A rapid construction method of high-reduction-degree building white model

Abstract

The application relates to the technical field of building modeling, in particular to a rapid construction method of a high-reduction building white model, which comprises the following steps: obtaining point cloud data of a target city; obtaining each closed edge contour at each height in the height direction of the building in the point cloud data, and taking the closed edge contour at the height of the position of the road surface as a ground contour; obtaining each profile surface in the target city and out-of-profile point cloud data of each profile surface; obtaining point cloud distribution characteristic values of each profile surface; obtaining each similar profile surface of each profile surface; obtaining the building feature similarity of each profile surface by the spatial distance between the ground contours and the similarity degree between the closed edge contours at each height for each profile surface and each similar profile surface thereof; obtaining the retainable degree of each profile surface, screening building point clouds and constructing a building white model. The application aims to improve the accuracy of the screened building point clouds and then improve the reduction degree of the subsequently constructed building white model.

Description

Technical Field

[0001] This application relates to the field of architectural modeling technology, specifically to a method for rapidly constructing high-fidelity architectural white models. Background Technology

[0002] Urban white models are a fundamental modeling method in the field of 3D modeling. They primarily refer to 3D models that contain only the shape and height information of buildings, without texture or detail information. Constructing an urban white model is a complex process involving data collection, processing, and 3D modeling. Its core elements include building outline data and building height data. While obtaining 2D building outline data is relatively easy and has various methods, obtaining building height data is more difficult. How to quickly and accurately obtain building height information has become a hot research topic.

[0003] Currently, the data sources used for constructing white models of urban buildings mainly consist of dense point cloud data of urban buildings obtained through technologies such as oblique photogrammetry, laser point cloud, and satellite stereo imaging. These point cloud data are then filtered using point cloud data filtering algorithms to separate the terrain and building point clouds, thus obtaining the building's point cloud data for white model construction. However, in actual modeling, the interaction between buildings and terrain, such as buildings situated on hillsides or those subject to shadows, can interfere with the extraction of building outlines, leading to a decrease in the accuracy and fidelity of the white model. Summary of the Invention

[0004] In view of the above, it is necessary to provide a rapid construction method for high-fidelity architectural white models. Compared with traditional rapid construction methods for architectural white models, this method improves the fidelity of the subsequently constructed architectural white model by increasing the accuracy of the selected architectural point cloud.

[0005] The rapid construction method for high-fidelity architectural white models proposed in this application adopts the following technical solution:

[0006] One embodiment of this application provides a rapid method for constructing high-fidelity architectural white models, the method comprising the following steps:

[0007] Acquire point cloud data of the target city and label the elevation of the road surface location in the point cloud data;

[0008] By analyzing the distribution of point cloud data in the target city, we obtain the closed edge contours at each height along the building height direction in the point cloud data. The closed edge contour at the height where the road surface is located is recorded as the ground contour. By analyzing the positional relationship between the closed edge contours at the height where the road surface is located and the heights of other heights, we obtain the contour surfaces in the target city. By analyzing the distance between the point cloud data of the target city and each contour surface, we obtain the outer point cloud data of each contour surface.

[0009] By analyzing the spatial distribution of point cloud data within each contour surface and combining it with clustering algorithms, we obtain the classification of point cloud data within each contour surface. By analyzing the quantity of point cloud data outside each contour surface and the dispersion of point cloud data within each contour surface along the building height direction, combined with the intra-class distance and inter-class distance of point cloud data within each contour surface, we obtain the point cloud distribution feature values ​​of each contour surface.

[0010] By comparing the ground contours of each contour surface with the other contour surfaces, we obtain the similar contour surfaces of each contour surface; for each contour surface and its similar contour surfaces, we obtain the architectural feature similarity of each contour surface by comparing the spatial distance between the ground contours and the similarity between the closed edge contours at each height.

[0011] By comparing the similarity between the point cloud distribution feature values ​​and the building features, the retainability of each contour surface is obtained. Based on the retainability, building point clouds are filtered from the point cloud data of the target city. A white model of the building is constructed based on the filtering results.

