Building feature line extraction method based on unmanned aerial vehicle inclined three-dimensional model

Abstract

The invention discloses a building characteristic line extraction method based on an unmanned aerial vehicle inclined three-dimensional model, and relates to the technical field of building characteristic analysis, and the method comprises the following steps: employing an unmanned aerial vehicle equipped with an inclined photography system to shoot a building from a plurality of angles, and obtaining the data of a plurality of photos of the building; and preprocessing the acquired photo data, and importing the photo data into three-dimensional modeling software to construct an inclined three-dimensional model of the building. The building is shot from different angles, the multi-view image data of the building can be obtained, the inclined three-dimensional model of the building can be constructed by combining the three-dimensional modeling technology, on the basis, the feature lines of the building can be extracted by utilizing the feature line extraction algorithm, and compared with traditional manual measurement or two-dimensional image extraction, the method has the advantages that the construction efficiency is improved, and the construction cost is reduced. The precision of feature extraction is greatly improved, the working efficiency is remarkably improved, more details of the building can be captured, and feature line extraction is more accurate.

Description

Technical Field

[0001] The present invention relates to the technical field of building feature analysis, and in particular to a method for extracting building feature lines based on an inclined three-dimensional model of an unmanned aerial vehicle. Background Art

[0002] UAV oblique photogrammetry technology is a high-tech that has developed rapidly in the international photogrammetry field in recent years. This technology uses a multi-lens camera carried by a drone to synchronously collect images from five different vertical and four oblique perspectives, thereby obtaining high-resolution textures of the top and sides of the building. This multi-angle, high-resolution image data provides a rich information basis for the extraction of building feature lines.

[0003] Buildings usually have complex geometric shapes and texture features. Especially when buildings have complex topological relationships such as self-intersection and mutual occlusion, it will affect the accuracy of the three-dimensional model and cause great interference to the extraction of feature lines. Therefore, how to optimize the extracted feature lines to improve the extraction accuracy of building feature lines is the problem we need to solve. To this end, a building feature line extraction method based on UAV tilted three-dimensional model is proposed. Summary of the invention

[0004] The purpose of the present invention is to provide a method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle to solve the problems raised in the above-mentioned background technology.

[0005] In order to solve the above technical problems, the technical solution adopted by the present invention is: A method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle comprises the following steps: Step 1: Use a drone equipped with a tilt photography system to photograph the building from multiple angles to obtain multiple photo data of the building; Step 2: pre-process the collected photo data and import them into 3D modeling software to construct a tilted 3D model of the building; Step 3: Convert the constructed tilted 3D model into point cloud data, analyze the point cloud data, and identify the category of the point cloud data; Step 4: Based on the point cloud data category analysis, the point cloud data is processed using a point cloud processing algorithm to extract the characteristic lines of the building, and the characteristic line matching index is calculated to evaluate the matching degree of the characteristic lines; Step 5: Reconstruct the 3D model using 3D modeling software based on the extracted building feature lines; Step 6: Use the texture mapping algorithm to collect texture information and map it to the new building three-dimensional model, and then optimize the effect of the mapped three-dimensional model.

[0006] A further improvement of the technical solution of the present invention is that in step 1, the process of acquiring the building photo data is as follows: Conduct on-site inspections of the location and scope of buildings to identify the location and scope of buildings to be photographed, and develop a detailed shooting plan based on shooting requirements. Select a drone platform based on the shooting range and building height, and equip it with an oblique camera with high resolution, wide-angle lens and stable shooting function to ensure that the drone and camera have good performance and meet shooting requirements; Use flight control software to plan the flight path of the drone according to the shooting range, building height and drone performance, ensure that the route covers the entire shooting area, and set the flight altitude, speed, and camera angle parameters to ensure the shooting quality. According to the drone's range and shooting requirements, plan the number and length of routes to ensure that the shooting mission can be completed smoothly; Preset ground control points in the shooting area, mark them and record their location information for later image correction and 3D reconstruction. The ground control points should be evenly distributed and sufficient in number, and be checked before the flight. The UAV performs flight missions according to the planned route. The drone takes photos according to the planned flight path, and adjacent images have sufficient overlap, with a heading overlap greater than 75% and a lateral overlap greater than 30%, to ensure the quality of later image stitching and three-dimensional reconstruction. The oblique photography camera takes photos of the building from multiple angles (20-30 degrees) to obtain multiple photo data of the building.

