Ancient building monitoring scanning method and system based on three-dimensional modeling

Abstract

This invention relates to the field of 3D modeling technology, specifically to a method and system for monitoring and scanning ancient buildings based on 3D modeling. The method includes the following steps: acquiring point cloud data of the ancient building through 3D laser scanning; grouping and processing to calculate the deviation between the normal vector and the radius of curvature to generate a curvature difference sequence; matching the normal vector deviation angle input with principal component analysis to extract the curvature gradient direction and generate a 3D mesh model; extracting node position tensors and rearranging them beyond thresholds to correct the 3D model; calculating Gaussian curvature and normal vector offset rate to identify damaged areas; performing tensor coupling to evaluate the monitoring status and outputting a 3D monitoring model of the ancient building. In this invention, a correlation sequence between curvature difference and normal vector deviation is established through point cloud spatial neighborhood grouping to identify minor geometrical abrupt changes in complex surfaces; principal component analysis maintains structural continuity; dynamic node rearrangement corrects geometric drift; and multi-feature fusion achieves damage identification and status classification, improving model stability and recognition reliability.

Description

Technical Field

[0001] This invention relates to the field of 3D modeling technology, and in particular to a method and system for monitoring and scanning ancient buildings based on 3D modeling. Background Technology

[0002] The field of 3D modeling technology involves a comprehensive technical system for spatial data acquisition, digital reconstruction, and structural representation of objects or scenes. Its core aspects include the acquisition of spatial geometric information, processing of point cloud data, model surface reconstruction, and texture mapping. This technical field is widely used in engineering surveying, cultural relic preservation, architectural restoration, and virtual simulation, aiming to reflect the geometric structure and morphological characteristics of real space with high-precision data models.

[0003] Traditional ancient building monitoring and scanning methods refer to the technical means of collecting structural morphology data and monitoring changes in ancient buildings with historical value and complex structural features. These methods use terrestrial laser scanners, photogrammetric equipment, or structured light scanning devices to acquire point cloud and image information of the building's exterior surface. Spatial registration and data stitching are then performed by fixing the location of the measuring station or setting up marker points to establish a three-dimensional model of the ancient building and to enable subsequent morphological comparison and deformation analysis.

[0004] Traditional methods for monitoring ancient buildings are susceptible to the effects of viewpoint obstruction and differences in reflectivity during point cloud acquisition, leading to the loss of local surface information. This results in faults or discontinuities in complex structural areas of the model. The registration process relies on fixed marker points, which are prone to global coordinate errors due to positional deviations, causing morphological distortion and spatial misalignment on the model surface. Subsequent comparisons tend to focus on the overall deformation trend while ignoring subtle curvature changes and local deformation features, making it difficult to effectively identify minor damage such as early cracks and erosion. This results in insufficient spatial accuracy of the monitoring results and unstable damage assessment, affecting the continuity and reliability of long-term data analysis. Summary of the Invention

[0005] To address the technical problems of traditional ancient building monitoring methods, such as the loss of local surface information due to viewpoint obstruction and reflectivity differences during point cloud acquisition, leading to discontinuities or faults in complex structural areas, the reliance on fixed marker points in the registration process, which can easily cause global coordinate errors due to positional deviations, resulting in morphological distortion and spatial misalignment of the model surface, and the tendency to focus on overall deformation trends in subsequent comparisons while ignoring subtle curvature changes and local deformation features, making it difficult to effectively identify early cracks, erosion, and other minor damage, resulting in insufficient spatial accuracy and unstable damage assessment of monitoring results, and affecting the continuity and reliability of long-term data analysis, this invention provides a scanning method for ancient building monitoring based on 3D modeling.

[0006] To achieve the above objectives, this invention employs a three-dimensional modeling-based method for monitoring and scanning ancient buildings, comprising the following steps: S1: Obtain point cloud data of the ancient building surface through 3D laser scanning, perform spatial adjacency grouping processing, calculate the weighted combination of the normal vector angle and the radius of curvature of the points, and generate a curvature difference sequence; S2: Call the curvature difference sequence, obtain the normal vector deviation angle for pair matching, input it into the principal component analysis model to perform direction sorting, extract the curvature gradient direction change rate, and generate a three-dimensional mesh model based on point cloud coordinates; S3: Call the three-dimensional mesh model, extract the spatial position tensor gradient of the mesh nodes, and when the ratio of the rate of change to the position tensor gradient exceeds a preset threshold, perform node coordinate rearrangement, perform geometric correction on the three-dimensional mesh model, and generate a corrected three-dimensional model. S4: Call the corrected 3D model, extract the Gaussian curvature, normal vector offset rate and interior angle gradient of the boundary points of the damaged area, input them into the support vector machine model to calculate the difference ratio, mark the location of the damaged area, and generate a 3D model of the damaged area. S5: Call the three-dimensional model of the damaged marker, perform tensor coupling operation on the Gaussian curvature of the damaged boundary, the normal vector offset rate, and the interior angle gradient, conduct monitoring status assessment of the damaged area and mark early warning, and output the three-dimensional monitoring model of the ancient building.

[0007] As a further embodiment of the present invention, the curvature difference sequence includes curvature change amplitude, normal vector difference degree and spatial continuity coefficient; the three-dimensional mesh model includes curvature gradient distribution, point cloud topology relationship and mesh structure parameters; the corrected three-dimensional model includes geometric correction coefficient, node rearrangement matrix and spatial optimization parameters; the damage identification three-dimensional model includes damage area distribution, boundary feature parameters and difference ratio mapping results; and the ancient building three-dimensional monitoring model includes damage status assessment indicators, risk warning information and structural health monitoring parameters.

[0008] As a further aspect of the present invention, the specific steps of S1 are as follows: S101: Obtain point cloud data of ancient building surface through 3D laser scanning, perform neighborhood distance calculation on the 3D coordinates of the points, aggregate points whose Euclidean distance is lower than the preset distance threshold, and remove isolated points whose number of neighborhood points is lower than the preset threshold, generating a cluster of spatially adjacent points; The preset distance threshold for Euclidean distance is determined based on the statistical results of the average point spacing of the scanned point cloud on the surface of the ancient building; The threshold for the number of neighboring points is determined based on the spatial consistency statistical results of the point cloud density distribution. S102: Based on the spatial adjacent point cluster, extract the point normal vector, calculate the angle between the normal vectors of adjacent points, and extract the curvature radius change rate of the corresponding point. After normalizing the angle value and the curvature change rate, perform weighted summation to generate a normal curvature deviation dataset. S103: Based on the normal curvature deviation dataset, the deviation data of each point cluster is serialized and rearranged according to the point index order, and numerical difference operation is performed on adjacent deviation values ​​to calculate the change amplitude between continuous points and reorganize them according to the index order to generate a curvature difference sequence.