[0012] In one embodiment, the process of obtaining the contour surface is as follows:

[0013] The Euclidean distance between the center points of any closed edge contour at any height and each closed contour edge at any adjacent height is denoted as the center distance. The minimum value of all the center distances corresponding to any closed edge contour is calculated. When the minimum value is less than a preset threshold, it is determined that the closed edge contour corresponding to the minimum value at any adjacent height belongs to the same contour surface as the closed edge contour.

[0014] If the projection of any closed edge contour onto the X0Y plane of the ground coordinate system lies within the projection of any ground contour onto the X0Y plane, then the closed edge contour and the ground contour belong to the same contour surface.

[0015] In one embodiment, the process of acquiring the outline point cloud data is as follows:

[0016] Obtain the point cloud data outside any contour surface from the point cloud data of the target city, and denot them as candidate out-of-contour point cloud data. Calculate the minimum value of the Euclidean distance between each candidate out-of-contour point cloud data and all point cloud data on the contour surface. Using the minimum value of all candidate out-of-contour point cloud data, obtain the segmentation threshold of the minimum value of all candidate out-of-contour point cloud data by cross-validation. The candidate out-of-contour point cloud data whose minimum value is less than or equal to the segmentation threshold are taken as the out-of-contour point cloud data of the contour surface.

[0017] In one embodiment, the process of obtaining the point cloud distribution feature values ​​is as follows:

[0018] Calculate the mean Euclidean distance between any two categories of point cloud data within each contour plane;

[0019] Calculate the average Euclidean distance between any two point cloud data within each category, and calculate the sum of the average values ​​for all categories of point cloud data within each contour surface;

[0020] Calculate the product of the quantity and the mean;

[0021] Calculate the sum of the dispersion and the sum;

[0022] The point cloud distribution feature value can be further obtained through the product and the accumulated value.

[0023] In one embodiment, the point cloud distribution characteristic value is the ratio of the product to the accumulated value.

[0024] In one embodiment, the process of obtaining each similar contour surface of each contour surface is as follows:

[0025] The similarity coefficient of the ground contour between any two contour surfaces is obtained by contour matching algorithm; the contour surfaces whose normalized value of the similarity coefficient between each contour surface is greater than or equal to a preset threshold are regarded as similar contour surfaces.

[0026] In one embodiment, the process of obtaining the architectural feature similarity is as follows:

[0027] Obtain the mean Euclidean distance between each contour surface and all its similar contour surfaces on the ground contour.

[0028] Calculate the similarity coefficient of the closed edge profile at each height between each profile surface and each of its similar profile surfaces; calculate the sum of the similarity coefficients of the closed edge profile at all heights between each profile surface and all its similar profile surfaces.

[0029] The similarity of architectural features can be further determined by the mean calculation result and the summation.

[0030] In one embodiment, the building feature similarity is the ratio of the sum to the mean.

[0031] In one embodiment, the retainability is the product of the point cloud distribution feature value and the building feature similarity.

[0032] In one embodiment, the method for filtering the building point cloud is as follows:

[0033] By using the retainability of all contour surfaces, cross-validation is employed to obtain the segmentation threshold for the retainability of all contour surfaces. Contour surfaces with a retainability greater than or equal to the segmentation threshold are selected, and the point cloud data on the selected contour surfaces, the point cloud data within the contour surfaces, and the point cloud data outside the contour surfaces are combined to form the building point cloud.

[0034] This application has at least the following beneficial effects:

[0035] This application analyzes the positional relationship of closed edge contours at different heights to accurately extract contour surfaces, thereby better reflecting the actual shape and structure of the building and improving the fidelity of the white model. By extracting the outer contour point cloud data, it can more accurately reflect the actual boundary of the building, ensuring the integrity of the building point cloud and avoiding modeling errors caused by missing boundary points. By considering the differences in distribution characteristics between the building's point cloud data and the point cloud data of the terrain and landform area, it calculates the point cloud distribution characteristic value, which can quantify the probability that each contour surface and its subordinate point cloud data belong to the building point cloud, helping to more accurately determine which point cloud data belong to the building point cloud.