[0007] A further improvement of the technical solution of the present invention is that in step 2, the process of constructing the inclined three-dimensional model of the building is: Download the building photos stored in the drone to the computer, screen the building photo data, and remove the building photos of unqualified quality, including photos that are overlapped, blurred, overexposed or underexposed, and contain obvious defects (such as lens stains, scratches, etc.); Performing preprocessing operations on the filtered building photo data, wherein the preprocessing operations include removing redundancy and noise and image registration and correction steps; Import the pre-processed photo data into the 3D modeling software Pix4D to ensure that the data format is compatible with the software, and perform preliminary 3D modeling processing in the software, including point cloud registration, point cloud filtering, and point cloud classification steps. Use the algorithm in Pix4D to match the feature points in multiple building photos to generate preliminary point cloud data, filter the point cloud data, remove noise points, retain valid buildings and terrain points, and then classify the point cloud data into different categories, such as buildings, ground, vegetation, etc., for subsequent processing; Oblique photography data is used to generate a three-dimensional surface model containing buildings and vegetation surface objects. Ground points are extracted from the point cloud data to generate a three-dimensional surface model containing only the ground to reflect the true shape of the ground. The outline and feature points of the building are extracted from the classified point cloud data, and then the algorithm in Pix4D is used to generate an inclined three-dimensional model of the building.

[0008] A further improvement of the technical solution of the present invention is that in step 3, the analysis process of the point cloud data category is: Use the 3D modeling software Pix4D to convert the model to point cloud, and export the 3D model of the building generated by oblique photogrammetry into point cloud format for subsequent analysis and processing; Use the point cloud processing software PCL to load the exported point cloud data and pre-process the point cloud data, including the steps of removing noise, filling holes, and smoothing, to improve the data quality and the accuracy of the analysis. The filtering algorithms (statistical filtering, voxel filtering, etc.) in the point cloud processing software PCL are used to remove noise points, and the hole areas in the point cloud are filled with the surface reconstruction algorithm. The smoothing algorithm of the moving minimum filter is applied to reduce the local sharp changes in the point cloud data and improve the continuity and smoothness of the data. The point cloud data are divided into different categories, including buildings, ground, and vegetation, using the Euclidean clustering segmentation classification algorithm in the point cloud processing software PCL.

[0009] A further improvement of the technical solution of the present invention is that in step 4, the evaluation process of the characteristic line matching degree is: Based on the category analysis results of point cloud data, point cloud data belonging to buildings are identified. In the point cloud data of buildings, the key feature points of the buildings are identified using a feature point extraction algorithm based on curvature. The key feature points include edge points and corner points of the buildings. Edge points are usually located on the outline of the building, while corner points are the points with the most dramatic changes on the outline of the building. Based on the extracted feature points, the feature lines of the building are generated by connecting adjacent feature points, and the linear interpolation technology is applied to make the feature lines smooth and continuous; Optimize the generated feature lines by comparing the length, direction, and position relationship between line segments, check the generated feature lines, remove redundant, repeated, or erroneous line segments, and adjust the shape and position of the feature lines according to the actual outline of the building so that the feature lines accurately reflect the outline of the building; Based on the CAD model, the building contour line is obtained as the reference feature line, the distance from each feature point to the nearest point on the reference feature line is calculated, and the deviation between the optimized feature line and the reference feature line is compared; Based on the calculation result of the deviation, the characteristic line matching index M is calculated to evaluate the matching degree of the characteristic line. If the characteristic line matching index is low, it means that the deviation is small, and the matching degree of the characteristic line is high; conversely, if the matching index is high, it means that the deviation is large, and the matching degree of the characteristic line is low; Combine the feature line matching index M to preset the feature line distinction threshold ,when When it is close to 1, it means that the matching degree of the feature line is high, which means that the distance deviation from the feature point to the reference feature line is small, the feature line is very close to or coincides with the reference feature line, and the feature line is directly accepted as a valid matching result for subsequent analysis and application. When it approaches 0, it indicates that the matching degree of the feature line is low, which means that the distance deviation from the feature point to the reference feature line is large, and there is a large deviation between the feature line and the reference feature line. For the feature lines with low matching degree, further optimization and adjustment are performed. The processing methods include: readjusting the position of the feature points or re-extracting the feature points, using optimization algorithms (such as genetic algorithms, simulated annealing, etc.) to fine-tune the feature lines to reduce the deviation, and eliminate the feature lines with too large deviation to avoid their influence on the overall matching results.

