Method for generating high-precision three-dimensional terrain model

Abstract

The invention discloses a method for generating a high-precision three-dimensional terrain model. The method comprises the following steps: S1, collecting geographic information data; collecting geographic information data of the target area by using the unmanned aerial vehicle; s2, data preprocessing, wherein the data preprocessing comprises precision correction, image splicing and three-dimensional reconstruction; s3, data storage and management; storing the processed data in a spatial database; step S4, generating a grid; discretizing the elevation data to generate a regular quadrilateral grid; step S5, carrying out elevation interpolation; elevation interpolation calculation is carried out on each grid node; step S6, constructing a three-dimensional terrain; a three-dimensional terrain model is generated according to the grid nodes and the elevation interpolation result; and S7, optimization and refinement are carried out. And optimizing the generated three-dimensional terrain model. Through the method provided by the invention, the efficiency and precision of three-dimensional terrain model generation are improved, so that the generated terrain model is finer, and actual terrain features can be truly reflected.

Description

Technical Field

[0001] The invention belongs to the technical field of three-dimensional terrain modeling, and in particular relates to a method for generating a high-precision three-dimensional terrain model. Background Art

[0002] With the development of geographic information systems (GIS) and remote sensing technology, three-dimensional terrain models are increasingly used in urban planning, disaster warning, environmental monitoring and other fields. Existing technologies mainly obtain terrain data through satellite images, aerial photogrammetry, etc., and then perform data processing and model construction. However, these methods often have problems such as insufficient accuracy, long time consumption, and high cost during data acquisition and processing, making it difficult to quickly generate high-precision three-dimensional terrain models.

[0003] In the prior art, for example, the technical solution described in the patent with publication number: CN112489213A: a three-dimensional terrain model generation method, device, electronic device and storage medium, whose main technical principle is to obtain contour lines constituting a target model, uniformly sample the contour lines to obtain three-dimensional sampling points, and build a grid based on the three-dimensional sampling points; distribute the three-dimensional sampling points in the grid of the grid, and use the Z coordinate value of the three-dimensional sampling point falling into the grid as the basic elevation value of the grid; based on the existing basic elevation value on the grid, insert the newly added elevation value in the blank grid of the grid using a preset interpolation algorithm; and generate a three-dimensional terrain model on the grid based on the basic elevation value and the newly added elevation value.

[0004] In the prior art, common methods for generating three-dimensional terrain models include LiDAR scanning, drone photogrammetry, and satellite image analysis. Although these methods can obtain a wide range of terrain data, they often require complex algorithms and a large amount of computing resources during data processing, and the accuracy of the model is limited by the resolution and accuracy of the data source. Summary of the invention

[0005] The purpose of the present invention is to provide a method for generating a high-precision three-dimensional terrain model to solve the problem that in the prior art proposed in the background technology, when generating a three-dimensional terrain model, complex algorithms and a large amount of computing resources are required in the data processing process, and the model accuracy is limited by the resolution and accuracy of the data source.

[0006] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0007] A method for generating a high-precision three-dimensional terrain model comprises the following steps:

[0008] Step S1, geographic information data collection: using a drone to collect geographic information data of a target area, including drone orthophoto data, terrain elevation data, and latitude and longitude data of the terrain range of the area to be generated;

[0009] Step S2, data preprocessing, data preprocessing includes precision correction, image stitching and three-dimensional reconstruction;

[0010] Step S3, data storage and management: storing the processed data in a spatial database;

[0011] Step S4, mesh generation: discretize the elevation data, generate a regular quadrilateral mesh, and optimize the mesh using the Delaunay triangulation method;

[0012] Step S5, elevation interpolation: performing elevation interpolation calculation on each grid node, using a cubic spline interpolation method to improve accuracy;

[0013] Step S6, three-dimensional terrain construction: generate a three-dimensional terrain model based on the grid nodes and the elevation interpolation results, use three-dimensional modeling software for visualization, map the grid nodes to a three-dimensional coordinate system, and generate vertices V (x, y, z), where z is the elevation value;

[0014] Step S7, optimization and refinement: optimizing the generated three-dimensional terrain model.

