High-reduction-degree simulation analysis method based on bim+gis technology
By combining BIM and GIS technologies in point cloud registration, calculating saliency and spatial structure similarity, and using a weighted ICP objective function for point cloud registration, the problem of insufficient reliability in existing point cloud registration technologies is solved, and the accuracy and reliability of high-fidelity simulation analysis are achieved.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- BEIJING FENGDA TECHNOLOGY CO LTD
- Filing Date
- 2026-02-06
- Publication Date
- 2026-07-21
AI Technical Summary
Existing point cloud registration technology relies solely on the similarity between the geometric coordinates of point cloud data for registration. This results in insufficient reliability of the matching results when factors such as beam cutting after construction, wall thickening, and system drift between GIS aerial survey and BIM coordinate systems occur, which can easily lead to inaccurate simulation analysis results.
A high-fidelity simulation analysis method based on BIM+GIS technology is adopted. By calculating the saliency of the target point and the pixel value of the points in the local neighborhood, saliency clusters are obtained. Combining the spatial distance and attribute similarity of the first-class and second-class points, the comprehensive spatial structure similarity is calculated. The weighted ICP objective function is used to perform point cloud registration and generate a 3D model.
This improves the accuracy of simulation analysis results, enhances the reliability of point cloud registration, and ensures the precision and reliability of simulation analysis under different construction changes.
Smart Images
Figure CN121683539B_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of point cloud registration technology, specifically to a high-fidelity simulation analysis method based on BIM+GIS technology. Background Technology
[0002] In fields such as smart cities, digital twins, and full lifecycle management of projects, high-fidelity simulation analysis is the core support for achieving accurate decision-making, risk prevention and control, and efficiency optimization. The technical synergy between BIM (Building Information Modeling) and GIS (Geographic Information System) breaks through the application limitations of single technologies, ensuring the integrity, accuracy, and practicality of simulation analysis from multiple levels, including spatial scale, data dimensions, and analysis scenarios.
[0003] In the process of achieving seamless integration between BIM and GIS models, point cloud registration technology is required to register BIM point cloud data and GIS point cloud data. However, existing point cloud registration technology relies solely on the similarity between the geometric coordinates of the point cloud data for registration. When factors such as beam cutting after construction, wall thickening, or system drift between GIS aerial survey and BIM coordinate systems occur, the reliability of the matching results is insufficient, which can easily lead to inaccurate simulation analysis results. Summary of the Invention
[0004] This application provides a high-fidelity simulation analysis method based on BIM+GIS technology to solve the problem that point cloud registration technology relies solely on the similarity between geometric coordinates of point cloud data for registration, resulting in insufficient reliability and inaccurate simulation analysis results. The specific technical solution adopted is as follows: One embodiment of this application provides a high-fidelity simulation analysis method based on BIM+GIS technology, which includes the following steps: Collect BIM point cloud data and GIS point cloud data; Denote any point in the BIM point cloud data and GIS point cloud data as the target point. Denote any point in the BIM point cloud data as a Class I point and any point in the GIS point cloud data as a Class II point. Calculate the saliency of the target point based on the pixel values of points in its local neighborhood. Cluster all points in the BIM and GIS point cloud data to obtain saliency clusters. Calculate the saliency distribution similarity between Class I and Class II points based on the difference in the ratio of the number of points belonging to the same saliency cluster in the local neighborhoods of Class I and Class II points, as well as the difference in saliency between the BIM and GIS point cloud data. Calculate the attribute similarity between Class I and Class II points by combining the spatial distance between them and the difference in saliency. The local neighborhood of the target point is divided into different local sub-neighborhoods. Based on the distance between the target point and the surface fitting results of all points in the local sub-neighborhood, the shortest spatial distance of the target point in the local sub-neighborhood is calculated. Based on the difference in the shortest spatial distance between Class I and Class II points in the corresponding local sub-neighborhoods, the difference in the number of points belonging to the same saliency cluster, and the saliency distribution similarity between Class I and Class II points, the comprehensive spatial structure similarity between Class I and Class II points is calculated. Based on the similarity of point attributes and comprehensive spatial structure between BIM point cloud data and GIS point cloud data, point cloud registration between BIM point cloud data and GIS point cloud data is achieved, and a 3D model is generated based on the point cloud registration results.
