A digitalization-assisted repair system for historic buildings

Abstract

The present application relates to a kind of digitalization auxiliary repair system for historical building, the system includes building data acquisition module, data preliminary processing module, automatic damage assessment module, repair auxiliary model construction module and historical building repair auxiliary module.The present application obtains original data by data acquisition;Adopt the preliminary processing method of multi-source data alignment, building value assessment, point cloud feature extraction and damage pre-labeling;Adopt improved clustering model to carry out automatic damage assessment, introduce adaptive clustering mechanism, combined with building historical value, provide scientific basis for repair decision;Adopt repair decision transformer model as repair auxiliary model, convert cultural relic protection specification into prior constraint, combined with polar coordinate position coding, strengthen the perception to building symmetry structure, finally realize the collaborative optimization of three-dimensional target.

Description

Technical Field

[0001] This invention relates to the field of digital assisted restoration technology for historical buildings, specifically to a digital assisted restoration system for historical buildings. Background Technology

[0002] Digital-assisted restoration of historical buildings utilizes modern digital technology to comprehensively record, analyze, and design restoration plans for historical buildings. This allows for a precise understanding of the building's current condition, assessment of the extent of structural and material damage, and the development of scientifically sound restoration plans. This technology not only preserves the historical value and cultural heritage of buildings but also minimizes human intervention that could damage them, thus improving the accuracy and efficiency of restoration.

[0003] However, traditional digital-assisted restoration systems for historical buildings suffer from several technical problems. Firstly, manual visual inspections are inefficient and subject to subjective experience, making it difficult to systematically quantify the extent of damage. Secondly, conventional algorithms cannot effectively integrate heterogeneous data from multiple sources, such as infrared thermal imaging and geometric point clouds, resulting in a lack of correlation analysis between internal defects and surface damage. Furthermore, they lack the ability to provide detailed assessments of high-value areas, easily overlooking hidden damage or over-intervening in low-value areas, thus affecting the accuracy of restoration plans. Thirdly, traditional digital-assisted restoration systems for historical buildings have weak capabilities in modeling the relationships between architectural spatial structures, neglecting the topological relationships of damaged areas. They often assess multiple objectives such as cost, technology, and value preservation in a fragmented manner, failing to quantify the impact of dynamic risks on decision variables, easily leading to resource imbalances or over-renovation. Finally, traditional models lack domain knowledge embedding mechanisms, making it difficult to ensure that restoration measures comply with cultural relic protection guidelines. Summary of the Invention

[0004] To address the aforementioned issues and overcome the shortcomings of existing technologies, this invention provides a digital assisted restoration system for historical buildings. Traditional digital assisted restoration systems for historical buildings suffer from problems such as low efficiency and reliance on subjective experience in manual visual inspections, difficulty in systematically quantifying damage levels, inability of conventional algorithms to effectively integrate multi-source heterogeneous data such as infrared thermal imaging and geometric point clouds leading to a lack of correlation analysis between internal defects and surface damage, insufficient ability to finely assess high-value areas, and tendency to overlook hidden damage or over-intervene in low-value areas, affecting the accuracy of restoration plans. This invention creatively employs an automatic damage assessment module, constructs an improved clustering model, and performs automatic damage assessment. Through multi-scale fusion, it captures damage characteristics from micro-texture to macro-structure, dynamically adjusts the recognition sensitivity based on the quantitative weight of the building's historical value, and introduces an adaptive clustering mechanism to optimize region segmentation based on building value density, providing a basis for subsequent restoration decisions. The solution provides objective and quantifiable scientific evidence. However, traditional digital auxiliary restoration systems for historical buildings suffer from weaknesses in modeling the spatial structure of buildings, neglecting the topological relationships of damaged areas, and often evaluating multiple objectives such as cost, technology, and value preservation in a fragmented manner. They fail to quantify the impact of dynamic risks on decision variables, easily leading to resource allocation imbalances or over-restoration. Furthermore, traditional models lack domain knowledge embedding mechanisms, making it difficult to ensure that restoration measures comply with cultural relic protection guidelines. This solution creatively adopts a restoration decision transformer model as an auxiliary model. Through multimodal damage fusion and domain knowledge graphs, cultural relic protection standards are transformed into prior constraints, ensuring that decisions comply with regulations. Combined with polar coordinate position coding, it accurately models the geometric and topological relationships of buildings, strengthening the perception of symmetrical structures. Simultaneously, a risk perception mechanism dynamically correlates damage severity with decision output, achieving synergistic optimization of the three-dimensional objectives of restoration technology, cost control, and value preservation.

[0005] The technical solution adopted by the present invention is as follows: The present invention provides a digital auxiliary repair system for historical buildings, including a building data acquisition module, a preliminary data processing module, an automatic damage assessment module, a repair auxiliary model construction module, and a historical building repair auxiliary module;

[0006] The building data acquisition module acquires data to obtain the original dataset of building repairs and sends the original dataset of building repairs to the data preliminary processing module.

[0007] The preliminary data processing module uses multi-source data alignment, building value assessment, point cloud feature extraction and damage pre-labeling to obtain historical preliminary datasets and current preliminary datasets, and sends the historical preliminary datasets and the current preliminary datasets to the automatic damage assessment module.

