Urban albuginea large-range rapid extraction method and system

Abstract

The invention provides an urban white film large-range rapid extraction method and system, and belongs to the technical field of satellite remote sensing image processing and information extraction. According to the method, the resource No.3 satellite stereo image pair is used as a data source, so that the cost can be reduced, the monitoring range can be expanded, and the development of domestic satellite application can be promoted. In addition, the invention provides a ground feature elevation estimation method based on fusion of digital surface model DSM data and building two-dimensional contour data, probability density estimation and quantile truncation mean value calculation are combined, building roof elevation data and building ground elevation data are accurately extracted, and the difference value between the two elevation data is calculated. And obtaining height data of all buildings. According to the method, the process of separating the terrain height from the building height in a conventional method is not needed, the complicated steps of point cloud editing, filtering, hole interpolation and the like are avoided, the operation is simple, and the efficiency of building height extraction is greatly improved while the operation cost is reduced.

Description

Technical Field

[0001] The invention relates to the technical field of satellite remote sensing image processing and information extraction, and in particular to a method and system for rapid large-scale extraction of urban white film. Background Art

[0002] Urban white film is a basic modeling method in the field of 3D modeling. It mainly refers to a 3D model that only contains the shape and height information of the building, but does not contain texture and detail information. This model has a wide range of application value in urban planning, public management, disaster prevention and mitigation, navigation, and smart city construction. It is of great significance for sustainable development, efficient use of urban resources, and provision of public services to residents.

[0003] The production of urban white film is a complex process involving data collection, processing and three-dimensional modeling. Its core elements include building outline data and building height data. There are various ways to obtain the two-dimensional outline data of the building, which is relatively easy to obtain, while the building height data is more difficult to obtain. How to quickly and accurately obtain building height information has become a hot topic in current research.

[0004] Traditional building height measurement methods, such as outdoor measurement by surveyors, laser radar, and aerial stereo imaging, have the advantages of mature technology and high precision, but they have shortcomings such as high cost, complex processing, and poor timeliness. With the development of remote sensing technology, high-resolution satellite image stereo mapping technology has been rapidly developed and widely used in the field of building height extraction. Compared with manual outdoor measurement, aerial images, and laser radar point cloud data, satellite remote sensing has the advantages of large monitoring area, wide coverage, low data cost, rapid acquisition, and continuous monitoring and updating, and has gradually become an important data source for building height extraction.

[0005] At present, the methods for extracting building heights based on high-resolution satellite images are mainly divided into two categories: shadow method and stereo pair method.

[0006] The shadow-based height estimation method can estimate the building height based on a single remote sensing image according to the length of the building shadow in the high-resolution optical image and its geometric relationship with the sun and the sensor. Commonly used algorithms include edge detection method, threshold segmentation method, etc. Such methods are suitable for tall buildings, large gaps between adjacent buildings, and no obstructions around. When the shadow of the building is obscured or confused with other objects, the estimation accuracy is poor.

[0007] The stereo reconstruction method based on satellite stereo image pairs uses the three-dimensional geographic information contained in the satellite stereo image pairs to estimate the building height. This method is based on the principle of binocular stereo vision and uses kernel line resampling stereo matching and dense point cloud reconstruction to generate DSM (Digital Surface Model) data. The data contains two parts: terrain height and object height. The height of the building can be obtained by further separating the object height. This method is not affected by the surrounding environment of the building. When the buildings are densely distributed and of uniform height, its universality is stronger.

[0008] At present, the method of making urban white film based on satellite stereo image pairs mainly uses sub-meter high-resolution stereo image pairs. First, deep learning technology or machine learning methods for object classification are used to extract buildings, and then combined with manual correction to obtain the two-dimensional contour data of buildings; then based on the principle of stereo vision, the kernel line resampling stereo matching and dense point cloud reconstruction are used to calculate and generate kernel line pair images through rational function models to obtain DSM data, and then use point cloud data filtering algorithms to obtain DEM (Digital Elevation Model) data, and then DSM data and DEM data are superimposed and analyzed, and the difference is calculated to obtain the surface object height layer; finally, the two-dimensional contour data of the building and the surface object height layer are spatially superimposed to obtain the height information of each building spot, and finally realize the three-dimensional model of urban buildings. The core of this technical method is to use the point cloud data filtering algorithm to separate the terrain height from the object height, so as to obtain the height of the building. The separation process involves the classification of point cloud data, the editing and filtering of point cloud data, the filling of holes interpolation, and the image base map assistance.

[0009] It should be noted that DSM contains a model of the altitude information of the tops of all objects on the surface (such as buildings, vegetation, roads, etc.). DSM not only records the ups and downs of the terrain itself (such as hills and valleys), but also contains the height information of the objects. DEM only represents the elevation information of the exposed terrain on the earth's surface (such as mountains, plains, etc.), and does not include the height of objects such as buildings and vegetation (the elevation of the exposed terrain on the earth's surface needs to be generated by subtracting the height of the objects from the DSM).

