Roadbed parameterized reverse modeling method and system based on point cloud geometry parameter regression
By preprocessing and feature extraction of roadbed point cloud data, combined with parametric regression methods, the problems of slow roadbed model reconstruction speed and low fit were solved, realizing fast and high-precision roadbed entity model reconstruction, supporting construction and analysis needs.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- RAILWAY CONSTR RES INST OF CHINA ACAD OF RAILWAY SCI CO LTD
- Filing Date
- 2026-05-15
- Publication Date
- 2026-07-31
AI Technical Summary
The creation of existing roadbed models relies on forward design, which results in slow reconstruction speed. Furthermore, it is difficult to guarantee the degree of geometric fit between the physical model and the actual roadbed geometry, making it difficult to meet the needs of geometric representation and business information management during the change process of construction components.
By preprocessing the original point cloud data of the roadbed, extracting the apparent feature information, calculating the spatial geometric parameters, creating a parametric component model, building a primitive model library, and reconstructing the roadbed entity model in combination with the spatial geometric parameters.
It enables the rapid creation of roadbed entity models that closely match the real surface, supports operation, maintenance, and simulation analysis, and has high reusability and flexible parameter adjustment capabilities.
Smart Images
Figure CN122492976A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of BIM model reverse reconstruction technology, and in particular to a roadbed parameterized reverse modeling method and system based on point cloud geometric parameter regression. Background Technology
[0002] As a core technology carrier for the digital transformation of transportation infrastructure, roadbed BIM models play a multi-dimensional enabling role in the entire life cycle management. They can not only intuitively present the spatial relationship between the route's horizontal and vertical profiles and topographic features, assisting in optimizing cut and fill schemes, slope protection methods, and drainage system layouts, but also conduct simulation analyses using finite element analysis tools to calculate operation and maintenance costs and damage levels.
[0003] Existing component identification and modeling methods are susceptible to problems such as missing small components and confusion between similar components due to data noise and occlusion, leading to model identification errors, inaccurate parameters, and component disconnection. Most current research focuses on improving modeling efficiency and parametric construction of specialized components, neglecting the dynamic updating of digital models and business information during railway construction. This makes it difficult to meet the needs of geometric representation and business information management for changes in construction components. Therefore, there is an urgent need for a method that can quickly create solid BIM models that match actual dimensions.
[0004] Reverse engineering, a digital technology that derives 3D models from physical objects or scene data, is based on the ability to quickly acquire high-precision point cloud data of target objects through 3D laser scanning, structured light scanning, or photogrammetry. Combined with point cloud processing and surface reconstruction algorithms, it generates editable 3D models, breaking through the limitations of traditional forward design that relies on original drawings or experience-based modeling. It can quickly create solid models with real 3D dimensions, providing a data foundation for construction and simulation analysis. Summary of the Invention
[0005] The purpose of this invention is to provide a roadbed parametric inverse modeling and system based on point cloud geometric parameter regression. It aims to solve the problems of existing roadbed model creation relying on forward design and slow reconstruction speed, as well as the difficulty in ensuring the degree of fit between the physical model and the real geometry of the roadbed. By leveraging the advantages of parametric regression inverse modeling, it enables the rapid reconstruction of roadbed physical models that fit the real surface, thereby providing data support for the operation, maintenance, and simulation analysis of the roadbed.
[0006] The embodiments of the present invention are implemented as follows:
[0007] A roadbed parameterization inverse modeling and system based on point cloud geometric parameter regression, comprising:
[0008] Preprocessing the original point cloud data of the roadbed yields a three-dimensional point cloud with uniform orientation.
[0009] Based on the directionally unified 3D point cloud, the apparent feature information of the roadbed structure is extracted, and the apparent feature structure points are calculated.
[0010] Based on the apparent feature structure points, the spatial geometric parameters of the roadbed are obtained by regressing the geometric parameters according to the spatial structure of the roadbed.
[0011] Create a parametric component model of the roadbed and build a parametric primitive model library for the roadbed.
[0012] Based on the roadbed parameterized primitive model library, the roadbed spatial geometric parameters are coupled to create a roadbed entity model.
[0013] In a preferred embodiment of the present invention, the above-described method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression, wherein the preprocessing of the original roadbed point cloud data to obtain a directionally unified three-dimensional point cloud includes:
[0014] The PDAL library was used to read the original point cloud data of the roadbed.
