A method and system for intelligent generation of new building models in historical and cultural districts

Abstract

This invention provides an intelligent generation method and system for new building models in historical and cultural blocks. The method includes: acquiring data on historical buildings in a target block to obtain a standard dataset of historical buildings based on individual building data; inputting this dataset into a display characteristic extraction model to output the dominant morphological genes of the components contained in each historical building; extracting the recessive quantitative genes of the components from the components contained in the individual historical building data; constructing a gene-parameter mapping rule base based on the architectural genes of historical buildings and the planning indicator constraints of the historical and cultural block to be built; constructing a parametric control model based on the gene-parameter mapping rule base; and generating a new building model of the historical and cultural block to be built by perturbing the parameters based on the parametric control model. The fusion of the dominant morphological genes and the recessive quantitative genes of the components achieves complementarity of geometric and textural information, laying the foundation for accurate mapping of architectural genes and parameters.

Description

Technical Field

[0001] This invention relates to the field of architectural design for new buildings in historical and cultural districts, and specifically to an intelligent method and system for generating architectural models for new buildings in historical and cultural districts. Background Technology

[0002] Historic and cultural districts are the core carriers of urban cultural heritage. Their spatial forms and architectural styles embody the historical memories and cultural characteristics of a specific region, making them an indispensable and important component of sustainable urban development. In recent years, as my country's urbanization process has entered the stock renewal stage, the protection and revitalization of historic and cultural districts has become a key issue in urban construction. Against this backdrop, how to achieve organic synergy between new buildings and existing historical features, avoid the homogenization problem of uniformity across cities, and simultaneously meet the requirements of modern living functions and planning indicators has become a core challenge for planners and architects.

[0003] In the process of developing this invention, the applicant discovered at least the following problems in the prior art: Currently, the style control of newly constructed buildings in historical and cultural blocks generally faces a dual dilemma: rigid constraints and vague flexible guidance. Traditional planning often relies on macro-indicators for rigid restrictions, such as specifying quantitative parameters like building height, floor area ratio, and setback distance. While this can control building volume to some extent, it is difficult to accurately grasp the core characteristics of architectural style, resulting in some new buildings having compliant volume but jarring style. On the other hand, flexible guidance based on expert experience, such as qualitative requirements like style coordination and textural integration, often suffers from strong subjectivity and poor implementation due to the lack of unified and quantifiable evaluation standards and implementation basis, making it difficult to effectively guide architectural design practice. Summary of the Invention

[0004] This invention provides an intelligent generation method and system for new building models in historical and cultural blocks, which can solve the above-mentioned technical problems in the prior art.

[0005] To achieve the above objectives, in a first aspect, embodiments of the present invention provide an intelligent generation method for new building models in historical and cultural districts, comprising: Acquire data on historical buildings in the target block, preprocess the data on historical buildings in the target block, and obtain a standard dataset of historical buildings based on individual historical building data; The fused images from the standard dataset of historical buildings are input into the display feature extraction model, which outputs the dominant morphological genes of the components contained in the individual historical buildings; the recessive quantitative genes of the components are extracted from the components contained in the data of individual historical buildings; and the style genes of historical buildings are constructed, which include the dominant morphological genes and the recessive quantitative genes of the components. The architectural features of historical buildings are encoded to obtain encoded architectural features, and an architectural features gene library is constructed based on the encoded architectural features. Based on the architectural features of historical buildings and the planning indicators of the proposed historical and cultural blocks, a gene and parameter mapping rule base is constructed. The gene and parameter mapping rule base includes the construction of qualitative mapping rules, quantitative mapping rules and boundary mapping rules. A parameterized control model is constructed based on a gene-parameter mapping rule base. Based on the parametric control model, a new building model of the historical and cultural block to be built is generated by perturbing the parameters.

[0006] Secondly, embodiments of the present invention provide an intelligent generation system for new building models in historical and cultural blocks, comprising: The data acquisition and processing unit is used to acquire data on historical buildings in the target block, preprocess the data on historical buildings in the target block, and obtain a standard dataset of historical buildings based on individual historical building data. The landscape gene extraction unit is used to input the fused images from the standard dataset of historical buildings into the display feature extraction model, output the dominant morphological genes of the components contained in the individual historical buildings; extract the recessive quantitative genes of the components from the components contained in the individual historical building data; and construct the landscape gene of the historical buildings, which includes the dominant morphological genes and the recessive quantitative genes of the components. The feature gene bank construction unit is used to encode the feature genes of historical buildings, obtain the encoded feature genes, and construct the feature gene bank based on the encoded feature genes. The mapping unit is used to construct a gene and parameter mapping rule base based on the style genes of historical buildings and the planning index constraints of the historical and cultural blocks to be built. The gene and parameter mapping rule base includes the construction of qualitative mapping rules, quantitative mapping rules and boundary mapping rules. A parameterized control model building unit, used to build parameterized control models based on a gene-parameter mapping rule base; The building model generation unit is used to generate new building models of historical and cultural blocks to be built based on parametric control models by perturbing the parameters.

[0007] The above technical solution has the following beneficial effects: the fusion of the dominant morphological genes and the recessive quantitative genes of the components achieves the complementarity of geometric and textural information, avoids feature fragmentation, and lays the foundation for the accurate mapping of appearance genes and parameters. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart of a method for intelligently generating models of newly constructed buildings in historical and cultural districts; Figure 2 It is a structural diagram of an intelligent system for generating models of newly constructed buildings in historical and cultural districts; Figure 3 This is a flowchart of the intelligent building generation process based on landscape gene extraction and parameterized control according to an embodiment of the present invention. Figure 4 This is a flowchart of the appearance gene extraction and encoding process according to an embodiment of the present invention; Figure 5 Constructing and mapping diagrams for parameterized rules; Figure 6 This is a block diagram of the landscape synergy quantitative evaluation index system according to an embodiment of the present invention; Figure 7 This is a block diagram of the system device configuration according to an embodiment of the present invention. Detailed Implementation

