Ancient building masonry wall weathering damage intelligent assessment method based on big data

Abstract

The invention discloses an ancient building masonry wall weathering damage intelligent assessment method based on big data. The method comprises the steps of establishing a multi-source data acquisition system, and integrating three-dimensional laser scanning, unmanned aerial vehicle aerial photography, panoramic photography and mobile crowd sensing technologies. Damage types such as weathering, saltpetering and green plant coverage are identified through a deep learning automatic classification algorithm, subjective errors of manual labeling are eliminated, labeling efficiency is improved by multiple times, mass data can be processed, and the method is suitable for large-scale ancient building group evaluation requirements; pixel coordinates, area and morphological characteristics of a damaged area are extracted based on computer vision, quantitative indexes such as weathering rate and saltpetering rate are accurately calculated, an objective basis is provided for the damage degree, and fuzziness of traditional experience judgment is avoided; historical repair records and environment monitoring data (such as temperature, humidity and pollution) are fused, a wall damage evolution model is constructed, the future damage trend is predicted, a preventive protection scheme is formulated in advance, and the service life of an ancient building is prolonged.

Description

Technical Field

[0001] This invention relates to the field of wall weathering damage assessment technology, specifically to an intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data. Background Technology

[0002] Ancient building walls, as important carriers of cultural heritage, are exposed to the natural environment for extended periods and commonly suffer from weathering, efflorescence, and biological erosion. Statistics show that over 60% of my country's ancient brick and stone buildings exhibit varying degrees of wall damage, with weathering-induced material flaking accounting for over 45% and efflorescence-induced salt crystallization accounting for 32%. Traditional assessment methods rely on manual inspections and experience-based judgment, which have the following drawbacks: 1. Manual inspection has a long cycle (usually once every 2-3 years), making it difficult to capture the dynamic evolution of damage; 2. Historical repair records and environmental monitoring data (such as temperature, humidity, and salt precipitation concentration) are usually stored independently, lacking spatiotemporal alignment and correlation analysis, resulting in a lack of quantitative basis for repair decisions; 3. Existing assessment reports are mostly presented in text or two-dimensional tables, which cannot intuitively show the spatial distribution of damage and future expansion trends, increasing the difficulty for decision-makers to understand. Summary of the Invention

[0003] To address the aforementioned technical problems, this paper provides an intelligent assessment method for weathering damage to ancient building brick and stone walls based on big data. This technical solution solves the problems mentioned in the background section.

[0004] To achieve the above objectives, the technical solution adopted by the present invention is as follows: In a first aspect of the present invention, a method for intelligent assessment of weathering damage to brick and stone walls of ancient buildings based on big data is provided, comprising: Establish a multi-source data acquisition system that integrates 3D laser scanning, drone aerial photography, panoramic photography and mobile crowd sensing technologies to acquire geometric and physical information and surface damage images of ancient building brick and stone walls. A weathering damage sample database was constructed, and the collected images were automatically classified and labeled using deep learning algorithms to form a standardized dataset containing typical damage types such as weathering, efflorescence, and vegetation cover. Develop a damage assessment model based on computer vision and image processing technology, extract pixel coordinates, area and morphological features of the damaged area, and calculate quantitative indicators such as weathering rate, efflorescence rate and green coverage rate. Generate a visual assessment report that integrates historical repair records and environmental monitoring data, and use digital twin technology to predict the evolution trend of wall damage and optimize repair plans.

[0005] Preferably, the establishment of the multi-source data acquisition system specifically includes: Use a handheld or fixed laser scanner to acquire point cloud data of the wall surface with an accuracy of no less than 0.1mm; A drone equipped with a high-resolution camera and an oblique photography system was used to capture images of the wall in a 720-degree full coverage. Temperature and humidity sensors, anemometers, and salt spray detection devices are integrated near the brick and stone walls of ancient buildings to collect physicochemical parameters of the environment in which the walls are located.

[0006] Preferably, the construction of the weathering damage sample database specifically includes: The original images were denoised, enhanced, and normalized to a resolution of 4096×4096 pixels. An improved YOLOv4 target detection algorithm, combined with K-means clustering analysis, is used to automatically identify weathered loose layers, efflorescence crystals, and areas covered by green vegetation. Based on the damage classification standards in the "Technical Specification for Repair of Exposed Walls of Historical Buildings", the severity and distribution range of each type of damage are marked, and the severity includes mild, moderate and severe. Enhanced samples are generated by rotating, flipping, and adjusting brightness, so that the database capacity is no less than 100,000 labeled images.