[0036] Furthermore, considering the existence of old buildings in the target city, the similarity of closed edge contours at different heights is analyzed. Combined with the point cloud distribution feature values, the retainability of each contour surface is calculated, which can more accurately assess the degree of each contour surface and its subordinate point cloud data of building point clouds. By filtering building point clouds through retainability, the architectural features of the contour surfaces can be more comprehensively evaluated, and the point cloud data of actual urban buildings and topography can be distinguished. This allows for the removal of topography point clouds from the urban point cloud data, improving the accuracy of the filtered building point clouds, and thus improving the fidelity of the subsequently constructed white model of buildings. Attached Figure Description

[0037] To more clearly illustrate the technical solutions and advantages in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 A flowchart illustrating the steps of a rapid construction method for high-fidelity architectural white models provided in this application;

[0039] Figure 2 A schematic diagram of point cloud data for building walls;

[0040] Figure 3 This is a schematic diagram of the outline surface;

[0041] Figure 4 A schematic diagram of the outer point cloud data;

[0042] Figure 5 This is a schematic diagram of the process for obtaining retainability. Detailed Implementation

[0043] In the description of the embodiments in this application, the words "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design scheme described as "exemplary" or "for example" in the embodiments of this application should not be construed as being more preferred or advantageous than other embodiments or design schemes. Specifically, the use of the words "exemplary," "or," and "for example" is intended to present the relevant concepts in a specific manner.

[0044] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. It should be understood that, unless otherwise stated, " / " in this application means "or".

[0045] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0046] The following description, in conjunction with the accompanying drawings, details a specific scheme for a rapid construction method for high-fidelity architectural white models provided in this application.

[0047] This application provides a method for rapidly constructing a high-fidelity architectural white model, specifically, the method described below. Please refer to [link to relevant documentation]. Figure 1 The method includes the following steps:

[0048] Step 1: Obtain point cloud data of the target city and mark the height of the road surface in the point cloud data.

[0049] Point cloud data of the target city for which architectural models need to be constructed is acquired using laser scanning technology. A point cloud filtering algorithm is then used to denoise the point cloud data, removing noise and outliers. Finally, the elevation of road surfaces in the point cloud data is labeled.

[0050] In this embodiment, the ROR (radius outlier removal) filtering algorithm is used to denoise the point cloud data of the target city. The ROR filtering algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to denoise the point cloud data of the target city, the implementer may use other existing technologies, such as the SOR (statistical outlier removal) statistical filtering algorithm, which will not be described in detail in this application.

[0051] Step 2: Denote the closed edge contour at the height of the road surface as the ground contour; obtain the contour surfaces in the target city by the positional relationship between the closed edge contours at the height of the road surface and other heights; obtain the outer point cloud data of each contour surface; classify the point cloud data within each contour surface; obtain the point cloud distribution feature value of each contour surface by the number of outer point cloud data of each contour surface and the dispersion of the point cloud data within each contour surface in the building height direction, combined with the intra-class distance and inter-class distance of the point cloud data within each contour surface.

[0052] Since urban white models only represent the shape and height of buildings, the construction process only requires the support of point cloud data related to the buildings. However, in addition to buildings, cities also contain many hills, small mountains, and slopes, among other topographical areas. During data collection, it is difficult to remove point cloud data from these areas using conventional thresholding or outlier removal methods. This can easily lead to confusion between the point cloud data of buildings and the point cloud data of topographical areas during the construction of the building white model, resulting in inaccuracies in the white model. Therefore, it is necessary to distinguish between the point cloud data of buildings and the point cloud data of topographical areas.

[0053] Step 2.1: Based on the distribution of point cloud data of the target city, obtain the closed edge contours at each height in the building height direction in the point cloud data, and record the closed edge contour at the height of the road surface as the ground contour; based on the positional relationship between the closed edge contours at the height of the road surface and other heights, obtain each contour surface in the target city; based on the distance between the point cloud data of the target city and each contour surface, obtain the outer point cloud data of each contour surface.

