Building facade structure extraction method and device, electronic equipment and medium

Abstract

The invention relates to a building facade structure extraction method and device, electronic equipment and a medium, and belongs to the technical field of building structure detection.The method comprises the steps that first point cloud data of a building facade are acquired, and a three-dimensional space is established according to the first point cloud data; dividing the three-dimensional space into a plurality of voxels, and performing super voxel clustering on the plurality of voxels to obtain at least one first super voxel; for each first super voxel, determining a difference between normal vectors of two adjacent points in the first super voxel; taking the points of which the differences are greater than a preset difference threshold as noisy points, and removing the noisy points to obtain second point cloud data; and determining the facade structure of the building according to the second point cloud data. According to the method, clustering is performed firstly, and then the noisy points in the point clouds are screened out according to the normal vectors of the point clouds in the first super voxels obtained through clustering, so that the validity of the point cloud data can be improved, and the accuracy of the facade structure can be improved when the facade structure of the building is constructed through the point cloud data.

Description

Technical Field

[0001] The present invention relates to the technical field of building structure detection, and in particular to a method, device, electronic equipment and medium for extracting building facade structure. Background Art

[0002] The facade structure of a building refers to the appearance and structure of the building's exterior walls, which includes walls, windows, doors, decorative elements, etc. It is the part of the building that is in direct contact with the external environment. Accurately extracting the facade structure of a building is the basis for research such as building image stitching, 3D laser point cloud feature extraction, and 3D reconstruction.

[0003] In the prior art, the extraction of building facade structures mostly uses 3D laser scanning technology to obtain point cloud data of the building facade, and then extracts feature points or contour points from the point cloud according to the properties of the point cloud itself, such as coordinates, reflection intensity, etc. to construct the building facade structure. However, in the prior art, the point cloud data of the building obtained often contains many noise points, which are difficult to effectively remove, resulting in the facade structure constructed based on the point cloud data. The structure is not accurate enough. Summary of the invention

[0004] In view of this, it is necessary to provide a building facade structure extraction method, device, electronic equipment and medium to solve the problem that the facade structure constructed is not accurate enough when the building facade structure is constructed based on point cloud data.

[0005] In order to solve the above problems, in a first aspect, the present invention provides a method for extracting a building facade structure, comprising: Acquire first point cloud data of a building facade, and establish a three-dimensional space according to the first point cloud data; Dividing the three-dimensional space into a plurality of voxels, and performing supervoxel clustering on the plurality of voxels according to point clouds in the plurality of voxels to obtain at least one first supervoxel; For each of the first supervoxels, determining a difference between normal vectors of two adjacent points in the first supervoxel; The points with a difference greater than a preset difference threshold are regarded as noise points, and the data of the noise points in the first point cloud data are removed to obtain the second point cloud data; The facade structure of the building is determined according to the second point cloud data.

[0006] Optionally, the performing supervoxel clustering on the plurality of voxels according to the point clouds in the plurality of voxels to obtain at least one first supervoxel comprises: Determine a point cloud in each voxel according to the first point cloud data; Determine a first similarity between voxels according to the point cloud in each voxel; the first similarity is a position similarity; The plurality of voxels are clustered into supervoxels according to the first similarity to obtain at least one first supervoxel.

[0007] Optionally, determining the facade structure of the building according to the second point cloud data includes: Determine a point cloud in each voxel according to the second point cloud data; Determine a second similarity between voxels according to the point cloud in each voxel; the second similarity is determined according to position similarity, color similarity and normal vector similarity; Performing supervoxel clustering on the plurality of voxels according to the second similarity to obtain at least one second supervoxel; The facade structure of the building is determined according to the second supervoxel.

[0008] Optionally, determining the second similarity between voxels according to the point cloud in each voxel includes: Determine a target voxel and a search range; wherein the target voxel is a preset voxel and an adjacent voxel with the second smallest similarity among adjacent voxels of the preset voxel; Performing a second similarity calculation on the target voxel and adjacent voxels within a search range of the target voxel; The calculation method of the second similarity is as follows:

[0009] Among them, D is the second similarity, D c is the color difference between two voxels, D s is the distance difference between two voxels, D n is the difference in the normal vector of the point cloud between two voxels, R s is the search range, w c 、w s 、w n They are the influence weights of color difference, distance difference, and normal vector difference respectively.