[0010] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0011] like Figure 1 As shown, in conjunction with embodiments of the present invention, a method for intelligently generating new building models in historical and cultural districts is provided, comprising: S101: Obtain data on historical buildings in the target block, preprocess the data on historical buildings in the target block, and obtain a standard dataset of historical buildings based on individual historical building data; S102: Input the fused images from the standard dataset of historical buildings into the display feature extraction model, and output the dominant morphological genes of the components contained in the individual historical buildings; extract the recessive quantitative genes of the components from the components contained in the individual historical building data; construct the style genes of historical buildings, which include the dominant morphological genes and the recessive quantitative genes of the components. S103: Encode the style genes of historical buildings to obtain the encoded style genes, and construct a style gene library based on the encoded style genes. S104: Construct a gene and parameter mapping rule base based on the style genes of historical buildings and the planning index constraints of the historical and cultural blocks to be built. The gene and parameter mapping rule base includes the construction of qualitative mapping rules, quantitative mapping rules and boundary mapping rules. S105: Constructing a parameterized control model based on a gene-parameter mapping rule base; S106: Based on the parametric control model, a new building model of the historical and cultural block to be built is generated by perturbing the parameters.

[0012] Overall flowchart as follows Figure 3 The diagram shown is a flowchart of intelligent building generation based on landscape gene extraction and parameterized control.

[0013] S101: Obtain data on historical buildings in the target block, preprocess the data to obtain a standard dataset of historical buildings based on individual building data, including: S101-1: Employs a RIEGL VZ-4000 terrestrial lidar with a point cloud density of 100 points / m², a scanning accuracy of 0.001m, and a maximum scanning distance of 2000m. The system is set to scan the facades and roofs of all historical buildings in the target block at a 100m scanning distance, generating lidar point cloud data. Scanning stations are set at 5m intervals to ensure unobstructed views. The GPS coordinates of each lidar station are recorded simultaneously (accuracy ±0.05m). It supports all-weather operation and solves the obstruction problem in complex streetscapes.

[0014] S101-2: Using a drone equipped with a full-frame camera, images of the facades of all historical buildings in the target block are captured along a planned flight path, forming drone image data. Simultaneously, the location coordinates of the facades are obtained through positioning. Specifically, a DJI Matrice 350 RTK drone, equipped with a Zenmuse P1 full-frame camera, is used. The flight altitude is 50m, with a forward overlap of 85% and a lateral overlap of 80%. The image resolution is 0.01m / pixel, and the pixel count is 50 megapixels, clearly capturing the texture and color information of the building facades. At least three images from different angles are obtained for each facade. RTK positioning (positioning accuracy ±1cm) is simultaneously enabled to obtain the geographic coordinates of the facades corresponding to the images.

[0015] S101-3: Collect historical textual data on historical buildings in the target neighborhood, including at least one of the following: survey reports, repair drawings, and planning documents, to complete the historical data. This data is used to extract attribute information such as the building's construction date and style, for subsequent gene classification analysis.

[0016] S101-4: Construct data on the historical buildings of the target block, including laser point cloud data, drone imagery data, and historical text data.

[0017] The data of historical buildings in the target block are sequentially subjected to point cloud denoising, image correction, and point cloud-image fusion to generate textured 3D point cloud models of the historical buildings in the target block. The textured 3D point cloud models are then automatically segmented into individual buildings to generate fused images of each historical building. These fused images of individual historical buildings form a standard dataset of historical buildings. Specifically: First, point cloud noise reduction: Ground and high-altitude noise are removed by direct filtering (Z-axis range 0.5-50m); isolated noise points are removed by statistical filtering (setting the number of neighboring points to 50 and the standard deviation multiple to 1.5), so that the point cloud noise rate is less than 0.1%.

[0018] Second, image correction: The camera intrinsic parameter matrix calibration is used to correct the distortion of the images in the UAV image data, as well as georegistration. Georegistration refers to converting the geographic coordinates of the corresponding elevation of the image from RTK coordinates to WGS84 coordinate system to ensure that the geographic coordinates of the image are consistent with the geographic coordinates of the point cloud.

[0019] Third, point cloud and image fusion: Common feature points, such as wall corners and eaves, are extracted from the laser point cloud data and the UAV image data. Mismatched points are removed using the RANSAC algorithm (matching accuracy ≥95%). Then, coordinate transformation maps the image texture from the UAV image data to the laser point cloud surface, generating a textured 3D point cloud model of the historical building. This achieves the fusion of the 3D geometric information of the laser point cloud data with the texture information of the UAV image data.

[0020] Fourth, data standardization and segmentation: The textured 3D point cloud models of historical buildings are unified to the WGS84 coordinate system with a resolution of 0.01m. A region-growing-based segmentation algorithm (with a growth threshold of 0.1m) is used to automatically segment the textured 3D point cloud models of historical buildings by individual buildings, outputting fused images of individual historical buildings to form a standard dataset of historical buildings (standardized fused dataset). The naming rule for the individual historical building dataset is: block number-building number-data type to ensure traceability.

[0021] By collecting and preprocessing multi-source data, a standard dataset is formed, providing high-quality data with "geometric accuracy and complete texture" for subsequent gene extraction.

[0022] like Figure 4 This is a flowchart of the process for extracting and encoding features genes.

[0023] In S102, the fused images from the standard dataset of historical buildings are input into the display feature extraction model, which outputs the dominant morphological genes of the components contained in the individual historical buildings, including: The fused images of individual historical buildings are input into the display feature extraction model, which outputs the dominant morphological genes of the components contained in each building. These genes include the core feature category, location coordinates, and confidence level. Core features with a confidence level greater than a threshold are retained. These core features include at least one of the following: gable roof, hip roof, brick facade, stone facade, arched window, rectangular window, column style, and decorative elements. Specifically, 100 building samples from typical historical districts are selected, and core features are labeled according to their characteristics, generating 50,000 labeled images to form a labeled dataset. The labeled data is divided into training, validation, and test sets in a 7:2:1 ratio. The YOLOv8 model is trained on the training set (100 iterations, learning rate 0.01). After training, the display feature extraction model achieves a recognition accuracy of over 92%, significantly higher than the accuracy of existing manual labeling and coarse extraction techniques.

