Real scene three-dimensional modeling method and system based on unmanned aerial vehicle oblique photogrammetry

Feature points are extracted and point cloud data are optimized through the SLFT algorithm, combined with Poisson reconstruction and grid simplification algorithm, and the problem of low modeling quality in drone tilt photogrammetry is solved, achieving efficient three-dimensional modeling process and accurate Terrain reconstruction.

CN118967980AInactive Publication Date: 2024-11-15CHINA UNIV OF GEOSCIENCES (WUHAN)

Patent Information

Application Number
CN202411448491.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-10-17
Publication Date
2024-11-15
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The existing real-life three-dimensional modeling method for drone tilt photogrammetry has the problem of low modeling quality, long processing cycles and inconsistent data, and conventional methods fail to effectively build point clouds, resulting in a decline in model geometric quality.

Method used

The SLFT algorithm is used to extract key feature points, generate point cloud data through the closest neighbor matching method and optimize it, combine Poisson reconstruction and grid simplification algorithm to generate a three-dimensional model, and optimize the model through the triangular edge collapse method to reduce redundant information and details.

Benefits of technology

Improves the construction accuracy and overall quality of the model, reduces processing cycles, and enhances the geometric details and accuracy of the model.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118967980A_ABST
    Figure CN118967980A_ABST
Patent Text Reader

Abstract

The invention provides a live-action three-dimensional modeling method and system based on unmanned aerial vehicle oblique photogrammetry, and relates to the field of live-action modeling methods.The method comprises the steps that an obtained actual oblique photogrammetry image is preprocessed, and preprocessing comprises distortion correction, brightness and contrast adjustment and denoising processing; extracting key feature points of the preprocessed actual oblique photography image by adopting an SLFT algorithm; on the basis of the key feature points, feature point matching is carried out by adopting a nearest neighbor matching method, and point cloud data is generated and optimized; performing surface reconstruction based on the optimized point cloud data to generate a three-dimensional model; mapping the texture of the actual oblique photography image into the three-dimensional model to obtain a preliminary model; and based on a triangular edge collapsing method, optimizing the preliminary model to obtain a terrain live-action three-dimensional model, and completing live-action three-dimensional modeling of unmanned aerial vehicle oblique photogrammetry.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of real-scene modeling methods, and in particular to a real-scene three-dimensional modeling method and system based on unmanned aerial vehicle oblique photogrammetry. Background Art

[0002] UAV oblique photogrammetry is a technology based on aerial photogrammetry using a UAV equipped with an oblique photography system. The UAV can capture images of different angles and directions on the ground by carrying an oblique photography system, thereby achieving all-round, multi-angle high-resolution image acquisition of the target area. The real-scene 3D modeling method based on UAV oblique photogrammetry is a method of subsequent modeling based on the obtained image data. In the actual modeling process, the raw data obtained through drone oblique photography is relatively large, and the conventional processing and conversion methods take a long time. The data is inconsistent, and the conventional modeling method does not build a point cloud based on the data and perform subsequent point cloud resampling. Therefore, the processed raw data contains a lot of redundant information and unnecessary details, which will reduce the geometric quality of the detailed model. Therefore, it is urgent to propose a real-scene 3D modeling method and system based on UAV oblique photogrammetry. Summary of the invention

[0003] The purpose of the present invention is to provide a real-scene 3D modeling method and system based on UAV tilt photogrammetry in order to solve the problem of modeling quality disclosure in the conventional real-scene 3D modeling method based on UAV tilt photogrammetry.

[0004] The above-mentioned purpose of the present application is achieved through the following technical solutions: S1: Acquire the actual oblique photography image of the modeling area; S2: Preprocessing the actual oblique photography image, including: distortion correction, brightness and contrast adjustment, and denoising; S3: The SLFT algorithm is used to extract the key feature points of the actual oblique photography image after preprocessing; based on the key feature points, the nearest neighbor matching method is used to match the feature points, generate point cloud data and optimize it; S4: Surface reconstruction is performed based on the optimized point cloud data to generate a three-dimensional model; the texture of the actual oblique photography image is mapped to the three-dimensional model to obtain a preliminary model; S5: Based on the triangle edge collapse method, the preliminary model is optimized to obtain the real-scene 3D model of the terrain, completing the real-scene 3D modeling of the UAV oblique photogrammetry.