[0007] Preferably, the improved YOLOv4 target detection algorithm, combined with K-means clustering analysis, is used to automatically identify weathered loose layers, efflorescence crystals, and vegetation-covered areas, specifically including: Using CSPDarknet53 as the backbone network, a hybrid attention mechanism module is embedded in its residual blocks. This module consists of a channel attention submodule and a spatial attention submodule connected in series. The channel attention submodule generates channel weight vectors through global average pooling, while the spatial attention submodule generates spatial weight maps through convolutional layers with 3×3 kernels. The two are fused by element-wise multiplication and then multiplied with the original feature map to enhance the feature responses of weathered and loose layers, efflorescence crystals, and vegetation-covered areas. Cluster analysis of labeled damage samples was performed based on the K-means++ algorithm. Intersection over Union (IoU) was used as the distance metric to iteratively optimize the cluster centers. Considering the large differences in the scale of damage to ancient buildings, nine anchor frames were generated, with aspect ratios covering the range of [0.25, 0.5, 1, 2, 4] and area ranges covering [16×16, 32×32, 64×64] pixels, ensuring adaptability to both micro-weathering cracks and large-area efflorescence areas. The CIoU loss function is used to replace the original YOLOv4 IoU loss, and a normalized distance and aspect ratio consistency penalty term based on the minimum bounding rectangle (MBR) is introduced. The calculation formula is as follows: ; in, (.) represents the Euclidean distance. and , where are the center points of the predicted bounding box and the ground truth bounding box, respectively, and c is the length of the diagonal of the minimum bounding rectangle. These are the weighting coefficients. Aspect ratio consistency parameter; Meanwhile, Focal Loss is introduced to address the class imbalance problem, and its formula is: ; in, For category weights, To predict probabilities, The adjustment factor has a value of 2. Redundant convolutional channels in CSPDarknet53 were removed by channel pruning, with a pruning rate of 40%, while channels that contribute highly to the damage features were retained. 8-bit fixed-point quantization was used to convert floating-point weight parameters to integers, reducing model storage space.

[0008] Preferably, the calculation steps of the damage assessment model include: The GrabCut algorithm was used to segment damaged regions in an image, and contour coordinates were extracted using the OpenCV library. The porosity of the damaged regions was then calculated through pixel area statistics and morphological analysis. Crack density and the distribution density of crystals ; Based on the porosity of the damaged area Crack density and the distribution density of crystals The weathering rate was obtained. Alkali bloom rate A and green coverage rate Three major quantitative indicators; The damage levels are classified based on the three quantitative indicators mentioned above, which guides the prioritization of repairs.

[0009] Preferably, the porosity The calculation formula is: ; in, The pixel area of ​​the pore region. This represents the total pixel area of ​​the damaged region; The crack density The calculation formula is: ; in, The pixel length of the crack skeleton line is converted to the actual length through pixel size calibration. This represents the total pixel area of ​​the damaged region; The distribution density of the crystals The calculation formula is: ; in, The pixel area of ​​the crystalline region; This represents the total pixel area of ​​the wall surface.

[0010] Preferably, the weathering rate The calculation formula is: ; in, and Here are the weighting coefficients. =0.6, =0.4; This is a conversion factor for crack length units, such as from pixels to meters; The formula for calculating the alkali bloom rate A is: A ; The formula for calculating the green coverage rate V is: ; in, The pixel area represents the area of ​​the green plants. This represents the total pixel area of ​​the wall surface; In the green coverage rate V, the root system of the plants causes cracks in the wall. A correction term can be introduced, which would be: ; in, The factor representing the aggravation effect of vegetation on cracks. Crack density.

[0011] Preferably, the generation of the visual assessment report integrates historical repair records and environmental monitoring data, and uses digital twin technology to predict the evolution trend of wall damage and optimize repair plans, specifically including: Calculated weathering rate Alkali bloom rate A and green coverage rate Spatial distribution data were used to convert each parameter value into RGB color values ​​using the Jet color mapping algorithm, generating a three-dimensional damage heat map of the wall surface, and overlaying location labels in the CGCS2000 coordinate system. Based on the areas with weathering rate >25% or efflorescence rate >15% in the damage heat map, and in conjunction with the recommended matching materials in the "Standard for Materials for the Repair of Ancient Buildings", and based on the "Construction Engineering Quantity List Pricing Specification", the repair budget for a single area is estimated, forming a priority ranking table that includes damage level, recommended process and total budget. Driven by time-series data of damage scores predicted by LSTM neural network for the next 5 years, the dynamic process of crack propagation, crystal growth and root erosion of the wall surface is simulated by Unity3D engine, generating an MP4 format evolution video with a frame rate of 30fps, and the predicted year and damage score change value are marked in the key frame. Damage heat map data from historical assessment periods are retrieved and displayed on the same screen as the current heat map using transparency overlay technology. The reduction in weathering rate (ΔW) and efflorescence rate (ΔA) after repair are quantified intuitively through color differences, and comparative analysis charts containing statistical significance (p < 0.05) are generated.

