Methods, devices, and media for heterogeneous 3D road modeling based on 2D slice data

Abstract

This invention discloses a method, device, and medium for modeling heterogeneous 3D roads based on 2D slice data. The method includes: classifying and assigning values ​​based on material visual features; converting a 2D road cross-section image into a cross-section matrix, where each unit in the cross-section matrix represents the material category of a pixel; extracting aggregate feature sets based on the cross-section matrix; generating 2D aggregate contours by boundary segmentation of the aggregate pixels extracted from the cross-section matrix; stretching the 2D aggregate contours into 3D aggregate voxels; registering and fusing aggregate voxels in adjacent slices of the 3D aggregate voxels to generate a 3D voxel field; extracting triangular meshes based on the 3D voxel field; and constructing a heterogeneous 3D road model based on the triangular meshes and the cross-section matrix. This improves the balance between the accuracy of 3D heterogeneous structure reconstruction and the ability to characterize damage features, ensuring a comprehensive improvement in modeling efficiency, realism, and damage characterization accuracy in 3D road model construction tasks.

Description

Technical Field

[0001] This invention relates to the field of road engineering technology, and in particular to a method, device, and medium for modeling heterogeneous three-dimensional roads based on two-dimensional slice data. Background Technology

[0002] In the field of road engineering, constructing heterogeneous three-dimensional road models that can realistically reflect internal damage and hidden defects is the data foundation for intelligent assessment of long-term pavement performance, prediction of defect evolution, and targeted maintenance. The mechanical behavior of pavement materials such as asphalt mixtures and cement concrete is significantly affected by aggregate morphology, spatial distribution, binder filling state, and internal microcracks, voids, and other damage defects. Especially in actual road conditions, many defects originate from the accumulation of heterogeneity and localized damage in the material's internal microstructure, exhibiting significant concealment and gradual development characteristics. Therefore, establishing heterogeneous three-dimensional realistic road models that can accurately characterize the internal composition, defect distribution, and hidden damage state of materials is a crucial foundation for road material simulation analysis, service safety evaluation, and maintenance decisions.

[0003] However, existing 3D road modeling methods face significant challenges when dealing with real road structures with hidden damage, including high modeling difficulty, insufficient realism, and difficulty in mass production. Current mainstream solutions rely on two types of modeling methods not based on 2D slice data. While these methods can achieve basic road model construction and preliminary representation of material composition, they struggle to meet the core engineering requirements of "realism, efficiency, and adaptability" when facing real road scenarios with internal damage and hidden defects. Furthermore, they lack adaptive modeling schemes based on multi-section data, failing to accurately characterize the heterogeneous features and damage distribution of materials.

[0004] Specifically, one type is the simulation software creation method such as PFC3D: although the component parameters can be controlled by manually adjusting the parameters, the model is too idealized. It cannot accurately reproduce the microscopic characteristics of real road materials (such as irregular shape of aggregates and heterogeneous distribution of components), nor can it simulate the geometric characteristics of hidden damage such as microcracks and voids, resulting in significant deviations from actual road conditions. At the same time, the parameter debugging and component reconstruction process is cumbersome, the amount of calculation is huge, which leads to slow model generation, the modeling of a single sample takes a long time, and it relies on experience debugging, making it impossible to quickly generate a large number of heterogeneous samples that reflect different damage states.

[0005] Another type is CT and laser scanning modeling: although it can rely on real material scanning to ensure the basic authenticity of the model, high-precision scanning is more expensive for samples with damage and hidden defects, and the preprocessing process such as noise removal of damaged areas and segmentation of components and damage boundaries in the scan data is more complicated. At the same time, the model has weak generalization ability and can only be adapted to specific scan samples. The scanning and reconstruction cycle of a single sample is long, making it difficult to support the construction of a large-scale road model library with multiple damage states, which restricts the application of disease statistical analysis and life prediction. Summary of the Invention

[0006] The main objective of this invention is to provide a method, device, and medium for modeling heterogeneous three-dimensional roads based on two-dimensional slice data. This invention aims to solve the technical problems of existing technologies, which are unable to meet the core engineering requirements when facing real road scenarios with internal damage and hidden defects, and lack adaptive modeling schemes based on multi-section data, thus failing to accurately characterize the heterogeneous features and damage distribution of materials.

[0007] To achieve the above objectives, the present invention provides a method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data, the method comprising the following steps:

[0008] A two-dimensional structure of road materials is constructed, and a standardized two-dimensional cross-sectional view is generated. The two-dimensional structure of road materials includes aggregates, binders, and voids.

[0009] Based on the material visual feature classification assignment, the two-dimensional cross-section image is converted into a cross-section matrix, where each cell in the cross-section matrix represents the material category to which the pixel belongs.

[0010] Based on the cross section matrix, an aggregate feature set is extracted, and an aggregate probability distribution model is generated based on the aggregate feature set. A three-dimensional voxel field is generated based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of the aggregate and Fourier descriptors. The three-dimensional voxel field is generated by registering and fusing three-dimensional aggregate voxel blocks in multiple adjacent slices. The three-dimensional aggregate voxel blocks are obtained by stretching the initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of the aggregate pixels extracted from the cross section matrix.

[0011] Triangular meshes are extracted based on the three-dimensional voxel field, and a heterogeneous three-dimensional road model is constructed based on the triangular meshes and the sectional matrix.

[0012] Optionally, the step of extracting aggregate feature set based on the cross-section matrix, generating aggregate probability distribution model based on the aggregate feature set, and generating three-dimensional voxel field based on the aggregate probability distribution model includes:

[0013] The adjacent different aggregate pixels in the sectional matrix are divided to obtain multiple initial aggregate regions, and noise is removed from the initial aggregate regions to obtain aggregate connected regions, which represent a two-dimensional aggregate cross section.

[0014] Two-dimensional morphological features and Fourier descriptors of each aggregate connected component are extracted, and an aggregate feature set is constructed. The Fourier descriptors are extracted based on the following formula:

[0015]

[0016] in, Indicates the first Each aggregate connectivity domain The Fourier descriptor, This represents the index of the sampling point on the aggregate profile. This represents the total number of sampling points on the aggregate profile. Indicates the first On the outline of the aggregate Complex coordinates of each sampling point This indicates the order of the Fourier descriptor. The basis functions represent the discrete Fourier transform;

[0017] Based on the aggregate feature set of each aggregate connected domain, shape regularity analysis is performed to generate an aggregate probability distribution model;

[0018] Three-dimensional aggregate voxel blocks are generated based on the aggregate probability distribution model.

[0019] Based on the aggregate feature set, the aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block are registered and fused to generate a three-dimensional voxel field.

[0020] Optionally, the aggregate probability distribution model includes a particle size distribution model, an aspect ratio distribution model, and an orientation distribution model; the step of generating the aggregate probability distribution model by analyzing the shape regularity of the aggregate feature set of each aggregate connected domain includes:

[0021] Kernel density analysis is performed on each aggregate connected region based on the aggregate feature set to obtain the equivalent radius. The two-dimensional morphological features in the aggregate feature set include the area, perimeter, minor axis, and orientation angle of the aggregate connected region. The equivalent radius is calculated with reference to the following formula:

[0022]

[0023] in, Indicates the first The equivalent radius of a connected domain of aggregates Indicates the first The area of ​​each aggregate connected domain on a two-dimensional cross-section;

[0024] A particle size distribution model is constructed based on the equivalent radius of each aggregate connected domain, referring to the following formula:

[0025]

[0026] in, The probability density function represents the equivalent radius of the aggregate, and the probability density function of the equivalent radius of the aggregate is a particle size distribution model. This indicates the total number of aggregate samples. Indicates kernel density bandwidth. Represents the kernel function. The independent variable representing the equivalent radius, This indicates that the equivalent radius has been standardized for calculating the sample point's position relative to the current value. The contribution of the probability density at that location;

[0027] The aspect ratio distribution parameters of each aggregate connected region are determined based on the major axis and minor axis of the least circumscribed ellipse of each aggregate connected region, and an aspect ratio distribution model is constructed based on the aspect ratio distribution parameters of each aggregate connected region. The aspect ratio distribution parameters are calculated based on the following formula:

[0028]

[0029] in, Indicates the first The minimum circumscribed ellipse major axis of each aggregate connected domain Indicates the first The minimum circumscribed ellipse minor axis of a connected domain of aggregates Indicates the first The aspect ratio distribution parameters of each aggregate connected domain;

[0030] A directional distribution model is constructed based on the directional angles of each aggregate's connected domains.

