An Automatic Identification Method and System for Mural Conservation and Restoration Based on Image Analysis

Abstract

This invention discloses an automatic identification method and system for mural protection and restoration based on image analysis. The method includes: generating the degradation state of the mural at different time points using an image simulation algorithm; converting two-dimensional damage images into a three-dimensional spatial model using 3D mapping technology to accurately pinpoint high-risk damage points; further integrating scattered damage points using a clustering analysis algorithm to extract the properties and spatial characteristics of damage groups; and automatically identifying damage boundaries by analyzing texture continuity and line direction using a curve feature detection algorithm. Dynamic sequence analysis tracks inter-frame differential changes to reveal the degradation propagation trajectory; encoding technology efficiently compresses and stores damage information; and historical comparison analysis uncovers matching degradation patterns and updates potential damage areas. This invention, with 3D spatial modeling as its core, integrates multi-dimensional analysis and efficient storage to achieve precise location, dynamic monitoring, and preventative protection of mural damage, significantly improving the scientific rigor and efficiency of cultural relic protection.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of cultural relic protection and restoration, and particularly relates to a mural protection and restoration area automatic identification method and system based on image analysis. BACKGROUND

[0002] As an important carrier of cultural heritage, murals carry the value of history and art, and their protection and restoration are directly related to the integrity of cultural heritage. Over time and environmental impact, murals often have cracks, paint peeling, discoloration and other problems, which need to be accurately identified to develop a repair plan. Image analysis technology has become a research hotspot in the field of mural protection due to its non-contact and high-precision characteristics. However, existing methods often struggle to balance comprehensiveness and refinement when dealing with complex damage scenarios, resulting in inaccurate repair area positioning and affecting protection effectiveness.

[0003] Current image analysis methods in mural protection rely mainly on single-time-point image data, making it difficult to capture the dynamic degradation process of the mural surface over time. For example, the expansion of cracks or the gradual peeling of paint often requires long-term monitoring to be discovered, but existing technologies lack effective means to identify such gradual changes. In addition, the complex texture and curve features of the mural surface increase the difficulty of analysis, and traditional methods struggle to accurately distinguish between natural texture and abnormal changes caused by damage. This makes it necessary for repair personnel to spend a lot of time manually checking damage areas in actual operations, which is inefficient and prone to missing potential problems. The identification of dynamic degradation processes is due to the complexity of the mural surface state. The interaction between material aging and environmental factors of murals leads to unpredictable spatial distribution and change trajectory of damage areas. For example, moisture can cause small cracks in the wall, which can expand over time and affect the stability of the surrounding paint. Existing technologies lack effective continuity analysis tools when dealing with such multi-time-point, multi-dimensional damage changes, making it difficult to accurately locate the spatial position and evolution trend of damage areas. In addition, the continuity analysis of the curved texture of the mural surface is also a major difficulty, as damage can cause texture breakage or deformation, but existing methods struggle to accurately distinguish between normal texture and the boundaries of damaged areas.

[0004] Therefore, how to use image analysis technology to achieve dynamic monitoring and accurate positioning of mural damage areas has become a key problem in the field of mural protection and restoration. This problem not only involves capturing the gradual changes in the mural surface, but also requires accurate identification of damage boundaries in complex texture backgrounds to ensure the relevance and efficiency of repair plans. SUMMARY

[0005] To solve the above technical problems, the present application provides a mural protection and restoration area automatic identification method and system based on image analysis. The mural protection and restoration area automatic identification method based on image analysis comprises:

[0006] Generate the degradation state images of the mural at different time nodes through the image simulation algorithm, obtain the contrast difference between the current mural state and the simulation image, and determine the distribution range and severity characteristics of the potential damage area;

[0007] Convert the determined potential damage area from a two-dimensional image to a three-dimensional space model using a three-dimensional mapping algorithm, obtain the spatial coordinates and depth information of the damage area, and mark it as a high-risk point if the spatial coordinate deviation exceeds the preset threshold to obtain a three-dimensional representation of the damage area;

[0008] Based on the three-dimensional representation of the damage area, the scattered damage points are processed by a clustering analysis algorithm to obtain the distribution of the clustered damage groups and determine the nature and spatial distribution characteristics of the damage groups;

[0009] According to the nature and spatial distribution characteristics of the damage groups, a curve feature detection algorithm is used to analyze the texture continuity and line direction of the mural surface, obtain the curve fracture or deformation position caused by damage, and identify the boundary range if the curve continuity is interrupted to obtain an automatic recognition result of the damage boundary.

[0010] According to the automatic recognition result, the mural images collected periodically are organized into a time sequence sequence through a dynamic sequence, the inter-frame difference change is obtained, and the expansion track and spread range of the progressive degradation process are determined.

[0011] According to the expansion track and spread range of the progressive degradation process, the identified damage area is encoded using an encoding technology, the compressed damage information data is obtained, and the encoding parameters are adjusted if the data compression rate is lower than the preset threshold to obtain a damage archive.

[0012] According to the damage archive, the past damage information is retrieved through historical comparison analysis to obtain a mode matching the current degradation process, and the update position of the potential damage area that needs to be prevented and repaired is determined.

[0013] Compared with the prior art, the present application has the following advantages and technical effects:

[0014] This invention discloses an intelligent monitoring and protection technology for mural degradation damage, addressing the difficulty in accurately identifying, locating, and preventing degradation damage caused by environmental factors during long-term preservation. The invention generates degradation states of murals at different time points using image simulation algorithms, and converts damaged areas from two-dimensional images into three-dimensional spatial models using a three-dimensional mapping algorithm, accurately acquiring the spatial coordinates and depth information of the damage and marking high-risk points. Furthermore, cluster analysis integrates scattered damage points to determine the nature and distribution characteristics of damage groups, and curve feature detection algorithms analyze texture continuity to automatically identify damage boundaries. Based on dynamic sequence analysis and inter-frame difference, the degradation propagation trajectory is tracked, and high-precision encoding technology compresses damage information to form an efficient storage archive. Historical comparative analysis matches degradation patterns and updates potential damage areas. This invention achieves precise location, dynamic monitoring, and preventative protection of mural damage, significantly improving the efficiency and accuracy of cultural heritage protection. Attached Figure Description

[0015] The accompanying drawings, which form part of this application, are used to provide a further understanding of this application. The illustrative embodiments and descriptions of this application are used to explain this application and do not constitute an undue limitation of this application. In the drawings:

[0016] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0017] Figure 2 This is a schematic diagram of the system structure according to an embodiment of the present invention. Detailed Implementation

[0018] It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other. This application will now be described in detail with reference to the accompanying drawings and embodiments.

