Unmanned aerial vehicle naked eye 3D two-dimensional and three-dimensional digital mapping method

Abstract

The invention provides an unmanned aerial vehicle naked eye 3D two-dimensional and three-dimensional digital mapping method, and provides three-dimensional contour mapping based on point cloud, multi-layer building digital mapping based on a live-action model and ground object and landform digital mapping based on a digital orthoimage, so that rapid mapping of landforms, ground objects and landforms in a construction area is realized, the mapping precision is better than or equal to 5 cm, and the mapping efficiency is greatly improved. The large-scale digital mapping requirement is met, the construction deployment efficiency of an engineering project is improved, and the method is suitable for three-dimensional topographic mapping of all construction areas such as railways, highways and municipal administration. Belongs to the surveying field.

Description

Technical Field

[0001] The present invention relates to a naked-eye 3D two- and three-dimensional digital mapping method for an unmanned aerial vehicle (UAV), which is particularly suitable for full digital mapping of complex terrain, objects, and landforms, and belongs to the field of surveying and mapping. Background Art

[0002] Topographic maps play a crucial role in engineering construction. During the design phase, they are used to determine the specific locations of various projects and corresponding buildings. They help designers find suitable locations on the map, stake out various facilities, measure distances and elevations, and determine the positioning and orientation of the project, including slopes. During the construction phase, topographic maps serve as foundational drawings to ensure that construction activities proceed according to design and avoid accidents or delays caused by misinterpretations of the terrain. Due to the time lag between the design and construction phases and the influence of the natural environment, once construction units begin work, the topography may differ from that of the design phase. To ensure construction safety, the construction site must be resurveyed and mapped.

[0003] Traditional topographic mapping relies primarily on RTK surveying, which presents challenges such as heavy fieldwork, low mapping efficiency, and significant climatic influences. With the rapid adoption of drone technology in recent years, the industry's main approaches include: First, manually measuring ground points on a real-world model, reconstructing a triangulated network from these points, and then establishing three-dimensional contour lines. This approach is labor-intensive, places high demands on surveyors, and is difficult to scale. Second, loading digital orthophotos (DOMs) into CAD allows for mapping of roads, rivers, farmland, and built-up areas. However, this approach cannot accurately measure the area of multi-story buildings, making it unsuitable for direct land acquisition and demolition mapping and management. Furthermore, direct CAD mapping is inefficient and requires multiple redrawings of the common edges of farmland, which can easily lead to mapping errors. Summary of the Invention

[0004] The present invention provides a method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle (UAV) to improve the efficiency of topographic mapping and ensure that the mapping accuracy can meet construction requirements.

[0005] To solve the above problems, a UAV naked-eye 3D two- and three-dimensional digital mapping method is proposed, which specifically includes: 1) UAV data collection and processing Deploy image control points around the mapping, measure the coordinates of each image control point using a positioning system, collect drone images, and perform two aerial triangulation solutions using aerial survey software to generate LAS format point cloud data, OSGB format 3D real scene models, and TIFF format DOM images. 2) 3D contour mapping based on point cloud data The acquired point cloud data is sequentially classified using mathematical morphology filtering, slope filtering, and cloth simulation filtering. The ground point cloud data is converted into TIN triangulation data using the Delaunay triangulation method. 3D contour line data in DXF format is generated based on the TIN triangulation data. 3) Naked-eye 3D mapping of landforms and features based on digital orthophotos Import DOM images into ArcGIS to create vectorized layers for woodlands, roads, farmlands, rivers, buildings, and greenhouses. Use topological analysis to establish relationships between vector graphics and raster graphics, creating a raster-to-vector - raster layer - vectorization - raster-to-vectorization workflow. Use Python scripts to calculate land attributes, area, and perimeter, and generate a table of project quantities. 4) Naked-eye 3D mapping of multi-story buildings based on real-scene models Import the 3D real-world model in OSGB format into CASS 3D software. By linking the 3D model with the 2D image, draw closed polygons along the building's exterior walls while viewing the results in the 3D model. Convert the survey results into DXF format and import them into ArcGIS. Through data connection design and field creation, statistical analysis of the multi-story building area is performed. 5) Topographic map merging and accuracy verification The vectorized files generated in the above steps are merged in GIS and directly output with the DOM image; the accuracy of the naked-eye 3D digital mapping is evaluated by the accuracy of the drone image control points.

