A multi-source point cloud feature cascade optimization ancient building reverse modeling method and system

Abstract

This invention relates to the field of surveying and mapping technology, and discloses a method and system for reverse modeling of ancient buildings based on multi-source point cloud feature cascade optimization. The key technical points are: acquisition of multi-source heterogeneous point cloud data and adaptive feature cascade extraction; high-precision registration of multi-source point clouds based on cascade feature cross-optimization; quantitative extraction of microscopic and macroscopic damage to ancient buildings based on cascade features; deformation-perception reverse modeling and supplementary volumetric micro-element quantization integration; and quantitative evaluation decision-making and conversion of physical entity repair instructions. The method employs local tensor calculation to extract underlying physical features such as surface roughness for full-process cascade coupling. Through adaptive manifold downsampling, cross-constraint weight registration, and spatial micro-element Riemann integral technology, it achieves high-precision seamless fusion of multi-source heterogeneous surveying and mapping data and automated quantitative diagnosis of damage, ultimately converting it into physical entity repair instructions that directly guide CNC machining and on-site construction.

Description

Technical Field

[0001] This invention relates to the field of surveying and mapping technology, and more specifically, to a method and system for reverse modeling ancient buildings using multi-source point cloud feature cascade optimization. Background Technology

[0002] Traditional surveying methods, as well as single-source 3D laser scanning or UAV photogrammetry, often suffer from limitations in perspective and incomplete data coverage. While multi-source point cloud fusion is used to obtain complete data, existing registration algorithms (such as traditional ICP) are prone to getting stuck in local optima due to their reliance solely on spatial distance measurements when dealing with complex overlapping areas like eaves and brackets in ancient architecture. This leads to severe misalignment and clipping issues. Furthermore, existing reverse modeling and damage assessment processes generally suffer from "information silos." Point cloud downsampling, data registration, and subsequent defect extraction (such as crack detection and skew settlement analysis) are often disconnected, failing to effectively reuse underlying physical geometric features. Moreover, the estimation of supplementary volume and deformation deviation analysis of physical components heavily relies on subjective human experience, lacking rigorous micro-element quantization integration and mathematical cascade support. The generated models are often "idealized" models that mask actual deformation, making them difficult to directly translate into precise physical repair instructions for on-site construction or CNC machining.

[0003] Therefore, this invention provides a method and system for reverse modeling of ancient buildings using multi-source point cloud feature cascade optimization, which improves the above-mentioned technical problems. Summary of the Invention

[0004] This invention aims to address the shortcomings of existing technologies by providing a method and system for reverse modeling ancient buildings using multi-source point cloud feature cascade optimization. The invention employs local tensor calculation to extract underlying physical features such as surface roughness for full-process cascade coupling. Through adaptive manifold downsampling, cross-constraint weight registration, and spatial infinitesimal Riemann integral technology, it achieves high-precision seamless fusion of multi-source heterogeneous surveying and mapping data and automated quantitative diagnosis of damage and defects. Ultimately, it transforms these data into physical entity repair instructions that directly guide CNC machining and on-site construction.

[0005] To achieve the above objectives, the present disclosure proposes the following technical solutions:

[0006] In a first aspect, this disclosure proposes a method for reverse modeling ancient buildings using multi-source point cloud feature cascade optimization, comprising the following steps:

[0007] S1. Obtain multi-source heterogeneous point cloud data of ancient buildings, construct a local covariance tensor for each point in the point cloud data to solve the surface roughness factor and normal vector; perform feature-guided adaptive manifold downsampling based on the surface roughness factor, output the downsampled source point cloud and target point cloud, and record the side length of the retained local spatial micro-element voxels.

[0008] S2. When searching for matching point pairs between the downsampled source point cloud and the target point cloud, the surface roughness factor and normal vector are introduced into the cross physical constraint weight function to calculate the matching comprehensive weight, construct the weighted target error function and solve the optimal transformation matrix to achieve registration between the downsampled source point cloud and the target point cloud, and output the fused point cloud.