[0005] Optionally, step S1 includes: S11: Determine the flight path of the drone based on the modeling area; divide the flight path into a preset number of terminals based on the flight path; confirm the shooting point of each terminal based on the set terminals; S12: Calculate the flight parameters of the UAV; the flight parameters include: flight altitude and speed; Calculate flight altitude based on the topographic relief of the modeled area and the required image resolution; Assume that the terrain height range of the modeling area is arrive , the required image resolution is , the flight altitude is calculated as:

[0006] Assume the length of the flight path is , the flight altitude is , speed , then the speed is calculated as:

[0007] In the formula, is the total time from takeoff to landing; S13: At each shooting point, the fisheye lens of the drone is used to perform oblique photography to obtain an actual oblique photography image of the modeling area.

[0008] Optionally, step S2 includes: S21: Use the fisheye lens correction algorithm to perform distortion correction on the actual oblique photography image as follows: If the distortion of the drone fisheye lens is radially symmetrical, the distortion correction calculation process is:

[0009] In the formula, is the radial distance after distortion, is the radial angle before distortion, is the radial distortion coefficient; If the distortion of the drone fisheye lens is linear, the distortion correction calculation formula is:

[0010] In the formula, is the radial distance after distortion, is the radial angle before distortion, and is the radial distortion coefficient; S22: using a histogram equalization method to adjust the brightness and contrast of the actual oblique photography image; The pixel values ​​of the actual oblique photography image are redistributed by the cumulative distribution function as follows:

[0011] In the formula, is the pixel value The corresponding transformed pixel value, is the gray level of the actual oblique photography image, and Represents width and height respectively, Is the pixel value The number of pixels; S23: Using a median filter processing method to perform denoising on the actual oblique photography image.

[0012] Optionally, step S3 includes: S31: The key feature points of the actual oblique photography image after preprocessing are extracted using the SLFT algorithm, as follows: S31a: construct a Gaussian difference pyramid; detect the extreme points of the scale space in the actual oblique photography image after preprocessing through the Gaussian difference pyramid, that is, the candidate key points; the Gaussian difference function is specifically defined as:

[0013] In the formula, is a Gaussian function, is the scale factor, is the actual oblique photography image, Represents the convolution operation; S31b: locate and describe the candidate key points to obtain key feature points; Candidate key point positioning: accurately locate the position, scale and direction of candidate key points; Candidate key point description: Calculate the descriptor of the candidate key point in the neighborhood around the candidate key point, use the feature vector to describe the local image features of the candidate key point neighborhood, and extract the distinguishable key feature points in the actual oblique photography image; S32: Based on the key feature points, a nearest neighbor matching method is used to match the feature points to obtain a feature point matching pair; S32a: Given two actual oblique photography images and ,image Key features of There is a corresponding feature descriptor ;image Key features of There is a corresponding feature descriptor ; The Euclidean distance method is used to calculate two feature descriptors. and The similarity between:

[0014] In the formula, is the dimension of the feature descriptor, and Represents the first dimension; represents the similarity function; S32b: constructing feature point matching pairs according to the calculated similarity; S33: For each feature point matching pair, based on the camera parameters and parallax of the feature point, calculate the corresponding point of the feature point matching pair in the three-dimensional space; and combine the corresponding points into an overall point cloud data; S33a: Calculate each feature point matching pair The disparity vector , the disparity vector In actual oblique photography images and The difference in pixel coordinates of the feature points in ; S33b: Camera parameters based on feature points, feature points and The pixel coordinates and disparity in the image plane are combined with the triangulation method to calculate the feature points and Corresponding points in 3D space ; Combine the corresponding points of each feature point matching pair into an overall point cloud data; S34: registering the point cloud data under multiple viewing angles so that the point cloud data are aligned in the same coordinate system; The nearest point iteration algorithm is used to optimize the registration of point cloud data by minimizing the error between point clouds. The optimization objective function is expressed as:

