Unmanned aerial vehicle oblique photography modeling method based on sensitive factor identification

Abstract

The invention discloses an unmanned aerial vehicle oblique photography modeling method based on sensitive factor identification, and relates to the technical field of unmanned aerial vehicle oblique photography modeling, and the method comprises the following steps: obtaining the orthographic and oblique photography data of a target area; performing light and color uniformity and distortion correction preprocessing on the acquired orthographic and oblique photography data to obtain first data; performing image segmentation on the first data, and marking preset sensitive factors in an image with different colors; performing aerial triangulation on the preprocessed image data to generate a basic three-dimensional model; based on the basic three-dimensional model, texture mapping is carried out on the preprocessed image data and the image data marked with the sensitive factors, and a live-action three-dimensional model of the target area and a three-dimensional model marked with the sensitive factors are generated. The three-dimensional live-action model is obtained through the oblique photography data of the unmanned aerial vehicle, meanwhile, the three-dimensional model marking the sensitive factors and corresponding to the three-dimensional live-action model is obtained, and the influence of the sensitive factors on engineering implementation is analyzed more truly and visually.

Description

Technical Field

[0001] The present invention relates to the fields of three-dimensional modeling technology and image processing technology, and more specifically, to a drone oblique photography modeling method based on sensitive factor recognition. Background Art

[0002] In recent years, with the development of drone technology, oblique photography technology and 3D modeling technology, 3D modeling based on drone oblique photography has become a new technology in the field of surveying and mapping. The 3D real scene model has become an important geographic information data for regional construction due to its three-dimensional intuitive characteristics. The 3D real scene model constructed by drone oblique photography technology has the measurement characteristics of large range, high precision and high definition, which can intuitively reflect the appearance, position, height and other attributes of the objects, effectively improve the measurement effect and surveying and mapping accuracy, and reduce the cost of 3D modeling. Drone oblique photography measurement technology has been widely used in the construction management of highway, railway, municipal and other line projects. However, in some specific fields, such as power grid infrastructure, simply restoring the 3D model cannot meet the project needs. The sensitive factors (houses, rivers and roads, etc.) in the target area seriously affect the planning of transmission lines. Therefore, it is particularly important to mark sensitive factors in the 3D real scene model.

[0003] The above disclosed technical solutions have at least the following technical problems: in certain specific fields, such as power grid infrastructure, simply restoring the three-dimensional model cannot meet the project requirements, and the sensitive factors (houses, rivers, roads, etc.) in the target area seriously affect the planning of transmission lines. In view of the above problems, the present invention proposes a solution. Summary of the invention

[0004] In order to overcome the above-mentioned defects of the prior art, an embodiment of the present invention provides a UAV oblique photography modeling method based on sensitive factor identification, which obtains a three-dimensional model with marked sensitive factors corresponding to the three-dimensional real scene model to solve the problem that simply restoring the three-dimensional model cannot meet the project requirements and the sensitive factors (houses, rivers and roads, etc.) in the target area seriously affect the planning of transmission lines.

[0005] To achieve the above object, the present invention provides the following technical solutions: A method for modeling unmanned aerial vehicle oblique photography based on sensitive factor identification comprises the following steps: acquiring orthophoto and oblique photography data of a target area; performing uniform light and color and distortion correction preprocessing on the acquired orthophoto and oblique photography data to obtain first data; performing image segmentation on the first data, and marking preset sensitive factors in the image with different colors; performing aerial triangulation on the preprocessed image data to generate a basic three-dimensional model; and performing texture mapping on the preprocessed image data and the image data after marking sensitive factors based on the basic three-dimensional model to generate a real-scene three-dimensional model of the target area and a three-dimensional model after marking sensitive factors.

[0006] In a preferred embodiment, the obtaining of orthophoto and oblique photography data of the target area is specifically: implementing drone operations to obtain orthophoto and oblique photography data of the target area, and checking the obtained data, and supplementing the images in areas where image information is missing to obtain complete image data of the target area.

[0007] In a preferred embodiment, the distortion correction includes radial distortion processing and tangential distortion correction processing.

