BIM-based building model data accurate conversion processing method

Abstract

The invention relates to the technical field of building three-dimensional data processing, in particular to a BIM-based building model data accurate conversion processing method. According to the method, the local density feature and the local curvature of each voxel point in the three-dimensional point cloud are analyzed, the influence degree of construction site noise on each voxel point is obtained based on the two features, and the filling necessity of each empty voxel block is obtained based on the overall influence degree distribution of neighborhood non-empty voxel blocks in a second neighborhood range of the empty voxel blocks. Whether the empty voxel blocks need to be filled or not is judged according to the filling necessity, and then a final filling result is obtained for BIM model construction and data conversion. According to the method, the finally constructed BIM building model is more accurate, the accuracy and adaptability of the model construction process under the interference of construction site environmental factors are improved, and data conversion is more accurate.

Description

Technical Field

[0001] This invention relates to the field of architectural 3D data processing technology, specifically to a method for accurate conversion and processing of BIM-based architectural model data. Background Technology

[0002] Currently, Building Information Modeling (BIM) technology is widely used in the construction industry. It enables comprehensive digital modeling, management, and collaboration of buildings, serving as a core element for project collaboration and intelligent construction. Its core lies in converting design models (such as Revit and ArchiCAD) into site-level data suitable for construction management, cost accounting, and schedule simulation. The BIM modeling process involves model building, data conversion, and information addition. During model building or updating, point cloud data serves as the foundation for these processes. Octree downsampling is an efficient, practical, and widely used preprocessing technique for processing 3D point clouds. By performing regular sampling in 3D space, it significantly reduces data complexity while effectively preserving the structural information of the building structure, providing a reliable data foundation for subsequent BIM modeling, deformation monitoring, ancient building protection, and digital twin applications. Octree downsampling, by simplifying and homogenizing massive point clouds, serves as a crucial preprocessing step, greatly optimizing the data conversion process from real-world scenarios to BIM construction site building models.

[0003] However, for construction sites, due to the special nature of construction, there are a lot of dust, specular reflection, and noise during construction. These factors can cause scattering and occlusion on the laser beam path of the laser scanner during the 3D point cloud acquisition process, thereby reducing scanning accuracy and affecting the quality and accuracy of point cloud data. This can lead to discontinuous point cloud data and missing data, which in turn affects the accuracy of octree downsampling, resulting in an incomplete BIM model and affecting the accuracy of the data conversion process from the real scene to the BIM construction site building model. Summary of the Invention

[0004] To address the technical problem that the acquisition of 3D point clouds of construction sites is affected by site noise, leading to incomplete octree downsampling results and impacting the final BIM model data conversion process, this invention aims to provide a precise BIM-based method for converting and processing building model data. The specific technical solution adopted is as follows: This invention proposes a method for accurate conversion and processing of BIM-based building model data, the method comprising: A laser scanner is used to acquire a 3D point cloud of the construction site, and the initial sampling result of the 3D point cloud after octree downsampling is obtained. The initial sampling result includes empty voxel blocks and non-empty voxel blocks. For each voxel in the 3D point cloud, the local density feature of each voxel is obtained based on the distance between the voxel and neighboring voxels within a preset first neighborhood range; the local curvature of the voxel is obtained based on the spatial distribution relationship between the voxel and its neighboring voxels; and the degree of influence of construction site noise on each voxel is obtained based on the local density feature and the local curvature. The influence degree of each voxel point in the non-empty voxel block is statistically analyzed to obtain the overall influence degree of each non-empty voxel block; for any empty voxel block, the overall influence degree of the neighboring non-empty voxel blocks within the preset second neighborhood range is statistically analyzed, and the filling necessity of each empty voxel block is obtained by combining the distance between the empty voxel block and the neighboring non-empty voxel blocks. Determine whether empty voxels need to be filled based on the filling necessity; use the final filling result as an optimized octree network; and use the optimized octree network to construct the BIM model and perform data conversion.

[0005] Furthermore, the method for obtaining the local density features includes: For a target voxel, obtain the Euclidean distance between the target voxel and each neighboring voxel within the first neighborhood, as well as the difference in laser reflection intensity; obtain the first distribution disorder of the laser reflection intensity difference; obtain the second distribution disorder of the Euclidean distance; obtain the neighborhood information distribution disorder based on the first and second distribution disorder; perform negative correlation mapping and normalization on the neighborhood information distribution disorder to obtain the local density feature.