[0009] As a further aspect of the present invention, the specific steps of S2 are as follows: S201: Obtain the curvature difference sequence, calculate the curvature difference between adjacent sampling points and form a curvature direction vector set, extract the direction cosine parameter, compare it with the point cloud normal vector component to calculate the angle deviation, match the angle deviation data group according to the spatial index, and generate a normal vector deviation matching set. S202: Call the normal vector deviation matching set, extract the direction cosine parameter components and calculate the covariance, input the covariance result into the principal component analysis model to perform feature component decomposition, extract the direction sequence index according to the principal component weight, and obtain the direction sorting index sequence. S203: Sort the index sequence according to the direction, calculate the directional difference of the curvature gradient components of adjacent index nodes, extract the curvature gradient direction change rate, match the change rate with the point cloud coordinate index, and generate a three-dimensional mesh model according to the coordinate weight.

[0010] As a further aspect of the present invention, the specific steps of S3 are as follows: S301: Call the three-dimensional mesh model, retrieve the three-dimensional coordinate information of the mesh nodes, calculate the coordinate difference between adjacent nodes and normalize it to a unit length, construct the node position gradient tensor, and associate it with the curvature gradient direction change rate of the corresponding node to generate the node spatial position tensor gradient set. S302: Based on the set of gradients of the node spatial position tensor, extract the associated rate of change of curvature gradient direction, calculate the ratio of the rate of change to the gradient of the position tensor and compare it with the judgment threshold, record the index of the node exceeding the threshold and extract the offset to obtain the set of node change differences. S303: Based on the node change difference set, rearrange the coordinate components of the marked nodes, adjust the spatial position of the nodes according to the offset, recalculate the mesh topology connection relationship and update the geometric structure distribution to obtain the corrected three-dimensional model.

[0011] As a further aspect of the present invention, adjusting the node spatial position based on the offset refers to scaling the offset vector according to a set ratio and then superimposing it with the original node coordinates to complete the linear correction of the node spatial position.

[0012] As a further aspect of the present invention, the specific steps of S4 are as follows: S401: Call the corrected three-dimensional model, scan the coordinate set of boundary points of the damaged area, calculate the curvature radius difference of adjacent grid units to obtain the curvature change rate, extract Gaussian curvature feature parameters based on the change results, and generate a Gaussian curvature feature set. S402: Based on the Gaussian curvature feature set, calculate the difference between the direction component of the boundary point normal vector and the direction cosine of the surface normal, perform direction difference normalization, accumulate the offset rate component, and generate a normal vector offset rate feature set. S403: Call the normal vector offset rate feature set, calculate the gradient sequence of the interior angles of adjacent triangular faces based on the node topology of the corrected 3D model, and input it together with the Gaussian curvature feature set and the normal vector offset rate feature set into the support vector machine to perform difference ratio clustering to obtain the damaged 3D model.

[0013] As a further aspect of the present invention, the specific steps of S5 are as follows: S501: Call the Gaussian curvature of the damaged boundary, the normal vector offset rate and the interior angle gradient data in the three-dimensional model of the damaged identifier, perform matching and normalization of the component parameters on the node coordinate index, and aggregate the processing results into a three-dimensional tensor unit set according to the node sequence to generate a damaged feature tensor set. S502: Based on the damaged feature tensor set, extract the principal direction component of the tensor unit, perform weighted integral calculation on the coupling parameters between the Gaussian curvature component, the normal offset component and the interior angle gradient component, and generate a tensor coupling strength set. S503: Based on the tensor coupling strength set, compare the coupling strength value of the damaged fragment with the damage state threshold, record the unit index above the threshold and mark the damage warning, map it to the three-dimensional coordinate field, and form a three-dimensional monitoring model of ancient buildings.

[0014] As a further aspect of the present invention, the normalized component parameters refer to mapping the Gaussian curvature, normal vector offset, and interior angle gradient data to a unified standardized interval and performing boundary correction on outlier components that exceed the interval, so that the normalized components remain within the quantizable range and are aggregated according to the node sequence to form a three-dimensional tensor unit set. The damaged state threshold refers to a dynamic threshold set based on the statistical distribution results of the tensor coupling strength set, according to the mean square error range of the coupling strength values ​​of the damaged segments.

[0015] A 3D modeling-based ancient building monitoring and scanning system includes: The point cloud processing module acquires point cloud data of the ancient building surface through 3D laser scanning, performs spatial adjacency grouping processing, calculates the weighted combination of the normal vector angle and the radius of curvature of the points, generates a curvature difference sequence, and transmits it to the spatial curvature calculation module. The spatial curvature calculation module calls the curvature difference sequence, obtains the normal vector deviation angle for pair matching, inputs it into the principal component analysis model to perform direction sorting, extracts the curvature gradient direction change rate, generates a three-dimensional mesh model based on point cloud coordinates, and passes it to the direction feature extraction module. The orientation feature extraction module calls the three-dimensional mesh model to extract the spatial position tensor of the mesh nodes. When the ratio of the rate of change to the position tensor exceeds a preset threshold, the node coordinates are rearranged to perform geometric correction on the three-dimensional mesh model, generate a corrected three-dimensional model, and pass it to the geometric correction module. The geometric correction module calls the corrected 3D model, extracts the Gaussian curvature, normal vector offset rate and interior angle gradient of the boundary points of the damaged area, inputs them into the support vector machine model to calculate the difference ratio, identifies the location of the damaged area, generates a 3D model of damage identification, and transmits it to the damage monitoring module. The damage monitoring module calls the three-dimensional model of the damage marker, performs tensor coupling operations on the Gaussian curvature, normal vector offset, and interior angle gradient of the damage boundary, evaluates the monitoring status of the damage area and marks it for early warning, and outputs the three-dimensional monitoring model of the ancient building.

[0016] Compared with the prior art, the advantages and positive effects of the present invention are as follows: In this invention, by grouping the spatial neighborhood of point cloud data and establishing a correlation sequence based on curvature difference and normal vector deviation, minute geometrical abrupt changes can be identified on complex surfaces, enhancing the accuracy of curvature gradient changes. The directional sorting of principal component analysis ensures that the model maintains structural continuity and spatial coordination during generation. The dynamic rearrangement of node coordinates can correct local geometrical drift caused by scanning errors, improving the model's matching degree in morphological restoration. Through multi-feature fusion analysis of Gaussian curvature, normal vector offset, and interior angle gradient, accurate boundary identification and state classification of damaged areas can be achieved, thereby obtaining higher stability and early warning accuracy in building morphology monitoring and ensuring the geometric consistency and identification reliability of the model in long-term observation. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the accompanying drawings without creative effort.

[0018] Figure 1 This is a schematic diagram of the steps of the present invention; Figure 2 This is a detailed schematic diagram of S1 of the present invention; Figure 3 This is a detailed schematic diagram of S2 of the present invention; Figure 4 This is a detailed schematic diagram of S3 of the present invention; Figure 5 This is a detailed schematic diagram of S4 of the present invention; Figure 6 This is a detailed schematic diagram of S5 of the present invention; Figure 7 This is a system module diagram of the present invention. Detailed Implementation

[0019] The technical solution of the present invention will now be described with reference to the accompanying drawings.

[0020] In embodiments of the present invention, words such as "exemplarily," "for example," etc., are used to indicate that something is an example, illustration, or description. Any embodiment or design described as "exemplary" in the present invention should not be construed as being more preferred or advantageous than other embodiments or designs. Specifically, the use of the word "exemplary" is intended to present the concept in a concrete manner. Furthermore, in embodiments of the present invention, the meaning expressed by "and / or" can be both, or either one.

