A method and device for automatically generating a three-dimensional building model using satellite images

Abstract

The present invention provides a method and device for automatically generating a three-dimensional model of a building using satellite images, the method comprising: collecting satellite images containing buildings, and preprocessing the satellite images; dividing the preprocessed satellite images into satellite blocks to identify the roof outline of the building; extracting the building body outline of the building based on the preprocessed satellite images; correcting the roof position and calculating the building height according to the roof outline and the building body outline of the building; and using the corrected roof position and building height to generate a three-dimensional model of the building using a three-dimensional reconstruction algorithm. The present invention realizes automatic recognition of the building roof outline and the building body mapping outline through high-precision image recognition and processing technology, and corrects the position of the building roof according to the shape and position information of these outlines, and calculates the height of the building, thereby automatically generating a three-dimensional model.

Description

Technical Field

[0001] The present invention relates to the technical field of three-dimensional modeling, and in particular to a method and a device for automatically generating a three-dimensional model of a building using satellite images. Background Art

[0002] With the continuous evolution of the construction industry, digital cities and smart cities are the focus of urban development and management, and surveying and mapping technology is an important technology for building digital cities and smart cities. Especially in important links such as building design, construction and management, three-dimensional modeling methods are particularly important.

[0003] In the existing technology, the production of three-dimensional models of buildings mainly relies on manual measurement and modeling. For example, laser beam scanning is used to record the laser beam reflection information, and then the point cloud data of the building is obtained to achieve three-dimensional modeling. However, the laser scanning method also requires manual operation to realize the transmission of signals, or the laser radar method is used to obtain the point cloud data of the building. However, the above methods require a lot of manpower and material resources, and are easily affected by human errors and subjective judgments, resulting in low production efficiency and low model accuracy. In addition, manual modeling also has the problem of being difficult to achieve large-scale, automated and rapid production requirements. Summary of the invention

[0004] In view of the above problems, the present invention proposes a method and device for automatically generating a three-dimensional building model using satellite images, which overcomes the above problems or at least partially solves the above problems.

[0005] According to one aspect of the present invention, a method for automatically generating a three-dimensional building model using satellite images is provided, the method comprising:

[0006] Collecting satellite images containing buildings and preprocessing the satellite images;

[0007] The pre-processed satellite images are divided into satellite image blocks to identify the roof outlines of buildings;

[0008] Extracting the building outlines of buildings based on preprocessed satellite images;

[0009] Correct the roof position and calculate the building height according to the building roof outline and building body outline;

[0010] Using the corrected roof position and building height, a 3D reconstruction algorithm is used to generate a 3D model of the building.

[0011] Optionally, dividing the pre-processed satellite image into satellite image blocks to identify the roof outline of the building includes:

[0012] Divide the preprocessed satellite image into satellite blocks to obtain a divided satellite image;

[0013] Inputting the block satellite image into a SAM model trained and improved by a supervised learning method, and outputting a grid semantic map of the semantic information of the building top surface after analysis by the SAM model;

[0014] Connected domain analysis is used to extract the contour to obtain the roof contour of the building;

[0015] Select the overlapping building objects in blocks and save them in vector format to obtain the digital line map of the building top surface.

[0016] Optionally, extracting the building outline of the building based on the preprocessed satellite image includes:

[0017] The SAM model improved by supervised training is used to identify and extract the building outline of the building in the satellite image. The building outline of the building includes all pixels from the base to the top of the building in the image and is saved as vector information.

[0018] Optionally, the correcting the roof position and calculating the building height according to the building roof outline and the building body outline comprises:

[0019] According to the roof outline of the building and the building body outline, a shape matching algorithm is used to iteratively translate the roof outline of the building within the building body outline to find out the difference between the roof outline of the building and the building body outline and the maximum distance that the digital line map of the top surface of the building can be moved, thereby correcting the position of the roof of the building to its foundation position;

[0020] Estimate the height of buildings based on the scale and corrected roof positions in satellite imagery.