[0010] At present, the use of sub-meter-level high-resolution satellite stereo image pairs to produce urban white film has the following two main drawbacks: The first is the satellite data source, which usually uses 0.3m-0.5m resolution satellite stereo image pairs (such as Pleiades, Geoeye, worldview-2 and worldview-3) and 0.7m resolution GF-7 satellite stereo image pairs. The advantage of this type of data source is that the higher the resolution, the higher the accuracy of building outline extraction and building height information extraction, which can meet the surveying and mapping requirements of 1:2000 to 1:10000 scale. However, the disadvantages are also very obvious. First, the price of this type of data is very expensive, and the cost is extremely high. Second, the width of this type of data is small, and the width of a single scene does not exceed 20 kilometers, and the monitoring range is small. The second step is to extract building height information, which is also the most critical part. Usually, DSM data is generated using stereo image pair data, and then the terrain height and object height are separated through the point cloud filtering algorithm to obtain DEM data. Finally, the height information of the building is obtained by overlay analysis. The core process involves the classification of point cloud data, editing and filtering of point cloud data, interpolation and filling of holes, and image base map assistance. Although the height information of the building can be extracted relatively accurately, the process involves manual editing, which has difficulties and shortcomings such as high requirements for the professional ability of technicians, low editing efficiency, and low degree of automation. In addition, filtering of point cloud data and interpolation and filling of holes will further increase the computing cost. Summary of the invention

[0011] The purpose of the present invention is to overcome the above-mentioned technical deficiencies and provide a method and system for rapid large-scale extraction of urban white film, so as to solve the problem that in the process of generating urban white film in the related technology, the process of extracting building height information involves manual editing, which has high requirements on the professional ability of technicians, low editing efficiency and low degree of automation.

[0012] In order to achieve the above technical objectives, the present invention adopts the following technical solutions: According to a first aspect of the present invention, a method for rapid large-scale extraction of urban white film is provided, comprising: Step S11, according to the scope of the working area, collecting satellite image data of the Resource-3 satellite stereo image pair covering the working area, wherein the satellite image data at least includes forward-looking image data and rear-looking image data; Step S12, pre-processing the rear-view image data to obtain a satellite rear-view image base map; Step S13, for the rear view image base map, using an object-oriented classification machine learning method to extract building spot vectors, combined with manual visual correction and adjustment, to obtain two-dimensional building contour data; Step S14, using a preset satellite processing module to import the forward image data and the rearward image data, and extract digital surface model DSM data; Step S15, extracting building roof elevation data from the DSM data using a quantile truncation method based on the two-dimensional building profile data; Step S16, extracting building ground elevation data from the DSM data based on the two-dimensional building contour data by combining probability density estimation with the quantile truncated mean method; Step S17, performing a difference operation on the building roof elevation data and the building ground elevation data to obtain height data of all buildings; Step S18: Generate an urban white film based on the two-dimensional outline data of the building and the height data of all buildings.

[0013] Preferably, the step S12 comprises: Step S121, radiation correction processing: using the radiation calibration module and the atmospheric correction module in the remote sensing image processing and analysis software ENVI, the rear-view image data of the Ziyuan-3 satellite stereo image pair are sequentially subjected to radiation calibration and atmospheric correction to eliminate the influence of atmospheric scattering and absorption on radiation; Step S122, orthorectification processing: using the orthorectification module in the remote sensing image processing and analysis software ENVI, orthorectify the rearview image that has been processed by the radiation correction in step S121 to eliminate the influence of geometric distortion caused by the terrain undulation and obtain the satellite rearview image base map.

[0014] Preferably, the step S13 comprises: Step S131, for the satellite rearview image base map, using the supervised classification module in the remote sensing image processing and analysis software ENVI, select an appropriate amount of building samples, adopt an object-oriented classification method, classify and extract the building spots, and obtain the building spot vector; Step S132, combining the satellite rear-view image base map, in the geographic information system software ArcGIS Pro software, using the method of manual visual correction, manually edit and correct the building patch vectors for missed or erroneous patches, and obtain the two-dimensional contour data of the building.

[0015] Preferably, the step S14 comprises: Step S141, importing satellite stereo data: using the SAT Master 10.0 satellite processing module in the photogrammetry and remote sensing data processing software Trimble InphoPhotogrammetry 10.0, by at least including the steps of creating a project, importing the front view image data and the rear view image data, importing RPC information, creating a scene, and checking the project, the import of satellite stereo image pair data is completed; Step S142, project image preprocessing: open the project created in step S141, perform at least image pyramid creation and image radiation correction processing operations, and complete the project processing settings; Step S143, aerial triangulation processing: using the MATCH-AT satellite triangulation module in the Trimble Inpho Photogrammetry 10.0 software, automatically identify the image points with the same name in multiple satellite stereo image pairs, and extract these points as the basic control data for aerial triangulation; and using the results of aerial triangulation, iteratively optimize the strict geometric model parameters RPCs of the satellite image to improve the geographic positioning accuracy of the image; Step S144, generate DSM data: use the Match-T DSM module in the Trimble Inpho Photogrammetry 10.0 software, use the corrected RPCs, and combine the same-name image points matched by multi-view images to generate DSM data including the building roof height.

[0016] Preferably, the step S15 comprises: Step S151, according to the two-dimensional outline data of the building, based on the geographic information system software ArcGIS Pro software, the pixel-by-pixel elevation value of the area where each building is located is obtained from the DSM data; Step S152: Use the quantile truncation method to sort the DSM elevation data in the building patch area from small to large, and select the elevation value at the 98% quantile as the roof elevation to effectively remove the noise and abnormal elevation values ​​that may exist in the DSM data.

[0017] Preferably, the step S151 is specifically as follows: If the building pattern Included The corresponding elevation value is , the elevation value extraction formula is: , where 1≤i≤n, are the coordinates of each pixel in the DSM; The step S152 is specifically as follows: The sorted elevation values ​​are recorded as , define the score as: , where p represents the quantile ratio and the value range is ; Take the 98% quantile elevation value Indicates the roof elevation: .

[0018] Preferably, the step S16 comprises: Step S161: Expand each building spot by creating a 20-meter buffer zone according to the two-dimensional building contour data. Within this range, extract the DSM elevation data of the building spot and its surrounding area to ensure complete coverage of the terrain changes around the building. Step S162: The DSM data extracted in the buffer are analyzed by a probability density estimation method to construct a distribution model of the elevation data, and the quantile truncated mean method is used to optimize the data.

