset constants according to the actual modeling situation and are used to adjust the importance of each sampling information factor in the comprehensive benefit; ω4, ω5, ω6 are the weight coefficients of flight distance, flight tortuosity and environmental information, respectively, which are used to adjust the importance of each mapping cost factor in the comprehensive benefit. It is a preset constant used to set the denominator not to 0.

[0033] is a function for processing pixel clarity, where is the maximum value of the pixel resolution of all surveying points. is the mean value of pixel clarity, is the standard deviation of pixel clarity; is a function for processing texture richness, where is the maximum value of texture richness of all surveying points, is the mean value of texture richness, is the standard deviation of texture richness; is a function for processing feature saliency, where is the maximum value of the feature significance of all surveying points, is the mean of the feature significance, is the standard deviation of feature significance; is the function for processing the flight distance, where is the maximum flight distance of all surveying points, is the mean flight distance, is the standard deviation of the flight distance; is the function for processing the flight tortuosity, where is the maximum value of the flight tortuosity of all surveying points, is the mean value of the flight tortuosity, is the standard deviation of flight tortuosity; is the function for processing the flight distance, where is the maximum flight distance of all surveying points, is the mean flight distance, is the standard deviation of the flight distance.

[0034] The aerial photography path planning step is to divide the basic mapping points into regions and form several key mapping areas through the mapping point planning strategy according to the structural characteristics, prioritize the basic mapping points in each key mapping area according to the comprehensive benefit value, and select the basic mapping points according to the priority ranking result to complete the drone aerial photography path planning. The aerial photography path planning step also includes dividing the key mapping areas according to the spatial coordinates of the basic mapping points through the distribution density and the structural characteristics of the modeling target, constructing a comprehensive benefit matrix for the basic mapping points in each key mapping area based on the comprehensive benefit value, confirming the weight of each basic mapping point by analyzing the comprehensive benefit matrix, and sorting the basic mapping points from high to low according to the weight.

[0035] For each basic surveying point in the key surveying area, a comprehensive benefit matrix M is constructed. The rows of the matrix represent the basic surveying points, and the columns represent the factors that affect the comprehensive benefits, such as pixel clarity C, texture richness T, feature prominence S, flight distance D, flight tortuosity W, and environmental information E. That is, M ij It represents the quantitative value of the i-th basic surveying and mapping point on the j-th influencing factor.

[0036] The principal component analysis method is used to analyze the comprehensive income matrix M. PCA transforms the original multiple correlated variables into a new set of uncorrelated variables, namely the principal components, through linear transformation. First, the covariance matrix of the comprehensive income matrix M is calculated, and then the eigenvalues ​​λ1 to λm of the covariance matrix are solved (m is the number of influencing factors) and the corresponding eigenvectors to The size of the eigenvalue reflects the contribution of the principal component to the data variance. The eigenvalue is normalized to obtain the weight of each principal component from τ1 to τ m ,Right now , these weights represent the relative importance of different influencing factors in the comprehensive benefits.

[0037] For each basic surveying and mapping point, the comprehensive score is calculated according to its value in the comprehensive benefit matrix and the weight of each factor, and the basic surveying and mapping points in each key surveying and mapping area are sorted from high to low according to the comprehensive score. The sorting result will serve as an important basis for UAV aerial photography path planning, and basic surveying and mapping points with high comprehensive scores will be given priority for photography to improve aerial photography efficiency and modeling quality.