[0009] S3. Geometric fitting of load-bearing wooden columns of ancient buildings based on fused point cloud to extract macroscopic deformation parameters; and retrieve the surface roughness factor corresponding to each point, extract the set of points with excessive roughness and perform topological clustering to obtain the set of microscopic defects.

[0010] S4. Spatial comparison is performed between the preset ideal BIM theoretical model of ancient buildings and the fused point cloud. The physical deviation of normal deformation of each point relative to the ideal surface along the normal direction is calculated. The side length of local spatial micro-element voxels is geometrically coupled with the physical deviation of normal deformation. The volume of solid repair and replacement is calculated by using the discrete volume integral equation for the set of micro-defect points.

[0011] S5. Based on the comprehensive macroscopic deformation parameters, microscopic defect point set, and physical repair and replenishment volume, calculate the structural repair priority index; and based on the structural repair priority index and the reverse mesh boundary of the microscopic defect point set, generate three-dimensional cutting instructions to drive CNC machining equipment to perform physical cutting on the replacement wood, or generate physical support parameters to guide the on-site scaffold support correction.

[0012] As a preferred embodiment of the present invention, S1, which involves calculating the surface roughness factor and the normal vector, and performing adaptive manifold downsampling guided by the surface roughness factor, specifically includes:

[0013] Select any point in the point cloud data Search its spatial neighborhood Construct a local covariance matrix using the nearest neighbor points:

[0014] ;

[0015] For the local covariance matrix Eigenvalue decomposition yields three non-negative eigenvalues. And extract the minimum eigenvalue. The corresponding unit eigenvector is used as the normal vector. ;

[0016] Using formula Calculate surface roughness factor ;

[0017] Using formula Solving the side length of local spatial infinitesimal elements ;

[0018] In the formula, For this The three-dimensional centroid coordinate vector of each neighboring point This is the transpose of the coordinate deviation vector; This is the preset limit physical resolution size. To smooth the amplification factor, This is the roughness sensitivity adjustment factor.

[0019] As a preferred embodiment of the present invention, in S2, the surface roughness factor and the normal vector are introduced into the cross-physical constraint weight function to calculate the matching comprehensive weight, construct the weighted objective error function, and solve the optimal transformation matrix, specifically including:

[0020] Calculate any pair of matching points Matching comprehensive weight :

[0021] ;

[0022] Match the overall weight Substitute into the objective function of minimizing the weighted point-to-surface error:

[0023] ;

[0024] In the formula, Source Point Cloud The coordinates of the i-th point in the array. and Here are its corresponding roughness factor and normal vector; For target point cloud The coordinates of its corresponding point and Here are its corresponding roughness factor and normal vector; The square of the Euclidean distance between the two points; This is the average distance tolerance threshold for the current iteration step; This is a preset tolerance for the included normal angle; This is the roughness difference penalty factor; The weighted total error of the registration process; This represents the total number of valid point pairs that have been successfully matched. Let be a rotation matrix. Let it be a translation vector; solve iteratively to make The minimum R and t are taken as the optimal transformation matrix.

[0025] As a preferred embodiment of the present invention, in step S3, the extraction of macroscopic deformation parameters specifically includes:

[0026] Spatial geometric fitting is performed on the load-bearing wooden columns in the fused point cloud to extract the actual centerline vector. and the three-dimensional coordinates of the column base center ;

[0027] Using formula Calculate the tilt angle of ancient buildings In the formula, The absolute gravity is the vertically downward reference vector;

[0028] Least square fitting is performed on the three-dimensional coordinates of the center of the base of multiple columns on the same elevation ring to obtain the actual foundation settlement reference surface, and the elevation settlement difference of each column base relative to the actual foundation settlement reference surface is calculated.

[0029] As a preferred embodiment of the present invention, in step S3, obtaining the set of microscopic disease points specifically includes:

[0030] Traverse and merge point clouds to extract the required parameters. The characteristic points are used as suspected disease points, among which The surface roughness factor of the feature point. A roughness threshold preset based on healthy component samples;

[0031] A region growing algorithm based on Euclidean distance is used to perform topological clustering on suspected disease points, and continuous sets of surface peeling or crack points are extracted as micro-disease point sets. .

