Aerial image-based historic site three-dimensional monitoring data processing method

Abstract

The invention relates to the technical field of image processing, in particular to an ancient site three-dimensional monitoring data processing method based on aerial images. The method comprises the following steps: acquiring three-dimensional point cloud data of an ancient site; processing the three-dimensional point cloud data by using a clustering algorithm to obtain a plurality of target clusters; the plurality of target clusters correspond to a plurality of structure regions of the historic site; calculating the damage scores of all angular points in the structure area by calculating the difference value of the normal vector included angles of the fitting curved surfaces where the angular points and the symmetric points are located, and finally evaluating the overall damage condition of the historic site according to the damage proportion of the angular points. The damage score is determined by calculating the difference value of the normal vector included angles of the fitting curved surfaces where the angular points and the symmetric points of the angular points are located, the symmetric characteristic of the historic site structure is effectively utilized, tiny damage is converted into quantized data, structural deformation caused by natural or human factors can be sensitively captured, misjudgment of a curvature complex area and a damage area is avoided, and the accuracy of damage is improved. And the accuracy and reliability of damage identification are improved.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a method for processing three-dimensional monitoring data of ancient monuments based on aerial images. Background Art

[0002] As important carriers of human history and culture, historical monuments carry a wealth of historical information and cultural value. However, over the long years, they have been continuously subjected to natural erosion and human damage. Many have suffered varying degrees of damage, including cracks, peeling, and collapse. This damage not only affects the appearance and structural stability of the monuments, but can also lead to the loss of their historical and cultural value.

[0003] With the development of aerial photogrammetry and 3D reconstruction technology, the acquisition of 3D point cloud data of historical monuments provides powerful data support for in-depth analysis of their damage. Therefore, by analyzing 3D point cloud data, we can more intuitively and accurately understand the extent and location of damage to historical monuments, providing a scientific basis for decision-making on preservation and restoration.

[0004] However, in practical applications, accurately identifying damaged areas on monuments faces numerous challenges. For one thing, after natural disasters or human damage, monuments suffer complex and irregular surface deformations. Furthermore, the architectural structure of the monuments themselves contains areas with significant curvature variations, such as overhanging eaves, angled corners, and carved columns. These areas exhibit similar geometric characteristics to damaged areas in 3D point cloud data, making it difficult for traditional 3D point cloud-based damage identification methods to effectively distinguish damaged areas from areas with normal curvature variations, leading to misjudgments or omissions of damage. This not only leads to unnecessary waste in restoration costs, but also may delay the repair of truly damaged areas due to incorrect assessments, thus affecting the accuracy and effectiveness of monument protection and restoration efforts. Summary of the Invention

[0005] To solve the technical problem of difficulty in effectively distinguishing damaged areas from areas with normal curvature changes in the above-mentioned historical monument monitoring, the present invention provides a method for processing data for three-dimensional monitoring of historical monuments based on aerial images, the method comprising: Collect three-dimensional point cloud data of the monument; use a clustering algorithm to process the three-dimensional point cloud data to obtain multiple target clusters; the multiple target clusters correspond to multiple structural areas of the monument; calculate the damage scores of all corner points in any structural area, including: obtaining any corner point and its symmetrical point of the current structural area, when the symmetrical point is another corner point of the structural area, determine a first fitting surface according to the data points of the corner point and its neighborhood window, determine a second fitting surface according to the data points of the symmetrical point and its neighborhood window, calculate the difference between the angle between the normal vectors of any pair of points on the first fitting surface and the angle between the normal vectors of the symmetrical point pair on the second fitting surface, and mark it as an angle difference value, the damage score is equal to the sum of the angle difference values ​​of all point pairs on the first fitting surface; evaluate the damage condition of the monument according to the damage scores of all corner points of the multiple structural areas.