[0038] Specifically, the path interference compensation step further includes: Figure 3As shown, the theoretical target model is extracted according to the theoretical target image at the corresponding position of the basic surveying and mapping points of the sampling image, and the image features of the sampling image and the theoretical target image are respectively extracted by edge detection and pixel coordinate calibration. The image features include the contour perimeter, area, shape descriptor, texture feature and key point position of the modeling target in the sampling image. The edge contour of the theoretical target image and the modeling target in the sampling image is extracted by the Canny algorithm to obtain the contour perimeter. The difference between the shape of the modeling target in the image obtained during the actual shooting process and the theoretical modeling can be judged by the contour perimeter deviation between the sampling image and the theoretical target image. Moreover, accurate coordinate calibration is performed for each pixel in the image by pixel coordinate calibration to determine the position information of each element in the image, so that the area and shape descriptor of the modeling target in the sampling image and the theoretical target image can be obtained. The area difference can also determine whether the modeling target is deformed in the image, and the deviation of its shape feature can be more intuitively reflected according to the shape descriptor. The texture feature is extracted by the gray level co-occurrence matrix method, and the detail features of the modeling target are reflected in detail by the gray level co-occurrence matrix elements. The key point positions in the image that have a significant impact on the modeling are extracted by the SIFT algorithm. After obtaining the various features of the two images, the spatial distribution of the key point position deviations in the two images is statistically matched to obtain the spatial distribution value. The spatial distribution value can intuitively show the distribution state of the key point position deviation in the entire modeling target space. The deviation coefficient is calculated by the interference deviation model according to the image feature difference between the sampled image and the theoretical target image, and the first-level deviation threshold, the second-level deviation threshold and the spatial distribution threshold are set. The deviation threshold represents the deviation value of the image features between the two images. By quantifying the deviation, it is convenient to judge the specific deviation degree of the image. When the deviation coefficient is higher than the first-level deviation threshold and lower than the second-level deviation threshold, and the spatial distribution value is lower than the spatial distribution threshold, it indicates that there is a certain degree of difference between the sampled image and the theoretical target image, but this difference is mainly concentrated in the local area and has a relatively small impact on the overall modeling. Therefore, there is a microscopic deviation in outputting the basic surveying and mapping point. At this time, the subsequent aerial photography path can be fine-tuned by the microscopic path adjustment strategy to ensure the accuracy of the aerial photography data.When the deviation coefficient is higher than the secondary deviation threshold and the spatial distribution value is higher than the spatial distribution threshold, it means that there is a large difference between the sampled image and the theoretical target image, and this difference is widely distributed in space, which may have a serious impact on the overall modeling. Therefore, there is a global deviation in the output of the basic surveying and mapping point. In this case, it is necessary to start the global path adjustment strategy to re-plan and adjust the entire aerial photography path to ensure that aerial photography data that meets the modeling requirements is obtained. Through the above-mentioned path interference compensation step, the present invention can judge the deviation type occurring in the drone aerial photography process in real time and accurately, and adopt corresponding adjustment strategies according to different deviation types, which effectively improves the accuracy of drone aerial photography path planning and the reliability of dynamic modeling, and provides more efficient and accurate technical support for applications in related fields.

[0039] Specifically, the interference deviation model is configured as follows: ; Among them, D is the deviation coefficient, which is used to judge the degree of deviation between the sampled image and the theoretical target image. The main function of the interference deviation model is to calculate a deviation coefficient D, which can comprehensively and comprehensively reflect the degree of difference between the sampled image and the theoretical target image. During the drone aerial photography process, due to the interference of various factors, such as changes in ambient light and unstable drone posture, the sampled image may deviate from the theoretical target image. The deviation coefficient D calculated by the interference deviation model can quantify this deviation, so as to help us accurately judge whether micro-path adjustment or global path adjustment is needed, and ensure the accuracy of aerial photography data and the reliability of modeling; n represents the number of image logarithms at the modeling target in the image, usually 1. When the basic surveying and mapping point needs to be sampled continuously, n is the number of corresponding sampled image and theoretical target image logarithms; and are the contour perimeters of the i-th sampled image and the theoretical target image, respectively. In order to eliminate the influence of the difference in the numerical values ​​of the contour perimeters between different images, max ( , ) Normalization is performed by the maximum value so that the contour perimeters of different images can be compared at a unified scale; and are the area components of the i-th sampling image and the theoretical target image in the j direction, respectively. The area component represents the length of the modeling target after mapping in the j direction. Multiple area components are obtained by decomposing the area of ​​the modeling target from multiple directions. m represents the number of area components. and They represent the norms of the area vectors of the i-th sampled image and the theoretical target image, respectively. They are comprehensive measures of multiple area components. For irregularly shaped modeling targets, their area measurement values ​​can be calculated through area components, thereby more comprehensively reflecting the area characteristics of the modeling targets. and denote the shape descriptors of the i-th sample image and the theoretical target image respectively. The shape descriptors are used to describe the shape features of the modeling target. H Represents the Hausdorff distance between shape descriptors, represents the distance metric between shape descriptors, is a way to measure the distance between two points, and is used to measure the difference between shape descriptors. ; and Respectively represent the texture feature values ​​of the i-th sampled image and the theoretical target image, It is the definition domain of texture features. When calculating the difference of texture features, the comparison is limited to the definition domain to ensure the rationality and effectiveness of the calculation.