[0006] Furthermore, the first distribution disorder is the result of normalizing the standard deviation of the laser reflection intensity difference.

[0007] Furthermore, the method for obtaining the second distribution disorder includes: Based on the Euclidean distance, a clustering algorithm is used to classify voxel points within the first neighborhood to obtain multiple voxel point categories. In each voxel point category, the corresponding Euclidean distance is negatively correlated to obtain a first distance weight, and the standard deviation of the first distance weight is used as the initial distribution disorder of the voxel point category. The cumulative value of the initial distribution disorder of all voxel point categories is used as the second distribution disorder.

[0008] Furthermore, the method for obtaining the local curvature includes: The voxel category to which the target voxel belongs is taken as the target category. For voxels in the target category, the curvature obtained by the PCA-based three-dimensional spatial curvature algorithm is taken as the local curvature.

[0009] Furthermore, the method for obtaining the overall degree of influence includes: For each voxel, after negatively correlated mapping of the local density features, the product of the mapping result and the local curvature is used as the degree of influence of the voxel. The average influence of all voxel points in the non-empty voxel block is taken as the overall influence.

[0010] Furthermore, the method for obtaining the necessity of filling includes: Based on the spatial distance between the target empty voxel and its corresponding neighboring non-empty voxel, a second distance weight is obtained for each neighboring non-empty voxel; the second distance weight is positively correlated with the spatial distance. For all neighborhood non-empty voxels, the overall influence level is used as a weight and averaged with the second distance weight to obtain the neighborhood information of the target empty voxel which is not referenced by construction site noise. The non-referenceability is negatively correlated and normalized to obtain the filling necessity of the target empty voxel block.

[0011] Furthermore, the method for obtaining the second distance weight includes: For each neighboring non-empty voxel, obtain the center distance between the neighboring non-empty voxel and the target empty voxel; obtain the average normal vector in the neighboring non-empty voxel; obtain the difference vector between the center coordinate vector of the neighboring non-empty voxel and the center coordinate vector of the target empty voxel; use the dot product of the average normal vector and the difference vector as the vector distance weight; normalize the center distance and multiply it with the vector distance weight to obtain the second distance weight.

[0012] Further, determining whether an empty voxel block needs to be filled based on the filling necessity includes: If the filling necessity is greater than the preset necessity threshold, then it is determined that the corresponding empty voxel block needs to be filled.

[0013] The present invention also proposes an automatic conversion system for BIM-based construction site building model data, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements any of the steps of the BIM-based building model data accurate conversion processing method.

[0014] The present invention has the following beneficial effects: This invention considers that construction site noise can lead to the loss of voxel information, resulting in the irregular disappearance of some structures and producing irregular results such as sharp edges or jagged edges. The voxels scanned from these structures with missing information exhibit isolated distribution and prominent shapes. Therefore, this invention analyzes the local density and curvature of each voxel in the 3D point cloud. Based on these two features, it determines the degree of influence of construction site noise on each voxel, avoiding misjudgments caused by judging the degree of influence solely based on a single dimension. Furthermore, for the initial sampling results of octree downsampling, empty voxel blocks represent regions with missing information. The filling process of these regions is affected by the degree of dust influence on their location; the less dust influence, the better the filling result, and the more necessary it is to fill the corresponding empty voxel block. Conversely, if the dust influence is low, the empty voxel block should not be filled. Therefore, this invention determines the necessity of filling each empty voxel block based on the overall influence distribution of the non-empty voxel blocks in the second neighborhood of the empty voxel block. By utilizing the necessity of filling to determine whether empty solid blocks need to be filled, the final filling result is obtained for BIM model construction and data conversion. This makes the final BIM building model more accurate, improves the accuracy and adaptability of the model construction process under the interference of site environmental factors, and makes data conversion more precise. Attached Figure Description

[0015] 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.

[0016] Figure 1 This is a flowchart of a BIM-based building model data accurate conversion and processing method provided in one embodiment of the present invention. Detailed Implementation

[0017] 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 BIM-based building model data accurate conversion and processing method 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.

[0018] 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.