[0005] Furthermore, the specific method for calculating the salience of the target point is as follows: Points with the same pixel value as the target point in the local neighborhood of the target point are denoted as adjacent homogeneous points of the target point. The number of points with the largest pixel value in the local neighborhood of the target point is denoted as the maximum number of adjacent pixels of the target point. The ratio of the number of adjacent homogeneous points of the target point to the number of the largest neighboring pixels is denoted as the first ratio of the target point. The negative correlation result of the first ratio of the target point is denoted as the saliency of the target point.
[0006] Furthermore, the specific method for obtaining the saliency distribution similarity between the first-class points and the second-class points is as follows: The ratio of the number of points belonging to the same salient cluster within the local neighborhood of Class I points and Class II points is denoted as the first ratio of the same salient cluster. The ratio of the number of points belonging to the same salient cluster within BIM point cloud data and GIS point cloud data is denoted as the second ratio of the same salient cluster. The difference between the second ratio and the first ratio of the salient cluster is denoted as the first difference of the salient cluster. The negative correlation result of the first difference of all significant clusters is denoted as the significant distribution similarity between Class I and Class II points.
[0007] Furthermore, the method for calculating the attribute similarity between the first-class points and the second-class points is as follows: The normalized value of the spatial distance between Class I points and Class II points is denoted as the relative distance between Class I points and Class II points; The absolute value of the difference between the significance of Class I points and Class II points is denoted as the significance difference between Class I points and Class II points; The negative correlation between the relative distance, significance distribution similarity, and significance difference between Class I and Class II points is denoted as the attribute similarity between Class I and Class II points.
[0008] Furthermore, the method for calculating the shortest spatial distance of the target point in its local sub-neighborhood is as follows: For each local sub-neighborhood of the target point, perform surface fitting on all points contained therein to obtain the fitted surface within the local sub-neighborhood of the target point. The shortest Euclidean distance from the target point to the fitted surface within the local sub-neighborhood of the target point is denoted as the shortest spatial distance of the target point in the local sub-neighborhood.
[0009] Furthermore, the method for obtaining the comprehensive spatial structure similarity of the first-class points and the second-class points is as follows: Based on the difference in the shortest spatial distance between Class I and Class II points in their corresponding local sub-neighborhoods, calculate the spatial distance difference between Class I and Class II points in their corresponding local sub-neighborhoods; the negative correlation processing result of the spatial distance difference between Class I and Class II points in all their corresponding local sub-neighborhoods is denoted as the spatial structure similarity between Class I and Class II points. Based on the difference in the number of points belonging to the same salient cluster contained in the corresponding local sub-neighborhoods of Class I and Class II points, the difference in the number of points belonging to the same salient cluster in the corresponding local sub-neighborhoods of Class I and Class II points is calculated; the negative correlation processing result of the difference in the number of points belonging to the same salient cluster in all corresponding local sub-neighborhoods of Class I and Class II points is denoted as the salient spatial structure similarity between Class I and Class II points. The spatial structure similarity and salient spatial structure similarity are weighted and summed using the salient distribution similarity of Class I and Class II points. The result of the weighted summation is denoted as the comprehensive spatial structure similarity of Class I and Class II points.
[0010] Furthermore, the method for obtaining the spatial distance difference is as follows: The absolute value of the difference between the shortest spatial distances of Class I points and Class II points in their corresponding local sub-neighborhoods is denoted as the spatial distance difference between Class I points and Class II points in their corresponding local sub-neighborhoods.
[0011] Furthermore, the method for obtaining the difference in the number of members in the same cluster is as follows: The absolute value of the difference between the number of points belonging to the same salient cluster contained in the corresponding local sub-neighborhood of Class I points and Class II points is denoted as the difference in the number of points belonging to the same salient cluster in the corresponding local sub-neighborhood of Class I points and Class II points.