[0015] In the formula, and Respectively represent the points to be registered in the two point clouds, is the transformation to be optimized; It represents the parameter value at which a function takes its minimum value in its domain; S35: Evaluate the registered point cloud data, compare the point cloud features before and after the registration, and obtain an evaluation result; S36: Adjust the registered point cloud data according to the evaluation result.

[0016] Optionally, step S4 includes: S41: meshing the point cloud data using a Poisson reconstruction algorithm to obtain a triangular mesh of the point cloud data; S42: using a mesh simplification algorithm to optimize the triangular mesh to obtain a three-dimensional model; Using the grid simplification algorithm, the objective function is calculated to minimize the error metric, as follows:

[0017] In the formula, is the set of vertices of the triangle mesh, is the vertex, are the vertices in the triangle mesh The error measure of Represents the objective function When the minimum value is taken, the independent variable The value of S43: Mapping the texture of the actual oblique photography image to the three-dimensional model to obtain a preliminary model.

[0018] Optionally, step S5 includes: S51: Use median filtering to remove noise from the preliminary model; S52: Use Alpha Shape algorithm to fill holes in the preliminary model; S53: Based on the triangle edge collapse method, the preliminary model after filling holes and removing noise is optimized; The method of triangle edge collapse is as follows:

[0019] In the formula, is the vertex of the edge to be collapsed, yes The adjacent vertices of neighbor) is Error measure between adjacent vertices; Represents the objective function When the minimum value is taken, the independent variable The value of .

[0020] A real-scene three-dimensional modeling system based on drone oblique photogrammetry, the system comprising: drone equipment, a computer, and a display; The drone equipment is used to collect actual oblique photography images; The computer is used to pre-process the actual oblique photography image, and the pre-processing includes distortion correction, brightness and contrast adjustment, and noise removal; The computer is also used to extract key feature points of the actual oblique photography image after preprocessing by using the SLFT algorithm; based on the key feature points, the nearest neighbor matching method is used to match the feature points, generate point cloud data and optimize it; The computer is also used to perform surface reconstruction based on the optimized point cloud data to generate a three-dimensional model; map the texture of the actual oblique photography image to the three-dimensional model to obtain a preliminary model; The computer is also used to optimize the preliminary model, obtain a real-scene three-dimensional model of the terrain, and complete the real-scene three-dimensional modeling of the drone oblique photogrammetry; The display is used for three-dimensionally visualizing the terrain real-scene three-dimensional model.

[0021] A computer-readable storage medium stores instructions. When the instructions are executed, a real-scene three-dimensional modeling method based on unmanned aerial vehicle oblique photogrammetry is executed.

[0022] The beneficial effects of the technical solution provided by this application are: 1. Use mesh simplification algorithm to optimize the triangular mesh, remove unnecessary triangles, reduce mesh complexity, and improve model construction accuracy.

[0023] 2. Based on the triangle edge collapse method, the preliminary model after filling holes and removing noise is optimized to reduce the complexity of the preliminary model and improve the overall quality of the model. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] The present application will be further described below with reference to the accompanying drawings and embodiments, in which: Figure 1 is a step diagram in an embodiment of the present application; Figure 2 It is a system module diagram in an embodiment of the present application. DETAILED DESCRIPTION

[0025] In order to have a clearer understanding of the technical features, purposes and effects of the present application, the specific implementation methods of the present application are now described in detail with reference to the accompanying drawings.

[0026] The embodiment of the present application provides a real-scene three-dimensional modeling method based on drone oblique photogrammetry.