[0021] In the embodiments of this invention, the terms "image" and "picture" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning. Similarly, the terms "of," "corresponding (relevant)," and "corresponding" may sometimes be used interchangeably. It should be noted that, without emphasizing the distinction between them, they convey the same meaning.

[0022] In this embodiment of the invention, sometimes the subscript such as W1 is written in a non-subscript form such as W1. When the difference is not emphasized, the meaning they express is the same.

[0023] To make the technical problems, technical solutions and advantages of the present invention clearer, a detailed description will be given below in conjunction with the accompanying drawings and specific embodiments.

[0024] Please see Figure 1 This invention provides a method for monitoring and scanning ancient buildings based on three-dimensional modeling, including the following steps: S1: Obtain point cloud data of the ancient building surface through 3D laser scanning, perform spatial adjacency grouping processing, calculate the weighted combination of the normal vector angle and the radius of curvature of the points, and generate a curvature difference sequence; S2: Call the curvature difference sequence, obtain the normal vector deviation angle for pair matching, input it into the principal component analysis model to perform direction sorting, extract the curvature gradient direction change rate, and generate a three-dimensional mesh model based on point cloud coordinates; S3: Call the 3D mesh model, extract the spatial position tensor gradient of the mesh nodes, and when the ratio of the rate of change to the position tensor gradient exceeds the preset threshold, perform node coordinate rearrangement, perform geometric correction on the 3D mesh model, and generate a corrected 3D model. S4: Call the corrected 3D model, extract the Gaussian curvature, normal vector offset rate and interior angle gradient of the boundary points of the damaged area, input them into the support vector machine model to calculate the difference ratio, mark the location of the damaged area, and generate a 3D model of the damaged area. S5: Call the 3D model of the damaged area, perform tensor coupling operation on the Gaussian curvature, normal vector offset and interior angle gradient of the damaged boundary, conduct monitoring status assessment of the damaged area and mark early warning, and output the 3D monitoring model of the ancient building.

[0025] The curvature difference sequence includes the curvature change amplitude, normal vector difference degree, and spatial continuity coefficient; the three-dimensional mesh model includes the curvature gradient distribution, point cloud topology, and mesh structure parameters; the corrected three-dimensional model includes geometric correction coefficients, node rearrangement matrix, and spatial optimization parameters; the damage identification three-dimensional model includes the damage area distribution, boundary feature parameters, and difference ratio mapping results; and the ancient building three-dimensional monitoring model includes damage status assessment indicators, risk warning information, and structural health monitoring parameters.

[0026] Please see Figure 2 The specific steps of S1 are as follows: S101: Obtain point cloud data of ancient building surface through 3D laser scanning, perform neighborhood distance calculation on the 3D coordinates of the points, aggregate points whose Euclidean distance is lower than the preset distance threshold, and remove isolated points whose number of neighborhood points is lower than the preset threshold, generating a cluster of spatially adjacent points; The preset distance threshold for Euclidean distance is determined based on the statistical results of the average point spacing of the scanned point cloud on the surface of the ancient building; The threshold for the number of neighboring points is determined based on the spatial consistency statistics of the point cloud density distribution; Three-dimensional laser scanning is used to acquire surface point cloud data of the brackets and beams of ancient buildings, such as a Ming Dynasty wooden archway. The data is stored in the form of three-dimensional coordinate points. For example, the coordinates of a point P1 are... The unit is millimeters. Determining the preset distance threshold first requires statistically analyzing the average point spacing of all point clouds collected from the dougong (bracket set) area. This is done by calculating the Euclidean distance between each point and its nearest neighbor, and then averaging these distances. Assuming the statistically obtained average point spacing on the surface of the dougong is... Based on this statistical result, the preset distance threshold is set to the average distance between points. times, that is The determination of the threshold for the number of neighboring points is based on the spatial consistency statistical results of the point cloud density distribution. After performing density analysis on the point cloud data of the entire archway scan, it was found that in areas with intact and flat structures, taking any point as the center and a radius of... The number of points typically contained within a sphere is... There are more than one point, while at the edge or in sparsely noisy regions, the number of points within a sphere of the same radius may be less than one. Therefore, to effectively distinguish between main structural points and isolated noise points, the threshold for the number of neighboring points is set to [number]. In specific execution, for point P1 Calculate the three-dimensional Euclidean distance between it and the other points. For example, there exists a point P2 with coordinates of... The Euclidean distance between the two points is: ; The distance is less than the preset value. The threshold is used, so P2 is considered a neighbor of P1, while the coordinates of the other point P3 are... The calculated distance between it and P1 is much greater than If the point is not considered a neighboring point, then after traversing the points, the total number of points in the neighborhood of P1 is calculated. One point, this quantity Greater than the preset threshold for the number of neighboring points Therefore, P1 is preserved; conversely, if there exists an isolated point P4, then P1 is preserved. Only found within the neighborhood The number of neighboring points is less than the threshold. If P4 is an isolated point, it will be removed as such. This is done by calculating the neighborhood distance and determining the number of neighboring points, and then removing points with a distance less than the required Euclidean distance. And the number of neighboring points is not less than Points are aggregated together to generate spatially adjacent point clusters.

[0027] S102: Based on the cluster of spatially adjacent points, extract the point normal vector, calculate the angle between the normal vectors of adjacent points, and extract the rate of change of the radius of curvature of the corresponding points. After normalizing the angle value and the rate of change of curvature, perform weighted summation to generate a normal curvature deviation dataset. Based on the generated set of spatially adjacent point clusters, such as the point cluster formed after removing isolated points from the aforementioned archway bracket section, a point is extracted from it. and a neighboring point in the same cluster First, for the point Extracting point normal vectors, this process involves selecting... Surrounding radius For each neighboring point within a given area, principal component analysis is performed on the set of 3D coordinates of that point to obtain a local plane that best fits the point. The normal vector of this plane is then used as the point's normal vector. point normal vector Assuming The unit vector coordinates are Using the same method, for adjacent points Calculate its point normal vector Assuming its coordinates are Next, calculate the normal vectors of these two adjacent points. and The angle between them is calculated by taking the inverse cosine of the dot product, i.e. The included angle value is approximately Simultaneously, the rate of change of the radius of curvature at the corresponding point is extracted. This process first calculates the point... The curvature value at a given point is determined by the coefficients of the quadratic surface fitted to its neighborhood point set, assuming that the calculated value is... The curvature is The corresponding radius of curvature is Calculate using the same method The curvature is The corresponding radius of curvature is approximately The rate of change of the radius of curvature is obtained by calculating the ratio of the difference between the radii of curvature at two points to their average value, i.e. Subsequently, the included angle value With rate of change of curvature Normalization is performed, and the normalization reference range for the included angle values ​​is set to... to , where greater than This is considered a characteristic mutation and is directly assigned a value. ,but The normalized value is The normalized reference range for the rate of change of curvature is set to to ,but The normalized value is Finally, a weighted sum is performed on the two normalized values. Here, to highlight the sensitivity of the normal vector change to edge features, the weight of the included angle value is set. for Weight of the rate of change of curvature for The summation result is Use this value as a point Its neighboring points The normal curvature deviation between points is calculated, and this process is repeated for adjacent point pairs in the point cluster to generate a normal curvature deviation dataset.