[0032] As a preferred embodiment of the present invention, in S4, the physical deviation of the normal deformation of each point relative to the ideal surface along the normal direction is calculated, and the volume of physical repair and replacement is calculated by solving the discrete volume integral equation for the set of microscopic defects, specifically including:

[0033] Using formula Solving for any point in the fused point cloud Physical deviation of normal deformation ;

[0034] Using the discrete volume integral equation: Solve the volume of solid repair and replacement ;

[0035] In the formula, The three-dimensional coordinates of the center of the base of the cylinder are: This refers to the unit vector of the vertical centerline specified in the ideal BIM theoretical model of ancient buildings. To create the ideal radius specified in the standard; This belongs to the microscopic disease point set The point in the middle, Let its normal absolute concavity depth be denoted as ; let the side length of the local spatial voxel preserved at each point in the downsampled source point cloud output in S1 be . , The angle between the actual normal vector at that point and the normal vector of the ideal surface.

[0036] As a preferred technical solution of the present invention, in S5, the structural repair priority index is specifically calculated using the following formula:

[0037] ;

[0038] In the formula, This serves as a priority index for structural repair. To standardize the permissible safety limit for tilt angle; For the maximum flexural deviation, To standardize the maximum allowable deformation limit; To compensate for the volume, This refers to the complete total volume recorded in the component's historical archives; The risk weight coefficient is preset for the system.

[0039] Secondly, this disclosure proposes a multi-source point cloud fusion and ancient building reverse modeling system based on feature cascade optimization. The system includes: a data acquisition and cascade feature extraction module, a feature cross-optimization registration module, a macro-micro disease extraction module, a reverse modeling and supplementary volume quantization module, and a physical entity repair execution module.

[0040] The data acquisition and cascaded feature extraction module is used to acquire multi-source heterogeneous point cloud data of ancient buildings, calculate the surface roughness factor and normal vector, perform adaptive manifold downsampling based on the surface roughness factor, and record the side length of local spatial micro-element voxels.

[0041] The feature cross-optimized registration module is used to solve the optimal transformation matrix by introducing the surface roughness factor and normal vector into the cross physical constraint weight function when searching for matching point pairs between the downsampled source point cloud and the target point cloud, and output the fused point cloud.

[0042] The macro-micro defect extraction module is used to extract macro deformation parameters of ancient buildings by geometric fitting based on fused point cloud, and to extract the point set with excessive roughness by topological clustering to obtain the point set of micro defects.

[0043] The reverse modeling and volume quantification module is used to compare the physical deviation of normal deformation caused by the disease with the preset ideal BIM theoretical model of ancient buildings, and to geometrically couple the side length of local spatial micro-element voxels and the physical deviation of normal deformation through discrete volume integral equations to calculate the volume of solid repair and replacement.

[0044] The physical entity repair execution module is used to calculate the structural repair priority index by comprehensively considering the above parameters, and generate three-dimensional cutting instructions to drive CNC machining equipment to perform physical cutting or generate physical support parameters to guide on-site support correction based on the calculation results.

[0045] Thirdly, this disclosure proposes a computer device, including a memory and a processor. The memory stores a computer program, and the processor executes the computer program to implement a method for reverse modeling ancient buildings by multi-source point cloud feature cascade optimization.

[0046] Fourthly, this disclosure proposes a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements a method for reverse modeling ancient buildings by multi-source point cloud feature cascade optimization.

[0047] In summary, the present invention has the following beneficial effects:

[0048] Firstly, this invention extracts the "surface roughness" and "normal deviation" of the local tensor of the point cloud as the underlying physical genes throughout the entire process, and constructs a cross-physical constraint weight function based on this. When processing the fusion of UAV and ground point clouds, this mechanism can effectively attenuate the matching weights of discrete noise points such as the eaves of ancient buildings and weeds, avoiding the "misalignment" and "pattern crossing" traps caused by traditional registration algorithms (such as ICP) that rely solely on spatial distance metrics and are prone to local optima, thus achieving high-precision seamless stitching of cross-source heterogeneous data.