[0024] The fused images from the standard dataset of historical buildings are input into the display feature extraction model, which outputs the dominant morphological genes of the components contained in the individual historical buildings. The dominant morphological genes of the components include the core feature category, the location coordinates of the core feature, and the confidence level. Core features with a confidence level greater than the confidence level threshold (e.g., confidence level ≥ 0.8) are included in the landscape gene library.

[0025] In S102, recessive quantitative genes of components are extracted from the components contained in the data of individual historical buildings, including: S102-2: Segment the point cloud subsets of components contained in the historical building unit data and calculate the geometric quantization parameters of the components contained in the historical building unit. Specifically, using the PointNet++ point cloud segmentation algorithm, segment the point cloud subsets of components contained in the historical building unit data, such as the point cloud subsets of roof, facade, and doors and windows, and calculate the geometric quantization parameters of the components contained in the historical building unit, such as facade height-to-width ratio (facade height / width), roof slope (angle between the roof slope and the horizontal plane, calculated through the point cloud normal vector), door and window ratio (door and window height / width), and component spacing (such as column spacing and window spacing). The calculation accuracy reaches ±0.02m, which is far higher than the accuracy of manual annotation and rough extraction in existing technologies.

[0026] S102-3: Extract the component areas contained in the data of individual historical buildings, convert the construction areas to the HSV color space, extract the mean and standard deviation of hue, saturation and brightness from the HSV color space, and use the mean and standard deviation of hue, saturation and brightness of the components as the color quantification parameters of the components contained in the individual historical buildings.

[0027] S102-4: The stylistic genes of historical buildings include dominant morphological genes and implicit quantitative genes of components. The implicit quantitative genes of components include geometric quantitative parameters, color quantitative parameters, and material quantitative parameters. Specifically, material types are identified through texture features (energy and entropy values ​​of the gray-level co-occurrence matrix) (e.g., the texture entropy value range of blue bricks is 0.8-1.2, and that of stone is 1.3-1.8), thus forming the material quantitative parameters of the components. Extracting implicit quantitative genes accurately reflects the intrinsic characteristics of historical style, making the stylistic expression more complete and systematic.

[0028] The fusion of the dominant morphological genes and the implicit quantitative genes of components achieves complementarity of geometric and textural information, avoiding feature fragmentation and laying the foundation for accurate mapping of style genes and parameters. This solves the technical problem in existing technologies that "only intuitive surface features such as roof form and material are extracted without deeply exploring the intrinsic relationships behind these features (such as facade height-to-width ratio, component size proportions, and color brightness ranges), resulting in the inability to form inherited style genes, leading to fragmented extracted features and difficulty in supporting in-depth style synergy."

[0029] S103: Encode the stylistic features of historical buildings to obtain the encoded stylistic features genes, and construct a stylistic features gene library based on the encoded stylistic features genes, including: S103-1: Encoding the stylistic genes of historical buildings. The gene encoding includes a core feature category code, a sub-category code, and quantification parameters included in the implicit quantification genes, totaling 16 characters. For example, "double-pitched roof + 30° slope" is encoded as WT01-01-030.0, where WT represents the roof type, W is the roof itself, T is the type, 01 is the double-pitched roof sub-code, and 030.0 is the slope quantification value; "blue brick facade + aspect ratio 1.5 + hue 200°" is encoded as EL01-01-1.5_200, where E represents the facade, L is the type, 01 is the blue brick sub-code, 1.5 is the aspect ratio, and 200 is the hue value. Encoding ensures that each stylistic gene is directly mapped to a parameter.

[0030] S103-2: Construct a basic information table for historical buildings, which includes the block number, building number, building type, and building year; S103-3: Construct a dominant morphogenetic gene table based on dominant morphogenetic genes. The dominant morphogenetic gene table includes building number, core feature category, gene code, and core feature location coordinates. S103-4: Construct a recessive quantitative gene table based on recessive quantitative genes. The recessive quantitative gene table includes building number, quantitative parameters and calculation precision. Save the basic information table of historical buildings that form associations using building numbers, the dominant morphological gene table and the recessive quantitative gene table to the landscape gene bank.

[0031] S103-4: Construct a recessive quantification gene table based on recessive quantification genes. The recessive quantification gene table includes building number, quantification parameters and calculation precision. S103-5: Save the basic information table of historical buildings, the dominant morphological gene table, and the recessive quantitative gene table, which are associated using building numbers, to the landscape gene database. Specifically, the landscape gene database is constructed using a MySQL database.

[0032] The system enables linked queries across three tables via building numbers, supporting the retrieval of landscape genes by street block, building type, and core feature type. This achieves a digital representation of landscape genes. Verification is performed through expert review and comparison with historical data. If verification is successful, a landscape gene database is output; otherwise, it is re-extracted. This ensures the accuracy of gene extraction.

[0033] S104: Constructing a rule base for mapping genes and parameters based on the stylistic features of historical buildings and the planning indicators of the proposed historical and cultural blocks, including: Figure 5 This paper constructs a parameterized rule construction and mapping framework, assigns corresponding gene weights to the style genes in the style gene library, and builds a gene-parameter mapping rule library based on the style genes of historical buildings and the planning indicator constraints of the proposed historical and cultural blocks. Specifically, this includes constructing qualitative mapping rules, quantitative mapping rules, and boundary mapping rules. Based on the gene-parameter mapping rule library, a parameterized control model and rule set are constructed, serving as the core bridge connecting gene extraction and building generation. The planning indicator constraints of the proposed historical and cultural blocks include, for example, height and floor area ratio.

[0034] The association between dominant morphological genes and recessive quantitative genes was established, forming a morphological gene system with inheritance and systematicity, realizing the leap from feature accumulation to gene extraction, and providing a complete core basis for subsequent parameterized control.

[0035] S104-1: Construct qualitative mapping rules to convert dominant morphological genes into morphological type parameters for the parameterized control model. For example, mapping a gable roof to Grasshopper is: the roof type parameter is gable, and it is associated with a preset roof ridge generation logic; mapping an arched window to Grasshopper is: the window / door style parameter is arched, and the arched outline generation algorithm is automatically invoked. The mapping rules are stored in XML files with the format <gene code, parameter name, parameter value, association algorithm>. Automatic parsing and matching of gene codes to Grasshopper parameters is achieved through XML and Python scripts.

