Lightweight BIM model construction method for urban civil engineering quantity calculation

Abstract

The present application relates to the technical field of three-dimensional modeling, and particularly relates to a lightweight BIM model construction method for urban civil engineering quantity calculation, which solves the technical problem of large engineering quantity calculation error caused by loss of complex region details and data redundancy of simple region due to fixed size downsampling in the prior art. The method comprises the following steps: obtaining a point cloud data set of a urban civil engineering region; determining the neighborhood complexity of each point cloud according to the point cloud data set, and performing region segmentation on the point cloud data according to the neighborhood complexity to obtain a plurality of sub-regions; the neighborhood complexity is used to represent the geometric feature complexity of the surrounding of the point cloud; performing dimension reduction processing on the point cloud data of the sub-region according to the neighborhood complexity of the plurality of point clouds in the sub-region to obtain lightweight point cloud data; and constructing a lightweight building information model of the urban civil engineering region according to the lightweight point cloud data.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, specifically to a lightweight BIM model construction method for calculating urban civil engineering quantities. Background Technology

[0002] Building Information Modeling (BIM) has the ability to automatically extract quantities of engineering works, which can promote the construction of smart construction sites and the development of smart cities. When using BIM models to calculate urban civil engineering quantities, lightweight BIM models are usually constructed by reducing the sampling amount in order to improve calculation efficiency.

[0003] Because civil engineering projects involve a wide geographical area and diverse structural types, with significant differences in structural characteristics across different regions, high-density point cloud data is required for complex structures (such as doors, windows, beam-column connections, and other areas with significant curvature changes) to ensure minimal errors in subsequent construction simulations and quantity calculations. Conversely, for simpler structures (such as walls and floors with flat surfaces), the point cloud data density can be appropriately reduced to minimize data redundancy and meet the lightweight requirements of BIM models. However, current technologies for constructing lightweight BIM models typically divide the point cloud space into fixed-size 3D grids for downsampling, failing to adaptively adjust based on regional structural complexity. This results in oversampling of complex areas, leading to the loss of crucial detail information, and undersampling of simple areas, generating substantial data redundancy. Ultimately, this affects the accuracy of urban civil engineering quantity calculations, making it difficult to meet the precision and efficiency requirements of practical engineering applications. Summary of the Invention

[0004] To address the technical problem in existing technologies where fixed-size downsampling leads to the loss of details in complex areas and data redundancy in simple areas, resulting in significant errors in quantity calculations, this invention aims to provide a lightweight BIM model construction method for urban civil engineering quantity calculations. The specific technical solution adopted is as follows:

[0005] Firstly, a lightweight BIM model construction method for calculating urban civil engineering quantities is provided, comprising: acquiring a point cloud data set of an urban civil engineering area; determining the neighborhood complexity of each point cloud based on the point cloud data set, and dividing the point cloud data into multiple sub-regions based on the neighborhood complexity; the neighborhood complexity is used to characterize the geometric feature complexity around the point cloud; performing dimensionality reduction processing on the point cloud data of the sub-regions based on the neighborhood complexity of multiple point clouds in the sub-regions to obtain lightweight point cloud data; and constructing a lightweight building information model of the urban civil engineering area based on the lightweight point cloud data.

[0006] Based on the above technical solution, in the lightweight BIM model construction method for urban civil engineering quantity calculation provided by this invention, after acquiring point cloud data, adaptive region segmentation and dimensionality reduction of the point cloud are achieved based on neighborhood complexity, replacing the traditional fixed-size voxel filtering downsampling method. This allows areas with complex geometric features in urban civil engineering (such as beam-column nodes) to avoid errors in quantity calculation by retaining sufficient point cloud details, while areas with simple geometric features (such as walls) reduce redundancy due to data dimensionality reduction. The resulting lightweight BIM model not only ensures the accuracy of civil engineering quantity calculation but also improves the efficiency of model construction and calculation, effectively overcoming the technical problems of lost details in complex areas and data redundancy in simple areas in the prior art.

[0007] In conjunction with the first aspect above, in one possible implementation, the method for determining the neighborhood complexity of each point cloud based on the point cloud dataset specifically includes: constructing a neighborhood of each point cloud based on the distribution characteristics of its neighboring point clouds; and determining the neighborhood complexity based on the spatial distribution characteristics of multiple point clouds within the neighborhood.

[0008] In conjunction with the first aspect mentioned above, in one possible implementation, the method for constructing the neighborhood of each point cloud based on the distribution characteristics of its neighboring point clouds specifically includes: determining the Euclidean distance between each point cloud and multiple neighboring point clouds, with each point cloud as the center; sorting the point clouds in ascending order according to their Euclidean distances, and determining the initial neighborhood radius of each point cloud based on the Euclidean distances corresponding to the first preset number of neighboring point clouds; determining the target neighborhood radius of each point cloud based on its initial neighborhood radius, a preset density scaling factor, and a preset minimum radius constraint; and using the location of each point cloud as the center of a sphere, the spherical region formed by the target neighborhood radius is taken as the neighborhood of each point cloud.

[0009] In conjunction with the first aspect mentioned above, in one possible implementation, the method for determining neighborhood complexity based on the spatial distribution characteristics of multiple point clouds within the neighborhood specifically includes: extracting feature values ​​of multiple point cloud data within the neighborhood using a preset feature value analysis algorithm; using the feature values ​​to characterize the spatial distribution characteristics of the multiple point clouds; and obtaining the neighborhood complexity by calculating curvature entropy based on the feature values ​​of the multiple point cloud data within the neighborhood.

[0010] In conjunction with the first aspect above, in one possible implementation, the aforementioned preset eigenvalue analysis algorithm is principal component analysis, and the eigenvalues ​​include multiple dimensions; the aforementioned method for obtaining neighborhood complexity by calculating curvature entropy based on the eigenvalues ​​of multiple point cloud data in the neighborhood specifically includes: normalizing the eigenvalues ​​of multiple dimensions; and calculating curvature entropy based on the normalized eigenvalues ​​of multiple dimensions as neighborhood complexity.

[0011] In conjunction with the first aspect mentioned above, in one possible implementation, the method for dividing point cloud data into multiple sub-regions based on neighborhood complexity specifically includes: determining a neighborhood complexity difference threshold based on the maximum or minimum value of the neighborhood complexity of multiple point clouds; starting from any unvisited point cloud, performing region growth based on the neighborhood complexity difference threshold to obtain multiple initial sub-regions; the region growth includes merging point clouds whose neighborhood complexity difference with the starting point is less than the neighborhood complexity difference threshold into one sub-region; and for the multiple initial sub-regions, merging initial sub-regions with a number of point clouds less than a preset point cloud number threshold into adjacent initial sub-regions to obtain multiple sub-regions.

[0012] In conjunction with the first aspect above, in one possible implementation, the method further includes: determining the local point cloud density based on the Euclidean distance between point clouds within the initial sub-region; determining a first quantity threshold based on a preset density safety factor, the local point cloud density, and a preset minimum structure area; determining a second quantity threshold based on the local point cloud density and the minimum sub-region area among multiple initial sub-regions; and determining the larger of the first quantity threshold and the second quantity threshold as the preset point cloud quantity threshold.

[0013] In conjunction with the first aspect mentioned above, in one possible implementation, the method for reducing the dimensionality of point cloud data in a sub-region based on the neighborhood complexity of multiple point clouds in the sub-region to obtain lightweight point cloud data specifically includes: centralizing the point cloud data of each sub-region; determining the number of lightweight dimensions for each sub-region based on the neighborhood complexity of the point cloud in each sub-region, as well as a preset minimum number of retained dimensions and a preset maximum number of retained dimensions; and projecting the centralized point cloud data of each sub-region onto a low-dimensional space corresponding to the number of lightweight dimensions to obtain the dimensionality-reduced point cloud data.