[0019] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0020] Example 1

[0021] like Figure 1 As shown, this embodiment provides an automatic identification method for mural protection and restoration areas based on image analysis, including:

[0022] The degradation state images of the murals at different time points are generated by the image simulation algorithm. The difference between the current state of the murals and the simulated images is obtained to determine the distribution range and severity characteristics of the potential damage areas.

[0023] The determined potential damage area is converted from a two-dimensional image to a three-dimensional space model by using a three-dimensional mapping algorithm, spatial coordinates and depth information of the damage area are obtained, if the spatial coordinate deviation exceeds a preset threshold, it is marked as a high-risk point, and a three-dimensional representation of the damage area is obtained;

[0024] Based on the three-dimensional representation of the damage area, the scattered damage points are regionally processed by a clustering analysis algorithm, the distribution of the clustered damage groups is obtained, and the nature and spatial distribution characteristics of the damage groups are determined;

[0025] According to the nature and spatial distribution characteristics of the damage groups, a curve feature detection algorithm is used to analyze the texture continuity and line direction of the mural surface, the curve fracture or deformation position caused by damage is obtained, if the curve continuity is interrupted, it is marked as a boundary range, and an automatic recognition result of the damage boundary is obtained;

[0026] According to the automatic recognition result, the mural images collected periodically are organized into a time sequence by a dynamic sequence, the inter-frame difference change is obtained, and the extension trajectory and spread range of the progressive degradation process are determined;

[0027] According to the extension trajectory and spread range of the progressive degradation process, the identified damage area is encoded by using an encoding technology, the compressed damage information data is obtained, if the data compression rate is lower than a preset threshold, the encoding parameters are adjusted, and a damage archive is obtained;

[0028] According to the damage archive, the past damage information is searched by historical comparison and analysis, the mode matched with the current degradation process is obtained, and the update position of the potential damage area needing preventive protection and repair is determined.

[0029] Further, the process of determining the distribution range and severity characteristics of the potential damage area includes:

[0030] Generate the degradation state images of the mural at multiple time nodes by an image simulation algorithm;

[0031] An image generation model based on a convolutional neural network is used, the initial image of the mural and the time node parameters are input, and the simulation degradation image corresponding to the time node is output;

[0032] Degradation features are extracted from the simulation degradation image, if the pixel value difference between the simulation image and the initial image exceeds a preset threshold, it is marked as a degradation feature, and a degradation feature set is obtained;

[0033] Compare the current mural image with the simulation degradation image by an image comparison algorithm, calculate the pixel-level difference between the images by a structural similarity index algorithm, and determine the potential damage area;

[0034] The distribution range is extracted from the potential damage area, and an image segmentation algorithm is used to spatially cluster the damage area to obtain the boundary and range of the damage area;

[0035] The severity is analyzed based on the distribution range, the pixel value change range of each damaged area is calculated, and the severity level is determined by combining the area of ​​the region.

[0036] Damage detection results are extracted from severity levels, and areas with severity exceeding a preset threshold are marked to obtain a distribution map of potential damage areas.

[0037] By integrating damage detection results through state analysis and employing a weighted average method, combined with the distribution range and severity, a comprehensive assessment result of the mural degradation status is generated.

[0038] Specifically, this embodiment focuses on the simulation and detection of mural degradation. Based on a convolutional neural network image generation model, it employs a generative adversarial network architecture. Inputting an initial image of the mural and time parameters (e.g., 50 years, 100 years), it generates corresponding degradation images. The initial image is a high-resolution photograph of the mural, with pixel values ​​stored in RGB format. The model learns from historical mural degradation data to simulate features such as color fading and crack formation. The generated simulated images reflect the possible degradation states of the mural at different points in time, such as a 20% fading of color or a 0.5 mm increase in crack width. This method can predict future degradation trends, providing a basis for conservation measures.

[0039] In one possible implementation, this embodiment sets a pixel value difference threshold of 15% when extracting degradation features from a simulated degraded image. For example, if the pixel values ​​of a certain region in the initial image are [200, 150, 100], and the corresponding region in the simulated image is [170, 120, 80], and the difference exceeds the threshold, it is marked as a degradation feature, such as fading or peeling. The extracted feature set includes fading areas, crack locations, etc., and is stored as a feature vector. This feature extraction method facilitates the quantification of the degree of degradation and provides data support for subsequent analysis.

[0040] Specifically, the image comparison employs a structural similarity index algorithm to calculate the similarity between the current mural image and the simulated degraded image. For example, if the structural similarity index of a certain mural area is 0.85, which is below the preset threshold of 0.9, it indicates the presence of a potential damaged area. The algorithm locates the pixel range of the damaged area by comparing brightness, contrast, and structural differences, such as a region occupying 5% of the image area. This method can accurately identify the location of damage and improve detection efficiency.

[0041] In one embodiment, the image segmentation algorithm of this embodiment uses K-means clustering to spatially cluster potential damage regions.

[0042] For example, a damaged area containing 1000 pixels can be clustered into three sub-regions with clear boundaries, ranging from 300 to 400 pixels. The extracted distribution range can quantify the extent of damage spread, providing a basis for prioritizing repairs.

[0043] For example, when analyzing the severity of damage, the magnitude of pixel value change in each damaged area is calculated. If an area experiences an average change of 20% and occupies 2% of the image, it is classified as moderate damage. Combining the area and the magnitude of change, a severity level is assigned, such as mild, moderate, and severe. This grading method helps in developing targeted repair plans.

[0044] In one possible implementation, the damage detection results generate a distribution map by marking areas where the severity exceeds a threshold.

[0045] For example, severely damaged areas are marked in red, occupying 1% of the image area, and are located in the upper left corner of the mural. This distribution map visually shows the location of the damage, making it easier for restoration personnel to quickly locate it.

[0046] Specifically, the state analysis involved in this embodiment adopts a weighted average method, which takes into account both the distribution range and the severity.

[0047] For example, the weights for severe damage are 0.6, moderate damage is 0.3, and mild damage is 0.1, resulting in a comprehensive degradation index of 0.75, indicating that the murals are generally severely degraded. This assessment method integrates multi-dimensional information, providing a scientific basis for cultural relic protection.