[0028] S103: Based on the normal curvature deviation dataset, the deviation data of each point cluster is serialized and rearranged according to the point index order, and numerical difference operation is performed on adjacent deviation values ​​to calculate the change amplitude between consecutive points and reorganize them according to the index order to generate a curvature difference sequence. Based on the normal curvature deviation dataset, which records the geometric feature differences of each pair of adjacent points in the bracket set of ancient building archways, for example, for a set containing... For a local point cluster of points, the normal curvature deviation values ​​between consecutive point pairs are, in order of their index order on the scan path, as follows: Next, the deviation data of each point cluster is serialized and rearranged according to the point index order to ensure that the order of data processing is consistent with the spatial arrangement order of the points. Then, numerical difference operation is performed on adjacent deviation values. This operation is achieved by calculating the absolute value of the difference between the next deviation value and the previous deviation value in the sequence, so as to quantify the severity of the deviation change. The specific calculation process is shown in Table 1.

[0029] Table 1: Example Table for Calculating Curvature Difference Sequence

[0030] As shown in Table 1, when calculating the variation range between consecutive points, the difference between the deviation values ​​of indices (2-3) and (1-2) is... The difference between the deviation values ​​of indices (3-4) and (2-3) is These two differences are relatively small, where "relatively small" is defined as a difference below a preset mutation threshold. The difference between the deviation values ​​of indices (4-5) and (3-4) is This value is greater than the mutation detection threshold. This indicates a sharp change in geometric feature near index 4, possibly corresponding to a crack or the starting point of a carved edge on the bracket component. Similarly, the difference in deviation values ​​between indices (6-7) and (5-6) is... Similarly, it is greater than the threshold. The mutation value will be used to calculate the magnitude of the change, i.e.: ; Reorganize according to the original point index order to generate a curvature difference sequence.

[0031] Please see Figure 3 The specific steps of S2 are as follows: S201: Obtain the curvature difference sequence, calculate the curvature difference between adjacent sampling points and form a curvature direction vector set, extract the direction cosine parameter, compare it with the point cloud normal vector component to calculate the angle deviation, match the angle deviation data group according to the spatial index, and generate a normal vector deviation matching set. Obtain the generated curvature difference sequence, such as the sequence for the edge of a crack in a section of the archway brackets. And simultaneously retrieve the original point cloud 3D coordinates of the corresponding index of the sequence, such as point Coordinates are ,point Coordinates are The process involves performing a difference operation between adjacent sampling points to calculate the displacement vector between them. Specifically, this involves subtracting the coordinate components of the previous point from the coordinate components of the subsequent point to obtain the point. arrive displacement vector The unit is millimeters. This coordinate difference operation is repeated for adjacent point pairs in the sequence to form a set of displacement vectors representing the direction of curvature change, constituting a curvature direction vector set. Then, the direction cosine parameter of each displacement vector is extracted. First, the magnitude of the vector is calculated, for example... The modulus is Then, each component of the vector is divided by its magnitude to obtain the normalized direction cosine parameter. Its direction cosine is Then, the extracted direction cosine parameters are compared with the corresponding normal vector components of the point cloud to calculate the angle deviation. Here, the point is retrieved. point normal vector Its unit vector coordinates are By calculating the dot product of the direction cosine vector and the normal vector, the cosine value of the angle between the two vectors is obtained. Finally, the angle deviation is obtained by taking its inverse cosine. This angular deviation value With point The spatial index is associated, and the complete process of extracting direction cosine, comparing normal vectors and calculating angle deviation is repeated for the vectors in the curvature direction vector set. Each calculated angle deviation value is stored in correspondence with the spatial index of the starting point to generate a normal vector deviation matching set.

[0032] S202: Call the normal vector deviation matching set, extract the direction cosine parameter components and calculate the covariance, input the covariance result into the principal component analysis model to perform feature component decomposition, extract the direction sequence index according to the principal component weight, and obtain the direction sorting index sequence. The generated normal vector deviation matching set is invoked. This matching set contains the direction cosine parameters associated with each point cloud index. For example, for a local area near the crack in the archway, the direction cosine parameters of three consecutive index points are extracted, as shown in Table 2. First, the covariance matrix of this set of direction cosine parameter components is calculated. Each component needs to be calculated first. The average value of ) , , Then, the variance and covariance are calculated based on the mean, for example, The variance of the components is: ; and and The covariance of the components is: ; After completing the calculation of component pairs, a result is obtained. The covariance matrix is ​​then substituted into a computational process for performing eigencomponent decomposition. This process obtains three eigenvalues ​​by solving for the eigenvalues ​​and eigenvectors of the matrix, let's assume they are... and the corresponding three orthogonal eigenvectors The magnitude of the eigenvalues ​​represents the weights of the principal components. The largest, its corresponding eigenvector The direction of the greatest data variation, i.e., the direction of features such as cracks, was identified. The direction sequence index was extracted based on the principal component weights. Specifically, the original cosine vector of each direction (as shown in Table 2) was used to... ) respectively with the main feature vector When performing a dot product operation, the larger the absolute value of the dot product, the closer the direction vector is to the principal direction. Let's assume the calculated absolute values ​​of the dot products are... , , Sort the original point indices from largest to smallest based on the dot product value. The corresponding dot product is the largest, followed by Finally Therefore, a sorted index sequence is obtained, for example This yields the directional sorted index sequence.

[0033] Table 2: Example Table of Direction Cosine Parameters

[0034] As shown in Table 2, the table lists the direction cosine vectors of three example points extracted from the normal vector deviation matching set. These vectors will be used for subsequent covariance calculation and principal component analysis.

[0035] S203: Sort the index sequence by direction, calculate the direction difference of the curvature gradient components of adjacent index nodes, extract the curvature gradient direction change rate, match the change rate with the point cloud coordinate index, and generate a three-dimensional mesh model based on the coordinate weight. Sort the index sequence according to the generated direction, for example This sequence defines the traversal order of points along the main features (such as cracks) of the ancient building surface. Next, the directional differences of the curvature gradient components between adjacent index nodes are calculated. First, the curvature gradient vector is calculated for each point in the sequence. This vector is determined by analyzing the distribution of curvature values ​​between that point and its neighbors, pointing in the direction of the fastest curvature increase. Assuming the calculated points... The curvature gradient vector is Its neighboring points in the sequence The curvature gradient vector is The difference in direction between the two gradient vectors is quantified by calculating the angle between them, i.e. Substitute the values ​​to perform the dot product operation. Assuming both gradient vectors are unit vectors, the difference in direction is: Then, extract the rate of change of curvature gradient direction, divide the previously calculated directional difference value by the Euclidean distance between the two points, assuming point... and The spatial distance between them is The rate of change is This rate of change value is matched to the starting point of the point pair. On the coordinate index, this process is repeated for adjacent point pairs in the sorted index sequence along the entire direction. The calculated rate of change is matched with the corresponding point cloud coordinate index. Next, a 3D mesh model is generated based on the coordinate weights. Here, the coordinate weights are directly determined by the rate of change of the curvature gradient direction. The "higher" the weight value, the more drastic the geometric structure change in the region where the point is located. Among them, a rate of change greater than... The region was identified as a high-weight region, with a rate of change of [missing information]. to The region between these two is the medium-weighted region, and the region below this weight is... For low-weight regions, during mesh construction, for example using a sphere-rolling connection method, different rolling sphere radii are set for regions with different weights. In high-weight regions, a smaller radius is used, for example... This encourages the formation of smaller, denser triangular patches between points, while in low-weight regions, a larger radius is used, for example... This process forms large triangular patches, which are then used to connect point cloud data into a complete three-dimensional surface, generating a three-dimensional mesh model.