[0006] In the aforementioned method, drone images are collected through a combination of fully automatic and manual flight modes; Professional aerial survey software is used for drone data processing. First, the original drone images are added, an aerial triangulation solution algorithm is established, and the photos are converted into aerial triangulation point cloud data. Secondly, the coordinates of the image control points are added, and the puncture point method is used to assign the image control point coordinates to the aerial triangulation point cloud data. Through the aerial triangulation optimization method and the image control point adjustment report, the errors of the aerial survey results are queried. For image control points with a position error greater than 5cm, the points need to be re-punctured and the drone data is converted to the construction coordinate system to ensure the accuracy of drone digital mapping. Finally, through the second aerial triangulation solution algorithm, high-precision dense point cloud data, 3D real-life model, and DOM image are respectively established.

[0007] In the above method, drone images are collected through a combination of fully automatic and manual flight modes; professional aerial survey software is used for drone data processing. First, the original drone images are added, an aerial triangulation solution algorithm is established, and the photos are converted into aerial three-dimensional point cloud data; secondly, the coordinates of the image control points are added, and the puncture point method is used to assign the image control point coordinates to the aerial three-dimensional point cloud data. Through the aerial three-dimensional optimization method and the image control point adjustment report, the error of the aerial survey results is queried. For image control points with a position error greater than 5cm, it is necessary to re-puncture the points and convert the drone data into the construction coordinate system to ensure the accuracy of drone digital mapping; finally, through the second aerial triangulation solution algorithm, high-precision dense point cloud data, three-dimensional real scene model, and DOM image are respectively established.

[0008] In the above method, the point cloud is filtered and classified. First, the mathematical morphology filtering classification method is used to perform overall filtering and classification on the mapping point cloud data to obtain a coarse classification point cloud, which quickly separates large areas of ground and non-ground points; secondly, the slope filtering classification method is used to re-select areas with larger voids in the point cloud data, usually mountains with larger slopes or steep areas, and re-optimize the classification results of the ground point cloud by optimizing the slope classification parameters; finally, the cloth simulation filtering classification method is used to solve the problem of misclassification or classification anomalies of the concrete surface.

[0009] In the above method, a TIN triangulation network is established, and the classified point cloud data is saved as a new point cloud data file. Through the Delaunay triangulation method, the maximum side length of the triangulation network does not exceed 50m, and the display resolution is X=1m, Y=1m. The ground point cloud data is converted into an irregular triangulation network, namely TIN triangulation network data, and the triangulation network file is saved separately; three-dimensional contour lines are established. According to construction requirements, the topographic map scale is 1 / 500, the corresponding main contour interval is 2.5m, the contour interval is 0.5m, the contour line smoothing type is selected as B-spline area, and the smoothing degree is 70% to 85%. Three-dimensional contour line data is established based on the TIN triangulation network, and the data format file is dxf.

[0010] In the above method, vectorization parameters are established, DOM images are imported into ArcGIS, coordinate bases are set, and shapefile tools are used to create new vectorization parameters. Vectorization layer parameters for forest land, roads, farmland, rivers, buildings, and greenhouses are established respectively. Different colors are used to distinguish different types of land features through layer style settings. Raster layer vectorization: Use ArcGIS's topology analysis function to perform topological analysis on raster data. Based on the topological analysis results, establish the relationship between vector graphics and raster graphics, forming a raster vector - raster layer - vectorization - raster - vectorization operation process. For areas with non-overlapping boundaries, use the area analysis tool in the corresponding layer to directly perform vectorization processing; Extract engineering quantities, select vectorized layer data, use data design to establish land attributes, units, quantity, area and perimeter fields, perform statistical analysis on the vectorized layer, and obtain the engineering quantity table.