[0006] The present invention collects three-dimensional point cloud data and clusters and divides the structural areas to achieve accurate deconstruction of the monument structure. The damage score is determined by calculating the angle difference between the normal vectors of the fitted surface at the corner point and its symmetrical point, effectively utilizing the symmetrical characteristics of the monument structure to convert subtle damage into quantitative data. Compared with traditional methods, this method can keenly capture structural deformations caused by natural or human factors, avoid misjudgment of complex curvature areas and damaged areas, and improve the accuracy and reliability of damage identification. Finally, the damage scores of the corner points of each structural area are combined to evaluate the overall damage of the monument, providing scientific and accurate data support for the protection and restoration of the monument, and improving the efficiency of monument monitoring and the scientific nature of protection decisions.

[0007] As a further improvement of the method of the present invention, the angle difference value ;in, It is In the structural area The first pair of points on the fitting surface corresponding to the corner points The cosine similarity of the normal vector, is the first pair of points on the fitting surface The cosine similarity of the normal vectors of symmetrical point pairs on the second fitting surface, is the inverse cosine function.

[0008] This method accurately measures the difference in angles between normal vectors of fitted surfaces at corner points of monuments. Using cosine similarity as a quantification basis, it quantifies damage into numerical values, effectively capturing surface changes caused by damage. This method can more precisely distinguish between damaged and normal areas, reducing false positives and missed detections. It provides highly accurate quantitative metrics for monument damage assessment, enhancing the scientific nature and accuracy of monument damage identification.

[0009] As a further improvement of the method of the present invention, the method of obtaining any corner point and its symmetrical point in the current structural area includes: determining the coordinate point of any corner point in the structural area by a corner point detection algorithm; calculating the coordinate point of the symmetrical point of the corner point; ;in, It is In the structural area The coordinates of the symmetrical points of the corner points, It is The coordinates of the center point of the structure area, It is In the structural area The coordinates of the corner points.

[0010] Corner points are precisely located using a corner detection algorithm, providing a reliable foundation for subsequent analysis. Specific formulas are used to calculate the positions of symmetrical points, enabling rapid and accurate determination based on the center point of the structural region. This process improves the efficiency and accuracy of obtaining the positions of corner points and their symmetrical points, providing accurate data for subsequent operations such as calculating damage scores based on corner and symmetrical points, thereby enhancing the accuracy and reliability of monument damage assessments.

[0011] As a further improvement to the method of the present invention, when the symmetrical point is not another corner point of the structural region, the damage score is set equal to 1.

[0012] This method simplifies the handling of special cases, quickly determines the damage score in atypical symmetrical point scenarios, avoids complex calculations, improves the efficiency of damage assessment, and ensures the consistency of the assessment process.

[0013] As a further improvement of the method of the present invention, the three-dimensional point cloud data is processed using a clustering algorithm to obtain multiple target clusters, including: constructing multiple initial clusters based on the three-dimensional point cloud data; applying an agglomerative hierarchical clustering algorithm to the multiple initial clusters to obtain multiple basic clusters; and allocating data points in the three-dimensional point cloud data that are not selected as initial clusters to corresponding basic clusters to obtain target clusters.

[0014] In the processing of three-dimensional monitoring data of historical sites, the present invention constructs initial clusters and applies an agglomerative hierarchical clustering algorithm to effectively aggregate similar data, reduce the complexity of three-dimensional point cloud data, and highlight the structural characteristics of the historical sites. The remaining data points are subsequently allocated and the target clusters are precisely divided, achieving detailed distinctions between the structural areas of the historical sites, significantly improving data processing efficiency and laying an accurate and orderly data foundation for damage assessment. At the same time, to address the problem that hierarchical clustering is prone to poor clustering effects at the intersection of different structural areas due to differences in point cloud features, by analyzing the differences between the point cloud data and its local features, initial clusters that affect aggregation are eliminated, further improving the accuracy of the division, ensuring that the clustering results are more consistent with the actual structure of the historical sites, and significantly enhancing the scientific nature and accuracy of historical site damage assessments.

[0015] As another improvement of the method of the present invention, the method of constructing multiple initial clusters based on the three-dimensional point cloud data includes: calculating the grayscale feature similarity between any data point in the point cloud data and the data points in its neighborhood; when the grayscale feature similarity of the data point is greater than its preset similarity threshold, the data point is used as the initial cluster; the grayscale feature similarity ;in, It is the first Similarity of grayscale features of data points; It is Gray value of data points; It is In the neighborhood of the data point Gray value of data points; is the number of data points in the neighborhood; is the absolute value symbol; is the natural exponential function, is the maximum function.