[0010] A further improvement of the technical solution of the present invention is that the calculation process of the characteristic line matching index is: Calculate the average distance from all feature points to the reference feature line to obtain the baseline value of the feature point ; Based on the baseline value of the feature point and the distance from the ith feature point to the nearest point of the reference feature line, the root mean square deviation of the distance from all feature points to the reference feature line is calculated, and then the root mean square deviation is standardized to associate it with the range of the maximum and minimum distances to obtain the standardized root mean square deviation term ; Combine the distance from the i-th feature point to the nearest point of the reference feature line, the baseline value of the feature point, the maximum and minimum distances of all feature points to the reference feature line, and calculate the exponential penalty term. , to penalize the feature point with the largest deviation from the reference feature line; The feature line matching index M is calculated by combining the standardized root mean square deviation and the exponential penalty term.

[0011] A further improvement of the technical solution of the present invention is that the calculation formula of the characteristic line matching index is: ; ; ; ; Where M is the feature line matching index, is the standardized root mean square deviation term, is the exponential penalty term, N is the total number of feature points, is the distance from the i-th feature point to the nearest point of the reference feature line, is the average value of the distance from all feature points to the reference feature line, which is used as the reference value of the feature points. is the maximum value of the distance from all feature points to the reference feature line, which is the maximum value obtained by traversing the distance of all feature points. , is the minimum value of the distance from all feature points to the reference feature line, which is the minimum value obtained by traversing the distance of all feature points. , the value range of M is between 0 and 1. When the distances from all feature points to the reference feature line are very close (i.e., the deviation is small), M is close to 1, indicating that the matching degree of the feature line is high. When there is a large deviation, M is close to 0, indicating that the matching degree of the feature line is low.

[0012] A further improvement of the technical solution of the present invention is that in step 5, the process of reconstructing the three-dimensional model is: Import the extracted building feature line data into the 3D modeling software, and create basic geometric bodies similar to the building shape in the 3D modeling software based on the feature line data, such as cubes, cylinders or NURBS curves, as the basis for building the 3D model, ensuring that the size and position of the basic geometric bodies match the feature line data; Simulate the preliminary shape of the building by moving, rotating and scaling the primitives to the correct position and setting their proportions and sizes to match the actual building; Use the editing tools of the 3D modeling software (push, pull, offset, trim, etc.) to adjust the shape and size of the basic geometric body, and gradually adjust the parts of the basic geometric body according to the feature line data to match the actual shape of the building; The adjusted basic geometric bodies are connected using the Boolean operation tools of the 3D modeling software to form a complete 3D model of the building.

[0013] A further improvement of the technical solution of the present invention is that in step 6, the process of optimizing the effect of the three-dimensional model after mapping is as follows: Based on the acquired data of multiple building photos, high-quality building texture images are extracted to ensure that the collected texture images are clear and complete and can accurately reflect the surface details of the building. The collected texture images are cropped, resized and pre-processed to meet the texture mapping requirements of the 3D model. Image editing software (Photoshop) is used to denoise and enhance the contrast of the texture images to improve the quality of the texture. In the 3D modeling software, the surface of the new 3D building model is divided into multiple areas, and corresponding texture coordinates are assigned to each area. The texture expansion operation is performed on the 3D building model to establish the coordinate mapping relationship between the 3D model and the texture space. Use the texture mapping algorithm (UV mapping) to map the preprocessed texture image onto the corresponding surface of the 3D model. During the mapping process, ensure that the texture image matches the surface shape and details of the 3D model to avoid problems such as stretching, distortion or dislocation; According to the mapping effect, use the texture editing tools in the 3D modeling software to adjust the brightness, contrast, and saturation of the texture image to coordinate with the overall effect of the model, and apply multiple texture mapping technology to further optimize the details of the 3D model, such as adding detail textures, adjusting the repetition pattern of textures, etc., to enhance the realism and detail performance of the model; Use a 3D rendering engine (Cycles for Blender) to render the mapped 3D model, check the texture mapping effect and quality, and further optimize the model based on the rendering results, adjusting the lighting, shadows, and reflection effects to improve the visual effect and realism of the model; Perform a comprehensive check on the mapped 3D model to ensure that the texture mapping effect is accurate, complete and as expected, check whether the texture has any stretching, distortion, or misalignment problems, and whether the texture fits tightly against the 3D model surface. Use the verification tool of the 3D modeling software to check the integrity of the model to ensure that there are no missing, overlapping, or erroneous geometries in the model, and verify whether the model's topology and texture mapping are correct.