[0012] Furthermore, the specific method for achieving point cloud registration between BIM point cloud data and GIS point cloud data based on the similarity of point attributes and comprehensive spatial structure between the two data includes: Based on the similarity of point attributes and the comprehensive spatial structure of points in BIM point cloud data and GIS point cloud data, the point cloud similarity between Class I points and Class II points is calculated. The comprehensive spatial structure similarity of two matched points during the point cloud registration process is used as the reliability weight to establish a weighted ICP objective function. The ICP point cloud registration algorithm is then used to perform point cloud matching on BIM point cloud data and GIS point cloud data to obtain the point cloud matching results.
[0013] Furthermore, the method for calculating the point cloud similarity is as follows: The mean of the attribute similarity and the comprehensive spatial structure similarity between Class I and Class II points is denoted as the point cloud similarity between Class I and Class II points.
[0014] The beneficial effects of this application are: This application considers that although the grayscale values and reflectance intensity values of objects in BIM point cloud data and GIS point cloud data reflect different object characteristics, there is still a certain correlation between them. Based on the fixed distribution of other objects within a fixed neighborhood of a certain location on a building, the salience of the target point is calculated, and salience clusters are obtained. Based on the similarity and spatial distance of the salience distribution of point cloud data within the local neighborhoods of Class I and Class II points, the probability that Class I and Class II points are corresponding points in BIM point cloud data and GIS point cloud data is evaluated, and the attribute similarity of Class I and Class II points is obtained. Furthermore, based on the similarity of spatial structures, the correspondence between different points in BIM point cloud data and GIS point cloud data is further analyzed. Based on the distance between the surface fitting results of all points within the local sub-neighborhood of the target point, the salience of the target point is calculated. The method calculates the shortest spatial distance between points in their local sub-neighborhoods, and based on the similarity of these shortest spatial distances and the similarity of the distribution of points belonging to the same salient cluster within the local sub-neighborhoods, it calculates the comprehensive spatial structure similarity between Class I and Class II points. Finally, based on the attribute similarity and comprehensive spatial structure similarity of points in BIM and GIS point cloud data, it comprehensively evaluates the semantic similarity and geometric consistency of matched points in the BIM and GIS point cloud data. Based on the evaluation results, it performs point cloud registration between BIM and GIS point cloud data, and generates a 3D model based on the registration results. This addresses the problem that point cloud registration technology relies solely on the similarity between geometric coordinates of point cloud data, resulting in insufficient reliability and potentially inaccurate simulation analysis results, thus improving the accuracy of simulation analysis results. Attached Figure Description
[0015] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.
[0016] Figure 1 This is a schematic diagram of a high-fidelity simulation analysis method based on BIM+GIS technology provided in one embodiment of this application. Detailed Implementation
[0017] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.
[0018] Please see Figure 1 The diagram illustrates a flowchart of a high-fidelity simulation analysis method based on BIM+GIS technology according to an embodiment of this application. The method includes the following steps: Step S001: Collect BIM point cloud data and GIS point cloud data.
[0019] BIM point cloud data and GIS point cloud data were collected using BIM+GIS acquisition equipment. Specifically, Revit software was used to acquire the BIM model, and BIM point cloud data was extracted from the BIM model. LiDAR airborne lidar was used to acquire GIS point cloud data.
[0020] Denoising is performed on both BIM point cloud data and GIS point cloud data. Specifically, this embodiment uses a three-dimensional Gaussian filtering algorithm to achieve denoising.
[0021] It is understandable that BIM point cloud data can be used to obtain the X, Y, and Z coordinates of each point in the BIM point cloud data, as well as the grayscale value; similarly, GIS point cloud data can be used to obtain the X, Y, and Z coordinates of each point in the GIS point cloud data, as well as the reflection intensity value.
[0022] Dimensionless processing is performed on each type of data in both BIM point cloud data and GIS point cloud data. This embodiment uses the Z-Score standard normalization method for dimensionless processing. The Z-Score standard normalization method for dimensionless processing is a well-known technique and will not be elaborated further.
[0023] At this point, BIM point cloud data and GIS point cloud data have been obtained.