[0019] Preferably, the step S161 is specifically as follows: If the building pattern The buffer is , the buffer radius is d, and its geometric expression is: ; Among them, (x, y) represent the buffer zone The horizontal and vertical coordinate values ​​of any pixel in the image coordinate system, (x r ,y r ) represents the building pattern The horizontal and vertical coordinate values ​​of the center point pixel in the image coordinate system; Extract elevation data sets within the buffer area : ; Among them, z i Representing the buffer The image coordinates are (x i ,y i ) pixel points correspond to the elevation values ​​in the DSM data; In step S162, the quantile truncated mean method is used to optimize the data, including: arranging all the extracted elevation values ​​in ascending order, selecting the elevation values ​​in the 5%-10% interval for mean calculation, and using this as the building ground elevation, specifically: Elevation values ​​extracted from the buffer , sorted from small to large, recorded as ; Select the elevation value in the 5%-10% range , calculate the mean as the ground elevation value: ,in, - +1; Using probability density estimation Construct a distribution model for elevation data: ; in, is the kernel function, the Gaussian kernel is commonly used: , The bandwidth determines the smoothness of the distribution and is determined based on historical experience or experimental data.

[0020] Preferably, the step S17 is specifically as follows: In the geographic information system software ArcGIS Pro, the difference calculation is performed between the value corresponding to the building roof elevation field and the value corresponding to the building ground elevation field in the building map vector data attribute table to obtain the height data of the building map.

[0021] According to a second aspect of the present invention, a system for rapid large-scale extraction of urban white film is provided, comprising: A collection module, used for collecting satellite image data of the Resource-3 satellite stereo image pair covering the working area according to the working area range, wherein the satellite image data at least includes forward-looking image data and rear-looking image data; A processing module, used for pre-processing the rear-view image data to obtain a satellite rear-view image base map; It is also used to extract building spot vectors from the rear view image base map using an object-oriented classification machine learning method, and obtain two-dimensional building contour data in combination with manual visual correction and adjustment; An extraction module is used to use a preset satellite processing module to import the forward image data and the rearward image data, and extract digital surface model DSM data; It is also used to extract building roof elevation data from DSM data using a quantile truncation method based on the two-dimensional building contour data; It is also used to extract building ground elevation data from DSM data based on the two-dimensional building contour data by combining probability density estimation with quantile truncated mean method; A calculation module, used for performing difference calculation on the building roof elevation data and the building ground elevation data to obtain the height data of all buildings; The generating module is used to generate the urban white film based on the two-dimensional outline data of the building and the height data of all the buildings.

[0022] The technical solution provided by the embodiments of the present invention may have the following beneficial effects: By adopting the stereo image pair of the Ziyuan-3 satellite as the data source, it is the first civilian high-resolution stereo mapping satellite independently designed and launched by my country, with a resolution of 2.1m and a width of 52km. It has the advantages of low acquisition cost, large data collection volume and wide coverage. It can meet the surveying and mapping requirements of 1:50,000 scale and fully meet the three-dimensional modeling of urban white film level. The use of this data can not only reduce costs and expand the monitoring scope, but also promote the development of domestic satellite applications.

[0023] In addition, the present invention also proposes a method for estimating the elevation of land objects based on the fusion of digital surface model DSM data and building two-dimensional contour data, combining probability density estimation with quantile truncated mean calculation, accurately extracting building roof elevation data and building ground elevation data, and calculating the difference between the two to obtain the height data of all buildings. This method does not require the process of separating terrain height from building height in conventional methods, avoiding complicated steps such as point cloud editing, filtering, and hole interpolation, and is simple to operate. While reducing computing costs, it also greatly improves the efficiency of building height extraction. This method has a high level of automation and can further realize the automated production of urban white film. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 is a flow chart of a method for rapid large-scale extraction of urban white film according to an exemplary embodiment; Figure 2 is a schematic block diagram of a system for rapid large-scale extraction of urban white film according to an exemplary embodiment; Figure 3 is a diagram showing the effect of generating a white film in a city according to an exemplary embodiment; Figure 4 The present invention is a flow chart showing a method for rapid large-scale extraction of urban white film according to another exemplary embodiment. DETAILED DESCRIPTION

[0025] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work should fall within the scope of protection of the present application.

[0026] Embodiment 1 Figure 1 is a flowchart of a method for rapid large-scale extraction of urban white film according to an exemplary embodiment, see Figure 1 , the method comprising: Step S11, according to the scope of the working area, collecting satellite image data of the Resource-3 satellite stereo image pair covering the working area, wherein the satellite image data at least includes forward-looking image data and rear-looking image data; Step S12, pre-processing the rear-view image data to obtain a satellite rear-view image base map; Step S13, for the rear view image base map, using an object-oriented classification machine learning method to extract building spot vectors, combined with manual visual correction and adjustment, to obtain two-dimensional building contour data; Step S14, using a preset satellite processing module to import the forward image data and the rearward image data, and extract digital surface model DSM data; Step S15, extracting building roof elevation data from the DSM data using a quantile truncation method based on the two-dimensional building profile data; Step S16, extracting building ground elevation data from the DSM data based on the two-dimensional building contour data by combining probability density estimation with the quantile truncated mean method; Step S17, performing a difference operation on the building roof elevation data and the building ground elevation data to obtain height data of all buildings; Step S18: Generate an urban white film based on the two-dimensional outline data of the building and the height data of all buildings.

[0027] It should be noted that the technical solution provided in this embodiment is loaded and run in an electronic device in specific practice, and the electronic device includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in this embodiment. The electronic device includes but is not limited to: a client computer, and / or a local server, and / or a cloud server.