[0014] Due to the adoption of the above technical solution, the present invention has the following technical advances compared with the prior art: The present invention provides a method for extracting feature lines of buildings based on an inclined three-dimensional model of an unmanned aerial vehicle. By photographing the building from different angles, multi-view image data of the building can be obtained. Combined with the three-dimensional modeling technology, an inclined three-dimensional model of the building can be constructed. On this basis, a feature line extraction algorithm is used to extract the feature lines of the building. Compared with traditional manual measurement or two-dimensional image extraction, not only the accuracy of feature extraction is greatly improved, but also the work efficiency is significantly improved, more details of the building can be captured, and feature line extraction is more accurate.

[0015] The present invention provides a method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle. The three-dimensional model of the building is further enriched and improved through the extracted building feature lines. The extracted feature lines can be used as an important reference for model construction to help refine the geometric shape and texture information of the model, so that the ultimately generated three-dimensional model is more realistic and can more truly reflect the actual form and structure of the building. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0017] Figure 1 is a flow chart of the method of the present invention; Figure 2 A flowchart of constructing a tilted three-dimensional model of a building according to the present invention; Figure 3 This is a flow chart for evaluating the matching degree of characteristic lines of the present invention. DETAILED DESCRIPTION

[0018] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are 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.

[0019] Embodiment 1, as Figure 1 , Figure 2 As shown, the present invention provides a method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle, comprising the following steps: Step 1. Use a drone equipped with an oblique photography system to photograph buildings from multiple angles, obtain multiple photo data of the buildings, conduct on-site inspections of the location and range of the buildings, clarify the location and range of the buildings to be photographed, and develop a detailed shooting plan based on shooting requirements. Select a drone platform based on the shooting range and building height, and equip it with an oblique photography camera with high resolution, wide-angle lens and stable shooting function to ensure that the drone and camera have good performance and meet the shooting requirements. Use flight control software to plan the flight path of the drone based on the shooting range, building height and drone performance to ensure that the route covers the entire shooting area, set the flight altitude, speed, and camera angle parameters to ensure shooting quality, and plan the number and length of routes based on the drone's range and shooting requirements to ensure that the shooting mission can be completed smoothly. Preset ground control points in the shooting area, mark the ground control points and record their location information for later image correction and 3D reconstruction. The ground control points should be evenly distributed and sufficient in number. Perform a pre-flight inspection. Before takeoff, conduct a comprehensive inspection of the drone to ensure that the GPS, compass, and airspeed tube are in good condition, check whether the camera lens is clean, the battery is sufficient, and the memory card is full, etc. Ensure that the take-off and landing areas are safe and obstacle-free, perform flight missions according to the planned route, and shoot according to the planned flight path. At the same time, adjacent images have sufficient overlap, with a heading overlap greater than 75% and a lateral overlap greater than 30% to ensure the quality of later image stitching and 3D reconstruction. The oblique camera shoots the building from multiple angles (20-30 degrees) to obtain multiple photo data of the building; Step 2: Preprocess the collected photo data and import them into the 3D modeling software to construct a tilted 3D model of the building. Download the building photos stored in the drone to the computer, screen the building photo data, and eliminate unqualified building photos, including overlapping, blurred, overexposed or underexposed photos, and photos containing obvious defects (such as lens stains, scratches, etc.). Preprocess the screened building photo data, where the preprocessing operation includes removing redundancy and noise and image registration and correction steps, specifically: removing overlapping, blurred or poor quality images from the collected photo data to reduce the burden of subsequent processing, removing background noise that will affect the accuracy of the 3D model, such as trees, pedestrians, etc., using ground control point information to register the collected photos to ensure that they are aligned in the same spatial reference system, and perform geometric correction on the registered photos to eliminate misalignment, distortion and Perspective effect to ensure the accuracy of subsequent processing, import the pre-processed photo data into the 3D modeling software Pix4D, ensure that the data format is compatible with the software, and perform preliminary 3D modeling processing in the software, including point cloud registration, point cloud filtering, and point cloud classification steps. Use the algorithm in Pix4D to match the feature points in multiple building photos to generate preliminary point cloud data, filter the point cloud data to remove noise points, retain valid buildings and terrain points, and then classify the point cloud data into different categories, such as buildings, ground, vegetation, etc., for subsequent processing, use oblique photography data to generate a 3D surface model containing buildings and vegetation surface objects, extract ground points from the point cloud data, generate a 3D surface model containing only the ground, reflect the true shape of the ground, and extract the outline and feature points of the building from the classified point cloud data, and then use the algorithm in Pix4D to generate an oblique 3D model of the building; Step 3: Convert the constructed inclined 3D model into point cloud data, analyze the point cloud data, clarify the category of point cloud data, use the 3D modeling software Pix4D to convert the model to point cloud, export the inclined 3D model of the building generated by the inclined photogrammetry into a point cloud format for subsequent analysis and processing, use the point cloud processing software PCL to load the exported point cloud data, and pre-process the point cloud data, including the steps of removing noise, filling holes, and smoothing to improve data quality and analysis accuracy. Among them, the filtering algorithm (statistical filtering, voxel filtering, etc.) in the point cloud processing software PCL is used to remove noise points, and the hole area in the point cloud is filled with a surface reconstruction algorithm. The smoothing algorithm of the moving minimum filter is applied to reduce the local sharp changes in the point cloud data and improve the continuity and smoothness of the data. The classification algorithm of the Euclidean clustering segmentation in the point cloud processing software PCL is used to classify the point cloud data into different categories, including buildings, ground, and vegetation; Step 4: Based on the point cloud data category analysis, the point cloud data is processed using a point cloud processing algorithm to extract the characteristic lines of the building, and the characteristic line matching index is calculated to evaluate the matching degree of the characteristic lines; Step 5: Reconstruct the 3D model using 3D modeling software based on the extracted building feature lines; Step 6: Use the texture mapping algorithm to collect texture information and map it to the new building three-dimensional model, and then optimize the effect of the mapped three-dimensional model.