[0048] Understandably, the above-mentioned technical process, through prediction, detection, and evaluation, forms a closed-loop analysis system that can effectively guide mural conservation work and extend the lifespan of cultural relics.

[0049] Furthermore, the process of obtaining a three-dimensional representation of the damaged area includes:

[0050] A three-dimensional mapping algorithm is used to extract potential damage areas from two-dimensional image data, generate an initial three-dimensional spatial model, and obtain a preliminary three-dimensional representation containing spatial coordinates and depth information.

[0051] Geometric correction is performed on the preliminary three-dimensional representation using spatial coordinate information, and the model accuracy is optimized using a stereoscopic microscopy algorithm to obtain the corrected three-dimensional spatial model.

[0052] Spatial coordinate information is extracted from the corrected 3D spatial model, and the deviation between each coordinate point and the reference point is calculated to obtain a set of deviation values.

[0053] If the deviation of any coordinate point in the deviation value set exceeds the preset threshold, it is marked as a high-risk point, and a high-risk point set is obtained.

[0054] Based on the set of high-risk points, the corrected 3D spatial model is segmented into regions, and the damage region is divided using a region growing algorithm to obtain a set of segmented damage regions.

[0055] The depth information acquisition technology is used to perform depth analysis on the segmented set of damaged regions, calculate the depth distribution of each region, and obtain the three-dimensional damaged region.

[0056] Based on the three-dimensional damage area, a three-dimensional region representation is generated, and the spatial coordinates and depth information are stored to obtain the final three-dimensional damage area model.

[0057] For example, in the field of mural damage detection, when using 3D mapping algorithms to extract potential damage areas from 2D image data, this embodiment uses stereo vision technology to convert 2D images into 3D point cloud data. Assuming there is an ancient mural, two high-resolution cameras are used to acquire image pairs from different angles during photography. Based on the principle of parallax, the spatial depth of each pixel is calculated, generating a preliminary 3D representation containing X, Y, and Z coordinates. This method can effectively capture the spatial features of minute undulations and cracks on the mural surface, helping to discover damage that is difficult to detect in planar images.

[0058] In one possible implementation, this embodiment optimizes the model accuracy through geometric correction after generating a preliminary three-dimensional spatial model.

[0059] For example, reference point cloud data acquired using a laser scanner can be registered with the preliminary model to correct for deformation caused by camera angle deviations. Suppose the preliminary model of a certain area on the mural surface shows a thickness of 2 mm, while the reference data is 1.8 mm; geometric correction can reduce the error to within 0.1 mm. This correction significantly improves the reliability of the model, providing accurate spatial information for subsequent analysis.

[0060] Specifically, this embodiment further optimizes the model accuracy using stereomicroscopy algorithms. For example, for fine cracks on the surface of a mural, the algorithm analyzes changes in crack width and depth by magnifying the depth information of a local area. Assuming the depth of a crack area changes from 0.5 mm to 1 mm, the algorithm can accurately pinpoint its boundaries. This high-precision analysis helps identify subtle features of the damage, providing an accurate basis for repair.

[0061] For example, when extracting spatial coordinate information from the corrected 3D model, the deviation of each coordinate point from the ideal plane is calculated. If the deviation of a coordinate point in a certain area is 0.3 mm, exceeding a preset threshold of 0.2 mm, it is marked as a high-risk point. By statistically analyzing the set of high-risk points, the distribution range of the damaged area can be preliminarily determined. This method can quickly locate potential damage, providing data support for subsequent segmentation.

[0062] In one possible implementation, this embodiment utilizes a region growing algorithm to segment the damaged region.

[0063] For example, based on a set of high-risk points, starting from the origin of a crack, the algorithm expands the region according to depth and color changes to delineate a complete damage boundary. Assuming a damage area is 10 square centimeters with clear boundaries, this segmentation can intuitively display the extent and shape of the damage, facilitating the design of repair plans.

[0064] Specifically, this embodiment analyzes the depth distribution of the segmented region using depth information acquisition technology.

[0065] For example, for a specific area of ​​spalling, the algorithm calculates the distribution of its depth from 0.2 mm to 1.5 mm, generating a precise three-dimensional labeling of the damaged area. This analysis can reveal the severity of the damage and provide a basis for prioritizing repair areas.

[0066] For example, when generating the final 3D damage area model, the spatial coordinates and depth information are stored in point cloud format, making it compatible with subsequent visualization tools. Suppose the model shows a region with significant depth variations, covering an area of ​​15 square centimeters, this region is marked as a high-priority repair area. This 3D model provides the repair team with an intuitive spatial reference, improving repair efficiency and accuracy.

[0067] Furthermore, the process of determining the nature and spatial distribution characteristics of lesion clusters includes:

[0068] Acquire three-dimensional representation data of the damaged area, and generate point cloud data using stereomicroscopy or three-dimensional scanning technology to obtain an initial set of damaged points;

[0069] The initial set of damage points is processed using the K-means clustering algorithm, and the spatial proximity between damage points is calculated based on Euclidean distance to obtain preliminary cluster groups.

[0070] If the density of damage points in the initial cluster is lower than a preset threshold, the DBSCAN algorithm is used to perform secondary clustering on the low-density areas to obtain the optimized damage clusters.

[0071] Based on the optimized damage groups, calculate the spatial bounding box and center point coordinates of each group to determine the spatial distribution characteristics of the groups;

[0072] Extract the geometric attributes of the group from its spatial distribution characteristics to determine the group's properties.

[0073] Principal component analysis algorithm is used to reduce the dimensionality of the property features, extract the main distribution patterns, and obtain the feature vector of the damage group;

[0074] Spatial correlation between groups is calculated using feature vectors. If the correlation is higher than a preset threshold, the related groups are merged to obtain the final distribution of damage groups.

[0075] For example, in this embodiment, when generating point cloud data using a stereomicroscope or 3D scanning technology, a high-resolution stereomicroscope is used to scan the surface of the mural, generating an initial set of damage points containing millions of points, each containing three-dimensional coordinate information. The stereomicroscope, through the principle of binocular parallax, captures the depth and location of minute cracks or corrosion points on the mural surface, generating high-precision point cloud data.

[0076] For example, scanning a mural with sides of 0.5 meters yields a point cloud of approximately 5 million points with a spacing of about 0.01 millimeters, ensuring the capture of subtle damage features.

[0077] In one possible implementation, this embodiment sets the number of clusters to 5 when processing the initial set of damage points based on the K-means clustering algorithm, and calculates the distance from each point to the cluster center based on Euclidean distance.