[0036] Please see Figure 4 The specific steps of S3 are as follows: S301: Call the 3D mesh model, retrieve the 3D coordinate information of the mesh nodes, calculate the coordinate difference between adjacent nodes and normalize it to a unit length, construct the node position gradient tensor, and associate it with the curvature gradient direction change rate of the corresponding node to generate the node spatial position tensor gradient set. The generated 3D mesh model, which details the geometric shape of the bracket system of the ancient architectural archway, is invoked. First, the 3D coordinate information of the nodes in the mesh is retrieved, and then one node is selected. Its coordinates are The unit is millimeters, and it retrieves a node that is directly adjacent to it on the grid topology. Its coordinates are Next, calculate the coordinate difference between these two adjacent nodes, that is... Subtract the coordinate components The coordinate components are used to obtain a difference vector: ; Subsequently, the difference vector is normalized by first calculating its magnitude. Then, each component of the vector is divided by the magnitude to obtain a normalized vector of unit length. For nodes Adjacent nodes, for example Repeat this coordinate difference and normalization calculation to obtain a set of unit vectors pointing to multiple neighboring nodes, such as... Wait, and use this set of unit vectors as nodes. The position gradient tensor describes The spatial orientation around the point, the next step is to correlate and calculate the values ​​corresponding to the node. The rate of change of the curvature gradient direction, assuming The rate of change at that point is This rate of change value is compared with the node The location gradient tensor is bound to the node. Specifically, a data structure is created that contains the node index, the location gradient tensor of the node (i.e., a set of unit vectors), and the rate of change of its curvature gradient direction. For each node in the 3D mesh model, the above process of searching for neighboring nodes, calculating coordinate differences, normalizing, constructing the location gradient tensor and associating the rate of change is performed. This generates a data record containing the spatial location gradient and the rate of change of geometric features for each node in the entire model, generating a set of node spatial location tensor gradients.

[0037] S302: Based on the set of gradients of the node spatial location tensor, extract the rate of change of the associated curvature gradient direction, calculate the ratio of the rate of change to the gradient of the location tensor and compare it with the judgment threshold, record the index of the node that exceeds the threshold and extract the offset to obtain the set of node change differences. Based on the generated set of node spatial location tensor gradients, this dataset provides detailed geometric and spatial variation information for each node in the archway bracket model. First, a node is extracted from the dataset, for example, node... And retrieve the rate of change of its associated curvature gradient direction, which is... Next, the ratio of this rate of change to the location tensor gradient is calculated. Here, the "location tensor gradient" is not a single numerical value, but rather represents the local geometric stability of the node. The actual calculation process is as follows: calculate the node... Directly adjacent nodes (e.g.) The average rate of change of the curvature gradient direction of ), assuming that the values ​​of this term are respectively , , The local average rate of change is Then calculate the nodes The ratio of the rate of change to this local mean is obtained. This ratio Quantified nodes Compared to the degree of geometric abrupt change in its surrounding environment, the calculated ratio will then be... The ratio is compared with a preset decision threshold, which is set based on statistical analysis of the ratio of nodes in the model. For example, the threshold is set as the mean of the ratio samples plus twice the standard deviation. Assuming the calculated threshold is... ,because Greater than Therefore, node A node identified as exceeding the threshold is recorded at its index. Next, the offset of this node is extracted. The offset is a vector, its direction set to the average direction of the unit vectors in the node's gradient tensor. Its magnitude is calculated based on how much the ratio exceeds the threshold. The specific formula is: Offset magnitude = (Ratio - Threshold) × Scaling factor, where the scaling factor is set based on the average side length of the grid. Assuming the average side length is... The scaling factor is one-tenth of it, that is ,but The offset size is The magnitude of this offset is combined with the average direction vector to obtain the final offset vector, as shown in Table 3. This process is repeated for nodes to record the index of the nodes that exceed the threshold and their corresponding offset vectors, thus obtaining the node change difference set.

[0038] Table 3: Example Table for Calculating Node Change Differences

[0039] As shown in Table 3, this table lists the filtering process for some nodes, where nodes and Because its calculated ratio is higher than the threshold The nodes were marked as exceeding the threshold, and their corresponding offsets were calculated.

[0040] S303: Based on the node change difference set, rearrange the coordinate components of the marked nodes, adjust the spatial position of the nodes according to the offset, recalculate the mesh topology connection relationship and update the geometric structure distribution to obtain the corrected 3D model. Based on the obtained set of node change differences, this set is explicitly marked as follows: For nodes exhibiting significant geometric abrupt changes and their offset vectors requiring adjustment, the coordinate components of the marked nodes are first adjusted. This "rearrangement" is actually a position update operation, taking the nodes as an example. For example, its original coordinates are The offset vector is obtained from the set of differences in node changes. The original coordinates are added to the corresponding components of the offset vector to obtain the new node coordinates. This coordinate adjustment operation is performed on all marked nodes, moving them to their new spatial locations. Subsequently, due to the node movement, the original mesh topology connections may no longer be ideal and need to be recalculated. Specifically, during execution, the nodes are traversed and compared with the moved nodes (e.g., ...). Connected triangular facets, calculate the shape quality of each triangle, for example, for a triangle composed of nodes For the triangle formed, calculate its three interior angles. If one interior angle becomes too small, for example, less than... Or too large, for example, greater than If the triangle is of poor quality, it is considered to need optimization. One optimization operation is edge flipping, checking for triangles that share the longest side with the poor-quality triangle (e.g., ...). The adjacent triangles of the side are assumed to be By deleting shared edges and create a new connection. This forms two new triangles. and The interior angles of the two new triangles are recalculated. If the minimum angle has increased compared to before, the flip is accepted; otherwise, it is undone. By performing this type of local topology adjustment in the affected area, the geometric structure distribution of the entire model is updated. Finally, after adjusting the coordinates of the marked nodes and recalculating and optimizing the subsequent mesh topology, a corrected 3D model with more accurate geometric features and better mesh quality is obtained.