[0016] This method constructs initial clusters by calculating the grayscale feature similarity between a point cloud data point and its neighboring points. This method effectively selects data points with similar grayscale features as clustering starting points. This method uses a specific formula to accurately quantify similarity, avoiding the blindness of randomly selecting initial clusters. When the similarity is greater than a threshold, the initial cluster is determined, which can make the clustering more consistent with the intrinsic distribution characteristics of the data, improve the rationality and scientific nature of the initial cluster construction, and lay a good foundation for subsequent operations such as agglomerative hierarchical clustering, thereby improving the accuracy and reliability of the clustering division of the three-dimensional point cloud data of the ancient monuments.

[0017] As another improvement of the method of the present invention, the allocation of data points in the three-dimensional point cloud data that are not selected as the initial clustering cluster to the corresponding basic clustering cluster includes: calculating the Euclidean distance from any data point that is not selected as the initial clustering cluster to the center of all basic clustering clusters; and allocating the data points that are not selected as the initial clustering cluster to the basic clustering cluster with the closest distance.

[0018] By calculating the Euclidean distance from unselected data points to the center of the underlying cluster and assigning them to the nearest underlying cluster, we can improve clustering results in a simple and efficient manner. This distance-based allocation strategy ensures more reasonable data point attribution, enhances the accuracy and completeness of clustering, and facilitates more precise demarcation of historical structures, providing a well-organized data foundation for subsequent damage analysis.

[0019] As another improvement of the method of the present invention, the damage condition of the monument is evaluated based on the damage scores at all corner points of the multiple structural areas, including: when the damage score at the corner point is greater than a preset damage acceptability, marking the corner point; when the number of marked corner points accounts for more than a preset ratio of the total number of corner points of the monument, marking the monument as severely damaged.

[0020] The damage to monuments is assessed by comparing corner damage scores with pre-set standards. Corner points exceeding the threshold are marked, and then the proportion of marked corner points is used to determine whether the damage is severe. This method quantifies and standardizes damage assessments, enabling rapid and accurate determination of the overall extent of damage to monuments, providing a clear and intuitive basis for decision-making regarding monument preservation.

[0021] As another improvement of the method of the present invention, the determination of the first fitting surface and the second fitting surface includes: determining a mathematical model of the fitting surface, the mathematical model including a quadratic surface and a parametric surface; for the mathematical model, using the least squares method to calculate its parameters to obtain the fitting surface.

[0022] Mathematical models such as quadratic and parametric surfaces were selected, and the least squares method was used to calculate parameters and determine the fitted surface. This method provides a scientific means for analyzing the surface morphology near the corners of the monument. It can accurately fit the actual surface and lay a reliable geometric analysis foundation for accurately calculating the damage score of the corner points and assessing the damage situation.

[0023] As another improvement of the method of the present invention, the collecting of three-dimensional point cloud data of the monument includes: using an aerial photography device to capture an image of the monument; constructing a three-dimensional structure of the monument using an MVS algorithm to obtain three-dimensional point cloud data of the monument.

[0024] Using aerial photography equipment combined with the MVS algorithm to obtain 3D point cloud data of ancient monuments allows for rapid and comprehensive collection of spatial information about the monuments. This method captures and constructs 3D structures from multiple aerial angles, effectively covering the entire monument and obtaining high-precision point cloud data, providing a rich and accurate data source for subsequent damage analysis and other processes.