[0031] Optionally, generating three-dimensional aggregate voxel blocks based on the aggregate probability distribution model includes:

[0032] An initial two-dimensional aggregate profile is generated based on the aggregate probability distribution model.

[0033] The initial two-dimensional aggregate profile is noise-enhanced using multi-scale Perlin noise to generate aggregate profile perturbation, as shown in the following formula:

[0034]

[0035] in, Indicates the first aggregate in the connected domain The perturbation amount of each contour point Indicates the number of noise scales. Indicates the first Weights for each noise scale, Represents the Perlin noise function. Indicates the first The angle of each contour point Indicates the noise frequency;

[0036] The initial two-dimensional aggregate profile after noise enhancement is stretched into a three-dimensional structure, and the three-dimensional structure is converted into a three-dimensional aggregate voxel block. The profile stretching is performed according to the following formula:

[0037]

[0038] in, Indicates control height The stretching parameters of the aggregate profile are used to control the scaling and shape changes of the stretched aggregate profile. This indicates that it follows a Gaussian process. Represents the mean function, Represents the kernel function, used to describe height. With height The similarity between them.

[0039] Optionally, the step of registering and fusing aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block based on the aggregate feature set to generate a three-dimensional voxel field includes:

[0040] Based on the aggregate feature set, the aggregate similarity between aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block is calculated, and the aggregate similarity is calculated with reference to the following formula:

[0041]

[0042] in, This represents the first adjacent slice in a three-dimensional aggregate voxel block. The aggregate carboxyl block and the first Similarity between individual aggregate voxel blocks and They represent the first One aggregate carboxyl block and the first Two-dimensional shape characteristics of an aggregate voxel block Indicates the first The aggregate carboxyl block and the first Differences in shape characteristics between individual aggregate voxel blocks This represents a preset hyperparameter used to control the sensitivity of morphological feature differences to the similarity score. and They represent the first One aggregate carboxyl block and the first Fourier descriptor of a single aggregate voxel block The similarity function representing Fourier descriptors;

[0043] Based on the aggregate similarity, registerable voxel pairs are selected from the three-dimensional aggregate voxel blocks. The registerable voxel pairs consist of two aggregate voxel blocks whose slice positions are adjacent and whose aggregate similarity exceeds the similarity threshold.

[0044] Based on the aggregate similarity, the registration weights of each pair of voxels to be registered are determined, and the registration of the voxels to be registered is optimized based on the registration weights to obtain the registration result. The registration optimization refers to the following formula:

[0045]

[0046] in, Represents spatial transformation Find the minimum value. Representing spatial transformation The corresponding total error function, Indicates the volume pairs to be registered Registration weights, This indicates the volume element alignment deviation term. Indicates the aggregate volume element Apply spatial transformation , Indicates the relationship with aggregate volume elements Aggregate volumetric elements in adjacent slice layers, Represents the coefficient of the regularization term. Indicates a smoothing regularization term;

[0047] A conflict penalty term is constructed, and the registration result is corrected based on the conflict penalty term. Target registration voxel pairs are then determined based on the corrected registration results, and these target voxel pairs are fused to generate a three-dimensional voxel field. The conflict penalty term is defined by the following formula:

[0048]

[0049] in, This represents the total error function after adding conflict penalty. Indicates the conflict penalty coefficient. Indicates the overlap tolerance threshold. Indicates aggregate volume and The intersection volume represents the volume of the overlapping portion. This represents the conflict exponential function.

[0050] Optionally, the step of extracting a triangular mesh based on the three-dimensional voxel field and constructing a heterogeneous three-dimensional road model based on the triangular mesh and the tangent matrix includes:

[0051] The three-dimensional voxel field is binarized;

[0052] Isosurfaces are extracted from the binarized 3D voxel field to obtain a triangular mesh;

[0053] The triangular meshes are integrated to obtain a continuous three-dimensional aggregate skin;

[0054] Two-dimensional distribution information of binder, aggregate and void is extracted from the cross-section matrix. Combined with the spatial position information of the slice matrix, the two-dimensional distribution information is mapped to three-dimensional space to obtain the three-dimensional distribution area of ​​binder, aggregate and void.

[0055] Based on the three-dimensional aggregate skin, the three-dimensional distribution areas of binder, aggregate and voids are structurally filled to construct a heterogeneous three-dimensional road model.

[0056] Furthermore, to achieve the above objectives, the present invention also proposes a heterogeneous three-dimensional road modeling device based on two-dimensional slice data, the heterogeneous three-dimensional road modeling device based on two-dimensional slice data comprising:

[0057] A two-dimensional slice data generation module is used to construct the two-dimensional structure of road materials and generate standardized two-dimensional cross-sectional images. The two-dimensional structure of road materials includes aggregates, binders, and voids.

[0058] The matrix transformation module is used to classify and assign values ​​based on material visual features, and to convert the two-dimensional cross-section image into a cross-section matrix, where each unit in the cross-section matrix represents the material category to which the pixel belongs;

[0059] The feature extraction module is used to extract aggregate feature set based on the cross section matrix, generate aggregate probability distribution model based on the aggregate feature set, and generate three-dimensional voxel field based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of aggregate and Fourier descriptors. The three-dimensional voxel field is generated by registering and fusing three-dimensional aggregate voxel blocks in multiple adjacent slices. The three-dimensional aggregate voxel blocks are obtained by stretching the initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of aggregate pixels extracted from the cross section matrix.

[0060] A three-dimensional road model construction module is used to extract triangular meshes based on the three-dimensional voxel field, and to construct a heterogeneous three-dimensional road model based on the triangular meshes and the tangent matrix.

[0061] Furthermore, to achieve the above objectives, this application also proposes a heterogeneous three-dimensional road modeling device based on two-dimensional slice data. The device includes: a memory, a processor, and a heterogeneous three-dimensional road modeling program based on two-dimensional slice data stored in the memory. The processor is used to run the heterogeneous three-dimensional road modeling program based on two-dimensional slice data. The computer program is configured to implement the steps of the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described above.

[0062] In addition, to achieve the above objectives, this application also proposes a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps of the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described above.

[0063] In addition, to achieve the above objectives, this application also provides a computer program product, which includes a computer program that, when executed by a processor, implements the steps of the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described above.