[0041] Please see Figure 5 The specific steps of S4 are as follows: S401: Call the corrected 3D model, scan the coordinate set of boundary points of the damaged area, calculate the curvature radius difference between adjacent grid units to obtain the curvature change rate, extract Gaussian curvature feature parameters based on the change results, and generate a Gaussian curvature feature set; The corrected 3D model is then retrieved. First, a pre-defined, manually delineated area suspected of being damaged, such as a weathered area at the mortise and tenon joint of a memorial archway, is scanned for its boundary point coordinates. This scanning process involves sequentially extracting the coordinates of each mesh node forming the edge line along the region's edge. Assuming the coordinates are extracted sequentially up to the nodes... Then, the curvature radius difference between adjacent mesh cells is calculated, with nodes... For example, a vertex can belong to two adjacent mesh triangles. and First calculate exist The curvature at that point, through analysis A quadratic surface is fitted with the coordinates of its annular neighboring points, and the principal curvature is calculated from it. Assuming the calculated values ​​are... exist The principal curvature value at that point is The corresponding radius of curvature is Calculate using the same method exist The curvature value at that point is The corresponding radius of curvature is The difference in radius of curvature is divided into The rate of change of curvature is the ratio of the difference value to the distance between the two points. Here, it is simplified to the change at the nodes, and the difference value is used directly. As a measure of the severity of change, Gaussian curvature feature parameters are extracted based on this change result. This process involves calculating the Gaussian curvature value for each point on the boundary. Its Gaussian curvature Through its surrounding triangular facets The calculation is based on the sum of the vertex angles of the points, assuming a surrounding area. The sum of the vertex angles of the five triangular facets is ,Right now The radius of the arc, and the area of ​​the local region formed by the surrounding surfaces is Then Gaussian curvature The calculated Gaussian curvature value As a node Feature parameters for boundary points Repeat the above calculations and associate each point with its corresponding Gaussian curvature value to generate a Gaussian curvature feature set.

[0042] S402: Based on the Gaussian curvature feature set, calculate the difference between the direction component of the boundary point normal vector and the direction cosine of the surface normal, perform direction difference normalization, accumulate the offset rate component, and generate the normal vector offset rate feature set. Based on the boundary point sequence contained in the generated Gaussian curvature feature set, for example Calculate the difference between the direction components of the boundary point normal vector and the direction cosine of the surface normal, using the nodes... For example, the surface normals, i.e., vertex normals, in correcting a 3D model are obtained by averaging the normal vectors of the triangles containing them. Let's assume the unit normal vector is... The direction component of the boundary point normal vector, defined here as the tangent direction of that point on the boundary line, is calculated... Its two neighboring points on the boundary and The coordinates are obtained by difference and average, assuming Coordinates are , Coordinates are Then the tangent vector is approximately: After normalization, it becomes Calculate the normal vector tangent vector The dot product, i.e., the direction cosine value, The direction cosine difference is Here, it is assumed that the tangent and normal of the ideal smooth boundary are orthogonal, and their dot product is... Then, directional difference normalization is performed, and this difference is... Mapping to the interval [0, 1], since the direction cosine value ranges from [-1, 1] and the difference value ranges from [0, 1], this value can be used directly. As a result of normalization, the next step is to accumulate the offset component, a process performed along the sequence of boundary points, starting with the first point. Its offset rate is the normalized difference value, assumed to be calculated as follows: For the second point Its cumulative offset rate is The cumulative offset plus Its own normalized difference, i.e. For the third point Assuming its normalized difference is Then its cumulative offset rate is Similarly, each boundary point and its corresponding cumulative offset value are stored to generate a normal vector offset feature set.

[0043] S403: Call the normal vector offset rate feature set, calculate the gradient sequence of the interior angles of adjacent triangular faces based on the node topology of the corrected 3D model, and input it into the support vector machine together with the Gaussian curvature feature set and the normal vector offset rate feature set to perform difference ratio clustering to obtain the 3D model of the damaged label. The generated normal vector offset feature set is called, and combined with the corrected 3D model, the gradient sequence of the interior angles of adjacent triangular faces is first calculated based on the node topology of the model, taking a node on the boundary of the damaged region as an example. For example, this node consists of two triangular faces. and The two triangles share a common vertex and an edge. ,calculate Center and edge Opposite interior angles Assuming , then calculate Center and edge Opposite interior angles Assuming Then the two triangular faces are on the common edge The gradient of the interior angle at point is , take this gradient value As a node One feature is then used to extract the interior angle gradient feature from the node set extracted from the Gaussian curvature feature set. Gaussian curvature value (e.g.) ), and nodes extracted from the normal vector offset feature set. The cumulative offset value (e.g.) These components, together, form a multi-dimensional feature vector, as shown in Table 4. This vector, containing multiple feature components, is then input into a pre-trained support vector machine classifier. This classifier has been built by learning from a large amount of sample data of ancient architectural components labeled as "damaged" or "intact." For example, when the input node... eigenvectors In this process, the Support Vector Machine (SVM) calculates the position of the feature vector in the feature space using its internal decision function and determines whether it belongs to the "damaged" or "intact" category. This process is called difference ratio clustering, where "difference ratio" refers to the multiple input features themselves, and "clustering" here specifically manifests as a binary classification operation, grouping points with similar damage features into one class. If the SVM outputs "damaged," then the nodes are classified in the model. The nodes are marked, for example, by assigning them a red attribute. This feature extraction and classification process is repeated for nodes in the detection area. Finally, the nodes classified as "damaged" and their triangular facets are highlighted to obtain a 3D model of the damaged identifier.

[0044] Table 4: Example of Input Features for Support Vector Machines

[0045] As shown in Table 4, this table displays the feature vectors of the three nodes prepared for the input support vector machine. Each vector consists of three components: interior angle gradient, Gaussian curvature, and cumulative offset of the normal vector.

[0046] Please see Figure 6The specific steps of S5 are as follows: S501: Call the Gaussian curvature of the damaged boundary, the normal vector offset rate and the interior angle gradient data in the 3D model of the damaged identifier, perform matching and normalization of the component parameters on the node coordinate index, and aggregate the processing results into a 3D tensor unit set according to the node sequence to generate a damaged feature tensor set. The generated 3D model of the damaged boundary uses Gaussian curvature, normal vector offset, and interior angle gradient data. First, for a boundary node marked as "damaged," such as node P at the mortise and tenon joint of a bracket set, its coordinate index is used to match and retrieve its corresponding three feature data, specifically the Gaussian curvature value. Cumulative offset of normal vector and interior angle gradient Next, the three component parameters are normalized. This process maps data of different dimensions and ranges to a unified interval. Taking the Gaussian curvature component as an example, the nodes on the damaged boundary are traversed first, and the maximum value of the Gaussian curvature is statistically obtained. The minimum value is The normalized Gaussian curvature value of node P is... Similarly, assume the global range of the normal vector offset rate is... Then the normalized value of this term for node P is For the interior angle gradient, assume its global range is... Then the normalized value of this term for node P is These three dimensionless results were processed. The nodes are aggregated into a three-dimensional vector, which is the three-dimensional tensor unit of node P. This vector comprehensively reflects the degree of geometric and topological anomalies of the point. For nodes on the damaged boundary, the nodes are arranged according to their connection order on the boundary, for example... The above data matching, component normalization and aggregation operations are performed one by one, and all the obtained three-dimensional tensor units are organized according to the node sequence to generate a set of damaged feature tensors.