[0078] For example, the surface of a mural may contain cracks, corrosion pits, and wear areas. The K-means algorithm divides the point cloud into five initial groups, each representing a type of damage. Assuming a group contains 100,000 points with the center point at coordinates (100, 50, 20), through iterative optimization, the average distance between points within the group and the center is reduced to 0.05 millimeters, improving clustering accuracy.

[0079] For example, in this embodiment, when using the DBSCAN algorithm for secondary clustering in low-density areas, the neighborhood radius is set to 0.02 mm and the minimum number of points is 50. Low-density areas may correspond to the edge points of micro-cracks. DBSCAN identifies these points through density connectivity, generating more refined damage groups.

[0080] For example, a low-density region containing 2000 points is divided into 3 subgroups after DBSCAN processing, each corresponding to a different direction of crack propagation, thus avoiding misjudging noise points as damage points.

[0081] In one possible implementation, this embodiment calculates the spatial bounding box and center point coordinates of the optimized group by using the maximum and minimum values ​​of the point cloud coordinates to determine the bounding box.

[0082] For example, the bounding box of a certain damage group has a range of 10-12 mm on the x-axis, 5-6 mm on the y-axis, and 0-0.5 mm on the z-axis, with the center point coordinates being (11, 5.5, 0.25). This information reflects the spatial distribution of damage on the mural surface, facilitating subsequent analysis.

[0083] For example, when extracting the geometric properties of a group, if the volume of a group is calculated to be 2 cubic millimeters, the surface area to be 15 square millimeters, and the shape factor to be 0.8, it indicates that the damage is a flat corrosion pit rather than a deep crack. The shape factor, calculated as the ratio of surface area to volume, reflects the morphological characteristics of the damage and helps distinguish between different types of damage.

[0084] In one possible implementation, this embodiment uses principal component analysis to reduce the dimensionality of property features, compressing high-dimensional features such as volume, surface area, and shape factor into two-dimensional feature vectors.

[0085] For example, the feature vector of a certain group is (1.2, 0.3), which reflects its main distribution pattern and facilitates subsequent correlation analysis. After dimensionality reduction, the data dimensionality is reduced, and the computational efficiency is improved.

[0086] For example, in this embodiment, when calculating the spatial correlation between groups, the Pearson correlation coefficient is used, and the threshold is set to 0.9. If the correlation coefficient between two groups is 0.95, it indicates that they may belong to the extension of the same crack, and they are then merged into one group.

[0087] For example, the merged group contains 150,000 points, covering an area of ​​5 square centimeters on the mural surface, reflecting a large area of ​​continuous damage.

[0088] Understandably, the final damage cluster distribution is used for mural maintenance decisions.

[0089] For example, the distribution of damage clusters in a mural shows that the center point is close to the seam of the mural, indicating that cracks may have appeared and should be repaired first. This analysis, through progressive processing from point cloud to cluster, ensures accurate identification and classification of damaged areas.

[0090] Furthermore, the process of obtaining the automatic identification results of the damage boundary includes:

[0091] Acquire an image of the mural surface and generate the first image data;

[0092] The Canny edge detection algorithm is used to process the first image data and extract curve features to obtain the second image data;

[0093] Based on the curve characteristics of the second image data, analyze the continuity of the curve. If the continuity is interrupted, record the coordinates of the interruption point to obtain a set of curve interruption locations.

[0094] The SIFT algorithm is used to detect the line direction of the second image data, determine whether the lines deviate from the preset trajectory, and obtain a set of curve deformation positions.

[0095] The set of curve interruption locations and the set of curve deformation locations are merged to generate a set of damage boundary points;

[0096] A region growing algorithm is used to process the damage boundary point set, connect the boundary points, and generate the damage boundary contour.

[0097] Based on the boundary contour data, an automatic identification result of the damaged boundary is generated.

[0098] For example, in this embodiment, when acquiring images of the mural surface using a high-resolution imaging device, an industrial-grade CMOS camera with a resolution of 50 megapixels is used, along with a macro lens, to ensure the capture of subtle textures on the mural surface. The imaging process must be conducted under a constant light source to avoid light and shadow interference, generating the first image data. The data is stored in RGB format with a resolution of 4000×3000 pixels, covering a 1-square-meter area of ​​the mural surface. This setup can clearly record minute cracks or faded areas on the mural surface, providing high-quality basic data for subsequent analysis.

[0099] In one possible implementation of this embodiment, when processing the first image data using the Canny edge detection algorithm, the image is first converted to grayscale, and then smoothed using Gaussian filtering to reduce noise interference. A low threshold of 50 and a high threshold of 150 are set to extract the edges of cracks or peeling areas on the mural surface, generating the second image data. The edge detection results are presented as a binary image, with crack line widths approximately 2-5 pixels. This method effectively highlights the contours of surface damage on the mural, facilitating subsequent feature analysis.

[0100] Specifically, in this embodiment, when analyzing the curve continuity of the second image data, the adjacency relationship of edge pixels is tracked to determine whether the curve is interrupted. If the pixel spacing of a certain curve segment exceeds 3 pixels, it is recorded as an interruption point, and a set of interruption point coordinates is generated.

[0101] For example, in an image of a mural, 10 breakpoints were detected, with coordinates distributed in the upper left corner of the image, indicating that there may be a large crack there. This analysis helps to accurately locate the damaged area.

[0102] For example, in this embodiment, when using the SIFT algorithm to detect the direction of lines, key points and descriptors are extracted to determine whether the lines deviate from a preset trajectory. The preset trajectory can be based on the geometric rules of the original pattern of the mural, such as straight lines or arcs. If the key point offset angle exceeds 15 degrees, it is recorded as a deformed position.

[0103] For example, in a floral pattern on a mural, five instances of line misalignment were detected, concentrated at the edges of the petals, indicating that deformation may have occurred due to material aging. This method can effectively identify pattern distortion caused by damage.

[0104] In one possible implementation, when merging the set of curve interruption locations and the set of curve deformation locations in this embodiment, spatial proximity analysis is used to group points with a distance of less than 10 pixels into the same damage region, generating a set of damage boundary points.

[0105] For example, after merging, three main damage areas were obtained, containing 20, 15, and 10 points respectively, distributed in different areas of the mural. This merging method can integrate scattered damage features to form a unified analytical object.

[0106] Specifically, in this embodiment, when using the region growing algorithm to connect the set of damaged boundary points, the boundary points are used as seed points, the growth threshold is set to 5 pixels, and neighboring points are connected to generate a complete contour.

