A Method for Constructing Urban 3D Models Based on High-Resolution Satellite Imagery

Abstract

This application relates to the field of remote sensing image information processing, specifically disclosing a method for constructing a 3D urban model based on high-resolution stereo imaging satellite imagery. The method includes: extracting the top building outline of a target building from the satellite imagery; obtaining the top height information of the target building based on the corresponding Digital Surface Model (DSM) imagery; offsetting the top building outline to obtain the bottom building outline of the target building; and establishing a 3D model of the target building based on the bottom building outline and the top height information. This application utilizes satellite imagery to achieve rapid, efficient, and accurate construction of a 3D urban model at LOD 1.3.

Description

Technical Field

[0001] This application relates to the technical field of remote sensing image information processing, and in particular to a method for constructing a three-dimensional urban model based on high-resolution satellite imagery. Background Technology

[0002] With the development of urbanization, industrialization, and modernization, large and mega-cities are constantly emerging in my country, and urban construction has achieved rapid progress. This has also brought severe challenges to urban governance. In recent years, the emergence of information and intelligent technologies has provided a technological foundation for promoting urban reform and governance towards a more human-centered and intelligent direction. In the field of smart city construction, the construction of 3D city models is becoming an important component, playing a significant role in areas such as map applications, urban planning, and virtual events.

[0003] my country's remote sensing Earth observation technology has been continuously developing, with breakthroughs in high-resolution remote sensing satellite technology, an increasing number of civilian and commercial satellites being launched, and new highs being achieved in the scale and quality of images, laying the foundation for the application of remote sensing images in the construction of urban 3D models.

[0004] In traditional urban 3D model building, building outlines are often drawn manually. This method has low automation, is time-consuming, labor-intensive, and costly. Furthermore, it is slow to update and cannot keep track of real-time changes in urban buildings.

[0005] With the application and development of deep learning technology in the field of remote sensing, deep learning-based remote sensing image building outline extraction technology has increasingly served many fields such as urban construction, land monitoring, and target surveillance, and has also provided new methods for constructing urban 3D models. Utilizing a "remote sensing + AI" approach for building outline extraction and urban 3D construction has become a hot topic. Building height, shape, and location information are the foundation for constructing urban 3D models. Using remote sensing imagery to comprehensively extract building height, shape, and location information to achieve rapid, end-to-end construction of high-precision urban 3D models has become an industry focus.

[0006] However, a single remote sensing image cannot provide sufficient information about a building's height. Furthermore, the building top outlines obtained using deep learning-based building extraction techniques are often irregular and fail to reflect the building's true shape. Additionally, during building outline extraction, variations in satellite sensor attitude often cause top-to-bottom offsets in the image, resulting in discrepancies between the automatically extracted top outline and the actual bottom of the building, thus affecting the building's positional accuracy.

[0007] These issues have all contributed to the difficulty of constructing high-precision 3D models of cities. The challenge of rapidly acquiring building height, shape, and location information using remote sensing imagery has always been a hurdle. The need to quickly, efficiently, and accurately construct LOD 1.3 level 3D city models using satellite imagery is becoming increasingly urgent (LOD 1.0 cannot represent building forms, LOD 1.1 can represent a rough form, LOD 1.2 can represent a detailed form but not the height differences between adjacent buildings, and LOD 1.3 can represent both the detailed form and the height differences between adjacent buildings). Summary of the Invention

[0008] This application provides a method for constructing urban 3D models, utilizing satellite imagery to achieve rapid, efficient, and accurate construction of urban 3D models. This application combines DSM automatic generation, building outline extraction and position correction, building outline fusion, and building white-film production to construct a rapid and automated framework for producing urban LOD1.3 level 3D models, achieving high-precision building white-film production and rapid construction of urban 3D models.

[0009] Firstly, a method for constructing a 3D urban model is provided, including:

[0010] Extract the top outline of the target building from satellite imagery;

[0011] Based on the digital surface model (DSM) image corresponding to the target building, obtain the top height information of the target building;

[0012] The bottom building outline of the target building is obtained by offsetting the top building outline;

[0013] A three-dimensional model of the target building is created based on the bottom building outline and the top height information.