[0049] Secondly, this invention employs a feature-guided adaptive manifold downsampling strategy to dynamically allocate the edge lengths of local spatial micro-element voxels. This allows voxels in smooth wall surfaces or foundation areas to be automatically enlarged to significantly compress data redundancy, while voxels in crack edges and complex bracket carving areas approach the limit of physical resolution. This design completely overcomes the technical bias of traditional uniform downsampling, which tends to "smooth out" microscopic damage features, greatly improving the accuracy of subsequent defect extraction.

[0050] Third, this invention transforms the "adaptive voxel side length" of the front-end downsampling into the micro-area of ​​the local tangent plane, and geometrically couples it with the "normal deformation depth" of the back-end, constructing a discrete volume equation based on the Riemann integral of spatial micro-elements. This multi-step parameter cascade closed loop achieves extremely rigorous and accurate quantitative calculation of the missing volume of wood and the degree of weathering of cracks, providing scientific mathematical and physical support for the restoration of the original appearance of ancient buildings. Attached Figure Description

[0051] Figure 1 The flowchart illustrates a method for reverse modeling ancient buildings using multi-source point cloud feature cascade optimization, as provided in this embodiment of the invention. Detailed Implementation

[0052] The present application will now be described in detail with reference to specific embodiments. These embodiments will help those skilled in the art to further understand the present application, but do not limit the present application in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present application. These all fall within the protection scope of the present application.

[0053] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.

[0054] Unless otherwise defined, all technical and scientific terms used in this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in this specification is for the purpose of describing particular embodiments only and is not intended to limit the scope of this application. The term "and / or" as used in this specification includes any and all combinations of one or more of the associated listed items.

[0055] Furthermore, the technical features involved in the various embodiments of this application described below can be combined with each other as long as they do not conflict with each other.

[0056] This disclosure aims to address the problems in existing ancient building surveying and restoration, such as the ease of misalignment and overlap in multi-source heterogeneous point cloud registration, reliance on subjective human judgment for damage extraction, and the lack of quantitative deformation closure analysis in inverse models. Therefore, this disclosure proposes a method and system for inverse modeling of ancient buildings using multi-source point cloud feature cascade optimization, for high-precision evaluation of the actual deformation and health status of ancient building structures. This method employs computational techniques such as local tensor solving, adaptive manifold downsampling, and spatial infinitesimal Riemann integrals. By extracting and cascading the underlying physical genes (such as surface roughness and normal deviation) throughout the entire process, it addresses the technical bottlenecks of local optimum traps in cross-source heterogeneous data registration and the difficulty in accurately estimating the volume of entity replacement. This achieves seamless fusion of multi-source surveying data, fully automated quantitative diagnosis of damage, and direct guidance for the generation of physical restoration instructions.

[0057] Please refer to Figure 1 , Figure 1 The flowchart illustrates a method for reverse modeling ancient buildings using multi-source point cloud feature cascade optimization, as described in an embodiment of this disclosure. The overall process mainly includes the following five steps:

[0058] S1. Acquisition of multi-source heterogeneous point cloud data and adaptive feature cascade extraction.

[0059] S1.1 Multi-source data synchronous acquisition and preprocessing: High-density point cloud of ancient buildings is obtained using a terrestrial 3D laser scanner. Macroscopic point clouds are acquired using drones equipped with lidar and inertial measurement units (IMUs). The Statistical Outlier Removal (SOR) algorithm is used to remove noise.

[0060] S1.2 Construction of Local Covariance Tensor: For any point in the merged point cloud set Using KD-trees to search its spatial neighborhood Construct a local covariance matrix using the nearest neighbor points. :

[0061]

[0062] In the formula: The number of neighboring points (preferably set to 20); For the neighboring region Three-dimensional spatial coordinate vectors of points; For this The three-dimensional spatial centroid coordinate vector of each neighboring point; This is the transpose of the coordinate deviation vector.