[0027] Please refer to Figure 1 , Figure 1 1 is a step diagram of a real scene 3D modeling method based on drone oblique photogrammetry in an embodiment of the present application, comprising: S1: Acquire the actual oblique photography image of the modeling area; Step S1 includes: S11: Determine the flight path of the drone based on the modeling area; divide the flight path into a preset number of terminals based on the flight path; confirm the shooting point of each terminal based on the set terminals; S12: Calculate the flight parameters of the UAV; the flight parameters include: flight altitude and speed; Calculate flight altitude based on the topographic relief of the modeled area and the required image resolution; Assume that the terrain height range of the modeling area is arrive , the required image resolution is , the flight altitude is calculated as:

[0028] Assume the length of the flight path is , the flight altitude is , speed , then the speed is calculated as:

[0029] In the formula, is the total time from takeoff to landing; S13: At each shooting point, the fisheye lens of the drone is used to perform oblique photography to obtain an actual oblique photography image of the modeling area.

[0030] S2: Preprocessing the actual oblique photography image, including: distortion correction, brightness and contrast adjustment, and denoising; Step S2 includes: S21: Use the fisheye lens correction algorithm to perform distortion correction on the actual oblique photography image as follows: If the distortion of the drone fisheye lens is radially symmetrical, the distortion correction calculation process is:

[0031] In the formula, is the radial distance after distortion, is the radial angle before distortion, is the radial distortion coefficient; If the distortion of the drone fisheye lens is linear, the distortion correction calculation formula is:

[0032] In the formula, is the radial distance after distortion, is the radial angle before distortion, and is the radial distortion coefficient; Specifically, based on the calculation results of the fisheye lens correction algorithm, the obtained results are used to describe the distortion characteristics of the fisheye lens, and the actual oblique photography image is subjected to distortion correction so that the obtained image is close to the real scene for subsequent use in three-dimensional modeling.

[0033] S22: using a histogram equalization method to adjust the brightness and contrast of the actual oblique photography image; The pixel values ​​of the actual oblique photography image are redistributed by the cumulative distribution function as follows:

[0034] In the formula, is the pixel value The corresponding transformed pixel value, is the gray level of the actual oblique photography image, and Represents width and height respectively, Is the pixel value The number of pixels; Specifically, a histogram equalization method is used to calculate the histogram of the image, and the pixel values ​​of the image are redistributed based on the histogram equalization formula to adjust the brightness and contrast of the image and enhance the clarity of the details in the image for subsequent feature extraction and registration processing.

[0035] S23: Using a median filtering method to perform denoising on the actual oblique photography image.

[0036] Specifically, The coordinates in the image are The pixel value of the median filter is calculated as follows:

[0037] In the formula, and is the pixel index of the neighborhood, Represents the operation of sorting pixel values ​​and taking the median. The median value of the pixel's surrounding neighborhood is used to replace the current pixel's value, reducing the impact of noise in the image.

[0038] S3: The SLFT algorithm is used to extract the key feature points of the actual oblique photography image after preprocessing; based on the key feature points, the nearest neighbor matching method is used to match the feature points, generate point cloud data and optimize it; Step S3 includes: S31: The key feature points of the actual oblique photography image after preprocessing are extracted using the SLFT algorithm, as follows: S31a: construct a Gaussian difference pyramid; detect the extreme points of the scale space in the actual oblique photography image after preprocessing through the Gaussian difference pyramid, that is, the candidate key points; the Gaussian difference function is specifically defined as:

[0039] In the formula, is a Gaussian function, is the scale factor, is the actual oblique photography image, Represents the convolution operation; Specifically, extreme points are found in the Gaussian difference pyramid, and local maximum or minimum values ​​are selected as candidate key points.