[0064] This invention constructs a two-dimensional structure of road materials and generates standardized two-dimensional cross-sectional images. The two-dimensional structure includes aggregates, binders, and voids. Based on visual feature classification and assignment of materials, the two-dimensional cross-sectional images are converted into a cross-sectional matrix. Each unit in the cross-sectional matrix represents the material category of a pixel. An aggregate feature set is extracted based on the cross-sectional matrix, and an aggregate probability distribution model is generated based on this feature set. A three-dimensional voxel field is then generated based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of the aggregates and Fourier descriptors. The three-dimensional voxel field is formed by registering and fusing three-dimensional aggregate voxel blocks from multiple adjacent slices. The three-dimensional aggregate voxel block is generated by stretching an initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of aggregate pixels extracted from the sectional matrix. Triangular meshes are extracted based on the three-dimensional voxel field, and a heterogeneous three-dimensional road model is constructed based on the triangular meshes and the sectional matrix. This breaks through the limitations of existing single empirical parameters or single sectional surfaces, optimizes the balance between the utilization efficiency of two-dimensional slice data, the accuracy of three-dimensional heterogeneous structure restoration, and the ability to characterize damage features, and ensures that in the three-dimensional road model construction task, a comprehensive improvement is achieved in modeling efficiency, model realism, damage characterization accuracy, and engineering adaptability. Attached Figure Description

[0065] 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, for those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0066] Figure 1 This is a schematic diagram of the structure of a heterogeneous 3D road modeling device based on 2D slice data, which is part of the hardware operating environment of the embodiment of the present invention.

[0067] Figure 2 This is a flowchart illustrating the first embodiment of the heterogeneous 3D road modeling method based on 2D slice data according to the present invention.

[0068] Figure 3 This is a schematic diagram of a two-dimensional cross-section in one embodiment of the non-homogeneous three-dimensional road modeling method based on two-dimensional slice data of the present invention;

[0069] Figure 4 This is a structural block diagram of the first embodiment of the heterogeneous three-dimensional road modeling device based on two-dimensional slice data of the present invention.

[0070] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0071] It should be understood that the specific embodiments described herein are for illustrative purposes only and are not intended to limit the scope of the invention.

[0072] Reference Figure 1 , Figure 1 This is a schematic diagram of the structure of a heterogeneous 3D road modeling device based on 2D slice data, which is part of the hardware operating environment of the embodiment of the present invention.

[0073] like Figure 1As shown, the heterogeneous 3D road modeling device based on 2D slice data may include: a processor 1001, such as a central processing unit (CPU), a communication bus 1002, a user interface 1003, a network interface 1004, and a memory 1005. The communication bus 1002 is used to enable communication between these components. The user interface 1003 may include a display screen or an input unit such as a keyboard; optionally, the user interface 1003 may also include a standard wired interface or a wireless interface. The network interface 1004 may optionally include a standard wired interface or a wireless interface (such as a Wireless-Fidelity (Wi-Fi) interface). The memory 1005 may be high-speed random access memory (RAM) or stable non-volatile memory (NVM), such as a disk storage device. Optionally, the memory 1005 may also be a storage device independent of the aforementioned processor 1001.

[0074] Those skilled in the art will understand that Figure 1 The structure shown does not constitute a limitation on a heterogeneous 3D road modeling device based on 2D slice data, and may include more or fewer components than shown, or combine certain components, or have different component arrangements.

[0075] like Figure 1 As shown, the memory 1005, which is a computer-readable storage medium, may include an operating system, a network communication module, a user interface module, and a non-homogeneous three-dimensional road modeling program based on two-dimensional slice data.

[0076] exist Figure 1 In the heterogeneous 3D road modeling device based on 2D slice data shown, the network interface 1004 is mainly used for data communication with the network server; the user interface 1003 is mainly used for data interaction with the user; the processor 1001 and memory 1005 in the heterogeneous 3D road modeling device based on 2D slice data of the present invention can be set in the heterogeneous 3D road modeling device based on 2D slice data. The heterogeneous 3D road modeling device based on 2D slice data calls the heterogeneous 3D road modeling program based on 2D slice data stored in the memory 1005 through the processor 1001 and executes the heterogeneous 3D road modeling method based on 2D slice data provided in the embodiment of the present invention.

[0077] This invention provides a method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data, referring to... Figure 2 , Figure 2This is a flowchart illustrating the first embodiment of the non-homogeneous three-dimensional road modeling method based on two-dimensional slice data according to the present invention.

[0078] In this embodiment, the method for modeling heterogeneous 3D roads based on 2D slice data includes the following steps:

[0079] Step S10: Construct a two-dimensional structure for road materials and generate a standardized two-dimensional cross-sectional view. The two-dimensional structure of the road materials includes aggregates, binders, and voids.

[0080] It should be understood that the executing entity of this embodiment can be a computing service device with data processing, network communication, and program execution functions, such as a tablet computer, personal computer, or mobile phone, or a terminal electronic device capable of performing the above functions. The following description uses a non-homogeneous 3D road modeling device based on 2D slice data (hereinafter referred to as the modeling device) as an example to illustrate this embodiment and the following embodiments.

[0081] It should be noted that the two-dimensional structure of road materials can be a two-dimensional digital model constructed by simulation software, which can reflect the spatial distribution relationship of the core components (aggregate, binder, and voids) of road materials. It is the basic carrier for generating standardized two-dimensional cross-sectional diagrams and has the characteristics of being adjustable and repeatable.

[0082] Standardized two-dimensional cross-sectional images can be images of road material cross-sections with uniform size, resolution, coordinate specifications, and parameter annotations. Their core function is to eliminate the format differences of different cross-sectional data and provide consistent input for subsequent batch data processing.

[0083] In practical implementation, the modeling equipment can use the discrete element method software PFC2D (Particle Flow Code in 2 Dimensions) to construct a two-dimensional structure of road materials, including aggregates, binders, and voids, directly generating standardized two-dimensional cross-sectional diagrams. This provides a controllable and precise input source for subsequent slice matrix data reading, referencing... Figure 3 , Figure 3 This is a schematic diagram of a two-dimensional cross-section in one embodiment.

[0084] Step S20: Based on the material visual feature classification and assignment, the two-dimensional cross-sectional image is converted into a cross-sectional matrix, where each unit in the cross-sectional matrix represents the material category to which the pixel belongs.

[0085] It should be noted that the visual characteristics of materials can be the distinguishable visual attributes of each component of the road material in a two-dimensional cross-sectional image, such as grayscale value, texture roughness, color depth, etc., which are the core basis for accurately distinguishing material categories.

[0086] A cross-section matrix is ​​a digital carrier that uses a two-dimensional matrix as its data structure to store the material category of each pixel in a two-dimensional cross-section image. The row and column indices of the matrix correspond to the pixel coordinates of the cross-section image, and the numerical values ​​of the matrix cells directly represent the material category at the corresponding position, which facilitates rapid numerical calculations and feature extraction by computers.

[0087] In practice, the modeling equipment takes the interface diagram of the road structure cross section, including materials such as asphalt, aggregate, cement, and air, as input. Using the pixels of the real road structure cross section diagram as the unit, it classifies and assigns values ​​to each component based on the visual characteristics (color, texture) of the materials and converts them into matrix form for storage.

[0088] The section matrix is ​​denoted as:

[0089]

[0090] in, , This indicates the height and width of the cross-section. Indicates the material category to which the pixel belongs, such as:

[0091] : Represents the matrix number line, number The material corresponding to the column position is void;

[0092] : Represents the matrix number line, number The material corresponding to the column position is a binder;

[0093] : Represents the matrix number line, number The material corresponding to the column position is aggregate.

[0094] Step S30: Extract aggregate feature set based on the cross-section matrix, generate aggregate probability distribution model based on the aggregate feature set, and generate three-dimensional voxel field based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of aggregate and Fourier descriptors. The three-dimensional voxel field is a set composed of multiple sliced ​​aggregate voxel blocks.

[0095] It should be noted that the aggregate feature set is a combination of features used to comprehensively characterize aggregate properties. Its core includes two-dimensional morphological features that reflect the macroscopic geometric shape of aggregates and Fourier descriptors that characterize the microscopic texture of aggregate contours. It is the core basis for realizing aggregate identification, matching and three-dimensional generation.

[0096] Two-dimensional morphological aggregates refer to the macroscopic geometric properties presented in a two-dimensional cross-section. They can intuitively reflect the size, shape, and spatial orientation of the aggregates and are the basic characteristics for distinguishing different aggregates.