[0012] In a second aspect of the invention, a smart assessment system for weathering damage to ancient building brick and stone walls based on big data is also provided, comprising: The acquisition module is used to establish a multi-source data acquisition system, integrating three-dimensional laser scanning, UAV aerial photography, panoramic photography and mobile crowd sensing technology to acquire geometric and physical information and surface damage images of the brick and stone walls of ancient buildings. The sample database construction module is used to construct a weathering damage sample database. It uses deep learning algorithms to automatically classify and label the collected images, forming a standardized dataset containing typical damage types such as weathering, efflorescence, and vegetation cover. The assessment module is used to develop a damage assessment model. Based on computer vision and image processing technology, it extracts the pixel coordinates, area and morphological features of the damaged area and calculates quantitative indicators such as weathering rate, efflorescence rate and green coverage rate. The assessment report generation module is used to generate a visual assessment report, which integrates historical repair records and environmental monitoring data, and uses digital twin technology to predict the evolution trend of wall damage and optimize repair plans.

[0013] In a third aspect of the invention, an electronic device is also provided. The electronic device includes at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of the first aspect of the invention.

[0014] Compared with existing technologies, this invention provides an intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data, which has the following beneficial effects: This invention integrates 3D laser scanning, UAV aerial photography, panoramic photography, and mobile swarm sensing to achieve full-dimensional data coverage from macro to micro and from overall to local dimensions, avoiding blind spots of single technologies. It is particularly suitable for detecting complex structures or areas inaccessible from high altitudes. Through deep learning-based automated classification algorithms, it identifies damage types such as weathering, efflorescence, and vegetation coverage, eliminating subjective errors from manual annotation, increasing annotation efficiency several times over, and handling massive amounts of data to meet the needs of large-scale ancient building complex assessments. Based on computer vision, it extracts pixel coordinates, area, and morphological features of damaged areas, accurately calculating quantitative indicators such as weathering rate and efflorescence rate, providing objective evidence of damage severity and avoiding the ambiguity of traditional experience-based judgments. By integrating historical repair records and environmental monitoring data (such as temperature, humidity, and pollution), it constructs a wall damage evolution model, predicts future damage trends, and allows for the development of preventative protection plans to extend the lifespan of ancient buildings. This method, through technological integration and innovation, achieves precise, intelligent, and dynamic assessment of weathering damage to ancient building brick and stone walls, providing an efficient, scientific, and sustainable solution for cultural heritage protection, and has significant industry promotion value and social significance. Attached Figure Description

[0015] Figure 1 This is a schematic diagram of the method flow for S101-S104 in this invention; Figure 2 This is a schematic diagram of the method flow for S201-S203 in this invention; Figure 3 This is a schematic diagram of the method flow for S301-S304 in this invention; Figure 4 This is a schematic diagram of the method flow for S401-S404 in this invention; Figure 5 This is a schematic diagram of the method flow for S501-S503 in this invention; Figure 6 This is a schematic diagram of the method flow of S601-S604 in this invention. Detailed Implementation

[0016] The following description is intended to disclose the invention and enable those skilled in the art to implement it. The preferred embodiments described below are merely examples, and other obvious variations will occur to those skilled in the art.

[0017] Example 1 Please refer to Figure 1 As shown, in a first aspect of the present invention, a method for intelligent assessment of weathering damage to brick and stone walls of ancient buildings based on big data is provided, comprising: S101. Establish a multi-source data acquisition system, integrating 3D laser scanning, UAV aerial photography, panoramic photography and mobile crowd sensing technologies to obtain geometric and physical information and surface damage images of ancient building brick and stone walls. S102. Construct a weathering damage sample database. Use deep learning algorithms to automatically classify and label the collected images to form a standardized dataset containing typical damage types such as weathering, efflorescence, and vegetation cover. S103. Develop a damage assessment model based on computer vision and image processing technology, extract pixel coordinates, area and morphological features of the damaged area, and calculate quantitative indicators such as weathering rate, efflorescence rate and green coverage rate. S104. Generate a visual assessment report that integrates historical repair records and environmental monitoring data, and uses digital twin technology to predict the evolution trend of wall damage and optimize repair plans.