[0019] The following description, in conjunction with the accompanying drawings, details a specific scheme for a BIM-based method for accurately converting and processing building model data provided by the present invention.

[0020] This invention is implemented in the BIM model construction process of a construction site. Octree downsampling is a crucial preprocessing step in this process. Octree downsampling acts as an efficient bridge between point clouds and BIM, sacrificing unnecessary microscopic details for a more efficient and accurate representation of the macroscopic structure. The simplified steps of octree downsampling include: 1. Spatial partitioning and octree construction; 2. Sampling within leaf nodes; 3. Outputting the downsampled point cloud. During octree downsampling, when a voxel is subdivided, the algorithm determines its sub-cubes: if the sub-cube contains at least one or more point cloud voxel points, it is marked as a non-empty voxel block; if the sub-cube contains no point cloud data points, it is marked as an empty voxel block. These empty voxel blocks are generated solely by construction site noise such as occlusion and dust scattering. It should be noted that the specific octree downsampling process is a well-known technique and will not be elaborated upon or limited here. The embodiments of the present invention mainly determine whether empty voxel blocks need to be filled by analyzing the degree of influence of construction site noise on voxel points in a three-dimensional point cloud.

[0021] Please see Figure 1 The diagram illustrates a flowchart of a BIM-based building model data accurate conversion and processing method according to an embodiment of the present invention. The method includes: Step S1: Use a laser scanner to acquire the three-dimensional point cloud of the construction site, and obtain the initial sampling result of the three-dimensional point cloud after octree downsampling. The initial sampling result includes empty voxels and non-empty voxels.

[0022] In this embodiment of the invention, step S1 is a data preparation step, which can obtain the initial sampling results of 3D point cloud and octree downsampling using existing equipment and well-known algorithms. The specific process is a well-known technique to those skilled in the art and will not be described in detail here.

[0023] Step S2: For each voxel in the 3D point cloud, obtain the local density feature of each voxel based on the distance between the voxel and the neighboring voxels within the preset first neighborhood range; obtain the local curvature of the voxel based on the spatial distribution relationship between the voxel and the neighboring voxels; and obtain the degree of influence of construction site noise on each voxel based on the local density feature and the local curvature.

[0024] When collecting point cloud data at construction sites, dust and other suspended particles scatter the laser beam, reducing its penetration and scanning quality. This can lead to increased noise and missing data in obscured areas. Inaccurate point cloud data acquisition affects the accuracy of BIM building model construction and data conversion. When data is missing, the originally complete structure of the building will exhibit irregular "damage," resulting in sharp and prominent voxel points. Specifically, the scanned voxel points will show a localized low density, incompletely continuous dense distribution, and significant sharp curvature. Therefore, this invention analyzes each voxel point in the 3D point cloud individually, setting a first neighborhood range centered on the voxel point. Other voxel points within the first neighborhood range are considered neighboring voxel points. Then, within the first neighborhood range, the local density characteristics and local curvature of the voxel points are determined based on the spatial relationship between the voxel point and its neighboring voxel points. The greater the local density and local curvature, the sharper the spatial characteristics of the voxel, and the more likely it is to be a voxel of a building structure affected by construction site noise, meaning the voxel is more affected by construction site noise.

[0025] In this embodiment of the invention, the first neighborhood range is a spherical range centered on the target voxel point, with a radius of 2 meters in the real-world coordinate system. It should be noted that the calibration method between the real-world coordinate system and the three-dimensional point cloud coordinate system is a well-known technical means among those skilled in the art. In other specific implementations of this invention, the size of the first neighborhood range can be set differently according to the size of the building, which will not be elaborated or limited here.

[0026] Preferably, in this embodiment of the invention, the method for obtaining local density features includes: For a target voxel, the Euclidean distance between the target voxel and each neighboring voxel within the first neighborhood is obtained, along with the difference in laser reflection intensity. In this embodiment of the invention, the difference in laser reflection intensity is the absolute value of the difference in reflection intensity between two voxels.

[0027] The first distribution disorder of the laser reflection intensity difference is obtained. The smaller the first distribution disorder, the more uniform the reflection intensity of each voxel point is within the first neighborhood of the target voxel point, and there will be no chaotic changes. Therefore, it is less affected by dust scattering.