[0107] For example, in a mural, the algorithm connects scattered crack points into three closed contours, each corresponding to a different damaged area. This method can generate continuous damage boundaries, facilitating subsequent automated processing.

[0108] For example, in this embodiment, when generating the automatic identification result of the damage boundary, the contour data is converted into a vector format, and the length and area of ​​each contour are recorded.

[0109] For example, a damaged area with a contour length of 200 pixels and an area of ​​approximately 5000 square pixels indicates a severely flaked area. This result can be directly used to formulate a mural restoration plan, improving the targeting and efficiency of the restoration process.

[0110] Furthermore, the process of determining the trajectory and extent of the progressive degradation process includes:

[0111] Acquire regularly collected mural images and generate a standardized first image sequence through preprocessing;

[0112] An edge detection algorithm is used to extract the damage boundaries of each frame in the first image sequence to obtain a set of boundary points;

[0113] For the boundary point set, it is organized into a dynamic sequence in chronological order, and the difference of boundary points between adjacent frames is calculated to obtain the changed region.

[0114] By analyzing the geometric features of the changed regions, a clustering algorithm is used to determine the expansion trajectory of the degradation process;

[0115] If the area of ​​the continuous frame change region of the extended trajectory exceeds a preset threshold, it is judged as significant degradation, and a boundary description of the degradation region is generated.

[0116] Based on the boundary description, an interpolation algorithm is used to calculate the spatial distribution of the degraded region and obtain the spread range;

[0117] By overlaying the temporal extent of the degradation process, a dynamic expansion path is generated.

[0118] For example, in this embodiment, when periodically acquiring images of the murals, a high-resolution camera is used under fixed lighting conditions to capture the entire mural once a month, ensuring that the image resolution reaches at least 30 million pixels to capture subtle damage details. In the preprocessing stage, the images undergo standardization processes such as grayscale conversion, noise filtering, and brightness equalization to generate a unified first image sequence.

[0119] For example, in monitoring an ancient temple mural, the acquired image sequence is filtered using Gaussian filtering to remove noise, generating a standardized grayscale image set for easier subsequent analysis. In the edge detection stage, this embodiment uses the Sobel algorithm to extract the damage boundaries of each frame. The Sobel algorithm highlights the edge features of the damaged area by calculating the image's grayscale gradient.

[0120] For example, regarding a crack in a mural, this embodiment uses the Sobel algorithm to detect abrupt changes in grayscale at the crack edge, forming a set of boundary points. Assuming 1000 boundary points are detected in a given frame, their coordinates are recorded to form point set data. Dynamic sequence analysis of the boundary point set is then performed, arranging the boundary points of each frame in chronological order and calculating the differences between adjacent frames.

[0121] For example, in a six-month image sequence, the coordinates of boundary points in adjacent frames are compared to identify newly added or moved boundary points, generating changed regions. Suppose the difference in boundary points between the third and fourth months shows a 2-centimeter increase in crack length, indicating damage propagation. In geometric feature analysis, this embodiment uses the K-means clustering algorithm to classify the changed regions and determine the propagation trajectory of the degradation process.

[0122] For example, clustering results might show that the boundary points of a certain area are concentrated and extend horizontally, indicating that cracks are spreading laterally along the surface of the mural. Assuming the areas of the changing regions in three consecutive frames are 5, 7, and 10 square centimeters respectively, exceeding a preset threshold of 8 square centimeters, it is judged as significant degradation.

[0123] Specifically, for the boundary description of the degraded region, this embodiment generates a closed contour through the geometric distribution of boundary points.

[0124] For example, the boundary points of a crack region are fitted to an elliptical profile using the least squares method to describe its shape and extent. The spatial distribution of the degraded region is then further calculated using a spline interpolation algorithm.

[0125] For example, for a crack, interpolation calculations show that its spread range expands from an initial 5 cm to 8 cm, covering a specific area of ​​the mural. In the temporal overlay stage, the spread ranges of each frame are superimposed to generate a dynamic expansion path of the degradation process.

[0126] For example, the crack in a mural shows a radial pattern of expansion from the center to the edge, with the path length increasing by about 1 centimeter per month. This dynamic path helps to visually demonstrate the degradation trend, facilitating the development of subsequent conservation measures.

[0127] For example, when monitoring a thousand-year-old mural, the above method can clearly track the process of cracks from initial tiny fissures to significant expansion, and the generated dynamic path provides a precise basis for restoration work.

[0128] Preferably, by combining image data from different time points, the analysis results can be further correlated with environmental factors such as humidity and temperature changes, providing a comprehensive reference for mural conservation.

[0129] Furthermore, the process of obtaining damage records includes:

[0130] By analyzing the progressive degradation process, data on the expansion trajectory and spread range are obtained to determine the boundary range of the degradation area;

[0131] The degraded region is encoded using an encoding technique to obtain initially compressed damage data;

[0132] If the compression ratio of the initially compressed damaged data is lower than a preset threshold, the optimized compressed data can be obtained by adjusting the encoding parameters.

[0133] An initial damage profile is generated based on the optimized compressed data;

[0134] By comparing the initial damage profile with the expansion trajectory data, it can be determined whether the profile completely covers the degraded area;

[0135] If the archive does not fully cover the degraded area, incremental coding technology is used to supplement the coding, resulting in an updated damaged archive;

[0136] Based on the updated damage profile, generate the final storage profile.

[0137] In one possible implementation, this embodiment analyzes the progressive degradation process by using periodically collected mural image data, combined with image processing techniques, to extract the expansion trajectory and spread range of the degradation area.

[0138] For example, in a conservation project for murals on an ancient building, images of the murals were collected over 12 consecutive months, each at a resolution of 3000×2000 pixels, generating a standardized image sequence. Edge detection technology was used to identify the boundary point sets of cracks and peeling areas on the mural surface, recording the boundary coordinates at each time point to form a dynamic boundary sequence. By comparing the boundary point sets of adjacent time points, the length of newly formed cracks and the area changes of peeling regions were calculated, yielding the expansion trajectory and spread range.

[0139] For example, a crack in a certain area extends from an initial 5 cm to 8 cm, and the area of ​​spread expands from 20 square centimeters to 35 square centimeters. This method facilitates the quantification of the degradation process and provides a data foundation for subsequent coding.

[0140] Specifically, in this embodiment, when using encoding technology to perform high-precision encoding of degraded regions, a region-based compression algorithm is selected.