[0028] It should be noted that the Ziyuan-3 satellite is equipped with three cameras: forward-looking, front-looking, and rear-looking, forming a multi-angle stereoscopic observation capability: Orthographic image: taken vertically downward, used for extracting basic geographic information (such as orthophoto).

[0029] Front / rear view images: Oblique shooting, used to construct stereo image pairs, and generate high-precision DSM (digital surface model) and terrain elevation data through multi-view image matching.

[0030] In the stereo image pair of Ziyuan-3 satellite, the forward-looking image data and the rear-looking image data are image data of different perspectives acquired based on the satellite multi-angle imaging technology. Their specific meanings are as follows: 1. Forward image data Definition: A forward-looking image is an image taken by a satellite obliquely from the side and front in the forward flight direction (i.e. the direction of the satellite's orbit), which is usually related to the tilt angle of the forward-looking camera (e.g., about 22°).

[0031] Function: Provides the tilted viewing angle information of surface objects in the direction of satellite travel. Combined with the forward image and the orthographic image (vertically downward), it can form the geometric relationship of a stereo image pair, which is used to calculate the elevation information of ground points.

[0032] 2. Rear view image data Definition: A rearview image is an image taken from the side and rear of a satellite in the direction of its backward flight (i.e., the direction in which the satellite orbit retreats), usually related to the tilt angle of the rearview camera (e.g., about 22°).

[0033] Function: Provides the oblique viewing angle information of the surface objects in the direction of flight behind the satellite, and combines it with the rearview image and the frontview image to further supplement the geometric constraints of the stereo image pair and enhance the accuracy and reliability of 3D reconstruction.

[0034] It can be understood that the technical solution provided in this embodiment uses the stereo image pair of the Ziyuan-3 satellite as a data source. It is the first civilian high-resolution stereo mapping satellite independently designed and launched by my country, with a resolution of 2.1m and a width of 52km. It has the advantages of low acquisition cost, large data collection volume, and wide coverage. It can meet the surveying and mapping requirements of 1:50,000 scale and fully meet the three-dimensional modeling of urban white film level. The use of this data can not only reduce costs and expand the monitoring scope, but also promote the development of domestic satellite applications.

[0035] In addition, the technical solution provided in this embodiment also proposes a method for estimating the elevation of land objects based on the fusion of digital surface model DSM data and building two-dimensional contour data, combining probability density estimation with quantile truncated mean calculation, accurately extracting building roof elevation data and building ground elevation data, and calculating the difference between the two to obtain the height data of all buildings. This method does not require the process of separating terrain height from building height in conventional methods, avoiding complicated steps such as point cloud editing, filtering, and hole interpolation, and is simple to operate. While reducing computing costs, it also greatly improves the efficiency of building height extraction. This method has a high level of automation and can further realize the automated production of urban white film.

[0036] In specific practice, in step S12, based on previous data processing experience, the DSM data produced using the ZY-3 satellite stereo image pair is consistent with the rear-view image of the stereo image pair. Therefore, in specific practice, only the rear-view image data is preprocessed, including radiation correction, orthorectification and other steps, to obtain the satellite rear-view image base map, which is used as the base map for subsequent extraction of building two-dimensional contour data.

[0037] Therefore, see Figure 4 , the step S12 comprises: Step S121, radiation correction processing: using the radiation calibration module and the atmospheric correction module in the remote sensing image processing and analysis software ENVI, the rear-view image data of the Ziyuan-3 satellite stereo image pair are sequentially subjected to radiation calibration and atmospheric correction to eliminate the influence of atmospheric scattering and absorption on radiation; Step S122, orthorectification processing: using the orthorectification module in the remote sensing image processing and analysis software ENVI, orthorectify the rearview image that has been processed by the radiation correction in step S121 to eliminate the influence of geometric distortion caused by the terrain undulation and obtain the satellite rearview image base map.

[0038] It should be noted that ENVI (The Environment for Visualizing Images) is a professional remote sensing image processing and analysis software, which is widely used in geographic information systems (GIS), environmental monitoring, agriculture, forestry, urban planning, disaster assessment and other fields. It provides a complete set of tools for extracting information and analyzing images obtained from satellites, aviation or drones.

[0039] ENVI supports the reading, correction, fusion and mosaicking of various sensor data (such as Landsat, Sentinel, MODIS, Ziyuan-3 and other optical satellites, as well as SAR radar data). It provides pre-processing functions such as radiation calibration, atmospheric correction, and geometric correction to ensure data quality.

[0040] In practice, see Figure 4 , the step S13 comprises: Step S131, for the satellite rearview image base map, using the supervised classification module in the remote sensing image processing and analysis software ENVI, select an appropriate amount of building samples, adopt an object-oriented classification method, classify and extract the building spots, and obtain the building spot vector; Step S132, combining the satellite rear-view image base map, in the geographic information system software ArcGIS Pro software, adopting the method of manual visual correction, manually edit and correct the missed and erroneous building spot vectors, and obtain the two-dimensional building outline data (i.e., building spot vector data).

[0041] It should be noted that ArcGIS Pro is a new generation of professional geographic information system (GIS) software developed by Esri, which integrates map making, spatial data analysis, 3D visualization, data management and other functions, and is widely used in urban planning, natural resource management, environmental protection, emergency response, transportation and logistics, etc. It is the core desktop tool of the ArcGIS ecosystem, supporting the whole process from simple map drawing to complex spatial modeling.