[0040] and Respectively represent the position vector of the kth key point of the i-th sampled image and the theoretical target image, p is the number of key points, and Respectively represent the norm of the set of all key point position vectors of the i-th sample image and the theoretical target image, and represent the sum of the distances from all key points to the centroid. By using the key point position difference, the deviation between images can be judged more accurately. α, β, γ, δ and ε represent the weight factors of each parameter, where the value of ε is greater than α, β, γ, δ, and α+β+γ+δ+ε=1.

[0041] During the drone aerial photography process, due to interference from various factors, such as slight airflow effects, sensor errors, etc., there may be certain differences between the sampled image and the theoretical target image, but this difference is mainly concentrated in the local area and has little impact on the overall modeling, that is, micro-deviation occurs. The main function of the micro-path adjustment strategy is to calculate the offset of the basic surveying and mapping points based on these deviation information and adjust their position coordinates, so that the subsequently collected sampled images are closer to the theoretical target image and reduce the impact of deviation on modeling. This strategy can quickly and effectively correct local deviations without large-scale path replanning, thereby improving aerial photography efficiency and modeling accuracy. Specifically, the micro-path adjustment strategy includes: calculating the basic mapping point offset according to the contour perimeter, area, shape descriptor, texture features and key point position deviation value of the modeling target in the sampling image and the theoretical target image, adjusting the basic mapping point position coordinates according to the basic mapping point offset, and re-acquiring the modeling target sampling image of the updated basic mapping point position, respectively extracting the contour perimeter of the modeling target in the sampling image and the theoretical target image through the edge detection algorithm, and then calculating their difference as the contour perimeter deviation value, using pixel coordinate calibration and area calculation method to obtain the area of ​​the modeling target in the sampling image and the theoretical target image, and calculating the area deviation value. The shape descriptor of the sampling image and the theoretical target image is obtained by the shape description method of the Hu moment, and the difference between them is calculated by the Hausdorff distance metric to obtain the shape descriptor deviation value. The basic mapping point offset calculation configuration includes: ; in, is a comprehensive correction vector, which indicates the offset of the coordinates of the basic surveying and mapping points. (v PAx 、v PAy ) is the first correction vector based on the deviation of contour perimeter and area, where , , k P is the contour perimeter deviation correction coefficient, k A is the area deviation correction factor, is the contour perimeter deviation, is the area deviation, P0 is the reference contour perimeter of the modeled target in the theoretical target image, A0 is the reference area of ​​the modeled target in the theoretical target image, and θ represents the correction direction angle calculated based on the deviation of the contour perimeter and area. Jx 、v Jy ) is the second correction vector based on the deviation of the shape descriptor, , , where ΔH i is the i-th element in the difference vector matrix, ω iis the weight coefficient of the i-th element in the x direction, ω ＇ i is the weight coefficient of the i-th element in the y direction, (v Xx 、v Xy ) is the third correction vector based on the deviation of texture features, , , where k X is the texture deviation correction coefficient, α is the correction direction angle determined according to the texture deviation distribution, (Δx, Δy) is the key point position deviation; The main function of the global path adjustment strategy is to re-evaluate and adjust the comprehensive benefits of the basic surveying and mapping points when global deviations occur, so as to re-plan the aerial photography path and improve the quality of aerial photography data and the accuracy of modeling. In the actual aerial photography process, due to the influence of various complex factors, such as environmental changes, large deviations in the flight attitude of the drone, etc., the sampled image may be greatly different from the theoretical target image, affecting the modeling effect. Through the global path adjustment strategy, the priority of the basic surveying and mapping points can be dynamically adjusted according to the deviation situation, so that the drone can collect data more targetedly, reduce invalid shooting, and improve the efficiency of aerial photography and the quality of modeling. Specifically, the global path adjustment strategy includes calculating the benefit adjustment factor of the basic surveying and mapping point corresponding to the sampling image according to the deviation of the contour perimeter, area, shape descriptor, texture feature and key point position, and the benefit adjustment factor includes texture richness adjustment factor, feature saliency adjustment factor and flight distance adjustment factor. These deviation values ​​reflect the degree of difference between the sampled image and the theoretical target image in various aspects, and are an important basis for calculating the benefit adjustment factor. The texture richness adjustment factor is used to measure the influence of the difference between the modeling target texture feature in the sampled image and the theoretical target image on the benefit of the basic surveying and mapping point. If the texture features of the sampled image are too simple or complex compared to the theoretical target image, it may affect the accuracy of modeling, so the benefits of the basic mapping points need to be adjusted. The feature saliency adjustment factor takes into account the impact of the difference between the shape, contour and other features of the modeled target in the sampled image and the theoretical target image on the benefits of the basic mapping points. The flight distance adjustment factor takes into account the impact of the flight distance of the drone from the current position to the basic mapping point on the benefits, and recalculates the comprehensive benefit adjustment value of the basic mapping point through the comprehensive benefit recalculation model according to the comprehensive benefit adjustment factor. The recalculated comprehensive benefit adjustment value replaces the comprehensive benefit value of the basic mapping point as the final comprehensive benefit value, and re-prioritizes the basic mapping points in each key mapping area according to their comprehensive benefit values, and re-selects the basic mapping points for the second aerial photography path planning according to the priority ranking results. The global path adjustment strategy is adaptive and flexible, and can dynamically adjust the aerial photography path according to the actual global deviation situation to improve the quality of aerial photography data. This strategy is suitable for various complex dynamic modeling scenarios, especially when the environment changes greatly and the UAV's flight posture is prone to large deviations, such as when modeling in mountainous areas and urban areas with high-rise buildings. It can effectively deal with global deviations and ensure the smooth progress of modeling work.