[0028] The second distribution disorder of the Euclidean distance is obtained. The smaller the second distribution disorder, the more concentrated the position distribution of each voxel point is within the first neighborhood of the target voxel point, and the smaller the degree of information loss caused within the first neighborhood.

[0029] The neighborhood information distribution disorder is obtained based on the first and second distribution disorder. A greater neighborhood information distribution disorder indicates disorder and instability within the first neighborhood in terms of both reflected light intensity and voxel spatial distribution, thus indicating a smaller local density feature for the target voxel. Therefore, the neighborhood information distribution disorder is negatively correlated and normalized to obtain the local density feature.

[0030] In this embodiment of the invention, the normalization method can be linear normalization, or the range standardization method or the sigmoid function mapping method can be used to achieve the normalization purpose. Subtracting the normalized result from the positive integer 1 can achieve negative correlation mapping and normalization.

[0031] Preferably, in this embodiment of the invention, the first distribution disorder is the result of normalizing the standard deviation of the laser reflection intensity difference. That is, the larger the standard deviation, the more disordered the laser reflection intensity difference, and the greater the first distribution disorder.

[0032] Furthermore, in this embodiment of the invention, the method for obtaining the second distribution disorder includes: In order to analyze the distribution of voxel points within the first neighborhood more precisely, this embodiment of the invention uses distance clustering to initially classify the voxel points within the neighborhood, and then analyzes them separately in each category.

[0033] First, based on the Euclidean distance, a clustering algorithm is used to classify the voxel points within the first neighborhood, obtaining multiple voxel point categories. In this embodiment of the invention, the K-means clustering algorithm is selected, and the K value can be determined according to the elbow method. The specific process is a well-known technique to those skilled in the art and will not be elaborated here.

[0034] Secondly, within each voxel category, a negative correlation mapping is performed on the corresponding Euclidean distance to obtain a first distance weight. The standard deviation of the first distance weight is used as the initial distribution disorder of the voxel category. It should be noted that since neighboring voxels and the target voxel are not the same voxel, their Euclidean distance cannot be zero. Therefore, the negative correlation mapping here can directly use the reciprocal of the Euclidean distance as the first distance weight. The larger the first distance weight, the closer the corresponding neighboring voxel is to the target voxel. The larger the standard deviation of the first distance weight within a voxel category, the more disordered the voxel distribution within that category. In this embodiment, K-means clustering is first used to group voxels with similar Euclidean distances into the same category. Then, the degree of disorder in the distribution of voxels within the same voxel category is determined according to the above calculation method. Under normal conditions without construction site noise, the distribution of voxels within the same voxel category is relatively regular and the degree of disorder is low. Therefore, if the degree of disorder in the distribution of voxels within the same voxel category is high, it indicates a greater influence from construction site noise.

[0035] Then, the initial distribution disorder of all voxel point categories is statistically analyzed, and the cumulative value of the initial distribution disorder of all voxel point categories is taken as the second distribution disorder.

[0036] As a specific example, in one implementation of this invention, the formula for the local density feature can be expressed as: ;in, Let norm() represent the local density feature of the i-th voxel, and let norm() denote the normalization function. Let be the difference in laser reflection intensity between the i-th voxel and its m-th neighboring voxels, and M be the number of neighboring voxels. For the first distribution disorder, K is the number of voxel categories. Let B be the Euclidean distance between the i-th voxel and the b-th neighboring voxel in the k-th voxel category, where B is the number of neighboring voxels in the k-th voxel category. This represents the disorder of the second distribution. To mitigate the chaotic distribution of neighborhood information, an averaging method is used to fuse these two parameters. The normalization operation employed by the normalization function has already been described in the above description of the embodiments of this invention and will not be repeated here.

[0037] Preferably, in this embodiment of the invention, the method for obtaining local curvature includes: The voxel category to which the target voxel belongs is taken as the target category. For voxels within the target category, the curvature obtained using a PCA-based 3D spatial curvature algorithm is taken as the local curvature. It should be noted that the PCA-based 3D spatial curvature algorithm is a commonly used spatial curvature estimation algorithm in the field of spatial curvature estimation, and its specific details will not be elaborated upon here.