[0054] Typically, buildings in a city have similar shapes, generally resembling cuboids. Therefore, point cloud data is usually distributed on the walls perpendicular to the ground, and the distribution of point cloud data on these walls is relatively even. However, due to the complexity of individual building wall structures and unavoidable errors during point cloud data acquisition, the projection of the point cloud data onto the X0Y plane of the ground coordinate system for a single wall may not be a straight line. Point cloud data may exist around the projection, and this surrounding point cloud data is also considered part of the wall structure, representing its characteristics to some extent. A schematic diagram of building wall point cloud data is shown below. Figure 2 As shown, Figure 2 In the diagram, 1 represents point cloud data and 2 represents the actual wall surface.

[0055] As for urban topography, due to the irregularity of its shape and the presence of a lot of vegetation, the point cloud data appears more cluttered and isolated in its overall appearance. Furthermore, because the vegetation is distributed relatively dispersedly, the point cloud data of the topographic area also exhibits the characteristics of local clustering and overall dispersion. Compared to buildings, the shape of the topographic area is similar to a pyramid, with the outline gradually decreasing from bottom to top. Therefore, when viewed from above, the point cloud data of the topographic area is more concentrated within the ground outline.

[0056] Based on the above analysis, all point cloud data at the road surface height are obtained from the point cloud data of the target city. This point cloud data is then projected onto the X0Y plane to generate a two-dimensional image. The width and length of the two-dimensional image are determined by the range of all point cloud data at the road surface height. The grayscale value of the projected points of the point cloud data on the two-dimensional image is set to 0, and the grayscale value of the remaining points is set to 255. Using the two-dimensional image, an edge detection algorithm is employed to obtain all edge points at the road surface height. Then, a contour tracking algorithm is used to obtain all closed edge contours. The contour tracking algorithm is a well-known technique and will not be described in detail here. Each closed edge contour at the road surface height is denoted as a local surface contour, i.e., the ground contour of a single building or topographical area.

[0057] In this embodiment, the Canny operator is used to obtain edge points. The Canny operator is a well-known technology and will not be described in detail in this application. As other implementation methods, based on the ability to obtain all edge points at the road surface position height, implementers may use other existing technologies, such as the Sobel operator, etc. This application does not impose any special restrictions.

[0058] Furthermore, the closed edge contour at the height of the road surface is denoted as the ground contour. By analyzing the positional relationship between the closed edge contours at the height of the road surface and other heights, the contour surfaces in the target city are obtained. Specifically: the Euclidean distance between the center points of any closed edge contour at any height and the edges of closed contours at any adjacent height is denoted as the center distance. The minimum value of all center distances corresponding to any closed edge contour is calculated. When the minimum value is less than a preset threshold, the closed edge contour corresponding to the minimum value at any adjacent height is determined to belong to the same contour surface as the closed edge contour. When the projection of any closed edge contour on the X0Y plane is within the projection of any ground contour on the X0Y plane, the closed edge contour is determined to belong to the same contour surface as the ground contour. A schematic diagram of the contour surface is shown below. Figure 3 As shown, Figure 3 In the diagram, 3 represents the ground outline, and 4 represents the outline surface.

[0059] In this embodiment, the preset threshold value is 1.5m. The preset threshold value is preset by humans, and the implementer can set it according to the actual situation. This application does not impose any special restrictions.

[0060] Theoretically, the projection of a single wall of a building onto the X0Y plane should be a line. However, due to noise and acquisition accuracy, the "line" projected onto the X0Y plane by the point cloud data of a single wall has a certain width. Therefore, the point cloud data around each contour surface may also be the point cloud data of the building. Thus, taking the q-th contour surface as an example, we acquire the point cloud data outside the contour surface from the point cloud data of the target city, denoted as each candidate outer contour point cloud data of the contour surface. We calculate the minimum Euclidean distance between each candidate outer contour point cloud data and all point cloud data on the contour surface. Using the minimum value of all candidate outer contour point cloud data, we obtain a segmentation threshold for the minimum value of all candidate outer contour point cloud data through cross-validation. Candidate outer contour point cloud data whose minimum value is less than or equal to the segmentation threshold are used as the outer contour point cloud data of the q-th contour surface. The acquisition of outer contour point cloud data should not exceed the extension line of the ground contour. A schematic diagram of the outer contour point cloud data is shown below. Figure 4 As shown, Figure 4 In the diagram, 3 represents the ground contour, 5 represents the point cloud data within the contour surface, 6 represents the point cloud data outside the contour, and 7 represents the point cloud data outside the contour. For each contour surface, both the point cloud data within the contour surface and the point cloud data outside the contour are considered part of the point cloud data for that contour surface. If a single contour surface is the contour of a building, then the point cloud data on the contour surface, the point cloud data within the contour surface, and the point cloud data outside the contour are all considered part of the building's point cloud. The cross-validation method is a well-known technique and will not be elaborated upon in this application.