[0036] S104-2: Construct quantitative mapping rules to directly convert recessive quantified genes into numerical parameters for the parameterized control model. These numerical parameters include dimensional parameters, which are optimized using gene weights. For example, a facade height-to-width ratio of 1.5 (weight 0.2) and an adjacent building height-to-width ratio of 1.6 (weight 0.1) are used to calculate the new building's height-to-width ratio parameter as 1.53 using a weighted average. A roof slope of 30° (weight 0.25) is directly used as the new building's roof slope parameter. Numerical parameters are preserved to two decimal places to ensure model accuracy.

[0037] The study establishes a three-layer structure: a target layer (style coordination), a criterion layer (roof, facade, doors and windows, materials), and an indicator layer (specific style genes). Five architecture and planning experts were invited to score the importance of each style gene using the Analytic Hierarchy Process (AHP) to generate a judgment matrix, which serves as the first sub-weight. The information entropy of each style gene in the street sample was calculated using the entropy weight method, which is the second sub-weight; the higher the frequency and the lower the information entropy value, the higher the second sub-weight. The first and second sub-weights are then combined in a 6:4 ratio to obtain the weight of the style gene. For example, in a certain blue brick street, the weight of the blue brick facade gene is 0.25, and the weight of the double-pitched roof is 0.20.

[0038] S104-3: Construct boundary mapping rules to transform the planning constraints of the proposed historical and cultural district into boundary parameters of the parametric control model. For example, a building height ≤ 15m corresponds to a maximum Z-axis coordinate ≤ 15, a plot ratio ≤ 1.2 is transformed into a volume constraint parameter through building volume / land area ≤ 1.2, and a setback distance ≥ 3m corresponds to a collision detection parameter where the distance between the building outline and the land boundary line is ≥ 3m. Boundary parameters are embedded into the model as hard constraints to ensure that the generated scheme does not violate planning requirements.

[0039] The parameters of a parametric control model include: morphological parameters, numerical parameters, and boundary parameters.

[0040] S104-4: Save the qualitative mapping rules, quantitative mapping rules, and boundary mapping rules to the gene and parameter mapping rule library.

[0041] Three types of mapping rules are constructed to achieve seamless integration between landscape genes and parametric design software (Grasshopper). This enables fully automated conversion of genes into numerical parameters of the parametric control model, improving conversion efficiency by more than 10 times. It avoids parameter biases caused by the strong subjectivity of human experience, ensuring the consistency of landscape feature transmission. It supports the addition of new gene and parameter types, adapting to historical blocks of different regions and styles (such as Jiangnan water towns and northern courtyard houses), demonstrating good versatility. This improved efficiency ensures the accuracy and consistency of landscape feature transmission, resulting in stable landscape synergy effects.

[0042] S105: Constructing a parameterized control model based on a gene-parameter mapping rule base, including: Based on the Grasshopper platform, a control model is built using a combination of Python scripts and visual programming. The parameterized control model includes a data interface module, a parameter-driven module, a collision detection module, and an output module. S105-1: Construct a data interface module. Within the data interface module, read the landscape gene database data through the MySQL API and read the planning indicator constraints of the historical and cultural blocks to be built through the Excel interface. Automatically parse the read data and match it with the mapping rules corresponding to the gene and parameter mapping rule library. Transform the read data through the corresponding mapping rules to obtain the parameters of the parameterized control model.

[0043] S105-2: Construct a parameter-driven module to transform the parameters of the parameterized control model, including morphological parameters (roof type, facade segmentation, etc.), dimensional parameters (height, width, etc.), and boundary parameters (setback, floor area ratio, etc.). Through the parameter-driven module, the parameters of the parameterized control model are associated with the parameter generation and update logic of the parameterized control model. When the landscape gene pool data and the planning indicator constraints of the historical and cultural blocks to be built are updated, the parameterized control model is updated in real time.

[0044] S105-3: Construct a collision detection module. The collision detection module is used to detect the collision relationship between the generated building model of the historical and cultural block to be built and the land boundary line and adjacent buildings in real time. If the boundary parameters are violated, the size parameters will be automatically adjusted (such as reducing the height and shrinking the plane).

[0045] S105-4: Construct the output module. The output module exports the generated parametric model to obtain a 3D architectural model of the historical and cultural block to be built. The format is Rhino 3D model (.3dm). At the same time, it outputs a parameter list (.xlsx). The parameter list records the source and gene weight of the landscape genes corresponding to each parameter.

[0046] The model integrates landscape genes, gene weights, and planning indicators (height, floor area ratio, etc.) into the generation logic of the parametric control model, achieving integrated control with parameter-driven, real-time updates, and constraint verification. The parametric control model is verified; if verification passes, it is output; otherwise, it is returned to the gene and parameter mapping rule base for readjustment.

[0047] S106: Based on a parametric control model, a new building model for the proposed historical and cultural district is generated by perturbing the parameters, including: S106-1: The perturbation range of parameters is determined by the quantitative parameters and gene weights included in the recessive quantification genes. The perturbation range for parameters corresponding to high-weight genes (such as facade material parameters of the core block) is ±5%; for medium-weight parameters (such as roof slope), the perturbation range is ±10%; and for low-weight parameters (such as window spacing), the perturbation range is ±15%, ensuring the differentiation of the schemes and not deviating from the core style. Using the Latin hypercube sampling algorithm, multiple sets (e.g., 50 sets) of non-repeating parameter combinations are randomly generated within one hour within the parameter perturbation range. Each set of parameter combinations is input into the parametric control model to generate the corresponding 3D architectural model of the historical and cultural block to be built, while simultaneously recording the gene matching of parameters within each set. Generation efficiency is improved by more than 50 times. The parameter perturbation range is linked to gene weights, ensuring consistency in core style features (such as materials and roofs) and achieving differentiation in secondary features (such as window spacing and decoration) while maintaining diversity, avoiding homogenization or style deviation. There is no need to manually construct multiple sets of parameters, reducing repetitive work and lowering the threshold for historical block architectural design. S106-2: A preliminary screening of the 3D architectural models of the proposed historical and cultural blocks is conducted. Models violating boundary parameters are eliminated, and the remaining models are used as candidate models. Candidate models meeting the evaluation requirements are then used as the new building models for the proposed historical and cultural blocks. Specifically, 3D architectural models violating hard constraints (i.e., exceeding height limits or floor area ratio limits) are eliminated, retaining approximately 30 sets of 3D architectural models as candidate models. The collision detection module verifies the compliance of the scheme in real time based on hard constraints, ensuring that the generated scheme is stylistically harmonious and meets the standards, achieving joint control of style and planning. Gene weight calculation gives core genes (such as blue brick facades) a higher proportion in parameter allocation, prioritizing the matching of the dominant style of the block in the generated model, avoiding style ambiguity caused by averaged feature matching. The parameter-driven module supports single-parameter fine-tuning and multi-parameter combination adjustment, with a response speed ≤10 seconds / time. Secondary optimization can be performed based on the generated results, balancing automation and creative needs, and improving the model's controllability and flexibility.