[0008] The automatic damage assessment module is used to automatically assess the damage status of historical buildings. By constructing an improved clustering model, it performs automatic damage assessment to obtain a dataset to be processed, an auxiliary repair training set, and an auxiliary repair test set. The dataset to be processed is sent to the historical building repair auxiliary module, and the auxiliary repair training set and the auxiliary repair test set are sent to the repair auxiliary model construction module.

[0009] The repair auxiliary model construction module constructs a repair decision transformer model as a repair auxiliary model and sends the repair auxiliary model to the repair auxiliary model construction module.

[0010] The historical building renovation assistance module specifically involves using the renovation assistance model to assist in the renovation of historical buildings and obtain reference results for building renovation decisions.

[0011] Furthermore, in the building data acquisition module, the original dataset for building repair specifically includes a historical original dataset for building repair and a current original dataset for building repair. Both the historical original dataset for building repair and the current original dataset for building repair include building optical image data, building infrared image data, building basic information data, building material data, and building 3D point cloud data. The historical original dataset for building repair also includes labeled data.

[0012] Furthermore, in the preliminary data processing module, the multi-source data alignment is used to unify the data spatial coordinate system, specifically by performing pixel-level alignment of building optical image data, building infrared image data, and building 3D point cloud data through feature point matching;

[0013] The building value assessment is used to quantify the historical value of a building. Specifically, it involves converting data on the building's construction date, the building's cultural relic protection level, and the types of building techniques into weights using preset rules and a multilayer perceptron, thereby constructing a building value weight matrix.

[0014] The point cloud feature extraction is used to characterize the geometric properties of the building surface. Specifically, it involves processing the 3D point cloud data of the building through normal vector calculation and depth projection to generate point cloud geometric features that are aligned with the building's optical image data and infrared image data.

[0015] The damage pre-labeling is used to pre-label the building damage level. Specifically, it involves analyzing the differences between building optical image data and building infrared image data to pre-generate a building damage binary mask. Then, the building damage binary mask is manually corrected and the damage level is divided to obtain the building damage level label.

[0016] By performing multi-source data alignment, building value assessment, point cloud feature extraction, and damage pre-labeling, preliminary data processing is carried out on the original historical dataset and the original current dataset of building repair to obtain a preliminary historical dataset and a preliminary current dataset.

[0017] Furthermore, in the automatic damage assessment module, the automatic identification and assessment of damaged areas of historical buildings is carried out by analyzing and extracting multi-scale damage features, integrating the building value weight matrix and constructing an improved clustering model for processing, and obtaining a spatial heat map of damage levels.

[0018] The automatic damage assessment module specifically includes multi-scale feature extraction, feature enhancement, improved clustering model design, damage heatmap generation, and building damage assessment.

[0019] The multi-scale feature extraction is used to capture damage features of different sizes. Specifically, it involves analyzing the fusion information of architectural optical image data and architectural infrared image data through a convolutional neural network to extract cross-scale feature maps from fine-grained textures to macroscopic structures, resulting in cross-scale features, including:

[0020] Pyramid convolution is used to extract scale-sensitive features. Specifically, it fuses architectural optical features and architectural infrared features through channel splicing operations, and extracts multi-granularity features from the fused features using three different sizes of convolution kernels to obtain multi-scale features.

[0021] Cross-scale feature fusion is used to integrate multi-granularity information. Specifically, it involves fusing small-scale features with large-scale features through upsampling and weighted addition operations to obtain cross-scale features.

[0022] The feature enhancement is used to integrate the historical value and damage characteristics of a building. Specifically, it expands the building value weight to the same dimension as the cross-scale feature through tensor broadcasting operation, and multiplies it element-wise with the cross-scale feature to obtain the value enhancement feature.

[0023] The improved clustering model design, used for damage region segmentation, specifically achieves adaptive segmentation of damage regions through superpixel segmentation and dynamic density clustering, obtaining pixel-level damage level labels, including:

[0024] Superpixel segmentation is used to reduce computational complexity. Specifically, it uses a simple linear iterative clustering algorithm to segment the input image data into multiple homogeneous texture regions. Each homogeneous texture region is represented by a superpixel unit, resulting in a set of superpixel units.

[0025] Region feature aggregation is used to extract representative features of each homogeneous texture region. Specifically, it calculates the mean of the value enhancement features within each superpixel unit through average pooling operation to obtain the representative features of the superpixel unit.

[0026] Dynamic density clustering is used for adaptive region partitioning. Specifically, it dynamically adjusts the search radius of DBSCAN clustering by building value weights, and clusters superpixel units based on the representative features of superpixel units to obtain cluster labels for damaged regions.

[0027] Damage level mapping is used to generate automatic damage assessment results. Specifically, it maps the clustering labels of the damaged area to five-level building damage level labels through a small number of labeled samples, resulting in pixel-level damage level labels.

[0028] The damage heatmap generation is used to visualize the automatic damage assessment results. Specifically, it involves superimposing pixel-level damage level labels and building value weights onto the original input image data through color mapping to obtain a spatial heatmap of damage levels.

[0029] The building damage assessment specifically involves using the historical preliminary dataset and the current preliminary dataset as input to the automatic damage assessment module. Through multi-scale feature extraction, feature enhancement, improved clustering model design, and damage heatmap generation, automatic building damage assessment is performed to obtain a dataset to be processed and a historical assessment dataset. The historical assessment dataset is then segmented to obtain an auxiliary repair training set and an auxiliary repair test set.