[0061] Step 2.2: Based on the spatial distribution of point cloud data within each contour surface, and combined with a clustering algorithm, obtain the classification of point cloud data within each contour surface; based on the quantity of point cloud data outside each contour surface, and the dispersion of point cloud data within each contour surface along the building height direction, and combined with the intra-class distance and inter-class distance of point cloud data within each contour surface, obtain the point cloud distribution feature value of each contour surface.

[0062] Furthermore, by analyzing the spatial distribution of point cloud data within each contour surface and combining it with a clustering algorithm, the classifications of point cloud data within each contour surface are obtained, specifically:

[0063] Using the 3D coordinates of the point cloud data within each contour surface as input, a clustering algorithm is employed to output the classification of the point cloud data within each contour surface. If the q-th contour surface is a terrain contour, the distribution of the point cloud data within the contour surface is relatively discrete, with a high degree of clustering in small areas and a discrete overall clustering. However, for building contours, the point cloud data within the contour surface typically represents the roof of the building, with similar overall heights and a uniform distribution.

[0064] Based on the above analysis, by using the number of point cloud data outside each contour surface and the dispersion of the point cloud data within each contour surface along the building height direction, combined with the intra-class distance and inter-class distance of the point cloud data within each contour surface, the point cloud distribution characteristic value of each contour surface is obtained, which is used to characterize the building point cloud features of each contour surface. The expression is:

[0065] ; This represents the point cloud distribution characteristic value of the q-th contour surface; This represents the number of point cloud data points outside the q-th contour surface; It represents the mean of the Euclidean distances between any two categories of point cloud data within the q-th contour surface; J represents the dispersion of the z-coordinate of the point cloud data within the q-th contour plane; J represents the number of classifications of the point cloud data within the q-th contour plane. This represents the average Euclidean distance between any two point cloud data points within the j-th category of the point cloud data on the q-th contour surface. Specifically, the Euclidean distance between the two categories is the Euclidean distance between the centers of the two categories.

[0066] In this embodiment, the dispersion is variance. As other implementation methods, based on the ability to measure the uneven distribution of the z-coordinate of point cloud data, implementers may use other existing technologies, such as standard deviation, coefficient of variation, etc. This application does not impose any special restrictions.

[0067] It should be noted that for the q-th contour surface, the more point cloud data outside the contour surface, the greater the dispersion of the z-coordinate of the point cloud data within the q-th contour surface, and the smaller the intra-class distance but the larger the inter-class distance of the point cloud data within the q-th contour surface, the more likely the point cloud data on the q-th contour surface, the point cloud data within the q-th contour surface, and the point cloud data outside the q-th contour surface are to belong to the building point cloud. Therefore, they should be preserved during the construction of the building white model.

[0068] Step 3: Obtain each similar contour surface by measuring the similarity between each contour surface and the ground contours of the other contour surfaces; for each contour surface and its similar contours, obtain the architectural feature similarity of each contour surface by measuring the spatial distance between the ground contours and the similarity between the closed edge contours at each height.

[0069] Because a certain number of old buildings, such as ancient buildings and tiled-roof houses, still exist in the city, their overall architectural design may be limited by the design methods of that time. Therefore, these buildings tend to have less regularity in form and differ significantly from the appearance of modern residential buildings. Simply calculating the point cloud distribution characteristic values ​​of each contour surface might misclassify the point cloud data of these buildings as point cloud data of topographical areas, leading to their exclusion. This ultimately results in low fidelity and reduced accuracy in the construction of the urban building white model, thus requiring further analysis.