[0015] According to the above technical solution, in step S2, the precision correction is specifically as follows:

[0016] Calibration using ground control points (GCP) ensures high accuracy of measurement results; calibration using ground control points (GCP) includes the following steps:

[0017] In step S2, the precision correction is specifically to ensure the high precision of the measurement results; the correction using ground control points (GCP) includes the following steps:

[0018] Step S201, selecting a suitable GCP location, the location of the GCP should meet the following requirements: the GCP should be distributed throughout the entire operation area to ensure the geographic reference accuracy of the entire area; and a location that is easily identifiable in the image should be selected;

[0019] Step S202, laying out GCP marks, using markers to mark GCP positions; using GNSS equipment to measure the coordinates of each GCP to obtain high-precision three-dimensional coordinate data;

[0020] Step S203, data recording and management; record coordinates: record the three-dimensional coordinates (X, Y, Z) of each GCP, and ensure that these coordinates correspond to the on-site landmarks; compile a number for each GCP for subsequent matching and calibration in image processing;

[0021] Step S204, UAV flight and image acquisition; flight altitude and overlap rate: the flight altitude and image overlap rate of the UAV are set according to the size and requirements of the operation area; the UAV automatically takes ground images during flight to ensure that the GCP mark is clearly visible in the image;

[0022] Step S205, image processing and GCP correction; image import and preliminary processing: import the images collected by the drone into the image processing software for preliminary stitching and alignment; manually calibrate the GCP position, input the actual coordinates of each GCP, and use these GCPs to perform geometric correction and accuracy improvement of the image;

[0023] Step S206, orthophoto generation; Generate orthophoto: after the GCP correction is completed, generate an orthophoto with high geographic accuracy.

[0024] According to the above technical solution, in step S2, after precision correction, multiple overlapping images are stitched into a complete orthophoto map to complete image stitching; the following steps are included:

[0025] Geometric correction: Perform geometric correction on each image, including removing distortion and aligning to the same geographic coordinate system;

[0026] Radiometric correction: Perform radiometric correction based on the image metadata to ensure that the radiometric brightness of each image is consistent;

[0027] Image Registration: Feature Extraction and Matching: Use computer vision techniques to extract feature points in an image and match them to corresponding feature points in adjacent images;

[0028] Geometric transformation: Based on the matching relationship of feature points, the geometric transformation between images is calculated and the images are registered to the same coordinate system;

[0029] Image fusion: Using the geometric transformation relationship after image registration, all images are superimposed and fused;

[0030] Overlapping area processing: Use weighted average or select the best pixel value to process the pixel values ​​in the overlapping area;

[0031] Projection transformation: Orthographic projection is performed based on the geographic information and geometric transformation parameters of each image, so that the final stitched image has a unified geographic reference;

[0032] Correct edge effects: For black edges or transition issues that may exist on the edges of orthophotos, perform appropriate processing and correction.

[0033] According to the above technical solution, in step S2, the three-dimensional reconstruction specifically includes the following steps:

[0034] Image feature extraction and matching: Use computer vision technology to extract features from images;

[0035] Feature point matching: Match the corresponding feature points in different images to establish the corresponding relationship between the images;

[0036] Stereo vision and triangulation: Using the matched feature points, the three-dimensional position and shape of objects in the image are calculated through stereo vision and triangulation techniques;

[0037] Sparse point cloud generation: Based on the triangulation results of feature points, an initial sparse point cloud is generated to represent the shape of the object surface;

[0038] Image matching and depth estimation: Using higher-density feature matching and depth estimation technology, a denser point cloud is generated from a sparse point cloud to reflect a more detailed morphology of the surface;

[0039] Surface reconstruction: Based on dense point cloud data, surface reconstruction is performed to generate an accurate surface model;

[0040] DEM generation: Extract elevation information from the surface model and generate a digital elevation model (DEM).