[0038] In other specific implementations of this invention, the local curvature can also be calculated directly using the plane normal vector formed by the point cloud. This is also a technical means well known to those skilled in the art, and will not be limited or elaborated here.

[0039] Preferably, in this embodiment of the invention, since the degree of influence of construction site noise on voxel points should be negatively correlated with local density features and positively correlated with local curvature, and considering that the local density features have been normalized, the positive integer 1 can be directly subtracted from the local density features, i.e., the disordered distribution of neighborhood information in the above formula for obtaining local density features, as the negative correlation mapping result. After multiplying it with the local curvature, the degree of influence of construction site noise on voxel points is obtained.

[0040] Step S3: Calculate the influence degree of each voxel point in the non-empty voxel block to obtain the overall influence degree of each non-empty voxel block; for any empty voxel block, calculate the overall influence degree of the neighboring non-empty voxel blocks within the preset second neighborhood range, and combine the distance between the empty voxel block and the neighboring non-empty voxel blocks to obtain the filling necessity of each empty voxel block.

[0041] Based on the degree of influence of construction site noise on each voxel point obtained in step S2, the voxel blocks of the initial sampling results in octree downsampling can be analyzed. For empty voxel blocks, the data may not have been collected due to dust during data acquisition. To effectively fill them, it is necessary to consider the possibility that other non-empty voxel blocks in the local voxel space of the empty voxel block are coplanar with it. The greater the probability that the data points in other non-empty voxel blocks are on the same plane as the center point of the current empty voxel block, the greater the possibility that the data points in the current empty voxel block are missing due to dust during data acquisition. Therefore, data supplementation can be considered. At the same time, when considering the possibility of coplanarity between other non-empty voxel blocks and the current empty voxel block, the overall influence of construction site noise on the non-empty voxel blocks can be used. The greater the overall influence, the lower the accuracy of the data. Therefore, the confidence level or weight should be lower when comparing it with the current empty voxel block.

[0042] Therefore, in this embodiment of the invention, the influence degree of each voxel point in each non-empty voxel block is first calculated to obtain the overall influence degree of each non-empty voxel block. In this embodiment of the invention, the average influence degree of all voxel points in the non-empty voxel block can be directly used as the overall influence degree.

[0043] Secondly, this embodiment of the invention further analyzes any empty volume block, statistically analyzes the overall influence of neighboring non-empty volume blocks within a preset second neighborhood range, and combines the distance between the empty volume block and the neighboring non-empty volume blocks to obtain the filling necessity of each empty volume block. That is, the closer the neighboring non-empty volume blocks are to the empty volume block, the greater the probability of them being coplanar, and the smaller the overall influence of construction site noise on the neighboring non-empty volume blocks, the more likely the empty volume block can be filled, and the greater the filling necessity.

[0044] In this embodiment of the invention, the second neighborhood range is similar to the first neighborhood range, also being a spherical range with a radius set to 5 meters in the real world.

[0045] Preferably, in this embodiment of the invention, the method for obtaining the necessity of filling includes: Based on the spatial distance between the target empty voxel and its corresponding neighboring non-empty voxels, a second distance weight is obtained for each neighboring non-empty voxel; the second distance weight is positively correlated with the spatial distance. That is, the larger the second distance weight, the greater the distance between the neighboring non-empty voxels and the target empty voxel; the greater the distance between them, the less the target empty voxel should be filled, i.e., the less necessary it is to be filled.

[0046] For all neighboring non-empty voxels, the overall impact level is used as a weight, and the second distance weight is used as a weighted average to obtain the unreliability of the target empty voxel's neighborhood information due to construction site noise. Because a greater overall impact level indicates a greater influence of dust on neighboring non-empty voxels, resulting in lower information reliability, and the farther the neighboring non-empty voxels are from the target empty voxel, the lower their information reliability also becomes. Therefore, a greater overall impact level and a larger second distance weight correspond to a greater unreliability.

[0047] The non-referenceability is negatively correlated and normalized to obtain the filling necessity of the target empty voxel block. It should be noted that the negative correlation mapping and normalization here is the same method used for the local density features mentioned above, and will not be repeated here.