[0018] As will be understood by those skilled in the art, this invention integrates 3D laser scanning, UAV aerial photography, panoramic photography, and mobile swarm intelligence sensing to achieve full-dimensional data coverage from macro to micro and from whole to part, avoiding blind spots of single technologies. It is particularly suitable for detecting complex structures or areas inaccessible from high altitudes. Through deep learning-based automated classification algorithms, it identifies damage types such as weathering, efflorescence, and vegetation coverage, eliminating subjective errors from manual annotation, increasing annotation efficiency several times over, and processing massive amounts of data to meet the needs of large-scale ancient building complex assessments. Based on computer vision, it extracts pixel coordinates, area, and morphological features of damaged areas, accurately calculating quantitative indicators such as weathering rate and efflorescence rate, providing objective evidence for the degree of damage and avoiding the ambiguity of traditional experience-based judgments. By integrating historical repair records and environmental monitoring data (such as temperature, humidity, and pollution), it constructs a wall damage evolution model, predicts future damage trends, and formulates preventative protection plans in advance to extend the lifespan of ancient buildings. This method, through technological integration and innovation, achieves precise, intelligent, and dynamic assessment of weathering damage to ancient building brick and stone walls, providing an efficient, scientific, and sustainable solution for cultural heritage protection, and has significant industry promotion value and social significance.

[0019] Please refer to Figure 2 As shown, a multi-source data acquisition system is established, specifically including: S201. Use a handheld or fixed laser scanner to acquire point cloud data of the wall surface with an accuracy of not less than 0.1mm; S202: A drone equipped with a high-resolution camera and oblique photography system is used to achieve 720-degree full coverage of wall images; S203. Integrate temperature and humidity sensors, anemometers, and salt spray detection devices near the brick and stone walls of ancient buildings to collect physicochemical parameters of the environment in which the walls are located.

[0020] Please refer to Figure 3 As shown, the construction of the weathering damage sample database specifically includes: S301. Denoise, enhance and normalize the acquired raw image to unify the resolution to 4096×4096 pixels; S302. An improved YOLOv4 target detection algorithm is adopted, combined with K-means clustering analysis, to automatically identify weathered loose layer, efflorescence crystals and green vegetation coverage areas; S303. Based on the damage classification standards in the "Technical Specification for Repair of Exposed Walls of Historical Buildings", the severity and distribution range of each type of damage shall be marked. The severity includes mild, moderate and severe. S304. Enhanced samples are generated by rotating, flipping, and adjusting brightness, so that the database capacity is no less than 100,000 labeled images.

[0021] Those skilled in the art will understand that eliminating feature extraction bias caused by resolution differences in multi-source images ensures accurate identification of details such as damage morphology and texture by subsequent algorithms, avoids misjudgment caused by image blur or scale inconsistencies, eliminates interference from the shooting environment (such as lighting and shadows) through noise reduction algorithms, enhances the contrast between damaged areas and healthy walls, improves the recognition sensitivity of target detection algorithms, and is particularly advantageous for capturing early minor damage. Combining target detection algorithms with unsupervised clustering analysis, the system automatically locates damaged areas such as weathered loose layers and efflorescence crystals, adapting to the rapid assessment needs of large-scale ancient building complexes. Based on the "Technical Specification for Repair of Exposed Walls of Historical Buildings", the severity is divided into mild, moderate and severe, and the distribution range is marked, providing a quantitative grading basis for subsequent assessment models, avoiding assessment bias caused by subjective experience, and improving the reproducibility of results. By simulating different shooting angles and environmental conditions through affine transformations such as rotation, flipping, and brightness adjustment, the diversity of the dataset is expanded, effectively alleviating the overfitting problem of deep learning models to small samples and significantly improving the model's generalization ability in complex scenes. The inclusion of low-frequency damage (such as local peeling caused by extreme weather) in the samples enables the model to identify rare damage and enhances the robustness of the evaluation system.