[0097] Fourier descriptors can be frequency domain feature parameters obtained by performing a discrete Fourier transform on the aggregate profile curve. They have rotation, translation, and scaling invariance and can accurately characterize the micro-texture and irregular details of the profile.

[0098] A three-dimensional voxel field is a three-dimensional digital space set formed by registering, optimizing and integrating multiple sliced ​​aggregate voxel blocks. It contains the three-dimensional morphology, spatial location and distribution relationship of all aggregates and is the core data carrier for constructing a three-dimensional road model.

[0099] Aggregate voxel blocks are aggregate entities composed of basic units (voxels) in three-dimensional space. They are discretized representations of the three-dimensional morphology of aggregates and can accurately depict the three-dimensional geometric contours and spatial proportions of aggregates.

[0100] In a specific implementation, the three-dimensional voxel field is generated by registering and fusing three-dimensional aggregate voxel blocks in multiple adjacent slices. The three-dimensional aggregate voxel blocks are obtained by stretching the initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of the aggregate pixels extracted from the sectional matrix.

[0101] Understandably, this embodiment achieves comprehensive capture of the macroscopic morphology and microscopic texture of aggregates through multi-dimensional feature extraction, providing accurate feature basis for subsequent 3D generation; based on probability distribution model and natural noise enhancement technology, it ensures that the generated 3D aggregate voxel blocks conform to the statistical laws of real aggregates, improving the realism of the model; through registration optimization and conflict detection, it ensures the spatial continuity and non-overlap of the 3D voxel field, providing high-quality and highly reliable core data support for subsequent 3D modeling.

[0102] Furthermore, in order to accurately extract aggregate features, step S30 above may include:

[0103] Step S301: Divide adjacent different aggregate pixels in the section matrix to obtain multiple initial aggregate regions, and remove noise from the initial aggregate regions to obtain aggregate connected regions, wherein the aggregate connected regions represent a two-dimensional aggregate cross section.

[0104] Step S302: Extract the two-dimensional morphological features and Fourier descriptors of each aggregate connected domain, and construct an aggregate feature set;

[0105] Step S303: Analyze the shape regularity based on the aggregate feature set of each aggregate connected domain to generate an aggregate probability distribution model;

[0106] Step S304: Generate a three-dimensional aggregate voxel block based on the aggregate probability distribution model;

[0107] Step S305: Based on the aggregate feature set, register and fuse the aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block to generate a three-dimensional voxel field.

[0108] In the specific implementation, for each connected component Extract the following morphological features:

[0109] area: ;

[0110] perimeter: ;

[0111] Minimum circumscribed ellipse major axis: ;

[0112] Minimum circumscribed minor axis of the ellipse: (Find an ellipse that "just covers the aggregate". The longest diameter of the ellipse is equal to the major axis and the shortest diameter is equal to the minor axis - these two numbers are used to describe the shape proportions of the aggregate.)

[0113] Direction angle: (The "angle of inclination" of the aggregate in the two-dimensional plane), the orientation angle will be used during registration to make the orientation of the three-dimensional aggregate consistent with the tangent.

[0114] Simultaneously, the Fourier descriptor of the contour is calculated:

[0115]

[0116] in, Indicates the first Each aggregate connectivity domain The Fourier descriptor, This represents the index of the sampling point on the aggregate profile. This represents the total number of sampling points on the aggregate profile. Indicates the first On the outline of the aggregate Complex coordinates of each sampling point (two-dimensional coordinates) , ) represents a complex number: , (It is the imaginary unit) This indicates the order of the Fourier descriptor (corresponding to the "frequency component" in the frequency domain). The smaller the value, the lower the frequency. These represent the basis functions of the Discrete Fourier Transform, used to realize the conversion from the spatial domain to the frequency domain.

[0117] Calculating the Fourier descriptor is equivalent to performing a Fourier transform on the aggregate boundary, which is used to describe the "frequency domain feature fingerprint" of the contour shape.

[0118] Assuming aggregate connected domain The outline is an elliptical shape. Its Fourier descriptor is: high low-frequency components (because the shape is smooth) and low high-frequency components (the boundaries are not jagged).

[0119] If the contour is very complex (like a pebble with many uneven edges): then: a higher high-frequency component indicates a higher "surface roughness");

[0120] Fourier descriptors are used for morphological similarity calculation during subsequent registration;

[0121] The purpose of constructing the aggregate feature set here is to connect the domains of each aggregate piece. Transform into a set of numerical features:

[0122]

[0123] Furthermore, in order to achieve comprehensive capture of the macroscopic morphology and microscopic texture of the aggregate, in one embodiment, the aggregate probability distribution model includes a particle size distribution model, an aspect ratio distribution model, and an orientation distribution model. Step S303 may include:

[0124] Step S3031: Perform kernel density analysis on each aggregate connected region based on the aggregate feature set of each aggregate connected region to obtain the equivalent radius. The two-dimensional morphological features in the aggregate feature set include the area, perimeter, minimum circumscribed ellipse major axis, minimum circumscribed ellipse minor axis, and orientation angle of the aggregate connected region.

[0125] Step S3032: Construct a particle size distribution model based on the equivalent radius of each aggregate connected domain;

[0126] Step S3033: Determine the aspect ratio distribution parameters of each aggregate connected region based on the minimum circumscribed ellipse major axis and minimum circumscribed ellipse minor axis of each aggregate connected region, and construct an aspect ratio distribution model based on the aspect ratio distribution parameters of each aggregate connected region;

[0127] Step S3034: Construct an orientation distribution model based on the orientation angles of each aggregate connected domain.

[0128] It is understood that this embodiment is based on the construction of a three-dimensional aggregate generation parameter distribution model (based on the aggregate morphology characteristics of all two-dimensional cross-sections, the "shape pattern" of the aggregate is statistically determined, and these patterns are transformed into a probability model required for the random generation of three-dimensional aggregates).

[0129] By statistically analyzing the aggregate characteristics of all slices, a random generation rule was established.

[0130] 1. Particle size distribution model construction:

[0131] For equivalent radius samples Perform kernel density estimation:

[0132]

[0133] in, Indicates the first The equivalent radius of a connected domain of aggregate (the radius of the circle after the two-dimensional cross-sectional area of ​​the aggregate is equivalent to a circle). Indicates the first The area of ​​each aggregate connected domain on a two-dimensional cross-section;

[0134] The size distribution of all stones is statistically analyzed to form the probability density function of the particle size distribution model:

[0135]

[0136] in, The probability density function represents the equivalent radius of the aggregate, and the probability density function of the equivalent radius of the aggregate is a particle size distribution model. This indicates the total number of aggregate samples. Indicates kernel density bandwidth. Represents the kernel function. The independent variable representing the equivalent radius, This indicates that the equivalent radius has been standardized for calculating the sample point's position relative to the current value. The contribution of the probability density at that location;

[0137] The aspect ratio distribution parameter is calculated based on the following formula:

[0138]

[0139] in, Indicates the first The minimum circumscribed ellipse major axis of each aggregate connected domain Indicates the first The minimum circumscribed ellipse minor axis of a connected domain of aggregates Indicates the first The aspect ratio distribution parameters of each aggregate connected domain;

[0140] Established through sample statistics Distribution model;

[0141] opposite direction angle Establish a distribution model ;

[0142] The probability density function of the orientation angle describes different orientation angles (the variable is denoted as...). The probability density of occurrence.

[0143] It is understandable that the purpose of this embodiment is to construct the following three probability models, as shown in Table 1, which is a schematic table of distribution model parameters:

[0144] Table 1. Schematic diagram of distribution model parameters

[0145]

[0146] The above probability distribution model is used to randomly generate three-dimensional aggregates, so that the generated three-dimensional stones are consistent with the statistical regularity of the actual cross-section.