[0141] For example, using region-segmentation-based JPEG2000 encoding, the boundary point set and pixel information of the degraded region are compressed into initial damaged data. If the initial compression ratio is lower than a preset threshold, such as a target compression ratio of 80% but only 60% is achieved, the encoding parameters are adjusted, such as increasing quantization accuracy or optimizing the segmentation region size, and the compressed data is regenerated.

[0142] For example, in one mural case, the initial compressed data occupied 10MB of storage space. After adjusting the parameters, the optimized compressed data was reduced to 6MB while retaining 90% of the damage details. This optimization improved storage efficiency and facilitated long-term preservation.

[0143] Preferably, in this embodiment, when generating a damage archive with high efficiency, the optimized compressed data is combined with metadata to form a structured archive file.

[0144] For example, the archive contains the boundary coordinates of the degraded areas, timestamps, and compressed image data, stored in a database, occupying approximately 5MB per frame. When comparing the damaged archive with the extended trajectory data, geometric matching is used to verify the archive's coverage integrity.

[0145] For example, when comparing the boundary of a degraded area recorded in an archive with trajectory data, it was found that 10% of the edge crack area was missing. In this case, incremental encoding technology is used to supplement the pixel information of the missing area and update the archive.

[0146] For example, incremental encoding adds 0.5MB of data to ensure that the archive fully covers all degraded areas.

[0147] In one embodiment, multi-resolution storage technology is used to further optimize the final, high-efficiency storage archive.

[0148] For example, combining high-resolution damage data with low-resolution preview data generates layered archives, facilitating rapid retrieval and viewing. In a mural conservation project, the final archive supports quickly locating the degradation status of a specific area in a given month; for instance, querying the expansion path of a crack over six months takes only seconds to load. This method improves the archive's usability, facilitating subsequent analysis and restoration decisions.

[0149] For example, this embodiment implements incremental coding supplementation through a dynamic update mechanism for efficient management. Suppose that five new cracks appear in a degraded area of ​​a mural within three months. Incremental coding only compresses the newly added areas, generating 0.3MB of supplementary data, which is then merged into the original archive. This method reduces the overhead of redundant coding, ensuring that the archive always reflects the latest degradation state.

[0150] Furthermore, the process of identifying the location of potential damaged areas requiring preventative protection and repair includes:

[0151] Historical damage data is obtained from damage archives, and database query technology is used to retrieve records related to the current degradation pattern to obtain a historical damage dataset;

[0152] For historical damage datasets, the k-means clustering algorithm is used to analyze the feature distribution of degradation patterns and obtain classification results for similar degradation processes;

[0153] Based on the classification results of similar degradation processes, information on the damaged areas in each category is obtained, and the distribution pattern of potential damaged areas is determined through statistical analysis.

[0154] If the overlap between the distribution pattern of potential damage areas and the current degradation mode exceeds a preset threshold, a decision tree algorithm is used to analyze the damage probability of the areas and obtain a list of high-risk areas.

[0155] Obtain the spatial location information of the high-risk area list, and determine the updated location of potential damage areas through geometric mapping technology;

[0156] Based on the update location, time series analysis is used to predict future degradation trends and obtain a priority ranking for prevention, protection and restoration.

[0157] The highest priority region is obtained from the priority ranking, and a region update scheme for prevention, protection and repair is generated through a spatial optimization algorithm.

[0158] For example, in this embodiment, when retrieving historical damage data from the damage archive, database query technology is used to perform precise retrieval targeting specific degradation patterns. The damage archive stores years of corrosion and crack data on the mural surface. The database may contain fields such as time, damage type, location, and environmental conditions. The query technology can use SQL statements to filter out corrosion records related to the current high temperature and high humidity environment.

[0159] For example, querying the corrosion data of murals over the past 5 years yields a historical damage dataset containing 1,000 records.

[0160] It should be noted that queries must ensure data integrity to avoid missing key records.

[0161] Specifically, this embodiment uses the k-means clustering algorithm to analyze the distribution of degradation pattern features in a historical damage dataset. k-means divides the data into several categories by calculating damage features such as area, depth, and expansion rate. Assuming the dataset is divided into three categories: minor corrosion, moderate cracking, and severe spalling, the clustering results show that 70% of the records belong to minor corrosion, 20% to moderate cracking, and 10% to severe spalling. This classification helps identify similar degradation processes.

[0162] In one embodiment, damage area information for each category is extracted based on the classification results, and distribution patterns are determined through statistical analysis. It is assumed that statistics show minor corrosion is mostly distributed in the upper and lower edges of the mural and is positively correlated with humidity. If the current degradation mode is corrosion under high temperature and high humidity conditions, and the overlap with the minor corrosion category reaches 80%, then the distribution pattern is considered highly correlated.

[0163] It should be noted that the overlap threshold can be set to 75% to ensure the reliability of the analysis.

[0164] For example, for degradation patterns with high overlap, this embodiment uses a decision tree algorithm to analyze the damage probability. The decision tree calculates the damage risk for each area based on features such as humidity, temperature, and material aging. For instance, if a mural in a certain area has a 90% probability of damage due to long-term exposure to high humidity, it is included in the high-risk area list. This list provides a priority basis for subsequent maintenance.

[0165] Specifically, the spatial location of high-risk areas can be determined using geometric mapping techniques. Assuming the mural surface is divided into a 100x100 cm grid, the mapping technique positions the high-risk area at grid coordinates (20,30) to (25,35). By analyzing historical data, the updated location may be extended to (30,40). This mapping ensures the precise positioning of the damaged area.

[0166] In one embodiment, time series analysis is used to predict future degradation trends. Based on the corrosion expansion rate over the past 5 years, it is predicted that high-risk areas may expand by 10 centimeters within the next year. Priority ranking shows that the edge areas of the mural are ranked first due to their rapid expansion rate.

[0167] It should be noted that time series analysis can be combined with the ARIMA model to enhance prediction accuracy.

[0168] For example, for the highest priority area, this embodiment uses a spatial optimization algorithm to generate a preventative protection plan. Assuming the algorithm optimization recommends adding an anti-corrosion coating to the area between (20,30) and (25,35) of the mural, and adjusting the maintenance cycle to once every 3 months, this plan significantly improves maintenance efficiency through precise positioning and priority ranking.