[0042] Specifically, the comprehensive income recalculation model is configured as follows: ; The core function of the comprehensive benefit recalculation model is to recalculate the comprehensive benefit value of the basic surveying and mapping points according to the deviations between the sampled image and the theoretical target image in multiple aspects in the global path adjustment strategy. The application of this model enables aerial photography path planning to adapt more accurately to the actual shooting situation, improve the quality of aerial photography data and the accuracy of modeling. In the actual aerial photography process, due to the interference of various complex factors, global deviations may occur. At this time, the original comprehensive benefit value can no longer accurately reflect the actual value of the basic surveying and mapping points. Recalculating the comprehensive benefit adjustment value through this model can provide a more scientific basis for subsequent path replanning. is the comprehensive benefit adjustment value, g1 is the texture richness adjustment factor, which is used to indicate the impact of the deviation value on the texture richness in the comprehensive benefit value. , , They are the adjustment factors for the contour perimeter and area, which may increase in some modeling scenes that require high texture details. and The value of g1 is used to more sensitively reflect the impact of the deviation of texture features on the comprehensive benefits. If the contour perimeter and area of ​​the sampled image are significantly different from those of the theoretical target image, the value of g1 will increase accordingly, thereby reducing the comprehensive benefits of the basic mapping point in terms of texture richness. g2 is the feature saliency adjustment factor , , is an adjustment factor used to represent the impact of the deviation value on the feature significance in the comprehensive return value. For modeling targets with regular shapes, it may increase The weight of the shape descriptor bias; for targets with large contour changes, it may increase The weight of the deviation of the contour perimeter. Through the calculation of g2, the value of the basic surveying points in terms of feature significance can be more accurately evaluated, so as to reasonably adjust the comprehensive benefits.