[0040] S31b: locate and describe the candidate key points to obtain key feature points; Candidate key point positioning: accurately locate the position, scale and direction of candidate key points; Candidate key point description: Calculate the descriptor of the candidate key point in the neighborhood around the candidate key point, use the feature vector to describe the local image features of the candidate key point neighborhood, and extract the distinguishable key feature points in the actual oblique photography image; S32: Based on the key feature points, a nearest neighbor matching method is used to match the feature points to obtain a feature point matching pair; S32a: Given two actual oblique photography images and ,image Key features of There is a corresponding feature descriptor ;image Key features of There is a corresponding feature descriptor ; The Euclidean distance method is used to calculate two feature descriptors. and The similarity between:

[0041] In the formula, is the dimension of the feature descriptor, and Represents the first dimension; represents the similarity function; S32b: constructing feature point matching pairs according to the calculated similarity; S33: For each feature point matching pair, based on the camera parameters and parallax of the feature point, calculate the corresponding point of the feature point matching pair in the three-dimensional space; and combine the corresponding points into an overall point cloud data; S33a: Calculate each feature point matching pair The disparity vector , the disparity vector In actual oblique photography images and The difference in pixel coordinates of the feature points in ; S33b: Camera parameters based on feature points, feature points and The pixel coordinates and disparity in the image plane are combined with the triangulation method to calculate the feature points and Corresponding points in 3D space ; Combine the corresponding points of each feature point matching pair into an overall point cloud data; S34: registering the point cloud data under multiple viewing angles so that the point cloud data are aligned in the same coordinate system; Specifically, multi-view point cloud coarse registration refers to the process of preliminarily aligning point cloud datasets from multiple perspectives. It is usually used when the point cloud dataset is large, the initial position is unknown, or there is a large position offset.

[0042] The nearest point iteration algorithm is used to optimize the registration of point cloud data by minimizing the error between point clouds. The optimization objective function is expressed as:

[0043] In the formula, and Respectively represent the points to be registered in the two point clouds, is the transformation to be optimized; It represents the parameter value at which a function takes its minimum value in its domain; S35: Evaluate the registered point cloud data, compare the point cloud features before and after the registration, and obtain an evaluation result; S36: Adjust the registered point cloud data according to the evaluation result.

[0044] S4: Surface reconstruction is performed based on the optimized point cloud data to generate a three-dimensional model; the texture of the actual oblique photography image is mapped to the three-dimensional model to obtain a preliminary model; Step S4 includes: S41: meshing the point cloud data using a Poisson reconstruction algorithm to obtain a triangular mesh of the point cloud data; S42: using a mesh simplification algorithm to optimize the triangular mesh to obtain a three-dimensional model; Specifically, the triangular mesh is optimized, unnecessary triangles are removed, and the mesh complexity is reduced.

[0045] Using the grid simplification algorithm, the objective function is calculated to minimize the error metric, as follows:

[0046] In the formula, is the vertex set of the triangle mesh, is the vertex, are the vertices in the triangle mesh The error measure of Represents the objective function When the minimum value is taken, the independent variable The value of S43: Mapping the texture of the actual oblique photography image to the three-dimensional model to obtain a preliminary model.

[0047] Specifically, for each triangle in the three-dimensional model, based on its position in the original image, the texture is intercepted from the corresponding position and mapped to the corresponding three-dimensional model to complete the preliminary modeling process.

[0048] S5: Based on the triangle edge collapse method, the preliminary model is optimized to obtain the real-scene 3D model of the terrain, completing the real-scene 3D modeling of the UAV oblique photogrammetry.

[0049] Step S5 includes: S51: Use median filtering to remove noise from the preliminary model; S52: Use Alpha Shape algorithm to fill holes in the preliminary model; Specifically, to identify holes in the preliminary model, Alpha Shape is a geometric structure based on point sets, which is used to describe the geometric shape around the point set. It is defined as the intersection between the convex hull of the point set and the largest empty sphere of each point in the point set. The construction formula is as follows:

[0050] In the formula, is a set of points, is a parameter, express of Triangulation, represents the largest empty sphere with edge as diameter; S53: Based on the triangle edge collapse method, the preliminary model after filling holes and removing noise is optimized; The method of triangle edge collapse is as follows:

[0051] In the formula, is the vertex of the edge to be collapsed, yes The adjacent vertices of neighbor) is Error measure between adjacent vertices; Represents the objective function When the minimum value is taken, the independent variable The value of .

