A method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar

Abstract

The present invention discloses a method for automatically identifying and modeling urban three-dimensional textures based on millimeter-wave radar. The method comprises the following steps: establishing intelligent air-ground data acquisition equipment to collect three-dimensional data of a target city; constructing an urban three-dimensional texture recognition algorithm to automatically identify the target city's three-dimensional texture; generating and modeling the target city's three-dimensional texture using a target detection algorithm; displaying the target city's three-dimensional texture through a visualization platform, and digitally interacting with the target city's three-dimensional texture in combination with a wearable virtual reality device; and printing the target city's three-dimensional texture using a 3D printing device to provide an intuitive display for city managers and urban designers. By leveraging millimeter-wave radar, a high-precision, full-dimensional data acquisition technology, the present invention addresses the problems of low data resolution and missing three-dimensional data in previous research, effectively improving the accuracy and dimensionality of urban texture research.

Description

Technical Field

[0001] The present invention relates to the technical field of urban planning, and in particular to a method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar. Background Art

[0002] Urban three-dimensional texture is one of the important research contents of urban planning and urban renewal, and is of great significance to urban development. Currently, there are the following limitations:

[0003] First, acquiring data on the three-dimensional texture of a city is complex and expensive, usually requiring the use of high-cost technologies such as LiDAR or high-precision aerial photogrammetry, and a lot of effort is needed to complete this basic work of data collection. The labor cost, economic cost, and time cost are high, and a lot of human and material resources are required.

[0004] Secondly, the accuracy of urban 3D texture recognition is limited. Due to data acquisition limitations and insufficient detail representation, the accuracy of urban 3D texture models is generally low. In particular, the accuracy of building geometry, height, and location may not meet the requirements of some high-precision applications.

[0005] Third, there are few interactive displays of urban three-dimensional texture evaluation. Although urban three-dimensional texture identification has broad application prospects in urban planning, architectural design, virtual reality and other fields, it currently cannot enable managers, decision makers and the public to intuitively understand the problems of urban three-dimensional texture and its influencing factors. Summary of the Invention

[0006] In order to solve the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a method for automatic recognition and modeling of urban three-dimensional texture based on millimeter wave radar.

[0007] The purpose of the present invention can be achieved by the following technical solution: A method for automatic recognition and modeling of urban three-dimensional texture based on millimeter wave radar, comprising the following steps:

[0008] S1 Intelligent Collection of Urban 3D Texture Data: Build an intelligent air-ground dual-dimensional data collection device equipped with millimeter-wave radar to collect real-world scenes of buildings and open spaces in the target city, with both two-dimensional and three-dimensional data. The air-ground and ground data are translated and stored in a unified dataset to construct a basic dataset for the target city.

[0009] S2 Automatic recognition of urban 3D texture: Call the target city basic dataset constructed in step S1, generate urban building 3D texture for building entities, rasterize the open space virtual bodies and construct spatial triangulations for each grid, calculate the height of each grid with the millimeter wave radar intelligent ranging equipment, generate urban open space 3D texture, and store the generated urban building 3D texture and urban open space 3D texture in the target city 3D texture dataset;

[0010] S3 Automatic modeling of urban 3D texture: Call the urban 3D texture dataset calculated in step S2, uniformly translate different types of data into 3D spatial coordinates, call the image data collected in the urban basic data in step S1, use the target detection algorithm to automatically identify the 3D texture plane and facade element information, and unify the coordinates of all data. Use the laser VR scanner to generate a 3D real scene of the urban 3D texture identified in step S2;

[0011] S4: Using a handheld controller to display the three-dimensional urban texture model in step S3, and using VR glasses and virtual reality gloves to interact with the model.

[0012] S5 Results Printing: Print the three-dimensional texture model of the target city generated in step S4 through a large desktop 3D printing device and a drawing data output device; the printing includes two types: plan view and three-dimensional model. The three-dimensional model is colored using coloring software.