[0024] Step S002: Denote any point in the BIM point cloud data and GIS point cloud data as the target point, any point in the BIM point cloud data as a Class I point, and any point in the GIS point cloud data as a Class II point. Calculate the saliency of the target point based on the pixel values of points within its local neighborhood. Cluster all points in the BIM and GIS point cloud data to obtain saliency clusters. Calculate the saliency distribution similarity between Class I and Class II points based on the difference in the ratio of the number of points belonging to the same saliency cluster within their local neighborhoods and within the BIM and GIS point cloud data. Combine the spatial distance between Class I and Class II points with the difference in saliency to calculate the attribute similarity between them.
[0025] In BIM point cloud data, the grayscale value of a point reflects the surface color of the corresponding object, while in GIS point cloud data, the reflectance value reflects the internal composition of the corresponding object. For high-fidelity simulation analysis of buildings, to achieve thermal insulation, waterproofing, aesthetics, and durability, different building materials typically correspond to different colors and properties. For example, building exterior walls usually require thermal insulation, so light colors are generally chosen, and materials such as polystyrene boards or rock wool boards are commonly used. Roofs typically require waterproofing and rainproofing, using materials such as waterproof coatings, waterproof membranes, and waterproof mortar, and colors such as black or dark colors are generally chosen. Therefore, although grayscale values and reflectance values reflect different object characteristics, there is a certain correlation between them.
[0026] The location distribution of other objects within a fixed neighborhood of a certain location on a building is fixed. Therefore, the number of other points within the neighborhood of the corresponding point in BIM point cloud data and GIS point cloud data should be similar.
[0027] Any point in the BIM point cloud data and GIS point cloud data is designated as the target point. The spherical neighborhood range of the target point with a first preset length equal to the neighborhood radius is designated as the local neighborhood of the target point. Points within the local neighborhood of the target point with the same pixel value as the target point are designated as adjacent homogeneous points of the target point. The variance of the pixel values within the local neighborhood of the target point is designated as the maximum number of adjacent pixels of the target point. The ratio of the number of adjacent homogeneous points of the target point to the maximum number of adjacent pixels is designated as the first ratio of the target point. The negative correlation result of the first ratio of the target point is designated as the saliency of the target point.
[0028] It is understood that negative correlation processing is applied to the first ratio of the target point, that is, to ensure that the first ratio of the target point is negatively correlated with the significance of the target point. It is understood that the negative correlation in this application refers to the relationship between the independent variable and the dependent variable, where the independent variable is the first ratio of the target point and the dependent variable is the significance of the target point. The negative correlation means that the dependent variable decreases (increases) as the independent variable increases (decreases), and can be an inverse relationship, a subtraction relationship, etc.
[0029] Preferably, as an embodiment of this application, the difference between the number 1 and the first ratio of the target point is recorded as the significance of the target point.
[0030] In this embodiment, the first preset length is set to 17 pixels. The pixel values of the points in the BIM point cloud data and GIS point cloud data are grayscale values and reflection intensity values, respectively. There may be multiple points with the largest pixel value in the local neighborhood of the target point. For example, when the maximum pixel value in the local neighborhood of the target point is 50, the maximum number of adjacent pixels of the target point is the number of points with a pixel value of 50 in the local neighborhood of the target point.
[0031] The saliency of any point in BIM point cloud data and GIS point cloud data can be obtained using the same method.
[0032] When two points in BIM point cloud data and GIS point cloud data correspond to the same location point, the distribution of the two corresponding points in their local neighborhood point cloud data is similar, and the significance values of the two corresponding points are also similar. Based on this, the correlation between different points in BIM point cloud data and GIS point cloud data is further analyzed.
[0033] Using the significant differences between different points as a metric for distance, all points in both BIM point cloud data and GIS point cloud data are clustered together to obtain significant clusters.
[0034] In this embodiment, the DBSCAN clustering algorithm is used to cluster all points in BIM point cloud data and GIS point cloud data.
[0035] Any point in the BIM point cloud data is designated as a Class I point, and any point in the GIS point cloud data is designated as a Class II point. The ratio of the number of points belonging to the same salient cluster within the local neighborhood of Class I and Class II points is designated as the first ratio of the same salient cluster. The ratio of the number of points belonging to the same salient cluster in the BIM and GIS point cloud data is designated as the second ratio of the same salient cluster. The difference between the second ratio and the first ratio of the salient cluster is designated as the first difference of the salient cluster. The negative correlation result of the first differences of all salient clusters is designated as the salient distribution similarity between Class I and Class II points.