[0043] g3 is the flight distance adjustment factor , used to indicate the impact of the deviation value on the flight distance in the comprehensive benefit value. During the aerial photography process, the longer the flight distance, the more energy the drone consumes and the longer it takes to shoot, which will reduce the actual benefit of the basic mapping point. Therefore, g3 is used to quantify this impact, so that when replanning the aerial photography path, basic mapping points with a shorter flight distance can be given priority to improve the efficiency of aerial photography. g4 is the flight tortuosity adjustment factor, which is used to indicate the impact of the deviation value on the flight tortuosity in the comprehensive benefit value. A tortuous flight path will increase the difficulty and time cost of the drone's flight, and also affect the stability and data quality of the shooting. By considering the flight tortuosity, when recalculating the comprehensive benefit, it is possible to avoid selecting basic mapping points with too tortuous flight paths, thereby optimizing the aerial photography path.

[0044] The above shows and describes the basic features, principles and advantages of the present invention. It should be noted that the present invention is not limited to the above embodiments, which are only partial embodiments. Under the premise of not departing from the spirit and scope of the present invention, several improvements and supplements made are considered as the protection scope of the present invention.

Claims

1. A UAV aerial photography path planning method for dynamic modeling, characterized in that: The steps include: A target data acquisition step, acquiring a remote sensing image of the modeling target as a model analysis image, and obtaining the size, shape and structural features of the modeling target as basic model information based on the model analysis image; A modeling structure analysis step, in which the structural type of the modeling target is determined through a structural analysis model according to the basic model information, and a theoretical target model is constructed, and basic surveying and mapping points of the modeling target are obtained according to the structural type combined with the basic model information analysis; A surveying and mapping point benefit calculation step, obtaining sampling information and surveying and mapping costs according to the basic surveying and mapping points through a surveying and mapping benefit calculation strategy, and obtaining a comprehensive benefit value of the basic surveying and mapping points through a comprehensive benefit model according to the sampling information and surveying and mapping costs of the basic surveying and mapping points; The aerial photography path planning step includes dividing the basic surveying points into regions and forming several key surveying areas according to the structural features through the surveying point planning strategy, prioritizing the basic surveying points in each key surveying area according to the comprehensive benefit value, and selecting the basic surveying points according to the priority ranking result to complete the drone aerial photography path planning; The path interference compensation step is that when the UAV reaches the basic surveying and mapping point in the aerial photography path, a sample image of the modeling target is acquired in real time, and a feature comparison is performed between the sample image and the theoretical target model, and a deviation coefficient is calculated according to the feature comparison result through the interference deviation model. A micro deviation or a global deviation is output according to the comparison result between the deviation coefficient and the deviation threshold. When the output is a micro deviation, a micro path adjustment strategy is entered; when the output is a global deviation, a global path adjustment strategy is entered.

2. The method for drone aerial photography path planning for dynamic modeling according to claim 1, characterized in that: The structural analysis model obtains the structural type of the modeling target according to the shape and structural feature analysis of the modeling target, and the structural type includes a conventional three-dimensional structure and a special-shaped three-dimensional structure; when the structural analysis model determines that the similarity between the shape of the modeling target and the preset geometric parameters is greater than a preset threshold, and the number of symmetrical structures and structural repeating units in the structural features of the modeling target is greater than the preset threshold, the output structural type is a conventional three-dimensional structure, otherwise the output structural type is a special-shaped three-dimensional structure. When the structural type is a conventional three-dimensional structure, basic surveying and mapping points are selected from the geometric center plane, connection points and key feature points of each geometric shape of the modeling target according to the shape and size of the modeling target; when the structural type is a special-shaped three-dimensional structure, basic surveying and mapping points are set at special nodes according to the shape information of the modeling target, including densely selecting basic surveying and mapping points at positions where the curvature of the curve changes drastically, and setting basic surveying and mapping points at the turning points of the broken line according to the model analysis image analysis.