[0063] S1.3, Cascaded physical characteristic parameter calculation: For Eigenvalue decomposition yields three non-negative eigenvalues. ; Calculate the surface roughness factor :

[0064]

[0065] In the formula: For this point The surface roughness factor at that location ranges from [0, 1 / 3]. These are the minimum, middle, and maximum eigenvalues ​​of the covariance matrix, respectively.

[0066] At the same time, extract the minimum eigenvalue. The corresponding unit eigenvector is used as the normal vector of that point. .

[0067] S1.4 Feature-guided adaptive manifold downsampling: based on roughness factor Calculate the point Adaptive voxel side length of the region :

[0068]

[0069] In the formula: This refers to the side length of the local spatial voxel retained after downsampling; The set limit physical resolution (preferably 2mm); The smoothing amplification factor is preferably 15. It is a roughness sensitivity adjustment factor (preferably 5).

[0070] according to After downsampling, the downsampled UAV source cloud is output separately. Ground target point cloud .

[0071] S2. High-precision registration of multi-source point clouds based on cascaded feature cross-optimization.

[0072] S2.1 Construction of Cross-Physical Constraint Weight Function: This involves using the nearest point iterative algorithm to find... and When there are matching point pairs, calculate any pair of matching points. Overall weight :

[0073]

[0074] In the formula: Optimize the weights for matching point pairs; Source Point Cloud The coordinates of the i-th point in the array. and Here are its corresponding roughness factor and normal vector; For target point cloud The coordinates of its corresponding point. and Here are its corresponding roughness factor and normal vector; The square of the Euclidean distance between the two points; This is the average distance tolerance threshold for the current iteration step; The preset normal angle tolerance (preferably 0.1 radians); The roughness difference penalty factor is 10 (preferably).

[0075] S2.2, Solving the weighted objective error function: The weights... Substitute into the objective function of minimizing the point-to-surface error:

[0076]

[0077] In the formula: The weighted total error of the registration process; This represents the total number of valid point pairs that were successfully matched. Let be the 3×3 spatial rotation matrix to be solved; Let be the 3×1 spatial translation vector to be solved; through iterative solution, we can achieve... By minimizing R and t, multi-source point clouds are transformed to a unified coordinate system, outputting a high-precision, seamlessly fused point cloud. .

[0078] S3. Quantitative extraction of micro and macroscopic damage to ancient buildings based on cascaded features.

[0079] S3.1 Macroscopic Damage: Extraction of Column Grid Overturning Angle. Spatial geometric fitting is performed on the load-bearing wooden columns to extract the actual centerline vector. Combined with the vertical gravity reference vector provided by the IMU (Integrated Measurement Unit) of the data acquisition device. Calculate the tilt angle of the column. :

[0080]

[0081] In the formula: This represents the actual tilt angle of the wooden pillar; This is the actual centerline direction vector of the wooden pillar obtained through fitting; It is the unit vector of absolute gravity pointing vertically downwards; and These are the magnitudes of the two vectors mentioned above.

[0082] S3.2, Microscopic Damage: Spatial topological clustering of the affected area. Retrieve and merge point clouds. Roughness factor corresponding to each point Extracting the desired result Region-based growth clustering is performed on the points to obtain a continuous set of disease (cracks or decay) points. ,in This is a roughness threshold preset based on healthy smooth samples.

[0083] S4. Deformation-sensing inverse modeling and volumetric infinitesimal quantization integration.

[0084] S4.1 Extraction of Physical Deviation of Normal Deformation: Align the historical standard theoretical model of ancient buildings to the current coordinate system and calculate... any point in the middle Absolute deviation depth relative to the ideal surface along the normal direction :

[0085]

[0086] In the formula: For point The depth of the normal concavity or bulge at the location; The three-dimensional coordinates of the center of the bottom surface of the cylinder are extracted using S3.1 fitting. The unit vector of the ideal vertical midline as defined in the theoretical model; To create the ideal radius of the cylinder as specified in the standard.