[0036] Preferably, as an embodiment of this application, the absolute value of the sum of the first differences of all saliency clusters is denoted as the first sum of the first-class points and the second-class points, and the difference between the number 1 and the first sum of the first-class points and the second-class points is denoted as the saliency distribution similarity of the first-class points and the second-class points.
[0037] In the process of calculating the first ratio and the second ratio, in order to avoid the denominator being zero, a preset value needs to be added to the denominator. In this embodiment, the preset value is 0.1.
[0038] The greater the similarity in the saliency distributions of Class I and Class II points, the more similar the saliency distributions of the point cloud data within their local neighborhoods, and the greater the probability that Class I and Class II points are corresponding points in BIM point cloud data and GIS point cloud data.
[0039] The normalized value of the spatial distance between Class I and Class II points is denoted as the relative distance between Class I and Class II points.
[0040] In this embodiment, the sigmoid function is used to calculate the normalized value. The sigmoid function is a well-known technique and will not be described in detail here. As other implementation methods, implementers can use other methods of the prior art, such as the tanh function.
[0041] The absolute value of the difference in significance between Class I and Class II points is denoted as the significance difference between Class I and Class II points; the negative correlation between the relative distance between Class I and Class II points, the similarity of significance distribution, and the significance difference is denoted as the attribute similarity between Class I and Class II points.
[0042] Preferably, as an embodiment of this application, the saliency distribution similarity between Class I and Class II points is used as a weight to perform a weighted summation of the saliency differences between Class I and Class II points. The sum of the weighted summation of the saliency distribution similarity and saliency differences between Class I and Class II points and the relative distance between Class I and Class II points is recorded as the first sum value of Class I and Class II points. The difference between the number 1 and the normalized value of the first sum value of Class I and Class II points is recorded as the attribute similarity between Class I and Class II points.
[0043] The greater the similarity in attributes between Class I and Class II points, the more similar the saliency distribution of the point cloud data within the local neighborhood of Class I and Class II points, the closer the spatial locations of Class I and Class II points, and the greater the probability that Class I and Class II points are corresponding points in BIM point cloud data and GIS point cloud data.
[0044] The same method can be used to obtain the attribute similarity between any point in BIM point cloud data and any point in GIS point cloud data.
[0045] This completes the comparison of the attributes of any point in the BIM point cloud data with any point in the GIS point cloud data.
[0046] Step S003: Divide the local neighborhood of the target point into different local sub-neighborhoods. Calculate the shortest spatial distance of the target point in the local sub-neighborhood based on the distance between the surface fitting results of all points in the local sub-neighborhood of the target point. Calculate the comprehensive spatial structure similarity of the first-class and second-class points based on the difference in the shortest spatial distance between the first-class and second-class points in their corresponding local sub-neighborhoods, the difference in the number of points belonging to the same saliency cluster, and the saliency distribution similarity between the first-class and second-class points.
[0047] When two points in BIM point cloud data and GIS point cloud data correspond to the same location point, the spatial structure of the point cloud data of the two corresponding points in their local neighborhood is similar. Therefore, based on the similarity of spatial structure, the correspondence between different points in BIM point cloud data and GIS point cloud data is further analyzed to improve the reliability of point cloud registration results.
[0048] The local neighborhood of the target point is divided into 8 local sub-neighborhoods by spatial orthogonal segmentation. Surface fitting is performed on all points contained in each local sub-neighborhood of the target point to obtain the fitted surface in the local sub-neighborhood of the target point. The shortest Euclidean distance from the target point to the fitted surface in the local sub-neighborhood of the target point is denoted as the shortest spatial distance of the target point in the local sub-neighborhood.
[0049] Optionally, when the number of points contained in a local sub-neighborhood is less than a preset threshold (e.g., 3), the shortest spatial distance of that local sub-neighborhood is the preset maximum value (or 0). Alternatively, the Euclidean distance from the point to the target point can be used instead.
[0050] In this embodiment, the Poisson surface fitting algorithm is used to achieve surface fitting.