3. The method for drone aerial photography path planning for dynamic modeling according to claim 1, characterized in that: The surveying and mapping benefit calculation strategy includes extracting the pixel clarity, texture richness and feature significance of the area where the basic surveying and mapping points are located in the model analysis image as sampling information according to the structure type, and analyzing the UAV flight distance, the flight tortuosity of the UAV at key points and environmental information according to the size and shape of the modeling target to obtain the surveying and mapping cost, and calculating the sampling information of each surveying and mapping point and the surveying and mapping cost through a comprehensive benefit model to obtain the comprehensive benefit value of each basic surveying and mapping point.

4. The method for drone aerial photography path planning for dynamic modeling according to claim 3 is characterized in that: The comprehensive income model configuration is: ； Among them, R i is the comprehensive benefit value of the i-th basic surveying and mapping point, C i is the pixel definition of the i-th basic mapping point, T i is the texture richness of the i-th basic mapping point, S i is the feature significance of the i-th basic mapping point, indicating the prominence of the feature of the mapping point; D i is the flight distance from the UAV to the i-th basic mapping point, W i is the flight tortuosity of the UAV at the i-th basic mapping point, indicating the flight difficulty of the UAV at this point; E i is the environmental impact factor of the i-th basic surveying and mapping point.

5. The method for drone aerial photography path planning for dynamic modeling according to claim 1, characterized in that: The aerial photography path planning step also includes dividing the key surveying and mapping areas according to the spatial coordinates of the basic surveying and mapping points through distribution density and the structural characteristics of the modeling target, constructing a comprehensive benefit matrix based on the comprehensive benefit value of the basic surveying and mapping points in each key surveying and mapping area, confirming the weight of each basic surveying and mapping point by analyzing the comprehensive benefit matrix, and sorting the basic surveying and mapping points from high to low according to the weight.

6. The method for drone aerial photography path planning for dynamic modeling according to claim 1, characterized in that: The path interference compensation step also includes extracting the theoretical target image at the position corresponding to the theoretical target model according to the basic surveying and mapping points of the sampling image, extracting the image features of the sampling image and the theoretical target image respectively through edge detection and pixel coordinate calibration, the image features including the contour perimeter, area, shape descriptor, texture features and key point positions of the modeling target in the sampling image, statistically matching the spatial distribution of the key point position deviations in the two images to obtain a spatial distribution value, calculating the deviation coefficient through the interference deviation model according to the image feature difference between the sampling image and the theoretical target image, and setting a primary deviation threshold, a secondary deviation threshold and a spatial distribution threshold, when the deviation coefficient is higher than the primary deviation threshold and lower than the secondary deviation threshold, and the spatial distribution value is lower than the spatial distribution threshold, outputting that the basic surveying and mapping point has a microscopic deviation, and when the deviation coefficient is higher than the secondary deviation threshold, and the spatial distribution value is higher than the spatial distribution threshold, outputting that the basic surveying and mapping point has a global deviation.

7. The method for drone aerial photography path planning for dynamic modeling according to claim 1, characterized in that: The interference deviation model configuration includes: ； Where D is the deviation coefficient, which is used to determine the degree of deviation between the sampled image and the theoretical target image; n represents the number of image pairs at the modeled target in the image, which is usually 1. When the basic mapping point needs to be sampled continuously, n is the number of corresponding sampled image pairs and theoretical target image pairs; and are the contour perimeters of the i-th sampled image and the theoretical target image, respectively, max( , ) means normalization by the maximum value; and are the area components of the i-th sample image and the theoretical target image in the j direction, respectively. The area component represents the length of the modeled target after mapping in the j direction. m represents the number of area components. and Represent the norm of the area vector of the i-th sampled image and the theoretical target image respectively; and denote the shape descriptors of the i-th sampled image and the theoretical target image, respectively, and d H Represents the Hausdorff distance between shape descriptors, which represents the distance measure between shape descriptors; and Respectively represent the texture feature values ​​of the i-th sampled image and the theoretical target image, is the domain of texture features; and Respectively represent the position vector of the kth key point of the i-th sampled image and the theoretical target image, p is the number of key points, and represents the norm of the set of all key point position vectors of the i-th sampled image and the theoretical target image, represents the sum of the distances from all key points to the centroid, α, β, γ, δ and ε represent the weight factors of each parameter, where the value of ε is greater than α, β, γ, δ, and α+β+γ+δ+ε=1.