[0022] Please refer to Figure 4 As shown, an improved YOLOv4 target detection algorithm, combined with K-means clustering analysis, is used to automatically identify weathered loose layers, efflorescence crystals, and vegetation-covered areas, specifically including: S401. Using CSPDarknet53 as the backbone network, a hybrid attention mechanism module is embedded in its residual block. This module consists of a channel attention submodule and a spatial attention submodule connected in series. The channel attention submodule generates channel weight vectors through global average pooling, and the spatial attention submodule generates spatial weight maps through convolutional layers with 3×3 kernels. The two are fused by element-wise multiplication and then multiplied with the original feature map to enhance the feature response of weathered loose layer, alkali crystals and green vegetation coverage area. S402. Cluster analysis is performed on the labeled damage samples based on the K-means++ algorithm. The intersection-union ratio (IoU) is used as the distance metric to iteratively optimize the cluster centers. Considering the large differences in the scale of damage to ancient buildings, nine anchor frames are generated, with aspect ratios covering the range of [0.25, 0.5, 1, 2, 4] and area ranges covering [16×16, 32×32, 64×64] pixels, ensuring adaptability to both micro-weathering cracks and large-area efflorescence areas. S403. The CIoU loss function is used to replace the original YOLOv4 IoU loss, and a normalized distance and aspect ratio consistency penalty term of the minimum bounding rectangle (MBR) are introduced. The calculation formula is as follows: ; in, (.) represents the Euclidean distance. and , where are the center points of the predicted bounding box and the ground truth bounding box, respectively, and c is the length of the diagonal of the minimum bounding rectangle. These are the weighting coefficients. Aspect ratio consistency parameter; Meanwhile, Focal Loss is introduced to address the class imbalance problem, and its formula is: ; in, For category weights, To predict probabilities, The adjustment factor has a value of 2. S404. Redundant convolutional channels in CSPDarknet53 are removed through channel pruning. The pruning rate is set to 40%, and channels that contribute highly to the damage features are retained. 8-bit fixed-point quantization is used to convert floating-point weight parameters to integers to reduce model storage space.

[0023] Those skilled in the art will understand that the beneficial effects are as follows: (1) Hybrid attention mechanism enhances feature extraction and improves the accuracy of micro-damage recognition. Channel weight vectors are generated using global average pooling to focus on channel features of damage such as weathered porous layers and efflorescence crystals. A spatial weight map is generated using a 3×3 convolutional layer to accurately locate the spatial distribution of the damage area. The two weight maps are fused and multiplied with the original feature map to improve the model's ability to detect minute weathering cracks (width < 2 mm) and early efflorescence spots (area < 1 cm²). 2 The feature response is enhanced by more than 30%, significantly reducing the false negative rate. In the identification of green plant-covered areas, the attention mechanism can effectively distinguish the similar features of wall texture and plant leaves, reducing false detections by 25% (such as misjudging the shadow of vines as damage).

[0024] (2) K-means++ clustering optimizes anchor frame design to adapt to multi-scale damage detection To address the significant scale variations in damage to ancient buildings (ranging from hair-thin cracks to efflorescence covering several square meters), the K-means++ algorithm iteratively optimizes cluster centers based on the intersection-over-union (IoU) ratio, generating nine anchor boxes covering aspect ratios of [0.25, 0.5, 1, 2, 4] and area ranges of [16×16, 32×32, 64×64] pixels. Experiments show that this design improves the recall rate of large-scale damage (such as extensive vegetation cover) by 18% and the localization accuracy of small-scale damage (such as micro-cracks) by 22%, while reducing redundant computation: compared to the fixed anchor boxes of the original YOLOv4, the optimized anchor boxes have a higher degree of matching with the actual damage size, reducing the generation of invalid candidate boxes by 40% and accelerating model inference speed.

[0025] (3) Improved loss function to solve the problem of class imbalance in bounding box regression. By optimizing the bounding box regression process through the normalized distance and aspect ratio consistency penalty term of the minimum bounding rectangle (MBR), CIoU improves the overlap (IoU) between the predicted box and the ground truth box by 12% in the detection of weathered and porous layers. It is especially accurate in locating damage with tilted or irregular shapes (such as spalling blocks). Focal Loss alleviates class imbalance: for the problem that there are far more normal area samples than damaged samples in the ancient building dataset, Focal Loss dynamically adjusts the class weights (γ=2) to make the model pay more attention to hard-to-classify samples (such as the boundary between mild efflorescence and healthy walls).