[0015] The original point cloud data of the roadbed is simplified by voxel downsampling with a fixed grid size.
[0016] Statistical filtering is performed on the downsampled point cloud data to remove outliers with an average distance greater than a threshold, resulting in denoised clean point cloud data.
[0017] Based on the denoised clean point cloud data, PCA principal component analysis is performed to calculate the principal feature orientation of the point cloud.
[0018] A rotation matrix is constructed based on the principal feature direction, and the point cloud is rotated so that the principal feature direction is parallel to the X-axis of the global coordinate system, resulting in a three-dimensional point cloud with uniform orientation.
[0019] In a preferred embodiment of the present invention, in the above-mentioned roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, the step of extracting the apparent feature information of the roadbed structure based on the directional unified three-dimensional point cloud and calculating the apparent feature structure points includes:
[0020] Based on the three-dimensional point cloud with unified orientation, slices are made along the X-axis according to a set length threshold. The three-dimensional point cloud in each slice is projected onto the YZ plane to obtain a two-dimensional point cloud for characterizing the cross-sectional morphology of the roadbed.
[0021] For the two-dimensional point cloud of each cross section, the KNN algorithm is used to search for the set of neighboring points on the YZ plane. The covariance matrix is calculated based on the set of neighboring points. The eigenvector corresponding to the smallest eigenvalue of the covariance matrix is selected as the direction of the two-dimensional normal vector to construct the normal vector field of the cross section point cloud.
[0022] Based on the normal vector field of the cross-sectional point cloud, the rate of change of the angle between the normal vectors is calculated point by point, and the extreme point of the rate of change of the normal vector is found as the initial point of interest.
[0023] The initial interest points obtained through screening are filtered based on the prior elevation information of the roadbed structure, and points whose elevation values exceed the threshold are removed to obtain the apparent feature structure points that characterize the boundary positions of each structure of the roadbed.
[0024] In a preferred embodiment of the present invention, in the above-described method for roadbed parameterization inverse modeling based on point cloud geometric parameter regression, the step of calculating the rate of change of the normal vector angle point by point according to the normal vector field of the cross-sectional point cloud, and finding the extreme point of the rate of change of the normal vector as the initial point of interest includes:
[0025] Based on the normal vector field of the cross-sectional point cloud, the angle between the normal vector and the Y-axis direction is calculated point by point.
[0026] For each center point, calculate the variance of the angle between the center point and the normal vector of its neighboring points.
[0027] The variance value is used to characterize the degree of change of the angle between the normal vectors at the center point. Local extreme points of change are selected as initial points of interest to characterize the abrupt change in the geometry of the cross section.
[0028] In a preferred embodiment of the present invention, in the above-described method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression, the step of obtaining the roadbed spatial geometric parameters by regressing geometric parameters according to the spatial structure of the roadbed based on the apparent feature structure points includes:
[0029] Based on the apparent feature points, clustering is performed according to the distribution range of the Y-direction coordinates, and the apparent feature points that characterize different structural parts in the roadbed cross section, such as the shoulder and the toe of the slope, are respectively assigned to the corresponding groups.
[0030] For each group of apparent feature points, three-dimensional coordinate smoothing is performed to eliminate local fluctuations in the discrete point cloud, resulting in a smoothed spatial position sequence of each group of apparent feature points.
[0031] The YZ coordinates of each set of apparent feature points after smoothing are mapped to specific geometric corner points of the parameterized roadbed cross-section model, thus transforming the discrete feature points into a set of corner point coordinates with clear geometric meaning.
[0032] The spatial geometric parameters of the roadbed are obtained by calculating and regressing the set of corner coordinates based on the spatial structural relationship of the roadbed.
[0033] In a preferred embodiment of the present invention, in the above-mentioned roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, the roadbed spatial geometric parameters include at least one of roadbed height, top surface width, and elevation difference between adjacent roadbed segments.
[0034] In a preferred embodiment of the present invention, the above-described method for reverse modeling of roadbed parametric parameters based on point cloud geometric parameter regression includes the following steps: creating a roadbed parametric component model and constructing a roadbed parametric primitive model library.
[0035] Based on the roadbed structural characteristics defined by standard design specifications or existing engineering drawings for railway engineering, the roadbed is decomposed into multiple standardized structural levels and functional components.
[0036] Based on the subgrade decomposition results, an independent parametric component model is created for each structural component.