[0042] In practice, see Figure 4 , the step S14 comprises: Step S141, importing satellite stereo data: using the SAT Master 10.0 satellite processing module in the photogrammetry and remote sensing data processing software Trimble InphoPhotogrammetry 10.0, by at least including the steps of creating a project, importing the front view image data and the rear view image data, importing RPC information, creating a scene, and checking the project, the import of satellite stereo image pair data is completed; Step S142, project image preprocessing: open the project created in step S141, perform at least image pyramid creation and image radiation correction processing operations, and complete the project processing settings; Step S143, aerial triangulation processing: using the MATCH-AT satellite triangulation module in the Trimble Inpho Photogrammetry 10.0 software, automatically identify the image points with the same name in multiple satellite stereo image pairs, and extract these points as the basic control data for aerial triangulation; and using the results of aerial triangulation, iteratively optimize the strict geometric model parameters RPCs of the satellite image to improve the geographic positioning accuracy of the image; Step S144, generate DSM data: use the Match-T DSM module in the Trimble Inpho Photogrammetry 10.0 software, use the corrected RPCs, and combine the same-name image points matched by multi-view images to generate DSM data including the building roof height.

[0043] It should be noted that Trimble Inpho Photogrammetry 10.0 is a professional photogrammetry and remote sensing data processing software developed by Trimble. It focuses on extracting high-precision geographic information from aerial, satellite or drone images, and generating terrain models, orthophotos (DOM), three-dimensional maps and other results. It is widely used in surveying and mapping, engineering planning, natural resource management, disaster assessment and other fields, and is particularly suitable for processing large-scale high-resolution image data. The software generates high-resolution DSM (including the height of all objects on the surface) and DEM (exposed terrain elevation) data based on image matching results. The software supports filtering and classification to separate ground points from non-ground points (such as buildings and vegetation).

[0044] In step S142, completing the project processing settings (such as image pyramid creation, radiation correction, etc.) is the key basis for the efficient operation of subsequent processes and the generation of high-precision DSM data.

[0045] 1. The role of image pyramid creation Accelerated Image Processing: The image pyramid stores data in multiple resolutions in hierarchical form (e.g., from low resolution to high resolution), and in subsequent processing (such as matching and adjustment), low-resolution images can be quickly called for rough calculations, reducing the amount of computation and improving efficiency.

[0046] Support for Large-Scale Data Management: For a large amount of satellite images (such as the multi-scene mosaicked data of Ziyuan-3), the pyramid structure can effectively manage memory and storage resources, avoiding performance bottlenecks when directly processing the original high-resolution images.

[0047] Adaptation to Multi-Scale Analysis: In aerial triangulation (step S143) and DSM generation (step S144), different stages may require image data of different resolutions (e.g., low resolution for rough matching and high resolution for fine matching), and the pyramid structure can flexibly support this.

[0048] 2. The Role of Image Radiometric Calibration Eliminating Sensor Noise and Distortion: Satellite sensors may introduce noise or radiometric distortion (such as vignetting) during the imaging process due to factors such as illumination and atmospheric scattering. Radiometric calibration adjusts the image brightness and contrast to ensure radiometric consistency among multiple scenes of images and avoid matching errors.

[0049] Improving Image Matching Accuracy: Without radiometric calibration, it may be difficult to match homologous points between different images due to terrain shadows, cloud cover, or differences in sensor responses. After calibration, the texture features of the images are clearer, enhancing the robustness of matching algorithms (such as feature point extraction).

[0050] Supporting Multi-Temporal Data Fusion: In scenarios such as disaster monitoring that require comparing images of different times, radiometric calibration can eliminate radiometric changes caused by time differences and improve the reliability of change detection.

[0051] Aerial Triangulation in Step S143: The MATCH-AT module relies on the preprocessed image pyramid to quickly extract tie points and optimize the RPCs parameters. If the pyramid is not created and the original images are directly processed, it may lead to excessive computation time or even memory overflow.

[0052] DSM Generation in Step S144: The Match-T DSM module requires radiometrically consistent images as input; otherwise, the generated DSM may have noise such as stripes and spots, affecting the extraction of terrain details.

[0053] RPCs are rational polynomial coefficients provided by satellite manufacturers and are used to convert pixel / row-column coordinates to ground coordinates (latitude, longitude, and elevation).

[0054] Based on the control points (tie points and adjusted exterior orientation elements) obtained through aerial triangulation, the errors of the RPCs are inversely calculated and the model parameters are corrected.

[0055] The process of step S143 is: automatically extract connection points→generate initial matching point data→optimize RPCs→improve global geometric accuracy.

[0056] Function: Eliminate the systematic errors of the original RPCs (such as sensor distortion and positioning deviation caused by terrain undulation). If only relying on the original RPCs, large positioning errors may occur in mountainous areas or urban high-rise areas. By optimizing RPCs by connecting points, error transmission can be significantly reduced, ensuring that the generated DSM / DEM data is more authentic and reliable. Step S143 can improve the accuracy of subsequent DSM / DEM generation, especially in complex terrain or large areas.

[0057] The essence of the process from step S143 to S144 is to apply the encrypted geometric correction results of aerial triangulation to DSM generation. The core steps include: exporting aerial triangulation results (RPCs, exterior orientation elements); importing them into the DSM module and configuring parameters; performing dense matching to generate DSM. This process reflects the full process automation capability of Trimble Inpho software. Users only need to ensure the quality of aerial triangulation results to efficiently generate high-precision DSM.

[0058] In practice, see Figure 4 , the step S15 comprises: Step S151, according to the two-dimensional outline data of the building, based on the geographic information system software ArcGIS Pro software, the pixel-by-pixel elevation value of the area where each building is located is obtained from the DSM data; Step S152: Use the quantile truncation method to sort the DSM elevation data in the building patch area from small to large, and select the elevation value at the 98% quantile as the roof elevation to effectively remove the noise and abnormal elevation values ​​that may exist in the DSM data.