[0047] S502: Based on the damaged feature tensor set, extract the principal direction component of the tensor unit, perform weighted integral calculation on the coupling parameters between the Gaussian curvature component, the normal offset component and the interior angle gradient component, and generate a tensor coupling strength set. Based on the generated residual feature tensor set, this dataset provides a tensor unit containing three normalized components for each boundary point. First, the principal direction component of each tensor unit is extracted. This step is to determine which feature is most "prominent" at the current node. For the tensor unit of node P... By comparing the values ​​of the three components, the value of the third component (the interior angle gradient component) was found to be... The interior angle gradient is the largest, therefore, it is determined as the principal direction component of node P. Next, weighted integration is performed on the coupling parameters between the Gaussian curvature component, the normal offset component, and the interior angle gradient component. This calculation aims to fuse three discrete features into a single comprehensive index, namely the coupling strength. Weights need to be assigned to the three components. The weight settings refer to historical data analysis of the damage patterns of wooden ancient building structures. Topological fracture (corresponding to the interior angle gradient) has the highest direct correlation, followed by drastic changes in the surface normal (corresponding to the normal offset), while surface curvature (corresponding to Gaussian curvature) has the lowest correlation. Therefore, weights are set as shown in Table 5, and weighted integration is performed, i.e., a weighted summation is performed on the tensor elements of each node. For node P, its tensor coupling strength is... This weighted integral calculation is repeated for each tensor unit in the damaged feature tensor set, thereby calculating a corresponding coupling strength value for each boundary node and generating a tensor coupling strength set.

[0048] Table 5: Feature Component Weight Configuration Table

[0049] As shown in Table 5, the table lists the weights assigned to the three core feature components when calculating the tensor coupling strength, with a total weight of 1.0.

[0050] S503: Based on the tensor coupling strength set, compare the coupling strength value of the damaged fragment with the damage state threshold, record the unit index above the threshold and mark the damage warning, map it to the three-dimensional coordinate field, and form a three-dimensional monitoring model of ancient buildings; Based on the generated tensor coupling strength set, which contains the coupling strength value of each node along the damaged boundary of the archway brackets, the coupling strength value of each damaged segment is first compared with a preset damage state threshold. This threshold is set based on retrospective analysis of digital models of a large number of similar ancient architectural components that have undergone structural damage. Statistical analysis shows that when the coupling strength value exceeds a certain critical point, the probability of accelerated deterioration or fracture of the structure increases significantly. Therefore, the damage state threshold is set as follows: Taking node P as an example, its calculated coupling strength value is ,because Below the threshold The node was not marked as being in a warning state. Let's take another node Q, and assume its calculated coupling strength value is... ,because Above the threshold The node's cell index is recorded, and a damage warning is assigned to it. This is achieved by setting the warning status field of the node to "True" in the model data structure. Next, the node marked with a warning status is mapped to the 3D coordinate field. This involves retrieving the node's (e.g., node Q) precise 3D coordinates in the corrected 3D model and assigning special visual effects to the coordinates during visualization rendering, such as rendering them as a bright red. Simultaneously, the warning level and the node's coupling strength value are also considered. Information such as these are attached as attributes to the coordinate point. When a user queries the point in the software, relevant warning information will pop up. In this way, abstract risk assessment data is closely linked with the actual building space location to form a three-dimensional monitoring model of ancient buildings.

[0051] Please see Figure 7 A 3D modeling-based ancient building monitoring and scanning system includes: The point cloud processing module acquires point cloud data of the ancient building surface through 3D laser scanning, performs spatial adjacency grouping processing, calculates the weighted combination of the normal vector angle and the radius of curvature of the points, generates a curvature difference sequence, and transmits it to the spatial curvature calculation module. The spatial curvature calculation module calls the curvature difference sequence, obtains the normal vector deviation angle for pair matching, inputs it into the principal component analysis model to perform direction sorting, extracts the curvature gradient direction change rate, generates a three-dimensional mesh model based on point cloud coordinates, and passes it to the direction feature extraction module. The orientation feature extraction module calls the 3D mesh model, extracts the spatial position tensor of the mesh nodes, and performs node coordinate rearrangement when the ratio of the rate of change to the position tensor exceeds a preset threshold. It then performs geometric correction on the 3D mesh model, generates a corrected 3D model, and passes it to the geometric correction module. The geometric correction module calls the corrected 3D model, extracts the Gaussian curvature, normal vector offset rate and interior angle gradient of the boundary points of the damaged area, inputs them into the support vector machine model to calculate the difference ratio, identifies the location of the damaged area, generates a 3D model of damage identification, and transfers it to the damage monitoring module. The damage monitoring module calls the 3D model of the damage marker, performs tensor coupling operations on the Gaussian curvature, normal vector offset, and interior angle gradient of the damage boundary, assesses the monitoring status of the damage area and marks it for early warning, and outputs a 3D monitoring model of the ancient building.

[0052] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for monitoring scanning of an ancient building based on three-dimensional modeling, characterized by, Includes the following steps: S1: Obtain point cloud data of the ancient building surface through 3D laser scanning, perform spatial adjacency grouping processing, calculate the weighted combination of the normal vector angle and the radius of curvature of the points, and generate a curvature difference sequence; S2: Call the curvature difference sequence, obtain the normal vector deviation angle for pair matching, input it into the principal component analysis model to perform direction sorting, extract the curvature gradient direction change rate, and generate a three-dimensional mesh model based on point cloud coordinates; S3: Call the three-dimensional mesh model, extract the spatial position tensor gradient of the mesh nodes, and when the ratio of the rate of change to the position tensor gradient exceeds a preset threshold, perform node coordinate rearrangement, perform geometric correction on the three-dimensional mesh model, and generate a corrected three-dimensional model. S4: Call the corrected 3D model, extract the Gaussian curvature, normal vector offset rate and interior angle gradient of the boundary points of the damaged area, input them into the support vector machine model to calculate the difference ratio, mark the location of the damaged area, and generate a 3D model of the damaged area. S5: Call the three-dimensional model of the damaged marker, perform tensor coupling operation on the Gaussian curvature of the damaged boundary, the normal vector offset rate, and the interior angle gradient, conduct monitoring status assessment of the damaged area and mark early warning, and output the three-dimensional monitoring model of the ancient building.