[0037] All the created parametric component models are integrated and managed to obtain a roadbed parametric primitive model library containing multi-level components that can be adjusted by parameters.
[0038] In a preferred embodiment of the present invention, in the above-mentioned roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, the parameterized component model includes at least one of the following: embankment parameterized primitive model, subgrade bottom layer parameterized primitive model, subgrade surface layer parameterized primitive model, and fastener array and track primitive model.
[0039] In a preferred embodiment of the present invention, in the above-described method for reverse modeling of roadbed parametric parameters based on point cloud geometric parameter regression, the step of creating a roadbed entity model by coupling the roadbed spatial geometric parameters according to the roadbed parametric primitive model library includes:
[0040] Based on the roadbed spatial geometric parameters, a corresponding roadbed parametric component model is selected from the roadbed parametric primitive model library. Using the spatial position and azimuth orientation resolved by the roadbed spatial geometric parameters as a reference, an instance of the roadbed parametric component model is created in three-dimensional space.
[0041] The dimensional attributes in the roadbed spatial geometric parameters are assigned to the created roadbed parametric component model instance, and the spatial and geometric attributes are adjusted so that the dimensions of the reconstructed model fit the original point cloud data of the roadbed.
[0042] By attaching semantic attributes identifying the structure type to the parametric component model instance of the roadbed, a roadbed entity model is created.
[0043] A roadbed parameterized inverse modeling system based on point cloud geometric parameter regression, comprising:
[0044] The data processing module is used to preprocess the original point cloud data of the roadbed to obtain a three-dimensional point cloud with uniform orientation.
[0045] The feature calculation module is used to extract the apparent feature information of the roadbed structure based on the three-dimensional point cloud with unified orientation, and to calculate the apparent feature structure points.
[0046] The geometric parameter regression module is used to regress geometric parameters according to the spatial structure of the roadbed based on the apparent feature structure points to obtain the spatial geometric parameters of the roadbed.
[0047] The model library construction module creates parametric component models of the roadbed and builds a parametric primitive model library for the roadbed.
[0048] The parametric reverse reconstruction module is used to create a roadbed entity model by coupling the roadbed spatial geometric parameters according to the roadbed parametric primitive model library.
[0049] The beneficial effects of the embodiments of the present invention are:
[0050] This invention obtains the actual size parameters of the roadbed point cloud through a specific method, which can more realistically measure the true size of the roadbed point cloud and better restore the actual model size.
[0051] The parameterized roadbed primitive model constructed in this invention has strong reusability, and the model parameters can be flexibly adjusted and coupled with each other, allowing for overall adjustment with only a few parameters.
[0052] This method uses point cloud parameters as the data basis and roadbed component primitives as the model basis. It adopts a model-data coupling approach for rapid batch reconstruction, which maintains spatial geometric properties while reflecting the true size of the spatial model. Attached Figure Description
[0053] To more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of the present invention and should not be regarded as a limitation on the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0054] Figure 1 This is a schematic diagram of the roadbed parameterized inverse modeling method based on point cloud geometric parameter regression according to the present invention. Detailed Implementation
[0055] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0056] Please refer to Figure 1 The first embodiment of the present invention provides a roadbed parametric inverse modeling and system based on point cloud geometric parameter regression, comprising: preprocessing the original roadbed point cloud data to obtain a directional unified three-dimensional point cloud; extracting the apparent feature information of the roadbed structure based on the directional unified three-dimensional point cloud, and calculating the apparent feature structure points; regressing the geometric parameters according to the spatial structure of the roadbed based on the apparent feature structure points to obtain the roadbed spatial geometric parameters; creating a roadbed parametric component model and constructing a roadbed parametric primitive model library; and creating a roadbed entity model by coupling the roadbed spatial geometric parameters according to the roadbed parametric primitive model library.
[0057] In a preferred embodiment of the present invention, the above-described method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression includes the following steps: preprocessing the original roadbed point cloud data to obtain a three-dimensional point cloud with uniform orientation: reading the original roadbed point cloud data using the PDAL library; simplifying the original roadbed point cloud data using a voxel downsampling method with a fixed grid size to ensure uniform point cloud density; performing statistical filtering on the downsampled point cloud data to remove outliers with an average distance greater than a threshold, such as adding n times the standard deviation to the mean, to obtain denoised clean point cloud data; performing PCA principal component analysis based on the denoised clean point cloud data to calculate the principal feature orientation of the point cloud; constructing a rotation matrix based on the principal feature orientation to rotate the point cloud so that the principal feature orientation is parallel to the X-axis of the global coordinate system, thereby obtaining a three-dimensional point cloud with uniform orientation.