[0052] The optimized real-scene 3D model of the terrain is compared with the actual scene to check the consistency between the model and the actual scene. The precision evaluation of point cloud resampling is used as the model evaluation method to resample the point cloud data to reduce the density and complexity of the point cloud. The resampled point cloud data is compared with the actual scene to evaluate the precision and accuracy of the model. The root mean square error is used as the calculation method, specifically:

[0053] In the formula, is the number of points in the point cloud, It is The root mean square error is the average of the squares of the distances between the resampled position and the actual position of each point in the point cloud, and then the square root is taken. Therefore, the smaller the value obtained, the smaller the average deviation between the resampled model and the actual scene, which indicates that the overall accuracy of the model is higher.

[0054] Specifically, the mapped three-dimensional model is optimized to remove noise and fill holes to improve the overall quality of the model. The optimized model is compared with the actual scene to complete the model accuracy assessment.

[0055] Please refer to Figure 2 , Figure 2 It is a module diagram of a real-scene 3D modeling system based on drone oblique photogrammetry in an embodiment of the present application, the system comprising: drone equipment, a computer, and a display; The drone equipment is used to collect actual oblique photography images; The computer is used to pre-process the actual oblique photography image, and the pre-processing includes distortion correction, brightness and contrast adjustment, and noise removal; The computer is also used to extract key feature points of the actual oblique photography image after preprocessing by using the SLFT algorithm; based on the key feature points, the nearest neighbor matching method is used to match the feature points, generate point cloud data and optimize it; The computer is also used to perform surface reconstruction based on the optimized point cloud data to generate a three-dimensional model; map the texture of the actual oblique photography image to the three-dimensional model to obtain a preliminary model; The computer is also used to optimize the preliminary model, obtain a real-scene three-dimensional model of the terrain, and complete the real-scene three-dimensional modeling of the drone oblique photogrammetry; The display is used for three-dimensionally visualizing the terrain real-scene three-dimensional model.

[0056] Specifically, the photos taken by the drone equipment are transmitted to the computer via wireless transmission.

[0057] The present application also discloses a computer-readable storage medium, which stores a plurality of instructions, and the instructions are suitable for a processor to load to execute the above-mentioned real-scene three-dimensional modeling method based on drone oblique photogrammetry.

[0058] The above are only exemplary embodiments of the present disclosure and cannot be used to limit the scope of the present disclosure. That is, any equivalent changes and modifications made according to the teachings of the present disclosure are still within the scope of the present disclosure. After considering the disclosure of the specification and the truth of practice, those skilled in the art will easily think of other embodiments of the present disclosure.

[0059] This application is intended to cover any variation, use or adaptation of the present disclosure, which follows the general principles of the present disclosure and includes common knowledge or customary technical means in the art not described in the present disclosure. The description and examples are to be regarded as exemplary only, and the scope and spirit of the present disclosure are defined by the claims.

Claims

1. A real scene 3D modeling method based on UAV oblique photogrammetry, characterized in that: The method comprises the following steps: S1: Acquire the actual oblique photography image of the modeling area; S2: Preprocessing the actual oblique photography image, including: distortion correction, brightness and contrast adjustment, and denoising; S3: The SLFT algorithm is used to extract the key feature points of the actual oblique photography image after preprocessing; based on the key feature points, the nearest neighbor matching method is used to match the feature points, generate point cloud data and optimize it; S4: Surface reconstruction is performed based on the optimized point cloud data to generate a three-dimensional model; the texture of the actual oblique photography image is mapped to the three-dimensional model to obtain a preliminary model; S5: Based on the triangle edge collapse method, the preliminary model is optimized to obtain the real-scene 3D model of the terrain, completing the real-scene 3D modeling of the UAV oblique photogrammetry.