[0059] In specific practice, the step S151 is specifically as follows: If the building pattern Included The corresponding elevation value is , the elevation value extraction formula is: , where 1≤i≤n, are the coordinates of each pixel in the DSM; The step S152 is specifically as follows: The sorted elevation values ​​are recorded as , define the score as: , where p represents the quantile ratio and the value range is ; Take the 98% quantile elevation value Indicates the roof elevation: 。

[0060] In specific practice, referring to Figure 4 , the step S16 includes: Step S161: According to the two-dimensional contour data of the building, expand each building patch by creating a 20-meter buffer (expand the analysis range to avoid the influence of the building edge on the ground elevation estimation). Within this range, extract the DSM elevation data of the building patch and its surrounding area to ensure complete coverage of the terrain changes around the building; Step S162: Analyze the DSM data extracted in the buffer through the probability density estimation method, construct the distribution model of the elevation data, and optimize the data using the trimmed mean method of quantiles.

[0061] In specific practice, the step S161 is specifically: If the buffer of the building patch is , and the buffer radius is d, its geometric expression is: ; where (x, y) respectively represent the abscissa and ordinate values of any pixel in the buffer in the image coordinate system, and (x r , y r ) represent the abscissa and ordinate values of the central point pixel of the building patch in the image coordinate system; Extract the elevation data set within the buffer range: ; where z i represents the pixel point with the image coordinates (x , y i , y i ) in the buffer, and the corresponding elevation value in the DSM data; In the step S162, the trimmed mean method of quantiles is used to optimize the data, including: arranging all the extracted elevation values in ascending order, selecting the elevation values in the 5%-10% interval for mean calculation, and using this as the building ground elevation. Specifically: For the elevation values extracted in the buffer, sort them in ascending order and record them as ; select the elevation values in the 5%-10% interval, and calculate the mean as the ground elevation value: , where - +1; Using probability density estimation Construct a distribution model for elevation data: ; in, is the kernel function, the Gaussian kernel is commonly used: , The bandwidth determines the smoothness of the distribution and is determined based on historical experience or experimental data.

[0062] In specific practice, the step S17 is specifically as follows: In the geographic information system software ArcGIS Pro, the difference calculation is performed between the value corresponding to the building roof elevation field and the value corresponding to the building ground elevation field in the building spot vector data attribute table to obtain the height data of the building spot.

[0063] In summary, the technical solution provided in this embodiment aims to address the problems existing in the prior art in the production process of white film of urban buildings, such as high cost and small monitoring range caused by the expensive data source and narrow width of sub-meter stereo image pairs, as well as the high computing cost, low production efficiency, low automation level, and high professional ability requirements caused by the complicated steps of separating terrain height and object height in the process of extracting building height information. A method for rapid extraction of urban white film based on the stereo image pair of the Resource-3 satellite is proposed, which can reduce costs, improve efficiency, be simple to operate, have high precision and automation level, and can realize large-scale automated and rapid production of urban-level white film.

[0064] Embodiment 2 Figure 2 is a schematic block diagram of a system 100 for rapid extraction of urban white film over a large area according to an exemplary embodiment, see Figure 2 , the system 100 comprises: The acquisition module 101 is used to acquire satellite image data of the Resource-3 satellite stereo image pair covering the working area according to the working area range, wherein the satellite image data at least includes forward-looking image data and rear-looking image data; The processing module 102 is used to pre-process the rear-view image data to obtain a satellite rear-view image base map; It is also used to extract building spot vectors from the rear view image base map using an object-oriented classification machine learning method, and obtain two-dimensional building contour data in combination with manual visual correction and adjustment; An extraction module 103 is used to use a preset satellite processing module to import the forward-looking image data and the rearward-looking image data, and extract digital surface model DSM data; It is also used to extract building roof elevation data from DSM data using a quantile truncation method based on the two-dimensional building contour data; It is also used to extract building ground elevation data from DSM data based on the two-dimensional building contour data by combining probability density estimation with quantile truncated mean method; The calculation module 104 is used to perform a difference operation on the building roof elevation data and the building ground elevation data to obtain the height data of all buildings; The generating module 105 is used to generate the urban white film based on the two-dimensional outline data of the building and the height data of all the buildings.

[0065] It should be noted that the technical solution provided in this embodiment is loaded and run in an electronic device in specific practice, and the electronic device includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described in this embodiment. The electronic device includes but is not limited to: a client computer, and / or a local server, and / or a cloud server.

[0066] The implementation methods of the above modules refer to the implementation methods of the corresponding steps in Example 1, which will not be repeated in this example.

[0067] It can be understood that the technical solution provided in this embodiment uses the stereo image pair of the Ziyuan-3 satellite as a data source. It is the first civilian high-resolution stereo mapping satellite independently designed and launched by my country, with a resolution of 2.1m and a width of 52km. It has the advantages of low acquisition cost, large data collection volume, and wide coverage. It can meet the surveying and mapping requirements of 1:50,000 scale and fully meet the three-dimensional modeling of urban white film level. The use of this data can not only reduce costs and expand the monitoring scope, but also promote the development of domestic satellite applications.

[0068] In addition, the technical solution provided in this embodiment proposes a method for estimating the elevation of land objects based on the fusion of digital surface model DSM data and building two-dimensional contour data, combining probability density estimation with quantile truncated mean calculation, accurately extracting building roof elevation data and building ground elevation data, and calculating the difference between the two to obtain the height data of all buildings. This method does not require the process of separating terrain height from building height in conventional methods, avoiding complicated steps such as point cloud editing, filtering, and hole interpolation. It is simple to operate, reduces computing costs, and greatly improves the efficiency of building height extraction. This method has a high level of automation and can further realize the automated production of urban white film.

[0069] In specific practice, in order to verify the feasibility of the method and system for rapid extraction of urban white membranes provided by the present invention, the following takes the urban white membrane production case in Jingtai County, Gansu Province as an example to illustrate the method for rapid extraction of urban white membranes based on the stereo image pair of Ziyuan-3 satellite provided in the first embodiment.