[0058] For each point Calculate the average distance between the node and its k nearest neighbors, where k represents the number of points in the neighborhood. Point its neighboring points Calculate the Euclidean distance between all points. and standard deviation ;
[0059] If the average distance of a certain point Exceeding the threshold range ,or If it is an outlier, then it is determined to be an outlier.
[0060] In a preferred embodiment of the present invention, the above-mentioned method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression includes the following steps: extracting the apparent feature information of the roadbed structure based on the directionally unified 3D point cloud and calculating the apparent feature structure points: slicing the directionally unified 3D point cloud along the X-axis direction according to a set length threshold; projecting the 3D point cloud within each slice onto the YZ plane to obtain a 2D point cloud characterizing the cross-sectional morphology of the roadbed; for each cross-section of the 2D point cloud, searching for a set of neighboring points on the YZ plane using the KNN algorithm; calculating the covariance matrix based on the neighboring point set; selecting the eigenvector corresponding to the minimum eigenvalue of the covariance matrix as the 2D normal vector direction to construct a cross-sectional point cloud normal vector field; calculating the rate of change of the normal vector angle point by point based on the cross-sectional point cloud normal vector field; finding the extreme point of the rate of change of the normal vector as the initial interest point; filtering the selected initial interest points based on the prior elevation information of the roadbed structure, removing points whose elevation values exceed the threshold, and obtaining the apparent feature structure points characterizing the boundary positions of each structure of the roadbed.
[0061] Among them, the dimension with the largest variance is selected as the splitting dimension.
[0062] , Represents the eigenvalue set of the d-th dimension. This represents the variance of that dimension.
[0063] On the selected dimension d, the median is taken as the split point to ensure that the number of nodes in the left and right subtrees is balanced. Calculate Euclidean distance ,
[0064] Search path selection,
[0065] ,
[0066] Starting from the root node, select a subtree based on the value of the query point in the current split dimension. The pruning condition is...
[0067]
[0068] ,
[0069] If the distance from the query point to the splitting hyperplane is greater than the current best distance, then the other subtree cannot contain a closer neighbor.
[0070] Two-dimensional point cloud data ,in Calculate the mean and center , Construct the covariance matrix Describes the variance and covariance of the data distribution.
[0071] ,
[0072] in:
[0073] ,
[0074] ,
[0075] .
[0076] The formula for eigenvalue decomposition is: ,
[0077] in, This indicates that the eigenvalues are arranged in descending order. ,
[0078] The corresponding unit eigenvectors are orthogonal.
[0079] Project the original data onto the first k principal components to retain the eigenvectors with the smallest values. ,in, The data matrix is centered, with one point in each row. The data is after dimensionality reduction. If k=1, it is a one-dimensional projection.
[0080] In a preferred embodiment of the present invention, in the above-described roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, the step of calculating the rate of change of the normal vector angle point by point according to the normal vector field of the cross-sectional point cloud and finding the extreme points of the rate of change of the normal vector as initial points of interest includes: calculating the angle between the normal vector and the Y-axis direction point by point according to the normal vector field of the cross-sectional point cloud; calculating the variance of the angle between the normal vector and the normal vector of the neighboring points for each center point; using the value of the variance to characterize the degree of change of the normal vector angle at the center point, and selecting the local extreme points of the degree of change as initial points of interest characterizing the abrupt change in the cross-sectional geometry.
[0081] The variance of the angle between the normal vector of each two-dimensional point and the Z-axis direction is calculated as follows: .
[0082] If the difference between the current window variance and the variance of adjacent windows exceeds the threshold ϵ, it is marked as an extreme point. The window variance is... .
[0083] In a preferred embodiment of the present invention, the above-mentioned method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression, wherein obtaining the roadbed spatial geometric parameters by regressing geometric parameters according to the spatial structure of the roadbed based on the apparent feature points includes: clustering the apparent feature points according to the distribution interval of the Y-direction coordinates, and classifying the apparent feature points representing different structural parts such as shoulders and toes in the roadbed cross-section into corresponding groups; performing three-dimensional coordinate smoothing on each group of apparent feature points to eliminate local fluctuations in the discrete point cloud, and obtaining a smoothed spatial position sequence of each group of apparent feature points; mapping the YZ coordinates of each group of apparent feature points after smoothing to specific geometric corner points of the parametric roadbed cross-section model, and transforming the discrete feature points into a set of corner point coordinates with clear geometric meaning; and calculating and regressing the set of corner point coordinates according to the spatial structural relationship of the roadbed to obtain the roadbed spatial geometric parameters.