[0013] Furthermore, in step S1, an air-ground two-dimensional data intelligent collection device is constructed, which is equipped with a millimeter-wave radar with a transmission frequency of 77 Hz, a ranging range of 50 m, and a ranging angle of 60°. During aerial collection, the device is mounted on a drone, and during ground collection, the device is mounted around vehicles and buildings to collect real scenes of the target city. The collected data types include two categories: building entities and open spaces, and the collection dimensions include two categories: plane and three-dimensional. The air and ground data are translated and stored together in a unified data set NAS device to construct a basic data set of the target city. Among them, the plane data includes the perimeter, area, spatial coordinates, and fifth facade photos of the building entity; the length, width, spatial coordinates, and plane views of the open space; and the three-dimensional data includes the height of the building entity and the facade photos of the building.

[0014] Furthermore, in step S1, the aerial equipment uses a millimeter-wave radar drone, and the ground equipment uses a millimeter-wave radar scanner fixed on a vehicle and a millimeter-wave radar sensor installed around the building (structure). The specific functions are that the millimeter-wave radar drone is used to collect point cloud data of urban high-rise and super-high-rise buildings and take photos of the building facades, the millimeter-wave radar scanner fixed on the vehicle is used to collect point cloud data of urban single-story and multi-story buildings and structures of 3m and above, and the millimeter-wave radar sensors installed around the building (structure) are used to collect point cloud data of structures below 3m, plants and decorations in the virtual body of open space.

[0015] Furthermore, the step S2 specifically includes:

[0016] Step S2-1: Retrieve the perimeter and height information from the target city basic dataset constructed in step S1, and divide the city's three-dimensional texture into building entities and open space virtual bodies. Building entities must simultaneously meet the requirements of being a quadrilateral, having a shape integrity greater than 50%, and being greater than 3 meters in height.

[0017] Step S2-2: For building entities, call the building height information in the target city basic dataset in step S1 to generate a 3D texture of the city buildings and store it in the target city 3D texture dataset;

[0018] Step S2-3: For the open space virtual body, divide the open space plane into 10m×10m grids, and rename each grid in sequence as "Open Space-n"; where "n" represents the number, and delete grids with an area of ​​less than 10m×10m.

[0019] Step S2-4: Using an intelligent laser ranging device equipped with a millimeter-wave radar with a transmission frequency of 77 GHz, a ranging range of 40 m, and a lateral angle range of 150°, scan the vertices and midpoints of the four sides of the grid generated in step S2-3, construct eight spatial triangular faces, and sequentially name them "n spatial triangular facet-m"; where "n" is the same as the grid number and "m" represents the number of faces.

[0020] Step S2-5: Retrieve the target city building plan and height data obtained in step S1, combine it with the grid space triangles constructed in step S2-4 to calculate the open space height, generate the three-dimensional texture of the city open space, and store it in the target city three-dimensional texture dataset.

[0021] Furthermore, the spatial triangles of each grid are constructed by connecting the grid midpoint, the top edge of the adjacent building roof, and the vertex of the bottom edge of the adjacent building ground to obtain 8 "spatial triangles"; wherein, the adjacent building refers to a building entity with a width of less than or equal to 40m between the bottom edge of the building and the bottom edge of the open space; the building edge refers to the line connecting the top edge of the adjacent building roof and the vertex of the bottom edge of the ground, that is, the edge of the "spatial triangle" coincides with a certain facade of the building.

[0022] Furthermore, the calculation of the height of each grid by combining the millimeter wave radar intelligent ranging device means taking the line connecting the four corners of the grid as the midpoint of the grid, and calculating the height of each grid by combining the millimeter wave radar intelligent ranging device and the midpoint of the grid S M Use an intelligent outdoor laser ranging device equipped with a millimeter-wave radar with a transmission frequency of 77GHz, a ranging range of 40m, and a side angle range of 150° to draw a vertical line to the surrounding buildings, and record the vertical point as the building point S A , measure the distance from the grid midpoint to the building point; then draw a vertical line from the building point to the building facade, and record the intersection of the vertical line and the top edge of the building as the building top edge point S B ; Measure building point S A To the top edge point S of the building B The distance is recorded as the building point height H AB , grid height Where m∈(1,8).

