Knowledge graph-based automatic recognition and transfer method for mongolian architectural styles and features

Abstract

Disclosed in the present invention is a knowledge graph-based automatic recognition and transfer method for Mongolian architectural styles and features, the method comprising: Mongolian architecture model library construction, knowledge graph construction, gene clustering and recognition of styles and features, Mongolian architectural style transfer, and actual measurement and correction. The method constructs knowledge graphs of Mongolian architectural styles and features, recognizes style genes of architectures and constructs a library, thereby achieving transfer of Mongolian architectural styles and features, and performs evaluation and optimization by means of on-site construction. The present invention can achieve automatic recognition and transfer of the ethnic styles and features of Mongolian architectures; an element network constructed on the basis of knowledge graphs is more comprehensive, automatically recognized style genes are more scientific and automatically matched architectural styles and features are more accurate, and results can be visually presented, such that the present invention can provide better guidance for Mongolian architecture planning and construction.

Description

An Automatic Recognition and Transfer Method for Architectural Styles in Montenegro Based on Knowledge Graphs Technical Field

[0001] This invention relates to the field of architectural engineering, and specifically to an automatic identification and transfer method for Montenegrin architectural styles based on knowledge graphs. Background Technology

[0002] Urban architectural style is a comprehensive reflection of a city's physical form, socio-cultural, and economic characteristics; it serves as a carrier of urban character and culture. Shaping distinctive architectural styles helps to showcase a city's unique personality, meet people's material and spiritual needs, and achieve harmony between region, culture, and era. Traditional methods of transferring architectural styles typically rely on designers' practical experience and professional knowledge to develop design schemes, which suffer from excessive subjectivity and insufficient scientific rigor. Given the diverse and complex architectural elements of Montenegrin architecture, it is even more difficult to comprehensively and intuitively identify and transfer its stylistic features.

[0003] Architectural style transfer refers to identifying the colors, patterns, facade materials, etc. of one building and applying them to another building to make them present the same or similar architectural styles. Technical issues

[0004] To address the shortcomings of existing technologies, this invention provides an automatic identification and transfer method for architectural styles in Montenegro based on knowledge graphs. This method solves the problems of traditional architectural style identification lacking comprehensive consideration of multi-dimensional factors, insufficient scientific rationality and data quantification support, and the inability to dynamically and intuitively transfer architectural styles. Technical solutions

[0005] An automatic identification and transfer method for architectural styles in Monte Carlo based on knowledge graphs, the method includes the following steps:

[0006] (1) Construction of the Monte Carlo architectural model library:

[0007] Typical Montenegrin architecture was surveyed using a total station with a ranging accuracy of at least 2mm + 2ppm and an angle measurement accuracy of at least 2″. Based on the data collected from the field survey, a three-dimensional model was created by dividing the distinctive structural elements of Montenegrin architecture into layers and assigning them numbers. The distinctive structural elements include 15 elements in three categories: timber frame system, felt system, and rope system. The characteristic attributes and relationships of each distinctive structural element in the three-dimensional model were calculated using a workstation equipped with an RTX 3060 graphics card. The calculation results and the three-dimensional model were entered into a model library to construct a Montenegrin architecture model library.

[0008] (2) Construction of a knowledge graph of architectural styles in Montenegro:

[0009] Each distinctive structural element in step (1) is taken as an "entity", and the relationship between each distinctive structural element and other distinctive structural elements in the model is taken as a "relationship". The two are combined to form a knowledge graph ontology of "entity-relationship-entity". Then, the distinctive structural elements, the characteristic attributes of the distinctive structural elements and the relationship between the distinctive structural elements in the Monte Carlo architectural model library are generated into a data table to construct the knowledge graph of each Monte Carlo architectural model in the Monte Carlo architectural model library.

[0010] (3) Clustering identification of architectural style genes in Montenegro:

[0011] Using a workstation equipped with an RTX 3060 graphics card, the graphs of Monte Carlo buildings in the Monte Carlo architectural model library were clustered into multiple categories. The branch and bound algorithm was used to reduce the scope of entities layer by layer according to the importance of entities in the graph structure. The optimal structure of entities in each category of graph was obtained as the core subset, forming a Monte Carlo architectural style gene library.