[0010] Optionally, determining the facade structure of the building according to the second point cloud data includes: Performing plane fitting processing according to the second point cloud data to obtain a fitting plane; performing clustering processing on the second supervoxel according to the fitting plane and the second supervoxel to obtain a cluster block; The facade structure of the building is determined according to the cluster blocks.

[0011] Optionally, the clustering the second supervoxels according to the fitting plane and the second supervoxels to obtain cluster blocks includes: Determine a target fitting point corresponding to the fitting plane; Determine the projection point of the target fitting point on the corresponding fitting plane; The second supervoxels are clustered based on the projection points using a restricted plane cutting algorithm to obtain cluster blocks.

[0012] Optionally, determining the facade structure of the building according to the cluster blocks includes: Determining the concave-convex structural relationship of the cluster blocks by a restricted plane cutting algorithm; The facade structure of the building is determined according to the concave-convex structural relationship between the cluster blocks and the cluster blocks.

[0013] In a second aspect, the present invention further provides a building facade structure extraction device, comprising: A space creation module, used to obtain first point cloud data of the building facade and establish a three-dimensional space according to the first point cloud data; A clustering module, configured to divide the three-dimensional space into a plurality of voxels, and perform supervoxel clustering on the plurality of voxels according to point clouds in the plurality of voxels to obtain at least one first supervoxel; a normal vector difference determination module, configured to determine, for each of the first supervoxels, a difference between normal vectors of two adjacent points in the first supervoxel; a point cloud data filtering module, configured to treat points with a difference greater than a preset difference threshold as noise points, and remove the noise points in the first point cloud data to obtain second point cloud data; A facade structure determination module is used to determine the facade structure of the building according to the second point cloud data.

[0014] In a third aspect, the present invention also provides an electronic device comprising a memory and a processor, wherein the memory is used to store programs; and the processor is coupled to the memory and is used to execute the programs stored in the memory to implement the steps in the above-mentioned building facade structure extraction method.

[0015] In a fourth aspect, the present invention further provides a computer-readable storage medium for storing a computer-readable program, wherein the program or instructions, when executed by a processor, can implement the steps in the above-mentioned building facade structure extraction method.

[0016] The beneficial effects of the present invention are: The present invention first obtains the first point cloud data of the building facade, divides the point cloud in the three-dimensional space into each voxel according to the first point cloud data, and then clusters the voxels according to the point cloud in each voxel to obtain the first supervoxel; respectively determines the normal vector of each point in each first supervoxel, and when the difference between the normal vectors of two adjacent points in the first supervoxel is less than or equal to the preset difference threshold, it can be determined that the two points are points located on the same facade and are valid points that need to be retained; and when the difference between the normal vectors of two adjacent points in the first supervoxel is greater than the preset difference threshold, it can be determined that the two points are not points located on the same facade, so the two points can be treated as noise points and removed, so that the points in each first supervoxel are as much as possible points on the same facade, thereby improving the effectiveness and accuracy of the point cloud data. Furthermore, when the facade structure of the building is constructed by the second point cloud data with the noise data removed, the accuracy of the facade structure can be improved. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 A schematic flow chart of an embodiment of a building facade structure extraction method provided by the present invention; Figure 2 A schematic flow chart of another embodiment of the building facade structure extraction method provided by the present invention; Figure 3 A schematic diagram of a point cloud data preprocessing flow chart provided by the present invention; Figure 4 A schematic diagram of a flow chart of a point cloud segmentation method based on plane feature preservation provided by the present invention; Figure 5 A schematic flow chart of a point cloud clustering method based on a restricted plane cutting algorithm provided by the present invention; Figure 6 A schematic structural diagram of an embodiment of a building facade structure extraction device provided by the present invention. DETAILED DESCRIPTION

[0018] The technical solutions in the embodiments of the present invention will be described clearly and completely below in conjunction with the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of the present invention.

[0019] In the description of the embodiments of the present invention, unless otherwise specified, "multiple" means two or more than two. "And / or" describes the association relationship of associated objects, indicating that three relationships may exist. For example, "A and / or B" may mean: A exists alone, A and B exist at the same time, and B exists alone.

[0020] The descriptions of "first", "second", etc. involved in the embodiments of the present invention are only for descriptive purposes and cannot be understood as indicating or implying their relative importance or implicitly indicating the number of the indicated technical features. Therefore, the technical features defined as "first" and "second" may explicitly or implicitly include at least one of the features.