[0021] Optionally, the three-dimensional reconstruction algorithm includes a stereo matching algorithm or a surface reconstruction algorithm.

[0022] According to another aspect of the present invention, there is also provided a device for automatically generating a three-dimensional model of a building using satellite images, comprising a processor and a memory storing program instructions, wherein the processor is configured to execute a method for automatically generating a three-dimensional model of a building using satellite images as described in any one of the above items when executing the program instructions.

[0023] The present invention also provides a computer-readable storage medium, which is used to store program code, and the program code is used to execute any of the above-mentioned methods for automatically generating a three-dimensional building model using satellite images.

[0024] The present invention also provides a computing device, including the device for automatically generating a three-dimensional building model using satellite images.

[0025] The method and device for automatically generating a three-dimensional model of a building using satellite images provided by the present invention realizes automatic recognition of the outline of the building roof and the mapping outline of the building body through high-precision image recognition and processing technology, and corrects the position of the building roof according to the shape and position information of these outlines, and calculates the height of the building at the same time, thereby automatically generating a three-dimensional model. By introducing satellite images and AI algorithms, the automatic generation of a three-dimensional model of a building is realized, solving the problems of low production efficiency and low model accuracy caused by manual operation in the prior art. The scheme of the present invention has the following technical effects:

[0026] 1. Improved production efficiency: This patent realizes the automation and rapid production of three-dimensional models of buildings, greatly improving production efficiency.

[0027] 2. Improved model accuracy: This patent improves the accuracy and precision of model recognition by adopting high-precision image recognition and processing technology and shape matching algorithm, thereby improving the accuracy of the model.

[0028] 3. Reduced costs: This patent does not require a lot of manual intervention, reducing the cost of manpower and material resources.

[0029] 4. Scalability and flexibility: This patent has good scalability and flexibility, and can adapt to different scenarios and application requirements.

[0030] The above description is only an overview of the technical solution of the present invention. In order to more clearly understand the technical means of the present invention, it can be implemented according to the contents of the specification. In order to make the above and other purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are listed below.

[0031] Based on the following detailed description of specific embodiments of the present invention in conjunction with the accompanying drawings, those skilled in the art will become more aware of the above and other objects, advantages and features of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0032] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0033] Figure 1 A schematic flow chart of a method for automatically generating a three-dimensional building model using satellite images according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0034] The exemplary embodiments of the present invention will be described in more detail below with reference to the accompanying drawings. Although the exemplary embodiments of the present invention are shown in the accompanying drawings, it should be understood that the present invention can be implemented in various forms and should not be limited by the embodiments set forth herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present invention and to enable the scope of the present invention to be fully communicated to those skilled in the art.

[0035] The embodiment of the present invention provides a method for automatically generating a three-dimensional building model using satellite images. Figure 1 As shown, the method for automatically generating a three-dimensional building model using satellite images in this embodiment may include:

[0036] S1, collecting satellite images containing buildings and preprocessing the satellite images;

[0037] S2, divide the pre-processed satellite image into satellite image blocks to identify the roof outline of the building;

[0038] S3, extracting the building outline of the building based on the preprocessed satellite image;

[0039] S4, correcting the roof position and calculating the building height according to the building roof outline and the building body outline;

[0040] S5, using the corrected roof position and building height, a three-dimensional model of the building is generated by using a three-dimensional reconstruction algorithm. For example, the three-dimensional model of the building can be generated by raising the building roof DLG according to the building height.

[0041] The method for automatically generating a three-dimensional building model using satellite images in an embodiment of the present invention realizes the automated, high-precision and rapid production of a three-dimensional building model through the steps of satellite image acquisition and processing, roof outline DLG recognition, building outline recognition, roof position correction and building 3D modeling, thus solving the problems existing in the prior art. In addition, the embodiment of the present invention also has good scalability and flexibility, and can adapt to different scenarios and application requirements.