[0008] In a preferred embodiment, the acquired orthophoto and oblique photography data are subjected to uniform light and color and distortion correction preprocessing to obtain the first data, specifically: uniform light and color processing is performed by Wallis filtering method to unify the grayscale mean and grayscale variance of the target area image, reduce the contrast value of the high-contrast interval, and increase the contrast value of the low-contrast interval; and the distortion-corrected image is obtained by estimating and correcting the distortion parameters of each pixel point.

[0009] In a preferred embodiment, the image segmentation of the first data and marking of preset sensitive factors in the image with different colors are specifically as follows: preparing a number of oblique photography image data; importing the oblique photography image data into AnyLabeling; using the tools and interface provided by AnyLabeling to mark the sensitive factors in the image by drawing bounding boxes, marking key points, adding text labels, and other operations; after the marking is completed, verifying and editing the oblique photography image data; after the marking is completed, exporting the oblique photography image data into a format for image segmentation model training; using the labeled data set to train the image segmentation model, continuously updating the data set and retraining the model; using the trained model to perform image segmentation on the preprocessed image data, and marking the sensitive factors in the image with different colors.

[0010] In a preferred embodiment, the aerial triangulation of the preprocessed image data is performed to generate a basic three-dimensional model, specifically: automatically detecting and extracting feature points of the preprocessed image, wherein the feature points include pixel values; processing the scale, rotation, and illumination changes between images, and using feature descriptors to find and match the same feature points between different images; preliminarily estimating the camera's internal orientation elements and external orientation elements based on ground control points, wherein the internal orientation elements include focal length and principal points, and the external orientation elements include position and posture; obtaining feature points with the smallest intersecting error in three-dimensional space by adjusting the camera's position, posture parameters, and feature point coordinates; and generating a basic three-dimensional model based on the smallest feature point, internal orientation elements, and external orientation elements combined with collinearity equations.

[0011] In a preferred embodiment, the texture mapping is performed on the image data after preprocessing and after marking sensitive factors, respectively, to generate a real-life three-dimensional model of the target area and a three-dimensional model after marking sensitive factors, specifically as follows: according to the collinearity equation, the relationship between the external coordinate transformation to the image coordinate is constructed, and the three-dimensional point coordinates are reversely texture mapped; the texture coordinates and the rectangular texture space are defined based on a preset texture coordinate formula; the points in the spatial coordinate system are projected to the image space through external orientation elements, and then the corresponding relationship between these image space points and the original three-dimensional model space points is established; based on the texture coordinates and the rectangular texture space, the two-dimensional texture coordinates are mapped to the surface of the three-dimensional model to obtain the relationship between the texture coordinates and the three-dimensional space points; based on the coordinate conversion relationship from three-dimensional to two-dimensional and the corresponding relationship between the texture coordinates and the three-dimensional space points, the pixel values ​​are used as texture space points and mapped to the corresponding surface of the three-dimensional model to obtain the three-dimensional model.

[0012] The technical effects and advantages of the UAV oblique photography modeling method based on sensitive factor identification of the present invention are as follows: 1. The present invention obtains a three-dimensional real scene model through oblique photography data from a drone, and at the same time obtains a three-dimensional model with marked sensitive factors corresponding to the three-dimensional real scene model, which can not only restore the authenticity of the target area, but also more intuitively analyze the impact of sensitive factors on project implementation.

[0013] 2. Compared with the traditional 3D modeling method, the UAV oblique photography modeling technology of the present invention has lower costs. The traditional method requires a lot of manpower, material resources and time resources, while the UAV oblique photography technology reduces the dependence on manpower and reduces labor costs through automation and intelligence. At the same time, due to the widespread application of UAV platforms and the popularity of oblique photography measurement technology, the cost of related equipment and software is also gradually decreasing, making the entire modeling process more economical and efficient. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 The present invention is a structural schematic diagram of a UAV oblique photography modeling method based on sensitive factor identification.

[0015] Figure 2 The present invention is a flow chart of a method for modeling UAV oblique photography based on sensitive factor identification. DETAILED DESCRIPTION

[0016] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0017] Embodiment 1, Figure 1 The present invention provides a UAV oblique photography modeling method based on sensitive factor identification, which includes the following steps: S1, obtain the orthophoto and oblique photography data of the target area; The data requires selecting appropriate weather conditions to implement drone operations, obtain orthophoto and oblique photography data of the target area, check the acquired data, and perform additional photography in areas where image information is missing to obtain complete image data of the target area.