[0014] Compared with the prior art, the solution provided in this application has at least the following beneficial technical effects:

[0015] This application establishes a rapid high-precision urban 3D model construction method based on high-resolution stereo imaging satellite imagery. This method achieves a complete workflow framework design for high-precision building white-film production and urban 3D model construction through methods such as DSM image production based on stereo image pairs from remote sensing images, building outline extraction, distance and orientation calculation of building outline offsets, multi-angle building outline fusion, and building height acquisition. Specifically, the extracted building top outlines are automatically normalized and offset corrected, and multi-angle building outline fusion technology is used to obtain accurate building bottom outlines, improving the accuracy of building geographic location and laying the foundation for high-precision white-film production and urban 3D model construction.

[0016] In conjunction with the first aspect, in some implementations of the first aspect, obtaining the top height information of the target building based on the digital surface model (DSM) image corresponding to the target building includes:

[0017] Based on the set of geographic coordinate information and the set of height information in the DSM image, determine the set of pixel coordinate information corresponding to the DSM image;

[0018] Based on the set of pixel coordinate information, the DSM image is overlaid with the outline of the top building;

[0019] The top height information is obtained based on the height information indicated by the pixels located within the outline of the top building in the DSM image.

[0020] By overlaying DSM imagery with the top building outline, the height information corresponding to the top building outline can be accurately obtained. Additionally, other information contained in the DSM imagery can be mapped to the top building outline.

[0021] In conjunction with the first aspect, in some implementations of the first aspect, determining the set of pixel coordinate information corresponding to the DSM image based on the set of geographic coordinate information and the set of height information in the DSM image includes:

[0022] The geographic coordinate information set and the altitude information set are input into the rational function RPC (rational polynomial coefficients) model for orthographic projection transformation to obtain the pixel coordinate information set.

[0023] The RPC model can quickly convert between pixel coordinates and geographic coordinates with high accuracy and is relatively user-friendly for converting coordinates of multiple buildings.

[0024] In conjunction with the first aspect, in some implementations of the first aspect, the top height information is the median of all height values ​​indicated by pixels located within the top building outline in the DSM image.

[0025] The median height value can represent the average height of the building's top and has relatively high representativeness.

[0026] In conjunction with the first aspect, in certain implementations of the first aspect, the offsetting of the top building outline to obtain the bottom building outline of the target building includes:

[0027] Obtain the geographic coordinate information corresponding to the top building outline;

[0028] The digital elevation model (DEM) image corresponding to the target building is overlaid with the geographic coordinate information.

[0029] Based on the height information indicated by the pixels in the DEM image corresponding to the geographic coordinate information, the bottom height information of the target building is obtained;

[0030] The bottom building outline of the target building is determined based on the bottom height information and the geographic coordinate information.

[0031] By converting geographic coordinates and pixel coordinates between DSM and DEM models, coordinate offset information can be made more accurate.

[0032] In conjunction with the first aspect, in some implementations of the first aspect, obtaining the geographic coordinate information corresponding to the top building outline includes:

[0033] The geographic coordinate information is obtained based on the pixel coordinate information corresponding to the top building outline and the top height information.

[0034] Geographic coordinates are obtained by converting pixel coordinates to geographic coordinates, which has relatively high accuracy.

[0035] In conjunction with the first aspect, in some implementations of the first aspect, obtaining the geographic coordinate information based on the pixel coordinate information corresponding to the top building outline and the top height information includes:

[0036] The pixel coordinate information and the top height information are input into a rational function RPC model for inverse projection transformation to obtain the geographic coordinate information.

[0037] The RPC model can quickly convert between pixel coordinates and geographic coordinates with high accuracy and is relatively user-friendly for converting coordinates of multiple buildings.

[0038] In conjunction with the first aspect, in some implementations of the first aspect, obtaining the bottom building outline of the target building based on the bottom height information and the geographic coordinate information includes:

[0039] Based on the aforementioned geographic coordinate information, determine the centroid geographic coordinates;

[0040] The bottom centroid pixel coordinates are determined based on the centroid geographic coordinates and the bottom height information;

[0041] The top centroid pixel coordinates are determined based on the pixel coordinate information corresponding to the top building outline.