[0030] Furthermore, in the renovation auxiliary model construction module, the model used to construct the digital auxiliary renovation of historical buildings is specifically the renovation decision transformer model, which serves as the renovation auxiliary model.

[0031] The repair auxiliary model construction module specifically includes multimodal feature fusion, knowledge prior attention design, polar coordinate encoder design, building repair decision generation, and model construction and training.

[0032] The multimodal feature fusion is used to integrate multi-source repair data. Specifically, it involves aligning and fusing damage level spatial heatmaps, building material features, and point cloud geometric features through feature splicing and 1×1 convolution operations, and then compressing them to obtain multimodal fused features.

[0033] The aforementioned prior knowledge attention design is used to inject domain-specific prior knowledge. Specifically, it involves attention interaction between a predefined ancient architecture knowledge graph and multimodal fusion features to obtain knowledge reinforcement features. The predefined ancient architecture knowledge graph is used to transform known restoration experience in the field of ancient architecture into machine-understandable knowledge. Specifically, it is a knowledge graph constructed based on relevant technical regulations and management methods for ancient architecture. The prior knowledge attention design includes the following:

[0034] Knowledge graph embedding is used to incorporate restoration experience in the field of ancient buildings. Specifically, it involves processing the predefined ancient building knowledge graph through a graph neural network to obtain ancient building knowledge embedding.

[0035] Attention-based interaction is used for knowledge-guided feature correction. Specifically, it calculates the correlation between ancient building knowledge embedding and multimodal fusion features through an attention mechanism to obtain knowledge-enhanced features.

[0036] The polar coordinate encoder design is used to model architectural spatial relationships. Specifically, it captures the topological relationships of damaged areas in a building through rotationally symmetric polar coordinate position encoding and a six-layer Transformer architecture to obtain deep architectural features, including:

[0037] Polar coordinate position encoding is used to enhance geometry perception. Specifically, it replaces the standard position encoding with a polar coordinate function to obtain rotationally symmetric position embedding.

[0038] Deep feature extraction is used to extract deep-level historical building damage information. Specifically, it uses a six-layer Transformer architecture to process rotationally symmetric position embeddings to obtain deep building features.

[0039] The building repair decision generation is used to generate quantifiable decisions. Specifically, it outputs building repair decision prediction results through damage risk quantification and multi-objective prediction, including:

[0040] Damage risk quantification is used to calculate the degree of damage risk. Specifically, it obtains the damage risk coefficient by weighting the proportion of pixels with different damage levels.

[0041] Multi-objective prediction is used for preliminary decision-making prediction. Specifically, it uses three independent output heads to predict the type of restoration technology, the level of restoration cost, and the degree of preservation of the architectural and cultural heritage value.

[0042] Risk perception decision correction is used to integrate damage risk coefficient. Specifically, it dynamically adjusts the output of three decision dimensions—type of repair technology, level of repair cost, and degree of preservation of architectural and cultural heritage value—through the damage risk coefficient to obtain the prediction result of building repair decision.

[0043] The construction and training of the model specifically involves constructing a repair decision transformer model through multimodal feature fusion, knowledge prior attention design, polar coordinate encoder design, and building repair decision generation. The model is then trained and its performance is verified based on the auxiliary repair training set and the auxiliary repair test set to obtain the repair decision transformer model, which serves as the auxiliary repair model.

[0044] Furthermore, in the historical building renovation assistance module, the dataset to be processed is used as the input of the renovation assistance model to obtain the building renovation decision reference result, and the digital renovation of historical buildings is assisted based on the building renovation decision reference result.

[0045] The beneficial effects achieved by the present invention using the above solution are as follows:

[0046] (1) Traditional digital auxiliary repair systems for historical buildings suffer from low efficiency of manual visual inspection and are constrained by subjective experience. They are difficult to systematically quantify the degree of damage. Conventional algorithms cannot effectively integrate multi-source heterogeneous data such as infrared thermal imaging and geometric point clouds, resulting in the lack of correlation analysis between internal defects and surface damage. Furthermore, they lack the ability to conduct refined assessment of high-value areas, easily overlooking hidden damage or over-intervening in low-value areas, thus affecting the accuracy of repair plans. This solution creatively adopts an automatic damage assessment module, constructs an improved clustering model, and conducts automatic damage assessment. Through multi-scale fusion, it captures damage characteristics from micro-texture to macro-structure. Combined with the quantitative weight of the building's historical value, it dynamically adjusts the identification sensitivity. At the same time, it introduces an adaptive clustering mechanism to optimize regional segmentation based on the building's value density, providing an objective and quantifiable scientific basis for subsequent repair decisions.

[0047] (2) Traditional digital auxiliary repair systems for historical buildings have weak modeling capabilities for the spatial structure of buildings, neglect the topological correlation of damaged areas, and often evaluate multiple objectives such as cost, technology, and value preservation in a fragmented manner. They cannot quantify the impact of dynamic risks on decision variables, which can easily lead to resource allocation imbalances or over-repair. In addition, traditional models lack domain knowledge embedding mechanisms, making it difficult to ensure that repair measures comply with cultural relic protection guidelines. This solution creatively adopts a repair decision transformer model as a repair auxiliary model. Through multimodal damage fusion and domain knowledge graph, cultural relic protection standards are transformed into prior constraints to ensure that decisions comply with regulations. Combined with polar coordinate position coding, the geometric topological relationship of buildings is accurately modeled to enhance the perception of the symmetrical structure of buildings. At the same time, a risk perception mechanism is used to dynamically correlate damage severity with decision output, so as to achieve synergistic optimization of the three-dimensional objectives of repair technology, cost control, and value preservation. Attached Figure Description

[0048] Figure 1 A schematic diagram of a digital assisted restoration system for historical buildings provided by the present invention;

[0049] Figure 2 This is a flowchart illustrating the preliminary data processing module.