[0023] Furthermore, the step S3 specifically includes:

[0024] Step S3-1 calls the urban 3D texture data set calculated in step S2 to collect the building heights of the urban building 3D texture and the open space heights of the urban open space 3D texture, and uniformly translates the building height data and the open space height data into 3D space coordinates;

[0025] Step S3-2 calls the fifth facade image of the building entity, the open space plane photo, and the building (structure) facade photo data in the urban basic data set of step S1, and automatically identifies the image data using the R-FCN target detection algorithm; the recognition process includes: first preprocessing the photo to eliminate irrelevant information; importing the preprocessed image into the trained ResNet-101 classification network; there are three cognitions on the feature map obtained by the last convolutional layer of the pretrained network, the first branch performs RPN operation on the feature map to obtain the corresponding ROI, the second branch obtains a K*K*(C+1)-dimensional position-sensitive score map for classification on the feature map, and the third branch obtains a 4*K*K-dimensional position-sensitive score map for regression on the feature map; position-sensitive average pooling operations are performed on the K*K*(C+1)-dimensional position-sensitive score map and the 4*K*K-dimensional position-sensitive score map respectively to obtain corresponding categories and position information; the recognition content includes three-dimensional texture plane and facade element information; the recognition result is the coordinate information and spatial relationship information of the image;

[0026] Step S3-3 unifies the coordinate data of steps S3-1 and S3-2 into the WGS 84 coordinate system;

[0027] In step S3-4, the spatial coordinates after the unified coordinate system in step S3-3 and the point cloud data obtained by the millimeter-wave radar scanning in step S1 are processed to remove overlapping points and conflicting points, and a laser VR scanner is used to generate a three-dimensional real scene of the three-dimensional texture of the city.

[0028] Furthermore, the results display in step S4 uses a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, and a sand table to construct a three-dimensional digital sand table, and cooperates with a handheld controller to perform a virtual scene display of the three-dimensional texture model of the city in step S3, and uses VR glasses and virtual reality gloves to realize scene interaction with the model; wherein, an in-depth experience of the model is realized through the handheld controller; the selection, scaling, and modification of the model are realized through the virtual reality gloves; the interactive display of the model is realized through the VR glasses, and the model can be operated together with the virtual reality gloves.

[0029] Furthermore, in step S5, the result printing is performed by printing the target city three-dimensional texture model generated in S4 through a large-scale desktop 3D printing device and a drawing data output device; the interactive model described in step S4 is used to select the printing range and printing angle, and the selected city three-dimensional texture model is integrated using a data integration and transfer device, including two types of plan view and three-dimensional model. The selected model is 3D printed through a large-scale desktop 3D printing device, and the model scale is 1:10000. The printed white model is colored by coloring software; the plan view of the model is input through the drawing data output device, and the printing scale is 1:10000.

[0030] Beneficial effects of the present invention:

[0031] 1. High Data Collection Efficiency: This invention utilizes the strong penetration and high resolution of millimeter-wave radar to collect high-precision and high-efficiency data at the urban scale. This solves the problem of urban data collection being greatly affected by weather and the low resolution of collected data. The air-ground dual-dimensional data collection equipment can efficiently collect data on urban buildings and open spaces in one go, effectively avoiding data loss and repeated data collection and replenishment.

[0032] 2. Objectivity of the process: This invention uses spatial triangulation to calculate three-dimensional spatial information, generates three-dimensional architectural textures for building entities, and constructs spatial triangulations for virtual open spaces to generate three-dimensional textures for open spaces. In this process, the data acquired by millimeter-wave radar scanning and the height obtained by intelligent measurement together constitute the three-dimensional texture data of the city. Each step of data acquisition and result measurement is objective and credible.

[0033] 3. Interactive visibility: This invention constructs a three-dimensional digital sand table and uses a handheld controller to display the three-dimensional urban texture model in a virtual scene. VR glasses and virtual reality gloves are used to enable users to interact with the model in a scene-based manner, thus realizing the visualization and interactivity of the three-dimensional urban texture model. BRIEF DESCRIPTION OF THE DRAWINGS

[0034] Figure 1 It is a flow chart of the method of the present invention;

[0035] Figure 2 It is a flowchart for automatic recognition of urban three-dimensional texture. DETAILED DESCRIPTION

[0036] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making any creative efforts shall fall within the scope of protection of the present invention.