2. The three-dimensional modeling-based ancient building monitoring scanning method according to claim 1, wherein, The curvature difference sequence includes curvature change amplitude, normal vector difference degree, and spatial continuity coefficient; the three-dimensional mesh model includes curvature gradient distribution, point cloud topology relationship, and mesh structure parameters; the corrected three-dimensional model includes geometric correction coefficient, node rearrangement matrix, and spatial optimization parameters; the damage identification three-dimensional model includes damage area distribution, boundary feature parameters, and difference ratio mapping results; and the ancient building three-dimensional monitoring model includes damage status assessment indicators, risk warning information, and structural health monitoring parameters. 3.The three-dimensional modeling-based ancient building monitoring scanning method according to claim 1, wherein, The specific steps of S1 are as follows: S101: Obtain point cloud data of ancient building surface through 3D laser scanning, perform neighborhood distance calculation on the 3D coordinates of the points, aggregate points whose Euclidean distance is lower than the preset distance threshold, and remove isolated points whose number of neighborhood points is lower than the preset threshold, generating a cluster of spatially adjacent points; S102: Based on the spatial adjacent point cluster, extract the point normal vector, calculate the angle between the normal vectors of adjacent points, and extract the curvature radius change rate of the corresponding point. After normalizing the angle value and the curvature change rate, perform weighted summation to generate a normal curvature deviation dataset. S103: Based on the normal curvature deviation dataset, the deviation data of each point cluster is serialized and rearranged according to the point index order, and numerical difference operation is performed on adjacent deviation values ​​to calculate the change amplitude between continuous points and reorganize them according to the index order to generate a curvature difference sequence.

4. The three-dimensional modeling based ancient building monitoring scanning method according to claim 3, wherein, The specific steps of S2 are as follows: S201: Obtain the curvature difference sequence, calculate the curvature difference between adjacent sampling points and form a curvature direction vector set, extract the direction cosine parameter, compare it with the point cloud normal vector component to calculate the angle deviation, match the angle deviation data group according to the spatial index, and generate a normal vector deviation matching set. S202: Call the normal vector deviation matching set, extract the direction cosine parameter components and calculate the covariance, input the covariance result into the principal component analysis model to perform feature component decomposition, extract the direction sequence index according to the principal component weight, and obtain the direction sorting index sequence. S203: Sort the index sequence according to the direction, calculate the directional difference of the curvature gradient components of adjacent index nodes, extract the curvature gradient direction change rate, match the change rate with the point cloud coordinate index, and generate a three-dimensional mesh model according to the coordinate weight.

5. The ancient building monitoring and scanning method based on three-dimensional modeling according to claim 4, characterized in that, The specific steps for S3 are as follows: S301: Call the three-dimensional mesh model, retrieve the three-dimensional coordinate information of the mesh nodes, calculate the coordinate difference between adjacent nodes and normalize it to a unit length, construct the node position gradient tensor, and associate it with the curvature gradient direction change rate of the corresponding node to generate the node spatial position tensor gradient set. S302: Based on the set of spatial location tensor gradients of the nodes, extract the associated curvature gradient direction change rate, calculate the ratio of the change rate to the location tensor gradient and compare it with a preset threshold, record the index of nodes exceeding the threshold and extract the offset to obtain the set of node change differences. S303: Based on the node change difference set, rearrange the coordinate components of the marked nodes, adjust the spatial position of the nodes according to the offset, recalculate the mesh topology connection relationship and update the geometric structure distribution to obtain the corrected three-dimensional model.

6. The ancient building monitoring and scanning method based on three-dimensional modeling according to claim 5, characterized in that, The adjustment of node spatial position based on offset refers to scaling the offset vector according to a set ratio and superimposing it with the original node coordinates to complete the linear correction of the node spatial position.

7. The ancient building monitoring and scanning method based on three-dimensional modeling according to claim 5, characterized in that, The specific steps of S4 are as follows: S401: Call the corrected three-dimensional model, scan the coordinate set of boundary points of the damaged area, calculate the curvature radius difference of adjacent grid units to obtain the curvature change rate, extract Gaussian curvature feature parameters based on the change results, and generate a Gaussian curvature feature set. S402: Based on the Gaussian curvature feature set, calculate the difference between the direction component of the boundary point normal vector and the direction cosine of the surface normal, perform direction difference normalization, accumulate the offset rate component, and generate a normal vector offset rate feature set. S403: Call the normal vector offset rate feature set, calculate the gradient sequence of the interior angles of adjacent triangular faces based on the node topology of the corrected 3D model, and input it together with the Gaussian curvature feature set and the normal vector offset rate feature set into the support vector machine to perform difference ratio clustering to obtain the damaged 3D model.

8. The ancient building monitoring and scanning method based on three-dimensional modeling according to claim 7, characterized in that, The specific steps of S5 are as follows: S501: Call the Gaussian curvature of the damaged boundary, the normal vector offset rate and the interior angle gradient data in the three-dimensional model of the damaged identifier, perform matching and normalization of the component parameters on the node coordinate index, and aggregate the processing results into a three-dimensional tensor unit set according to the node sequence to generate a damaged feature tensor set. S502: Based on the damaged feature tensor set, extract the principal direction component of the tensor unit, perform weighted integral calculation on the coupling parameters between the Gaussian curvature component, the normal offset component and the interior angle gradient component, and generate a tensor coupling strength set. S503: Based on the tensor coupling strength set, compare the coupling strength value of the damaged fragment with the damage state threshold, record the unit index above the threshold and mark the damage warning, map it to the three-dimensional coordinate field, and form a three-dimensional monitoring model of ancient buildings.

9. The ancient building monitoring and scanning method based on three-dimensional modeling according to claim 8, characterized in that, The normalized component parameters refer to mapping the Gaussian curvature, normal vector offset, and interior angle gradient data to a unified standardized interval and performing boundary correction on outlier components that exceed the interval, so that the normalized components remain within the quantizable range and are aggregated according to the node sequence to form a three-dimensional tensor unit set. The damaged state threshold refers to a dynamic threshold set based on the statistical distribution results of the tensor coupling strength set, according to the mean square error range of the coupling strength values ​​of the damaged segments.

10. A monitoring and scanning system for ancient buildings based on 3D modeling, characterized in that, The system is used to implement the ancient building monitoring and scanning method based on three-dimensional modeling as described in any one of claims 1-9, and the system comprises: The point cloud processing module acquires point cloud data of the ancient building surface through 3D laser scanning, performs spatial adjacency grouping processing, calculates the deviation between the normal vector of the point in the group and the curvature radius of the adjacent point, generates a curvature difference sequence, and transmits it to the spatial curvature calculation module. The spatial curvature calculation module calls the curvature difference sequence, obtains the normal vector deviation angle for pair matching, inputs it into the principal component analysis model to perform direction sorting, extracts the curvature gradient direction change rate, generates a three-dimensional mesh model based on point cloud coordinates, and passes it to the direction feature extraction module. The orientation feature extraction module calls the three-dimensional mesh model to extract the spatial position tensor of the mesh nodes. When the difference between the rate of change and the position tensor exceeds a preset threshold, the node coordinates are rearranged to perform geometric correction on the three-dimensional mesh model, generate a corrected three-dimensional model, and pass it to the geometric correction module. The geometric correction module calls the corrected 3D model, extracts the Gaussian curvature, normal vector offset rate and interior angle gradient of the boundary points of the damaged area, inputs them into the support vector machine model to calculate the difference ratio, identifies the location of the damaged area, generates a 3D model of damage identification, and transmits it to the damage monitoring module. The damage monitoring module calls the three-dimensional model of the damage marker, performs tensor coupling operations on the Gaussian curvature, normal vector offset, and interior angle gradient of the damage boundary, evaluates the monitoring status of the damage area and marks it for early warning, and outputs the three-dimensional monitoring model of the ancient building.

Images