[0041] According to the above technical solution, in step S4, the elevation data is discretized as follows:

[0042] Select a set of mesh nodes on the quadrilateral mesh, each with the coordinates (x i ,y j ), where i and j represent the horizontal and vertical positions of the grid nodes respectively;

[0043] The two-dimensional plane area D is divided into n×m small areas, each of which consists of a grid node (x i ,y j ) definition, where i = 1, 2, ..., n and j = 1, 2, ..., m; the elevation data Zi at each node j Indicates the ground height at this node.

[0044] According to the above technical solution, in step S4, generating a regular quadrilateral mesh is specifically as follows:

[0045] On the discretized mesh nodes, a quadrilateral mesh is generated by connecting adjacent mesh nodes; if the connection between adjacent mesh nodes meets the requirements of a quadrilateral, the vertex coordinates and connection relationship of each quadrilateral can be defined; assuming that each quadrilateral consists of four vertices (x i ,y j ), (x i+1 ,y j ), (x i+1 ,yj+1 ), (x i ,y j+1 ) is composed; the mathematical expression is expressed as:

[0046] Grid node coordinates: (x i ,y j ), where i = 1, 2, ..., n-1 and j = 1, 2, ..., m-1;

[0047] The coordinates of the vertices of the quadrilateral: (x i ,y j ), (x i+1 ,y j ), (x i+1 ,y j+1 ), (x i ,y j+1 ).

[0048] According to the above technical solution, in step S5, the elevation interpolation is specifically expressed by the following formula:

[0049] S(x)=a0+a1(x-x0)+a2(x-x0) 2 +a3(x-x0) 3

[0050] Among them, S(x) is the interpolation function, and a0, a1, a2, a3 are coefficients.

[0051] According to the above technical solution, in step S6, the three-dimensional terrain construction specifically includes the following steps:

[0052] Step S601, triangular mesh generation: converting the interpolated terrain surface into a triangular mesh model;

[0053] S602, surface reconstruction, creating triangular patches of the terrain surface by connecting adjacent mesh nodes to form a continuous terrain model;

[0054] S603, texture mapping: adding texture information to the terrain model based on the original image or other texture data to enhance the visual effect.

[0055] According to the above technical solution, in step S7, the generated three-dimensional terrain model is optimized including simplifying polygons, and the simplified polygons are performed by the following formula:

[0056]

[0057] Among them, Q is the error metric and di is the error of the i-th vertex.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The method of the present invention improves the efficiency and accuracy of generating three-dimensional terrain models, making the generated terrain models more refined and able to truly reflect the actual terrain features. The method of the present invention is suitable for large-scale terrain data processing and has good scalability and practicality. Compared with a single high-cost surveying and mapping technology, the present invention reduces the overall cost through multi-source data fusion, and supports multiple data sources and formats for output, and is suitable for multiple fields such as urban planning, disaster warning, and environmental monitoring. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A schematic diagram is generated for the three-dimensional terrain model of the present invention. DETAILED DESCRIPTION

[0061] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0062] Embodiment 1

[0063] like Figure 1 As shown, a method for generating a high-precision three-dimensional terrain model comprises the following steps:

[0064] Step S1, geographic information data collection: using a drone to collect geographic information data of a target area, including drone orthophoto data, terrain elevation data, and latitude and longitude data of the terrain range of the area to be generated;

[0065] Step S2, data preprocessing, data preprocessing includes precision correction, image stitching and three-dimensional reconstruction;

[0066] Step S3, data storage and management: storing the processed data in a spatial database;

[0067] Step S4, mesh generation: discretize the elevation data, generate a regular quadrilateral mesh, and optimize the mesh using the Delaunay triangulation method;

[0068] Step S5, elevation interpolation: performing elevation interpolation calculation on each grid node, using a cubic spline interpolation method to improve accuracy;

[0069] Step S6, three-dimensional terrain construction: generate a three-dimensional terrain model based on the grid nodes and the elevation interpolation results, use three-dimensional modeling software for visualization, map the grid nodes to a three-dimensional coordinate system, and generate vertices V (x, y, z), where z is the elevation value;

[0070] Step S7, optimization and refinement: optimizing the generated three-dimensional terrain model.