[0050] Figure 3 This is a flowchart illustrating the automatic damage assessment module.

[0051] Figure 4 A flowchart illustrating the process of building a module for repair auxiliary models.

[0052] The accompanying drawings are provided to further illustrate the invention and form part of the specification. They are used together with the embodiments of the invention to explain the invention and do not constitute a limitation thereof. Detailed Implementation

[0053] The technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0054] In the description of this invention, it should be understood that the terms "upper", "lower", "front", "rear", "left", "right", "top", "bottom", "inner", "outer", etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are only for the convenience of describing this invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on this invention.

[0055] Example 1, see Figure 1 The present invention provides a digital auxiliary repair system for historical buildings, including a building data acquisition module, a preliminary data processing module, an automatic damage assessment module, a repair auxiliary model construction module, and a historical building repair auxiliary module.

[0056] The building data acquisition module acquires data to obtain the original dataset of building repairs and sends the original dataset of building repairs to the data preliminary processing module.

[0057] The preliminary data processing module uses multi-source data alignment, building value assessment, point cloud feature extraction and damage pre-labeling to obtain historical preliminary datasets and current preliminary datasets, and sends the historical preliminary datasets and the current preliminary datasets to the automatic damage assessment module.

[0058] The automatic damage assessment module is used to automatically assess the damage status of historical buildings. By constructing an improved clustering model, it performs automatic damage assessment to obtain a dataset to be processed, an auxiliary repair training set, and an auxiliary repair test set. The dataset to be processed is sent to the historical building repair auxiliary module, and the auxiliary repair training set and the auxiliary repair test set are sent to the repair auxiliary model construction module.

[0059] The repair auxiliary model construction module constructs a repair decision transformer model as a repair auxiliary model and sends the repair auxiliary model to the repair auxiliary model construction module.

[0060] The historical building renovation assistance module specifically involves using the renovation assistance model to assist in the renovation of historical buildings and obtain reference results for building renovation decisions.

[0061] Example 2, see Figure 1 This embodiment is based on the above embodiment. In the building data acquisition module, the original dataset of building repair specifically includes the original dataset of building repair history and the original dataset of building repair current. Both the original dataset of building repair history and the original dataset of building repair current include building optical image data, building infrared image data, building basic information data, building material data and building 3D point cloud data. The original dataset of building repair history also includes labeled data.

[0062] The architectural optical image data specifically refers to RGB image data covering the building facade and structural details, taken under different lighting conditions and from different angles; the architectural infrared image data specifically refers to architectural infrared image data that can reflect the differences in the overall temperature distribution of the building; the architectural basic information data specifically includes data on the building's construction date, the building's cultural relic protection level, and the types of construction techniques used; the annotation data specifically includes building damage level labels and building repair decision labels. The building damage level labels specifically include no damage, minor damage, moderate damage, severe damage, and complete destruction. The building repair decision labels specifically include three dimensions: repair technology type label, repair cost level label, and building cultural relic value preservation degree label.

[0063] Example 3, see Figure 1 and Figure 2 This embodiment is based on the above embodiment. In the data preliminary processing module, the multi-source data alignment is used to unify the data spatial coordinate system. Specifically, it is to perform pixel-level alignment of building optical image data, building infrared image data and building 3D point cloud data through feature point matching.

[0064] The building value assessment is used to quantify the historical value of a building. Specifically, it involves converting data on the building's construction date, the building's cultural relic protection level, and the types of building techniques into weights using preset rules and a multilayer perceptron, thereby constructing a building value weight matrix.

[0065] The point cloud feature extraction is used to characterize the geometric properties of the building surface. Specifically, it involves processing the 3D point cloud data of the building through normal vector calculation and depth projection to generate point cloud geometric features that are aligned with the building's optical image data and infrared image data.

[0066] The damage pre-labeling is used to pre-label the building damage level. Specifically, it involves analyzing the differences between building optical image data and building infrared image data to pre-generate a building damage binary mask. Then, the building damage binary mask is manually corrected and the damage level is divided to obtain a building damage level label. The damage pre-labeling includes fusion mask generation, morphological dilation processing, and manual correction and damage level division.

[0067] The fusion mask generation is used to detect abnormal regions in infrared and RGB images. Specifically, it generates a temperature difference mask and a color difference mask, and obtains the fusion mask through logical AND operations. The formula used is as follows:

[0068] ;

[0069] In the formula, Mt represents the temperature difference mask, and Mc represents the color difference mask. This represents the value of the building's infrared image at coordinates (a, b). This represents the average pixel value of the building's infrared image. This represents the standard deviation of pixels in building infrared images. This represents the value of the architectural optical image at coordinates (a, b). Indicates the reference pixel value for the building's healthy zone. This indicates the preset color difference threshold, and MF represents the fusion mask. This indicates the calculation of the L2 norm. This represents the logical AND operation;

[0070] The morphological dilation process is used to optimize the continuity of the damaged area. Specifically, a 3×3 circular structural element is used to dilate the fusion mask to obtain a binary mask of building damage.