2. A real-scene 3D modeling method based on drone oblique photogrammetry as claimed in claim 1, characterized in that: Step S1 includes: S11: Determine the flight path of the drone based on the modeling area; divide the flight path into a preset number of terminals based on the flight path; confirm the shooting point of each terminal based on the set terminals; S12: Calculate the flight parameters of the UAV; the flight parameters include: flight altitude and speed; Calculate flight altitude based on the topographic relief of the modeled area and the required image resolution; Assume that the terrain height range of the modeling area is arrive , the required image resolution is , the flight altitude is calculated as: Assume the length of the flight path is , the flight altitude is , speed , then the speed is calculated as: In the formula, is the total time from takeoff to landing; S13: At each shooting point, the fisheye lens of the drone is used to perform oblique photography to obtain an actual oblique photography image of the modeling area.

3. A real scene 3D modeling method based on drone oblique photogrammetry as claimed in claim 1, characterized in that: Step S2 includes: S21: Use the fisheye lens correction algorithm to perform distortion correction on the actual oblique photography image as follows: If the distortion of the drone fisheye lens is radially symmetrical, the distortion correction calculation process is: In the formula, is the radial distance after distortion, is the radial angle before distortion, is the radial distortion coefficient; If the distortion of the drone fisheye lens is linear, the distortion correction calculation formula is: In the formula, is the radial distance after distortion, is the radial angle before distortion, and is the radial distortion coefficient; S22: using a histogram equalization method to adjust the brightness and contrast of the actual oblique photography image; The pixel values ​​of the actual oblique photography image are redistributed by the cumulative distribution function as follows: In the formula, is the pixel value The corresponding transformed pixel value, is the gray level of the actual oblique photography image, and Represents width and height respectively, Is the pixel value The number of pixels; S23: Using a median filter processing method to perform denoising on the actual oblique photography image.

4. The real scene 3D modeling method based on drone oblique photogrammetry according to claim 1, characterized in that: Step S3 includes: S31: The key feature points of the actual oblique photography image after preprocessing are extracted using the SLFT algorithm, as follows: S31a: construct a Gaussian difference pyramid; detect the extreme points of the scale space in the actual oblique photography image after preprocessing through the Gaussian difference pyramid, that is, the candidate key points; the Gaussian difference function is specifically defined as: In the formula, is a Gaussian function, is the scale factor, is the actual oblique photography image, Represents the convolution operation; S31b: locate and describe the candidate key points to obtain key feature points; Candidate key point positioning: accurately locate the position, scale and direction of candidate key points; Candidate key point description: Calculate the descriptor of the candidate key point in the neighborhood around the candidate key point, use the feature vector to describe the local image features of the candidate key point neighborhood, and extract the distinguishable key feature points in the actual oblique photography image; S32: Based on the key feature points, a nearest neighbor matching method is used to match the feature points to obtain a feature point matching pair; S32a: Given two actual oblique photography images and ,image Key features of There is a corresponding feature descriptor ;image Key features of There is a corresponding feature descriptor ; The Euclidean distance method is used to calculate two feature descriptors. and The similarity between: In the formula, is the dimension of the feature descriptor, and Respectively represent the first dimension; represents the similarity function; S32b: constructing feature point matching pairs according to the calculated similarity; S33: For each feature point matching pair, based on the camera parameters and parallax of the feature point, calculate the corresponding point of the feature point matching pair in the three-dimensional space; and combine the corresponding points into an overall point cloud data; S33a: Calculate each feature point matching pair The disparity vector , the disparity vector In actual oblique photography images and The difference in pixel coordinates of the feature points in ; S33b: Camera parameters based on feature points, feature points and The pixel coordinates and disparity in the image plane are combined with the triangulation method to calculate the feature points and Corresponding points in 3D space ; Combine the corresponding points of each feature point matching pair into a whole point cloud data; S34: registering the point cloud data under multiple viewing angles so that the point cloud data are aligned in the same coordinate system; The nearest point iteration algorithm is used to optimize the registration of point cloud data by minimizing the error between point clouds. The optimization objective function is expressed as: In the formula, and Respectively represent the points to be registered in the two point clouds, is the transformation to be optimized; It represents the parameter value at which a function takes its minimum value in its domain; S35: Evaluate the registered point cloud data, compare the point cloud features before and after the registration, and obtain an evaluation result; S36: Adjust the registered point cloud data according to the evaluation result.