[0071] The method of the present invention improves the efficiency and accuracy of generating three-dimensional terrain models, making the generated terrain models more refined and able to truly reflect the actual terrain features. The method of the present invention is suitable for large-scale terrain data processing and has good scalability and practicality. Compared with a single high-cost surveying and mapping technology, the present invention reduces the overall cost through multi-source data fusion, and supports multiple data sources and formats for output, and is suitable for multiple fields such as urban planning, disaster warning, and environmental monitoring.

[0072] Embodiment 2

[0073] This embodiment is a further refinement of the first embodiment.

[0074] In step S2, the precision correction is specifically to ensure the high precision of the measurement results; the correction using ground control points (GCP) includes the following steps:

[0075] Step S201, select the appropriate GCP location. The location of the GCP should meet the following requirements: GCP should be distributed throughout the entire operation area to ensure the geographic reference accuracy of the entire area; select a location that is easy to identify in the image, such as a crossroads, a unique ground sign, etc. GCP should be set at different altitudes to reduce the impact of terrain undulations on image accuracy.

[0076] Step S202, the layout of the GCP mark, the layout of the GCP mark should meet the following requirements: use obvious markers (such as white crosses, rectangular marks) to mark the GCP position. The mark should be large enough to ensure that it is clearly visible in the drone image.

[0077] Precise positioning: Use high-precision GNSS (Global Navigation Satellite System) equipment to measure the coordinates of each GCP. Usually RTK (Real-time Kinematic Differential Kinematics) or PPK (Post-processing Kinematic Differential Kinematics) is used to obtain high-precision three-dimensional coordinate data.

[0078] Step S203, data recording and management; record coordinates: record the three-dimensional coordinates (X, Y, Z) of each GCP, and ensure that these coordinates correspond to the on-site landmarks. Compile numbers, each GCP should have a unique number to facilitate subsequent matching and calibration in image processing software.

[0079] Step S204, UAV flight and image acquisition; Flight altitude and overlap rate: Set the UAV flight altitude and image overlap rate (forward overlap and lateral overlap) according to the size and requirements of the operation area. Usually the forward overlap rate is 70%-80%, and the lateral overlap rate is 60%-70%. Image shooting: The UAV automatically shoots ground images during flight to ensure that the GCP mark is clearly visible in the image.

[0080] Step S205, image processing and GCP correction; image import and preliminary processing: import the images collected by the drone into image processing software (such as Pix4D, Agisoft Metashape, etc.) for preliminary stitching and alignment.

[0081] GCP matching and correction: Manually calibrate the GCP position in the software and enter the actual coordinates of each GCP. The software will use these GCPs to perform geometric correction and accuracy improvement on the image.

[0082] Accuracy assessment: Use software to assess the geographic accuracy of images to ensure that the error of orthophotos is within an acceptable range.

[0083] Step S206, orthophoto generation; Generate orthophoto: After the GCP correction is completed, the software will generate an orthophoto with high geographic accuracy.

[0084] Verification and Export: Verify the final orthophoto to ensure it meets the project requirements and export it for further analysis or application.

[0085] In step S2, after precision correction, multiple overlapping images are stitched into a complete orthophoto map to complete image stitching; the following steps are included:

[0086] Geometric correction: Perform geometric correction on each image, including removing distortion and aligning to the same geographic coordinate system;

[0087] Radiometric correction: Perform radiometric correction based on the image metadata to ensure that the radiometric brightness of each image is consistent;

[0088] Image Registration: Feature Extraction and Matching: Use computer vision techniques to extract feature points in an image and match them to corresponding feature points in adjacent images;

[0089] Geometric transformation: Based on the matching relationship of feature points, the geometric transformation between images is calculated and the images are registered to the same coordinate system;

[0090] Image fusion: Using the geometric transformation relationship after image registration, all images are superimposed and fused;

[0091] Overlapping area processing: Use weighted average or select the best pixel value to process the pixel values ​​in the overlapping area;

[0092] Projection transformation: Orthographic projection is performed based on the geographic information and geometric transformation parameters of each image, so that the final stitched image has a unified geographic reference;

[0093] Correct edge effects: For black edges or transition problems that may exist on the edges of orthophotos, appropriate processing and correction should be performed. When generating orthophotos, black edges or transition problems may sometimes appear on the edges of images. This is usually caused by image stitching, terrain occlusion, incomplete flight paths, or insufficient image overlap. In order to ensure the quality of the final orthophoto, the following methods can be used for processing and correction:

[0094] 1. Increase image overlap: In the UAV flight plan, increase the forward and lateral overlap of the image. Usually, the forward overlap rate is set at 70%-80%, and the lateral overlap rate is 60%-70%. This ensures that the image has enough data to cover the edge area when stitching, reducing the appearance of black edges.