[0070] For point clouds of buildings in cities, old buildings and new residential buildings typically form distinct zoning clusters. Old buildings are surrounded by other old buildings, and new residential buildings usually cluster within their respective areas, forming distinct building groups. The architectural structures, forms, and styles within each group are quite similar. Therefore, for point cloud data of old buildings, the point cloud features of a single building outline are quite similar to the point cloud features of the surrounding building outlines. Conversely, if there are specific topographical areas in a city, they are usually larger in area but more isolated, such as small hills or mountains, which are typically large in area but fewer in number within a given area.

[0071] Specifically, for point clouds of buildings, their ground contours are highly similar in shape to those of other surrounding ground contours, and their overall distribution is dense. Furthermore, the distribution characteristics of point clouds within the contour plane are highly similar to the distribution of point cloud data within other surrounding contour planes. In contrast, for the bottom contours of topographical areas, their distribution is more discrete, with very few similar ground contours. Moreover, due to differences in vegetation distribution and landform types, the distribution characteristics of point cloud data within the contour planes vary considerably across different terrains.

[0072] Based on the above analysis, similar contour surfaces are obtained by comparing the similarity of each contour surface with the ground contours of the other contour surfaces. Specifically:

[0073] The similarity coefficient of the ground contour between any two contour surfaces is obtained by contour matching algorithm. Contour surfaces whose normalized value of the similarity coefficient between each contour surface is greater than or equal to a preset threshold are regarded as similar contour surfaces.

[0074] In this embodiment, the contour matching algorithm is Fréchet distance, and the similarity coefficient is the reciprocal of the sum of the Fréchet distance between ground contours and a preset positive number. The calculation of Fréchet distance is a well-known technique, and will not be described in detail in this application. As other implementation methods, based on the ability to measure the similarity between ground contours, implementers may adopt other existing feasible techniques, such as Hausdorff distance, etc. This application does not impose any special restrictions.

[0075] In this embodiment, the Min-Max normalization method is used to obtain the normalized value of the similarity coefficient. The Min-Max normalization method is a well-known technique and will not be described in detail in this application.

[0076] In this embodiment, the preset threshold value is 0.95. The preset threshold value is preset by a person and can be adjusted by the implementer according to the actual accuracy requirements. This application does not impose any special restrictions.

[0077] Furthermore, for each contour surface and its similar contours, the architectural feature similarity of each contour surface is obtained by using the spatial distance between ground contours and the similarity between closed edge contours at each height. This similarity is used to characterize the authenticity of each contour surface as a building, and the expression is:

[0078] In the formula, This represents the similarity of architectural features of the q-th contour surface; This represents the number of similar contour surfaces to the q-th contour surface; calculate the similarity coefficient of the closed edge contours at each height between the q-th contour surface and its p-th similar contour surface. This represents the sum of similarity coefficients of the closed edge contours at all heights between the q-th contour surface and its p-th similar contour surface; This represents the average Euclidean distance between the q-th contour surface and all its similar contour surfaces. The Euclidean distance between two ground contours is calculated by calculating the Euclidean distance between the center point cloud data of the two ground contours.

[0079] It should be noted that the smaller the mean Euclidean distance between the q-th contour surface and its similar contour surfaces, and the higher the approximation of the closed edge contours at each height, the more likely the point cloud data on the q-th contour surface, the point cloud data within the q-th contour surface, and the point cloud data outside the q-th contour surface are to belong to the building point cloud. The higher the realism of the building features, the more they should be preserved during the construction of the building white model.

[0080] Step 4: Obtain the retainability of each contour surface by comparing the point cloud distribution feature value with the building feature similarity, and then filter the building point cloud from the point cloud data of the target city using the retainability.

[0081] Furthermore, the retainability of each contour surface is obtained by comparing the point cloud distribution feature values ​​of each contour surface with the building feature similarity, expressed as:

[0082] In the formula, This indicates the retainability of the q-th profile surface; This represents the point cloud distribution characteristic value of the q-th contour surface; This represents the architectural feature similarity of the q-th contour surface.

[0083] It should be noted that: the higher the point cloud distribution feature value of the q-th contour surface, and the greater the realism of the architectural features, the more likely the point cloud data on the q-th contour surface, the point cloud data within the q-th contour surface, and the point cloud data outside the q-th contour surface are to belong to architectural point clouds. This means they have a stronger feature representation capability in the architectural white model construction process and should be retained. A schematic diagram of the process for obtaining retainability is shown below. Figure 5 As shown.