[0042] In some embodiments, the above step S1 collects satellite images containing buildings, and preprocessing the satellite images may include: collecting satellite images containing buildings, and performing preprocessing, such as denoising, enhancement, etc. At the same time, according to the distribution of buildings in the satellite image range, select landmark buildings and obtain the height information of the landmark buildings through data query as reference data.

[0043] In some embodiments, the above step S2 of dividing the pre-processed satellite image into satellite image blocks to identify the roof outline of the building includes:

[0044] S2-1, the pre-processed satellite image is divided into satellite images to obtain a divided satellite image; optionally, according to the coordinate range of the satellite image, starting from the upper left corner of the satellite image, a 1024*1024 size image is cropped block by block from left to right and from top to bottom. To ensure the integrity of the building in the cropped image, the two adjacent images on the left and right and on the top and bottom overlap by 256 pixels (25%), thereby obtaining a divided satellite image.

[0045] S2-2, the block satellite image is input into the SAM model that has been trained and improved by the supervised learning method. Since the original SAM model is pre-trained with a large amount of data, it has a good representation ability for the compressed coding information of the input image, but it only divides by objects and does not generate specific categories. Therefore, the improved SAM model mainly uses the intermediate result image_embeddings of the original SAM model as the encoder to encode the input image information, and then builds a decoder neural network module to replace the decoder module in the SAM model, decodes the image_embeddings into a target semantic mask, and the SAM model analyzes and outputs a grid semantic map of the semantic information of the top surface of the building;

[0046] S2-3, using connected domain analysis to extract contours. Since the semantic information of the building top surface segmented by S2-2 is a two-dimensional mask image, and the building top surface exists in the form of patches with the same pixel values, the connected domain is used for analysis, that is, the eight pixels around a pixel with the same pixel value are the same connected domain. In this way, all patches are extracted separately according to the connected domain, and then their outer boundary contours are retained to obtain the building roof contour;

[0047] S2-4, select the overlapping building objects in blocks, save them in vector format to obtain the digital line drawing map of the building top surface. Specifically, the overlapping building objects in blocks are preferred. Due to the limitation of the size of the SAM network, S2-1 divides the image into 1024*1024 sizes, and there is a 25% overlap between adjacent images. The overlapping buildings may be recognized as slightly different results on multiple images. For multiple results determined to be the same building based on the intersection-union ratio, the building with the larger roof outline area is preferred as the final result, which is saved in vector format to obtain the digital line drawing map of the building top surface, i.e., the building top surface DLG.

[0048] The processed satellite images are divided into blocks and input into the SAM (Segment Anything Model) trained and improved by the supervised learning method to output the raster semantic map of the semantic information of the building top surface. Then, the connected domain analysis is used to extract the contours and optimize the overlapping building objects in the blocks. The blocks are saved in vector format as the digital line graphic (DLG) of the building top surface.

[0049] In some embodiments, the above-mentioned step S3 of extracting the building outline of the building based on the preprocessed satellite image includes: using the SAM model improved by supervised training to identify and extract the building outline of the building in the satellite image, wherein the building outline of the building includes all pixels from the base to the top surface of the building in the image and is saved as vector information.

[0050] In the same rooftop DLG extraction process, when identifying the building outline, the SAM model improved by supervised training is used to identify and extract the mapping outline of the building object in the satellite image. The improved SAM network has the same network structure as the DLG extraction network, but uses a deep learning model trained with a building data set. The satellite image cut into 1024*1024 blocks is input into the model to generate a semantic mask of the building. After connected domain processing, all pixels from the base to the top of the building on the image are extracted and saved as vector information. This embodiment uses high-precision image recognition and processing technology to automatically identify the building roof outline and the building mapping outline, thereby improving the accuracy and precision of recognition.