[0095] 2. Image cropping: In the image processing software, you can use the automatic cropping function to remove the black edges of the image. The software will automatically identify and crop the invalid area based on the image stitching results. Manual cropping: In the generated orthophoto, if the automatic cropping is not ideal, you can manually select the cropping area to remove the black edges or discontinuous transition areas.

[0096] 3. Use elevation data for adjustment; integrate DEM data: If the generated orthophoto uses a digital elevation model (DEM) for orthorectification, it can ensure that the image edge is consistent with the terrain undulations, reducing image stitching problems caused by terrain differences.

[0097] Adjust elevation differences: If there are transition issues at the edges, it may be due to elevation differences between different images. This can be corrected by adjusting the image projection method or using more accurate elevation data.

[0098] 4. Post-processing of images; Mask and gradient processing: In image processing software (such as Photoshop or GIMP), you can use mask or gradient tools to manually process the edges of images, smooth the transition area, and reduce visual abruptness. Color and brightness adjustment: For the transition between different images, you can adjust the color and brightness of the edge area to make it more consistent with the adjacent images, thereby alleviating the transition problem.

[0099] 5. Reprocess or re-fly; re-fly and shoot: If the edge problem is too serious, you can consider re-flying and shooting the problem area to obtain more image data to cover and correct these areas. Re-stitching: Based on the existing data, readjust the stitching parameters or use a higher-precision stitching algorithm to re-process the image to reduce the black edges or transition problems on the edges.

[0100] 6. Combine multi-source data; Multi-source data fusion: Fuse drone images with other geographic data sources (such as satellite images, existing high-resolution orthophotos, etc.), and use other data for supplementary coverage in marginal areas.

[0101] In step S2, the three-dimensional reconstruction specifically includes the following steps:

[0102] Image feature extraction and matching: Use computer vision technology to extract features from images;

[0103] Feature point matching: Match the corresponding feature points in different images to establish the corresponding relationship between the images;

[0104] Stereo vision and triangulation: Using the matched feature points, the three-dimensional position and shape of objects in the image are calculated through stereo vision and triangulation techniques;

[0105] Sparse point cloud generation: Based on the triangulation results of feature points, an initial sparse point cloud is generated to represent the shape of the object surface;

[0106] Image matching and depth estimation: Using higher-density feature matching and depth estimation technology, a denser point cloud is generated from a sparse point cloud to reflect a more detailed morphology of the surface;

[0107] Surface reconstruction: Based on dense point cloud data, surface reconstruction is performed to generate an accurate surface model;

[0108] DEM generation: Extract elevation information from the surface model and generate a digital elevation model (DEM).

[0109] The steps and methods for extracting elevation information from the surface model and generating a digital elevation model are as follows:

[0110] 1. Data preparation; understand the difference between surface model and DEM:

[0111] Digital Surface Model (DSM): DSM represents the elevation of all objects on the Earth's surface, including buildings, trees, etc. It is generated through methods such as remote sensing, LiDAR, or photogrammetry, and reflects the highest points on the surface.

[0112] Digital Elevation Model (DEM): DEM represents the pure terrain surface, that is, the elevation information of the earth's bare ground, excluding any surface cover.

[0113] Obtaining DSM data: First, high-quality DSM data is required. DSM can be generated through drone imagery, satellite imagery, LiDAR data, or aerial photography.

[0114] Land feature classification: In order to extract the bare surface from the DSM, different features on the surface (such as buildings, vegetation, etc.) need to be identified and classified.