[0014] In conjunction with the first aspect mentioned above, in one possible implementation, the method for constructing a lightweight building information model for urban civil engineering areas based on lightweight point cloud data specifically includes: classifying multiple sub-regions into high-complexity, medium-complexity, and low-complexity regions based on the average neighborhood complexity of the point cloud in each sub-region; if the sub-region is a high-complexity region, performing surface reconstruction based on the point cloud data to generate a triangular mesh model; if the sub-region is a low-complexity region, generating parametric planar components based on the point cloud data; if the sub-region is a medium-complexity region, performing cylindrical fitting based on the point cloud data to generate parametric cylindrical components; and combining the triangular mesh model, parametric planar components, and parametric cylindrical components to form a lightweight building information model.

[0015] In conjunction with the first aspect above, in one possible implementation, the method for obtaining a point cloud data set of urban civil engineering areas specifically includes: collecting raw point cloud data of urban civil engineering areas by using a drone equipped with a lidar device; preprocessing the raw point cloud data to obtain a point cloud data set; the preprocessing includes performing statistical filtering on the raw point cloud data to remove noise points that deviate from the normal distribution range.

[0016] Secondly, a lightweight BIM model building device for calculating urban civil engineering quantities is provided, comprising: a processor and a storage medium; the storage medium includes instructions, and the processor is used to execute the instructions to perform the actions described in the first aspect and any possible implementation thereof. This lightweight BIM model building device for calculating urban civil engineering quantities can be an electronic device or a chip within an electronic device.

[0017] Thirdly, a computer-readable storage medium is provided, in which instructions are stored, which, when executed on a lightweight BIM model building device for calculating urban civil engineering quantities, cause the lightweight BIM model building device for calculating urban civil engineering quantities to perform the actions described in the first aspect and any possible implementation thereof.

[0018] Fourthly, a computer program product containing instructions is provided, which, when run on a lightweight BIM model building device for calculating urban civil engineering quantities, causes the lightweight BIM model building device for calculating urban civil engineering quantities to perform the actions described in the first aspect and any possible implementation thereof.

[0019] The present invention has the following beneficial effects:

[0020] After acquiring point cloud data, adaptive region segmentation and dimensionality reduction of the point cloud are achieved based on neighborhood complexity, replacing the traditional fixed-size voxel filtering downsampling method. This allows areas with complex geometric features in urban civil engineering (such as beam-column nodes) to avoid errors in engineering quantity calculations by retaining sufficient point cloud details, while areas with simple geometric features (such as walls) reduce redundancy due to data dimensionality reduction. The resulting lightweight BIM model not only ensures the accuracy of civil engineering quantity calculations but also improves the efficiency of model construction and calculation, effectively overcoming the technical problems of lost details in complex areas and data redundancy in simple areas in existing technologies. Attached Figure Description

[0021] To more clearly illustrate the technical solutions and advantages in the embodiments of the present invention 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 the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A flowchart illustrating a lightweight BIM model construction method for calculating urban civil engineering quantities, provided as an embodiment of the present invention;

[0023] Figure 2 A flowchart illustrating a lightweight BIM model construction method for calculating urban civil engineering quantities, provided as an embodiment of the present invention.

[0024] Figure 3 A flowchart illustrating a lightweight BIM model construction method for calculating urban civil engineering quantities, provided as an embodiment of the present invention.

[0025] Figure 4 A flowchart illustrating a lightweight BIM model construction method for calculating urban civil engineering quantities, provided as an embodiment of the present invention.

[0026] Figure 5 A flowchart illustrating a lightweight BIM model construction method for calculating urban civil engineering quantities, provided as an embodiment of the present invention.

[0027] Figure 6 This is a schematic diagram of the hardware structure of a lightweight BIM model building device for calculating urban civil engineering quantities, provided as an embodiment of the present invention. Detailed Implementation

[0028] To further illustrate the technical means and effects adopted by the present invention to achieve its intended purpose, the following, in conjunction with the accompanying drawings and preferred embodiments, details the specific implementation, structure, features, and effects of a lightweight BIM model construction method for urban civil engineering quantity calculation proposed according to the present invention. In the following description, different "one embodiment" or "another embodiment" do not necessarily refer to the same embodiment. Furthermore, specific features, structures, or characteristics in one or more embodiments can be combined in any suitable form.

[0029] 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 invention pertains.

[0030] The following description, in conjunction with the accompanying drawings, details a specific scheme for a lightweight BIM model construction method for calculating urban civil engineering quantities provided by the present invention.

[0031] Please see Figure 1 This document illustrates a flowchart of a lightweight BIM model construction method for calculating urban civil engineering quantities, provided by an embodiment of the present invention. The method includes:

[0032] S1. Obtain the point cloud data set of the urban civil engineering area.

[0033] In some implementations, drones equipped with lidar devices collect raw point cloud data of urban construction areas. The raw point cloud data is then preprocessed to obtain a point cloud dataset. This preprocessing includes statistical filtering of the raw point cloud data to remove noise points that deviate from the normal distribution range.

[0034] Specifically, drones follow a gridded or zigzag flight path to repeatedly collect data on urban civil engineering areas (including building structures, ancillary facilities, and surrounding terrain) in multiple rounds to obtain comprehensive and dense raw point cloud data.

[0035] The raw point cloud data contains a large number of three-dimensional coordinate points, which are affected by moving pedestrians, vehicles, surrounding vegetation, atmospheric scattering, and the inherent errors of the lidar sensor, resulting in noisy points that deviate from the actual civil engineering structure. Therefore, statistical filtering is performed on the raw point cloud data. By analyzing the point cloud distribution characteristics, noisy points that deviate from the normal distribution range are removed, ultimately yielding a point cloud dataset that accurately reflects the geometric structure of the urban civil engineering area.

[0036] Point cloud data includes the coordinates of multiple dimensions of the point cloud.

[0037] S2. Determine the neighborhood complexity of each point cloud based on the point cloud dataset, and divide the point cloud data into regions based on the neighborhood complexity to obtain multiple sub-regions.

[0038] Neighborhood complexity is used to characterize the geometric feature complexity of the point cloud's surrounding area.

[0039] In some implementations, each point cloud in the point cloud dataset is taken as the center, and the neighborhood is determined by a neighborhood search method with a preset number of nearest neighbor points. Then, principal component analysis is used to extract the feature values ​​of the point cloud in the local neighborhood. After normalization, the curvature entropy is calculated to obtain the neighborhood complexity, thereby achieving accurate differentiation of the complexity of the geometric features around each point cloud.

[0040] Subsequently, the neighborhood complexity of all point clouds is calculated to determine the threshold for neighborhood complexity difference. Unvisited point clouds are used as seed points to form initial sub-regions through region growth. A preset point cloud quantity threshold is determined by combining the preset density safety factor, local point cloud density, preset minimum structure area, and minimum sub-region area. Initial sub-regions with a point cloud quantity less than this threshold are merged. Finally, multiple sub-regions that accurately match the characteristics of urban civil engineering structures are obtained, providing a foundation for subsequent targeted dimensionality reduction processing.

[0041] S3. Based on the neighborhood complexity of multiple point clouds in the sub-region, perform dimensionality reduction on the point cloud data of the sub-region to obtain lightweight point cloud data.