[0025] Beneficial effects of the present invention: The present invention fully considers the design characteristics and symmetry of the historic buildings, and uses specific algorithms and formulas to accurately calculate the position of the symmetrical points of each corner point. By analyzing the angular difference in the normal vector direction of the corner point's neighboring point cloud and its symmetrical point's neighboring point cloud on the fitting surface, the possibility of damage at each corner point can be quantified. This method can keenly capture subtle changes caused by damage, effectively distinguish actual damage from normal structural variations, avoid misjudgment and missed judgment, and greatly improve the accuracy of historic damage analysis. At the same time, the local grayscale feature similarity index is introduced, and high-quality initial clusters are selected based on a strict calculation and screening mechanism. This operation can avoid the adverse interference of low-similarity point clouds in the transition area on the clustering process, optimize the performance of the traditional AGNES algorithm in the division of complex historic structures, significantly improve the accuracy of clustering, and provide a more reliable data basis for subsequent historic damage assessment based on clustering results. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] Figure 1 The present invention provides a flowchart of a method for processing three-dimensional monitoring data of ancient monuments based on aerial images. DETAILED DESCRIPTION

[0027] The embodiment of the present invention provides a method for processing three-dimensional monitoring data of ancient monuments based on aerial images, such as Figure 1 As shown, the method includes steps S100 to S400: Step S100: Collect three-dimensional point cloud data of the historical site.

[0028] To elaborate, it is necessary to first use aerial photography equipment to capture images of the historic buildings. In order to establish the three-dimensional structure of the historic buildings, it is necessary to use aerial photography equipment to shoot or scan the historic buildings from multiple angles when collecting images to ensure that the details and structural information on the surface of the historic buildings can be obtained.

[0029] The MVS algorithm can recover the three-dimensional geometric information of a scene from multiple images with overlapping areas. In this invention, the MVS algorithm is used to construct the three-dimensional structure of the monument from the collected multi-angle images. The MVS algorithm will output the three-dimensional point cloud data of the monument.

[0030] Step S200: Process the three-dimensional point cloud data using a clustering algorithm to obtain a plurality of target clusters, where the plurality of target clusters correspond to a plurality of structural regions of the monument.

[0031] Specifically, the structure of a historic building often includes multiple elements, such as doors, windows, eaves, and columns, each with significantly different geometric and grayscale characteristics. If the entire 3D structure of a historic building is analyzed, the characteristics of these different structures will interfere with each other, making damage identification inaccurate. Therefore, it is necessary to segment the different structures of a historic building. In practice, clustering algorithms can be used to segment these different structures.

[0032] Taking the agglomerative hierarchical clustering algorithm as an example, the algorithm uses each feature point as an initial cluster and performs continuous iterative aggregation to obtain the target cluster. However, for the intersection of different structural areas, since the point cloud feature points of the intersection area are different from other local point cloud feature points, if the point cloud feature points of these intersection areas are aggregated as the initial cluster, the quality of the basic cluster obtained will be poor. Continuing iterative aggregation will continue to transfer low-similarity features to the target cluster obtained after aggregation, resulting in a poor clustering effect and an inability to accurately correspond to the actual different structural areas of the monument. Therefore, the present invention improves the clustering algorithm's division of the monument's structural areas.

[0033] Specifically, first, calculate the grayscale feature similarity between any data point in the point cloud data and the data points in its neighborhood; grayscale feature similarity: ; in, It is the first Similarity of grayscale features of data points; It is Gray value of data points; It is In the neighborhood of the data point Gray value of data points; is the number of data points in the neighborhood; is the absolute value symbol; is the natural exponential function, is the maximum function.

[0034] In this formula, the neighborhood is centered on the data point and has a size of The window size can be set according to the situation and needs of the historic building during actual use. Indicates the difference in grayscale features between the current 3D point cloud and the point clouds in its neighborhood. The smaller the value, the closer the corresponding grayscale values ​​of the local point clouds of the current 3D point cloud are, that is, the more similar the local grayscale features of the current 3D point clouds are, which means that the local point clouds of the current 3D point cloud are more likely to be distributed in the same monument structure area; conversely, the more significant the difference in corresponding grayscale values ​​of the local point clouds of the current 3D point cloud is, that is, the local grayscale features of the current 3D point cloud are mixed, which means that the current 3D point cloud may be located in the structural transition area.

[0035] Secondly, when the grayscale feature similarity of a data point is greater than its preset similarity threshold, the data point is used as the initial cluster.