[0071] By performing multi-source data alignment, building value assessment, point cloud feature extraction, and damage pre-labeling, preliminary data processing is carried out on the original historical dataset and the original current dataset of building repair to obtain a preliminary historical dataset and a preliminary current dataset.

[0072] Example 4, see Figure 1 and Figure 3 This embodiment is based on the above embodiment. In the automatic damage assessment module, it is used for the automatic identification and assessment of the damage area of ​​historical buildings. Specifically, it is used to analyze and extract multi-scale damage features, integrate the building value weight matrix and construct an improved clustering model for processing to obtain a spatial heat map of damage level.

[0073] The automatic damage assessment module specifically includes multi-scale feature extraction, feature enhancement, improved clustering model design, damage heatmap generation, and building damage assessment.

[0074] The multi-scale feature extraction is used to capture damage features of different sizes. Specifically, it involves analyzing the fusion information of architectural optical image data and architectural infrared image data through a convolutional neural network to extract cross-scale feature maps from fine-grained textures to macroscopic structures, resulting in cross-scale features, including:

[0075] Pyramid convolution is used to extract scale-sensitive features. Specifically, it fuses architectural optical and infrared features through channel splicing operations, and extracts multi-granularity features from the fused features using three different sized convolution kernels to obtain multi-scale features. The formula used is as follows:

[0076] ;

[0077] In the formula, Indicates fusion characteristics, This represents the concatenation operation function. Indicates the infrared characteristics of a building. Indicates architectural optical features, This represents a multi-scale feature with dimensions sc×sc. Represents a two-dimensional convolution function;

[0078] Cross-scale feature fusion is used to integrate multi-granularity information. Specifically, it fuses small-scale features with large-scale features through upsampling and weighted summation operations to obtain cross-scale features. The formula used is as follows:

[0079] ;

[0080] In the formula, This represents an upsampled feature representing a multi-scale feature of size 64×64. This represents an upsampled feature representing a multi-scale feature of size 128×128. Represents the upsampling function. This represents a multi-scale feature with dimensions of 64×64. This represents a multi-scale feature with dimensions of 128×128. Indicates cross-scale features, Represents the ReLU activation function. This indicates the first cross-scale fusion weight. This represents the second cross-scale fusion weight. This represents the third cross-scale fusion weight;

[0081] The aforementioned feature enhancement is used to integrate the historical value and damage characteristics of a building. Specifically, it involves expanding the building value weights to the same dimension as the cross-scale features through tensor broadcasting operations, and then multiplying them element-wise with the cross-scale features to obtain the value-enhanced features. The formula used is as follows:

[0082] ;

[0083] In the formula, Indicates value reinforcement characteristics, This represents the dimension expansion operation function. Indicates the building's value weight. This represents element-wise multiplication.

[0084] The improved clustering model design, used for damage region segmentation, specifically achieves adaptive segmentation of damage regions through superpixel segmentation and dynamic density clustering, obtaining pixel-level damage level labels, including:

[0085] Superpixel segmentation is used to reduce computational complexity. Specifically, it uses a simple linear iterative clustering algorithm to segment the input image data into multiple homogeneous texture regions. Each homogeneous texture region is represented by a superpixel unit, resulting in a set of superpixel units.

[0086] Region feature aggregation is used to extract representative features of each homogeneous texture region. Specifically, it calculates the mean of the value enhancement features within each superpixel unit through average pooling operation to obtain the representative features of the superpixel unit.

[0087] Dynamic density clustering is used for adaptive region partitioning. Specifically, it dynamically adjusts the search radius of DBSCAN clustering based on building value weights. Based on the representative features of superpixel units, it clusters superpixel units to obtain cluster labels for damaged regions. The dynamic adjustment of the search radius uses the following formula:

[0088] ;

[0089] In the formula, This represents the search radius of the k-th superpixel unit. Indicates the basic search radius. This represents the average building value weight calculated through average pooling within the k-th superpixel unit;

[0090] Damage level mapping is used to generate automatic damage assessment results. Specifically, it maps the clustering labels of the damaged area to five-level building damage level labels through a small number of labeled samples, resulting in pixel-level damage level labels.

[0091] The damage heatmap generation is used to visualize the automatic damage assessment results. Specifically, it involves superimposing pixel-level damage level labels and building value weights onto the original input image data through color mapping to obtain a spatial heatmap of damage levels.

[0092] The building damage assessment specifically involves using the historical preliminary dataset and the current preliminary dataset as input to the automatic damage assessment module. Through multi-scale feature extraction, feature enhancement, improved clustering model design, and damage heatmap generation, automatic building damage assessment is performed to obtain a dataset to be processed and a historical assessment dataset. The historical assessment dataset is then segmented to obtain an auxiliary repair training set and an auxiliary repair test set.

[0093] By performing the above operations, we can address the technical problems of traditional digital assisted restoration systems for historical buildings, such as low efficiency of manual visual inspection, susceptibility to subjective experience, difficulty in systematically quantifying damage levels, inability of conventional algorithms to effectively integrate multi-source heterogeneous data such as infrared thermal imaging and geometric point clouds, resulting in a lack of correlation analysis between internal defects and surface damage, insufficient ability to finely assess high-value areas, easy neglect of hidden damage or over-intervention in low-value areas, and impact on the accuracy of restoration plans. This solution creatively adopts an automatic damage assessment module, constructs an improved clustering model, and performs automatic damage assessment. Through multi-scale fusion, it captures damage characteristics from micro-texture to macro-structure, dynamically adjusts the identification sensitivity by combining the quantitative weight of the building's historical value, and introduces an adaptive clustering mechanism to optimize region segmentation based on building value density, providing an objective and quantifiable scientific basis for subsequent restoration decisions.