[0084] Furthermore, by using cross-validation to obtain the segmentation threshold for the retainability of all contour surfaces, the point cloud data on each contour surface, the point cloud data within the q-th contour surface, and the point cloud data outside the q-th contour surface are determined to belong to the building point cloud. Conversely, if the retainability of each contour surface, the point cloud data within the q-th contour surface, and the point cloud data outside the q-th contour surface are determined not to belong to the building point cloud, contour surfaces with a retainability greater than or equal to the segmentation threshold are selected, and the point cloud data on the selected contour surfaces, the point cloud data within the contour surfaces, and the point cloud data outside the contour surfaces are combined to form the building point cloud.

[0085] Step 5: Construct a white model of the building based on the selected building point cloud.

[0086] The white model is constructed using the filtered building point cloud, following these steps:

[0087] (1) Building outline extraction: Using the three-dimensional Hough transform algorithm, the edges or outlines of buildings are extracted by detecting the straight lines and planar features in the building point cloud.

[0088] (2) Building height extraction: By analyzing the elevation information of the building point cloud, the elevation values ​​of the top and bottom of the building are extracted, that is, the building height is obtained.

[0089] (3) White model construction: Based on the outline and height information of the building, use parametric modeling tools such as AutoCAD or Revit to quickly generate the white model of the building.

[0090] The detailed process of constructing a white model using the selected building point cloud is a well-known technique and will not be elaborated upon in this application.

[0091] In summary, this application, by analyzing the positional relationship of closed edge contours at different heights, can accurately extract contour surfaces, thereby better reflecting the actual shape and structure of the building and improving the fidelity of the white model. By extracting the outer contour point cloud data, it can more accurately reflect the actual boundary of the building, ensuring the integrity of the building point cloud and avoiding modeling errors caused by missing boundary points. By considering the differences in distribution characteristics between the point cloud data of the building and the point cloud data of the terrain area, and calculating the point cloud distribution characteristic value, it can quantify the probability that each contour surface and its subordinate point cloud data belong to the building point cloud, which helps to more accurately determine which point cloud data belong to the building point cloud.

[0092] Furthermore, considering the existence of old buildings in the target city, the similarity of closed edge contours at different heights is analyzed. Combined with the point cloud distribution feature values, the retainability of each contour surface is calculated, which can more accurately assess the degree of each contour surface and its subordinate point cloud data of building point clouds. By filtering building point clouds through retainability, the architectural features of the contour surfaces can be more comprehensively evaluated, and the point cloud data of actual urban buildings and topography can be distinguished. This allows for the removal of topography point clouds from the urban point cloud data, improving the accuracy of the filtered building point clouds, and thus improving the fidelity of the subsequently constructed white model of buildings.

[0093] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. In some alternative implementations, the functions marked in the blocks may occur in a different order than that shown in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different blocks may also occur in a different order than disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. Each block in a block diagram and / or flowchart, and combinations of blocks in a block diagram and / or flowchart, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.

[0094] It will be apparent to those skilled in the art that this application is not limited to the details of the exemplary embodiments described above, and that this application can be implemented in other specific forms without departing from its essential characteristics. Therefore, the embodiments described above should be considered exemplary and non-limiting in all respects.