[0042] In some implementations, the point cloud data of each sub-region is first centered to eliminate interference from different coordinate dimensions caused by units or offsets, ensuring the accuracy of subsequent dimensionality reduction analysis. Then, the average neighborhood complexity of all point clouds within each sub-region is calculated as an indicator of the overall geometric feature complexity of that sub-region. This is combined with a preset minimum number of retained dimensions (the minimum feature dimensions required for the simplest structure, ensuring no loss of basic geometric information) and a preset maximum number of retained dimensions (the feature dimensions required for the most complex structure, ensuring key details are preserved). Based on the relative relationship between the average neighborhood complexity of the sub-region and the maximum and minimum neighborhood complexities of all point clouds, the target dimensionality reduction dimension for that sub-region is determined. The higher the average neighborhood complexity of the sub-region (the more complex the geometric features), the closer the target dimensionality reduction dimension is to the preset maximum number of retained dimensions; conversely, the lower the average neighborhood complexity (the simpler the geometric features), the closer the target dimensionality reduction dimension is to the preset minimum number of retained dimensions.

[0043] Finally, a low-dimensional projection is performed on the centralized sub-region point cloud data to obtain lightweight point cloud data. The dimensionality-reduced point cloud data of all sub-regions are integrated to form the overall lightweight point cloud data. This data significantly reduces data redundancy while preserving the key geometric features of the civil engineering structure.

[0044] S4. Based on lightweight point cloud data, construct a lightweight building information model for urban civil engineering areas.

[0045] In some implementations, lightweight point cloud data, being essentially a discrete set of 3D coordinate points, lacks the structured engineering unit division, parametric geometric features (such as component dimensions and material properties), and topological relationships between components (such as connection logic and spatial constraints) required for BIM modeling. BIM components, however, can classify discrete points into standardized units based on the functional attributes of civil engineering (such as walls and beams / columns) through fitting and reconstruction, assigning them explicit geometric parameters (such as thickness and radius) and engineering attributes, and establishing spatial relationships between components. This allows subsequent modeling to form a structured BIM model that conforms to engineering logic and can be used for quantity calculations and construction analysis, avoiding model chaos, information loss, and functional failure caused by directly using point clouds. Therefore, it is necessary to first convert lightweight point cloud data into BIM components, and then construct a lightweight building information model that can fully represent the structural characteristics of urban civil engineering areas based on the spatial coordinate parameters of the components.

[0046] Based on the above technical solution, after acquiring point cloud data, adaptive region segmentation and dimensionality reduction of the point cloud are achieved based on neighborhood complexity, replacing the traditional fixed-size voxel filtering downsampling method. This allows areas with complex geometric features in urban civil engineering (such as beam-column nodes) to avoid errors in engineering quantity calculations by retaining sufficient point cloud details, while areas with simple geometric features (such as walls) reduce redundancy due to data dimensionality reduction. The resulting lightweight BIM model not only ensures the accuracy of civil engineering quantity calculations but also improves the efficiency of model construction and calculation, effectively overcoming the technical problems of lost details in complex areas and data redundancy in simple areas in existing technologies.

[0047] In one possible implementation, combining Figure 1 ,like Figure 2 As shown, the method for determining the neighborhood complexity of each point cloud based on the point cloud dataset in S2 above can be specifically implemented through the following S21 to S22, which are explained in detail below:

[0048] S21. Construct the neighborhood of each point cloud based on the distribution characteristics of its neighboring point clouds.

[0049] In some implementations, the Euclidean distances between each point cloud and multiple neighboring point clouds are first determined, with each point cloud as the center. Then, these point clouds are sorted in ascending order of Euclidean distance, and the initial neighborhood radius of each point cloud is determined based on the Euclidean distances of the first preset number of neighboring point clouds. Based on the initial neighborhood radius, a preset density scaling factor, and a preset minimum radius constraint, the target neighborhood radius of each point cloud is determined. The spherical region formed by the target neighborhood radius and the location of each point cloud is then considered the neighborhood of each point cloud.

[0050] Specifically, firstly, from the point cloud dataset, taking each independent point cloud as the center, we traverse the surrounding point clouds. Using a preset spatial distance calculation method, we determine the Euclidean distance (in meters) between the center point cloud and each other point cloud. Point clouds with a distance less than a preset distance threshold are defined as neighboring point clouds. The Euclidean distance between the center point cloud and neighboring point clouds can quantitatively characterize the spatial density between them. The preset distance threshold can be determined based on experience in different scenarios. For example, for fine point clouds of urban buildings (such as high-precision LiDAR scanning), a value of 0.05-0.1m is used, which is suitable for local point clouds of regular components such as walls and floors, ensuring that the neighborhood contains enough detailed points without cross-component association. For regular point clouds of urban buildings (such as ordinary resolution scanning), a value of 0.1-0.2m is used, which is suitable for medium-complexity components such as beams, columns, and pipes, balancing the number of neighborhood points and computational efficiency. For large-scale civil engineering scenarios (such as factories and municipal facilities), a value of 0.2-0.5m is used, which is suitable for large-scale, low-density point cloud areas, ensuring that the neighborhood contains enough point clouds to support feature calculation.

[0051] Next, the calculated Euclidean distances between all neighboring point clouds and the central point cloud are sorted in ascending order of numerical value. Based on the overall density characteristics and local geometric feature extraction requirements of the urban civil engineering area point cloud data, a preset number of neighboring points is set (this preset number must ensure that it covers enough point clouds to reflect the local geometric shape, avoiding the loss of local features due to too few points, or the introduction of irrelevant point clouds to interfere with feature extraction due to too many points; for example, the preset number can be 10). The k-nearest neighbor (KNN) algorithm is used to select 10 neighboring point clouds as the core neighboring points of the central point cloud.

[0052] In some implementations, the average Euclidean distance between the core neighboring points and the central point cloud is first taken to obtain the average neighborhood distance of the point cloud as the initial neighborhood radius. The average neighborhood distance is used to quantitatively characterize the density of points around the point cloud.

[0053] Subsequently, considering the uneven distribution of point clouds in urban civil engineering areas (e.g., dense point clouds in complex structures and sparse point clouds in simple structures), a preset density scaling factor and a preset minimum radius constraint are introduced. The preset density scaling factor dynamically adjusts the neighborhood size based on local point cloud density. For example, it appropriately reduces the neighborhood range in dense point cloud areas to avoid including too many redundant point clouds, and appropriately expands the neighborhood range in sparse point cloud areas to ensure sufficient feature point clouds are included. The preset minimum radius constraint prevents the initial neighborhood radius in sparse areas from being too small, resulting in insufficient point cloud data within the neighborhood and inability to effectively extract geometric features.

[0054] Next, based on the average neighborhood distance Preset density scaling factor and preset minimum radius constraint Calculate the target neighborhood radius :

[0055]

[0056] In the formula, the density scaling factor (Empirical value of 4) is used to calculate the average neighborhood distance. Scale adjustments are made to transform local density features into a basic scaling value for the neighborhood radius. If local points are extremely dense ( (small value) This ensures that the neighborhood is not excessively shrunk, covering sufficient detailed point clouds to guarantee the accuracy of engineering calculations; local point sparseness ( (High value) This avoids excessive expansion of the neighborhood and avoids including too many redundant point clouds, thus achieving lightweight design.

[0057] minimum radius (Empirical value of 0.04) This provides a safety net to ensure the effectiveness of the neighborhood in sparse regions. Too small, plus This ensures that the neighborhood radius is at least 0.04, preventing subsequent complexity calculations from failing due to an excessively small neighborhood.

[0058] In other implementations, the initial neighborhood radius of the center point cloud can be directly used as the Euclidean distance between the core neighboring points and the center point cloud. This initial neighborhood radius can then just cover the core neighboring points, initially forming a basic neighborhood range that reflects the local point cloud distribution. The initial neighborhood radius is calculated with a preset density scaling factor and compared with a preset minimum radius constraint; the larger of the two values ​​is taken as the target neighborhood radius of the center point cloud. If the calculated initial neighborhood radius is greater than the preset minimum radius constraint, the result is used as the target neighborhood radius; if the calculated initial neighborhood radius is less than or equal to the preset minimum radius constraint, the preset minimum radius constraint is used as the target neighborhood radius. This ensures that the target neighborhood radius can adapt to the local point cloud density while also meeting the minimum point cloud quantity requirement for feature extraction.