[0147] Furthermore, to ensure that the generated three-dimensional aggregate voxel blocks conform to the statistical laws of real aggregates and improve the realism of the model, in one embodiment, step S304 above may include:

[0148] Step S3041: Generate an initial two-dimensional aggregate profile based on the aggregate probability distribution model;

[0149] Step S3042: Noise enhancement is performed on the initial two-dimensional aggregate profile using multi-scale Perlin noise to generate aggregate profile perturbation on the initial two-dimensional aggregate profile;

[0150] Step S3043: Stretch the initial two-dimensional aggregate profile after noise enhancement into a three-dimensional structure, and convert the three-dimensional structure into a three-dimensional aggregate voxel block.

[0151] It is understood that this embodiment applies to each Construct three-dimensional aggregates based on the parameter distribution from the previous step. .

[0152] 1. Two-dimensional contour noise enhancement:

[0153] Original 2D contour It might be too regular; common problems:

[0154] (1) The outline is composed of pixels and is stepped;

[0155] (2) The curves are not natural enough;

[0156] (3) It lacks the "convex and concave" feel of real stone in real three-dimensional space;

[0157] Therefore, this embodiment adds a perturbation to the contour points, which is generated here using multi-scale Perlin noise:

[0158]

[0159] in, Indicates the first aggregate in the connected domain The perturbation amount of each contour point Indicates the number of noise scales. Indicates the first Weights for each noise scale, Represents the Perlin noise function. Indicates the first The angle of each contour point Indicates noise frequency; Perlin noise: Perlin noise is a type of natural noise. It is widely used for generating mountains, clouds, and realistic material textures to make curves or surfaces appear "natural and irregular."

[0160] 2. Extrude the contour into a three-dimensional block:

[0161] exist The direction is determined by establishing the height and shape perturbation using a Gaussian process:

[0162]

[0163] in, Indicates control height The stretching parameters of the aggregate profile are used to control the scaling and shape changes of the stretched aggregate profile. This indicates that it follows a Gaussian process. Represents the mean function, Represents the kernel function, used to describe height. With height Similarity between them;

[0164] This embodiment introduces the Gaussian process (GP) as a model for the profile's variation with height;

[0165] in, Control at height "The scaling and shape changes of the outline";

[0166] GP makes the changes continuous, smooth, and random;

[0167] GP kernel function This determines the "similarity between different heights," for example: neighboring slices are not very different, while the difference increases with distance, and the overall pattern shows a continuous curve variation.

[0168] At each height A perturbed equivalent contour is reconstructed, and all contours are stacked to form a 3D point cloud. Voxelization is then performed (converting the point cloud into a voxel model composed of a 3D lattice; voxelization makes the 3D shape a "solid" rather than a "collection of hollow points") to obtain a 3D voxel.

[0169]

[0170] Furthermore, to ensure the continuity of aggregates with different heights in three-dimensional space and improve the quality of three-dimensional modeling, in one embodiment, step S305 may include:

[0171] Step S3051: Calculate the aggregate similarity between aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block based on the aggregate feature set;

[0172] Step S3052: Based on the aggregate similarity, select the voxel pairs to be registered from the three-dimensional aggregate voxel blocks. The voxel pairs to be registered consist of two aggregate voxel blocks with adjacent slice positions and aggregate similarity exceeding the similarity threshold.

[0173] Step S3053: Determine the registration weight of each pair of voxels to be registered based on the aggregate similarity, and optimize the registration of the voxels to be registered based on the registration weight to obtain the registration result;

[0174] Step S3054: Construct a conflict penalty term, correct the registration result based on the conflict penalty term, determine the target registration voxel pair based on the corrected registration result, fuse the target registration voxel pair, and generate a three-dimensional voxel field.

[0175] It is understood that this embodiment is used to ensure the continuity of aggregates with different height sections in three-dimensional space.

[0176] Match aggregate voxels between adjacent slices and calculate similarity based on morphological features and Fourier descriptors:

[0177]

[0178] in, This represents the first adjacent slice in a three-dimensional aggregate voxel block. The aggregate carboxyl block and the first Similarity between individual aggregate voxel blocks and They represent the first One aggregate carboxyl block and the first Two-dimensional shape characteristics of an aggregate voxel block Indicates the first The aggregate carboxyl block and the first Differences in shape characteristics between individual aggregate voxel blocks This represents a preset hyperparameter used to control the sensitivity of morphological feature differences to the similarity score. and They represent the first One aggregate carboxyl block and the first Fourier descriptor of a single aggregate voxel block Let represent the similarity function of the Fourier descriptors.

[0179] Fourier descriptor similarity is used to measure the similarity of aggregate profiles at the "texture level";

[0180] Larger aggregates are more likely to belong to the same aggregate, ensuring that highly similar aggregates are matched first;

[0181] This is used to convert differences into similarities; the higher the shape similarity, the better. The closer the value is to 1; the lower the similarity, the better. The value is close to 0.

[0182] Voxel registration optimization solves for the optimal transformation by the following optimization objective. :

[0183]

[0184] in, Represents spatial transformation Find the minimum value. Representing spatial transformation The corresponding total error function, Indicates the volume pairs to be registered Registration weights, This indicates the volume element alignment deviation term. Indicates the aggregate volume element Apply spatial transformation , Indicates the relationship with aggregate volume elements Aggregate volumetric elements in adjacent slice layers, Represents the coefficient of the regularization term. This indicates a smoothing regularization term.

[0185] It is a comprehensive evaluation of "errors, discrepancies, overlap differences, and distortions," with the aim of addressing all possible issues. Inside, find the one that... The smallest one This refers to the optimal alignment.

[0186] The meaning of the registration optimization formula: Find a spatial transformation (Translation / rotation / scaling) allows the upper-level 3D volumetric elements to... Try to align with the next layer .

[0187] Voxel crossover ratio It is used to measure whether two voxels overlap and how much they overlap, and to determine whether they are the same aggregate. The more they overlap, the closer they are to 1. That is to After the change and Similarity;

[0188] It is the weight (derived from similarity) The higher the similarity, the higher the weight given to this change.

[0189] Smoothing regularization term to prevent transformation That's too extreme. It doesn't allow for 90° rotation or large-scale translation to ensure the registration result is "reasonable." For example, it restricts translation: ;

[0190] The conflict penalty is calculated using the following formula:

[0191]

[0192] in, This represents the total error function after adding conflict penalty. Indicates the conflict penalty coefficient. Indicates the overlap tolerance threshold. Indicates aggregate volume and The intersection volume represents the volume of the overlapping portion. This represents the conflict exponential function.

[0193] The conflict penalty term is used to avoid excessive overlap of voxels.

[0194] Conflict penalty formula restriction: Stones on the same or different layers cannot overlap or intersect too much.

[0195] ,if (If the overlap exceeds the allowed limit, this item equals 1 (indicating a conflict); otherwise, this item equals 0 (indicating no problem)). It is a tolerance threshold;

[0196] Also pursuing the minimum The goal is to align the stones properly without making them intersect.

[0197] Step S40: Extract triangular meshes based on the three-dimensional voxel field, and construct a heterogeneous three-dimensional road model based on the triangular meshes and the tangent matrix.

[0198] It should be noted that a triangular mesh can be a continuous curved surface structure formed by splicing multiple triangular facets according to certain rules. It is the mainstream way to represent the shape of three-dimensional objects. It has the characteristics of simple structure, efficient rendering and controllable precision, and can accurately restore the three-dimensional contour and surface features of aggregates.