[0042] The bottom building outline is obtained by offsetting the top building outline based on the offset between the bottom centroid pixel coordinates and the top centroid pixel coordinates.

[0043] The centroid can represent the location of a building's outline and has relatively high representativeness.

[0044] In conjunction with the first aspect, in some implementations of the first aspect, determining the bottom centroid pixel coordinates based on the centroid geographic coordinates and the bottom height information includes:

[0045] The centroid geographic coordinates and the bottom height information are input into the rational function RPC model for orthographic transformation to obtain the bottom centroid pixel coordinates.

[0046] The RPC model can quickly convert between pixel coordinates and geographic coordinates with high accuracy and is relatively user-friendly for converting coordinates of multiple buildings.

[0047] In conjunction with the first aspect, in some implementations of the first aspect, the bottom height information is the median of all height values ​​indicated by all pixels of the DEM image corresponding to the geographic coordinate information.

[0048] The median height value can represent the average height of the building's base and is relatively representative.

[0049] In a second aspect, an electronic device is provided, the electronic device comprising: one or more processors; one or more memories; the one or more memories storing one or more computer programs, the one or more computer programs including instructions that, when executed by the one or more processors, cause the electronic device to perform the method as described in any of the implementations of the first aspect above. Attached Figure Description

[0050] Figure 1 A schematic flowchart illustrating a method for constructing a three-dimensional city model, provided as an embodiment of this application;

[0051] Figure 2 A schematic diagram of the extracted top building outline;

[0052] Figure 3 A schematic diagram showing the overlay of DSM imagery and the top building outline;

[0053] Figure 4 A schematic diagram illustrating the regularization of the top building's outline;

[0054] Figure 5 A flowchart illustrating the offset of the top building outline to the bottom building outline;

[0055] Figure 6 This is a schematic diagram showing the offset of the top building outline to the bottom building outline;

[0056] Figure 7A flowchart illustrating the construction process of a 3D city model is provided in this embodiment of the application.

[0057] Figure 8 This is a schematic diagram of a three-dimensional city model. Detailed Implementation

[0058] This application provides a method for rapid construction of high-precision urban 3D models based on high-resolution satellite imagery of stereo imaging, which mainly includes: (1) generation of DSM based on stereo image pairs and acquisition of open source DEM data; (2) extraction and normalization of building top surface contours; (3) correction of top and bottom offset of building contour positions based on rational polynomial coefficient (RPC) model; (4) fusion of the corrected building bottom contours extracted from images from two (or more) angles, and obtaining more accurate building contour information based on confidence index; (5) acquisition of building height information based on DSM and DEM; and (6) automatic production of building white film and construction of urban 3D models by using the obtained building bottom contour positions combined with height information.

[0059] The present application will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0060] This application discloses a method for constructing a three-dimensional urban model, referring to... Figure 1 This includes the following steps:

[0061] 110. Extract the top building outline of the target building from the satellite image.

[0062] 120. Based on the digital surface model (DSM) image of the target building, obtain the top height information of the target building.

[0063] 130, offset the top building outline to obtain the bottom building outline of the target building.

[0064] 140. Based on the bottom building outline and top height information, create a three-dimensional model of the target building.

[0065] The urban 3D model construction method provided in this application embodiment can automatically regularize and correct the extracted top building outline of the target building, obtain the accurate bottom outline of the building, improve the geographical location accuracy of the building, and lay the foundation for high-precision white film production and urban 3D model construction.

[0066] The following is about Figure 1 The method for constructing a 3D city model is explained in detail.

[0067] In some embodiments provided in this application, the top building outline of a target building in satellite imagery can be extracted using a deep learning model. This extraction involves methods such as image vectorization and normalization to obtain a regularly shaped top building outline. Figure 2 As shown. The steps for automatic extraction of the top building outline can be as follows: (1) Obtain high-resolution remote sensing images of the target area taken from multiple angles; (2) Use a building extraction deep learning model to extract the target building in the target area and obtain the top patch of the target building; (3) Vectorize the top patch from the previous step to obtain the top building outline.