8. The method for drone aerial photography path planning for dynamic modeling according to claim 1, characterized in that: The microscopic path adjustment strategy includes calculating the basic surveying and mapping point offset according to the contour perimeter, area, shape descriptor, texture feature and key point position deviation value of the modeling target in the sampling image and the theoretical target image, adjusting the basic surveying and mapping point position coordinates according to the basic surveying and mapping point offset, and re-acquiring the modeling target sampling image of the updated basic surveying and mapping point position. The basic surveying and mapping point offset calculation configuration includes: ； in, is a comprehensive correction vector, which indicates the offset of the coordinates of the basic surveying and mapping points. (v PAx 、v PAy ) is the first correction vector based on the deviation of contour perimeter and area, where , , k P is the contour perimeter deviation correction coefficient, k A is the area deviation correction factor, is the contour perimeter deviation, is the area deviation, P0 is the reference contour perimeter of the modeled target in the theoretical target image, A0 is the reference area of ​​the modeled target in the theoretical target image, and θ represents the correction direction angle calculated based on the deviation of the contour perimeter and area. Jx 、v Jy ) is the second correction vector based on the deviation of the shape descriptor, , , where ΔH i is the i-th element in the difference vector matrix, ω i is the weight coefficient of the i-th element in the x direction, ω ＇ i is the weight coefficient of the i-th element in the y direction, (v Xx 、v Xy ) is the third correction vector based on the deviation of texture features, , , where k X is the texture deviation correction coefficient, α is the correction direction angle determined according to the texture deviation distribution, and (Δx, Δy) is the key point position deviation.

9. The method for drone aerial photography path planning for dynamic modeling according to claim 8, characterized in that: The global path adjustment strategy includes: obtaining the benefit adjustment factor of the basic surveying and mapping point at the corresponding position of the sampling image according to the deviation of the contour perimeter, area, shape descriptor, texture feature and key point position, the benefit adjustment factor includes a texture richness adjustment factor, a feature significance adjustment factor and a flight distance adjustment factor; recalculating the comprehensive benefit adjustment value of the basic surveying and mapping point through a comprehensive benefit recalculation model according to the comprehensive benefit adjustment factor; replacing the comprehensive benefit value of the basic surveying and mapping point with the recalculated comprehensive benefit adjustment value as the final comprehensive benefit value; re-prioritizing the basic surveying and mapping points in each key surveying and mapping area according to their comprehensive benefit values; and reselecting the basic surveying and mapping points for the second aerial photography path planning according to the priority ranking result.

10. The method for drone aerial photography path planning for dynamic modeling according to claim 9, characterized in that: The comprehensive income recalculation model is configured with: ； in, is the comprehensive benefit adjustment value, g1 is the texture richness adjustment factor, which is used to indicate the impact of the deviation value on the texture richness in the comprehensive benefit value. , , are the adjustment coefficients for the contour perimeter and area, respectively, and g2 is the feature significance adjustment factor , , is the adjustment factor, which is used to indicate the impact of the deviation value on the characteristic significance in the comprehensive benefit value, and g3 is the flight distance adjustment factor , which is used to indicate the influence of the deviation value on the flight distance in the comprehensive benefit value. g4 is the flight tortuosity adjustment factor, which is used to indicate the influence of the deviation value on the flight tortuosity in the comprehensive benefit value.

Images