[0199] A heterogeneous 3D road model is a 3D digital model that can fully reproduce the spatial distribution characteristics and heterogeneous properties of each component (aggregate, binder, voids) of road materials, and can realistically reflect the structural differences and potential damage state inside the road.

[0200] It is understandable that this embodiment achieves accurate characterization of the three-dimensional morphology of aggregates by transforming discrete three-dimensional voxel data into a continuous and visualized triangular mesh structure; by integrating the component information of the cross-section matrix, the heterogeneous characteristics of road materials are fully restored, improving the realism and comprehensiveness of the model; the generated standard format model can be directly imported into mainstream engineering simulation software (such as finite element software and discrete element software) without secondary format conversion, which greatly improves the engineering adaptability and application efficiency of the model.

[0201] Furthermore, in order to improve the integrity and aesthetics of aggregate morphology representation and avoid model distortion caused by mesh defects, step S40 above may include:

[0202] Step S401: Binarize the three-dimensional voxel field;

[0203] Step S402: Extract isosurfaces from the binarized 3D voxel field to obtain a triangular mesh;

[0204] Step S403: Integrate the triangular mesh to obtain a continuous three-dimensional aggregate skin;

[0205] Step S404: Extract the two-dimensional distribution information of binder, aggregate and void from the section matrix, and combine it with the spatial position information of the slice matrix to map the two-dimensional distribution information to three-dimensional space to obtain the three-dimensional distribution area of ​​binder, aggregate and void;

[0206] Step S405: Based on the three-dimensional aggregate skin, structural filling is performed on the three-dimensional distribution areas of the binder, aggregate and voids respectively to construct a heterogeneous three-dimensional road model.

[0207] In the specific implementation, all registered voxels are fused (the registered and fused voxel field (occupying the voxel set) is converted into a continuous, usable 3D mesh (e.g., STL / OBJ), and necessary post-processing is performed to meet the input requirements of visualization or engineering simulation (FEM / DEM):

[0208]

[0209] The final complete three-dimensional road aggregate voxel set;

[0210] : in the The first slice obtained from the slice Three-dimensional voxel blocks of block aggregate;

[0211] (That is, to cut different slices) All aggregates (3D voxel blocks) generated differently All of them are combined to obtain a complete three-dimensional voxel field.

[0212] The MarchingCubes algorithm is used, with an occupancy probability threshold (examining each 2×2×2 eight voxels, determining which positions are "inside / outside", and piecing together a triangle at the interface; all triangles together form the 3D skin of the aggregate):

[0213]

[0214] Occupancy probability First, decide which voxels are considered "solid," such as:

[0215]

[0216] After voxels are binarized, the MarchingCubes algorithm is used to extract isosurfaces from the occupied voxel field to obtain triangular meshes, generating the final 3D mesh model (STL, OBJ format, etc.).

[0217] Understandably, this embodiment achieves spatial correlation between heterogeneous elements and damage features among different slices by designing an adaptive matching algorithm for slice features. Simultaneously, it introduces a lightweight heterogeneous material distribution and damage state interpolation module to accurately reconstruct the irregular morphology of aggregates, heterogeneous component distribution, and damage features such as internal microcracks and voids without excessively increasing the computational burden. This scheme focuses on optimizing the balance between the efficiency of 2D slice data utilization, the accuracy of 3D heterogeneous structure reconstruction, and the ability to characterize damage features, ensuring a comprehensive improvement in modeling efficiency, model realism, damage characterization accuracy, and engineering adaptability in 3D road model construction tasks.

[0218] In the specific implementation, the two-dimensional distribution information of cementitious material (value 2) and voids (value 0) is extracted from the cross-section matrix. Combined with the spatial location information of the slices, the two-dimensional component distribution information is mapped to three-dimensional space to obtain the three-dimensional distribution areas of cementitious material and voids (the inside of the aggregate skin is the aggregate area, the outside of the skin and the non-void area is the cementitious material area, and the remaining area is the void area). In three-dimensional modeling software (such as Blender, ANSYS), based on the three-dimensional aggregate skin, the aggregate, cementitious material, and void areas are digitally filled (assigned different material attribute labels, such as red for aggregate, blue for cementitious material, and transparent for voids). The model is then mesh optimized (such as refining the mesh of key areas and simplifying the mesh of non-key areas) to ensure a balance between model accuracy and computational efficiency. Finally, a complete heterogeneous three-dimensional road model is generated.

[0219] Technical Effects: 1. The generated 3D aggregate model is highly consistent with the real road aggregate in terms of shape and distribution, significantly improving realism; 2. The registration continuity of aggregates in adjacent cross-sections is strong, with no breakage or misalignment issues; 3. The aggregate contours possess natural irregularity and high fidelity; 4. The model has no voxel overlap conflicts, resulting in high geometric accuracy; 5. The output STL / OBJ format mesh model can be directly used for engineering simulations such as FEM / DEM, demonstrating strong adaptability. Refer to Tables 2 and 3 below. Table 2 compares the modeling accuracy of the non-homogeneous 3D road modeling method based on 2D slice data in this embodiment with other modeling methods. Table 3 compares the modeling efficiency of this embodiment with other modeling methods.

[0220] Table 2. Comparison of Modeling Accuracy Data

[0221]

[0222] Table 3. Comparison of Modeling Efficiency Data

[0223]

[0224] This embodiment constructs a two-dimensional structure of road materials to generate standardized two-dimensional cross-sectional images. The two-dimensional structure includes aggregates, binders, and voids. Based on visual feature classification and assignment, the two-dimensional cross-sectional images are converted into a cross-sectional matrix. Each unit in the cross-sectional matrix represents the material category of a pixel. An aggregate feature set is extracted based on the cross-sectional matrix, and an aggregate probability distribution model is generated based on this feature set. A three-dimensional voxel field is then generated based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of the aggregates and Fourier descriptors. The three-dimensional voxel field is formed by registering and fusing three-dimensional aggregate voxel blocks from multiple adjacent slices. The three-dimensional aggregate voxel block is generated by stretching an initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of aggregate pixels extracted from the sectional matrix. Triangular meshes are extracted based on the three-dimensional voxel field, and a heterogeneous three-dimensional road model is constructed based on the triangular meshes and the sectional matrix. This breaks through the limitations of existing single empirical parameters or single sectional surfaces, optimizes the balance between the utilization efficiency of two-dimensional slice data, the accuracy of three-dimensional heterogeneous structure restoration, and the ability to characterize damage features, and ensures that in the three-dimensional road model construction task, a comprehensive improvement is achieved in modeling efficiency, model realism, damage characterization accuracy, and engineering adaptability.

[0225] Furthermore, embodiments of the present invention also propose a computer-readable storage medium storing a heterogeneous three-dimensional road modeling program based on two-dimensional slice data. When the heterogeneous three-dimensional road modeling program based on two-dimensional slice data is executed by a processor, it implements the steps of the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described above.

[0226] The computer-readable storage medium provided in this application may be, for example, a USB flash drive, but is not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this embodiment, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, system, or device. The program code contained on the computer-readable storage medium may be transmitted using any suitable medium, including but not limited to: wires, optical cables, RF (Radio Frequency), etc., or any suitable combination thereof.

[0227] The aforementioned computer-readable storage medium may be included in a heterogeneous 3D road modeling device based on 2D slice data; or it may exist independently and not assembled into a heterogeneous 3D road modeling device based on 2D slice data.

[0228] Furthermore, this invention also proposes a computer program product, including a heterogeneous three-dimensional road modeling program based on two-dimensional slice data. When the heterogeneous three-dimensional road modeling program based on two-dimensional slice data is executed by a processor, it implements the steps of the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described above.

[0229] The specific implementation of the computer program product of the present invention is basically the same as the embodiments of the above-described method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data, and will not be repeated here.