[0070] The working area is located in the urban area of Jingtai County, Gansu Province, where the building distribution is relatively dense. According to the production requirements of the project, it is necessary to produce urban white membranes that meet the scale of 1:50,000 for 3D simulation applications. Considering the project cost, archived data situation and accuracy requirements, the method of the present invention is adopted, and the stereo image pair data of Ziyuan-3 satellite is used to extract the building patch vectors and building height information of the working area, so as to establish the urban white membrane of the working area.

[0071] First, use the radiation correction module and orthorectification module of ENVI software to perform preprocessing such as radiometric calibration, atmospheric correction, and orthorectification on the rear-view image of the Ziyuan-3 stereo image pair in sequence; Secondly, use the supervised classification module of ENVI software, and adopt the object-oriented classification method, specifically the Support Vector Machine (SVM). Select an appropriate amount of building sample data to classify and extract the bottom map of the rear-view image, and output the building patch vectors; Then, in the ArcGIS Pro software, import the rear-view image and the building patch vectors extracted by ENVI, and use the manual visual method to check the overlapping effect of the two. For the cases of missed extraction and mis-extraction of building patches, perform manual editing and correction to obtain accurate building patch vector data; Next, use the SATMaster 10.0 satellite processing module in the Trimble Inpho Photogrammetry 10.0 software to process the Ziyuan-3 satellite stereo image pair to obtain the digital surface model (DSM) data of the working area; Subsequently, the work of extracting building height data was carried out, which mainly included the extraction of building roof elevation data and building ground elevation data: (1) The quantile truncation method was used to extract building roof elevation data. Specifically, based on the DSM data and building patch vector data, the DSM elevation values ​​corresponding to all pixels within each building patch were arranged from small to large, and the value at 98% was taken as the roof elevation value of the building, thereby obtaining the building roof elevation data; (2) The probability density estimation combined with the quantile truncation mean method was used to extract building ground elevation data. Specifically, the building patch vector was first expanded with a 20-meter buffer zone, and then the DSM elevation values ​​corresponding to all pixels within each building patch after expansion were arranged from small to large, and the DSM average value of pixels in the 5%-10% range was taken as the ground elevation value of the building. The building patch vector was then connected with its buffer zone vector using its attributes to obtain the building ground elevation data.

[0072] Again, in ArcGIS Pro software, raster calculation is performed on the attribute table of the building map vector data. The building roof elevation (Elevation field) is subtracted from the building ground elevation field (Ground_Ele field) to obtain the building height data (Height field). The results are shown in Table 1:

[0073] Table 1 Attribute structure table of building pattern vector data (partial pattern) The accuracy of the building height data extraction results in the present invention was verified by using the actual measured building point heights of the project. The verification results show that the building height data extracted by the present invention has an average absolute accuracy of 2.93 meters, which meets the requirements of 1:50,000 scale surveying and mapping. Details are shown in Table 2.

[0074]

[0075] Table 2 Accuracy verification table of building height extraction results of the present invention Finally, in GIS software or 3D simulation software, the urban white film can be generated using the building pattern vector data and its corresponding height data field. In this case, the urban white film is generated in ArcGIS Pro, and the effect is as follows: Figure 3 shown.

[0076] The results show that the method for rapid extraction of urban white film based on the stereo image pair of the Resource-3 satellite proposed in the present invention can produce urban white film on a large scale, at low cost, quickly and efficiently, and is simple to operate, with high precision and automation level. It can further realize the large-scale automated rapid production of urban-level building white film and has high practical application value.

[0077] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0078] If the integrated units in the above embodiments are implemented in the form of software functional units and sold or used as independent products, they can be stored in the above computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or all or part of the technical solution can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling one or more computer devices (which may be personal computers, servers or network devices, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application.

[0079] In the above embodiments of the present application, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, please refer to the relevant description of other embodiments.

[0080] In the several embodiments provided in the present application, it should be understood that the disclosed client can be implemented in other ways. Among them, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation. For example, multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of units or modules, which can be electrical or other forms.

[0081] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0082] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0083] The above is only a preferred implementation of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A large-scale rapid extraction method of urban white film, characterized in that: include: Step S11, according to the scope of the working area, collecting satellite image data of the Resource-3 satellite stereo image pair covering the working area, wherein the satellite image data at least includes forward-looking image data and rear-looking image data; Step S12, pre-processing the rear-view image data to obtain a satellite rear-view image base map; Step S13, for the rear view image base map, using an object-oriented classification machine learning method to extract building spot vectors, combined with manual visual correction and adjustment, to obtain two-dimensional building contour data; Step S14, using a preset satellite processing module to import the forward image data and the rearward image data, and extract digital surface model DSM data; Step S15, extracting building roof elevation data from the DSM data using a quantile truncation method based on the two-dimensional building profile data; Step S16, extracting building ground elevation data from the DSM data based on the two-dimensional building contour data by combining probability density estimation with the quantile truncated mean method; Step S17, performing a difference operation on the building roof elevation data and the building ground elevation data to obtain height data of all buildings; Step S18: Generate an urban white film based on the two-dimensional outline data of the building and the height data of all buildings.

2. The method according to claim 1, characterized in that The step S12 comprises: Step S121, radiation correction processing: using the radiation calibration module and the atmospheric correction module in the remote sensing image processing and analysis software ENVI, the rear-view image data of the Ziyuan-3 satellite stereo image pair are sequentially subjected to radiation calibration and atmospheric correction to eliminate the influence of atmospheric scattering and absorption on radiation; Step S122, orthorectification processing: using the orthorectification module in the remote sensing image processing and analysis software ENVI, orthorectify the rearview image that has been processed by the radiation correction in step S121 to eliminate the influence of geometric distortion caused by the terrain undulation and obtain the satellite rearview image base map.