[0018] In this embodiment, the high-resolution quarry surveying data provided by Wingtra UAV Manufacturing Company is taken as an example to construct a real-scene 3D model of the area and a 3D model with sensitive factors marked. A total of 941 oblique photography data of the target area are collected with a resolution of 7952*5304.

[0019] S2, performing uniform light, uniform color and distortion correction preprocessing on the acquired orthophoto and oblique photography data to obtain first data; The obtained data is subjected to preprocessing of uniform light and color and distortion correction, specifically: The Wallis filter method is used to implement uniform light and color processing, unify the grayscale mean and grayscale variance of the target area image, reduce the contrast value of the high-contrast interval, and increase the contrast value of the low-contrast interval; By estimating the distortion parameters and performing correction calculations on each pixel, a distortion-corrected image is obtained.

[0020] The Wallis filter is used for uniform light and color processing. The Wallis filter can map the grayscale mean and variance of the local image to the given grayscale mean and variance values, so that the grayscale variance and grayscale mean at different positions of the image have approximately equal values, that is, the contrast of the area with small contrast in the image is increased, and the contrast of the area with large contrast is reduced, so that the tiny information of grayscale in the image is enhanced. The filter can be expressed as: (1) In the formula, is the gray value of the original image; is the gray value of the image after Wallis transformation; is the local gray value of the original image; is the local grayscale standard deviation of the original image; is the target value of the local grayscale mean of the transformed image; is the target value of the local grayscale standard deviation of the transformed image; is the expansion coefficient of image variance; is the brightness coefficient of the image.

[0021] The distortion correction process includes radial distortion and tangential distortion. Radial distortion is caused by the shape of the lens. The reason for this is that light is more bent at a position far from the center of the lens than at a position close to the center. In actual situations, radial distortion is relatively small. Typical radial distortions include barrel distortion and pincushion distortion. The first three terms of the Taylor series expansion around the position are used to quantitatively describe the radial distortion of the lens. The specific radial distortion processing formula is: (2) in: is the coordinate before radial correction, is the radially corrected position coordinate, For point The distance from the principal point of the image, is the radial distortion parameter. And (3) Tangential distortion is caused by defects in lens manufacturing that make the lens itself non-parallel to the image plane. The specific tangential distortion correction formula is: (4) in, is the coordinate before tangential correction, is the position coordinate after tangential correction, For point The distance from the principal point of the image, is the tangential distortion parameter.

[0022] Combining equations (2) and (4), we can get the position coordinates after distortion correction: , and the relationship is shown in equation (5). Distortion is the change in distance, and the change in distance is reflected in the coordinate value as an addition and subtraction relationship.

[0023] (5) There are several main benefits of preprocessing the acquired data for uniform light, color and distortion correction: Improve image quality: Light uniformity processing can eliminate brightness variations caused by uneven light sources, making the image look more uniform and the details clearer. Color uniformity processing can correct color casts, making the image color more natural and realistic, which is convenient for subsequent analysis.

[0024] Improved data consistency: Corrected images are more consistent in brightness and color, which is especially important for multi-image comparison, classification, or clustering analysis, and helps improve the reliability of the results.

[0025] Reduce the effects of distortion: Distortion correction corrects image distortions caused by lens characteristics such as barrel or pincushion distortion, ensuring that the acquired image more accurately reflects the actual scene.

[0026] Improve the results of subsequent analysis: The preprocessed data is more suitable for further image processing and analysis, such as feature extraction, target detection, etc., which can improve the accuracy and efficiency of the algorithm.

[0027] Facilitates automated processing: Consistent and standardized data makes subsequent automated analysis and processing more efficient, reducing the need for human intervention.

[0028] Improve visual presentation: For images that need to be displayed (such as in reports or presentations), images that have been uniformly lit, color-uniformed, and distortion-corrected are usually more attractive and readable, improving the audience experience.

[0029] In short, performing these preprocessing steps can significantly improve the availability of data and the effectiveness of analysis, and is an important part of image processing and analysis.