[0059] Finally, using the spatial location of each central point cloud as the center of a sphere and the radius of the determined target neighborhood as the radius, a spherical region is constructed, which is the neighborhood of the central point cloud.

[0060] S22. Determine the neighborhood complexity based on the spatial distribution characteristics of multiple point clouds within the neighborhood.

[0061] In some implementations, feature values ​​of multiple point cloud data within a neighborhood are first extracted using a pre-defined feature value analysis algorithm. These feature values ​​characterize the spatial distribution features of the multiple point clouds. Then, based on the feature values ​​of the multiple point cloud data within the neighborhood, the neighborhood complexity is obtained by calculating the curvature entropy.

[0062] Furthermore, if the preset eigenvalue analysis algorithm is principal component analysis (PCA), the eigenvalues ​​include multiple dimensions. It is necessary to first normalize the eigenvalues ​​of multiple dimensions, and then calculate the curvature entropy based on the normalized eigenvalues ​​of multiple dimensions as the neighborhood complexity.

[0063] Specifically, let's take the example of taking three dimensions for the feature values ​​(corresponding to the three-dimensional coordinate data of the point cloud):

[0064] Principal component analysis is used to extract the 3D coordinate data of all points in the neighborhood of each point cloud and perform centering processing (i.e., the coordinates of each point are subtracted from the mean of the coordinates of all points in the neighborhood) to eliminate the influence of coordinate offset on feature extraction.

[0065] A covariance matrix is ​​constructed based on the centered point set. This matrix reflects the dispersion and correlation of the point set in each dimension of three-dimensional space. The three eigenvalues ​​of the covariance matrix are solved through eigenvalue decomposition. ≥ ≥ ≥0, where the magnitude of the eigenvalue directly characterizes the degree of dispersion of the point set along the corresponding principal component direction. If ≥ ≈ ≈0 indicates that the point set is approximately linearly distributed (such as the region along the axis of a pipe); if ≈ ≥ ≈0 indicates that the point set is approximately planar (e.g., a wall area); if ≈ ≈ This indicates that the point set is approximately spherically distributed (such as complex regions like beam-column nodes). Therefore, these three eigenvalues ​​together constitute a quantitative indicator of the spatial distribution characteristics of the point cloud within the neighborhood.

[0066] Subsequently, curvature entropy is calculated based on the eigenvalues ​​to determine the neighborhood complexity. Curvature entropy quantifies the geometric complexity of the neighborhood into a continuous value. The larger the entropy value, the more irregular and detailed the neighborhood distribution in three-dimensional space, i.e., the higher the neighborhood complexity; conversely, the lower the complexity. Therefore, curvature entropy can serve as an effective measure of neighborhood complexity.

[0067] Since the magnitudes of the three eigenvalues ​​may differ due to variations in point cloud density, they must first be normalized.

[0068]

[0069] In the formula, and Indicates the first The eigenvalue and the eigenvalue .

[0070] Used for the sum of eigenvalues. If =0 (theoretically, this only holds true when there is only one point in the neighborhood, but in practical applications, it can be avoided by using the minimum radius constraint), so directly assigning the neighborhood complexity is 0.

[0071] Indicates the first The normalization result of each eigenvalue can reflect the relative differences between the eigenvalues.

[0072] Next, the curvature entropy is calculated based on the normalized eigenvalues. :

[0073]

[0074] In the formula, the uniformity of the feature value distribution is quantified based on the information entropy formula. The more uniform the feature value distribution (complex region, such as spherical or irregular curved surface), the larger the c value; the more concentrated the feature value distribution (simple region, such as plane or line), the smaller the c value, ultimately realizing the quantitative expression of the geometric complexity of the neighborhood.

[0075] Finally, output a point cloud dataset with an appended neighborhood complexity label. ,Right now:

[0076]

[0077] in, Indicates the first Point cloud data; Indicates the first Neighborhood complexity of a point cloud; This represents the total amount of point cloud data.

[0078] Based on the above technical solution, by constructing a neighborhood adapted to the local point cloud distribution, and quantifying the geometric complexity of the neighborhood based on the feature value analysis and curvature entropy calculation of the point cloud within the neighborhood, it is possible to accurately distinguish between complex structures (such as beam-column nodes) and simple structures (such as walls) in urban civil engineering areas. This provides a quantitative basis for subsequent adaptive dimensionality reduction, preservation of key details, and lightweighting for areas of different complexity. It effectively solves the problem of difficulty in balancing the accuracy and efficiency of point cloud processing caused by the lack of regional complexity differentiation in existing technologies, ensuring that the subsequent lightweight BIM model construction retains the details required for engineering quantity calculation while reducing redundant data through dimensionality reduction.

[0079] In one possible implementation, combining Figure 2 ,like Figure 3As shown, the method in S2 above for segmenting point cloud data into multiple sub-regions based on neighborhood complexity can be specifically implemented through the following S23 to S25, which are explained in detail below:

[0080] S23. Determine the neighborhood complexity difference threshold based on the maximum and minimum values ​​of the neighborhood complexity of multiple point clouds.

[0081] First, calculate the maximum value of the neighborhood complexity. (The point cloud corresponding to the most geometrically complex feature in the civil engineering area, such as the point cloud at beam-column joints) and minimum value (This corresponds to the simplest point cloud with the best geometric features, such as the point cloud over a flat wall surface). Then, combining the distribution pattern of the complexity of urban civil engineering point clouds, an empirical formula is used to determine the threshold for the difference in neighborhood complexity. :

[0082]

[0083] In the formula, The coefficient is an empirical coefficient with a value range of [0.1, 0.3]. (If the coefficient is too small, it may lead to overly fine region segmentation, increasing the burden of subsequent processing; if the coefficient is too large, it may lead to excessive region merging, resulting in the loss of key geometric details.) This threshold is used to determine whether the difference in neighborhood complexity between point clouds is within an acceptable range, providing a quantitative judgment standard for subsequent complexity-based point cloud region segmentation.

[0084] S24. Starting from any unvisited point cloud, perform region growth based on the neighborhood complexity difference threshold to obtain multiple initial sub-regions.

[0085] Region growing involves merging point clouds whose neighborhood complexity difference from the starting point is less than a neighborhood complexity difference threshold into a sub-region.

[0086] First, from the point cloud dataset with added neighborhood complexity labels In the middle, randomly select a point cloud that has not been marked as "visited". This serves as the starting point for region growing, and its neighborhood complexity is extracted. Simultaneously, the point cloud is marked as "visited" to avoid repeated selections later.

[0087] Starting point Centered on a point cloud, spatial indexing techniques are used to retrieve all unvisited point clouds in its vicinity (the retrieval range can be based on a pre-defined neighborhood or a preset spatial distance threshold to ensure coverage of potentially similarly complex regions). For each retrieved unvisited point cloud, its neighborhood complexity is extracted. The computational complexity of the neighborhood of the starting point. absolute value of the difference Then take the absolute value of the difference. Less than the neighborhood complexity difference threshold Point clouds The point cloud to be merged is determined to be a point cloud to be merged. From the starting point They are grouped into the same initial sub-region. The region of belonging Recorded as:

[0088]

[0089] At the same time, the point clouds to be merged Marked as "visited" and with To establish a new center, repeat the above steps, continuously searching for unvisited point clouds until no new unvisited point clouds that meet the criteria can be included in the sub-region, thus obtaining an initial sub-region.