[0011] In the above method, the digital mapping of multi-story buildings is carried out by importing a three-dimensional real-life model in osgb format through CASS 3D software, setting the layer of the building, linking the three-dimensional model browsing with the two-dimensional image, and drawing closed polygons along the outer walls of the building through the multi-building area drawing method, while viewing the results in the three-dimensional model; for engineering quantity calculation, the file after mapping the multi-story building is converted into dxf format, imported into ArcGIS, and the area statistical analysis of the multi-story building is realized through data connection design and field creation.

[0012] In the above method, vectorized files in the formats of shp, dxf, and dwg are directly merged through GIS. The merged files are directly used for road line selection and earthwork calculation, directly added to the Aowei software for viewing, and directly merged and output with high-resolution DOM images for use by construction personnel.

[0013] Compared with the existing technology, the present invention comprehensively adopts methods such as naked-eye 3D mapping of three-dimensional contour lines based on point cloud data, naked-eye 3D mapping of landforms and topography based on digital orthophotos, and naked-eye 3D mapping of multi-story buildings based on real-scene models to achieve rapid vectorized mapping of terrain, landforms, and topography within the construction area. The mapping accuracy is better than within 5 cm, meeting the requirements of large-scale digital mapping, improving the efficiency of construction deployment of engineering projects, improving the accuracy and precision of digital mapping, reducing the field workload of surveyors, and achieving high efficiency in both indoor and outdoor work. Efficient modeling is achieved in areas with complex terrain, with strong digital interactivity, and improving the level of digital management of engineering projects. It constitutes a technical closed loop of "high-precision data acquisition-intelligent processing and analysis-engineering application adaptation", which not only breaks through the multiple bottlenecks of traditional drone mapping in accuracy, efficiency, and scene adaptability, but also constructs a digital mapping technology standard suitable for modern engineering construction. It has significant industry demonstration effect and industrial promotion value, and is suitable for three-dimensional terrain mapping in all construction areas such as railways, highways, and municipal projects. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] Figure 1 It is a flow chart of the present invention. DETAILED DESCRIPTION

[0015] In order to make the objectives, technical solutions and advantages of the present invention more clear, the present invention will be further described in detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not intended to limit the present invention.

[0016] Example

[0017] Refer to the attached Figure 1 This embodiment improves the mapping method. Specifically, a method for naked-eye 3D two- and three-dimensional digital mapping using a drone is proposed, including: (1) UAV data collection and processing; In order to improve the accuracy of UAV mapping, after the design unit handed over the control point coordinates, based on previous UAV mapping experience, a new method was proposed for every 1km 2 The density standard of deploying three image control points improves the control accuracy by 2-3 times compared with traditional UAV mapping (usually deploying one point every 2-4km²), ensuring the coordinate conversion error is ≤5cm from the source. Image control points are deployed around the mapping, and the coordinates of each control point are measured using the Beidou positioning system. The coordinate reference adopts the national CGCS2000 coordinate system, and the elevation reference is the 1985 Yellow Sea elevation system. A unified framework of millimeter-level geographic reference is constructed, which solves the positioning deviation problem of traditional GNSS signals in areas blocked by obstructions, realizes seamless connection between the construction coordinate system and the national reference coordinate system, and avoids the cumulative error of subsequent coordinate conversion.

[0018] For UAV data collection, a combination of fully automatic and manual flight is adopted. Based on the experience of digital mapping, the parameters can be set as follows: the flight altitude should be controlled within 80m~100m, the overlap rate of UAV aerial photos should be 65%~85%, and the UAV lens should be 8°~20°. After setting the flight parameters, high-resolution UAV aerial photo data for mapping is established.