[0048] By constructing a method that satisfies planning constraints (building height, floor area ratio, land use, etc.) and architectural style parameters, the system automatically adjusts architectural details such as facade subdivision, roof combination, and component arrangement while meeting rigid planning indicators. This results in architectural schemes that are differentiated yet consistent with the surrounding architectural style, providing a wide range of choices.

[0049] Figure 6 The framework diagram of the quantitative evaluation index system for landscape coordination is shown. The evaluation index candidate model is used for scoring (total score 100 points). The weights of the evaluation indexes are gene matching degree (40 points), proportional coordination (30 points), color similarity (20 points), and planning compliance (10 points). Each type of index includes sub-items (such as gene matching degree including core and secondary gene matching rates).

[0050] The matching rate between the candidate model and the core genes of the neighborhood is calculated using a gene matching index. The score is calculated as follows: core gene (weight ≥ 0.2) matching rate × 25 points + minor gene (weight 0.1-0.2) matching rate × 15 points. The matching rate = number of matched genes / total number of genes × 100%. For example, if the core gene matching rate is 90% and the minor gene matching rate is 80%, the score is 90% × 25 + 80% × 15 = 34.5 points.

[0051] The deviation rate between the candidate model and the street's genetic parameters is calculated using the proportional coordination index. Deviation rate = |scheme parameters - average genetic parameters| / average genetic parameters × 100%. For key proportions such as facade height-to-width ratio and roof slope, an average deviation rate ≤ 5% earns 30 points, 5%-10% earns 20 points, 10%-15% earns 10 points, and > 15% earns 0 points.

[0052] The deviation between the candidate model and the average color of the street is calculated using the color similarity index (based on the HSV color space). A hue deviation ≤ 5° and a saturation deviation ≤ 5% earn 20 points, a hue deviation of 5°-10° or a saturation deviation of 5%-10% earns 10 points, and anything exceeding that earns 0 points. Material texture similarity is calculated using cosine similarity, and a similarity ≥ 0.8 earns 5 points.

[0053] 10 points are awarded for fully complying with all planning indicators; 0 points are awarded for violating one hard constraint (such as exceeding the height limit); and 5 points are awarded for violating one soft constraint (such as slightly insufficient setback).

[0054] Candidate schemes are ranked from highest to lowest based on their total score, and the top 10 are selected as candidate schemes. Optimization suggestions are provided for indicators with lower scores. For example, if the color similarity score is low, it is recommended to adjust the facade color angle from 220° to 200° (the average color angle of the block); or if the proportion coordination is insufficient, it is recommended to adjust the facade height-to-width ratio from 1.8 to 1.5. A report is output, including a complete report with a list of schemes (including ranking and scores), a 3D model file, a parameter list, and optimization suggestions, supporting secondary adjustments. The comprehensiveness and objectivity of the evaluation are ensured, achieving a closed loop of scheme generation, evaluation, and optimization.

[0055] Transforming architectural style harmony into a 100-point quantitative index avoids the drawbacks of relying on subjective expert judgment and provides a quantitative basis for planning approval. Optimization suggestions are directly linked to the gene parameters corresponding to low-scoring indicators, making modifications to the building model more targeted and avoiding blind adjustments. Optimized schemes improve scores by an average of 15-20 points. The closed-loop system ensures the generation process continuously approaches the optimal architectural style solution. The top 10 building models ultimately meet the requirements of high architectural style harmony and high compliance, improving the quality of architectural style control for new buildings in historical districts. Through automatic evaluation of the quantitative assessment of architectural style harmony, the system automatically quantifies and ranks the architectural style harmony effects of generated schemes, providing objective and quantifiable decision-making basis for planning approval and design optimization, and improving the accuracy and scientific nature of architectural style control.

[0056] Appendix Figure 7 The system architecture is illustrated in the block diagram. Hardware components include data acquisition equipment (H1) comprising a LiDAR and UAV aerial surveying system; data processing equipment including a high-performance workstation and storage server for algorithm computation and data storage; and output devices including a plotter and a 3D printer for outputting the solution results. The high-performance workstation is equipped with an i9-13900K CPU and an RTX 4090 GPU, supporting parallel computing and improving the feature recognition speed of the YOLOv8 model to 0.5 seconds per image and the parametric model generation speed to 10 seconds per solution, demonstrating high processing speed. The software system includes data processing software, AI algorithm software, parametric design software, and evaluation and management software. Point cloud data is exported to Matlab via CloudCompare's Python API for point cloud segmentation and parameter calculation. A Python script in Grasshopper calls a PyTorch-trained model for real-time import of feature recognition results. Real-time interaction between gene bank data and the parametric model is achieved via MySQL Connector / Python. An Excel VBA script automatically reads the solution parameters, calls the Python-written evaluation algorithm to calculate scores, and generates a visual evaluation report. It achieves full automation from data collection to solution output, enabling intelligent generation and evaluation of multiple solutions without human intervention.