[0026] (4) Lightweight model design, balancing accuracy and deployment efficiency By removing 40% of the redundant convolutional channels in CSPDarknet53 and retaining channels that contribute highly to damage features, the model reduces the number of model parameters by 35% while maintaining 95% of the original accuracy, accelerating inference speed by more than 2 times. 8-bit fixed-point quantization compression converts floating-point weight parameters to integers, reducing the model storage space from 240MB to 60MB, enabling real-time operation on embedded devices (such as edge computing modules for drones) and meeting the needs of rapid on-site detection of ancient buildings.

[0027] Please refer to Figure 5 As shown, the calculation steps of the damage assessment model include: S501. Segment damaged regions in the image using the GrabCut algorithm, extract contour coordinates using the OpenCV library, and calculate the porosity of the damaged regions through pixel area statistics and morphological analysis. Crack density and the distribution density of crystals ; S502, Porosity based on damaged region Crack density and the distribution density of crystals The weathering rate was obtained. Alkali bloom rate A and green coverage rate Three major quantitative indicators; S503. Classify damage levels based on three major quantitative indicators to guide the priority ranking of repairs.

[0028] Porosity The calculation formula is: ; in, The pixel area of ​​the pore region. This represents the total pixel area of ​​the damaged region; Crack density The calculation formula is: ; in, The pixel length of the crack skeleton line is converted to the actual length through pixel size calibration. This represents the total pixel area of ​​the damaged region; Crystal distribution density The calculation formula is: ; in, The pixel area of ​​the crystalline region; This represents the total pixel area of ​​the wall surface.

[0029] weathering rate The calculation formula is: ; in, and Here are the weighting coefficients. =0.6, =0.4; This is a conversion factor for crack length units, such as from pixels to meters; The formula for calculating the alkali bloom rate A is: A ; The formula for calculating the green coverage rate V is: ; in, The pixel area represents the area of ​​the green plants. This represents the total pixel area of ​​the wall surface; In the green coverage rate V, the root system of the plants causes cracks in the wall. A correction term can be introduced, which would be: ; in, The factor representing the aggravation effect of vegetation on cracks. Crack density.

[0030] Please refer to Figure 6 As shown, a visual assessment report is generated, integrating historical repair records and environmental monitoring data. Digital twin technology is used to predict the evolution trend of wall damage and optimize repair plans, specifically including: S601, Calculated Weathering Rate Alkali bloom rate A and green coverage rate Spatial distribution data were used to convert each parameter value into RGB color values ​​using the Jet color mapping algorithm, generating a three-dimensional damage heat map of the wall surface, and overlaying location labels in the CGCS2000 coordinate system. S602. Based on the areas with weathering rate >25% or efflorescence rate >15% in the damage heat map, recommend matching materials in conjunction with the "Standard for Materials for Repairing Ancient Buildings", and estimate the repair budget for a single area in accordance with the "Construction Engineering Quantity List Pricing Specification", and form a priority ranking table including damage level, recommended process and total budget. S603. Driven by the time series data of damage scores predicted by the LSTM neural network for the next 5 years, the dynamic process of crack propagation, crystal growth and root erosion of the wall surface is simulated by the Unity3D engine, generating an MP4 format evolution video with a frame rate of 30fps, and marking the predicted year and damage score change value in the key frame. S604. Retrieve damage heat map data from historical assessment periods, use transparency overlay technology to display it on the same screen as the current heat map, and intuitively quantify the reduction in weathering rate (ΔW) and efflorescence rate (ΔA) after repair through color differences, and generate comparative analysis charts that include statistical significance (p < 0.05).

[0031] In a second aspect of the invention, a smart assessment system for weathering damage to ancient building brick and stone walls based on big data is also provided, comprising: The acquisition module is used to establish a multi-source data acquisition system, integrating 3D laser scanning, drone aerial photography, panoramic photography and mobile crowd sensing technologies to acquire geometric and physical information and surface damage images of ancient building brick and stone walls. The sample database construction module is used to build a weathering damage sample database. It uses deep learning algorithms to automatically classify and label the collected images, forming a standardized dataset that includes typical damage types such as weathering, efflorescence, and vegetation cover. The assessment module is used to develop damage assessment models. Based on computer vision and image processing technology, it extracts the pixel coordinates, area and morphological features of the damaged area and calculates quantitative indicators such as weathering rate, efflorescence rate and green coverage rate. The assessment report generation module is used to generate visual assessment reports, integrating historical repair records and environmental monitoring data. It uses digital twin technology to predict the evolution trend of wall damage and optimize repair plans.

[0032] In a third aspect of the invention, an electronic device is also provided. The electronic device includes at least one processor; and a memory communicatively connected to the at least one processor; the memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enable the at least one processor to perform the method of the first aspect of the invention.