[0090] Finally, after the growth process of an initial sub-region terminates, if there are still unvisited point clouds in the point cloud dataset, another unvisited point cloud is randomly selected as a new seed point, and the above region growth process is repeated until all point clouds are marked as "visited," ultimately resulting in multiple independent initial sub-regions. The difference in the neighborhood complexity of the point clouds within each initial sub-region is less than [value missing]. Furthermore, the difference in point cloud neighborhood complexity between sub-regions is not less than .

[0091] S25. For multiple initial sub-regions, merge the initial sub-regions with a point cloud quantity less than a preset point cloud quantity threshold into adjacent initial sub-regions to obtain multiple sub-regions.

[0092] In some implementations, the preset point cloud quantity threshold can be a fixed value, or it can be a dynamic threshold determined based on the area size and point cloud distribution of multiple initial sub-regions, as well as a preset density safety factor and a preset minimum structure area. Specifically, this includes:

[0093] First, the local point cloud density is determined based on the Euclidean distances between point clouds within the initial sub-region. Specifically, the mean of the Euclidean distances between point clouds within the initial sub-region is calculated, and then the reciprocal of the square of the mean is approximated as the local point cloud density.

[0094] It should be noted that point cloud density refers to the number of points per unit area, reflecting the density of point clouds in a local region. The mean Euclidean distance between point clouds within the initial sub-region is also considered. This is used to quantize the average interval of local point clouds. In 3D space, if a spherical neighborhood is constructed with the mean value as the radius, since the surface scan point cloud is distributed near the object surface, it can be considered to be mainly distributed near the cross-section of the sphere through the center (the center of the sphere is the center of the initial sub-region), and its area is... If the number of point clouds in the neighborhood is N, then the theoretical area density is... Therefore, the theoretical area density of the cross-section passing through the center of the sphere is considered as the local point cloud density. To simplify the calculation (or when the number of local point clouds N is relatively stable), the constant term can be ignored. The influence of N, using It approximates the density trend. The smaller (the denser the dots). The smaller, the reciprocal The larger the value, the higher the density. The larger the value (the sparser the dots). The larger it is, the more its reciprocal The smaller the value, the lower the density. Based on the above analysis, the reciprocal of the square of the mean is a simplified quantification method for local point cloud density. It intuitively reflects the density trend of the point cloud through the reciprocal of the square of the distance, and is efficient and reasonable in engineering applications.

[0095] Secondly, considering the size requirements of the smallest identifiable structures in urban civil engineering projects (such as wall corners, small pipes, etc.), a preset minimum structural area (e.g., 0.1㎡ for a brick, representing the projected area of ​​the smallest structure on a two-dimensional plane) and a preset density safety factor (within a range of [1.2, 1.5], used to ensure that the point cloud covers structural details) are set. Then, based on the preset density safety factor... Local point cloud density and preset minimum structural area Determine the first quantity threshold. :

[0096]

[0097] Ensure the first quantity threshold It can meet the lower limit of the number of point clouds required for minimum structural feature extraction.

[0098] Simultaneously, the areas of all initial sub-regions are calculated (taking the projected area of ​​the smallest bounding box of each region on the plane), and the smallest sub-region area is selected. This is based on the local point cloud density. and the smallest subregion area among multiple initial subregions Determine the second quantity threshold :

[0099]

[0100] Ensure the second quantity threshold It is adapted to the number of point clouds in the smallest region under the current point cloud distribution density.

[0101] Finally, the larger of the first and second quantity thresholds is determined as the preset point cloud quantity threshold, which ensures both the point cloud requirement for covering the minimum structure and the actual distribution density of the local point cloud.

[0102] Furthermore, all initial sub-regions are traversed, and the number of point clouds in each region is counted. If the number of point clouds in an initial sub-region is greater than or equal to its corresponding preset point cloud number threshold, then the region is retained as a sub-region to be optimized; if the number of point clouds is less than the preset threshold, it is determined to be an excessively small region and needs to be merged.

[0103] For excessively small regions, spatial indexing is used to retrieve all adjacent initial sub-regions (judged by the minimum bounding boxes of the two regions having spatial overlap or a distance less than a preset adjacency distance, such as 0.1 meters). The difference in the mean neighborhood complexity between the excessively small region and each adjacent region is calculated (i.e., the absolute value of the average neighborhood complexity of the excessively small region and the average neighborhood complexity of the adjacent regions). The adjacent region with the smallest difference is selected as the merging target, and all point clouds of the excessively small region are merged into this adjacent region. The boundary of the merged region (recalculating the minimum bounding box) and the mean neighborhood complexity are then updated.

[0104] Repeat the above merging process until the number of point clouds in all sub-regions is not less than their corresponding preset point cloud number thresholds. Finally, multiple spatially contiguous sub-regions with sufficient point cloud numbers and consistent internal neighborhood complexity characteristics are obtained, and the sub-region set is output. ( (Indicates the number of regions).

[0105] Based on the above technical solution, by determining the threshold of neighborhood complexity difference, generating initial sub-regions through region growth and merging excessively small regions, the precise partitioning of urban civil engineering point clouds according to geometric complexity is achieved. This ensures that the neighborhood complexity within each sub-region is consistent and that the number of point clouds meets the requirements for feature extraction. This avoids the subsequent processing burden caused by overly fine region division and prevents the loss of key geometric details due to excessively small region size. It provides a structurally reasonable and information-complete point cloud unit for subsequent adaptive dimensionality reduction for regions of different complexity, preservation of details required for engineering quantity calculation, and lightweighting. This effectively solves the problem of balancing accuracy and efficiency in point cloud processing in existing technologies.

[0106] In one possible implementation, combining Figure 1 ,like Figure 4 As shown, the method in S3 above can be specifically implemented through the following steps S31 to S33, which are explained in detail below:

[0107] S31. Centralize the point cloud data of each sub-region.

[0108] In some implementations, the mean of the point cloud data for each sub-region in three dimensions is calculated to obtain the geometric center coordinates of that sub-region. Then, the geometric center coordinates are subtracted from the three-dimensional coordinates of each point cloud to obtain the centered point cloud data.

[0109] Through the above processing, the point cloud data in the sub-region will be redistributed around the geometric center, eliminating the interference of coordinate system offset (such as overall offset to a certain coordinate quadrant) during the acquisition of the original point cloud on subsequent operations such as covariance calculation and feature vector extraction. This ensures that subsequent dimensionality reduction processing for different sub-regions depends only on the relative spatial distribution characteristics of the point cloud, rather than the absolute coordinate position, laying the foundation for unified feature analysis and model construction.

[0110] S32. Determine the number of lightweight dimensions for each sub-region based on the neighborhood complexity of the point cloud in each sub-region, as well as the preset minimum and maximum number of retained dimensions.

[0111] In some implementations, the average neighborhood complexity of all point clouds within the p-th sub-region is first calculated. It serves as a comprehensive indicator characterizing the geometric complexity of the entire sub-region. The larger the value, the more complex the point cloud distribution within the sub-region (such as the beam-column node region). The smaller the value, the more regular the geometric features of the sub-region (such as the wall surface or pipe axis area).

[0112] At the same time, set a preset minimum number of retained dimensions. and the preset maximum number of dimensions to retain .in, A value of 1 can be used to adapt to one-dimensional linear structures (such as pipe axes). Retaining the length dimension is sufficient to characterize its geometric features; only the first principal component (the dimension with the largest variance) needs to be retained. Point cloud data only uses three-dimensional coordinates for PCA dimensionality reduction, therefore... The value can be set to 3, with 3 principal components as the upper limit. This can cover the key geometric details of complex 3D structures (such as beam-column nodes and irregular curved surfaces), ensuring that the point cloud in highly complex regions still retains sufficient information after dimensionality reduction. Both values ​​together limit the effective range of the number of lightweight dimensions, ensuring that dimensionality reduction does not excessively lose key features.