[0020] Embodiment 2, as Figure 3 As shown, based on Example 1, the present invention provides a technical solution: Preferably, in step 4, the evaluation process of the characteristic line matching degree is: Based on the category analysis results of point cloud data, point cloud data belonging to buildings are identified. In the building point cloud data, the key feature points of the building are identified using a curvature-based feature point extraction algorithm. The key feature points include the edge points and corner points of the building. The edge points are usually located on the building outline, while the corner points are the points with the most dramatic changes on the building outline. On the basis of the extracted feature points, the feature lines of the building are generated by connecting adjacent feature points, and the linear interpolation technology is applied to make the feature lines smooth and continuous. The generated feature lines are optimized, and the generated feature lines are checked by comparing the length, direction, and position relationship between the line segments, and redundant, repeated or erroneous line segments are removed. , and adjust the shape and position of the feature line according to the actual outline of the building, so that the feature line accurately reflects the outline of the building. The building outline is obtained based on the CAD model as the reference feature line, and the distance from each feature point to the nearest point on the reference feature line is calculated. The deviation between the optimized feature line and the reference feature line is compared. Based on the calculation result of the deviation, the feature line matching index M is calculated to evaluate the matching degree of the feature line. If the feature line matching index is low, it means that the deviation is small, and the matching degree of the feature line is high; conversely, if the matching index is high, it means that the deviation is large, and the matching degree of the feature line is low. Combined with the feature line matching index M, the distinguishing threshold of the feature line is preset ,when When it is close to 1, it means that the matching degree of the feature line is high, which means that the distance deviation from the feature point to the reference feature line is small, the feature line is very close to or coincides with the reference feature line, and the feature line is directly accepted as a valid matching result for subsequent analysis and application. When it approaches 0, it indicates that the matching degree of the feature line is low, which means that the distance deviation between the feature point and the reference feature line is large, and there is a large deviation between the feature line and the reference feature line. For the feature line with low matching degree, further optimization and adjustment are performed. The processing methods include: re-adjusting the position of the feature point or re-extracting the feature point, using optimization algorithms (such as genetic algorithms, simulated annealing, etc.) to fine-tune the feature line to reduce the deviation, and eliminate the feature line with too large deviation to avoid its influence on the overall matching result; Furthermore, the calculation process of the characteristic line matching index is: Calculate the average distance from all feature points to the reference feature line to obtain the baseline value of the feature point , based on the baseline value of the feature point and the distance from the ith feature point to the nearest point of the reference feature line, the root mean square deviation of the distance from all feature points to the reference feature line is calculated, and then the root mean square deviation is standardized to associate it with the range of the maximum and minimum distances to obtain the standardized root mean square deviation term , combining the distance from the i-th feature point to the nearest point of the reference feature line, the baseline value of the feature point, the maximum and minimum distances from all feature points to the reference feature line, and calculating the exponential penalty term , to penalize the feature point with the largest deviation from the reference feature line, and combine the standardized root mean square deviation and exponential penalty term to calculate the feature line matching index M; Furthermore, the calculation formula of the characteristic line matching index is: ; ; ; ; Where M is the feature line matching index, is the standardized root mean square deviation term, is the exponential penalty term, N is the total number of feature points, is the distance from the i-th feature point to the nearest point of the reference feature line, is the average value of the distance from all feature points to the reference feature line, which is used as the reference value of the feature points. is the maximum value of the distance