[0230] Reference Figure 4 , Figure 4 This is a structural block diagram of the first embodiment of the heterogeneous three-dimensional road modeling device based on two-dimensional slice data of the present invention.

[0231] like Figure 4 As shown, the heterogeneous 3D road modeling device based on 2D slice data proposed in this embodiment of the invention includes:

[0232] The two-dimensional slice data generation module 10 is used to construct the two-dimensional structure of road materials and generate standardized two-dimensional cross-sectional images. The two-dimensional structure of road materials includes aggregates, binders, and voids.

[0233] Matrix transformation module 20 is used to classify and assign values ​​based on material visual features, and convert the two-dimensional cross-section image into a cross-section matrix, wherein each unit in the cross-section matrix represents the material category to which the pixel belongs;

[0234] The feature extraction module 30 is used to extract aggregate feature set based on the cross section matrix, generate aggregate probability distribution model based on the aggregate feature set, and generate three-dimensional voxel field based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of aggregate and Fourier descriptors. The three-dimensional voxel field is generated by registering and fusing three-dimensional aggregate voxel blocks in multiple adjacent slices. The three-dimensional aggregate voxel blocks are obtained by stretching the initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of aggregate pixels extracted from the cross section matrix.

[0235] The three-dimensional road model construction module 40 is used to extract triangular meshes based on the three-dimensional voxel field and construct a heterogeneous three-dimensional road model based on the triangular meshes and the tangent matrix.

[0236] This embodiment constructs a two-dimensional structure of road materials to generate standardized two-dimensional cross-sectional images. The two-dimensional structure includes aggregates, binders, and voids. Based on visual feature classification and assignment, the two-dimensional cross-sectional images are converted into a cross-sectional matrix. Each unit in the cross-sectional matrix represents the material category of a pixel. An aggregate feature set is extracted based on the cross-sectional matrix, and an aggregate probability distribution model is generated based on this feature set. A three-dimensional voxel field is then generated based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of the aggregates and Fourier descriptors. The three-dimensional voxel field is formed by registering and fusing three-dimensional aggregate voxel blocks from multiple adjacent slices. The three-dimensional aggregate voxel block is generated by stretching an initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of aggregate pixels extracted from the sectional matrix. Triangular meshes are extracted based on the three-dimensional voxel field, and a heterogeneous three-dimensional road model is constructed based on the triangular meshes and the sectional matrix. This breaks through the limitations of existing single empirical parameters or single sectional surfaces, optimizes the balance between the utilization efficiency of two-dimensional slice data, the accuracy of three-dimensional heterogeneous structure restoration, and the ability to characterize damage features, and ensures that in the three-dimensional road model construction task, a comprehensive improvement is achieved in modeling efficiency, model realism, damage characterization accuracy, and engineering adaptability.

[0237] The heterogeneous 3D road modeling device based on 2D slice data provided in this application, employing the heterogeneous 3D road modeling method based on 2D slice data in the above embodiments, can solve the technical problem of heterogeneous 3D road modeling based on 2D slice data. Compared with the prior art, the beneficial effects of the heterogeneous 3D road modeling device based on 2D slice data provided in this application are the same as the beneficial effects of the heterogeneous 3D road modeling method based on 2D slice data provided in the above embodiments, and other technical features in the heterogeneous 3D road modeling device based on 2D slice data are the same as the features disclosed in the methods of the above embodiments, and will not be repeated here.

[0238] It should be understood that the above are merely illustrative examples and do not constitute any limitation on the technical solutions of the present invention. In specific applications, those skilled in the art can make settings as needed, and the present invention does not impose any restrictions on this.

[0239] It should be noted that the workflow described above is merely illustrative and does not limit the scope of protection of this invention. In practical applications, those skilled in the art can select some or all of the workflow to achieve the purpose of this embodiment according to actual needs, and no restrictions are imposed here.

[0240] In addition, for technical details not described in detail in this embodiment, please refer to the non-homogeneous three-dimensional road modeling method based on two-dimensional slice data provided in any embodiment of the present invention, which will not be repeated here.

[0241] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or system that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or system. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or system that includes that element.

[0242] The sequence numbers of the above embodiments of the present invention are for descriptive purposes only and do not represent the superiority or inferiority of the embodiments.

[0243] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as read-only memory / random access memory, magnetic disk, optical disk) and includes several instructions to cause a terminal device (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0244] The above are merely preferred embodiments of the present invention and do not limit the scope of the patent. Any equivalent structural or procedural transformations made based on the description and drawings of the present invention, or direct or indirect applications in other related technical fields, are similarly included within the scope of patent protection of the present invention.

Claims

1. A method for modeling heterogeneous 3D roads based on 2D slice data, characterized in that, The heterogeneous 3D road modeling method based on 2D slice data includes: A two-dimensional structure of road materials is constructed, and a standardized two-dimensional cross-sectional view is generated. The two-dimensional structure of road materials includes aggregates, binders, and voids. Based on the material visual feature classification assignment, the two-dimensional cross-section image is converted into a cross-section matrix, where each cell in the cross-section matrix represents the material category to which the pixel belongs. Based on the cross section matrix, an aggregate feature set is extracted, and an aggregate probability distribution model is generated based on the aggregate feature set. A three-dimensional voxel field is generated based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of the aggregate and Fourier descriptors. The three-dimensional voxel field is generated by registering and fusing three-dimensional aggregate voxel blocks in multiple adjacent slices. The three-dimensional aggregate voxel blocks are obtained by transforming the initial two-dimensional aggregate contour, which is generated by noise enhancement and aggregate contour perturbation, into a three-dimensional structure after contour stretching. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of the aggregate pixels extracted from the cross section matrix. Triangular meshes are extracted based on the three-dimensional voxel field, and a heterogeneous three-dimensional road model is constructed based on the triangular meshes and the sectional matrix.

2. The method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data as described in claim 1, characterized in that, The step of extracting aggregate feature sets based on the cross-section matrix, generating an aggregate probability distribution model based on the aggregate feature sets, and generating a three-dimensional voxel field based on the aggregate probability distribution model includes: The adjacent different aggregate pixels in the sectional matrix are divided to obtain multiple initial aggregate regions, and noise is removed from the initial aggregate regions to obtain aggregate connected regions, which represent a two-dimensional aggregate cross section. Two-dimensional morphological features and Fourier descriptors of each aggregate connected component are extracted, and an aggregate feature set is constructed. The Fourier descriptors are extracted based on the following formula: in, Indicates the first Each aggregate connectivity domain The Fourier descriptor, This represents the index of the sampling point on the aggregate profile. This represents the total number of sampling points on the aggregate profile. Indicates the first On the outline of the aggregate Complex coordinates of each sampling point This indicates the order of the Fourier descriptor. The basis functions represent the discrete Fourier transform; Based on the aggregate feature set of each aggregate connected domain, shape regularity analysis is performed to generate an aggregate probability distribution model; Three-dimensional aggregate voxel blocks are generated based on the aggregate probability distribution model. Based on the aggregate feature set, the aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block are registered and fused to generate a three-dimensional voxel field.