[0030] S3, performing image segmentation on the first data, and marking preset sensitive factors in the image with different colors; The preprocessed data is segmented and sensitive factors in the image are marked with different colors, specifically: Prepare several oblique photography image data; Import oblique photography image data into AnyLabeling (automatic labeling software); Use the tools and interfaces provided by AnyLabeling to annotate sensitive factors in images by drawing bounding boxes, marking key points, adding text labels, etc. After the annotation is completed, the oblique photography image data is verified and edited; After the annotation is completed, the oblique photography image data is exported into a format for image segmentation model training; Use annotated datasets to train image segmentation models, continuously update datasets and retrain models; Use the trained model to perform image segmentation on the preprocessed image data, and mark the sensitive factors in the image with different colors.

[0031] In this embodiment, 170 oblique photography images are used for annotation and training, and the image segmentation model Deeplabv3+ is used for image segmentation, which accurately segments and marks sensitive factors in the pre-processed image with different colors.

[0032] S4, performing aerial triangulation on the pre-processed image data to generate a basic three-dimensional model; Aerial triangulation (AT) is a core process in photogrammetry, especially when using aerial or drone image data for 3D reconstruction. It mainly uses multiple images taken at different locations with overlapping areas to determine the 3D coordinates of corresponding points in these images through mathematical and computer vision techniques, and infer the precise position and attitude of the camera when these images were taken.

[0033] In this embodiment, firstly, aerial triangulation is performed with the help of the PixelGrid system to obtain exterior orientation elements, and the exterior orientation elements are combined with collinearity equations (the collinearity equations analyze the collinear relationship between ground points, camera image points and projection centers according to the principle of camera center projection during shooting) to generate a basic three-dimensional model.

[0034] The method of performing aerial triangulation on the pre-processed image data to generate a basic three-dimensional model is specifically as follows: Automatically detecting and extracting feature points from the preprocessed image, wherein the feature points include pixel values; By handling scale, rotation, and illumination changes between images, and using feature descriptors to find and match the same feature points between different images; Preliminarily estimate the camera's internal and external orientation elements based on the ground control points, the internal orientation elements including focal length and principal point, and the external orientation elements including position and posture; By adjusting the position, attitude parameters and feature point coordinates of the camera, the feature points with the smallest intersection error in the three-dimensional space are obtained; The basic three-dimensional model is generated according to the smallest feature points, internal and external elements combined with the collinear equation.

[0035] The collinear equation expression is: (5) The object space coordinates of the object point are , , Indicates that the object space coordinates of the camera point are , , Indicates that the internal orientation elements of the image are , , Indicates that the plane coordinates of the camera image point are expressed as , Indicates that the exterior orientation elements of the image are , , The three represent the heading angle, the sideways tilt angle and the image tilt angle respectively. The three direction cosines are expressed as , , express. S5, performing texture mapping on the image data after preprocessing and after marking sensitive factors based on the basic three-dimensional model, to generate a real-scene three-dimensional model of the target area and a three-dimensional model after marking sensitive factors; Texture Mapping, also known as texture mapping, is a process of mapping texture pixels in texture space to pixels in screen space (or 3D model surface). Simply put, it is to paste an image (texture) onto the surface of a 3D object to enhance the sense of reality. This technology can be combined with lighting calculation, image blending and other technologies to form many very beautiful and realistic visual effects.

[0036] In 3D graphics, texture mapping is the most widely used method, especially for describing realistic objects. Texture mapping can significantly improve the visual details and realism of the model without increasing the complexity of the model.

[0037] In this embodiment, texture coordinates are obtained through collinear equations, a relationship between two-dimensional texture space points and three-dimensional space surface points is created, the grayscale values ​​of the two-dimensional texture space points are mapped to the surface of the three-dimensional target area, and a three-dimensional real scene model of the target area and a three-dimensional model of the marked sensitive factors are obtained.