[0021] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present invention. The appearance of the phrase in various places in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0022] Reference Figure 1 , shows a flow chart of an embodiment of a building facade structure extraction method provided by the present invention, the method comprising: Step S101, obtaining first point cloud data of a building facade, and establishing a three-dimensional space according to the first point cloud data.

[0023] Step S102 : dividing the three-dimensional space into a plurality of voxels, and performing supervoxel clustering on the plurality of voxels according to point clouds in the plurality of voxels to obtain at least one first supervoxel.

[0024] Step S103 : for each first supervoxel, determine the difference between normal vectors of two adjacent points in the first supervoxel.

[0025] Step S104: taking points whose differences are greater than a preset difference threshold as noise points, and removing the data of the noise points in the first point cloud data to obtain second point cloud data.

[0026] Step S105: determining the facade structure of the building according to the second point cloud data.

[0027] In this embodiment, the building can be scanned by a 3D laser scanner to obtain the first point cloud data of the building facade; the first point cloud data may include the location data, color data, etc. of the point cloud. After obtaining the first point cloud data, the first point cloud data may be subjected to registration processing and preliminary denoising processing, and the preliminary denoising processing includes the removal of non-building facade point clouds, air noise removal, large ground object removal, and small ground object removal.

[0028] After that, the first point cloud data after the registration processing and the preliminary denoising processing can be voxelized. The process of voxelization includes: establishing a three-dimensional space according to the first point cloud data, so that the three-dimensional space includes all point clouds recorded in the first point cloud data, and then dividing the three-dimensional space into multiple voxels, so that the above point clouds are distributed in each voxel. Voxel is the abbreviation of "volume pixel", which is a basic unit in three-dimensional space, similar to a pixel in a two-dimensional image. Voxel is used to represent a point or a small volume area in a three-dimensional data set. In an example of this embodiment, the voxelization process can be specifically as follows: searching the first point cloud data for two points with the smallest and largest coordinates as two vertices of the three-dimensional space, the line connecting the vertices is the diagonal of the three-dimensional space, and the three-dimensional space contains all point clouds; then, the three-dimensional space is divided into multiple voxels according to the voxel resolution, each voxel is a subspace of the three-dimensional space, and the point clouds are distributed in each voxel; finally, the coordinate system of the three-dimensional space can be established according to the right-hand rule with the point with the smallest coordinate as the origin, the center of mass of each voxel is determined according to the point cloud in each voxel, and the coordinates of the center of mass of the voxel in the coordinate system are used as the coordinates of the voxel. It should be noted that those skilled in the art can also establish the three-dimensional space and divide the voxels according to other methods.

[0029] After the first point cloud data is voxelized, the voxels can be clustered into supervoxels. Specifically, the voxels can be clustered into supervoxels based on at least one factor of the position similarity, color similarity, and normal vector similarity of the point cloud in the voxels to obtain at least one first supervoxel. Supervoxel clustering is an image processing technology that divides an image into multiple regions consisting of similar pixels or similar voxels. These regions are called "supervoxels." Supervoxels do not simply divide the image into grids, but aggregate adjacent and similar pixels or voxels together based on the color, brightness, texture and other characteristics of the pixels or voxels to form regions with certain connectivity.

[0030] For each first supervoxel, the normal vector of each point in each first supervoxel can be calculated respectively, and then it is determined whether the difference between the normal vectors of two adjacent points in the first supervoxel is greater than a preset difference threshold. If it is less than or equal to, it can be determined that the two points are points located on the same facade and are valid points that need to be retained; if it is greater than, it can be determined that the two points are not points located on the same facade, so the two points can be treated as noise points and removed, so that the points in each first supervoxel are points on the same facade as much as possible, that is, the plane features in each first supervoxel are retained.

[0031] Finally, the noise data in the first point cloud data is removed to obtain the second point cloud data, and the facade structure of the building is determined according to the second point cloud data.

[0032] The embodiment of the present invention performs voxelization and supervoxel clustering on the first point cloud data of the building facade to obtain the first supervoxel; based on the difference in the normal vectors of two adjacent points in each first supervoxel, the noise point in each first supervoxel is determined, and then the facade structure of the building is constructed by using the second point cloud data with the noise point data removed, thereby improving the accuracy of the facade structure.

[0033] In one embodiment, step S102 may specifically include: determining a point cloud in each voxel according to the first point cloud data; determining a first similarity between voxels according to the point cloud in each voxel; the first similarity is position similarity; performing supervoxel clustering on multiple voxels according to the first similarity to obtain at least one first supervoxel. In this embodiment, voxels may be clustered only according to position similarity to simplify the clustering process.