[0094] Example 5, see Figure 1 and Figure 4 This embodiment is based on the above embodiment. In the repair auxiliary model construction module, a model is used to construct the digital auxiliary repair model required for historical buildings. Specifically, a repair decision transformer model is constructed as a repair auxiliary model.

[0095] The repair auxiliary model construction module specifically includes multimodal feature fusion, knowledge prior attention design, polar coordinate encoder design, building repair decision generation, and model construction and training.

[0096] The multimodal feature fusion is used to integrate multi-source repair data. Specifically, it involves aligning and fusing damage level spatial heatmaps, building material features, and point cloud geometric features through feature splicing and 1×1 convolution operations, and then compressing them to obtain multimodal fused features.

[0097] The aforementioned prior knowledge attention design is used to inject domain-specific prior knowledge. Specifically, it involves attention interaction between a predefined ancient architecture knowledge graph and multimodal fusion features to obtain knowledge reinforcement features. The predefined ancient architecture knowledge graph is used to transform known restoration experience in the field of ancient architecture into machine-understandable knowledge. Specifically, it is a knowledge graph constructed based on relevant technical regulations and management methods for ancient architecture. The prior knowledge attention design includes the following:

[0098] Knowledge graph embedding is used to incorporate restoration experience in the field of ancient buildings. Specifically, it involves processing the predefined ancient building knowledge graph through a graph neural network to obtain ancient building knowledge embedding.

[0099] Attention-based interaction, used for knowledge-guided feature correction, specifically calculates the correlation between ancient architecture knowledge embeddings and multimodal fusion features through an attention mechanism to obtain knowledge-enhancing features. The formula used is as follows:

[0100] ;

[0101] In the formula, Represents interactive attention. This represents the softmax activation function. This represents the attention query transformation matrix. Represents the attention key transformation matrix. This represents the attention value transformation matrix. Ar represents the multimodal fusion feature, and Ar represents the embedding of ancient architecture knowledge. The dimension representing the attention key. This indicates a knowledge reinforcement feature. Indicates the layer normalization function;

[0102] The polar coordinate encoder design is used to model architectural spatial relationships. Specifically, it captures the topological relationships of damaged areas in a building through rotationally symmetric polar coordinate position encoding and a six-layer Transformer architecture to obtain deep architectural features, including:

[0103] Polar coordinate position encoding is used to enhance geometry perception. Specifically, it replaces the standard position encoding with a polar coordinate function to obtain a rotationally symmetric position embedding. The formula used is as follows:

[0104] ;

[0105] In the formula, r represents the distance from a pixel in the input image data to the image center, a represents the column coordinate of a pixel in the input image data, b represents the row coordinate of a pixel in the input image data, A represents the height of the input image data, and B represents the width of the input image data. This represents the angle of a pixel in the input image data relative to the image center. This represents the positional encoding value in dimension 2c. This represents the positional code value at dimension 2c+1, where C represents the number of dimensions for the positional code. This indicates rotationally symmetric position embedding, where Pe represents position encoding;

[0106] Deep feature extraction is used to extract deep-level historical building damage information. Specifically, it uses a six-layer Transformer architecture to process rotationally symmetric position embeddings to obtain deep building features.

[0107] The building repair decision generation is used to generate quantifiable decisions. Specifically, it outputs building repair decision prediction results through damage risk quantification and multi-objective prediction, including:

[0108] Damage risk quantification is used to calculate the degree of damage risk. Specifically, it obtains a damage risk coefficient by weighting the proportion of pixels with different damage levels. The formula used is as follows:

[0109] ;

[0110] In the formula, This indicates the proportion of pixels occupied by damage level m. The element representing the pixel-level damage level label map is the building damage level label at coordinates (a, b) of the input image data. This represents an indicator function; its value is 1 when the building damage level label at coordinates (a, b) is m, and its value is 0 otherwise. Ri represents the damage risk coefficient. The risk weight represents the damage level m;

[0111] Multi-objective prediction, used for preliminary decision-making, specifically predicts the type of restoration technology, the level of restoration cost, and the degree of preservation of architectural and cultural heritage value through three independent output heads, using the following formula:

[0112] ;

[0113] In the formula, This indicates the predicted types of repair techniques. This indicates the predicted cost level for repairs. This indicates the predicted results of the preservation value of architectural heritage. The weight of the second fully connected layer in the output header represents the type of repair technology. This indicates the weight of the first fully connected layer in the output header, representing the type of repair technology. This indicates the first fully connected layer bias term in the output header representing the type of repair technology. The second fully connected layer bias term of the output head indicates the type of repair technology. This indicates the weight of the second fully connected layer in the repair cost level output header. This indicates the weight of the first fully connected layer in the repair cost level output header. This represents the first fully connected layer bias term of the repair cost level output header. This represents the second fully connected layer bias term of the repair cost level output header. The weights of the second fully connected layer in the output header represent the degree of preservation of the architectural heritage value. The weights of the first fully connected layer in the output header represent the degree of preservation of the architectural heritage value. The first fully connected layer bias term in the output header represents the degree of preservation of the architectural heritage value. The second fully connected layer bias term in the output head, representing the degree of preservation of architectural heritage value, This represents the sigmoid activation function. This represents the GELU activation function. This represents the global average pooling function. Indicates the deep features of the building;