[0051] The same method can be used to obtain the shortest spatial distance of any point in BIM point cloud data and GIS point cloud data in each local sub-neighborhood.
[0052] The absolute value of the difference between the shortest spatial distances of Class I and Class II points in their corresponding local sub-neighborhoods is denoted as the spatial distance difference between Class I and Class II points in their corresponding local sub-neighborhoods. The negative correlation result of the spatial distance differences between Class I and Class II points in all their corresponding local sub-neighborhoods is denoted as the spatial structure similarity between Class I and Class II points.
[0053] Preferably, as an embodiment of this application, the normalized value of the sum of the spatial distance differences between Class I and Class II points in all corresponding local sub-neighborhoods is recorded as the comprehensive spatial distance between Class I and Class II points, and the difference between the number 1 and the comprehensive spatial distance between Class I and Class II points is recorded as the spatial structural similarity between Class I and Class II points.
[0054] The same method can be used to obtain the spatial structural similarity between any point in BIM point cloud data and any point in GIS point cloud data.
[0055] This completes the acquisition of the spatial structural similarity between any point in the BIM point cloud data and any point in the GIS point cloud data.
[0056] The absolute value of the difference in the number of points belonging to the same salient cluster contained in the corresponding local sub-neighborhoods of Class I and Class II points is denoted as the difference in the number of points belonging to the same salient cluster in the corresponding local sub-neighborhoods of Class I and Class II points. The negative correlation result of the difference in the number of points belonging to the same salient cluster in all corresponding local sub-neighborhoods of Class I and Class II points is denoted as the salient spatial structure similarity between Class I and Class II points.
[0057] Preferably, as an embodiment of this application, the sum of the differences in the number of co-clusters of all corresponding salient clusters of Class I and Class II points in all their corresponding local sub-neighborhoods is recorded as the comprehensive difference in the number of co-clusters of Class I and Class II points, and the difference between the number 1 and the comprehensive difference in the number of co-clusters of Class I and Class II points is recorded as the salient spatial structural similarity of Class I and Class II points.
[0058] The spatial structure similarity and salient spatial structure similarity are weighted and summed using the salient distribution similarity of Class I and Class II points. The result of the weighted summation is denoted as the comprehensive spatial structure similarity of Class I and Class II points.
[0059] Specifically, the reciprocal of the sum of the saliency distribution similarities between the number 1 and the points of class I and class II is denoted as the first ratio. The ratio of the saliency distribution similarity between the points of class I and class II to the sum of the saliency distribution similarities between the number 1 and the points of class I and class II is denoted as the second ratio. The first ratio is used as the weight of the spatial structure similarity between the points of class I and class II, and the second ratio is used as the weight of the saliency spatial structure similarity between the points of class I and class II. The weighted sum of the spatial structure similarity and the saliency spatial structure similarity is denoted as the comprehensive spatial structure similarity between the points of class I and class II.
[0060] The same method can be used to obtain the comprehensive spatial structure similarity between any point in BIM point cloud data and any point in GIS point cloud data.
[0061] At this point, the comprehensive spatial structure similarity between any point in the BIM point cloud data and any point in the GIS point cloud data is obtained.
[0062] Step S004: Based on the similarity of point attributes and the comprehensive spatial structure of the points in the BIM point cloud data and the GIS point cloud data, perform point cloud registration between the BIM point cloud data and the GIS point cloud data, and generate a 3D model based on the point cloud registration results.
[0063] The mean of the attribute similarity and the comprehensive spatial structure similarity between Class I and Class II points is denoted as the point cloud similarity between Class I and Class II points.
[0064] The same method can be used to obtain the point cloud similarity between any point in BIM point cloud data and any point in GIS point cloud data.
[0065] To strengthen the leading role of reliable matching point pairs in registration optimization, when using the ICP point cloud registration algorithm to match BIM point cloud data and GIS point cloud data, point cloud similarity is introduced before the norm error term for each source point and target point matching pair to construct a weighted ICP objective function and obtain the point cloud matching results of BIM point cloud data and GIS point cloud data.
[0066] Specifically, the comprehensive spatial structural similarity of two matched points during point cloud registration is used as the reliability weight to obtain the weighted ICP objective function.