[0033] The foregoing has shown and described the basic principles, main features, and advantages of the present invention. Those skilled in the art should understand that the present invention is not limited to the above embodiments. The embodiments and descriptions in the specification are merely principles of the invention. Various changes and modifications can be made to the invention without departing from its spirit and scope, and all such changes and modifications fall within the scope of the claimed invention. The scope of protection claimed by the appended claims and their equivalents is defined.

Claims

1. A smart assessment method for weathering damage to brick and stone walls of ancient buildings based on big data, characterized in that, include: Establish a multi-source data acquisition system that integrates 3D laser scanning, drone aerial photography, panoramic photography and mobile crowd sensing technologies to acquire geometric and physical information and surface damage images of ancient building brick and stone walls. A weathering damage sample database was constructed, and the collected images were automatically classified and labeled using deep learning algorithms to form a standardized dataset containing typical damage types such as weathering, efflorescence, and vegetation cover. Develop a damage assessment model based on computer vision and image processing technology, extract pixel coordinates, area and morphological features of the damaged area, and calculate quantitative indicators such as weathering rate, efflorescence rate and green coverage rate. Generate a visual assessment report that integrates historical repair records and environmental monitoring data, and use digital twin technology to predict the evolution trend of wall damage and optimize repair plans.

2. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 1, characterized in that, The establishment of the multi-source data acquisition system specifically includes: Use a handheld or fixed laser scanner to acquire point cloud data of the wall surface with an accuracy of no less than 0.1mm; A drone equipped with a high-resolution camera and an oblique photography system was used to capture images of the wall in a 720-degree full coverage. Temperature and humidity sensors, anemometers, and salt spray detection devices are integrated near the brick and stone walls of ancient buildings to collect physicochemical parameters of the environment in which the walls are located.

3. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 2, characterized in that, The construction of the weathering damage sample database specifically includes: The original images were denoised, enhanced, and normalized to a resolution of 4096×4096 pixels. An improved YOLOv4 target detection algorithm, combined with K-means clustering analysis, is used to automatically identify weathered loose layers, efflorescence crystals, and areas covered by green vegetation. Based on the damage classification standards in the "Technical Specification for Repair of Exposed Walls of Historical Buildings", the severity and distribution range of each type of damage are marked, and the severity includes mild, moderate and severe. Enhanced samples are generated by rotating, flipping, and adjusting brightness, so that the database capacity is no less than 100,000 labeled images.

4. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 3, characterized in that, The improved YOLOv4 target detection algorithm, combined with K-means clustering analysis, automatically identifies weathered loose layers, efflorescence crystals, and vegetation-covered areas, specifically including: Using CSPDarknet53 as the backbone network, a hybrid attention mechanism module is embedded in its residual blocks. This module consists of a channel attention submodule and a spatial attention submodule connected in series. The channel attention submodule generates channel weight vectors through global average pooling, while the spatial attention submodule generates spatial weight maps through convolutional layers with 3×3 kernels. The two are fused by element-wise multiplication and then multiplied with the original feature map to enhance the feature responses of weathered and loose layers, efflorescence crystals, and vegetation-covered areas. Cluster analysis of labeled damage samples was performed based on the K-means++ algorithm. Intersection over Union (IoU) was used as the distance metric to iteratively optimize the cluster centers. Considering the large differences in the scale of damage to ancient buildings, nine anchor frames were generated, with aspect ratios covering the range of [0.25, 0.5, 1, 2, 4] and area ranges covering [16×16, 32×32, 64×64] pixels, ensuring adaptability to both micro-weathering cracks and large-area efflorescence areas. The CIoU loss function is used to replace the original YOLOv4 IoU loss, and a normalized distance and aspect ratio consistency penalty term based on the minimum bounding rectangle (MBR) is introduced. The calculation formula is as follows: ； in, (.) represents the Euclidean distance. and , where are the center points of the predicted bounding box and the ground truth bounding box, respectively, and c is the length of the diagonal of the minimum bounding rectangle. These are the weighting coefficients. Aspect ratio consistency parameter; Meanwhile, Focal Loss is introduced to address the class imbalance problem, and its formula is: ； in, For category weights, To predict probabilities, The adjustment factor has a value of 2. Redundant convolutional channels in CSPDarknet53 were removed by channel pruning, with a pruning rate of 40%, while channels that contribute highly to the damage features were retained. 8-bit fixed-point quantization was used to convert floating-point weight parameters to integers, reducing model storage space.

5. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 4, characterized in that, The calculation steps of the damage assessment model include: The GrabCut algorithm was used to segment damaged regions in an image, and contour coordinates were extracted using the OpenCV library. The porosity of the damaged regions was then calculated through pixel area statistics and morphological analysis. Crack density and the distribution density of crystals ; Based on the porosity of the damaged area Crack density and the distribution density of crystals The weathering rate was obtained. Alkali bloom rate A and green coverage rate Three major quantitative indicators; The damage levels are classified based on the three quantitative indicators mentioned above, which guides the prioritization of repairs.

6. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 5, characterized in that, porosity The calculation formula is: ； in, The pixel area of ​​the pore region. This represents the total pixel area of ​​the damaged region; The crack density The calculation formula is: ； in, The pixel length of the crack skeleton line is converted to the actual length through pixel size calibration. This represents the total pixel area of ​​the damaged region; The distribution density of the crystals The calculation formula is: ； in, The pixel area of ​​the crystalline region; This represents the total pixel area of ​​the wall surface.

7. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 6, characterized in that, The weathering rate The calculation formula is: ； in, and Here are the weighting coefficients. =0.6, =0.4; This is a conversion factor for crack length units, such as from pixels to meters; The formula for calculating the alkali bloom rate A is: A ; The formula for calculating the green coverage rate V is: ； in, The pixel area represents the area of ​​the green plants. This represents the total pixel area of ​​the wall surface; In the green coverage rate V, the root system of the plants causes cracks in the wall. A correction term can be introduced, which would be: ； in, The factor representing the aggravation effect of vegetation on cracks. Crack density.

8. The intelligent assessment method for weathering damage of ancient building brick and stone walls based on big data as described in claim 7, characterized in that, The generated visual assessment report integrates historical repair records and environmental monitoring data, and uses digital twin technology to predict the evolution trend of wall damage and optimize repair plans. Specifically, it includes: Calculated weathering rate Alkali bloom rate A and green coverage rate Spatial distribution data were used to convert each parameter value into RGB color values ​​using the Jet color mapping algorithm, generating a three-dimensional damage heat map of the wall surface, and overlaying location labels in the CGCS2000 coordinate system. Based on the areas with weathering rate >25% or efflorescence rate >15% in the damage heat map, and in conjunction with the recommended matching materials in the "Standard for Materials for the Repair of Ancient Buildings", and based on the "Construction Engineering Quantity List Pricing Specification", the repair budget for a single area is estimated, forming a priority ranking table that includes damage level, recommended process and total budget. Driven by time-series data of damage scores predicted by LSTM neural network for the next 5 years, the dynamic process of crack propagation, crystal growth and root erosion of the wall surface is simulated by Unity3D engine, generating an MP4 format evolution video with a frame rate of 30fps, and the predicted year and damage score change value are marked in the key frame. Damage heat map data from historical assessment periods are retrieved and displayed on the same screen as the current heat map using transparency overlay technology. The reduction in weathering rate (ΔW) and efflorescence rate (ΔA) after repair are quantified intuitively through color differences, and comparative analysis charts containing statistical significance (p < 0.05) are generated.

9. A big data-based intelligent assessment system for weathering damage of ancient building brick and stone walls, used to implement the big data-based intelligent assessment method for weathering damage of ancient building brick and stone walls as described in any one of claims 1-8, characterized in that, include: The acquisition module is used to establish a multi-source data acquisition system, integrating three-dimensional laser scanning, UAV aerial photography, panoramic photography and mobile crowd sensing technology to acquire geometric and physical information and surface damage images of the brick and stone walls of ancient buildings. The sample database construction module is used to construct a weathering damage sample database. It uses deep learning algorithms to automatically classify and label the collected images, forming a standardized dataset containing typical damage types such as weathering, efflorescence, and vegetation cover. The assessment module is used to develop a damage assessment model. Based on computer vision and image processing technology, it extracts the pixel coordinates, area and morphological features of the damaged area and calculates quantitative indicators such as weathering rate, efflorescence rate and green coverage rate. The assessment report generation module is used to generate a visual assessment report, which integrates historical repair records and environmental monitoring data, and uses digital twin technology to predict the evolution trend of wall damage and optimize repair plans.

10. An electronic device, comprising at least one processor; and a memory communicatively connected to said at least one processor; characterized in that, The memory stores instructions that can be executed by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-8.

Images