[0034] In one embodiment, after filtering out the noise points, the voxels can be clustered again according to the point cloud in the voxels, and step S105 can specifically include: determining the point cloud in each voxel according to the second point cloud data; determining the second similarity between voxels according to the point cloud in each voxel; the second similarity is determined according to the position similarity, color similarity and normal vector similarity; performing supervoxel clustering on multiple voxels according to the second similarity to obtain at least one second supervoxel; determining the facade structure of the building according to the second supervoxel. In this embodiment, on the one hand, the voxels are clustered again by filtering out the second point cloud data of the noise point data, and on the other hand, the process of re-clustering the voxels considers more factors compared to the previous clustering process, so that the clustering result of this time will be more accurate. When the facade structure of the building is determined according to the clustering result of this time, the accuracy of the facade structure can be improved.

[0035] In one embodiment, the step of determining the second similarity between voxels according to the point cloud in each voxel may specifically include: determining a target voxel and a search range; wherein the target voxel is a preset voxel and an adjacent voxel with the smallest second similarity among adjacent voxels of the preset voxel; performing a second similarity calculation on the target voxel and adjacent voxels within the search range of the target voxel; and the calculation method of the second similarity is as follows:

[0036] Among them, D is the second similarity, D c is the color difference between two voxels, D s is the distance difference between two voxels, D n is the difference in the normal vector of the point cloud between two voxels, R s is the search range, w c 、w s 、w nThey are the influence weights of color difference, distance difference, and normal vector difference respectively.

[0037] In this embodiment, a reasonable initial seed voxel spacing R_seed can be set first, and then a uniformly distributed initial seed voxel set is selected in three-dimensional space according to the spacing, and each initial seed voxel (preset voxel, i.e., target voxel) is distinguished from other initial seed voxels as a class. Since the point cloud cannot fill the voxel space, there are isolated seed voxels that do not contain any points in the initial seed voxel set. Therefore, it is necessary to filter and denoise the initial seed voxel set to remove the isolated seed voxels from the set, thereby improving the efficiency of supervoxel clustering. Then, the search range is obtained, and the initial seed voxel is used as the starting point to determine the adjacent voxels of the initial seed voxel within the search range, and the second similarity calculation is performed on the initial seed voxel and the adjacent voxel, and similar adjacent voxels are classified into the class of the initial seed voxel.

[0038] After traversing the neighboring voxels of all initial seed voxels, the neighboring voxel with the smallest similarity is marked, and the marked neighboring voxel (target voxel) is used as the search starting point to traverse the neighboring voxels in the search range until all neighboring voxels except the initial seed voxel complete the second similarity calculation.

[0039] Finally, the number of voxels contained in each class is judged. If it is less than the established minimum threshold, the class is merged into the adjacent class.

[0040] In one embodiment, the step of determining the facade structure of the building based on the second point cloud data may specifically include: performing plane fitting processing based on the second point cloud data to obtain a fitting plane; clustering the second supervoxels based on the fitting plane and the second supervoxels to obtain cluster blocks; and determining the facade structure of the building based on the cluster blocks.

[0041] After obtaining the second supervoxel, the second supervoxel needs to be clustered again to obtain the building facade. However, it has been found through research that when clustering the second supervoxel, the clustering result is that the large building facade surface will be segmented into small building facades, while the actual small building facade and the large building facade should be the same facade, and the small building facade should not be segmented separately. After studying this phenomenon, it was found that the root cause of the phenomenon is the rough surface of the large building facade. Therefore, in order to solve the above problem, in this embodiment, a plane fitting process is first performed according to the second point cloud data to obtain a smooth fitting plane, and then the second supervoxel is clustered based on the fitting plane to obtain a cluster block. The building facade point cloud clustering method based on plane fitting has a higher block recognition accuracy. The plane fitting algorithm can be principal component analysis (PCA).

[0042] In one embodiment, according to the fitting plane and the second supervoxel, the second supervoxel is clustered to obtain a cluster block, including: determining the target fitting point corresponding to the fitting plane; determining the projection point of the target fitting point on the corresponding fitting plane; clustering the second supervoxel based on the projection point by a constrained plane cutting algorithm to obtain a cluster block. Constrained Planar Cuts (CPC) algorithm. The projection point can be used as the input of the CPC algorithm, and the second supervoxel is clustered again by the CPC algorithm to obtain a cluster block, which is a large building facade. The projection point data has clear normal information, and at the same time filters out the influence of small area concave areas on the large building facade on block recognition. The building facade point cloud clustering method based on the CPC algorithm improved by plane fitting has higher block recognition accuracy, so that the integrity of building facade recognition is fundamentally solved.