[0087] S4.2 Cascade Riemann integral calculation of solid repair and patching volume: The adaptive voxel side length of S1.4 Depth of normal deviation from S4.1 Coupling, establishing a set of disease points Discrete volume integral equation:

[0088] ·

[0089] In the formula: This refers to the total volume of the repair materials that require physical replenishment. This belongs to the disease point set The i-th point in; This represents the absolute depth of the normal depression at that point; Let be the edge length of the local infinitesimal voxel corresponding to that point. This represents the infinitesimal projected area of ​​the point on the local tangent plane. The actual normal vector at that point The angle between the vector and the normal vector of the ideal cylindrical surface. Used to correct projection distortion of curved surfaces.

[0090] S5. Quantitative evaluation decision-making and physical entity repair instruction conversion.

[0091] S5.1 Structural Repair Priority Index (RPI) Calculation: A comprehensive risk index is calculated by a computer processing system, integrating macro and micro parameters.

[0092] ·

[0093] In the formula: This serves as a priority index for structural repair. To standardize the permissible safety limit for tilt angle; For the maximum flexural deviation, To standardize the maximum allowable deformation limit; To compensate for the volume, This refers to the complete total volume recorded in the component's historical archives; The system's preset risk weight coefficients satisfy... .

[0094] S5.2, Solid Repair CNC Instruction Conversion. The computer equipment triggers solid repair control actions based on the above data: when When the threshold is exceeded, the extracted disease point set is used directly. The corresponding reverse mesh boundary generates three-dimensional cutting G-code instructions that can be recognized by CNC machining equipment, driving a multi-axis machine tool to perform precise physical dimension machining on the replacement wood.

[0095] This disclosure also proposes a multi-source point cloud fusion and ancient building reverse modeling system based on feature cascade optimization, including: a data acquisition and cascade feature extraction module, a feature cross-optimization registration module, a macro-micro disease extraction module, a reverse modeling and supplementary volume quantization module, and a physical entity repair execution module;

[0096] The data acquisition and cascaded feature extraction module is used to acquire multi-source heterogeneous point cloud data of ancient buildings, calculate the surface roughness factor and normal vector, perform adaptive manifold downsampling based on the surface roughness factor, and record the side length of local spatial micro-element voxels.

[0097] The feature cross-optimized registration module is used to solve the optimal transformation matrix by introducing the surface roughness factor and normal vector into the cross physical constraint weight function when searching for matching point pairs between the downsampled source point cloud and the target point cloud, and outputs the fused point cloud.

[0098] The macro- and micro-deformation extraction module is used to extract macro-deformation parameters of ancient buildings by geometric fitting based on fused point cloud, and to extract the point set with excessive roughness by topological clustering to obtain the point set of micro-deformation.

[0099] The reverse modeling and volume quantification module is used to compare the physical deviation of normal deformation caused by defects with the preset ideal BIM theoretical model of ancient buildings, and to geometrically couple the side length of local spatial micro-element voxels with the physical deviation of normal deformation through discrete volume integral equations to calculate the volume of solid repair and replacement.

[0100] The physical entity repair execution module is used to calculate the structural repair priority index by comprehensively considering the above parameters, and generate three-dimensional cutting instructions to drive CNC machining equipment to perform physical cutting or generate physical support parameters to guide on-site support correction based on the calculation results.

[0101] The above description is merely a preferred embodiment of the present invention. The scope of protection of the present invention is not limited to the above embodiments. All technical solutions falling within the scope of the present invention's concept are within the scope of protection of the present invention. It should be noted that for those skilled in the art, any improvements and modifications made without departing from the principles of the present invention should also be considered within the scope of protection of the present invention.