[0036] Historical buildings often have two or more structural transition regions. Within these transition regions, the local grayscale features corresponding to the 3D point cloud are more mixed. Using the 3D point cloud feature points in these structural regions as initial clusters can mislead the subsequent clustering process, mixing different structures and reducing clustering accuracy.

[0037] Therefore, to prevent initial clustering from selecting 3D point cloud data points in structural transition regions, we calculate the grayscale feature similarity of each data point and set a similarity threshold to select data points with a similarity greater than the threshold as initial clusters. This ensures that initial clusters are not located in structural transition regions. The threshold can be set to 0.6 or adjusted as needed.

[0038] Then, an agglomerative hierarchical clustering algorithm is applied to the multiple initial clusters to obtain multiple basic clusters.

[0039] After the initial cluster screening is completed, a traditional agglomerative hierarchical clustering algorithm such as the AGNES algorithm is applied to the initial clusters. This algorithm can merge multiple initial clusters to obtain multiple basic clusters.

[0040] Finally, the data points in the 3D point cloud data that are not selected as the initial clustering cluster are assigned to the corresponding basic clustering cluster to obtain the target clustering cluster.

[0041] After completing the above operations, if there are still data points whose grayscale feature similarity is less than the similarity threshold, these data points need to be assigned to the basic clusters. This way, all data points in the 3D point cloud data have corresponding clusters, and multiple target clusters are obtained based on the 3D point cloud data. Based on the above clustering, it can be seen that the multiple target clusters correspond to different structural areas of the historic building.

[0042] The specific data point assignment rule is: for each 3D point cloud data point that is not selected as an initial cluster, obtain its Euclidean distance to the center of all basic clusters, select the basic cluster with the closest distance, and merge them until each data point is assigned to the corresponding basic cluster. This assignment rule can be adjusted according to actual needs and is not limited to the closest distance.

[0043] Step S300: Calculate the damage scores of all corner points in any structural area.

[0044] Specifically, corner points are points where curvature changes significantly. Natural disasters or human damage can cause varying degrees of surface damage to historic buildings. Damaged areas are significantly deformed compared to surrounding areas, and the curvature changes significantly. Historic buildings often have curved contours, such as arches and fan-shaped windows and doors, where curvature changes significantly. Therefore, calculating damage based solely on corner points can lead to misidentification of damaged areas.

[0045] For historical buildings, the construction process takes into account factors such as mechanical stability and aesthetics. Each structure has a certain degree of symmetry, which means that the symmetrical position points corresponding to the corner points of the normal building outline in the same structural area are also corner points, thus providing a reliable basis for distinguishing damaged areas.

[0046] Therefore, to calculate the damage degree of a structural area, we can start from the corner points and their symmetrical points. For any structural area, the corner points and symmetrical points are obtained as follows: Specifically, first, all corner points and their coordinates in the structure area are obtained using a corner detection algorithm such as the Harris corner detection algorithm, the Shi-Tomasi corner detection algorithm, or the FAST feature detection algorithm. Then, the center point of the structure area is calculated using the mean of the coordinates of all 3D point cloud data in the structure area, and finally the coordinates of the symmetric points are calculated: ; in, It is In the structural area The coordinate points corresponding to the symmetrical points of the corner points are It is The corresponding coordinate points of the center points of the structure areas, It is In the structural area It should be noted that the points mentioned in this formula are all three-dimensional coordinate points.

[0047] According to the properties of symmetry, the symmetrical points and the corner points are symmetrical about the center point, that is, the center point is equal to the average position of the corner point and the symmetrical point. It reflects the reverse extension of the vector of the current corner point relative to the center point to obtain the coordinate point of its symmetrical point.

[0048] The above describes how to obtain the structural area, corner points, and symmetric points. Next, we will explain how to calculate the damage score at the corner points.