[0068] In some embodiments, to optimize the shape of the top building outline, a regularization algorithm can be used to optimize the outline corners through methods such as corner thinning and outlier removal. Specifically, methods such as edge translation and rotation are used to adjust the outline shape, ultimately completing the regularization process of the top building outline and obtaining a regularized top building outline, the effect of which is as follows. Figure 4 As shown, the standardized top building outline can more closely resemble the actual building shape. Building a 3D city model based on the standardized top building outline can improve the reliability of the city 3D model.

[0069] In some embodiments, during execution Figure 1 Prior to the method shown, DSM imagery could be produced based on high-resolution stereo imaging satellite imagery. The DSM imagery production steps are as follows: First, acquire stereo image pairs of remote sensing images of the target area; that is, acquire two satellite remote sensing images corresponding to the target area, with these two images captured from different angles. Next, perform denoising and enhancement operations on the remote sensing images. Then, automatically generate the DSM imagery of the target area based on a binocular stereo matching algorithm. In one embodiment, the DSM imagery can also be cropped according to the extent of the target area to obtain a high-precision DSM imagery corresponding to the target area.

[0070] To ensure the accuracy of the positional correction of the building's top outline and the height accuracy of the target building, DSM imagery can be preprocessed. The resolution and quality of the DSM imagery are checked. If the resolution is inconsistent with the satellite imagery, interpolation is performed to match the resolution of the DSM imagery with the satellite remote sensing imagery. If outliers are found, they are first removed, and then interpolation is used to supplement them. If missing values ​​are found, interpolation is used to supplement them. This ensures the quality of the DSM imagery.

[0071] Pixel values ​​(brightness) in DSM imagery can indicate the height corresponding to that pixel. In one possible scenario, the higher the brightness of a pixel, the higher its corresponding height. By mapping the top building outline to the DSM imagery, the top height information h1 of the target building can be obtained from the pixel values ​​of the DSM imagery.

[0072] In some embodiments, obtaining the top height information of the target building may include: determining the set of pixel coordinate information corresponding to the DSM image based on the set of geographic coordinate information and the set of height information in the DSM image; overlaying the DSM image with the outline of the top building based on the set of pixel coordinate information; and obtaining the top height information based on the height information indicated by the pixels in the DSM image located within the outline of the top building.

[0073] The geographic coordinate information set can be the set of geographic coordinate information contained in DSM imagery. The geographic coordinate information set includes at least the geographic coordinate information corresponding to the target building. In one embodiment, the geographic coordinate information set may include all geographic coordinate information corresponding to the target area where the target building is located.

[0074] The height information set can be the set of height information contained in DSM imagery. The height information set includes at least the height information corresponding to the target building. In one embodiment, the height information set may include all height information corresponding to the target area where the target building is located.

[0075] The pixel coordinate information set can be the set of pixel coordinate information contained in the DSM image. The pixel coordinate information set includes at least the pixel coordinate information corresponding to the target building. In one embodiment, the pixel coordinate information set may include all pixel coordinate information corresponding to the target area where the target building is located.

[0076] To map the building's top outline to the DSM imagery, the pixel coordinates of the building's outline need to be mapped to the geographic coordinates contained in the DSM imagery. This is done by inputting the geographic coordinates and height information sets into a rational polynomial coefficient (RPC) model for orthographic transformation, converting latitude and longitude coordinates to pixel coordinates, thus obtaining the DSM imagery in pixel coordinates. This facilitates the mapping and alignment of the DSM imagery and the building's top outline in pixel coordinates, as shown below. Figure 3As shown. Overlaying the DSM image and the top building outline can refer to overlaying the top building outline itself onto the DSM image, or it can refer to overlaying one or more pixel coordinates indicated by the top building outline onto the DSM image. In one possible scenario, pixels in the DSM image located within the top building outline can indicate the top height information. When the satellite remote sensing image contains multiple buildings, the top building outlines of all buildings can be extracted at once. The RPC model can process the coordinate information of multiple buildings. In another possible scenario, the centroid of the top building outline can be overlaid onto the DSM image, and the pixels in the DSM image corresponding to the centroid of the top building outline can indicate the top height information.