[0114] Risk perception decision correction is used to integrate damage risk coefficients. Specifically, it dynamically adjusts the outputs of three decision dimensions—the type of repair technology, the level of repair cost, and the degree of preservation of architectural and cultural heritage value—based on the damage risk coefficients to obtain the predicted results of building repair decisions. The formula used is as follows:

[0115] ;

[0116] In the formula, This indicates the prediction results for building repair and maintenance decisions. This indicates the compensation coefficient for different types of repair techniques. This indicates the compensation coefficient for the level of repair costs. The coefficient representing the degree of preservation of the architectural and cultural heritage value is the compensation coefficient. Indicates the weight of learnable repair techniques;

[0117] The construction and training of the model specifically involves constructing a repair decision transformer model through multimodal feature fusion, knowledge prior attention design, polar coordinate encoder design, and building repair decision generation. The model is then trained and its performance is verified based on the auxiliary repair training set and the auxiliary repair test set to obtain the repair decision transformer model, which serves as the auxiliary repair model.

[0118] By performing the above operations, we found that traditional digital auxiliary restoration systems for historical buildings have weaknesses in modeling the relationship between the building's spatial structure, neglect the topological correlation of damaged areas, and often conduct fragmented assessments of multiple objectives such as cost, technology, and value preservation. They also fail to quantify the impact of dynamic risks on decision variables, which can easily lead to imbalances in resource allocation or over-restoration. Furthermore, traditional models lack domain knowledge embedding mechanisms, making it difficult to ensure that restoration measures comply with cultural relic protection guidelines. This solution creatively adopts a restoration decision transformer model as a restoration auxiliary model. Through multimodal damage fusion and domain knowledge graphs, cultural relic protection standards are transformed into prior constraints to ensure that decisions comply with regulations. Combined with polar coordinate position coding, it accurately models the geometric and topological relationships of the building, strengthens the perception of the building's symmetrical structure, and uses a risk perception mechanism to dynamically correlate damage severity with decision output, achieving synergistic optimization of the three-dimensional objectives of restoration technology, cost control, and value preservation.

[0119] Example 6, see Figure 1 This embodiment is based on the above embodiment. In the historical building renovation assistance module, the dataset to be processed is used as the input of the renovation assistance model to obtain the building renovation decision reference result. Based on the building renovation decision reference result, the digital renovation of historical buildings is assisted.

[0120] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus.

[0121] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention.

[0122] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A digitalization-assisted restoration system for historic buildings, characterized by: The system comprises a building data acquisition module, a data preliminary processing module, an automatic damage assessment module, a repair assistance model construction module and a historical building repair assistance module; The building data acquisition module obtains a building repair original data set by data acquisition, and the building repair original data set specifically comprises a building repair history original data set and a building repair current original data set; The data preliminary processing module adopts a multi-source data alignment, building value assessment, point cloud feature extraction and damage pre-labeling data preliminary processing method to obtain a history preliminary data set and a current preliminary data set; The automatic damage assessment module is used for automatic identification and assessment of damage areas of historical buildings, specifically by analyzing and extracting multi-scale damage features, fusing a building value weight matrix and constructing an improved clustering model for processing to obtain a damage grade spatial heat map, and the automatic damage assessment module specifically comprises multi-scale feature extraction, feature enhancement, improved clustering model design, damage heat map generation and building damage assessment; The multi-scale feature extraction is used for capturing damage features of different sizes, specifically by analyzing fusion information of building optical image data and building infrared image data through a convolutional neural network to extract cross-scale feature maps from fine-grained texture to macrostructure to obtain cross-scale features; The feature enhancement is used for fusing the historical value and damage features of the building, specifically by extending the building value weight dimension to the same as the cross-scale features through tensor broadcasting operation and multiplying the value enhanced features with the cross-scale features element by element; The improved clustering model design is used for damage area segmentation, specifically by realizing adaptive division of damage areas through superpixel segmentation and dynamic density clustering to obtain pixel-level damage grade labels; The damage heat map generation is used for visualizing the automatic damage assessment result, specifically by superimposing the pixel-level damage grade labels and the building value weight on the original input image data through color mapping to obtain a damage grade spatial heat map; The building damage assessment specifically takes the history preliminary data set and the current preliminary data set as inputs of the automatic damage assessment module, performs building damage automatic assessment through the multi-scale feature extraction, the feature enhancement, the improved clustering model design and the damage heat map generation, obtains a to-be-processed data set and a history evaluation data set, and performs data set segmentation on the history evaluation data set to obtain an auxiliary repair training set and an auxiliary repair test set; The repair assistance model construction module is used for constructing a model required for digital auxiliary repair of historical buildings, specifically a repair decision transformer model as a repair assistance model, and the repair assistance model construction module specifically comprises multi-modal feature fusion, knowledge prior attention design, polar coordinate encoder design, building repair decision generation and model construction and training; The multi-modal feature fusion is used for integrating multi-source repair basis, specifically by aligning and fusing the damage grade spatial heat map, building material features and point cloud geometric features through feature splicing and 1x1 convolution operation and compression to obtain multi-modal fusion features. The knowledge prior attention design is used for injecting field prior knowledge, specifically, attention interaction is performed between a pre-defined ancient building knowledge graph and multi-modal fusion features to obtain knowledge reinforced features. The polar coordinate encoder design is used for modeling building space relationships, specifically, a rotationally symmetric polar coordinate position encoding and a six-layer Transformer architecture are used to capture the topological correlation of building damage areas to obtain building deep features. The building repair decision generation is used for generating quantifiable decisions, specifically, building repair decision prediction results are generated through damage risk quantification and multi-objective prediction. The model is constructed and trained, specifically, the multi-modal feature fusion, the knowledge prior attention design, the polar coordinate encoder design, and the building repair decision generation are used to construct a repair decision transformer model, the model is trained based on an auxiliary repair training set and an auxiliary repair test set, and the model performance is verified to obtain a repair decision transformer model as a repair auxiliary model. The historical building repair auxiliary module is used for assisting the digital repair of historical buildings based on the building repair decision reference results.