[0043] In one embodiment, determining the facade structure of a building according to cluster blocks includes: determining the concave-convex structural relationship of the cluster blocks by a restricted plane cutting algorithm; and determining the facade structure of the building according to the concave-convex structural relationship between cluster blocks.

[0044] The CPC algorithm can determine the concave-convex structural relationship of cluster blocks based on the Euclidean edge cloud (EEC). The idea of ​​EEC is quite magical because the concave-convexity is defined on two adjacent "pieces", in other words, it is defined on the edge connecting the two adjacent "pieces". Each edge is abstracted into a point cloud, and a point cloud with concave-convex information is obtained. The concave edge is weighted 1, and the others are 0, and the point cloud is defined in red and blue colors. If there are more blue points, the more concave it is, the more it should be cut, and the problem is transformed into using blue points to find a plane. The RanSaC algorithm is used to use the point cloud to find the most likely plane, and an evaluation function is introduced to evaluate the quality of this segmentation. The evaluation function is: , where S m Represents the evaluation of the quality of this segmentation, P m represents the distance from the EEC point concentration point to the plane, w i The weight of each EEC point.

[0045] In the prior art, the extraction of building facade structures often relies on the spatial characteristics of point clouds, such as spatial resolution, etc. Therefore, for areas with uneven point cloud density, such methods may have problems of incorrect or missed structural line extraction. However, the present invention does not rely on spatial resolution and point cloud density when extracting building facade structures, so the accuracy will be higher.

[0046] Reference Figure 2 , shows a flow chart of another embodiment of the building facade structure extraction method provided by the present invention. The building facade point cloud data is collected, and the point cloud data is preprocessed such as point cloud denoising and point cloud registration. Then, the noise points in the point cloud are screened out and the valid points in the point cloud are retained by the supervoxel segmentation algorithm (supervoxel clustering algorithm) based on plane feature preservation. Finally, the facade structure of the building is determined by the CPC algorithm based on plane fitting.

[0047] Reference Figure 3 , shows a schematic diagram of a point cloud data preprocessing flow chart provided by the present invention. The preprocessing of point cloud data mainly includes point cloud denoising and point cloud registration processing.

[0048] Reference Figure 4 , showing a schematic flow chart of a point cloud segmentation method based on plane feature preservation provided by the present invention. Plane feature preservation means filtering out noise points that do not belong to the same facade in the supervoxel after supervoxel clustering, and retaining valid points belonging to the same facade, that is, maintaining the plane feature of the supervoxel.

[0049] Reference Figure 5 , shows a schematic flow chart of a point cloud clustering method based on a restricted plane cutting algorithm provided by the present invention. When the supervoxels are further clustered based on the CPC algorithm, the plane fitting algorithm can be combined to improve the accuracy of cluster block recognition.

[0050] Reference Figure 6 , shows a schematic structural diagram of an embodiment of a building facade structure extraction device provided by the present invention, the device comprising: The space creation module 10 is used to obtain first point cloud data of the building facade and establish a three-dimensional space according to the first point cloud data; A clustering module 20, configured to divide the three-dimensional space into a plurality of voxels, and perform supervoxel clustering on the plurality of voxels according to point clouds in the plurality of voxels to obtain at least one first supervoxel; A normal vector difference determination module 30 is used to determine, for each first supervoxel, a difference between normal vectors of two adjacent points in the first supervoxel; A point cloud data filtering module 40 is used to treat points with a difference greater than a preset difference threshold as noise points, and remove the noise points in the first point cloud data to obtain second point cloud data; The facade structure determination module 50 is used to determine the facade structure of the building according to the second point cloud data.

[0051] It should be noted that the implementation principles or implementation processes of the above modules can refer to the above-mentioned embodiment of building facade structure extraction, and will not be described one by one here.

[0052] Those skilled in the art will appreciate that all or part of the processes of the above-mentioned embodiments can be implemented by instructing related hardware through a computer program, and the program can be stored in a computer-readable storage medium, wherein the computer-readable storage medium is a disk, an optical disk, a read-only storage memory, or a random access memory, etc.

[0053] The above description is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily conceived by any technician familiar with the technical field within the technical scope disclosed by the present invention should be covered within the protection scope of the present invention.