[0012] (4) Relocation of the architectural style of Montenegro to the target building:

[0013] Using a total station with a ranging accuracy of no less than 2mm+2ppm and an angle measurement accuracy of no less than 2″, the target building is surveyed and modeled on-site to construct a knowledge graph of the target building. Then, the structure of the knowledge graph of the target building is matched with the Monte Carlo architectural style gene library to find the matching core subset. From the Monte Carlo architectural model library, a random set of characteristic structural elements, characteristic attributes of characteristic structural elements, and the correlation between characteristic structural elements corresponding to the core subset are extracted. The current structural elements of the target building are style-transformed to realize the migration of Monte Carlo architectural style.

[0014] (5) Actual measurement and correction:

[0015] The relocated building is constructed on-site. The load-bearing capacity is measured by a universal tensile testing machine, the thermal insulation performance is measured by thermocouples, and the wind resistance performance is measured by an anemometer. The local building construction standards are then compared. If the construction standards are not met, the process is to return to step (4) and randomly select a set of structural elements, element characteristic attributes and relationships. Step (5) is repeated until the relocated building fully meets the standards.

[0016] Preferably, in step (1), the characteristic attributes and relationships of each element in the calculation model refer to the quantifiable characteristic values ​​of each entity and the quantifiable characteristic values ​​of each relationship.

[0017] Preferably, in step (2), constructing the knowledge graphs of each Monte d'I architecture in the Monte d'I architecture model library means taking the aforementioned three categories of 15 elements—wooden frame system, felt system, and rope system—as entities, combining entity attributes and relationships, and inputting them into the knowledge graph construction platform; using the platform to digitally encode the structured data, encapsulating the data into entities according to the ontology, establishing relationships between entities through algorithms and visually expressing them, and generating the final knowledge graphs of each Monte d'I architecture in the Monte d'I architecture model library.

[0018] Preferably, in step (3), the step of using the branch and bound algorithm to progressively reduce the entity range based on the importance of entities in the graph structure, and obtaining the optimal structure of entities in each category of the graph as the core subset, refers to using the branch and bound algorithm in knowledge graphs. To extract the optimal structure of entities, first determine the maximum and minimum number of entities to select, based on the entities in the graph structure. Importance reduce Given the entity scope, obtain the core subgraph, whose objective function is: ;

[0019] Where, when parameter When the value is 0, it indicates that the entity was not selected; when the parameter... When the value is 1, it indicates that the entity is selected; Indicates importance The subgraph with the largest sum is used as the core subset.

[0020] Preferably, in step (4), matching the target building knowledge graph structure through the Monte Carlo architectural style gene library refers to calling an algorithm in the knowledge graph construction platform to calculate the cosine similarity of angles between multidimensional vectors, and selecting graph structures with a cosine similarity greater than 0.8 for matching; wherein, the cosine similarity calculation formula is as follows, and the calculation result is between -1 and 1, where -1 is completely different and 1 is completely similar: .

[0021] Preferably, in step (4), extracting a random set of characteristic structural elements, characteristic structural element attribute and relationship between characteristic structural elements corresponding to the core subset of the Monte Carlo architectural model library means searching for relevant data in the knowledge graph construction platform, linking the query results to the original image of the element, and outputting the element characteristic attribute table and element relationship table in descending order of similarity.

[0022] Preferably, in step (4), the current structural elements of the target building undergo style conversion, which means that based on the CycleGAN model, using two generator networks and two discriminator networks, the query image results are converted to the target building image, and the rendering of the target building is output to realize the migration of style. Beneficial effects

[0023] 1. This invention models 15 distinctive structural elements across three categories—timber frame system, felt system, and rope system—in separate layers, calculates the characteristic attributes and relationships of each element in the model, and constructs a Montenegrin architectural model library, ensuring data standardization and accuracy. Furthermore, it utilizes a branch and bound algorithm to calculate core subsets, accurately identifying and building a library of Montenegrin architectural style genes, maximizing the accuracy and scientific rigor of style gene identification while improving the efficiency of the process.