[0057] like Figure 4 As shown, in conjunction with embodiments of the present invention, an intelligent generation system for new building models in historical and cultural districts is provided, comprising: The data acquisition and processing unit 21 is used to acquire data on historical buildings in the target block, preprocess the data on historical buildings in the target block, and obtain a standard dataset of historical buildings based on individual historical building data. The landscape gene extraction unit 22 is used to input the fused image from the standard dataset of historical buildings into the display feature extraction model, output the dominant morphological genes of the components contained in the individual historical buildings; extract the recessive quantitative genes of the components from the components contained in the individual historical building data; and construct the landscape gene of the historical buildings, which includes the dominant morphological genes of the components and the recessive quantitative genes of the components. The feature gene bank construction unit 23 is used to encode the feature genes of historical buildings, obtain the encoded feature genes, and construct the feature gene bank based on the encoded feature genes. Mapping unit 24 is used to construct a gene and parameter mapping rule base based on the style genes of historical buildings and the planning index constraints of the historical and cultural blocks to be built. The gene and parameter mapping rule base includes constructing qualitative mapping rules, quantitative mapping rules and boundary mapping rules. The parameterized control model construction unit 25 is used to construct a parameterized control model based on a gene-parameter mapping rule base. Building model generation unit 26 is used to generate new building models of the historical and cultural blocks to be built by perturbing the parameters based on the parametric control model.

[0058] Preferably, the data acquisition and processing unit 21 is specifically used for: Using ground-based lidar at a preset point cloud density of 100 points / m², the facades and roofs of all historical buildings in the target block are scanned to generate lidar point cloud data. Using a full-frame camera mounted on a drone, images of the facades of all historical buildings in the target block are taken along a planned flight path, forming drone image data; at the same time, the location coordinates of the facades are obtained through the positioning function. Collect historical textual data on historical buildings in the target neighborhood. The historical textual data shall include at least one of the following: census report, repair drawings, and planning documents. The data for the historical buildings in the target block is constructed, including laser point cloud data, drone imagery data, and historical text data. The data of historical buildings in the target block are sequentially subjected to point cloud denoising, image correction, and point cloud fusion with images to generate textured 3D point cloud models of historical buildings in the target block. The textured 3D point cloud models are automatically segmented into individual buildings to generate fused images of individual historical buildings. A standard dataset of historical buildings is formed by fused images of individual historical buildings.

[0059] Preferably, the appearance gene extraction unit includes: The dominant morphological gene extraction unit is used to input the fused image of historical building individual data into the display feature extraction model and output the dominant morphological genes of the components contained in the historical building individual. The dominant morphological genes of the components include the core feature category, the location coordinates of the core feature and the confidence level. Core features with a confidence level greater than the confidence level threshold are retained. Among them, the core features include at least one of the following: double-sloped roof, four-sloped roof, blue brick facade, stone facade, arched window, rectangular window, column and floral decoration. The recessive quantization gene extraction unit is used to segment the point cloud subset of the components contained in the data of individual historical buildings and calculate the geometric quantization parameters of the components contained in the individual historical buildings. The component areas contained in the data of individual historical buildings are extracted, the construction areas are converted to the HSV color space, and the mean and standard deviation of hue, saturation and brightness are extracted from the HSV color space. The mean and standard deviation of hue, saturation and brightness of the components are used as the color quantification parameters of the components contained in the individual historical buildings. The implicit quantization genes of a component are constructed, which include the component's geometric quantization parameters, color quantization parameters, and material quantization parameters.

[0060] Preferably, the landscape gene bank construction unit 23 is specifically used for: The architectural features of historical buildings are encoded, including core feature category codes, sub-category codes, and quantitative parameters included in the recessive quantitative genes. Construct a basic information table for historical buildings, which includes the block number, building number, building type, and building year. A dominant morphogenetic gene table was constructed based on dominant morphogenetic genes. The dominant morphogenetic gene table includes building number, core feature category, gene code, and core feature location coordinates. A recessive quantitative gene table was constructed based on recessive quantitative genes. The recessive quantitative gene table includes building number, quantitative parameters, and calculation precision. The basic information table of historical buildings that form associations using building numbers, the dominant morphological gene table, and the recessive quantitative gene table were saved to the landscape gene bank.

[0061] Preferably, the mapping unit 24 is specifically used for: A qualitative mapping rule is constructed, which is used to transform dominant morphological genes into morphology type parameters of the parameterized control model; A quantitative mapping rule is constructed to directly convert recessive quantified genes into numerical parameters of a parameterized control model. The numerical parameters include size parameters, which are then optimized in conjunction with gene weights. Construct boundary mapping rules, which are used to transform the planning indicator constraints of the historical and cultural blocks to be built into the boundary parameters of the parametric control model; The parameters of the parametric control model include: morphological type parameters, numerical parameters, and boundary parameters; Qualitative mapping rules, quantitative mapping rules, and boundary mapping rules are saved to the gene and parameter mapping rule library.

[0062] Preferably, the parameterized control model construction unit 25 is specifically used for: A data interface module is constructed to read data from the landscape gene bank and planning indicator constraints of the historical and cultural blocks to be built; the read data is automatically parsed and matched with the mapping rules corresponding to the gene and parameter mapping rule library; the read data is transformed through the corresponding mapping rules to obtain the parameters of the parameterized control model. A parameter-driven module is constructed to associate the parameters of the parameterized control model with the parameter generation and update logic of the parameterized control model. When the data in the landscape gene bank and the planning indicator constraints of the historical and cultural blocks to be built are updated, the parameterized control model is updated in real time. A collision detection module is constructed to detect the collision relationship between the generated building model of the historical and cultural block to be built and the land boundary line and adjacent buildings in real time. If the boundary parameters are violated, the size parameters are automatically adjusted. An output module is constructed to export the generated parametric model, obtaining the original architectural model of the historical and cultural block to be built. At the same time, a parameter list is output, which records the source and gene weight of the landscape genes corresponding to each parameter.