[0115] 2. Filtering to remove non-terrain elements; Filtering methods: Commonly used methods include low-pass filtering, median filtering, etc., which remove the high-frequency part (usually representing non-terrain elements) in the DSM through smoothing.

[0116] Land feature classification and segmentation: Use image analysis and classification techniques to identify non-topographic features (such as buildings and trees) in the DSM and remove them from the data or process them separately.

[0117] 3. Generate a bare earth model; Classification-based removal: Use classification algorithms to identify and remove buildings and vegetation from the surface model, retaining the elevation information of the bare surface. These classification methods can include machine learning algorithms, rule sets, or expert systems. Interpolation processing: After removing non-topographic features, blank areas may be generated in the DEM. These gaps are filled by interpolation algorithms to generate a continuous DEM.

[0118] 4. DEM generation; data format conversion: After processing the DSM, the resulting bare earth elevation data is usually converted to a grid format (such as GeoTIFF) to form a DEM. Smoothing: The generated DEM is smoothed to reduce noise and irregularities and ensure the continuity and accuracy of the DEM.

[0119] 5. Quality check and accuracy verification; Comparison with existing DEM: Compare the generated DEM with the existing high-precision DEM to verify its accuracy and consistency. Terrain profile analysis: Analyze the terrain profile of the DEM to check whether there are outliers or incorrect elevation data.

[0120] 6. Output; DEM export: Export the generated DEM data in a suitable format (such as GeoTIFF, ASCII Grid, etc.) for use in GIS or other analysis software.

[0121] Photogrammetry and point cloud processing software: such as Agisoft Metashape, Pix4D, and Global Mapper, can also directly process DSM data and generate DEM.

[0122] In step S4, the elevation data is discretized as follows:

[0123] Select a set of mesh nodes on the quadrilateral mesh, each with the coordinates (x i ,y j ), where i and j represent the horizontal and vertical positions of the grid nodes respectively;

[0124] The two-dimensional plane area D is divided into n×m small areas, each of which consists of a grid node (x i ,y j ) definition, where i = 1, 2, ..., n and j = 1, 2, ..., m; the elevation data Z at each node ij Indicates the ground height at this node.

[0125] In step S4, generating a regular quadrilateral mesh is specifically as follows:

[0126] On the discretized mesh nodes, a quadrilateral mesh is generated by connecting adjacent mesh nodes; if the connection between adjacent mesh nodes meets the requirements of a quadrilateral, the vertex coordinates and connection relationship of each quadrilateral can be defined; assuming that each quadrilateral consists of four vertices (x i ,y j ), (x i+1 ,y j ), (x i+1 ,y j+1 ), (x i ,y j+1 ) is composed; the mathematical expression is expressed as:

[0127] Grid node coordinates: (x i ,y j ), where i = 1, 2, ..., n-1 and j = 1, 2, ..., m-1;

[0128] The coordinates of the vertices of the quadrilateral: (x i ,y j ), (x i+1 ,y j ), (x i+1 ,y j+1 ), (x i ,y j+1 ).

[0129] In step S5, the elevation interpolation is specifically expressed by the following formula:

[0130] S(x)=a0+a1(x-x0)+a2(x-x0) 2 +a3(x-x0) 3

[0131] Among them, S(x) is the interpolation function, and a0, a1, a2, a3 are coefficients.

[0132] In step S6, the three-dimensional terrain construction specifically includes the following steps:

[0133] Step S601, triangular mesh generation: converting the interpolated terrain surface into a triangular mesh model;

[0134] S602, surface reconstruction, creating triangular patches of the terrain surface by connecting adjacent mesh nodes to form a continuous terrain model;

[0135] S603, texture mapping: adding texture information to the terrain model based on the original image or other texture data to enhance the visual effect.

[0136] In step S7, the generated three-dimensional terrain model is optimized including polygon simplification, and the polygon simplification is performed by the following formula:

[0137]

[0138] Among them, Q is the error metric and di is the error of the i-th vertex.

[0139] Embodiment 3

[0140] This embodiment is a further refinement of the second embodiment.