[0037] like Figure 1 and 2 As shown, a method for automatic recognition and modeling of urban three-dimensional texture based on millimeter wave radar includes the following steps:

[0038] S1 Intelligent Collection of Urban 3D Texture Data: Build an intelligent air-ground dual-dimensional data collection device equipped with millimeter-wave radar to collect real-world scenes of buildings and open spaces in the target city, with both two-dimensional and three-dimensional data. The air-ground and ground data are translated and stored in a unified dataset to construct a basic dataset for the target city.

[0039] In step S1, an air-ground dual-dimensional data intelligent collection device equipped with a millimeter-wave radar with a transmission frequency of 77 Hz, a ranging range of 50 m, and a ranging angle of 60° is constructed. During aerial collection, the device is mounted on a drone, and during ground collection, the device is mounted around vehicles and buildings (structures) to collect real-life scenes of the target city. The collected data types include two categories: building entities and open spaces, and the collection dimensions include two categories: plane and three-dimensional. The air and ground data are translated and stored together in a unified data set NAS device to construct a basic data set of the target city. The plane data includes the perimeter, area, spatial coordinates, and fifth facade photos of the building entity; the length, width, spatial coordinates, and plane views of the open space; and the three-dimensional data includes the height of the building entity and building facade photos.

[0040] In step S1, the aerial equipment uses a millimeter-wave radar drone, and the ground equipment uses a millimeter-wave radar scanner fixed on a vehicle and a millimeter-wave radar sensor installed around buildings (structures). The specific functions are: the millimeter-wave radar drone is used to collect point cloud data of urban high-rise and super-high-rise buildings and take photos of the building facades; the millimeter-wave radar scanner fixed on the vehicle is used to collect point cloud data of urban single-story and multi-story buildings and structures of 3m and above; the millimeter-wave radar sensors installed around buildings (structures) are used to collect point cloud data of structures below 3m, plants and decorations in the virtual body of open space.

[0041] S2 Automatic recognition of urban 3D texture: Call the target city basic dataset constructed in step S1, generate urban building 3D texture for building entities, rasterize the open space virtual bodies and construct spatial triangulations for each grid, calculate the height of each grid with the millimeter wave radar intelligent ranging equipment, generate urban open space 3D texture, and store the generated urban building 3D texture and urban open space 3D texture in the target city 3D texture dataset;

[0042] The step S2 specifically includes:

[0043] Step S2-1: Retrieve the perimeter and height information from the target city basic dataset constructed in step S1, and divide the city's three-dimensional texture into building entities and open space virtual bodies. Building entities must simultaneously meet the requirements of being a quadrilateral, having a shape integrity greater than 50%, and being greater than 3 meters in height.

[0044] Step S2-2: For building entities, call the building height information in the target city basic dataset in step S1 to generate a 3D texture of the city buildings and store it in the target city 3D texture dataset;

[0045] Step S2-3: For the open space virtual body, divide the open space plane into 10m×10m grids, and rename each grid in sequence as "Open Space-n"; where "n" represents the number, and delete grids with an area of ​​less than 10m×10m.

[0046] Step S2-4: Using an intelligent laser ranging device equipped with a millimeter-wave radar with a transmission frequency of 77 GHz, a ranging range of 40 m, and a lateral angle range of 150°, scan the vertices and midpoints of the four sides of the grid generated in step S2-3, construct eight spatial triangular faces, and sequentially name them "n spatial triangular facet-m"; where "n" is the same as the grid number and "m" represents the number of faces.

[0047] Step S2-5: Retrieve the target city building plan and height data obtained in step S1, combine it with the grid space triangles constructed in step S2-4 to calculate the open space height, generate the three-dimensional texture of the city open space, and store it in the target city three-dimensional texture dataset.

[0048] The spatial triangles of each grid are constructed by connecting the grid midpoint, the top edge of the adjacent building roof, and the vertex of the bottom edge of the adjacent building ground to obtain 8 "spatial triangles"; wherein, the adjacent building refers to a building entity with a width of less than or equal to 40m between the bottom edge of the building and the bottom edge of the open space; the building edge refers to the line connecting the top edge of the adjacent building roof and the vertex of the bottom edge of the ground, that is, the edge of the "spatial triangle" coincides with a certain facade of the building.