[0049] Regarding the calculation basis of the damage score, the distribution of the point cloud usually reflects the geometric shape of the structural area, and the direction of the normal vector of the local point cloud fitting surface is closely related to the distribution of the point cloud. If the point cloud shows obvious curvature in a certain direction, the direction of the normal vector of the fitting surface will be related to the curvature direction. The symmetrical structure of the building means that the contour curves of its design have similar geometric characteristics at relatively symmetrical positions, which means that the overall changes in the direction angles of the point cloud normal vectors at the two symmetrical positions are relatively consistent. Once the overall direction angles of the two symmetrical positions differ greatly, it may indicate that the current area has been affected by external forces, resulting in local changes in the geometric shape (such as bending, deformation), and it is considered that the current reference point is more likely to be damaged.

[0050] It can be seen from this that in order to calculate the damage score, it is also necessary to construct fitting surfaces for the corner points and their symmetrical points respectively. The fitting surface is to find a mathematically defined surface when a set of three-dimensional data points is given, so that it can best approximate or pass through these data points. The method of constructing the fitting surface is as follows: first determine the data points. The present invention uses the corner points or symmetrical points and multiple points closest to the corner points or symmetrical points as data points for fitting. Five points are preset, and the actual number can be adjusted according to needs. Then, the mathematical model of the fitting surface is determined, such as the quadratic surface and the parametric surface; finally, the least squares method is used to calculate the parameters of the mathematical model to obtain the fitting surface. The determination of the fitting surface is a prior art and will not be explained here.

[0051] The calculation formula for the specific damage score is: ; in, ; in, Indicates the In the structural area The damage score at each corner point is represents the standard normalization function, Indicates the In the structural area The degree of change in the normal vector direction angle within the corner point and its symmetrical point neighborhood, Indicates the In the structural area The coordinate points corresponding to the symmetrical points of the corner points are Indicates the The coordinate points corresponding to all corner points of the structure area, is the inverse cosine function, Represents the index values ​​of the two data points in the point pair on the first fitting surface, Indicates the total number of data points of the first fitting surface or the second fitting surface, It is In the structural area The first pair of points on the fitting surface corresponding to the corner points The cosine similarity of the normal vector, is the first pair of points on the fitting surface The cosine similarity of the normal vectors of symmetrical point pairs on the second fitting surface, Indicates the angular difference value.

[0052] It should be noted that the are any two points on the first fitting surface, and the symmetrical point pair is the point pair on the second fitting surface with respect to the first fitting surface. Two points of symmetry.

[0053] is the difference between the angles of the normal vectors of two points on the first fitted surface and the angles of the normal vectors of two points on the second fitted surface.

[0054] It indicates that the symmetrical point is not another corner point of the structural area, indicating that the symmetry of the monument structure is destroyed. The structural area where the current corner point is located is directly determined to be a damaged area, and the damaged score is set to 1. Indicates that the symmetry point is the other corner point of the structure area, It reflects the sum of the differences in the normal vector direction angles between any two 3D point cloud data points on the fitted surface corresponding to the current corner point and its symmetrical points. The larger the value, the worse the consistency of the normal vector direction angles in the neighborhood of the current corner point and its symmetrical points, which means that the symmetry of the current corner point is destroyed and there are significantly different bending conditions locally. It is believed that the possibility of damage at the current corner point is greater.

[0055] Repeat the above calculation to obtain the damage score of each corner point in the area.

[0056] Step S400: Evaluate the damage status of the monument based on the damage scores at all corner points of multiple structural regions.

[0057] To elaborate, the above calculation method can be used to calculate the damage scores of all corner points in all structural regions. Based on practical experience, a reasonable damage acceptability is set, with an empirical value of 0.7 in this example. If the damage score at the current corner point in different structural regions of the monument is greater than the damage acceptability, it indicates that the structural region where the current corner point is located has a high probability of damage and may have already experienced structural damage or deformation. The current corner point is marked.

[0058] If the total number of marked corner points on a historic building exceeds a preset ratio of 5% to the total number of corner points on the building, the building is considered severely damaged and may have affected the overall structural stability of the building. An alert is immediately issued, notifying conservation staff to initiate timely repairs. This preset ratio can be adjusted based on the building's condition and actual needs.

[0059] The above are all preferred embodiments of the present invention, and are not intended to limit the scope of protection of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the scope of protection of the present invention.