2. The digitalization-assisted restoration system for historic buildings according to claim 1, characterized in that: The multi-scale feature extraction includes: Pyramid convolution is used to extract scale-sensitive features, specifically, building optical features and building infrared features are fused through channel concatenation operation, multi-granularity feature extraction is performed on the fused features through three different size convolution kernels to obtain multi-scale features. Cross-scale feature fusion is used to integrate multi-granularity information, specifically, small-scale features and large-scale features are fused through upsampling and weighted addition operation to obtain cross-scale features. The improved clustering model design includes: Superpixel segmentation is used to reduce computational complexity, specifically, input image data is segmented into multiple homogeneous texture regions through a simple linear iterative clustering algorithm, each homogeneous texture region is represented by a superpixel unit to obtain a superpixel unit set. Region feature aggregation is used to extract representative features of each homogeneous texture region, specifically, the mean value of value reinforcement features in each superpixel unit is calculated through average pooling operation to obtain superpixel unit representative features. Dynamic density clustering is used for adaptive region division, specifically, the search radius of DBSCAN clustering is dynamically adjusted based on building value weight, superpixel units are clustered based on superpixel unit representative features to obtain damage region clustering labels. Damage level mapping is used to generate automatic damage assessment results, specifically, a small amount of labeled samples are used to map damage region clustering labels to five-level building damage level labels to obtain pixel-level damage level labels.

3. The digitalization-assisted restoration system for historic buildings according to claim 1, characterized in that: In the knowledge prior attention design, the pre-defined ancient building knowledge graph is used to convert known ancient building repair experience into machine-understandable knowledge, specifically, a knowledge graph is constructed based on ancient building related technical regulations and management methods, and the knowledge prior attention design includes: The knowledge graph embedding is used for introducing repair experience in the field of ancient buildings, and specifically, the predefined ancient building knowledge graph is processed by a graph neural network to obtain ancient building knowledge embedding. The attention interaction is used for knowledge-guided feature correction, and specifically, the relevance of the ancient building knowledge embedding and the multi-modal fusion features is calculated by an attention mechanism to obtain knowledge-enhanced features. The polar coordinate encoder design includes: The polar coordinate position encoding is used for enhancing geometric perception, and specifically, a rotationally symmetric position embedding is obtained by replacing the standard position encoding with a polar coordinate function. The deep feature extraction is used for extracting deep historical building damage information, and specifically, a building deep feature is obtained by processing the rotationally symmetric position embedding through a six-layer Transformer architecture. The building repair decision generation includes: The damage risk quantification is used for calculating the damage risk degree, and specifically, a damage risk coefficient is obtained by weighting the pixel proportions of different damage levels. The multi-target prediction is used for preliminary decision prediction, and specifically, the repair technology type, repair cost level, and building cultural relic value retention degree are predicted by three independent output heads. The risk perception decision correction is used for fusing the damage risk coefficient, and specifically, the repair technology type, repair cost level, and building cultural relic value retention degree are dynamically adjusted by the damage risk coefficient to obtain a building repair decision prediction result.

4. The digitalization-assisted restoration system for historic buildings according to claim 1, characterized in that: The building repair historical original data set and the building repair current original data set both include building optical image data, building infrared image data, building basic information data, building material data, and building 3D point cloud data. The building repair historical original data set further includes labeled data.

5. The digital assisted restoration system for historic buildings of claim 1, wherein: In the data preliminary processing module, the multi-source data alignment is used for unifying the data space coordinate system, and specifically, the building optical image data, building infrared image data, and building 3D point cloud data are pixel-level aligned by feature point matching. The building value evaluation is used for quantifying the historical value of the building, and specifically, the building construction year data, building cultural relic protection level data, and building process type data are converted into weights by a pre-set rule combined with a multi-layer perception machine to construct a building value weight matrix. The point cloud feature extraction is used for representing the geometric characteristics of the building surface, and specifically, the building 3D point cloud data is processed by normal vector calculation and depth projection to generate point cloud geometric features aligned with the building optical image data and building infrared image data. The damage pre-labeling is used for pre-labeling the building damage level, and specifically, a building damage binary mask is pre-generated by analyzing the differences between the building optical image data and the building infrared image data, and then the building damage binary mask is manually corrected and divided into damage levels to obtain building damage level labels. The building repair historical original data set and the building repair current original data set are preliminarily processed by the multi-source data alignment, building value evaluation, point cloud feature extraction, and damage pre-labeling to obtain a historical preliminary data set and a current preliminary data set.

Images