[0063] Preferably, the building model generation unit 26 is used for: By using the quantitative parameters and gene weights included in the recessive quantification of genes, the perturbation range of the parameters is determined; Within the range of parameter disturbance, multiple sets of non-repeating parameter combinations are randomly generated. Each set of parameter combinations is input into the parameterized control model, and the corresponding three-dimensional architectural model of the historical and cultural block to be built is automatically generated through the parameterized control model. A preliminary screening of the 3D architectural models of the historical and cultural blocks to be built was conducted. 3D architectural models that violated the boundary parameters were eliminated, and the remaining 3D architectural models of the historical and cultural blocks to be built were used as candidate models. The candidate models that met the evaluation requirements were used as the new building models of the historical and cultural blocks to be built.

[0064] The beneficial technical effects achieved by the embodiments of the present invention are as follows: This invention transforms the abstract features of historical and cultural blocks into calculable and controllable algorithmic rules through data acquisition and fusion, feature gene extraction and encoding, parametric rule construction, intelligent generation of multiple schemes, and quantitative evaluation and optimization. Under planning indicator constraints, it enables the intelligent generation of new buildings, achieving deep matching between the generated new building features and the historical block's appearance. This solution integrates cross-disciplinary technologies such as laser point cloud processing, deep learning, and parametric modeling. Its core lies in constructing a closed-loop mapping relationship between data, genes, parameters, and models, completely resolving the problems of ambiguous feature coordination and reliance on manual intervention in existing technologies. It also addresses the issues of disconnected processes and lack of closed-loop optimization in existing technologies.

[0065] From raw data collection to final solution output, no manual intervention is required. The entire process time has been reduced from several days with existing technologies to several hours, significantly improving design and management efficiency. The system can be directly applied to the renewal planning of historical and cultural blocks at all levels, adapting to block projects of different sizes and styles. It provides standardized and intelligent technical tools for urban landscape protection, with broad market application prospects and strong practicality and promotional value.

[0066] It should be understood that the specific order or hierarchy of steps in the disclosed process is an example of an exemplary method. Based on design preferences, it should be understood that the specific order or hierarchy of steps in the process may be rearranged without departing from the scope of this disclosure. The appended method claims provide elements of various steps in an exemplary order and are not intended to limit the scope to the specific order or hierarchy described.

[0067] In the above detailed description, various features are combined together in a single embodiment to simplify this disclosure. This approach to disclosure should not be construed as reflecting an intention that embodiments of the claimed subject matter require more features than are explicitly stated in each claim. Rather, as reflected in the appended claims, the invention is presented with fewer features than all of the features of the single disclosed embodiment. Therefore, the appended claims are hereby explicitly incorporated into the detailed description, wherein each claim stands alone as a preferred embodiment of the invention.

[0068] To enable any person skilled in the art to implement or use the present invention, the disclosed embodiments have been described above. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein can be applied to other embodiments without departing from the spirit and scope of this disclosure. Therefore, this disclosure is not limited to the embodiments given herein, but is consistent with the widest scope of the principles and novel features disclosed in this application. The specific embodiments described above further illustrate the purpose, technical solutions, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for intelligently generating models of newly constructed buildings in historical and cultural districts, characterized in that, include: Acquire data on historical buildings in the target block, preprocess the data on historical buildings in the target block, and obtain a standard dataset of historical buildings based on individual historical building data; The fused images from the standard dataset of historical buildings are input into the display feature extraction model, which outputs the dominant morphological genes of the components contained in the individual historical buildings. The implicit quantitative genes of the components are extracted from the components contained in the individual historical building data. A style gene for the historical building is constructed, which includes the dominant morphological genes and the implicit quantitative genes of the components. The dominant morphological genes of the components include core feature categories, the location coordinates of the core features, and confidence levels. Core features with confidence levels greater than a confidence threshold are retained. The core features include at least one of the following: double-sloped roof, four-sloped roof, blue brick facade, stone facade, arched window, rectangular window, column, and decorative elements. The implicit quantitative genes of the components include the geometric quantitative parameters, color quantitative parameters, and material quantitative parameters of the components. The architectural features of historical buildings are encoded to obtain encoded architectural features, and an architectural features gene library is constructed based on the encoded architectural features. Based on the stylistic characteristics of historical buildings and the planning indicators of the proposed historical and cultural districts, a gene-parameter mapping rule base is constructed. This rule base includes qualitative mapping rules, quantitative mapping rules, and boundary mapping rules, including: A qualitative mapping rule is constructed, which is used to transform dominant morphological genes into morphology type parameters of the parameterized control model; A quantitative mapping rule is constructed to directly convert recessive quantified genes into numerical parameters of a parameterized control model. The numerical parameters include size parameters and are optimized in combination with gene weights. Construct boundary mapping rules, which are used to transform the planning indicator constraints of the historical and cultural blocks to be built into the boundary parameters of the parametric control model; The parameters of the parametric control model include: morphological type parameters, numerical parameters, and boundary parameters; Qualitative mapping rules, quantitative mapping rules, and boundary mapping rules are saved to the gene and parameter mapping rule library; A parameterized control model is constructed based on a gene-parameter mapping rule base. Based on a parametric control model, a new building model for the proposed historical and cultural district is generated by perturbing the parameters. Specifically, this includes: The perturbation range of the parameters is determined by the quantitative parameters and gene weights included in the recessive quantification gene; Within the range of parameter disturbance, multiple sets of non-repeating parameter combinations are randomly generated. Each set of parameter combinations is input into the parameterized control model, and the corresponding three-dimensional architectural model of the historical and cultural block to be built is automatically generated through the parameterized control model. A preliminary screening of the 3D architectural models of the historical and cultural blocks to be built was conducted. 3D architectural models that violated the boundary parameters were eliminated, and the remaining 3D architectural models of the historical and cultural blocks to be built were used as candidate models. The candidate models that met the evaluation requirements were used as the new building models of the historical and cultural blocks to be built.