[0077] Suppose there are N pixels in the DSM image located within the outline of a top building. The set of height information indicated by these N pixels can be SET_1. In one possible scenario, the height information corresponding to any one of the N pixels can serve as the top height information h1 of the target building. In another possible scenario, the top centroid pixel coordinates of the top building outline can be obtained, where the top centroid pixel coordinates can be (row_1, col_1). The height information h1 indicated by the top centroid pixel coordinates can serve as the top height information h1 of the target building. In yet another possible scenario, the median of the height information set SET_1 can serve as the top height information h1 of the target building.

[0078] In some embodiments provided in this application, the bottom building outline of the target building can be obtained by offsetting the top building outline. In one possible implementation, the offset of the top building outline can be obtained from the tilt angle corresponding to the satellite remote sensing image and the height of the target building.

[0079] In order to obtain a more accurate bottom building outline, in some embodiments provided in this application, the bottom building outline of the target building can be derived from the top building outline of the target building by means of digital elevation model (DEM) images.

[0080] To ensure the accuracy of the positional correction of the building's top outline and the height accuracy of the target building, the DEM image can be preprocessed. The resolution and quality of the DEM image are checked. If the resolution is inconsistent with the satellite image, interpolation is performed to match the resolution of the DEM image with the satellite remote sensing image. If outliers are found, they are first removed, and then interpolation is used to supplement them. If missing values ​​are found, interpolation is used to supplement them. This ensures the quality of the DEM image.

[0081] In some embodiments, after obtaining the geographic coordinate information corresponding to the top building outline, the digital elevation model (DEM) image corresponding to the target building is overlaid with the geographic coordinate information; the bottom height information h2 of the target building is obtained according to the height information indicated by the pixels of the DEM image corresponding to the geographic coordinate information; and the bottom building outline of the target building is determined according to the bottom height information h2 and the geographic coordinate information.

[0082] In one possible implementation, the geographic coordinates corresponding to the top building outline can be obtained from the pixel coordinates of the top building outline and the top height information h1. The geographic coordinates can be obtained by inputting the pixel coordinates and the top height information h1 into a rational function RPC model for inverse projection transformation.

[0083] Specifically, based on the top height information h1 and the pixel coordinate information (x, y) of the top building outline obtained in step 120, the pixel coordinates are transformed into geographic coordinates by inverse projection transformation using the RPC model, and the geographic coordinate information (lon, lat) of the target building under the DSM image height value h1 can be obtained.

[0084] In another possible implementation, the geographic coordinates (lon, lat) of the target building can be directly obtained from the DSM model. As described above, after overlaying the top building outline with the DSM model, the geographic coordinates of the target building can be obtained based on the correspondence between the top building outline and the DSM model. This implementation is typically used when there is only one building in the DSM model. By precisely cropping the high-precision DSM model, it is easier to obtain the geographic coordinates of the target building.

[0085] The geographic coordinate information (lon,lat) corresponding to the target building obtained through the above method can be correlated with the corresponding target building area on the DEM image. This allows the geographic coordinate information (lon,lat) to be overlaid with the target building in the DEM image. In one possible scenario, the geographic coordinate information (lon,lat) can be a set, including at least the geographic coordinates corresponding to multiple pixel coordinates of the top building outline. These multiple pixel coordinates can be the keypoint coordinates of the top building outline, or they can be the complete set of pixel coordinates of the top building outline. In another possible scenario, the geographic coordinate information (lon,lat) can be a single coordinate, for example, it can be obtained from the top height information h1 and the centroid pixel coordinates of the top building outline.

[0086] Pixel values ​​(brightness) in a DEM image can indicate the altitude corresponding to that pixel, thus indicating the base height of a target building. In one possible scenario, the higher the brightness of a pixel, the higher its corresponding altitude. By mapping geographic coordinates to the DEM image, the base height information h2 of the target building can be obtained from the pixel values ​​of the DEM image.

[0087] In a DEM image, pixels corresponding to geographic coordinates can indicate bottom height information. Assume there are M pixels in the DEM image corresponding to geographic coordinates. Then, the set of height information indicated by these M pixels can be SET_2. In one possible scenario, the height information corresponding to any one of the M pixels can serve as the bottom height information h2 of the target building. In another possible scenario, the centroid geographic coordinates of the geographic coordinate information (lon, lat) can be obtained, where the centroid geographic coordinates can be (center_lon, center_lat). The height information h2 indicated by the centroid geographic coordinates can serve as the bottom height information h2 of the target building. In yet another possible scenario, the median of the height information set SET_2 can serve as the bottom height information h2 of the target building.