[0049] The method of calculating the height of each grid by combining the millimeter wave radar intelligent ranging device is to take the line connecting the four corners of the grid as the midpoint of the grid, and calculate the height of each grid by combining the millimeter wave radar intelligent ranging device. M Use an intelligent outdoor laser ranging device equipped with a millimeter-wave radar with a transmission frequency of 77GHz, a ranging range of 40m, and a side angle range of 150° to draw a vertical line to the surrounding buildings, and record the vertical point as the building point S A , measure the distance from the grid midpoint to the building point; then draw a vertical line from the building point to the building facade, and record the intersection of the vertical line and the top edge of the building as the building top edge point S B ; Measure building point S A To the top edge point S of the building B The distance is recorded as the building point height H AB , grid height Where m∈(1,8).

[0050] S3 Automatic modeling of urban 3D texture: Call the urban 3D texture dataset calculated in step S2, uniformly translate different types of data into 3D spatial coordinates, call the image data collected in the urban basic data in step S1, use the target detection algorithm to automatically identify the 3D texture plane and facade element information, and unify the coordinates of all data. Use the laser VR scanner to generate a 3D real scene of the urban 3D texture identified in step S2;

[0051] The step S3 specifically includes:

[0052] Step S3-1 calls the urban 3D texture data set calculated in step S2 to collect the building heights of the urban building 3D texture and the open space heights of the urban open space 3D texture, and uniformly translates the building height data and the open space height data into 3D space coordinates;

[0053] Step S3-2 calls the fifth facade image of the building entity, the open space plane photo, and the building (structure) facade photo data in the urban basic data set of step S1, and automatically identifies the image data using the R-FCN target detection algorithm; the recognition process includes: first preprocessing the photo to eliminate irrelevant information; importing the preprocessed image into the trained ResNet-101 classification network; there are three cognitions on the feature map obtained by the last convolutional layer of the pretrained network, the first branch performs RPN operation on the feature map to obtain the corresponding ROI, the second branch obtains a K*K*(C+1)-dimensional position-sensitive score map for classification on the feature map, and the third branch obtains a 4*K*K-dimensional position-sensitive score map for regression on the feature map; position-sensitive average pooling operations are performed on the K*K*(C+1)-dimensional position-sensitive score map and the 4*K*K-dimensional position-sensitive score map respectively to obtain corresponding categories and position information; the recognition content includes three-dimensional texture plane and facade element information; the recognition result is the coordinate information and spatial relationship information of the image;

[0054] Step S3-3 unifies the coordinate data of steps S3-1 and S3-2 into the WGS 84 coordinate system;

[0055] In step S3-4, the spatial coordinates after the unified coordinate system in step S3-3 and the point cloud data obtained by the millimeter-wave radar scanning in step S1 are processed to remove overlapping points and conflicting points, and a laser VR scanner is used to generate a three-dimensional real scene of the three-dimensional texture of the city.

[0056] S4: Using a handheld controller to display the three-dimensional urban texture model in step S3, and using VR glasses and virtual reality gloves to interact with the model.

[0057] The results display in step S4 uses a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, and a sand table to construct a three-dimensional digital sand table, and cooperates with a handheld controller to perform a virtual scene display of the three-dimensional texture model of the city in step S3, and uses VR glasses and virtual reality gloves to realize scene interaction with the model; wherein, an in-depth experience of the model is realized through the handheld controller; the selection, scaling, and modification of the model are realized through the virtual reality gloves; the interactive display of the model is realized through the VR glasses, and the model can be operated together with the virtual reality gloves.

[0058] S5 Results Printing: Print the three-dimensional texture model of the target city generated in step S4 through a large desktop 3D printing device and a drawing data output device; the printing includes two types: plan view and three-dimensional model. The three-dimensional model is colored using coloring software.

[0059] The results of step S5 are printed by printing the target city three-dimensional texture model generated in step S4 through a large-scale desktop 3D printing device and a drawing data output device; the interactive model described in step S4 is used to select the printing range and printing angle, and the selected city three-dimensional texture model is integrated using a data integration and transfer device, including two types of plan views and three-dimensional models. The selected model is 3D printed through a large-scale desktop 3D printing device, and the model scale is 1:10000. The printed white model is colored by coloring software; the plan view of the model is output through the drawing data output device, and the printing scale is 1:10000.