from all feature points to the reference feature line, which is the maximum value obtained by traversing the distance of all feature points. , is the minimum value of the distance from all feature points to the reference feature line, which is the minimum value obtained by traversing the distance of all feature points. , the value range of M is between 0 and 1. When the distances from all feature points to the reference feature line are very close (i.e., the deviation is small), M is close to 1, indicating that the matching degree of the feature line is high. When there is a large deviation, M is close to 0, indicating that the matching degree of the feature line is low. Close to 0 (i.e. all near ),and When M is close to 0, it means that the matching degree of the feature line is high. Larger or When it is large, M is close to 0, indicating that the matching degree of the feature lines is low; In step 5, the process of reconstructing the 3D model is as follows: Import the extracted building feature line data into the 3D modeling software, and create basic geometric bodies similar to the building shape in the 3D modeling software according to the feature line data, such as cubes, cylinders or NURBS curves, as the basis for building the 3D model, ensure that the size and position of the basic geometric bodies match the feature line data, place the basic geometric bodies in the correct position through move, rotate and scale operations, and set the scale and size of the basic geometric bodies to match the actual building to simulate the preliminary shape of the building, use the editing tools of the 3D modeling software (push, pull, offset, trim, etc.) to adjust the shape and size of the basic geometric bodies, according to the feature line data, gradually adjust the parts of the basic geometric bodies to match the actual shape of the building, use the Boolean operation tools of the 3D modeling software to connect the adjusted basic geometric bodies to form a complete 3D model of the building, after connecting the geometric bodies, adjust the entire 3D model of the building as a whole so that its shape and size are completely consistent with the actual shape of the building, use the verification tools of the 3D modeling software to check the integrity, accuracy and consistency of the model to ensure that there are no overlapping, missing or wrong geometric bodies in the model; In step 6, the process of optimizing the 3D model effect after mapping is as follows: Based on the data of multiple building photos, high-quality building texture images are extracted to ensure that the collected texture images are clear and complete and can accurately reflect the surface details of the building. The collected texture images are cropped, resized and pre-processed to meet the texture mapping requirements of the three-dimensional model. Image editing software (Photoshop) is used to denoise and enhance the contrast of the texture images to improve the quality of the texture. In the three-dimensional modeling software, the surface of the new building three-dimensional model is divided into multiple areas, and corresponding texture coordinates are assigned to each area. The texture expansion operation is performed on the building three-dimensional model to establish the coordinate mapping relationship between the three-dimensional model and the texture space. , use the texture mapping algorithm (UV mapping) to map the preprocessed texture image to the corresponding surface of the 3D model. During the mapping process, ensure that the texture image matches the surface shape and details of the 3D model to avoid problems such as stretching, distortion or dislocation. According to the mapping effect, use the texture editing tool in the 3D modeling software to adjust the brightness, contrast and saturation of the texture image to coordinate with the overall effect of the model, and apply multiple texture mapping technology to further optimize the details of the 3D model, such as adding detail textures, adjusting the repetition pattern of the texture, etc., to enhance the realism and detail performance of the model. Use the 3D rendering engine (Cycles Use Blender to render the mapped 3D model, check the mapping effect and quality of the texture, and further optimize the model based on the rendering results, adjust the lighting, shadow, and reflection effects to improve the visual effect and realism of the model, and conduct a comprehensive inspection of the mapped 3D model to ensure that the texture mapping effect is accurate, complete, and meets expectations, check whether the texture has stretching, distortion, or misalignment problems, and whether the texture fits tightly to the 3D model surface. Use the verification tool of the 3D modeling software to check the integrity of the model to ensure that there are no missing, overlapping, or incorrect geometries in the model, and verify whether the topology and texture mapping of the model are correct.