[0051] In some embodiments, the above step S4 corrects the roof position and calculates the building height according to the building roof outline and the building body outline, including:

[0052] S4-1, based on the roof outline and the building body outline of the building, a shape matching algorithm is used to iteratively translate the roof outline of the building within the building body outline to find the difference between the roof outline of the building and the building body outline and the maximum distance that the digital line map of the top surface of the building can move, thereby correcting the position of the roof of the building to its foundation position.

[0053] In this embodiment, a shape matching algorithm is used based on the building roof outline (DLG) and the building outline. Since the building outline range includes its roof outline, the building outline and its roof outline can be paired according to the size of the vector intersection. The building roof outline is iteratively translated within the building outline range. Within the range where the intersection area of ​​the building outline vector and its roof outline vector is greater than 90% of the area of ​​the roof outline vector, the farthest position to which the roof outline can be moved is the foundation position. The straight-line distance between the foundation position and the position of the original roof outline is recorded as the building mapping height (pixel). The foundation position is recorded as the true reflected DLG of the building, thereby correcting the position of the building roof to its foundation position.

[0054] S4-2, estimate the height of the building according to the scale in the satellite image and the corrected roof position. Since the actual height of the landmark building is recorded in S1 and the mapped height of the building is calculated in S4-1, the ratio of the actual height of the landmark building to its mapped height can be used as the height scale of the building. At the same time, the average of the height scales of multiple landmark buildings is taken as the scale of the building height in the satellite image, and the product of the mapped height of each building and the building height scale is taken as the estimated height of the building.

[0055] This embodiment uses a shape matching algorithm to find the difference between the building roof outline and the building body outline, thereby correcting the position of the building roof and improving the accuracy of the model. The height of the building is calculated based on the scale in the image and the corrected roof position, further improving the accuracy of the model.

[0056] In some embodiments, the above step S5 uses the corrected roof position and building height to generate a three-dimensional model of the building using a three-dimensional reconstruction algorithm, wherein the three-dimensional reconstruction algorithm includes a stereo matching algorithm or a surface reconstruction algorithm, etc. In this embodiment, the corrected roof outline is used as the true-shot DLG of the building, and then a three-dimensional white mold of the building is generated using a heightening modeling algorithm. Since the DLG extraction process is composed of the outer frame of the patch connected domain, the vector vertices are relatively dense. Under the condition that three consecutive points are linear, the middle points are simplified to simplify the vector vertices; then for adjacent vector edges, if the angle is 90±5 degrees, it is forced to be adjusted to 90 degrees as a right-angle treatment. After linear simplification and right-angle treatment, the true-shot DLG is raised according to the estimated height to generate a three-dimensional white mold model. This embodiment uses a three-dimensional reconstruction algorithm to generate a three-dimensional model of the building according to the corrected roof position and the calculated building height, thereby achieving automated, rapid and high-precision production.

[0057] The embodiment of the present invention collects satellite images containing buildings and performs preprocessing operations, such as denoising and enhancement, to improve the quality and clarity of the images. Then, AI deep learning technology is used to train and predict the preprocessed satellite images to automatically identify the roof outline of the building and the mapping outline of the building body. After identifying the roof outline and the building body outline, a shape matching algorithm is used to find the difference between the roof outline of the building and the building body outline, so as to correct the position of the roof of the building. At the same time, the height of the building is calculated according to the scale in the image and the corrected roof position. Finally, according to the corrected roof position and the calculated building height, a three-dimensional reconstruction algorithm, such as a stereo matching algorithm, a surface reconstruction algorithm, etc., is used to generate a three-dimensional model of the building.

[0058] An embodiment of the present invention also provides a device for automatically generating a three-dimensional model of a building using satellite images, comprising a processor and a memory storing program instructions, wherein the processor is configured to execute the method for automatically generating a three-dimensional model of a building using satellite images as described in the above embodiment when executing the program instructions.

[0059] An embodiment of the present invention further provides a computer-readable storage medium, wherein the computer-readable storage medium is used to store program code, and the program code is used to execute the method for automatically generating a three-dimensional building model using satellite images described in the above embodiment.