5. The real scene 3D modeling method based on drone oblique photogrammetry according to claim 1, characterized in that: Step S4 includes: S41: meshing the point cloud data using a Poisson reconstruction algorithm to obtain a triangular mesh of the point cloud data; S42: using a mesh simplification algorithm to optimize the triangular mesh to obtain a three-dimensional model; Using the grid simplification algorithm, the objective function is calculated to minimize the error metric, as follows: In the formula, is the set of vertices of the triangle mesh, is the vertex, are the vertices in the triangle mesh The error measure of Represents the objective function When the minimum value is taken, the independent variable The value of S43: Mapping the texture of the actual oblique photography image to the three-dimensional model to obtain a preliminary model.

6. The real scene 3D modeling method based on drone oblique photogrammetry according to claim 1, characterized in that: Step S5 includes: S51: Use median filtering to remove noise from the preliminary model; S52: Use Alpha Shape algorithm to fill holes in the preliminary model; S53: Based on the triangle edge collapse method, the preliminary model after filling holes and removing noise is optimized; The method of triangle edge collapse is as follows: In the formula, is the vertex of the edge to be collapsed, yes The adjacent vertices of neighbor) is Error measure between adjacent vertices; Represents the objective function When the minimum value is taken, the independent variable The value of .

7. A real-scene 3D modeling system based on drone oblique photogrammetry, used to implement a real-scene 3D modeling method based on drone oblique photogrammetry as described in claims 1-6, characterized in that: The system includes: drone equipment, a computer, and a display; The drone equipment is used to collect actual oblique photography images; The computer is used to pre-process the actual oblique photography image, and the pre-processing includes distortion correction, brightness and contrast adjustment, and noise removal; The computer is also used to extract key feature points of the pre-processed actual oblique photography image using the SLFT algorithm; based on the key feature points, the nearest neighbor matching method is used to match the feature points, generate point cloud data and optimize it; The computer is also used to perform surface reconstruction based on the optimized point cloud data to generate a three-dimensional model; map the texture of the actual oblique photography image to the three-dimensional model to obtain a preliminary model; The computer is also used to optimize the preliminary model, obtain a real-scene three-dimensional model of the terrain, and complete the real-scene three-dimensional modeling of the drone oblique photogrammetry; The display is used for three-dimensionally visualizing the terrain real-scene three-dimensional model.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores instructions, and when the instructions are executed by a computer, the method according to any one of claims 1 to 6 is executed.

Citation Information

Patent Citations

  • Contact network three-dimensional reconstruction method based on SIFT and LBP point cloud registration

    CN104299260A

  • Distortion parameter based fisheye camera calibration method

    CN107248178A

  • Automatic modeling method and device, computer equipment and storage medium

    CN114140581A

  • Three-dimensional reconstruction method integrating laser radar and oblique photography

    CN116740288A

  • Method and system for fusing multi-channel continuous video and three-dimensional twinborn scene on traffic road

    CN117152400A

Cited By

  • Three-dimensional reconstruction and reproduction system and method for oblique photography

    CN119672250A

  • Image monomer method based on live-action three dimensions

    CN119762687A

  • News scene three-dimensional reconstruction and visualization method based on multi-source remote sensing data

    CN119904592A

  • 3D Reconstruction and Visualization Method of News Scenes Based on Multi-Source Remote Sensing Data

    CN119904592B

  • Three-dimensional model generation method based on oblique photogrammetry of unmanned aerial vehicle

    CN121120940A