[0141] Step S2, during data preprocessing, the collected elevation data is denoised and a Gaussian filter is used to smooth the data:

[0142]

[0143] Among them, G(x,y) is the Gaussian function and σ is the standard deviation.

[0144] Interpolate missing data using bilinear interpolation:

[0145] f(x, y) = f(x1, y1)·(1-t x )·(1-t y )+f(x2,y1)·t x ·(1-t y )+f(x1,y2)·(1-t x )·t y +f(x2,y2)·t x ·t y

[0146] where t x and t y is the interpolation factor.

[0147] Step S6: When constructing the three-dimensional terrain, the following software tools can be used:

[0148] GIS software: such as ArcGIS, QGIS, etc., provides advanced interpolation and terrain modeling capabilities.

[0149] 3D modeling software: such as Blender, 3ds Max, etc., used to create and render complex 3D models.

[0150] Computer-aided design (CAD) software, such as AutoCAD, can also be used to create three-dimensional terrain models.

[0151] It should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or device.

[0152] Finally, it should be noted that the above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art can still modify the technical solutions described in the aforementioned embodiments or replace some of the technical features therein by equivalents. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A method for generating a high-precision three-dimensional terrain model, characterized in that: The following steps are involved: Step S1, geographic information data collection: using a drone to collect geographic information data of a target area, including drone orthophoto data, terrain elevation data, and latitude and longitude data of the terrain range of the area to be generated; Step S2, data preprocessing, data preprocessing includes precision correction, image stitching and three-dimensional reconstruction; Step S3, data storage and management: storing the processed data in a spatial database; Step S4, mesh generation: discretize the elevation data, generate a regular quadrilateral mesh, and optimize the mesh using the Delaunay triangulation method; Step S5, elevation interpolation: performing elevation interpolation calculation on each grid node, using a cubic spline interpolation method to improve accuracy; Step S6, three-dimensional terrain construction: generate a three-dimensional terrain model based on the grid nodes and the elevation interpolation results, use three-dimensional modeling software for visualization, map the grid nodes to a three-dimensional coordinate system, and generate vertices V (x, y, z), where z is the elevation value; Step S7, optimization and refinement: optimizing the generated three-dimensional terrain model.

2. A method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S2, the precision calibration is specifically as follows: Calibration using ground control points (GCP) ensures high accuracy of measurement results; calibration using ground control points (GCP) includes the following steps: In step S2, the precision correction is specifically to ensure the high precision of the measurement results; the correction using ground control points (GCP) includes the following steps: Step S201, selecting a suitable GCP location, the location of the GCP should meet the following requirements: the GCP should be distributed throughout the entire operation area to ensure the geographic reference accuracy of the entire area; and a location that is easily identifiable in the image should be selected; Step S202, laying out GCP marks, using markers to mark GCP positions; using GNSS equipment to measure the coordinates of each GCP to obtain high-precision three-dimensional coordinate data; Step S203, data recording and management; record coordinates: record the three-dimensional coordinates (X, Y, Z) of each GCP, and ensure that these coordinates correspond to the on-site landmarks; compile a number for each GCP for subsequent matching and calibration in image processing; Step S204, UAV flight and image acquisition; flight altitude and overlap rate: the flight altitude and image overlap rate of the UAV are set according to the size and requirements of the operation area; the UAV automatically takes ground images during flight to ensure that the GCP mark is clearly visible in the image; Step S205, image processing and GCP correction; image import and preliminary processing: import the images collected by the drone into the image processing software for preliminary stitching and alignment; manually calibrate the GCP position, input the actual coordinates of each GCP, and use these GCPs to perform geometric correction and accuracy improvement of the image; Step S206, orthophoto generation; Generate orthophoto: after the GCP correction is completed, generate an orthophoto with high geographic accuracy.