Claims

1. A rapid construction method for high-fidelity architectural white models, characterized in that, The method includes the following steps: Acquire point cloud data of the target city and label the elevation of the road surface location in the point cloud data; By analyzing the distribution of point cloud data in the target city, we obtain the closed edge contours at each height along the building height direction in the point cloud data. The closed edge contour at the height where the road surface is located is recorded as the ground contour. By analyzing the positional relationship between the closed edge contours at the height where the road surface is located and the heights of other heights, we obtain the contour surfaces in the target city. By analyzing the distance between the point cloud data of the target city and each contour surface, we obtain the outer point cloud data of each contour surface. By analyzing the spatial distribution of point cloud data within each contour surface and combining it with clustering algorithms, we obtain the classification of point cloud data within each contour surface. By analyzing the quantity of point cloud data outside each contour surface and the dispersion of point cloud data within each contour surface along the building height direction, combined with the intra-class distance and inter-class distance of point cloud data within each contour surface, we obtain the point cloud distribution feature values ​​of each contour surface. By comparing the ground contours of each contour surface with the other contour surfaces, we obtain the similar contour surfaces of each contour surface; for each contour surface and its similar contour surfaces, we obtain the architectural feature similarity of each contour surface by comparing the spatial distance between the ground contours and the similarity between the closed edge contours at each height. By comparing the similarity between the point cloud distribution feature values ​​and the building features, the retainability of each contour surface is obtained. Based on the retainability, building point clouds are filtered from the point cloud data of the target city. A white model of the building is constructed based on the filtering results. The process for obtaining the point cloud distribution feature values ​​is as follows: Calculate the mean Euclidean distance between any two categories of point cloud data within each contour plane; Calculate the average Euclidean distance between any two point cloud data within each category, and calculate the sum of the average values ​​for all categories of point cloud data within each contour surface; Calculate the product of the quantity and the mean; Calculate the sum of the dispersion and the sum; The point cloud distribution feature value can be further obtained by the product and the accumulated value; The point cloud distribution characteristic value is the ratio of the product to the accumulated value; The process of obtaining the similarity of the building features is as follows: Obtain the mean Euclidean distance between each contour surface and all its similar contour surfaces on the ground contour. Calculate the similarity coefficient of the closed edge profile at each height between each profile surface and each of its similar profile surfaces; calculate the sum of the similarity coefficients of the closed edge profile at all heights between each profile surface and all its similar profile surfaces. The similarity of architectural features can be further determined by the mean calculation result and the summation. The building feature similarity is the ratio of the sum to the mean. The retention rate is the product of the point cloud distribution feature value and the building feature similarity.

2. The rapid construction method for a high-fidelity architectural white model as described in claim 1, characterized in that, The process of obtaining the contour surface is as follows: The Euclidean distance between the center points of any closed edge contour at any height and each closed contour edge at any adjacent height is denoted as the center distance. The minimum value of all the center distances corresponding to any closed edge contour is calculated. When the minimum value is less than a preset threshold, it is determined that the closed edge contour corresponding to the minimum value at any adjacent height belongs to the same contour surface as the closed edge contour. If the projection of any closed edge contour onto the X0Y plane of the ground coordinate system lies within the projection of any ground contour onto the X0Y plane, then the closed edge contour and the ground contour belong to the same contour surface.

3. The rapid construction method for a high-fidelity architectural white model as described in claim 1, characterized in that, The process of acquiring the outer point cloud data is as follows: Obtain the point cloud data outside any contour surface from the point cloud data of the target city, and denot them as candidate out-of-contour point cloud data. Calculate the minimum value of the Euclidean distance between each candidate out-of-contour point cloud data and all point cloud data on the contour surface. Using the minimum value of all candidate out-of-contour point cloud data, obtain the segmentation threshold of the minimum value of all candidate out-of-contour point cloud data by cross-validation. The candidate out-of-contour point cloud data whose minimum value is less than or equal to the segmentation threshold are taken as the out-of-contour point cloud data of the contour surface.

4. The rapid construction method for a high-fidelity architectural white model as described in claim 1, characterized in that, The process of obtaining the similar contour surfaces of each contour surface is as follows: The similarity coefficient of the ground contour between any two contour surfaces is obtained by contour matching algorithm; the contour surfaces whose normalized value of the similarity coefficient between each contour surface is greater than or equal to a preset threshold are regarded as similar contour surfaces.

5. The rapid construction method for a high-fidelity architectural white model as described in claim 1, characterized in that, The method for filtering the building point cloud is as follows: By using the retainability of all contour surfaces, cross-validation is employed to obtain the segmentation threshold for the retainability of all contour surfaces. Contour surfaces with a retainability greater than or equal to the segmentation threshold are selected, and the point cloud data on the selected contour surfaces, the point cloud data within the contour surfaces, and the point cloud data outside the contour surfaces are combined to form the building point cloud.

Images