[0169] Example 2

[0170] like Figure 2 As shown, based on the same inventive concept, this embodiment also provides an automatic identification system for mural protection and restoration areas based on image analysis, including:

[0171] The mural degradation state simulation module is used to generate images of the mural degradation state at different time points through image simulation algorithms, obtain the comparison difference between the current mural state and the simulated image, and determine the distribution range and severity characteristics of potential damage areas.

[0172] The 3D mapping module for the damaged area is used to convert the identified potential damaged area from a 2D image into a 3D spatial model using a 3D mapping algorithm, obtain the spatial coordinates and depth information of the damaged area, and mark it as a high-risk point if the spatial coordinate deviation exceeds a preset threshold, thus obtaining a 3D representation of the damaged area.

[0173] The damage point clustering analysis module is used to process scattered damage points into clusters based on the three-dimensional representation of the damage region using a clustering analysis algorithm, obtain the distribution of the clustered damage groups, and determine the nature and spatial distribution characteristics of the damage groups.

[0174] The curve feature detection module is used to analyze the texture continuity and line direction of the mural surface based on the nature and spatial distribution characteristics of the damage group, and to obtain the location of curve breakage or deformation caused by damage. If the curve continuity is interrupted, it is marked as the boundary range, and the automatic identification result of the damage boundary is obtained.

[0175] The dynamic sequence analysis module is used to organize periodically collected mural images into a time sequence based on the automatic recognition results, obtain inter-frame differential changes, and determine the expansion trajectory and spread range of the progressive degradation process.

[0176] The damage information encoding module is used to encode the identified damage areas based on the expansion trajectory and spread range of the progressive degradation process, obtain compressed damage information data, and adjust the encoding parameters if the data compression rate is lower than a preset threshold to obtain a damage profile.

[0177] The historical comparison analysis module is used to retrieve past damage information based on damage records, obtain patterns that match the current degradation process, and determine the updated locations of potential damaged areas that require preventive protection and repair.

[0178] The automatic identification system for mural protection and restoration areas based on image analysis provided in this embodiment has all the advantages of the automatic identification method for mural protection and restoration areas based on image analysis provided in Embodiment 1.

[0179] Example 3

[0180] This embodiment also discloses a computer device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in Embodiment 1.

[0181] Example 4

[0182] This embodiment also discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in Embodiment 1.

[0183] Example 5

[0184] This embodiment also discloses a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in Embodiment 1.

[0185] The above are merely preferred embodiments of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. An automatic identification method for mural protection and restoration areas based on image analysis, characterized in that, include: The degradation state images of the murals at different time points are generated by the image simulation algorithm. The difference between the current state of the murals and the simulated images is obtained to determine the distribution range and severity characteristics of the potential damage areas. A three-dimensional mapping algorithm is used to convert the identified potential damage area from a two-dimensional image into a three-dimensional spatial model, obtain the spatial coordinates and depth information of the damage area, and mark it as a high-risk point if the spatial coordinate deviation exceeds a preset threshold, thus obtaining a three-dimensional representation of the damage area. Based on the three-dimensional representation of the damage region, the scattered damage points are processed by clustering analysis algorithm to obtain the distribution of the clustered damage groups and determine the nature and spatial distribution characteristics of the damage groups. Based on the nature and spatial distribution characteristics of the damage groups, a curve feature detection algorithm is used to analyze the texture continuity and line direction of the mural surface, obtain the location of curve breakage or deformation caused by damage, and determine that if the curve continuity is interrupted, it is marked as the boundary range, thus obtaining the automatic identification result of the damage boundary. Based on the automatic recognition results, the periodically collected mural images are organized into a time sequence through dynamic sequence organization, and the inter-frame differential changes are obtained to determine the expansion trajectory and spread range of the progressive degradation process. Based on the expansion trajectory and spread range of the progressive degradation process, the identified damage areas are encoded using coding technology to obtain compressed damage information data. If the data compression rate is lower than a preset threshold, the coding parameters are adjusted to obtain a damage profile. Based on the damage records, historical damage information is retrieved through historical comparative analysis to obtain patterns that match the current degradation process, and the updated locations of potential damaged areas requiring preventive protection and repair are determined.

2. The method according to claim 1, characterized in that, The process of determining the distribution and severity characteristics of potential injury areas includes: Images of the mural's degradation state at multiple time points were generated using image simulation algorithms; An image generation model based on convolutional neural networks is used. The initial image of the mural and the time node parameters are input, and the simulated degradation image of the corresponding time node is output. Degradation features are extracted from the simulated degraded image. If the pixel value of the simulated image differs from the pixel value of the initial image by more than a preset threshold, it is marked as a degradation feature, thus obtaining a set of degradation features. The current mural image is compared with the simulated degraded image by an image comparison algorithm. The structural similarity index algorithm is used to calculate the pixel-level differences between the images and determine the potential damage area. The distribution range is extracted from the potential damage area, and an image segmentation algorithm is used to spatially cluster the damage area to obtain the boundary and range of the damage area; The severity is analyzed based on the distribution range, the pixel value change range of each damaged area is calculated, and the severity level is determined by combining the area of ​​the region. Damage detection results are extracted from severity levels, and areas with severity exceeding a preset threshold are marked to obtain a distribution map of potential damage areas. By integrating damage detection results through state analysis and employing a weighted average method, combined with the distribution range and severity, a comprehensive assessment result of the mural degradation status is generated.

3. The method according to claim 1, characterized in that, The process of obtaining a three-dimensional representation of the damaged area includes: A three-dimensional mapping algorithm is used to extract potential damage areas from two-dimensional image data, generate an initial three-dimensional spatial model, and obtain a preliminary three-dimensional representation containing spatial coordinates and depth information. The preliminary three-dimensional representation is geometrically corrected using spatial coordinate information, and the model accuracy is optimized using a stereomicroscopic algorithm to obtain the corrected three-dimensional spatial model. Spatial coordinate information is extracted from the corrected three-dimensional spatial model, and the deviation between each coordinate point and the reference point is calculated to obtain a set of deviation values. If the deviation of any coordinate point in the deviation value set exceeds a preset threshold, it is marked as a high-risk point, thus obtaining a high-risk point set; Based on the set of high-risk points, the corrected three-dimensional spatial model is segmented into regions, and the damaged regions are divided using a region growing algorithm to obtain a set of segmented damaged regions. The depth of the segmented damage region set is analyzed by depth information acquisition technology, and the depth distribution of each region is calculated to obtain the three-dimensional damage region. Based on the three-dimensional damage region, a three-dimensional region representation is generated, and the spatial coordinates and depth information are stored to obtain the final three-dimensional damage region model.