[0021] The above description is only a specific implementation manner of the present application, but the protection scope of the present application is not limited thereto. Any technician familiar with the 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.

Claims

1. A method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle, characterized in that: The following steps are involved: Step 1: Use a drone equipped with a tilt photography system to photograph the building from multiple angles to obtain multiple photo data of the building; Step 2: pre-process the collected photo data and import them into 3D modeling software to construct a tilted 3D model of the building; Step 3: Convert the constructed tilted 3D model into point cloud data, analyze the point cloud data, and identify the category of the point cloud data; Step 4: Based on the point cloud data category analysis, the point cloud data is processed using a point cloud processing algorithm to extract the characteristic lines of the building, and the characteristic line matching index is calculated to evaluate the matching degree of the characteristic lines; Step 5: Reconstruct the 3D model using 3D modeling software based on the extracted building feature lines; Step 6: Use the texture mapping algorithm to collect texture information and map it to the new building three-dimensional model, and then optimize the effect of the mapped three-dimensional model.

2. The method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle according to claim 1, characterized in that: In step 1, the process of acquiring building photo data is as follows: Investigate the location and range of the building on site, clarify the location and range of the building to be photographed, and formulate a shooting plan based on the shooting requirements. In particular, select a drone platform based on the shooting range and building height, and equip it with an oblique photography camera; Use flight control software to plan the flight path of the drone according to the shooting range, building height and drone performance, and set the flight altitude, speed, and camera angle parameters. Plan the number and length of routes according to the drone's range and shooting requirements; Preset ground control points in the shooting area, mark the ground control points and record their location information, and conduct a pre-flight check; The flight mission is carried out according to the planned route. The UAV takes pictures according to the planned flight path. The tilt camera takes pictures of the building from multiple angles to obtain multiple photo data of the building.

3. The method for extracting building feature lines based on the tilted three-dimensional model of an unmanned aerial vehicle according to claim 2, characterized in that: In the step 2, the process of constructing the inclined three-dimensional model of the building is as follows: Download the building photos stored in the drone to the computer, screen the building photo data, and eliminate the building photos with unqualified quality; Performing preprocessing operations on the filtered building photo data, wherein the preprocessing operations include removing redundancy and noise and image registration and correction steps; Import the pre-processed photo data into the 3D modeling software Pix4D, and perform preliminary 3D modeling processing in the software, including point cloud registration, point cloud filtering, and point cloud classification steps; Oblique photography data is used to generate a three-dimensional surface model containing buildings and vegetation surface objects. Ground points are extracted from the point cloud data to generate a three-dimensional surface model containing only the ground to reflect the true shape of the ground. The outline and feature points of the building are extracted from the classified point cloud data, and then the algorithm in Pix4D is used to generate an inclined three-dimensional model of the building.

4. The method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle according to claim 3, characterized in that: In step 3, the analysis process of point cloud data category is as follows: Use the 3D modeling software Pix4D to convert the model to point cloud, and export the tilted 3D model of the building generated by oblique photogrammetry into a point cloud format; Use the point cloud processing software PCL to load the exported point cloud data and pre-process the point cloud data, including the steps of removing noise, filling holes, and smoothing. The filtering algorithm in the point cloud processing software PCL is used to remove noise points, and the hole areas in the point cloud are filled using the surface reconstruction algorithm. The moving minimum filter smoothing algorithm is applied to reduce the local sharp changes in the point cloud data. The point cloud data are divided into different categories, including buildings, ground, and vegetation, using the Euclidean clustering segmentation classification algorithm in the point cloud processing software PCL.

5. The method for extracting building feature lines based on the tilted three-dimensional model of an unmanned aerial vehicle according to claim 4, characterized in that: In step 4, the evaluation process of the feature line matching degree is as follows: Based on the category analysis results of the point cloud data, point cloud data belonging to buildings are identified. In the point cloud data of buildings, a feature point extraction algorithm based on curvature is used to identify the key feature points of the buildings. The key feature points include the edge points and corner points of the buildings. Based on the extracted feature points, the feature lines of the building are generated by connecting adjacent feature points, and the linear interpolation technology is applied to make the feature lines smooth and continuous; Optimize the generated feature lines by comparing the length, direction, and position relationship between line segments, check the generated feature lines, remove redundant, repeated, or erroneous line segments, and adjust the shape and position of the feature lines according to the actual outline of the building so that the feature lines reflect the outline of the building; Based on the CAD model, the building contour line is obtained as the reference feature line, the distance from each feature point to the nearest point on the reference feature line is calculated, and the deviation between the optimized feature line and the reference feature line is compared; Based on the calculation results of the deviation, the characteristic line matching index M is calculated to evaluate the matching degree of the characteristic line; Combine the feature line matching index M to preset the feature line distinction threshold ,when When it is close to 1, it means that the matching degree of the feature line is high, the feature line is close to or coincides with the reference feature line, and the feature line is directly accepted as a valid matching result. When it approaches 0, it means that the matching degree of the feature line is low and there is a large deviation between the feature line and the reference feature line. For the feature lines with low matching degree, further optimization and adjustment are performed. The processing methods include: readjusting the position of feature points or re-extracting feature points, using optimization algorithms to fine-tune feature lines, and eliminating feature lines with excessive deviations.