[0067] The ICP point cloud registration algorithm using the weighted ICP objective function to register point clouds is a well-known technique. The reliability weights are existing parameters in the weighted ICP objective function and will not be elaborated further.
[0068] It is understandable that the greater the comprehensive spatial structural similarity between two matched points during point cloud registration, the higher the proportion of the geometric error objective function of the two matched points in the convergence optimization process. This can guide the registration towards a direction with higher semantic and geometric consistency, avoid treating the geometric errors of all matched point pairs equally, and prevent the problem of being easily affected by factors such as coordinate system drift. This ensures that the registration result simultaneously satisfies "geometric alignment" and "semantic consistency", thereby improving the accuracy and robustness of registration of two types of heterogeneous point clouds: BIM point cloud data and GIS point cloud data.
[0069] The point cloud data is fused based on the point cloud matching results of BIM point cloud data and GIS point cloud data, and a 3D model is generated using a point cloud reconstruction algorithm.
[0070] In this embodiment, the Poisson reconstruction algorithm is used to generate the three-dimensional model. In practical applications, as other implementation methods, in addition to achieving the purpose of generating the three-dimensional model, the implementer may use other existing methods such as the moving cube algorithm or RBF radial basis function reconstruction to generate the three-dimensional model. This application does not impose any special restrictions.
[0071] This completes the high-fidelity simulation analysis.
[0072] The above description is only a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, improvements, etc., made within the principles of this application should be included within the protection scope of this application.
Claims
1. A high-fidelity simulation analysis method based on BIM+GIS technology, characterized in that, The method includes the following steps: Collect BIM point cloud data and GIS point cloud data; Denote any point in the BIM point cloud data and GIS point cloud data as the target point. Denote any point in the BIM point cloud data as a Class I point and any point in the GIS point cloud data as a Class II point. Calculate the saliency of the target point based on the pixel values of points within its local neighborhood in the point cloud data. Cluster all points in the BIM and GIS point cloud data to obtain saliency clusters. Calculate the saliency distribution similarity between Class I and Class II points based on the difference in the ratio of the number of points belonging to the same saliency cluster within their local neighborhoods and within the BIM and GIS point cloud data. Combine the spatial distance between Class I and Class II points with the absolute value of the saliency difference to calculate the attribute similarity between Class I and Class II points. The local neighborhood of the point cloud data where the target point is located is divided into different local sub-neighborhoods. Based on the distance between the target point and the surface fitting results of all points in the local sub-neighborhood, the shortest spatial distance of the target point in the local sub-neighborhood is calculated. Based on the difference in the shortest spatial distance between Class I and Class II points in the corresponding local sub-neighborhoods, the difference in the number of points belonging to the same saliency cluster, and the saliency distribution similarity between Class I and Class II points, the comprehensive spatial structure similarity between Class I and Class II points is calculated. Based on the similarity of point attributes and comprehensive spatial structure between BIM point cloud data and GIS point cloud data, point cloud registration between BIM point cloud data and GIS point cloud data is achieved, and a 3D model is generated based on the point cloud registration results. The method for determining the salient clusters is as follows: The absolute value of the difference in significance between Class I and Class II points is denoted as the significance difference between Class I and Class II points; the significance difference between different points is used as a distance metric, and all points in the BIM point cloud data and GIS point cloud data are clustered together to obtain significant clusters; The method for calculating the attribute similarity between the first-class points and the second-class points is as follows: The absolute value of the difference in significance between Class I and Class II points is denoted as the significance difference between Class I and Class II points. The significance distribution similarity between Class I and Class II points is used as a weight to perform a weighted summation of the significance differences between Class I and Class II points. The sum of the weighted summation of the significance distribution similarity and significance difference between Class I and Class II points and the relative distance between Class I and Class II points is denoted as the first sum of Class I and Class II points. The difference between the number 1 and the normalized value of the first sum of Class I and Class II points is denoted as the attribute similarity between Class I and Class II points.
2. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 1, characterized in that, The specific method for calculating the significance of the target point is as follows: Points with the same pixel value as the target point in the local neighborhood of the target point are denoted as adjacent homogeneous points of the target point. The number of points with the largest pixel value in the local neighborhood of the target point is denoted as the maximum number of adjacent pixels of the target point. The ratio of the number of adjacent homogeneous points of the target point to the number of the largest neighboring pixels is denoted as the first ratio of the target point. The negative correlation result of the first ratio of the target point is denoted as the saliency of the target point.
3. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 1, characterized in that, The specific method for obtaining the saliency distribution similarity between the first-class points and the second-class points is as follows: The ratio of the number of points belonging to the same salient cluster within the local neighborhood of Class I points and Class II points is denoted as the first ratio of the same salient cluster. The ratio of the number of points belonging to the same salient cluster within BIM point cloud data and GIS point cloud data is denoted as the second ratio of the same salient cluster. The difference between the second ratio and the first ratio of the salient cluster is denoted as the first difference of the salient cluster. The negative correlation result of the first difference of all significant clusters is denoted as the significant distribution similarity between Class I and Class II points.
4. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 1, characterized in that, The method for calculating the shortest spatial distance of the target point in its local sub-neighborhood is as follows: For each local sub-neighborhood of the target point, perform surface fitting on all points contained therein to obtain the fitted surface within the local sub-neighborhood of the target point. The shortest Euclidean distance from the target point to the fitted surface within the local sub-neighborhood of the target point is denoted as the shortest spatial distance of the target point in the local sub-neighborhood.
5. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 1, characterized in that, The method for obtaining the comprehensive spatial structure similarity between the first-class points and the second-class points is as follows: Based on the difference in the shortest spatial distance between Class I and Class II points in their corresponding local sub-neighborhoods, calculate the spatial distance difference between Class I and Class II points in their corresponding local sub-neighborhoods; the negative correlation processing result of the spatial distance difference between Class I and Class II points in all their corresponding local sub-neighborhoods is denoted as the spatial structure similarity between Class I and Class II points. Based on the difference in the number of points belonging to the same salient cluster contained in the corresponding local sub-neighborhoods of Class I and Class II points, the difference in the number of points belonging to the same salient cluster in the corresponding local sub-neighborhoods of Class I and Class II points is calculated; the negative correlation processing result of the difference in the number of points belonging to the same salient cluster in all corresponding local sub-neighborhoods of Class I and Class II points is denoted as the salient spatial structure similarity between Class I and Class II points. The spatial structure similarity and salient spatial structure similarity are weighted and summed using the salient distribution similarity of Class I and Class II points. The result of the weighted summation is denoted as the comprehensive spatial structure similarity of Class I and Class II points.
6. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 5, characterized in that, The method for obtaining the spatial distance difference is as follows: The absolute value of the difference between the shortest spatial distances of Class I points and Class II points in their corresponding local sub-neighborhoods is denoted as the spatial distance difference between Class I points and Class II points in their corresponding local sub-neighborhoods.
7. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 5, characterized in that, The method for obtaining the difference in the number of members in the same cluster is as follows: The absolute value of the difference between the number of points belonging to the same salient cluster contained in the corresponding local sub-neighborhood of Class I points and Class II points is denoted as the difference in the number of points belonging to the same salient cluster in the corresponding local sub-neighborhood of Class I points and Class II points.
8. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 1, characterized in that, The method for achieving point cloud registration between BIM point cloud data and GIS point cloud data based on the similarity of point attributes and comprehensive spatial structure between the two data includes the following specific methods: Based on the similarity of point attributes and the comprehensive spatial structure of points in BIM point cloud data and GIS point cloud data, the point cloud similarity between Class I points and Class II points is calculated. The comprehensive spatial structure similarity of two matched points during the point cloud registration process is used as the reliability weight to establish a weighted ICP objective function. The ICP point cloud registration algorithm is then used to perform point cloud matching on BIM point cloud data and GIS point cloud data to obtain the point cloud matching results.
9. The high-fidelity simulation analysis method based on BIM+GIS technology according to claim 8, characterized in that, The method for calculating point cloud similarity is as follows: The mean of the attribute similarity and the comprehensive spatial structure similarity between Class I and Class II points is denoted as the point cloud similarity between Class I and Class II points.