[0024] 2. This invention calculates the characteristic attributes and relationships of each distinctive element and inputs them into a knowledge graph construction platform. It digitally encodes the structured data, encapsulates the data into entities based on the ontology, establishes the relationships between entities through algorithms, and visualizes them. Ultimately, it realizes the construction of a knowledge graph of the architectural style of Montenegro. Through the integration of multiple elements, it improves the accuracy of subsequent style recognition and migration.

[0025] 3. This invention uses knowledge graphs combined with multiple algorithms to match similar graph structures, and further utilizes the CycleGAN model to transform the query image results into target building images, outputting renderings of the target buildings, thus realizing intelligent migration of architectural styles. The whole process is more feasible and scientific, and also improves the displayability of the output results. Attached Figure Description

[0026] Figure 1 is a flowchart of the method of the present invention.

[0027] Figure 2 is a knowledge graph of the architectural styles of Montenegro constructed in this invention.

[0028] Figure 3 is a rendering of the Montenegrin architecture output by this invention. The best embodiment of the present invention

[0029] As shown in Figures 1 to 3, the method for automatic identification and transfer of architectural style patterns in Monte Carlo based on knowledge graphs includes the following steps:

[0030] (1) Construction of the Monte Carlo architectural model library:

[0031] Typical Montenegrin architecture was surveyed using a total station with a ranging accuracy of at least 2mm + 2ppm and an angle measurement accuracy of at least 2″. Based on the data collected from the field survey, a three-dimensional model was created by dividing the distinctive structural elements of Montenegrin architecture into layers and assigning them numbers. The distinctive structural elements include 15 elements in three categories: timber frame system, felt system, and rope system. The characteristic attributes and relationships of each distinctive structural element in the three-dimensional model were calculated using a workstation equipped with an RTX 3060 graphics card. The calculation results and the three-dimensional model were entered into a model library to construct a Montenegrin architecture model library.

[0032] (2) Construction of a knowledge graph of architectural styles in Montenegro:

[0033] Each distinctive structural element in step (1) is taken as an "entity", and the relationship between each distinctive structural element and other distinctive structural elements in the model is taken as a "relationship". The two are combined to form a knowledge graph ontology of "entity-relationship-entity". Then, the distinctive structural elements, the characteristic attributes of the distinctive structural elements and the relationship between the distinctive structural elements in the Monte Carlo architectural model library are generated into a data table to construct the knowledge graph of each Monte Carlo architectural model in the Monte Carlo architectural model library.

[0034] (3) Clustering identification of architectural style genes in Montenegro:

[0035] Using a workstation equipped with an RTX 3060 graphics card, the graphs of Monte Carlo buildings in the Monte Carlo architectural model library were clustered into multiple categories. The branch and bound algorithm was used to reduce the scope of entities layer by layer according to the importance of entities in the graph structure. The optimal structure of entities in each category of graph was obtained as the core subset, forming a Monte Carlo architectural style gene library.

[0036] (4) Relocation of the architectural style of Montenegro to the target building:

[0037] Using a total station with a ranging accuracy of no less than 2mm+2ppm and an angle measurement accuracy of no less than 2″, the target building is surveyed and modeled on-site to construct a knowledge graph of the target building. Then, the structure of the knowledge graph of the target building is matched with the Monte Carlo architectural style gene library to find the matching core subset. From the Monte Carlo architectural model library, a random set of characteristic structural elements, characteristic attributes of characteristic structural elements, and the correlation between characteristic structural elements corresponding to the core subset are extracted. The current structural elements of the target building are style-transformed to realize the migration of Monte Carlo architectural style.

[0038] (5) Actual measurement and correction:

[0039] The relocated building is constructed on-site. The load-bearing capacity is measured by a universal tensile testing machine, the thermal insulation performance is measured by thermocouples, and the wind resistance performance is measured by an anemometer. The local building construction standards are then compared. If the construction standards are not met, the process is to return to step (4) and randomly select a set of structural elements, element characteristic attributes and relationships. Step (5) is repeated until the relocated building fully meets the standards.