[0019] Professional aerial survey software is used for drone data processing, and a "two-step aerial triangulation solution" process is proposed for the first time. First, the original drone image is added, an aerial triangulation solution algorithm is established, and the photos are converted into aerial triangulation point cloud data. Secondly, the coordinates of the image control points are added, and the puncture point method is used to assign the control point coordinates to the aerial triangulation point cloud data. Through the aerial triangulation optimization method and the image control point adjustment report, the error of the aerial survey results is queried. For image control points with a position error greater than 5cm, the points need to be re-punctured. A high-density point cloud and model are generated through a secondary solution to convert the drone data into the construction coordinate system, ensuring the accuracy of drone digital mapping and facilitating use by construction personnel. Finally, through the second aerial triangulation solution algorithm, high-precision dense point cloud data (las format), 3D real-life model (osgb format), and DOM image (tiff format) are respectively established.

[0020] Compared with a single aerial triangulation solution, the plane / elevation accuracy estimate is improved by 40%, realizing the transformation from "post-inspection" to "process control", ensuring that the data results directly meet the 1 / 500 large-scale mapping requirements.

[0021] (2) Three-dimensional contour naked-eye 3D mapping based on point cloud data Point cloud filtering and classification processing. Because point cloud data contains a large number of non-ground points, such as buildings, vegetation, power facilities, and vehicles, these points affect the accuracy of terrain mapping, necessitating filtering and classification of the point cloud. First, a mathematical morphology filtering and classification method is used to perform overall filtering and classification on the mapped point cloud data, generating a coarsely classified point cloud. This method quickly separates large areas of ground and non-ground points, significantly improving processing efficiency compared to single algorithms. Second, a slope filtering and classification method is used to relocate areas with large voids in the point cloud data, typically steep mountainous areas or steep slopes. By optimizing the slope classification parameters, the classification results of the ground point cloud are re-optimized, eliminating issues such as inaccurate point cloud filtering classification results in mountainous areas, the presence of numerous voids in some mountainous areas, and the inability to accurately represent the original ground features. For mountainous / steep slope areas with slopes greater than 25°, the slope classification threshold is dynamically adjusted to eliminate most void artifacts. Finally, a fabric simulation filtering and classification method is used to eliminate issues such as indistinguishable concrete pavements from concrete roof slabs or classification anomalies within building complexes. Through the above three-level progressive point cloud filtering and classification algorithm, accurate classification of ground point cloud data is achieved, breaking through the adaptability bottleneck of traditional single filtering algorithms in complex scenarios. The accuracy of ground point cloud classification is significantly improved, laying the foundation for high-precision contour line generation.

[0022] Create a TIN triangulated network. Save the classified point cloud data as a new point cloud data file. Using the Delaunay triangulation method, with a maximum triangulated network side length of no more than 50m and a display resolution of 1m x and 1m y, convert the ground point cloud data into an irregular triangulated network (TIN) and save it as a TIN file.

[0023] Establish 3D contour lines. The method for constructing contour lines using a triangulated network (TIN) is based on construction requirements. Typically, projects use a topographic map with the highest precision of 1 / 500, corresponding to a main contour interval of 2.5m and a contour interval of 0.5m. The contour smoothing type is B-spline area, with a smoothing level of 70% to 85%. 3D contour data is established based on the TIN triangulation network, and the data file format is DXF.

[0024] (3) Naked-eye 3D mapping of landforms and features based on digital orthophotos Establish vectorization parameters. Import the DOM image into ArcGIS, set the coordinate base, and use the shapefile tool to create new vectorization parameters. Establish vectorization layer parameters for woodlands, roads, farmland, rivers, buildings, greenhouses, and other areas. Use layer style settings to distinguish different types of land features with different colors.

[0025] Vectorization of raster layers. Use the topological analysis function of ArcGIS to perform topological analysis on raster data, and establish the association relationship between vector graphics and raster graphics based on the topological analysis results, forming an operational process of raster vector-raster layer-vectorization-raster-vectorization. For areas with no overlapping boundaries, such as greenhouses, buildings, rivers, woodlands and other areas, use the area analysis tool in the corresponding layers to directly perform vectorization processing. For the vectorization of roads, there is an intersection relationship between roads, and it is necessary to realize the vectorization of roads through the topological intersection relationship. For the vectorization of farmland, there are a large number of common edges. If CAD is used for vectorization, all boundaries of each plot of land need to be drawn. In order to improve the efficiency of vectorization, topological adjacent analysis is used to directly extract the common edges between plots and realize the rapid vectorization of land.