[0038] The texture mapping is performed on the image data after preprocessing and after sensitive factors are marked, respectively, to generate a real-scene three-dimensional model of the target area and a three-dimensional model after sensitive factors are marked, specifically: According to the collinear equation, the relationship between the external coordinate transformation and the image coordinate transformation is constructed, and the reverse texture mapping of the three-dimensional point coordinates is performed; Defining texture coordinates and rectangular texture space based on a preset texture coordinate formula; Through the exterior orientation elements, the points in the spatial coordinate system are projected into the image space, and then the corresponding relationship between these image space points and the original three-dimensional model space points is established; Based on the texture coordinates and the rectangular texture space, the two-dimensional texture coordinates are mapped to the surface of the three-dimensional model to obtain the relationship between the texture coordinates and the three-dimensional space points; By integrating the coordinate conversion relationship from three-dimensional to two-dimensional and the correspondence between texture coordinates and three-dimensional space points, the pixel value is used as the texture space point and mapped to the corresponding surface of the three-dimensional model to obtain the three-dimensional model.

[0039] This process completes the texture mapping and finally generates and outputs a 3D model with rich texture information. The model not only has accurate geometric shapes, but also has realistic surface materials and texture effects.

[0040] Let the texture coordinates be , the texture space is a rectangular domain, where , , the texture coordinates of each point are (6) The coordinates of the image point in the upper left corner of the rectangular domain are Indicates that the image point coordinates of the model surface vertex are expressed as Indicates that the pixel height and width of the rectangular domain are respectively , express. Based on the external orientation elements, the image space point in the spatial coordinate system is obtained , and construct the point and the model space point relationship. According to the texture coordinates, we can get relationship. Based on the above two relationships, Pixel value as texture space point , realize texture mapping and output three-dimensional model.

[0041] The benefits of texture mapping for generating 3D models are as follows: Improve model realism: Texture mapping significantly improves the realism of the model by adding rich colors, textures and details to the surface of the 3D model. This makes the 3D model visually closer to objects in the real world, enhancing the user's sense of immersion and trust.

[0042] Simplify the modeling process: Without texture mapping, in order to simulate the real appearance of an object, modelers need to create a large number of polygons and details. Texture mapping allows designers to "paint" these details through images, which greatly simplifies the modeling process and reduces the workload and time of modeling.

[0043] Enhanced visual effects: Texture mapping not only provides the model with rich colors and texture features, but also can create various special effects and patterns through different texture combinations and mapping methods. This makes the 3D model more visually attractive and can meet the visual needs of different application scenarios.

[0044] Support for sensitive factor marking: When texture mapping is performed on image data with sensitive factors marked, the sensitive factors can be marked with specific textures or colors. In this way, the sensitive factors will be clearly visible in the generated 3D model, which is convenient for subsequent analysis, evaluation and management.

[0045] Improve the visualization of the model: Texture mapping makes the 3D model more outstanding in visualization. Whether in virtual reality (VR), augmented reality (AR) or other visualization applications, texture mapping can provide users with a more realistic and vivid visual experience.

[0046] Support for multi-field applications: Texture mapping technology is widely used in digital archiving of cultural relics, film, television, advertising production, game development, architecture and design, etc. In these fields, texture mapping provides important support for creating realistic and attractive visual effects.

[0047] The above formulas are all dimensionless and numerical calculations. The formula is a formula for the most recent real situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0048] The above embodiments may be implemented in whole or in part by software, hardware, firmware or any other combination. When implemented by software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0049] Those of ordinary skill in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0050] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0051] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

[0052] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the protection scope of the present invention.

Claims

1. A UAV oblique photography modeling method based on sensitive factor identification, characterized in that: The following steps are involved: Obtain orthophoto and oblique photography data of the target area; Performing uniform light, uniform color and distortion correction preprocessing on the acquired orthophoto and oblique photography data to obtain first data; Performing image segmentation on the first data, and marking preset sensitive factors in the image with different colors; Perform aerial triangulation on the pre-processed image data to generate a basic three-dimensional model; Based on the basic 3D model, texture mapping is performed on the image data after preprocessing and after marking sensitive factors, so as to generate a real-scene 3D model of the target area and a 3D model after marking sensitive factors.

2. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 1, characterized in that: The obtaining of orthophoto and oblique photography data of the target area is specifically as follows: Carry out UAV operations to obtain orthophoto and oblique photography data of the target area, check the acquired data, and take additional photos in areas where image information is missing to obtain complete image data of the target area.

3. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 2 is characterized in that: The distortion correction includes radial distortion processing and tangential distortion correction processing.

4. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 3 is characterized in that: The acquired orthophoto and oblique photography data are preprocessed by uniform light and color and distortion correction to obtain the first data, specifically: The Wallis filter method is used to implement uniform light and color processing, unify the grayscale mean and grayscale variance of the target area image, reduce the contrast value of the high-contrast interval, and increase the contrast value of the low-contrast interval; By estimating the distortion parameters and performing correction calculations on each pixel, a distortion-corrected image is obtained.

5. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 4 is characterized in that: The image segmentation is performed on the first data, and preset sensitive factors in the image are marked with different colors, specifically: Prepare several oblique photography image data; Import oblique photography image data into AnyLabeling; Use the tools and interfaces provided by AnyLabeling to annotate sensitive factors in images by drawing bounding boxes, marking key points, adding text labels, etc. After the annotation is completed, the oblique photography image data is verified and edited; After the annotation is completed, the oblique photography image data is exported into a format for image segmentation model training; Use annotated datasets to train image segmentation models, continuously update datasets and retrain models; Use the trained model to perform image segmentation on the preprocessed image data, and mark the sensitive factors in the image with different colors.

6. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 5 is characterized in that: The method of performing aerial triangulation on the pre-processed image data to generate a basic three-dimensional model is specifically as follows: Automatically detecting and extracting feature points from the preprocessed image, wherein the feature points include pixel values; By handling scale, rotation, and illumination changes between images, and using feature descriptors to find and match the same feature points between different images; Preliminarily estimate the camera's internal and external orientation elements based on the ground control points, the internal orientation elements including focal length and principal point, and the external orientation elements including position and posture; By adjusting the position, attitude parameters and feature point coordinates of the camera, the feature points with the smallest intersection error in the three-dimensional space are obtained; The basic three-dimensional model is generated according to the smallest feature points, internal and external elements combined with the collinear equation.

7. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 6 is characterized in that: The texture mapping is performed on the image data after preprocessing and after sensitive factors are marked, respectively, to generate a real-scene three-dimensional model of the target area and a three-dimensional model after sensitive factors are marked, specifically: According to the collinear equation, the relationship between the external coordinate transformation and the image coordinate transformation is constructed, and the reverse texture mapping of the three-dimensional point coordinates is performed; Defining texture coordinates and rectangular texture space based on a preset texture coordinate formula; Through the exterior orientation elements, the points in the spatial coordinate system are projected into the image space, and then the corresponding relationship between these image space points and the original three-dimensional model space points is established; Based on the texture coordinates and the rectangular texture space, the two-dimensional texture coordinates are mapped to the surface of the three-dimensional model to obtain the relationship between the texture coordinates and the three-dimensional space points; By integrating the coordinate conversion relationship from three-dimensional to two-dimensional and the correspondence between texture coordinates and three-dimensional space points, the pixel value is used as the texture space point and mapped to the corresponding surface of the three-dimensional model to obtain the three-dimensional model.

8. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 7 is characterized in that: The specific calculation formula of the Wallis filter is as follows: In the formula, is the gray value of the original image; is the gray value of the image after Wallis transformation; is the local gray value of the original image; is the local grayscale standard deviation of the original image; is the target value of the local grayscale mean of the transformed image; is the target value of the local grayscale standard deviation of the transformed image; is the expansion coefficient of image variance; is the brightness coefficient of the image.

9. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 8, characterized in that: The radial distortion processing formula is: The tangential distortion correction processing formula is: in, is the coordinate before radial correction, is the radially corrected position coordinate, For point The distance from the principal point of the image, is the radial distortion parameter is the coordinate before tangential correction, is the position coordinate after tangential correction, For point The distance from the principal point of the image, is the tangential distortion parameter.

10. The method for modeling UAV oblique photography based on sensitive factor identification according to claim 9, characterized in that: The collinear equation is specifically: The object space coordinates of the object point are , , Indicates that the object space coordinates of the camera point are , , Indicates that the internal orientation elements of the image are , , Indicates that the plane coordinates of the camera image point are expressed as , Indicates that the exterior orientation elements of the image are , , The three represent the heading angle, the sideways tilt angle and the image tilt angle respectively. The three direction cosines are expressed as , , express.

Images