[0113] Then calculate the number of lightweight dimensions for the p-th sub-region. :

[0114]

[0115] In the formula, and This represents the maximum and minimum neighborhood complexity of the point cloud. In practical applications, scenarios involve various structures, and the neighborhood complexity of the point cloud will definitely have multiple different values. Therefore, the difference between the maximum and minimum values ​​is not zero.

[0116] This value is used to normalize the complexity of the current sub-region, obtaining its relative proportion in the global complexity distribution, and is mapped to [0, 1]. The lower the complexity, the closer this value is to 0; the higher the complexity, the closer it is to 1.

[0117] Using the total span of dimensions Multiply by the relative proportion of complexity to quantify the number of dimensions that should be added due to the difference in complexity. The higher the complexity, the more dimensions.

[0118] Minimum dimensionality Based on this, the aforementioned complexity-driven dimension increments are added to obtain the initial adaptive dimension.

[0119] pass Rounding the result down to an integer ensures the discreteness of the number of dimensions, which meets the actual requirement in engineering that the dimensions must be integers.

[0120] Ultimately, the number of lightweight dimensions for each sub-region was limited to [ , Within the interval, and adapted to the geometric complexity of the sub-region.

[0121] S33. Project the point cloud data after centering each sub-region onto a low-dimensional space corresponding to the number of lightweight dimensions to obtain the dimensionality-reduced point cloud data.

[0122] In some implementations, for each sub-region that has undergone centralization, the centralized point cloud dataset for that sub-region is extracted. Dimensionality reduction is then performed using principal component analysis, taking into account the reduced dimensionality of the sub-region. First, a covariance matrix is ​​constructed based on the centralized point cloud dataset. This matrix reflects the dispersion and correlation of the point cloud across G dimensions (i.e., the initial number of dimensions of the point cloud). Then, eigenvalue decomposition is used to solve for the G eigenvalues ​​of the covariance matrix (the number of eigenvalues ​​is consistent with the number of dimensions) and the corresponding G eigenvectors. The magnitude of each eigenvalue represents the variance contribution along the corresponding principal component direction.

[0123] Then, select the previous The eigenvectors corresponding to the largest eigenvalues ​​are composed of dimensions G× The projection matrix is ​​then calculated. Finally, the G-dimensional point cloud vector of each centered point cloud data is multiplied by the projection matrix to obtain the projection matrix of each point cloud. The projected coordinates in dimensional space, and the set of all projected coordinates is the point cloud data of the sub-region after dimensionality reduction.

[0124] Through the above operations, the dimensionality of point cloud data was reduced from G dimensions to [missing value]. Dimension, in preserving the original data, is derived from the previous While retaining the main spatial distribution characteristics of each principal component, the data size was significantly reduced. Among these, high-complexity sub-regions... Larger dimensions (close to the preset maximum retention dimension) can retain more details, while lower complexity sub-regions... A smaller (close to the preset minimum retained dimension) size enables deep lightweighting.

[0125] It should be understood that dimensionality reduction is only used for feature extraction and data compression, but lightweight point cloud data still retains three-dimensional coordinate information.

[0126] Based on the above technical solution, the point clouds of each sub-region are centralized to eliminate coordinate offset interference. The appropriate number of lightweight dimensions is adaptively determined by the neighborhood complexity of each sub-region (more dimensions are retained for complex regions to preserve geometric details, while fewer dimensions are used for simple regions to achieve deep compression). The centralized point cloud is then projected onto the corresponding low-dimensional space to complete dimensionality reduction. These three processes work together to achieve accurate and lightweight point cloud data processing. This approach maximizes the preservation of the core spatial distribution characteristics of each sub-region, meeting the key information required for subsequent BIM model construction and quantity calculations, while significantly reducing redundant data, effectively balancing the accuracy and efficiency of point cloud processing.

[0127] In one possible implementation, combining Figure 1 ,like Figure 5 As shown, the method in S4 above can be specifically implemented through the following steps S41 to S45, which are explained in detail below:

[0128] S41. Based on the average neighborhood complexity of the point cloud in each sub-region, classify the multiple sub-regions into high-complexity regions, medium-complexity regions, and low-complexity regions.

[0129] In some implementations, the distribution range is based on the global neighborhood complexity. , Set the first threshold. Second threshold (in < < < Threshold calculation can employ a proportional division method, taking into account the actual distribution characteristics of point cloud complexity in urban civil engineering projects (the common proportions of high, medium, and low complexity regions), and setting the following:

[0130]

[0131]

[0132] Among them, 0.3 and 0.7 are empirical proportional coefficients, which can be adjusted according to the structural complexity of specific engineering scenarios (such as industrial plants, civil buildings, etc.) to ensure that the classification results fit the actual component characteristics.

[0133] Then, the average neighborhood complexity of each sub-region is compared with two thresholds to perform level classification. If ≥ Then the p-th sub-region is classified as a highly complex region, corresponding to component regions with complex geometric features in the civil engineering scenario (such as beam-column joints, equipment foundations, irregular curved surface components, etc.); if < < Then the p-th sub-region is classified as a medium-complex region, corresponding to component regions with a certain degree of curvature or local details (such as curved walls, pipe bends, stair treads, etc.); if ≤ If the p-th sub-region is classified as a low-complexity region, it corresponds to a component region with regular geometric features (such as flat wall surfaces, straight pipe axes, floor slab planes, etc.).

[0134] S42. If the sub-region is a highly complex region, perform surface reconstruction based on the point cloud data to generate a triangular mesh model.

[0135] Specifically, for sub-regions within highly complex regions, point cloud data (including point cloud coordinates and normal vectors) is extracted. First, an indicator function gradient field is constructed to transform the discrete point cloud p into a continuous vector field. :

[0136]

[0137] In the formula, Indicates the first The point cloud coordinates are used to indicate the spatial location of each point.

[0138] Represents the smoothing kernel function. The relative positions are smoothly weighted, with points that are closer to each other contributing more and points that are farther apart contributing less, thus achieving a smooth transition of the point cloud and ensuring the continuity of the reconstructed surface.

[0139] Indicates the first The normal vectors of the point cloud are used to guide the orientation and gradient direction of the surface, ensuring that the reconstructed surface is consistent with the geometric normal features of the point cloud.

[0140] Represents the neighborhood of the i-th point cloud. The size of is the number of points in the spherical neighborhood of the i-th point cloud.

[0141] This is used to perform neighborhood weighting on the normal vector of the i-th point cloud. The more neighborhood points there are, the lower the weight, balancing the influence of the normal vectors of local dense and sparse points, so that the gradient field can more accurately reflect the overall surface trend.

[0142] This represents the total number of point clouds in highly complex regions. The contributions of all highly complex point clouds to the current point p are accumulated. The gradient field vector of the point is generated by combining the normal vector and spatial location information, and finally the discrete point cloud is transformed into a continuous vector field.

[0143] Then, the Poisson equation is solved using iterative methods or the finite element method:

[0144]

[0145] In the formula, This represents the divergence of a vector field. For the Laplace operator.

[0146] This indicates an indicator function used to describe the probability that each point in space belongs to the interior (x=1) or exterior of an object, and its isosurface is the surface of the object to be reconstructed.

[0147] Finally, using the indicator function x as input, an isosurface extraction algorithm is employed to generate triangular patches at the isosurface of x. These patches are then combined to output a triangular network model, which is a triangular mesh model of highly complex regions. This model can fully preserve the geometric details of complex structures such as beam-column nodes and irregular curved surfaces.