Claims

1. A method for reverse modeling ancient buildings using multi-source point cloud feature cascade optimization, executed by computer equipment, characterized in that: Includes the following steps: S1. Obtain multi-source heterogeneous point cloud data of ancient buildings, construct a local covariance tensor for each point in the point cloud data to solve the surface roughness factor and normal vector; perform feature-guided adaptive manifold downsampling based on the surface roughness factor, output the downsampled source point cloud and target point cloud, and record the side length of the retained local spatial micro-element voxels. S2. When searching for matching point pairs between the downsampled source point cloud and the target point cloud, the surface roughness factor and normal vector are introduced into the cross physical constraint weight function to calculate the matching comprehensive weight, construct the weighted target error function and solve the optimal transformation matrix to achieve registration between the downsampled source point cloud and the target point cloud, and output the fused point cloud. S3. Geometric fitting of load-bearing wooden columns of ancient buildings based on fused point cloud to extract macroscopic deformation parameters; The surface roughness factor corresponding to each point is retrieved, and the set of points with excessive roughness is extracted and topologically clustered to obtain the set of micro-defect points. S4. Spatial comparison of the preset ideal BIM theoretical model of ancient buildings with the fused point cloud, and calculation of the physical deviation of the normal deformation of each point relative to the ideal surface along the normal direction. By geometrically coupling the side length of the local spatial micro-element voxel with the physical deviation of the normal deformation, the volume of solid repair and replacement is calculated by using the discrete volume integral equation for the set of micro-defect points. S5. Based on the comprehensive macroscopic deformation parameters, microscopic defect point set, and physical repair and replenishment volume, calculate the structural repair priority index; and based on the structural repair priority index and the reverse mesh boundary of the microscopic defect point set, generate three-dimensional cutting instructions to drive CNC machining equipment to perform physical cutting on the replacement wood, or generate physical support parameters to guide the on-site scaffold support correction.

2. The method for reverse modeling ancient buildings based on multi-source point cloud feature cascade optimization according to claim 1, characterized in that, In S1, the surface roughness factor and normal vector are calculated, and adaptive manifold downsampling guided by the surface roughness factor is performed, specifically including: Select any point in the point cloud data Search its spatial neighborhood Construct a local covariance matrix using the nearest neighbor points: ； For the local covariance matrix Eigenvalue decomposition yields three non-negative eigenvalues. And extract the minimum eigenvalue. The corresponding unit eigenvector is used as the normal vector. ; Using formula Calculate surface roughness factor ; Using formula Solving the side length of local spatial infinitesimal elements ; In the formula, For this The three-dimensional centroid coordinate vector of each neighboring point This is the transpose of the coordinate deviation vector; This is the preset limit physical resolution size. To smooth the amplification factor, This is the roughness sensitivity adjustment factor.

3. The method for reverse modeling ancient buildings based on multi-source point cloud feature cascade optimization according to claim 1, characterized in that, In S2, the surface roughness factor and normal vector are introduced into the cross-physical constraint weighting function to calculate the matching comprehensive weight, construct the weighted objective error function, and solve for the optimal transformation matrix. Specifically, this includes: Calculate any pair of matching points Matching comprehensive weight : ； Match the overall weight Substitute into the objective function of minimizing the weighted point-to-surface error: ； In the formula, Source Point Cloud The coordinates of the i-th point in the array. and Here are its corresponding roughness factor and normal vector; For target point cloud The coordinates of its corresponding point and Here are its corresponding roughness factor and normal vector; The square of the Euclidean distance between the two points; This is the average distance tolerance threshold for the current iteration step; This is a preset tolerance for the included normal angle; This is the roughness difference penalty factor; The weighted total error of the registration process; This represents the total number of valid point pairs that have been successfully matched. Let be a rotation matrix. Let it be a translation vector; solve iteratively to make The minimum R and t are taken as the optimal transformation matrix.

4. The method for reverse modeling ancient buildings based on multi-source point cloud feature cascade optimization according to claim 1, characterized in that, In S3, the extraction of macroscopic deformation parameters specifically includes: Spatial geometric fitting is performed on the load-bearing wooden columns in the fused point cloud to extract the actual centerline vector. and the three-dimensional coordinates of the column base center ; Using formula Calculate the tilt angle of ancient buildings In the formula, The absolute gravity is the vertically downward reference vector; Least square fitting is performed on the three-dimensional coordinates of the center of the base of multiple columns on the same elevation ring to obtain the actual foundation settlement reference surface, and the elevation settlement difference of each column base relative to the actual foundation settlement reference surface is calculated.