Claims

1. A method for extracting building facade structure, characterized in that: include: Acquire first point cloud data of a building facade, and establish a three-dimensional space according to the first point cloud data; Dividing the three-dimensional space into a plurality of voxels, and performing supervoxel clustering on the plurality of voxels according to point clouds in the plurality of voxels to obtain at least one first supervoxel; For each of the first supervoxels, determining a difference between normal vectors of two adjacent points in the first supervoxel; The points with a difference greater than a preset difference threshold are regarded as noise points, and the data of the noise points in the first point cloud data are removed to obtain the second point cloud data; The facade structure of the building is determined according to the second point cloud data.

2. The building facade structure extraction method according to claim 1, characterized in that: The step of performing supervoxel clustering on the plurality of voxels according to the point clouds in the plurality of voxels to obtain at least one first supervoxel comprises: Determine a point cloud in each voxel according to the first point cloud data; Determine a first similarity between voxels according to the point cloud in each voxel; the first similarity is a position similarity; The plurality of voxels are clustered into supervoxels according to the first similarity to obtain at least one first supervoxel.

3. The building facade structure extraction method according to claim 1, characterized in that: The determining the facade structure of the building according to the second point cloud data comprises: Determine a point cloud in each voxel according to the second point cloud data; Determine a second similarity between voxels according to the point cloud in each voxel; the second similarity is determined according to position similarity, color similarity and normal vector similarity; Performing supervoxel clustering on the plurality of voxels according to the second similarity to obtain at least one second supervoxel; The facade structure of the building is determined according to the second supervoxel.

4. The building facade structure extraction method according to claim 3, characterized in that: The determining the second similarity between voxels according to the point cloud in each voxel comprises: Determine a target voxel and a search range; wherein the target voxel is a preset voxel and an adjacent voxel with the second smallest similarity among adjacent voxels of the preset voxel; Performing a second similarity calculation on the target voxel and adjacent voxels within a search range of the target voxel; The calculation method of the second similarity is as follows: Among them, D is the second similarity, D c is the color difference between two voxels, D s is the distance difference between two voxels, D n is the difference in the normal vector of the point cloud between two voxels, R s is the search range, w c 、w s 、w n They are the influence weights of color difference, distance difference, and normal vector difference respectively.

5. The building facade structure extraction method according to claim 3, characterized in that: The determining the facade structure of the building according to the second point cloud data comprises: Performing plane fitting processing according to the second point cloud data to obtain a fitting plane; performing clustering processing on the second supervoxel according to the fitting plane and the second supervoxel to obtain a cluster block; The facade structure of the building is determined according to the cluster blocks.

6. The building facade structure extraction method according to claim 5, characterized in that: The step of clustering the second supervoxel according to the fitting plane and the second supervoxel to obtain a cluster block comprises: Determine a target fitting point corresponding to the fitting plane; Determine the projection point of the target fitting point on the corresponding fitting plane; The second supervoxels are clustered based on the projection points using a restricted plane cutting algorithm to obtain cluster blocks.

7. The building facade structure extraction method according to claim 6, characterized in that: The step of determining the facade structure of the building according to the cluster blocks comprises: Determining the concave-convex structural relationship of the cluster blocks by a restricted plane cutting algorithm; The facade structure of the building is determined according to the concave-convex structural relationship between the cluster blocks and the cluster blocks.

8. A building facade structure extraction device, characterized in that: include: A space creation module, used to obtain first point cloud data of the building facade and establish a three-dimensional space according to the first point cloud data; A clustering module, configured to divide the three-dimensional space into a plurality of voxels, and perform supervoxel clustering on the plurality of voxels according to point clouds in the plurality of voxels to obtain at least one first supervoxel; a normal vector difference determination module, configured to determine, for each of the first supervoxels, a difference between normal vectors of two adjacent points in the first supervoxel; a point cloud data filtering module, configured to treat points with a difference greater than a preset difference threshold as noise points, and remove the noise points in the first point cloud data to obtain second point cloud data; A facade structure determination module is used to determine the facade structure of the building according to the second point cloud data.

9. An electronic device, characterized in that: comprising a memory and a processor, wherein: The memory is used to store programs; The processor is coupled to the memory and is used to execute the program stored in the memory to implement the steps in the building facade structure extraction method described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer-readable program, wherein when the program or instruction is executed by a processor, the steps in the building facade structure extraction method described in any one of claims 1 to 7 can be implemented.

Images