[0148] S43. If the sub-region is a low-complexity region, generate parameterized planar components based on the point cloud data.

[0149] In some implementations, point cloud data (with geometric features and a small number of lightweight dimensions) is extracted from sub-regions of low-complexity areas. The least squares method is used to fit the plane: the plane equation is defined, and the coordinates of all point clouds within the region are substituted to construct an error function.

[0150]

[0151] In the formula, This represents the total number of point clouds in low-complexity regions.

[0152] , , Let be the coordinates of the i-th point cloud in the three-dimensional coordinate system.

[0153] Let be the general equation of the plane, and a, b, c, d be the parameters of the general equation of the plane. For the i-th point cloud within the region, The absolute value can be understood as a measure of the deviation of that point from the plane.

[0154] The sum of squares of the deviation measures for all point clouds is taken, which is the error function E. Its core function is to quantify the overall deviation between the point cloud set and the plane to be fitted.

[0155] By taking partial derivatives with respect to a, b, c, and d and setting the error function to zero, we obtain the plane parameters and determine the mathematical expression for the parameterized plane.

[0156] Based on this expression, parametric planar components (such as wall surfaces, floor slabs, etc.) are generated. Their size and position are determined by the extension range of the planar space and the topological relationship, which meets the engineering quantity calculation requirements of regular planar components.

[0157] S44. If the sub-region is of medium complexity, perform cylindrical fitting based on the point cloud data to generate a parameterized cylindrical component.

[0158] In some implementations, point cloud data (with cylindrical geometric features) is extracted from sub-regions within a medium-complex region. A cylindrical fitting algorithm is used: given the equation of the cylinder's axis and radius, the coordinates of all point clouds within the medium-complex region are substituted into the geometric constraint that the distance from the point to the axis is equal to the radius. Then, by minimizing the distance error from the point cloud to the cylinder surface, the axis parameters and radius are obtained, thus determining the mathematical model of the parameterized cylinder.

[0159] Based on this model, parameterized cylindrical components (such as pipes, cylindrical columns, etc.) are generated, and their parameters such as length and radius are directly used for engineering quantity calculations (such as pipe material usage, concrete usage).

[0160] S45, combining triangular mesh models, parametric planar components, and parametric cylindrical components to form a lightweight building information model.

[0161] The BIM environment is used to import triangular mesh models of highly complex areas, parametric planar components of low-complexity areas, and parametric cylindrical components of medium-complexity areas. These components are then assembled based on spatial topological relationships (such as wall-to-floor connections and pipe-to-beam / column intersections). The triangular mesh model retains complex details to ensure accuracy, while the parametric planar and cylindrical components are lightweight to improve efficiency, ultimately forming a lightweight building information model containing all civil engineering components. Based on this model, components such as walls, pipes, and beam-column nodes can be automatically identified, and quantities of concrete, pipes, and steel structures can be calculated based on their parameters (such as planar area, cylinder volume, and triangular mesh surface area), achieving a complete application from point cloud lightweighting to quantity calculation.

[0162] It should be noted that the generation of BIM components based on their complexity is determined by the contradiction between the geometric differences of different structures in urban civil engineering areas, the requirements for modeling accuracy, and the goal of lightweighting. A unified approach cannot simultaneously meet the core requirements of detail preservation, efficiency improvement, and engineering adaptation.

[0163] From the perspective of civil engineering structural characteristics, high-complexity areas (such as beam-column joints and door-window connections) have irregular geometric shapes and large curvature variations. It is necessary to preserve the surface details completely to ensure the accuracy of engineering quantity calculations (such as accurate statistics of concrete usage at joints). If a uniform method (such as plane fitting) is used, key details will be lost directly. Medium- and low-complexity areas (such as walls, floors, and cylindrical columns) have regular geometric shapes (planes and columns). It is only necessary to extract core parameters (such as wall thickness and cylinder radius) to meet the modeling requirements. If a uniform complex reconstruction method (such as Poisson surface reconstruction) is used, a large number of redundant triangular faces will be generated, which violates the goal of lightweighting.

[0164] From the perspective of modeling efficiency and engineering practicality, generating components using different methods can balance accuracy and efficiency: For highly complex areas, Poisson surface reconstruction generates triangular mesh models, accurately reproducing irregular structures; for medium- to low-complexity areas, planar / cylindrical fitting generates parametric components, quickly outputting standardized geometric parameters (such as wall elevation and cylinder centerlines). Furthermore, parametric components are more suitable for subsequent applications of the BIM model (e.g., component size parameters can be directly called during quantity calculations, and spatial coordinates can be quickly matched during collision detection). If a uniform method is adopted, either excessive simplification leads to insufficient accuracy in complex areas, or excessive refinement results in data redundancy in simple areas. This fails to meet the varying accuracy requirements of civil engineering for different structures, reduces modeling efficiency, and increases model size, contradicting the original intention of lightweight BIM model construction.

[0165] Based on the above technical solution, sub-regions are classified into high, medium, and low complexity regions according to the average complexity of their neighborhoods. For high complexity regions, Poisson surface reconstruction is used to generate triangular mesh models to preserve details. For low complexity regions, parametric planar components are generated, and for medium complexity regions, cylindrical fitting is used to generate parametric cylindrical components to achieve efficient modeling. Finally, these components are combined to form a lightweight building information model, which not only accurately restores the geometric features of various urban civil engineering areas (complete details in complex areas and efficient modeling in regular areas), but also automatically identifies components and calculates engineering quantities based on the model. This achieves lightweight and accurate full-process modeling from point cloud processing to engineering application, effectively balancing model accuracy and construction efficiency.

[0166] It should be noted that the order of the above embodiments of the present invention is merely for descriptive purposes and does not represent the superiority or inferiority of the embodiments. The processes depicted in the accompanying drawings do not necessarily require a specific or sequential order to achieve the desired result. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0167] The various embodiments in this specification are described in a progressive manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0168] In this embodiment of the invention, the lightweight BIM model building device for calculating urban civil engineering quantities can be divided into functional units according to the above method example. For example, each function can be divided into separate functional units, or two or more functions can be integrated into one processing unit. The integrated unit can be implemented in hardware or software. It should be noted that the unit division in this embodiment is illustrative and only represents one logical functional division; other division methods may be used in actual implementation.

[0169] This invention also provides a schematic diagram of the hardware structure of a lightweight BIM model building device for calculating urban civil engineering quantities. (See attached diagram.) Figure 6 The lightweight BIM model building device 600 for calculating urban civil engineering quantities includes a processor 601, and optionally, a memory 602 connected to the processor 601.

[0170] In the first possible implementation, see Figure 6 The lightweight BIM model building device 600 for calculating urban civil engineering quantities also includes a transceiver 603. The processor 601, memory 602, and transceiver 603 are connected via a bus. The transceiver 603 is used to communicate with other devices or communication networks. Optionally, the transceiver 603 may include a transmitter and a receiver. The device in the transceiver 603 that implements the receiving function can be considered as a receiver, which is used to perform the receiving steps in the embodiments of the present invention. The device in the transceiver 603 that implements the transmitting function can be considered as a transmitter, which is used to perform the transmitting steps in the embodiments of the present invention.

[0171] Based on the first possible implementation method Figure 6 The structural diagram shown can be used to illustrate the structure of the lightweight BIM model building device for calculating urban civil engineering quantities involved in the above embodiments.

[0172] in, Figure 6 The diagram can also illustrate the system chip in a lightweight BIM model building device used for calculating urban civil engineering quantities. In this case, the actions performed by the aforementioned lightweight BIM model building device for calculating urban civil engineering quantities can be implemented by this system chip. The specific actions performed are described above and will not be repeated here.