[0088] Then, based on the bottom height information h2 and geographic coordinates, the bottom outline of the target building can be determined. For details, please refer to... Figure 5 .

[0089] In one possible implementation, the bottom building outline can be obtained by calculating the offset of the top building outline and then offsetting it. Specifically, the centroid geographic coordinates (center_lon, center_lat) and the bottom height information h2 mentioned above are input into a rational function RPC model for orthographic transformation, that is, the latitude and longitude coordinates are converted into pixel coordinates, which yields the bottom centroid pixel coordinates (row_2, col_2) corresponding to the centroid geographic coordinates (center_lon, center_lat). By comparing the top centroid pixel coordinates (row_1, col_1) and the bottom centroid pixel coordinates (row_2, col_2) mentioned above, the offset of the top building outline can be obtained. Specifically, the horizontal offset of the top building outline Δx = row_2 - row_1; the vertical offset of the top building outline Δy = col_2 - col_1.

[0090] Offsetting the original pixel coordinates (x, y) of the top building outline in pixel coordinates yields the coordinates (x+Δx, y+Δy) of the bottom building outline. In one possible scenario, when satellite remote sensing imagery contains multiple buildings, the offset of the top building outlines of each building can be calculated separately to obtain multiple bottom building outlines corresponding to each building.

[0091] In another possible approach, the geographic coordinate information (lon,lat) and bottom height information h2 mentioned above can be input into the rational function RPC model for orthographic transformation, that is, the latitude and longitude coordinates can be converted into pixel coordinates, and the bottom building outline in pixel coordinates can be obtained.

[0092] Figure 6 The diagram shows the top building outline (solid line) before correction and the bottom building outline (dashed line) after correction. Then, based on the bottom building outline and the target building's top height information h1, a 3D model of the target building can be created. The 3D model construction steps may include fusing the bottom building outline and generating a 3D white membrane. Further accuracy optimization of the building's bottom outline is achieved using outline overlay combined with confidence level methods. Finally, by combining a 3D globe and the open-source map visualization library Cesium, the white membrane production and the construction of the city's 3D model are completed. Figure 7 This paper illustrates a process for constructing a three-dimensional city model according to an embodiment of this application. Figure 8 This paper shows a three-dimensional city model obtained by a three-dimensional city model construction method provided in the embodiments of this application (different gray levels represent different heights).

[0093] The process of fusing bottom building outlines can be as follows: First, obtain multiple bottom building outlines corresponding to the target building from multiple angles; then, overlay the multiple bottom building outlines; next, filter the high-quality bottom building outlines. For overlapping bottom building outlines, compare the confidence levels of the bottom building outlines, retain the bottom building outline with the highest confidence level as the accurate bottom building outline, and delete the remaining bottom building outlines; for non-overlapping bottom building outlines, retain them all; iterate through all bottom building outlines to complete the filtering of bottom building outlines and obtain the fused bottom building outline.

[0094] The generation of a 3D white model of a building can be specifically as follows: First, obtain the fused high-precision bottom building outline; use DSM imagery and DEM imagery to obtain nDSM imagery, and overlay the bottom building outline of the target building with the nDSM imagery. Use the median of the pixel value set within the bottom building outline as the height information of the target building (i.e., the height information of the target building is obtained by subtracting the DSM and DEM); combine the position, shape, height and other information of the bottom building outline of the target building, and use the open-source library Cesium for 3D globe and map visualization to perform 3D target building modeling and rendering, and finally obtain a 3D white model of the target building with high-precision position information, thus completing the rapid construction of a 3D city model.

[0095] This application also provides an electronic device, which includes: one or more processors; one or more memories; the one or more memories storing one or more computer programs, the one or more computer programs including instructions, which, when executed by the one or more processors, cause the electronic device to perform the following actions: Figure 1 The method shown.