[0084] In a preferred embodiment of the present invention, in the above-mentioned roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, the roadbed spatial geometric parameters include at least one of roadbed height, top surface width, and elevation difference between adjacent roadbed segments.
[0085] In a preferred embodiment of the present invention, the above-mentioned method for reverse modeling of roadbed parametric parameters based on point cloud geometric parameter regression, wherein the creation of roadbed parametric component models and the construction of a roadbed parametric primitive model library includes: decomposing the roadbed into multiple standardized structural levels and functional components based on the roadbed structural features defined by standard design specifications for railway engineering or existing engineering drawings; creating an independent parametric component model for each structural component according to the roadbed decomposition results; and integrating and managing all the created parametric component models to obtain a roadbed parametric primitive model library containing multi-level components that can be adjusted by parameters.
[0086] Specifically, based on the railway subgrade structure drawings, the cross-sections of the embankment, the bottom layer of the subgrade, and the surface layer of the subgrade are drawn, and their parameters are set, including the length, height, and slope of the top and bottom sides. The length of the bottom side can be calculated from the length of the top side, and the following spatial topological constraints exist.
[0087] Embankment top side length = roadbed bottom side width
[0088] ,
[0089] For the upper and lower surfaces of the subgrade, there exists a relationship. In this context, subscript 2 represents the width of the bottom surface of the roadbed, and subscript 1 represents the width of the top surface of the bottom layer of the roadbed.
[0090] For the subgrade surface layer, its structure consists of the shoulder and the subgrade surface layer.
[0091] ,
[0092] ,
[0093] Wherein, is the horizontal width corresponding to the bottom of the shoulder, is the width of the top surface of the shoulder, and is the width of the subgrade surface layer;
[0094] shoulder pads ,
[0095] in, It corresponds to the shoulder height, and Pt corresponds to the stacking slope of the surfaces of the two adjacent structural layers.
[0096] Several cross-sections are placed to create solid models. By setting the height difference between the front and rear cross-sections as the slope, segment solid primitive components with longitudinal slopes are created. For the spatial topological constraints between the instances, their respective parameter information is set, and the length coupling between each level is achieved. The instance models of each level are pieced together, and the dimensions of each part are adjusted by uniformly planning parameter constraints to ensure that the spatial geometric dimensions of each level model are consistent and form an overall model.
[0097] In a preferred embodiment of the present invention, in the above-mentioned roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, the parameterized component model includes at least one of the following: embankment parameterized primitive model, subgrade bottom layer parameterized primitive model, subgrade surface layer parameterized primitive model, and fastener array and track primitive model.
[0098] In a preferred embodiment of the present invention, the above-described method for reverse modeling of roadbed parametric parameters based on point cloud geometric parameter regression, wherein the step of creating a roadbed entity model by coupling the roadbed spatial geometric parameters with the roadbed parametric primitive model library includes: selecting a corresponding roadbed parametric component model from the roadbed parametric primitive model library based on the roadbed spatial geometric parameters; creating an instance of the roadbed parametric component model in three-dimensional space based on the spatial position and azimuth orientation resolved by the roadbed spatial geometric parameters; assigning the size attributes in the roadbed spatial geometric parameters to the created roadbed parametric component model instance; adjusting the spatial and geometric attributes to make the size of the reconstructed model fit the original point cloud data of the roadbed; and attaching semantic attributes identifying the structure type to the roadbed parametric component model instance to create a roadbed entity model.
[0099] Specifically, based on the semantic tags of different attributes, the parameter file originating from the point cloud is parsed and stored according to semantics; the corresponding parameters in the primitive model are assigned values according to the spatial coordinates of the semantic parameters and the geometric parameter attributes, and a model instance is created at the spatial position given by the parameters, while rotating to the corresponding orientation according to the parameters defined by the "rotation" semantics.