Claims

1. A method for processing three-dimensional monitoring data of ancient monuments based on aerial images, characterized in that: include: Collect 3D point cloud data of historical sites; Processing the three-dimensional point cloud data using a clustering algorithm to obtain a plurality of target clusters; The multiple target clusters correspond to multiple structural areas of the monument; Calculating the impairment scores of all corner points in any structural region, including: obtaining any corner point and its symmetrical point in the current structural region, when the symmetrical point is another corner point of the structural region, determining a first fitting surface based on the data points of the corner point and its neighborhood window, determining a second fitting surface based on the data points of the symmetrical point and its neighborhood window, calculating the difference between the angle between the normal vectors of any pair of points on the first fitting surface and the angle between the normal vectors of the symmetrical point pair on the second fitting surface, and recording the difference as the angle difference value. The impairment score is equal to the sum of the angle difference values ​​of all point pairs on the first fitting surface; The damage condition of the monument is evaluated according to the damage scores at all corner points of the multiple structural regions.

2. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: The angle difference value ; in, It is In the structural area The first pair of points on the fitting surface corresponding to the corner points The cosine similarity of the normal vector, is the first pair of points on the fitting surface The cosine similarity of the normal vectors of symmetrical point pairs on the second fitting surface, is the inverse cosine function.

3. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: The obtaining of any corner point and its symmetrical point in the current structural area includes: Determine the coordinates of any corner point of the structural area by using a corner detection algorithm; Calculate the coordinates of the symmetrical points of the corner points ; in, It is In the structural area The coordinates of the symmetrical points of the corner points, It is The coordinates of the center point of the structure area, It is In the structural area The coordinates of the corner points.

4. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: When the symmetrical point is not another corner point of the structural region, the damage score is set equal to 1.

5. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: The three-dimensional point cloud data is processed using a clustering algorithm to obtain a plurality of target clusters, including: Constructing a plurality of initial clusters based on the three-dimensional point cloud data; Applying an agglomerative hierarchical clustering algorithm to the multiple initial clusters to obtain multiple basic clusters; Allocate data points in the three-dimensional point cloud data that are not selected as initial clusters to corresponding basic clusters to obtain target clusters.

6. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 5, characterized in that: The constructing of a plurality of initial clusters based on the three-dimensional point cloud data comprises: Calculating the grayscale feature similarity between any data point in the point cloud data and the data points in its neighborhood; When the grayscale feature similarity of the data point is greater than a preset similarity threshold, the data point is used as an initial cluster; The grayscale feature similarity ; in, It is the first Similarity of grayscale features of data points; It is Gray value of data points; It is In the neighborhood of the data point Gray value of data points; is the number of data points in the neighborhood; is the absolute value symbol; is the natural exponential function, is the maximum function.

7. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 5, characterized in that: The allocating data points in the three-dimensional point cloud data that are not selected as initial clusters to corresponding basic clusters includes: Calculate the Euclidean distance between any data point that is not selected as the initial cluster and the centers of all basic clusters; The data points that are not selected as the initial cluster are assigned to the basic cluster with the closest distance.

8. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: The step of evaluating the damage condition of the monument according to the damage scores at all corner points of the plurality of structural regions includes: When the damage score at the corner point is greater than a preset damage acceptability, marking the corner point; When the ratio of the number of marked corner points to the total number of corner points of the monument is greater than a preset ratio, the monument is marked as severely damaged.

9. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: Determining the first fitting curved surface and the second fitting curved surface includes: Determining a mathematical model of the fitting surface, wherein the mathematical model includes a quadratic surface and a parametric surface; For the mathematical model, the least square method is used to calculate its parameters to obtain a fitting surface.

10. The method for processing data of three-dimensional monitoring of ancient monuments based on aerial images according to claim 1, characterized in that: The collecting of three-dimensional point cloud data of the historical site includes: Use aerial photography equipment to capture images of monuments; The three-dimensional structure of the monument is constructed using the MVS algorithm to obtain the three-dimensional point cloud data of the monument.

Images