Claims

1. A method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar, characterized in that: The following steps are involved: S1 Intelligent Collection of Urban 3D Texture Data: Build an intelligent air-ground dual-dimensional data collection device equipped with millimeter-wave radar to collect real-world scenes of buildings and open spaces in the target city, with both two-dimensional and three-dimensional data. The air-ground and ground data are translated and stored in a unified dataset to construct a basic dataset for the target city. S2 Automatic recognition of urban 3D texture: Call the target city basic dataset constructed in step S1, generate urban building 3D texture for building entities, rasterize the open space virtual bodies and construct spatial triangulations for each grid, calculate the height of each grid with the millimeter wave radar intelligent ranging equipment, generate urban open space 3D texture, and store the generated urban building 3D texture and urban open space 3D texture in the target city 3D texture dataset; S3 Automatic modeling of urban 3D texture: Call the urban 3D texture dataset calculated in step S2, uniformly translate different types of data into 3D spatial coordinates, call the image data collected in the urban basic data in step S1, use the target detection algorithm to automatically identify the 3D texture plane and facade element information, and unify the coordinates of all data. Use the laser VR scanner to generate a 3D real scene of the urban 3D texture identified in step S2; S4: Using a handheld controller to display the three-dimensional urban texture model in step S3, and using VR glasses and virtual reality gloves to interact with the model. S5 Results Printing: Print the three-dimensional texture model of the target city generated in step S4 through a large desktop 3D printing device and a drawing data output device; the printing includes two types: plan view and three-dimensional model. The three-dimensional model is colored using coloring software.

2. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 1 is characterized in that: In step S1, an air-ground dual-dimensional data intelligent collection device equipped with a millimeter-wave radar with a transmission frequency of 77 Hz, a ranging range of 50 m, and a ranging angle of 60° is constructed. The device is mounted on a drone for aerial collection, and is mounted around vehicles and buildings for ground collection to collect real scenes of the target city. The collected data types include two categories: building entities and open spaces, and the collection dimensions include two categories: plane and three-dimensional. The air and ground data are translated and stored together in a unified data set NAS device to construct a basic data set of the target city. The plane data includes the perimeter, area, spatial coordinates, and fifth facade photos of the building entity; the length, width, spatial coordinates, and plane views of the open space; and the three-dimensional data includes the height of the building entity and building facade photos.

3. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 2 is characterized in that: In step S1, the aerial equipment uses a millimeter-wave radar drone, and the ground equipment uses a millimeter-wave radar scanner fixed on a vehicle and millimeter-wave radar sensors installed around buildings. The specific functions are: the millimeter-wave radar drone is used to collect point cloud data of urban high-rise and super-high-rise buildings and take photos of building facades; the millimeter-wave radar scanner fixed on the vehicle is used to collect point cloud data of urban single-story and multi-story buildings and structures of 3m and above; the millimeter-wave radar sensors installed around buildings are used to collect point cloud data of structures below 3m, plants and decorations in open space virtual bodies.

4. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 3 is characterized in that: The step S2 specifically includes: Step S2-1: Retrieve the perimeter and height information from the target city basic dataset constructed in step S1, and divide the city's three-dimensional texture into building entities and open space virtual bodies. Building entities must simultaneously meet the requirements of being a quadrilateral, having a shape integrity greater than 50%, and being greater than 3 meters in height. Step S2-2: For building entities, call the building height information in the target city basic dataset in step S1 to generate a 3D texture of the city buildings and store it in the target city 3D texture dataset; Step S2-3: For the open space virtual body, divide the open space plane into 10m×10m grids, and rename each grid sequentially as "Open Space-n"; where "n" represents the number, delete grids with an area less than 10m×10m. Step S2-4: Using an intelligent laser ranging device equipped with a millimeter-wave radar with a transmission frequency of 77 GHz, a ranging range of 40 meters, and a lateral angle range of 150 degrees, scan the vertices and midpoints of the four sides of the grid generated in step S2-3, construct eight spatial triangular faces, and sequentially name them "n spatial triangular facet-m"; where "n" is the same as the grid number, and "m" represents the number of faces. Step S2-5: Retrieve the target city building plan and height data obtained in step S1, combine it with the grid space triangles constructed in step S2-4 to calculate the open space height, generate the three-dimensional texture of the city open space, and store it in the target city three-dimensional texture dataset.

5. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 4 is characterized in that: The spatial triangles of each grid are constructed by connecting the grid midpoint, the top edge of the adjacent building roof, and the vertex of the bottom edge of the adjacent building ground to obtain eight "spatial triangles." Among them, the adjacent building refers to a building entity with a width of less than or equal to 40 meters between the bottom edge of the building and the bottom edge of the open space. The building edge refers to the line connecting the top edge of the adjacent building roof and the vertex of the bottom edge of the ground, that is, the edge of the "spatial triangle" coincides with a certain facade of the building.

6. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 5 is characterized in that: The method of calculating the height of each grid by combining the millimeter wave radar intelligent ranging device is to take the line connecting the four corners of the grid as the midpoint of the grid, and calculate the height of each grid by combining the millimeter wave radar intelligent ranging device. M Use an intelligent outdoor laser ranging device equipped with a millimeter-wave radar with a transmission frequency of 77GHz, a ranging range of 40m, and a side angle range of 150° to draw a vertical line to the surrounding buildings, and record the vertical point as the building point S A , measure the distance from the grid midpoint to the building point; then draw a vertical line from the building point to the building facade, and record the intersection of the vertical line and the top edge of the building as the building top edge point S B ; Measure building point S A To the top edge point S of the building B The distance is recorded as the building point height H AB , grid height Where m∈(1,8).

7. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 6 is characterized in that: The step S3 specifically includes: Step S3-1 calls the urban 3D texture data set calculated in step S2 to collect the building heights of the urban building 3D texture and the open space heights of the urban open space 3D texture, and uniformly translates the building height data and the open space height data into 3D space coordinates; Step S3-2 calls the fifth facade image of the building entity, the open space plane photo, and the building facade photo data in the urban basic data set of step S1, and automatically identifies the image data using the R-FCN target detection algorithm; the recognition process includes: first preprocessing the photo to eliminate irrelevant information; importing the preprocessed image into the trained ResNet-101 classification network; there are three cognitions on the feature map obtained by the last convolutional layer of the pretrained network, the first branch performs RPN operation on the feature map to obtain the corresponding ROI, the second branch obtains a K*K*(C+1)-dimensional position-sensitive score map for classification on the feature map, and the third branch obtains a 4*K*K-dimensional position-sensitive score map for regression on the feature map; position-sensitive average pooling operations are performed on the K*K*(C+1)-dimensional position-sensitive score map and the 4*K*K-dimensional position-sensitive score map respectively to obtain corresponding categories and position information; the recognition content includes three-dimensional texture plane and facade element information; the recognition result is the coordinate information and spatial relationship information of the image; Step S3-3 unifies the coordinate data of steps S3-1 and S3-2 into the WGS 84 coordinate system; In step S3-4, the spatial coordinates after the unified coordinate system in step S3-3 and the point cloud data obtained by the millimeter-wave radar scanning in step S1 are processed to remove overlapping points and conflicting points, and a laser VR scanner is used to generate a three-dimensional real scene of the three-dimensional texture of the city.

8. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 7 is characterized in that: The results display in step S4 uses a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, and a sand table to construct a three-dimensional digital sand table, and cooperates with a handheld controller to perform a virtual scene display of the three-dimensional texture model of the city in step S3, and uses VR glasses and virtual reality gloves to realize scene interaction with the model; wherein, an in-depth experience of the model is realized through the handheld controller; the selection, scaling, and modification of the model are realized through the virtual reality gloves; the interactive display of the model is realized through the VR glasses, and the model can be operated together with the virtual reality gloves.

9. The method for automatic recognition and modeling of urban three-dimensional texture based on millimeter-wave radar according to claim 8, characterized in that: The results of step S5 are printed by printing the target city three-dimensional texture model generated in step S4 through a large-scale desktop 3D printing device and a drawing data output device; the interactive model described in step S4 is used to select the printing range and printing angle, and the selected city three-dimensional texture model is integrated using a data integration and transfer device, including two types of plan views and three-dimensional models. The selected model is 3D printed through a large-scale desktop 3D printing device, and the model scale is 1:10000. The printed white model is colored by coloring software; the plan view of the model is output through the drawing data output device, and the printing scale is 1:10000.

Images