[0048] As a specific example, in one implementation of this invention, the necessity of filling is expressed by the formula: ;in The necessity of filling the j-th empty voxel block, Let be the number of non-empty voxel blocks in the second neighborhood of the j-th empty voxel block. The overall influence of the nth non-empty voxel block in the neighborhood. Let be the second distance weight of the nth non-empty neighborhood voxel block, and norm() denotes the normalization function.

[0049] Preferably, in this embodiment of the invention, the method for obtaining the second distance weight includes: For each neighboring non-empty voxel, obtain the center-point distance between the neighboring non-empty voxel and the target empty voxel. The larger the center-point distance, the farther the two are, and therefore the greater the weight of the second distance should be.

[0050] Obtain the average normal vector of the neighborhood non-empty voxels; obtain the difference vector between the coordinate vector of the center point of the neighborhood non-empty voxels and the coordinate vector of the center point of the target empty voxel; this difference vector is the position vector between the two center points; use the absolute value of the dot product of the average normal vector and the difference vector as the vector distance weight. If two points and their normal vectors are coplanar, then the normal vector of one of the points should be perpendicular to the position vector between the two points. Therefore, the center points of the neighborhood non-empty voxels and the empty voxels are denoted as two representative points, and their position vectors are multiplied by the average normal vector of the neighborhood non-empty voxels. The closer the result is to 0, the greater the probability that the two points are coplanar. That is, the larger the vector distance weight, the less coplanar the two voxels are, and the larger the corresponding second distance weight should be.

[0051] The normalized center point distance is multiplied by the vector distance weight to obtain the second distance weight. In this embodiment of the invention, the normalization method for the center point distance can be range standardization, that is, the maximum and minimum values ​​among all center point distances are counted, and the center point distance is normalized using range standardization.

[0052] Step S4: Determine whether the empty voxel block needs to be filled based on the filling necessity; use the final filling result as an optimized octree network; use the optimized octree network to construct the BIM model and perform data conversion.

[0053] Finally, the fill necessity can be used to determine whether each empty voxel needs to be filled; the lower the fill necessity, the less likely the empty voxel needs to be filled. In this embodiment of the invention, an interpolation method can be used to interpolate the empty voxels based on the information of the non-empty voxels within the second neighborhood, resulting in a completed octree network, which serves as the optimized octree network. Existing feature extraction algorithms (such as Hough transform or RANSAC algorithm) are used to extract geometric features based on the optimized octree network, and BIM building models are constructed using Revit and ClearEdge3D to complete the automatic data conversion.

[0054] It should be noted that the interpolation can be achieved using trilinear interpolation. Other implementations of this invention may also employ radial basis function interpolation, natural neighborhood interpolation, or other methods. The specific filling method is a well-known technique among those skilled in the art, and can also be implemented using diffusion-based methods, deep learning-based methods, or other techniques familiar to those skilled in the art; details will not be elaborated upon here.

[0055] Preferably, in this embodiment of the invention, determining whether an empty voxel block needs to be filled based on the filling necessity includes: If the filling necessity is greater than a preset necessity threshold, then the corresponding empty voxel block is determined to need to be filled. It should be noted that, because the filling necessity has been normalized, in one specific implementation of this invention, the necessity threshold can be set to 0.75.

[0056] In summary, this invention analyzes the local density and curvature of each voxel in a 3D point cloud. Based on these two features, the degree of influence of construction site noise on each voxel is obtained. Based on the overall influence distribution of non-empty voxels within the second neighborhood of an empty voxel, the filling necessity of each empty voxel is determined. The filling necessity is used to determine whether an empty voxel needs to be filled, thus obtaining the final filling result for BIM model construction and data conversion. This invention makes the final constructed BIM building model more accurate, improves the accuracy and adaptability of the model construction process under the interference of construction site environmental factors, and makes data conversion more precise.

[0057] 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.

[0058] 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.