[0026] Extract construction quantities. Select the vectorized layer data and use data design to create land attribute, unit, quantity, area, and perimeter fields. All data types are single precision. Use Python as the calculation method. Perform statistical analysis on the vectorized layer to obtain the construction quantity table.

[0027] (4) Naked-eye 3D mapping of multi-story buildings based on real-scene models For digital mapping of multi-story buildings, a "3D model positioning + 2D image verification" linkage operation mode is proposed: through CASS 3D software, a 3D real-life model in OSGB format is imported, the building layer is set, and the 3D model is browsed and linked with the 2D image. Using the multi-building face region drawing method, closed polygons are drawn along the building's exterior wall, and the results are viewed on the 3D model at the same time. For single-story buildings, it is named F1, two-story buildings are named F2, three-story buildings are named F3, four-story buildings are named F4, and so on. N-story buildings are named FN accordingly, solving the problem of misjudgment of the number of building floors in traditional 2D mapping.

[0028] Quantity calculation. By converting the surveyed multi-story building file into DXF format and importing it into ArcGIS, we can perform area statistical analysis of the multi-story building through data connection design and field creation. This solves the problem of CASS not being able to directly calculate the area of multi-story buildings and obtain the construction quantity of the building.

[0029] In building complexes with a floor area ratio greater than 2.5, the mapping efficiency is several times higher than that of total station measurements, with floor recognition accuracy reaching 100% and area statistical error less than 1%.

[0030] (5) Topographic map merging and accuracy verification.

[0031] Since the coordinate bases of the three types of naked-eye 3D digital mapping mentioned above are consistent, vectorized files in formats such as shp, dxf, and dwg can be directly merged through GIS. The merged files can be directly used for temporary road line selection, earthwork calculation, etc., can be directly added to the Aowei software for viewing, and can be directly merged and output with high-resolution DOM images, which is convenient for construction personnel to use, breaking the problem of multi-software data islands in traditional mapping, greatly shortening the mapping cycle, and reducing the input of field personnel.

[0032] The accuracy of naked-eye 3D digital mapping is assessed by the accuracy of the drone image control points. Using the known point inspection method, three coordinate points are measured on the actual ground, imported into the digital mapping file, and the differences between the coordinate values before and after mapping are compared. Since the topographic map scale required in construction is 1 / 500, according to the 1 / 500 mapping accuracy requirement, the checkpoints only need to meet the requirements of a plane error of less than 5cm and an elevation error of less than 5cm.

[0033] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle, characterized in that: Specifically include: 1) UAV data collection and processing; Deploy image control points around the mapping, measure the coordinates of each image control point using a positioning system, collect drone images, and perform two aerial triangulation solutions using aerial survey software to generate LAS format point cloud data, OSGB format 3D real scene models, and TIFF format DOM images. 2) 3D contour mapping based on point cloud data; The acquired point cloud data is sequentially classified using mathematical morphology filtering, slope filtering, and cloth simulation filtering. The ground point cloud data is converted into TIN triangulation data using the Delaunay triangulation method. 3D contour line data in DXF format is generated based on the TIN triangulation data. 3) Naked-eye 3D mapping of landforms and features based on digital orthophotos; Import DOM images into ArcGIS to create vectorized layers for woodlands, roads, farmlands, rivers, buildings, and greenhouses. Use topological analysis to establish relationships between vector graphics and raster graphics, creating a raster-to-vector - raster layer - vectorization - raster-to-vectorization workflow. Use Python scripts to calculate land attributes, area, and perimeter, and generate a table of project quantities. 4) Naked-eye 3D mapping of multi-story buildings based on real-scene models; Import the 3D real-world model in OSGB format into CASS 3D software. By linking the 3D model with the 2D image, draw closed polygons along the building's exterior walls while viewing the results in the 3D model. Convert the survey results into DXF format and import them into ArcGIS. Through data connection design and field creation, statistical analysis of the multi-story building area is performed. 5) Topographic map merging and accuracy verification; The vectorized files generated in the above steps are merged in GIS and directly output with the DOM image; the accuracy of the naked-eye 3D digital mapping is evaluated by the accuracy of the drone image control points.