[0040] In step (1), the characteristic attributes and relationships of each element in the calculation model refer to the quantifiable characteristic values ​​of each entity in Table 1 and the quantifiable characteristic values ​​of each relationship in Table 2.

[0041] Table 1 Explanation of Quantitative Indicators for Entity Element Attributes

[0042]

[0043] Table 2 Explanation of Quantitative Indicators for Entity Attributes

[0044]

[0045] In step (2), constructing the knowledge graphs of each Monte Carlo building in the Monte Carlo architectural model library means taking the aforementioned three categories of 15 elements (wooden frame system, felt system, and rope system) as entities, combining entity attributes and relationships, and inputting them into the knowledge graph construction platform; using the platform to digitally encode the structured data, encapsulating the data into entities according to the ontology, establishing the relationships between entities through algorithms and visually expressing them, and generating the final knowledge graphs of each Monte Carlo building in the Monte Carlo architectural model library.

[0046] In step (3), the branch and bound algorithm is used to progressively reduce the scope of entities based on their importance in the graph structure, obtaining the optimal structure of entities in each category of the graph as the core subset. This refers to using the branch and bound algorithm in knowledge graphs. To extract the optimal structure of entities, first determine the maximum and minimum number of entities to select, based on the entities in the graph structure. Importance reduce Given the entity scope, obtain the core subgraph, whose objective function is: ;

[0047] Where, when parameter When the value is 0, it indicates that the entity was not selected; when the parameter... When the value is 1, it indicates that the entity is selected; Indicates importance The subgraph with the largest sum is used as the core subset.

[0048] In step (4), matching the target building knowledge graph structure through the Monte Carlo architectural style gene library refers to calling an algorithm in the knowledge graph construction platform to calculate the cosine similarity of angles between multidimensional vectors, and selecting graph structures with a cosine similarity greater than 0.8 for matching; wherein, the cosine similarity calculation formula is as follows, and the calculation result is between -1 and 1, where -1 is completely different and 1 is completely similar: .

[0049] In step (4), extracting a random set of distinctive structural elements, characteristic attributes of distinctive structural elements, and relationships between distinctive structural elements corresponding to the core subset of the Monte Carlo architectural model library means searching for relevant data in the knowledge graph construction platform, linking the query results to the original images of the elements, and sorting them from high to low similarity to output an element characteristic attribute table and an element relationship table.

[0050] In step (4), the current structural elements of the target building undergo style conversion, which means that based on the CycleGAN model, using two generator networks and two discriminator networks, the query image results are converted into the target building image, and the rendering of the target building is output to realize the transfer of style.

Claims

1. A method for automatic identification and transfer of architectural style patterns in Montenegro based on knowledge graphs, characterized in that, The method includes the following steps: (1) Construction of the Monte Carlo architectural model library: Typical Mongolian architecture was surveyed using a total station. Based on the data collected from the field survey, a three-dimensional model was created by dividing the distinctive structural elements of the Mongolian architecture into layers and assigning them numbers. The distinctive structural elements include 15 elements in three categories: timber frame system, felt system, and rope system. The characteristic attributes and relationships of each distinctive structural element in the three-dimensional model were calculated using a workstation. The calculation results and the three-dimensional model were then entered into the model library to construct a model library of Mongolian architecture. (2) Construction of a knowledge graph of architectural styles in Montenegro: Each distinctive structural element in step (1) is taken as an "entity", and the relationship between each distinctive structural element and other distinctive structural elements in the model is taken as a "relationship". The two are combined to form a knowledge graph ontology of "entity-relationship-entity". Then, the distinctive structural elements, the characteristic attributes of the distinctive structural elements and the relationship between the distinctive structural elements in the Monte Carlo architectural model library are generated into a data table to construct the knowledge graph of each Monte Carlo architectural model in the Monte Carlo architectural model library. (3) Clustering identification of architectural style genes in Montenegro: Using a workstation, the graphs of Monte Carlo buildings in the Monte Carlo architectural model library are clustered into multiple categories. The branch and bound algorithm is used to reduce the scope of entities layer by layer according to the importance of entities in the graph structure. The optimal structure of entities in each category is obtained as the core subset, forming a gene library of Monte Carlo architectural style. (4) Relocation of the architectural style of Montenegro to the target building: Using a total station, the target building is surveyed and modeled on-site to construct a knowledge graph of the target building. Then, the structure of the knowledge graph of the target building is matched with the Monte Carlo architectural style gene library to find the core subset that matches it. A random set of characteristic structural elements, characteristic attributes of characteristic structural elements, and the relationship between characteristic structural elements are extracted from the Monte Carlo architectural model library. The current structural elements of the target building are style-transformed to realize the migration of Monte Carlo architectural style. (5) Actual measurement and correction: The relocated building is constructed on-site. The load-bearing capacity is measured by a universal tensile testing machine, the thermal insulation performance is measured by thermocouples, and the wind resistance performance is measured by an anemometer. The local building construction standards are then compared. If the construction standards are not met, the process is to return to step (4) and randomly select a set of structural elements, element characteristic attributes and relationships. Step (5) is repeated until the relocated building fully meets the standards.