[0096] This application presents a method for rapidly constructing high-precision urban 3D models based on high-resolution stereo imaging satellite imagery. This method utilizes DSM imagery production from stereo image pairs based on remote sensing images, building outline extraction, distance and orientation calculation of building outline offsets, multi-angle building outline fusion, and building height acquisition to achieve a complete workflow framework design for high-precision building white-film production and urban 3D model construction. Specifically, the extracted building top outlines undergo automated normalization and offset correction, and multi-angle building outline fusion technology is employed to obtain accurate building bottom outlines, improving the accuracy of building geographic location and laying the foundation for high-precision white-film production and urban 3D model construction.

[0097] Although this application discloses preferred embodiments as described above, it is not intended to limit this application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of this application. Therefore, the scope of protection of this application should be determined by the scope defined in the claims of this application.

Claims

1. A method for constructing a three-dimensional urban model, characterized in that, include: Extract the top outline of the target building from satellite imagery; Based on the digital surface model (DSM) image corresponding to the target building, obtain the top height information of the target building; The method of offsetting the top building outline to obtain the bottom building outline of the target building includes: acquiring the geographic coordinate information corresponding to the top building outline; overlaying the digital elevation model (DEM) image corresponding to the target building with the geographic coordinate information; acquiring the bottom height information of the target building based on the height information indicated by the pixels in the DEM image corresponding to the geographic coordinate information; determining the bottom building outline of the target building based on the bottom height information and the geographic coordinate information; and establishing a three-dimensional model of the target building based on the bottom building outline and the top height information.

2. The method according to claim 1, characterized in that, The step of obtaining the top height information of the target building based on the digital surface model (DSM) image corresponding to the target building includes: Based on the set of geographic coordinate information and the set of height information in the DSM image, determine the set of pixel coordinate information corresponding to the DSM image; Based on the set of pixel coordinate information, the DSM image is overlaid with the outline of the top building; the top height information is obtained based on the height information indicated by the pixels in the DSM image located within the outline of the top building.

3. The method according to claim 2, characterized in that, The step of determining the set of pixel coordinate information corresponding to the DSM image based on the set of geographic coordinate information and the set of height information in the DSM image includes: The set of geographic coordinate information and the set of height information are input into a rational function RPC model for orthographic projection transformation to obtain the set of pixel coordinate information.

4. The method according to claim 2 or 3, characterized in that, The top height information is the median of all height values ​​indicated by pixels located within the top building outline in the DSM image.

5. The method according to claim 1, characterized in that, The step of obtaining the geographic coordinate information corresponding to the top building outline includes: The geographic coordinate information is obtained based on the pixel coordinate information corresponding to the top building outline and the top height information.

6. The method according to claim 5, characterized in that, The step of obtaining the geographic coordinate information based on the pixel coordinate information corresponding to the top building outline and the top height information includes: The pixel coordinate information and the top height information are input into a rational function RPC model for inverse projection transformation to obtain the geographic coordinate information.

7. The method according to any one of claims 1, 5, and 6, characterized in that, The step of obtaining the bottom building outline of the target building based on the bottom height information and the geographic coordinate information includes: determining the centroid geographic coordinates based on the geographic coordinate information; The bottom centroid pixel coordinates are determined based on the centroid geographic coordinates and the bottom height information; The top centroid pixel coordinates are determined based on the pixel coordinate information corresponding to the top building outline. The bottom building outline is obtained by offsetting the top building outline based on the offset between the bottom centroid pixel coordinates and the top centroid pixel coordinates.

8. The method according to claim 7, characterized in that, Determining the bottom centroid pixel coordinates based on the centroid geographic coordinates and the bottom height information includes: The centroid geographic coordinates and the bottom height information are input into the rational function RPC model for orthographic transformation to obtain the bottom centroid pixel coordinates.

9. The method according to claim 1, characterized in that, The bottom height information is the median of all height values ​​indicated by all pixels in the DEM image corresponding to the geographic coordinate information.

10. An electronic device, characterized in that, The electronic device includes: One or more processors; One or more memory units; The one or more memories store one or more computer programs, the one or more computer programs including instructions that, when executed by the one or more processors, cause the electronic device to perform the method as described in any one of claims 1 to 9.

Images