3. The method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S2, after precision correction, multiple overlapping images are stitched into a complete orthophoto map to complete image stitching; the following steps are included: Geometric correction: Perform geometric correction on each image, including removing distortion and aligning to the same geographic coordinate system; Radiometric correction: Perform radiometric correction based on the image metadata to ensure that the radiometric brightness of each image is consistent; Image Registration: Feature Extraction and Matching: Use computer vision techniques to extract feature points in an image and match them to corresponding feature points in adjacent images; Geometric transformation: Based on the matching relationship of feature points, the geometric transformation between images is calculated and the images are registered to the same coordinate system; Image fusion: Using the geometric transformation relationship after image registration, all images are superimposed and fused; Overlapping area processing: Use weighted average or select the best pixel value to process the pixel values ​​in the overlapping area; Projection transformation: Orthographic projection is performed based on the geographic information and geometric transformation parameters of each image, so that the final stitched image has a unified geographic reference; Correct edge effects: For black edges or transition issues that may exist on the edges of orthophotos, perform appropriate processing and correction.

4. The method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S2, the three-dimensional reconstruction specifically includes the following steps: Image feature extraction and matching: Use computer vision technology to extract features from images; Feature point matching: Match the corresponding feature points in different images to establish the corresponding relationship between the images; Stereo vision and triangulation: Using the matched feature points, the three-dimensional position and shape of objects in the image are calculated through stereo vision and triangulation techniques; Sparse point cloud generation: Based on the triangulation results of feature points, an initial sparse point cloud is generated to represent the shape of the object surface; Image matching and depth estimation: Using higher-density feature matching and depth estimation technology, a denser point cloud is generated from a sparse point cloud to reflect a more detailed morphology of the surface; Surface reconstruction: Based on dense point cloud data, surface reconstruction is performed to generate an accurate surface model; DEM generation: Extract elevation information from the surface model and generate a digital elevation model (DEM).

5. The method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S4, the elevation data is discretized as follows: Select a set of mesh nodes on the quadrilateral mesh, each with the coordinates (x i ,y j ), where i and j represent the horizontal and vertical positions of the grid nodes respectively; The two-dimensional plane area D is divided into n×m small areas, each of which consists of a grid node (x i ,y j ) definition, where i = 1, 2, ..., n and j = 1, 2, ..., m; the elevation data Z at each node ij Indicates the ground height at this node.

6. The method for generating a high-precision three-dimensional terrain model according to claim 5, characterized in that: In step S4, generating a regular quadrilateral mesh is specifically as follows: On the discretized mesh nodes, a quadrilateral mesh is generated by connecting adjacent mesh nodes; if the connection between adjacent mesh nodes meets the requirements of a quadrilateral, the vertex coordinates and connection relationship of each quadrilateral can be defined; assuming that each quadrilateral consists of four vertices (x i ,y j ), (x i+1 ,y j ), (x i+1 ,y j+1 ), (x i ,y j+1 ) is composed; the mathematical expression is expressed as: Grid node coordinates: (x i ,y j ), where i = 1, 2, ..., n-1 and j = 1, 2, ..., m-1; The coordinates of the vertices of the quadrilateral: (x i ,y j ), (x i+1 ,y j ), (x i+1 ,y j+1 ), (x i ,y j+1 ).

7. The method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S5, the elevation interpolation is specifically expressed by the following formula: <h2 style=";text-align:left;direction:ltr">S(x) = a0 + a1(x-x0) + a2(x-x0)<h2 style=";text-align:left;direction:ltr"> 2 <h2 style=";text-align:left;direction:ltr"> +a3(x-x0)<h2 style=";text-align:left;direction:ltr"> 3 Among them, S(x) is the interpolation function, and a0, a1, a2, a3 are coefficients.

8. The method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S6, the three-dimensional terrain construction specifically includes the following steps: Step S601, triangular mesh generation: converting the interpolated terrain surface into a triangular mesh model; S602, surface reconstruction, creating triangular patches of the terrain surface by connecting adjacent mesh nodes to form a continuous terrain model; S603, texture mapping: adding texture information to the terrain model based on the original image or other texture data to enhance the visual effect.

9. The method for generating a high-precision three-dimensional terrain model according to claim 1, characterized in that: In step S7, the generated three-dimensional terrain model is optimized including polygon simplification, and the polygon simplification is performed by the following formula: Among them, Q is the error metric and di is the error of the i-th vertex.

Images