3. The method for modeling heterogeneous 3D roads based on 2D slice data as described in claim 2, characterized in that, The aggregate probability distribution model includes a particle size distribution model, an aspect ratio distribution model, and a direction distribution model; the step of generating the aggregate probability distribution model by analyzing the shape regularity of the aggregate feature set of each aggregate connected domain includes: Kernel density analysis is performed on each aggregate connected region based on the aggregate feature set to obtain the equivalent radius. The two-dimensional morphological features in the aggregate feature set include the area, perimeter, minor axis, and orientation angle of the aggregate connected region. The equivalent radius is calculated with reference to the following formula: in, Indicates the first The equivalent radius of a connected domain of aggregates Indicates the first The area of ​​each aggregate connected domain on a two-dimensional cross-section; A particle size distribution model is constructed based on the equivalent radius of each aggregate connected domain, referring to the following formula: in, The probability density function represents the equivalent radius of the aggregate, and the probability density function of the equivalent radius of the aggregate is a particle size distribution model. This indicates the total number of aggregate samples. Indicates kernel density bandwidth. Represents the kernel function. The independent variable representing the equivalent radius, This indicates that the equivalent radius has been standardized for calculating the sample point's position relative to the current value. The contribution of the probability density at that location; The aspect ratio distribution parameters of each aggregate connected region are determined based on the major axis and minor axis of the least circumscribed ellipse of each aggregate connected region, and an aspect ratio distribution model is constructed based on the aspect ratio distribution parameters of each aggregate connected region. The aspect ratio distribution parameters are calculated based on the following formula: in, Indicates the first The minimum circumscribed ellipse major axis of each aggregate connected domain Indicates the first The minimum circumscribed ellipse minor axis of a connected domain of aggregates Indicates the first The aspect ratio distribution parameters of each aggregate connected domain; A directional distribution model is constructed based on the directional angles of each aggregate's connected domains.

4. The method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data as described in claim 3, characterized in that, The generation of three-dimensional aggregate voxel blocks based on the aggregate probability distribution model includes: An initial two-dimensional aggregate profile is generated based on the aggregate probability distribution model. The initial two-dimensional aggregate profile is noise-enhanced using multi-scale Perlin noise to generate aggregate profile perturbation, as shown in the following formula: in, Indicates the first aggregate in the connected domain The disturbance amount of each contour point Indicates the number of noise scales. Indicates the first Weights for each noise scale, Represents the Perlin noise function. Indicates the first The angle of each contour point Indicates the noise frequency; The initial two-dimensional aggregate profile after noise enhancement is stretched into a three-dimensional structure, and the three-dimensional structure is converted into a three-dimensional aggregate voxel block. The profile stretching is performed according to the following formula: in, Indicates control height The stretching parameters of the aggregate profile are used to control the scaling and shape changes of the stretched aggregate profile. This indicates that it follows a Gaussian process. Represents the mean function, Represents the kernel function, used to describe height. With height The similarity between them.

5. The method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data as described in claim 4, characterized in that, The process of registering and fusing aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block based on the aggregate feature set to generate a three-dimensional voxel field includes: Based on the aggregate feature set, the aggregate similarity between aggregate voxel blocks in adjacent slices of the three-dimensional aggregate voxel block is calculated, and the aggregate similarity is calculated with reference to the following formula: in, This represents the first adjacent slice in a three-dimensional aggregate voxel block. The aggregate carboxyl block and the first Similarity between individual aggregate voxel blocks and They represent the first One aggregate carboxyl block and the first Two-dimensional shape characteristics of an aggregate voxel block Indicates the first The aggregate carboxyl block and the first Differences in shape characteristics between individual aggregate voxel blocks This represents a preset hyperparameter used to control the sensitivity of morphological feature differences to the similarity score. and They represent the first One aggregate carboxyl block and the first Fourier descriptor of a single aggregate voxel block The similarity function representing Fourier descriptors; Based on the aggregate similarity, registerable voxel pairs are selected from the three-dimensional aggregate voxel blocks. The registerable voxel pairs consist of two aggregate voxel blocks whose slice positions are adjacent and whose aggregate similarity exceeds the similarity threshold. Based on the aggregate similarity, the registration weights of each pair of voxels to be registered are determined, and the registration of the voxels to be registered is optimized based on the registration weights to obtain the registration result. The registration optimization refers to the following formula: in, Represents spatial transformation Find the minimum value. Representing spatial transformation The corresponding total error function, Indicates the volume pairs to be registered Registration weights, This indicates the volume element alignment deviation term. Indicates the aggregate volume element Apply spatial transformation , Indicates the relationship with aggregate volume elements Aggregate volumetric elements in adjacent slice layers, Represents the coefficient of the regularization term. Indicates a smoothing regularization term; A conflict penalty term is constructed, and the registration result is corrected based on the conflict penalty term. Target registration voxel pairs are then determined based on the corrected registration results, and these target voxel pairs are fused to generate a three-dimensional voxel field. The conflict penalty term is defined by the following formula: in, This represents the total error function after adding conflict penalty. Indicates the conflict penalty coefficient. Indicates the overlap tolerance threshold. Indicates aggregate volume and The intersection volume represents the volume of the overlapping portion. This represents the conflict exponential function.

6. The method for modeling heterogeneous three-dimensional roads based on two-dimensional slice data as described in any one of claims 1 to 5, characterized in that, The step of extracting triangular meshes based on the three-dimensional voxel field and constructing a heterogeneous three-dimensional road model based on the triangular meshes and the tangent matrix includes: The three-dimensional voxel field is binarized; Isosurfaces are extracted from the binarized 3D voxel field to obtain a triangular mesh; The triangular meshes are integrated to obtain a continuous three-dimensional aggregate skin; Two-dimensional distribution information of binder, aggregate and void is extracted from the cross-section matrix. Combined with the spatial position information of the slice matrix, the two-dimensional distribution information is mapped to three-dimensional space to obtain the three-dimensional distribution area of ​​binder, aggregate and void. Based on the three-dimensional aggregate skin, the three-dimensional distribution areas of binder, aggregate and voids are structurally filled to construct a heterogeneous three-dimensional road model.

7. A non-homogeneous three-dimensional road modeling device based on two-dimensional slice data, characterized in that, The heterogeneous 3D road modeling device based on 2D slice data includes: A two-dimensional slice data generation module is used to construct the two-dimensional structure of road materials and generate standardized two-dimensional cross-sectional images. The two-dimensional structure of road materials includes aggregates, binders, and voids. The matrix transformation module is used to classify and assign values ​​based on material visual features, and to convert the two-dimensional cross-section image into a cross-section matrix, where each unit in the cross-section matrix represents the material category to which the pixel belongs; The feature extraction module is used to extract aggregate feature set based on the cross section matrix, generate aggregate probability distribution model based on the aggregate feature set, and generate three-dimensional voxel field based on the aggregate probability distribution model. The aggregate feature set includes two-dimensional morphological features of aggregate and Fourier descriptors. The three-dimensional voxel field is generated by registering and fusing three-dimensional aggregate voxel blocks in multiple adjacent slices. The three-dimensional aggregate voxel blocks are obtained by stretching the initial two-dimensional aggregate contour. The initial two-dimensional aggregate contour is generated based on the boundary segmentation of aggregate pixels extracted from the cross section matrix. A three-dimensional road model construction module is used to extract triangular meshes based on the three-dimensional voxel field, and to construct a heterogeneous three-dimensional road model based on the triangular meshes and the tangent matrix.

8. A non-homogeneous three-dimensional road modeling device based on two-dimensional slice data, characterized in that, The heterogeneous 3D road modeling device based on 2D slice data includes: a memory, a processor, and a heterogeneous 3D road modeling program based on 2D slice data stored in the memory. The processor is used to run the heterogeneous 3D road modeling program based on 2D slice data, and the heterogeneous 3D road modeling program based on 2D slice data is configured to implement the heterogeneous 3D road modeling method based on 2D slice data as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a heterogeneous three-dimensional road modeling program based on two-dimensional slice data. When the heterogeneous three-dimensional road modeling program based on two-dimensional slice data is executed by a processor, it implements the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described in any one of claims 1 to 6.

10. A computer program product, characterized in that, The computer program product includes a heterogeneous three-dimensional road modeling program based on two-dimensional slice data, which, when executed by a processor, implements the steps of the heterogeneous three-dimensional road modeling method based on two-dimensional slice data as described in any one of claims 1 to 6.

Images