5. The method for reverse modeling ancient buildings based on multi-source point cloud feature cascade optimization according to claim 1, characterized in that, In S3, obtaining the microscopic disease point set specifically includes: Traverse and merge point clouds to extract the required parameters. The characteristic points are used as suspected disease points, among which The surface roughness factor of the feature point. A roughness threshold preset based on healthy component samples; A region growing algorithm based on Euclidean distance is used to perform topological clustering on suspected disease points, and continuous sets of surface peeling or crack points are extracted as micro-disease point sets. .

6. The method for reverse modeling ancient buildings based on multi-source point cloud feature cascade optimization according to claim 5, characterized in that, In S4, the physical deviation of the normal deformation of each point relative to the ideal surface along the normal direction is calculated, and the volume of physical repair and replacement is calculated by solving the discrete volume integral equation for the set of micro-defect points. Specifically, this includes: Using formula Solving for any point in the fused point cloud Physical deviation of normal deformation ; Using the discrete volume integral equation: Solve the volume of solid repair and replacement ; In the formula, The three-dimensional coordinates of the center of the base of the cylinder are: This refers to the unit vector of the vertical centerline specified in the ideal BIM theoretical model of ancient buildings. To create the ideal radius specified in the standard; This belongs to the microscopic disease point set The point in the middle, Let its normal absolute concavity depth be denoted as ; let the side length of the local spatial voxel preserved at each point in the downsampled source point cloud output in S1 be . , The angle between the actual normal vector at that point and the normal vector of the ideal surface.

7. The method for reverse modeling ancient buildings based on multi-source point cloud feature cascade optimization according to claim 6, characterized in that, In S5, the structural repair priority index is calculated using the following formula: ； In the formula, This serves as a priority index for structural repair. To standardize the permissible safety limit for tilt angle; For the maximum flexural deviation, To standardize the maximum allowable deformation limit; To compensate for the volume, This refers to the complete total volume recorded in the component's historical archives; The risk weight coefficient is preset for the system.

8. A multi-source point cloud fusion and ancient building reverse modeling system based on feature cascade optimization, characterized in that, The system is used to implement the ancient building reverse modeling method of multi-source point cloud feature cascade optimization as described in any one of claims 1 to 7. The system includes: a data acquisition and cascade feature extraction module, a feature cross-optimization registration module, a macro-micro disease extraction module, a reverse modeling and supplementary volume quantization module, and a physical entity repair execution module. The data acquisition and cascaded feature extraction module is used to acquire multi-source heterogeneous point cloud data of ancient buildings, calculate the surface roughness factor and normal vector, perform adaptive manifold downsampling based on the surface roughness factor, and record the side length of local spatial micro-element voxels. The feature cross-optimized registration module is used to solve the optimal transformation matrix by introducing the surface roughness factor and normal vector into the cross physical constraint weight function when searching for matching point pairs between the downsampled source point cloud and the target point cloud, and output the fused point cloud. The macro-micro defect extraction module is used to extract macro deformation parameters of ancient buildings by geometric fitting based on fused point cloud, and to extract the point set with excessive roughness by topological clustering to obtain the point set of micro defects. The reverse modeling and volume quantification module is used to compare the physical deviation of normal deformation caused by the disease with the preset ideal BIM theoretical model of ancient buildings, and to geometrically couple the side length of local spatial micro-element voxels and the physical deviation of normal deformation through discrete volume integral equations to calculate the volume of solid repair and replacement. The physical entity repair execution module is used to calculate the structural repair priority index by comprehensively considering the above parameters, and generate three-dimensional cutting instructions to drive CNC machining equipment to perform physical cutting or generate physical support parameters to guide on-site support correction based on the calculation results.

9. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes a computer program, it implements the ancient building reverse modeling method of multi-source point cloud feature cascade optimization as described in any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the reverse modeling method for ancient buildings based on multi-source point cloud feature cascade optimization as described in any one of claims 1 to 7.

Images