3. The method according to claim 1, characterized in that The step S13 comprises: Step S131, for the satellite rearview image base map, using the supervised classification module in the remote sensing image processing and analysis software ENVI, select an appropriate amount of building samples, adopt an object-oriented classification method, classify and extract the building spots, and obtain the building spot vector; Step S132, combining the satellite rear-view image base map, in the geographic information system software ArcGIS Pro software, using the method of manual visual correction, manually edit and correct the building patch vectors for missed or erroneous patches, and obtain the two-dimensional contour data of the building.

4. The method according to claim 1, characterized in that The step S14 comprises: Step S141, importing satellite stereo data: using the SAT Master 10.0 satellite processing module in the photogrammetry and remote sensing data processing software Trimble InphoPhotogrammetry 10.0, by at least including the steps of creating a project, importing the front view image data and the rear view image data, importing RPC information, creating a scene, and checking the project, the import of satellite stereo image pair data is completed; Step S142, project image preprocessing: open the project created in step S141, perform at least image pyramid creation and image radiation correction processing operations, and complete the project processing settings; Step S143, aerial triangulation processing: using the MATCH-AT satellite triangulation module in the Trimble Inpho Photogrammetry 10.0 software, automatically identify the image points with the same name in multiple satellite stereo image pairs, and extract these points as the basic control data for aerial triangulation; and using the results of aerial triangulation, iteratively optimize the strict geometric model parameters RPCs of the satellite image to improve the geographic positioning accuracy of the image; Step S144, generate DSM data: use the Match-TDSM module in Trimble Inpho Photogrammetry 10.0 software, use the corrected RPCs, and combine the same-name image points matched by multi-view images to generate DSM data including the building roof height.

5. The method according to claim 1, characterized in that The step S15 comprises: Step S151, according to the two-dimensional outline data of the building, based on the geographic information system software ArcGIS Pro software, the pixel-by-pixel elevation value of the area where each building is located is obtained from the DSM data; Step S152: Use the quantile truncation method to sort the DSM elevation data in the building patch area from small to large, and select the elevation value at the 98% quantile as the roof elevation to effectively remove the noise and abnormal elevation values ​​that may exist in the DSM data.

6. The method according to claim 5, characterized in that The step S151 is specifically as follows: If the building pattern Included The corresponding elevation value is , the elevation value extraction formula is: , where 1≤i≤n, are the coordinates of each pixel in the DSM; The step S152 is specifically as follows: The sorted elevation values ​​are recorded as , define the score as: , where p represents the quantile ratio and the value range is ; Take the 98% quantile elevation value Indicates the roof elevation: .

7. The method according to claim 1, characterized in that The step S16 comprises: Step S161: Expand each building spot by creating a 20-meter buffer zone according to the two-dimensional building contour data. Within this range, extract the DSM elevation data of the building spot and its surrounding area to ensure complete coverage of the terrain changes around the building. Step S162: The DSM data extracted in the buffer are analyzed by a probability density estimation method to construct a distribution model of the elevation data, and the quantile truncated mean method is used to optimize the data.

8. The method according to claim 7, characterized in that The step S161 is specifically as follows: If the building pattern The buffer is , the buffer radius is d, and its geometric expression is: ; Among them, (x, y) represent the buffer zone The horizontal and vertical coordinate values ​​of any pixel in the image coordinate system, (x r ,y r ) represents the building pattern The horizontal and vertical coordinate values ​​of the center point pixel in the image coordinate system; Extract elevation data sets within the buffer area : ; Among them, z i Representing the buffer The image coordinates are (x i ,y i ) pixel points, corresponding to the elevation values ​​in the DSM data; In step S162, the quantile truncated mean method is used to optimize the data, including: arranging all the extracted elevation values ​​in ascending order, selecting the elevation values ​​in the 5%-10% interval for mean calculation, and using this as the building ground elevation, specifically: Elevation values ​​extracted from the buffer , sorted from small to large, recorded as ; Select the elevation value in the 5%-10% range , calculate the mean as the ground elevation value: ,in, - +1; Using probability density estimation Construct a distribution model for elevation data: ; in, is the kernel function, the Gaussian kernel is commonly used: , The bandwidth determines the smoothness of the distribution and is determined based on historical experience or experimental data.

9. The method according to claim 1, characterized in that: The step S17 is specifically as follows: In the geographic information system software ArcGIS Pro, the difference calculation is performed between the value corresponding to the building roof elevation field and the value corresponding to the building ground elevation field in the building map vector data attribute table to obtain the height data of the building map.

10. A system for rapid large-scale extraction of urban white film, characterized in that: include: A collection module, used for collecting satellite image data of the Resource-3 satellite stereo image pair covering the working area according to the working area range, wherein the satellite image data at least includes forward-looking image data and rear-looking image data; A processing module, used for pre-processing the rear-view image data to obtain a satellite rear-view image base map; It is also used to extract building spot vectors from the rear view image base map using an object-oriented classification machine learning method, and obtain two-dimensional building contour data in combination with manual visual correction and adjustment; An extraction module is used to use a preset satellite processing module to import the forward image data and the rearward image data, and extract digital surface model DSM data; It is also used to extract building roof elevation data from DSM data using a quantile truncation method based on the two-dimensional building contour data; It is also used to extract building ground elevation data from DSM data based on the two-dimensional building contour data by combining probability density estimation with quantile truncated mean method; A calculation module, used for performing difference calculation on the building roof elevation data and the building ground elevation data to obtain the height data of all buildings; The generating module is used to generate the urban white film based on the two-dimensional outline data of the building and the height data of all the buildings.

Images