[0100] A second embodiment of the present invention provides a roadbed parametric reverse modeling system based on point cloud geometric parameter regression, comprising: a data processing module for preprocessing the original roadbed point cloud data to obtain a directional unified three-dimensional point cloud; a feature calculation module for extracting the apparent feature information of the roadbed structure based on the directional unified three-dimensional point cloud and calculating the apparent feature structure points; a geometric parameter regression module for regressing geometric parameters according to the spatial structure of the roadbed based on the apparent feature structure points to obtain the roadbed spatial geometric parameters; a model library construction module for creating roadbed parametric component models and constructing a roadbed parametric primitive model library; and a parametric reverse reconstruction module for creating a roadbed entity model by coupling the roadbed spatial geometric parameters according to the roadbed parametric primitive model library.
[0101] The embodiments of this invention aim to protect a method and system for roadbed parameterized inverse modeling based on point cloud geometric parameter regression, which has the following effects:
[0102] 1. By performing preprocessing operations such as downsampling and ground filtering, cluttered ground features and non-ground point cloud data are removed from the roadbed point cloud to obtain roadbed surface point cloud data, thereby improving the accuracy of the point cloud.
[0103] 2. A highly reusable parametric primitive model library has been created, enabling the rapid creation of railway subgrade models with only a few parameter adjustments.
[0104] 3. Instance models are created using high-precision point cloud surface parameters and primitive models with reliable spatial topology. These models achieve realistic spatial representation while maintaining high real-world dimensions. Furthermore, semantic attribute information can be attached to the models, providing a reliable basis for the use and analysis of roadbed models.
[0105] The computer program product of the roadbed parameterized inverse modeling method and apparatus based on point cloud geometric parameter regression provided in this invention includes a computer-readable storage medium storing program code. The instructions included in the program code can be used to execute the methods in the preceding method embodiments. For specific implementation, please refer to the method embodiments, which will not be repeated here.
[0106] Specifically, the storage medium can be a general-purpose storage medium, such as a portable disk or hard disk. When the computer program on the storage medium is run, it can execute the above-mentioned roadbed parameterized inverse modeling method based on point cloud geometric parameter regression, thereby enabling rapid reconstruction of a roadbed entity model that fits the real surface.
[0107] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this invention, essentially, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0108] Finally, it should be noted that the above-described embodiments are merely specific implementations of the present invention, used to illustrate the technical solutions of the present invention, and not to limit it. The scope of protection of the present invention is not limited thereto. Although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments within the technical scope disclosed in the present invention, or make equivalent substitutions for some of the technical features; and these modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention, and should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.
Claims
1. A method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression, characterized in that, include: Preprocessing the original point cloud data of the roadbed yields a three-dimensional point cloud with uniform orientation; Based on the directional unified three-dimensional point cloud, the apparent feature information of the roadbed structure is extracted, and the apparent feature structure points are calculated. Based on the apparent feature structure points, the spatial geometric parameters of the roadbed are obtained by regressing the geometric parameters according to the spatial structure of the roadbed. Create parametric component models of the roadbed and build a parametric primitive model library for the roadbed; Based on the roadbed parameterized primitive model library, the roadbed spatial geometric parameters are coupled to create a roadbed entity model.
2. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 1, characterized in that, The preprocessing of the original roadbed point cloud data to obtain a 3D point cloud with uniform orientation includes: The PDAL library is used to read the original point cloud data of the roadbed; The original point cloud data of the roadbed is simplified by voxel downsampling with a fixed grid size; Statistical filtering is performed on the downsampled point cloud data to remove outliers with an average distance greater than a threshold, resulting in denoised clean point cloud data. Based on the denoised clean point cloud data, PCA principal component analysis is performed to calculate the principal feature orientation of the point cloud. A rotation matrix is constructed based on the principal feature direction, and the point cloud is rotated so that the principal feature direction is parallel to the X-axis of the global coordinate system, resulting in a three-dimensional point cloud with uniform orientation.
3. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 2, characterized in that, The apparent feature information of the roadbed structure is extracted from the three-dimensional point cloud based on orientation unification, and the apparent feature structure points are calculated, including: Based on the three-dimensional point cloud with unified orientation, slices are made along the X-axis according to a set length threshold. The three-dimensional point cloud in each slice is projected onto the YZ plane to obtain a two-dimensional point cloud for characterizing the cross-sectional shape of the roadbed. For the two-dimensional point cloud of each cross section, the KNN algorithm is used to search for the set of neighboring points on the YZ plane. The covariance matrix is calculated based on the set of neighboring points. The eigenvector corresponding to the smallest eigenvalue of the covariance matrix is selected as the direction of the two-dimensional normal vector to construct the normal vector field of the cross section point cloud. Based on the normal vector field of the cross-sectional point cloud, the rate of change of the angle between the normal vectors is calculated point by point, and the extreme point of the rate of change of the normal vector is found as the initial point of interest. The initial interest points obtained through screening are filtered based on the prior elevation information of the roadbed structure, and points whose elevation values exceed the threshold are removed to obtain the apparent feature structure points that characterize the boundary positions of each structure of the roadbed.
4. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 3, characterized in that, The step of calculating the rate of change of the angle between the normal vectors point by point based on the normal vector field of the cross-sectional point cloud, and finding the extreme points of the rate of change of the normal vectors as the initial points of interest includes: Based on the normal vector field of the cross-sectional point cloud, calculate the angle between the normal vector and the Y-axis direction point by point; For each center point, calculate the variance of the angle between it and the normal vectors of its neighboring points; The variance value is used to characterize the degree of change of the angle between the normal vectors at the center point. Local extreme points of change are selected as initial points of interest to characterize the abrupt change in the geometry of the cross section.
5. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 1, characterized in that, The process of obtaining the subgrade spatial geometric parameters based on the apparent feature structure points and the regression geometric parameters according to the subgrade spatial structure includes: Based on the apparent feature points, clustering is performed according to the distribution range of the Y-direction coordinates, and the apparent feature points that characterize different structural parts in the roadbed cross section are respectively assigned to the corresponding groups; For each group of apparent feature points, three-dimensional coordinate smoothing is performed to eliminate local fluctuations in the discrete point cloud, resulting in a smoothed spatial position sequence of each group of apparent feature points. After smoothing, the YZ coordinates of each group of apparent feature points are mapped to specific geometric corner points of the parameterized roadbed cross-section model, thus transforming the discrete feature points into a set of corner point coordinates. The spatial geometric parameters of the roadbed are obtained by calculating and regressing the set of corner coordinates based on the spatial structural relationship of the roadbed.
6. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 5, characterized in that, The roadbed spatial geometric parameters include at least one of the following: roadbed height, top surface width, and elevation difference between adjacent roadbed sections.
7. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 1, characterized in that, The creation of the roadbed parametric component model and the construction of the roadbed parametric primitive model library include: The roadbed is decomposed into multiple standardized structural levels and functional components; Based on the subgrade decomposition results, create an independent parametric component model for each structural component; All the created parametric component models are integrated and managed to obtain a roadbed parametric primitive model library.
8. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 7, characterized in that, The parametric component model includes at least one of the following: embankment parametric primitive model, subgrade bottom layer parametric primitive model, subgrade surface layer parametric primitive model, and fastener array and track primitive model.
9. The method for parametric inverse modeling of roadbed based on point cloud geometric parameter regression according to claim 1, characterized in that, The step of creating a roadbed solid model by coupling the roadbed spatial geometric parameters according to the roadbed parameterized primitive model library includes: Based on the roadbed spatial geometric parameters, a corresponding roadbed parametric component model is selected from the roadbed parametric primitive model library. Using the spatial position and azimuth orientation resolved by the roadbed spatial geometric parameters as a reference, an instance of the roadbed parametric component model is created in three-dimensional space. Assign the dimensional attributes in the roadbed spatial geometric parameters to the created roadbed parametric component model instance, and adjust the spatial and geometric attributes to make the dimensions of the reconstructed model fit the original point cloud data of the roadbed. By attaching semantic attributes identifying the structure type to the parametric component model instance of the roadbed, a roadbed entity model is created.
10. A roadbed parameterized inverse modeling system based on point cloud geometric parameter regression, characterized in that, include: The data processing module is used to preprocess the original point cloud data of the roadbed to obtain a three-dimensional point cloud with uniform orientation. The feature calculation module is used to extract the apparent feature information of the roadbed structure based on the three-dimensional point cloud with unified orientation, and to calculate the apparent feature structure points. The geometric parameter regression module is used to regress geometric parameters according to the spatial structure of the roadbed based on the apparent feature structure points to obtain the spatial geometric parameters of the roadbed. The model library construction module creates parametric component models of the roadbed and builds a parametric primitive model library for the roadbed. The parametric reverse reconstruction module is used to create a roadbed entity model by coupling the roadbed spatial geometric parameters according to the roadbed parametric primitive model library.