Claims

1. A method for accurate conversion and processing of BIM-based building model data, characterized in that, The method includes: A laser scanner is used to acquire a 3D point cloud of the construction site, and the initial sampling result of the 3D point cloud after octree downsampling is obtained. The initial sampling result includes empty voxel blocks and non-empty voxel blocks. For each voxel in the 3D point cloud, the local density feature of each voxel is obtained based on the distance between the voxel and neighboring voxels within a preset first neighborhood range; the local curvature of the voxel is obtained based on the spatial distribution relationship between the voxel and its neighboring voxels; and the degree of influence of construction site noise on each voxel is obtained based on the local density feature and the local curvature. The influence degree of each voxel point in the non-empty voxel block is statistically analyzed to obtain the overall influence degree of each non-empty voxel block; for any empty voxel block, the overall influence degree of the neighboring non-empty voxel blocks within the preset second neighborhood range is statistically analyzed, and the filling necessity of each empty voxel block is obtained by combining the distance between the empty voxel block and the neighboring non-empty voxel blocks. Determine whether empty voxels need to be filled based on the filling necessity; use the final filling result as an optimized octree network; and use the optimized octree network to construct the BIM model and perform data conversion.

2. The method for accurate conversion and processing of BIM-based building model data according to claim 1, characterized in that, The method for obtaining the local density features includes: For a target voxel, obtain the Euclidean distance between the target voxel and each neighboring voxel within the first neighborhood, as well as the difference in laser reflection intensity; obtain the first distribution disorder of the laser reflection intensity difference; obtain the second distribution disorder of the Euclidean distance; obtain the neighborhood information distribution disorder based on the first and second distribution disorder; perform negative correlation mapping and normalization on the neighborhood information distribution disorder to obtain the local density feature.

3. The method for accurate conversion and processing of BIM-based building model data according to claim 2, characterized in that, The first distribution disorder is the result of normalizing the standard deviation of the laser reflection intensity difference.

4. The method for accurate conversion and processing of BIM-based building model data according to claim 2, characterized in that, The methods for obtaining the second distribution disorder include: Based on the Euclidean distance, a clustering algorithm is used to classify voxel points within the first neighborhood to obtain multiple voxel point categories. In each voxel point category, the corresponding Euclidean distance is negatively correlated to obtain a first distance weight, and the standard deviation of the first distance weight is used as the initial distribution disorder of the voxel point category. The cumulative value of the initial distribution disorder of all voxel point categories is used as the second distribution disorder.

5. The method for accurate conversion and processing of BIM-based building model data according to claim 4, characterized in that, The method for obtaining the local curvature includes: The voxel category to which the target voxel belongs is taken as the target category. For voxels in the target category, the curvature obtained by the PCA-based three-dimensional spatial curvature algorithm is taken as the local curvature.

6. The method for accurate conversion and processing of BIM-based building model data according to claim 1, characterized in that, The methods for obtaining the overall degree of influence include: For each voxel, after negatively correlated mapping of the local density features, the product of the mapping result and the local curvature is used as the degree of influence of the voxel. The average influence of all voxel points in the non-empty voxel block is taken as the overall influence.

7. The method for accurate conversion and processing of BIM-based building model data according to claim 1, characterized in that, The method for obtaining the necessity of filling includes: Based on the spatial distance between the target empty voxel and its corresponding neighboring non-empty voxel, a second distance weight is obtained for each neighboring non-empty voxel; the second distance weight is positively correlated with the spatial distance. For all neighborhood non-empty voxels, the overall influence level is used as a weight and averaged with the second distance weight to obtain the neighborhood information of the target empty voxel which is not referenced by construction site noise. The non-referenceability is negatively correlated and normalized to obtain the filling necessity of the target empty voxel block.

8. The method for accurate conversion and processing of BIM-based building model data according to claim 7, characterized in that, The method for obtaining the second distance weight includes: For each neighboring non-empty voxel, obtain the center distance between the neighboring non-empty voxel and the target empty voxel; obtain the average normal vector in the neighboring non-empty voxel; obtain the difference vector between the center coordinate vector of the neighboring non-empty voxel and the center coordinate vector of the target empty voxel; use the dot product of the average normal vector and the difference vector as the vector distance weight; normalize the center distance and multiply it with the vector distance weight to obtain the second distance weight.

9. The method for accurate conversion and processing of BIM-based building model data according to claim 1, characterized in that, The step of determining whether an empty voxel block needs to be filled based on the filling necessity includes: If the filling necessity is greater than the preset necessity threshold, then it is determined that the corresponding empty voxel block needs to be filled.

10. The method for accurate conversion and processing of BIM-based building model data according to claim 1, characterized in that, An interpolation method is used to interpolate the empty voxels based on the information of the non-empty voxels in the second neighborhood to obtain the final filling result.

Images