4. The method according to claim 1, characterized in that, The process of determining the nature and spatial distribution characteristics of lesion clusters includes: Acquire three-dimensional representation data of the damaged area, and generate point cloud data using stereomicroscopy or three-dimensional scanning technology to obtain an initial set of damaged points; The initial set of damage points is processed using the K-means clustering algorithm, and the spatial proximity between damage points is calculated based on Euclidean distance to obtain preliminary cluster groups. If the density of damage points in the initial cluster group is lower than a preset threshold, the low-density region is clustered again using the DBSCAN algorithm to obtain an optimized damage group. Based on the optimized damage groups, calculate the spatial bounding box and center point coordinates of each group to determine the spatial distribution characteristics of the groups; Extract the geometric attributes of the groups from the spatial distribution characteristics to determine the properties of the groups; Principal component analysis was used to reduce the dimensionality of the properties and characteristics, extract the main distribution patterns, and obtain the feature vector of the damage group. The spatial correlation between groups is calculated using the feature vectors. If the correlation is higher than a preset threshold, the related groups are merged to obtain the final distribution of damage groups.

5. The method according to claim 1, characterized in that, The process of obtaining the automatic identification results of damage boundaries includes: Acquire an image of the mural surface and generate the first image data; The first image data is processed using the Canny edge detection algorithm to extract curve features, thereby obtaining the second image data; Based on the curve characteristics of the second image data, the continuity of the curve is analyzed. If the continuity is interrupted, the coordinates of the interruption point are recorded to obtain a set of curve interruption positions. The SIFT algorithm is used to detect the line direction of the second image data, determine whether the lines deviate from the preset trajectory, and obtain a set of curve deformation positions. The set of curve interruption locations and the set of curve deformation locations are merged to generate a set of damage boundary points; The damage boundary point set is processed using a region growing algorithm, and the boundary points are connected to generate the damage boundary contour. Based on the boundary contour data, an automatic identification result of the damaged boundary is generated.

6. The method according to claim 1, characterized in that, The process of determining the trajectory and extent of progressive degradation includes: Acquire regularly collected mural images and generate a standardized first image sequence through preprocessing; An edge detection algorithm is used to extract the damage boundaries of each frame in the first image sequence to obtain a set of boundary points; For the set of boundary points, they are organized into a dynamic sequence in chronological order, and the differences in boundary points between adjacent frames are calculated to obtain the changed regions. By analyzing the geometric features of the changed regions, a clustering algorithm is used to determine the expansion trajectory of the degradation process; If the area of ​​the continuous frame change region of the extended trajectory exceeds a preset threshold, it is judged as significant degradation, and a boundary description of the degradation region is generated. Based on the boundary description, an interpolation algorithm is used to calculate the spatial distribution of the degraded region to obtain the spread range; By overlaying the temporal extent of the degradation process, a dynamic expansion path is generated.

7. The method according to claim 1, characterized in that, The process of obtaining a damage record includes: By analyzing the progressive degradation process, data on the expansion trajectory and spread range are obtained to determine the boundary range of the degradation area; The degraded region is encoded using an encoding technique to obtain initially compressed damage data; If the compression ratio of the initially compressed damaged data is lower than a preset threshold, then optimized compressed data can be obtained by adjusting the encoding parameters. Based on the optimized compressed data, an initial damage profile is generated; By comparing the initial damage profile with the extended trajectory data, it is determined whether the profile completely covers the degraded area; If the archive does not fully cover the degraded area, incremental coding technology is used to supplement the coding, resulting in an updated damaged archive; Based on the updated damage profile, generate the final storage profile.

8. The method according to claim 1, characterized in that, The process of identifying potential damaged areas requiring preventative protection and repair includes: Historical damage data is obtained from the damage archive, and database query technology is used to retrieve records related to the current degradation mode to obtain a historical damage dataset; For the aforementioned historical damage dataset, the k-means clustering algorithm is used to analyze the feature distribution of degradation patterns and obtain classification results for similar degradation processes; Based on the classification results of the similar degradation processes, information on the damaged areas in each classification is obtained, and the distribution pattern of potential damaged areas is determined through statistical analysis. If the overlap between the distribution pattern of potential damage areas and the current degradation mode exceeds a preset threshold, a decision tree algorithm is used to analyze the damage probability of the areas and obtain a list of high-risk areas. Obtain the spatial location information of the high-risk area list, and determine the updated location of potential damage areas through geometric mapping technology; Based on the updated location, time series analysis is used to predict future degradation trends and obtain a priority ranking for prevention, protection and restoration. The highest priority region is obtained from the priority ranking, and a region update scheme for prevention, protection and repair is generated through a spatial optimization algorithm.

9. An automatic identification system for mural protection and restoration areas based on image analysis, characterized in that, include: The mural degradation state simulation module is used to generate images of the mural degradation state at different time points through image simulation algorithms, obtain the comparison difference between the current mural state and the simulated image, and determine the distribution range and severity characteristics of potential damage areas. The 3D mapping module for the damaged area is used to convert the identified potential damaged area from a 2D image into a 3D spatial model using a 3D mapping algorithm, obtain the spatial coordinates and depth information of the damaged area, and mark it as a high-risk point if the spatial coordinate deviation exceeds a preset threshold, thus obtaining a 3D representation of the damaged area. The damage point clustering analysis module is used to perform regional clustering processing on scattered damage points based on the three-dimensional representation of the damage region, obtain the distribution of clustered damage groups, and determine the nature and spatial distribution characteristics of the damage groups. The curve feature detection module is used to analyze the texture continuity and line direction of the mural surface based on the nature and spatial distribution characteristics of the damage group, and to obtain the location of curve breakage or deformation caused by damage. If the curve continuity is interrupted, it is marked as the boundary range, and the automatic identification result of the damage boundary is obtained. The dynamic sequence analysis module is used to organize periodically collected mural images into a time sequence based on the automatic recognition results, obtain inter-frame differential changes, and determine the expansion trajectory and spread range of the progressive degradation process. The damage information encoding module is used to encode the identified damage areas based on the expansion trajectory and spread range of the progressive degradation process, obtain compressed damage information data, and adjust the encoding parameters if the data compression rate is lower than a preset threshold to obtain a damage profile. The historical comparison analysis module is used to retrieve past damage information based on the damage file through historical comparison analysis, obtain patterns that match the current degradation process, and determine the updated locations of potential damage areas that require preventive protection and repair.

Images