2. The method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle according to claim 1, characterized in that: Collect drone images through a combination of fully automatic and manual flight modes; Professional aerial survey software is used for drone data processing. First, the original drone images are added, an aerial triangulation solution algorithm is established, and the photos are converted into aerial triangulation point cloud data. Secondly, the coordinates of the image control points are added, and the puncture point method is used to assign the image control point coordinates to the aerial triangulation point cloud data. Through the aerial triangulation optimization method and the image control point adjustment report, the errors of the aerial survey results are queried. For image control points with a position error greater than 5cm, the points need to be re-punctured and the drone data is converted to the construction coordinate system to ensure the accuracy of drone digital mapping. Finally, through the second aerial triangulation solution algorithm, high-precision dense point cloud data, 3D real-life model, and DOM image are respectively established.

3. The method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle according to claim 1, characterized in that: The point cloud is filtered and classified. First, the mathematical morphology filtering classification method is used to perform overall filtering and classification on the mapping point cloud data to obtain a coarse classification point cloud, which can quickly separate large areas of ground and non-ground points. Secondly, the slope filtering classification method is used to re-select areas with larger voids in the point cloud data, usually mountains or steep areas with larger slopes. By optimizing the slope classification parameters, the classification results of the ground point cloud are re-optimized. Finally, the fabric simulation filtering classification method is used to solve the problem of misclassification or classification anomalies of the concrete surface.

4. The method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle according to claim 1, characterized in that: Establish a TIN triangulated network and save the classified point cloud data as a new point cloud data file. Using the Delaunay triangulation method, the maximum side length of the triangulated network does not exceed 50m, and the display resolution is X=1m, Y=1m. The ground point cloud data is converted into an irregular triangulated network, namely TIN triangulated network data, and saved as a triangulated network file. Establish three-dimensional contour lines. According to construction requirements, use a topographic map scale of 1 / 500, corresponding to a main contour interval of 2.5m and a contour interval of 0.5m. Select the B-spline area as the smoothing type of the contour line, and the smoothing degree is 70% to 85%. Establish three-dimensional contour line data based on the TIN triangulated network, and the data format file is dxf.

5. The method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle according to claim 1, characterized in that: Establish vectorization parameters, import DOM images into ArcGIS, set the coordinate base, use shapefile tools to create new vectorization parameters, and establish vectorization layer parameters for woodland, roads, farmland, rivers, buildings, and greenhouses. Use different colors to distinguish different types of land features through layer style settings. Raster layer vectorization: Use ArcGIS's topology analysis function to perform topological analysis on raster data. Based on the topological analysis results, establish the relationship between vector graphics and raster graphics, forming a raster vector - raster layer - vectorization - raster - vectorization operation process. For areas with non-overlapping boundaries, use the area analysis tool in the corresponding layer to directly perform vectorization processing; Extract engineering quantities, select vectorized layer data, use data design to establish land attributes, units, quantity, area and perimeter fields, perform statistical analysis on the vectorized layer, and obtain the engineering quantity table.

6. The method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle according to claim 1, characterized in that: For digital mapping of multi-story buildings, CASS 3D software is used to import 3D real-life models in osgb format, set the building layer, and link the 3D model browsing with the 2D image. Using the multi-building area drawing method, closed polygons are drawn along the building's exterior wall while the results are viewed in the 3D model. For engineering quantity calculation, the files after mapping the multi-story building are converted into dxf format and imported into ArcGIS. Through data connection design and field creation, statistical analysis of the multi-story building's area is achieved.

7. The method for naked-eye 3D two- and three-dimensional digital mapping using an unmanned aerial vehicle according to claim 1, characterized in that: Vectorized files in shp, dxf, and dwg formats are directly merged through GIS. The merged files are directly used for road line selection and earthwork calculation, and are directly added to the Aowei software for viewing. They are directly merged and output with high-resolution DOM images for use by construction personnel.

Images