2. The method for automatic identification and transfer of Monte Carlo architectural style based on knowledge graph as described in claim 1, characterized in that, In step (1), the workstation calculates the characteristic attributes and relationships of each feature structure element in the three-dimensional model, which means that the quantifiable feature values ​​of each entity and the quantifiable feature values ​​of each relationship can be used.

3. The method for automatic identification and transfer of Monte Carlo architectural style based on knowledge graph as described in claim 1, characterized in that, In step (2), constructing the knowledge graph of each Monte architecture in the Monte architecture model library means taking the aforementioned three categories of 15 elements, namely the wooden frame system, the felt system, and the rope system, as entities and inputting them into the knowledge graph construction platform in combination with entity attributes and relationships. The platform is used to digitally encode structured data, encapsulate the data into entities based on ontology, establish relationships between entities through algorithms and visualize them, and generate knowledge graphs of various Monte Carlo buildings in the final Monte Carlo architectural model library.

4. The method for automatic identification and transfer of Monte Carlo architectural style based on knowledge graph as described in claim 1, characterized in that, In step (3), the branch and bound algorithm is used to progressively reduce the scope of entities based on their importance in the graph structure, obtaining the optimal structure of entities in each category of the graph as the core subset. This refers to using the branch and bound algorithm in knowledge graphs. To extract the optimal structure of entities, first determine the maximum and minimum number of entities to select, based on the entities in the graph structure. Importance reduce Given the entity scope, obtain the core subgraph, whose objective function is: ; Where, when parameter When the value is 0, it indicates that the entity was not selected; when the parameter... When the value is 1, it indicates that the entity is selected; Indicates importance The subgraph with the largest sum is used as the core subset.

5. The method for automatic identification and transfer of Monte Carlo architectural style based on knowledge graph as described in claim 1, characterized in that, In step (4), matching the target building knowledge graph structure through the Monte Carlo architectural style gene library refers to calling an algorithm in the knowledge graph construction platform to calculate the cosine similarity of angles between multidimensional vectors, and selecting graph structures with a cosine similarity greater than 0.8 for matching; wherein, the cosine similarity calculation formula is as follows, and the calculation result is between -1 and 1, where -1 is completely different and 1 is completely similar: .

6. The method for automatic identification and transfer of Monte Carlo architectural style based on knowledge graph as described in claim 1, characterized in that, In step (4), extracting a random set of distinctive structural elements, characteristic attributes of distinctive structural elements, and relationships between distinctive structural elements corresponding to the core subset of the Monte Carlo architectural model library means searching for relevant data in the knowledge graph construction platform, linking the query results to the original images of the elements, and sorting them from high to low similarity to output an element characteristic attribute table and an element relationship table.

7. The method for automatic identification and transfer of Monte Carlo architectural style based on knowledge graph as described in claim 1, characterized in that, In step (4), the current structural elements of the target building undergo style conversion, which means that based on the CycleGAN model, using two generator networks and two discriminator networks, the query image results are converted into the target building image, and the rendering of the target building is output to realize the transfer of style.

Images