[0173] In implementation, each step of the method provided in this embodiment can be completed by integrated logic circuits in the processor or by instructions in software form. The steps of the method disclosed in this embodiment can be directly manifested as being executed by a hardware processor, or being executed by a combination of hardware and software modules in the processor.

[0174] The processor in this invention may include, but is not limited to, at least one of the following: a central processing unit (CPU), a microprocessor, a digital signal processor (DSP), a microcontroller unit (MCU), or an artificial intelligence processor, etc., which are various computing devices that run software. Each computing device may include one or more cores for executing software instructions to perform calculations or processing. The processor may be a standalone semiconductor chip or integrated with other circuits into a single semiconductor chip. For example, it may be integrated with other circuits (such as encoding / decoding circuits, hardware acceleration circuits, or various bus and interface circuits) to form a System-on-a-Chip (SoC), or it may be integrated as a built-in processor within an ASIC. The ASIC with the integrated processor may be packaged separately or together with other circuits. In addition to the cores for executing software instructions to perform calculations or processing, the processor may further include necessary hardware accelerators, such as field-programmable gate arrays (FPGAs), programmable logic devices (PLDs), or logic circuits that implement dedicated logic operations.

[0175] The memory in the embodiments of the present invention may include at least one of the following types: read-only memory (ROM) or other types of static storage devices capable of storing static information and instructions; random access memory (RAM) or other types of dynamic storage devices capable of storing information and instructions; or electrically erasable programmable read-only memory (EEPROM). In some scenarios, the memory may also be a compact disc read-only memory (CD-ROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but is not limited thereto.

[0176] This invention also provides a computer-readable storage medium including instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0177] This invention also provides a computer program product containing instructions that, when run on a computer, cause the computer to perform any of the methods described above.

[0178] This invention also provides a chip, which includes a processor and an interface circuit. The interface circuit is coupled to the processor. The processor is used to run computer programs or instructions to implement the above-described method. The interface circuit is used to communicate with other modules outside the chip.

[0179] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented using software programs, implementation can be, in whole or in part, in the form of a computer program product. This computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of the present invention are generated. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, computer instructions can be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium can be any available medium accessible to a computer or a data storage device containing one or more servers, data centers, etc., that can be integrated with the medium. The available media can be magnetic media (e.g., floppy disks, hard disks, magnetic tapes), optical media (e.g., DVDs), or semiconductor media (e.g., solid-state disks (SSDs)).

[0180] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings and the disclosure, will understand and implement other variations of the disclosed embodiments in carrying out the claimed invention. In this invention, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. A single processor or other unit can implement several of the functions listed in this invention.

[0181] Although the invention has been described in conjunction with specific features and embodiments, it is obvious that various modifications and combinations can be made therein without departing from the spirit and scope of the invention. Accordingly, this specification and drawings are merely illustrative of the invention and are to be considered as covering any and all modifications, variations, combinations, or equivalents within the scope of the invention. Clearly, those skilled in the art can make various alterations and modifications to the invention without departing from its spirit and scope. Thus, if such modifications and modifications of the invention fall within the scope of the invention and its equivalents, the invention is also intended to include such modifications and modifications.

Claims

1. A lightweight BIM model construction method for urban civil engineering quantity calculation, characterized in that, The method comprises the following steps: acquiring a point cloud data set of a city civil engineering area; determining a neighborhood complexity of each point cloud according to the point cloud data set, and regionally segmenting the point cloud data according to the neighborhood complexity to obtain a plurality of sub-regions; the neighborhood complexity is used to represent the geometric feature complexity of the neighborhood of the point cloud; determining the neighborhood complexity of each point cloud comprises: determining the Euclidean distance of each point cloud from a plurality of adjacent point clouds with each point cloud as the center; sorting the Euclidean distances in ascending order, and determining the initial neighborhood radius of each point cloud according to the Euclidean distances of the first preset number of adjacent point clouds; determining the target neighborhood radius of each point cloud according to the initial neighborhood radius of each point cloud, a preset density scaling factor and a preset minimum radius constraint; taking the position of each point cloud as the center of a sphere, and taking the spherical region formed by the target neighborhood radius as the neighborhood of each point cloud; extracting feature values of a plurality of point cloud data in the neighborhood through a preset feature value analysis algorithm; the feature values are used to represent the spatial distribution characteristics of the plurality of point clouds; and obtaining the neighborhood complexity by calculating the curvature entropy according to the feature values of the plurality of point cloud data in the neighborhood; wherein the preset feature value analysis algorithm is principal component analysis, and the feature values include a plurality of dimensions; obtaining the neighborhood complexity by calculating the curvature entropy according to the feature values of the plurality of point cloud data in the neighborhood comprises: normalizing the feature values of the plurality of dimensions; and calculating the curvature entropy according to the normalized feature values of the plurality of dimensions as the neighborhood complexity; performing dimension reduction processing on the point cloud data of the sub-region according to the neighborhood complexity of the plurality of point clouds in the sub-region to obtain lightweight point cloud data; constructing a lightweight building information model of the city civil engineering area according to the lightweight point cloud data. 2.The method of claim 1, wherein, regionally segmenting the point cloud data according to the neighborhood complexity to obtain a plurality of sub-regions comprises: determining a neighborhood complexity difference threshold according to the maximum and minimum values of the neighborhood complexity of the plurality of point clouds; taking any unvisited point cloud as a starting point, and performing region growing according to the neighborhood complexity difference threshold to obtain a plurality of initial sub-regions; the region growing comprises merging point clouds with a neighborhood complexity difference less than the neighborhood complexity difference threshold into one sub-region; for the plurality of initial sub-regions, merging an initial sub-region with a point cloud number less than a preset point cloud number threshold to an adjacent initial sub-region to obtain the plurality of sub-regions. 3.The method of claim 2, wherein, The method further comprises the following steps: determining a local point cloud density according to the Euclidean distance between point clouds in the initial sub-region; determining a first number threshold according to a preset density safety factor, the local point cloud density and a preset minimum structure area; determining a second number threshold according to the local point cloud density and the minimum sub-region area in the plurality of initial sub-regions; determining the larger one of the first number threshold and the second number threshold as the preset point cloud number threshold. 4.The method of claim 1, wherein, performing centering processing on the point cloud data of each sub-region; ​ According to the neighborhood complexity of the point cloud in each sub-region, and the preset minimum reserved dimension number and the preset maximum reserved dimension number, a lightweight dimension number of each sub-region is determined; The point cloud data of each sub-region after the centering processing is projected to a low-dimensional space corresponding to the lightweight dimension number, to obtain reduced point cloud data. 5.The method of claim 1, wherein, According to the lightweight point cloud data, a lightweight building information model of the urban civil engineering area is constructed, including: According to the average value of the neighborhood complexity of the point cloud in each sub-region, the plurality of sub-regions are classified into high complexity regions, medium complexity regions and low complexity regions; If the sub-region is a high complexity region, surface reconstruction is performed according to the point cloud data to generate a triangular mesh model; If the sub-region is a low complexity region, a parameterized planar component is generated according to the point cloud data; If the sub-region is a medium complexity region, cylindrical fitting is performed according to the point cloud data to generate a parameterized cylindrical component; The triangular mesh model, the parameterized planar component and the parameterized cylindrical component are combined to form the lightweight building information model. 6.The method of claim 1, wherein, A point cloud data set of an urban civil engineering area is obtained, including: Raw point cloud data of the urban civil engineering area is collected by a laser radar device carried by a drone; The raw point cloud data is preprocessed to obtain the point cloud data set; the preprocessing includes statistical filtering processing of the raw point cloud data to remove noise points deviating from the normal distribution range.

Images