[0060] An embodiment of the present invention further provides a computing device, including the apparatus for automatically generating a three-dimensional building model using satellite images as described in the above embodiment.

[0061] Those skilled in the art can clearly understand that the specific working processes of the systems, devices, modules and units described above can refer to the corresponding processes in the aforementioned method embodiments, and for the sake of brevity, they are not further described here.

[0062] In addition, the functional units in various embodiments of the present invention may be physically independent of each other, or two or more functional units may be integrated together, or all functional units may be integrated into one processing unit. The above integrated functional units may be implemented in the form of hardware, or in the form of software or firmware.

[0063] Those skilled in the art can understand that if the integrated functional unit is implemented in the form of software and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention can essentially or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, which includes a number of instructions to enable a computing device (such as a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present invention when running the instructions. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes.

[0064] Alternatively, all or part of the steps of implementing the aforementioned method embodiments may be accomplished by hardware associated with program instructions (such as a computing device such as a personal computer, a server, or a network device), and the program instructions may be stored in a computer-readable storage medium. When the program instructions are executed by a processor of a computing device, the computing device executes all or part of the steps of the methods described in the embodiments of the present invention.

[0065] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that within the spirit and principles of the present invention, the technical solutions described in the aforementioned embodiments can still be modified, or some or all of the technical features therein can be replaced by equivalents. However, these modifications or replacements do not deviate from the protection scope of the present invention.

Claims

1. A method for automatically generating a three-dimensional building model using satellite images, characterized in that: The method comprises: Collecting satellite images containing buildings and preprocessing the satellite images; The pre-processed satellite images are divided into satellite image blocks to identify the roof outlines of buildings; Extracting the building outlines of buildings based on preprocessed satellite images; Correct the roof position and calculate the building height according to the building roof outline and building body outline; Using the corrected roof position and building height, a 3D reconstruction algorithm is used to generate a 3D model of the building; The method of dividing the preprocessed satellite image into satellite image blocks to identify the roof outline of the building comprises: dividing the preprocessed satellite image into satellite image blocks to obtain divided satellite images; inputting the divided satellite images into a SAM model trained and improved by a supervised learning method, and outputting a grid semantic map of the semantic information of the top surface of the building after analysis by the SAM model; extracting the contour by using a connected domain analysis to obtain the roof outline of the building; selecting the overlapping building objects in the blocks, and saving them in a vector format to obtain a digital line map of the top surface of the building; The method of correcting the roof position and calculating the building height according to the building roof outline and the building body outline comprises: according to the building roof outline and the building body outline, using a shape matching algorithm, iteratively translating the building roof outline within the building body outline, finding the difference between the building roof outline and the building body outline and the farthest distance that the digital line map of the building top surface can move, thereby correcting the position of the building roof to its foundation position; estimating the building height according to the scale in the satellite image and the corrected roof position.

2. The method according to claim 1, characterized in that Extracting the building outline of the building based on the pre-processed satellite image includes: The SAM model improved by supervised training is used to identify and extract the building outline of the building in the satellite image. The building outline of the building includes all pixels from the base to the top of the building in the image and is saved as vector information.

3. The method according to claim 1, characterized in that The three-dimensional reconstruction algorithm includes a stereo matching algorithm or a surface reconstruction algorithm.

4. A device for automatically generating a three-dimensional building model using satellite images, characterized in that: It comprises a processor and a memory storing program instructions, wherein the processor is configured to execute the method for automatically generating a three-dimensional building model using satellite images as described in any one of claims 1 to 2 when executing the program instructions.

5. A computer-readable storage medium, characterized in that: The computer-readable storage medium is used to store program codes, and the program codes are used to execute the method for automatically generating a three-dimensional building model using satellite images as described in any one of claims 1-2.

6. A computing device, characterized in that It includes the device for automatically generating a three-dimensional building model using satellite images as described in claim 4.

Images