6. The method for extracting building feature lines based on the tilted three-dimensional model of an unmanned aerial vehicle according to claim 5, characterized in that: The calculation process of the characteristic line matching index is: Calculate the average distance from all feature points to the reference feature line to obtain the baseline value of the feature point ; Based on the baseline value of the feature point and the distance from the ith feature point to the nearest point of the reference feature line, the root mean square deviation of the distance from all feature points to the reference feature line is calculated, and then the root mean square deviation is standardized to associate it with the range of the maximum and minimum distances to obtain the standardized root mean square deviation term ; Combine the distance from the i-th feature point to the nearest point of the reference feature line, the baseline value of the feature point, the maximum and minimum distances of all feature points to the reference feature line, and calculate the exponential penalty term. , to penalize the feature point with the largest deviation from the reference feature line; The feature line matching index M is calculated by combining the standardized root mean square deviation and the exponential penalty term.

7. The method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle according to claim 6, characterized in that: The calculation formula of the characteristic line matching index is: ； ; ； ； Where M is the feature line matching index, is the standardized root mean square deviation term, is the exponential penalty term, N is the total number of feature points, is the distance from the i-th feature point to the nearest point of the reference feature line, is the average value of the distance from all feature points to the reference feature line, which is used as the reference value of the feature points. is the maximum value of the distance from all feature points to the reference feature line, which is the maximum value obtained by traversing the distance of all feature points. , is the minimum value of the distance from all feature points to the reference feature line, which is the minimum value obtained by traversing the distance of all feature points. , the value range of M is between 0 and 1.

8. The method for extracting building feature lines based on the tilted three-dimensional model of an unmanned aerial vehicle according to claim 7, characterized in that: In step 5, the process of reconstructing the three-dimensional model is as follows: Import the extracted building feature line data into the 3D modeling software, and create basic geometric bodies similar to the building shape in the 3D modeling software based on the feature line data as the basis for building the 3D model; Simulate the preliminary shape of the building by moving, rotating and scaling the primitives to the correct position and setting their proportions and sizes to match the actual building; Use the editing tools of the 3D modeling software to adjust the shape and size of the basic geometric bodies, and gradually adjust the parts of the basic geometric bodies according to the feature line data to match the actual shape of the building; The adjusted basic geometric bodies are connected using the Boolean operation tools of the 3D modeling software to form a complete 3D model of the building.

9. The method for extracting building feature lines based on a tilted three-dimensional model of an unmanned aerial vehicle according to claim 8, characterized in that: In step 6, the process of optimizing the three-dimensional model effect after mapping is as follows: Based on the acquired data of multiple building photos, the building texture images are extracted, and the acquired texture images are cropped, resized and resolution preprocessed. The texture images are then denoised and contrast enhanced using image editing software. In the 3D modeling software, the surface of the new 3D building model is divided into multiple areas, and corresponding texture coordinates are assigned to each area. The texture expansion operation is performed on the 3D building model to establish the coordinate mapping relationship between the 3D model and the texture space. Using a texture mapping algorithm, the preprocessed texture image is mapped onto a corresponding surface of the three-dimensional model; According to the mapping effect, use the texture editing tools in the 3D modeling software to adjust the brightness, contrast and saturation of the texture image to make it coordinated with the overall effect of the model, and apply multiple texture mapping technology to further optimize the details of the 3D model; Use a 3D rendering engine to render the mapped 3D model, check the texture mapping effect and quality, and further optimize the model based on the rendering results; Perform a comprehensive inspection on the mapped 3D model to check whether the texture has any stretching, distortion, or misalignment issues, and whether the texture fits tightly against the 3D model surface. Use the verification tool of the 3D modeling software to check the integrity of the model to verify whether the model's topology and texture mapping are correct.

Images