2. The intelligent generation method for new building models of historical and cultural blocks according to claim 1, characterized in that, The process involves acquiring data on historical buildings in the target block, preprocessing the data to obtain a standard dataset of historical buildings based on individual building data, including: Using ground-based lidar at a preset point cloud density of 100 points / m², the facades and roofs of all historical buildings in the target block are scanned to generate lidar point cloud data. Using a full-frame camera mounted on a drone, images of the facades of all historical buildings in the target block are taken along a planned flight path to form drone image data; at the same time, the location coordinates of the facades are obtained through the positioning function. Collect historical textual data of historical buildings in the target neighborhood, including at least one of the following: census reports, repair drawings, and planning documents; The data for the historical buildings in the target block is constructed, including laser point cloud data, drone imagery data, and historical text data. The data of historical buildings in the target block are sequentially subjected to point cloud denoising, image correction, and point cloud fusion with images to generate textured 3D point cloud models of historical buildings in the target block. The textured 3D point cloud models are automatically segmented into individual buildings to generate fused images of individual historical buildings. A standard dataset of historical buildings is formed by fused images of individual historical buildings.

3. The intelligent generation method for new building models of historical and cultural blocks according to claim 2, characterized in that, The process of inputting fused images from a standard dataset of historical buildings into a display feature extraction model and outputting the dominant morphological genes of the components contained in a single historical building includes: The fused image of individual historical building data is input into the display characteristic extraction model, and the dominant morphological genes of the components contained in the individual historical building are output.

4. The intelligent generation method for new building models of historical and cultural blocks according to claim 3, characterized in that, The extraction of recessive quantitative genes from the components contained in the data of individual historical buildings includes: Segment the point cloud subsets of the components contained in the data of individual historical buildings, and calculate the geometric quantization parameters of the components contained in the individual historical buildings. The component areas contained in the data of individual historical buildings are extracted, and the construction areas are converted to the HSV color space. The mean and standard deviation of hue, saturation and brightness are extracted from the HSV color space. The mean and standard deviation of hue, saturation and brightness of the components are used as the color quantification parameters of the components contained in the individual historical buildings.

5. The intelligent generation method for new building models of historical and cultural blocks according to claim 4, characterized in that, The process of encoding the stylistic features of historical buildings to obtain encoded stylistic features, and constructing a stylistic features gene library based on these encoded stylistic features, includes: The architectural features of historical buildings are encoded, including core feature category codes, sub-category codes, and quantitative parameters included in the recessive quantitative genes. Construct a basic information table for historical buildings, which includes the block number, building number, building type, and building year. A dominant morphogenetic gene table is constructed based on dominant morphogenetic genes. The dominant morphogenetic gene table includes building number, core feature category, gene code, and core feature location coordinates. A recessive quantitative gene table is constructed based on recessive quantitative genes. The recessive quantitative gene table includes building number, quantitative parameters and calculation precision. The basic information table of historical buildings that form associations using building numbers, the dominant morphological gene table and the recessive quantitative gene table are saved to the landscape gene database.

6. The intelligent generation method for new building models of historical and cultural blocks according to claim 1, characterized in that, The parameterized control model constructed based on the gene-parameter mapping rule base includes: A data interface module is constructed, which is used to read data from the landscape gene bank and planning indicator constraints of the historical and cultural blocks to be built; the read data is automatically parsed and matched with the mapping rules corresponding to the gene and parameter mapping rule library; the read data is transformed through the corresponding mapping rules to obtain the parameters of the parameterized control model. A parameter-driven module is constructed, which is used to associate the parameters of the parameterized control model with the parameter generation and update logic of the parameterized control model, so as to realize the real-time update of the parameterized control model when the data of the landscape gene library and the planning indicator constraints of the historical and cultural blocks to be built are updated. A collision detection module is constructed, which is used to detect the collision relationship between the generated building model of the historical and cultural block to be built and the land boundary line and adjacent buildings in real time. If the boundary parameters are violated, the size parameters are automatically adjusted. An output module is constructed to export the generated parameterized model, thereby obtaining the original architectural model of the historical and cultural block to be built. At the same time, a parameter list is output, which records the source and gene weight of the landscape genes corresponding to each parameter.

7. An intelligent generation system for new building models in historical and cultural blocks, characterized in that, include: The data acquisition and processing unit is used to acquire data on historical buildings in the target block, preprocess the data on historical buildings in the target block, and obtain a standard dataset of historical buildings based on individual historical building data. The architectural style gene extraction unit is used to input the fused images from the standard dataset of historical buildings into the display feature extraction model, and output the dominant morphological genes of the components contained in the individual historical buildings; extract the recessive quantitative genes of the components contained in the data of individual historical buildings; construct the architectural style gene of the historical buildings, which includes the dominant morphological genes of the components and the recessive quantitative genes of the components; wherein, the dominant morphological genes of the components include the core feature category, the location coordinates of the core features and the confidence level, and retain the core features with a confidence level greater than the confidence level threshold, wherein the core features include at least one of the following: double-sloped roof, four-sloped roof, blue brick facade, stone facade, arched window, rectangular window, column, and floral decoration; the recessive quantitative genes of the components include the geometric quantitative parameters of the components, the color quantitative parameters of the components, and the material quantitative parameters of the components; The feature gene bank construction unit is used to encode the feature genes of historical buildings, obtain the encoded feature genes, and construct the feature gene bank based on the encoded feature genes. The mapping unit is used for: constructing qualitative mapping rules, which are used to transform dominant morphological genes into morphological type parameters of a parameterized control model; constructing quantitative mapping rules, which are used to directly transform recessive quantitative genes into numerical parameters of a parameterized control model, wherein the numerical parameters include size parameters and are optimized in conjunction with gene weights; and constructing boundary mapping rules, which are used to transform the planning indicator constraints of the proposed historical and cultural block into boundary parameters of the parameterized control model; wherein the parameters of the parameterized control model include: morphological type parameters, numerical parameters, and boundary parameters; and saving the qualitative mapping rules, quantitative mapping rules, and boundary mapping rules to a gene and parameter mapping rule library, which includes the construction of qualitative mapping rules, quantitative mapping rules, and boundary mapping rules. A parameterized control model building unit, used to build parameterized control models based on a gene-parameter mapping rule base; The building model generation unit is used to: determine the perturbation range of parameters by using the quantification parameters and gene weights included in the recessive quantification genes; randomly generate multiple sets of non-repeating parameter combinations within the parameter perturbation range, input each set of parameter combinations into the parameterized control model, and automatically generate the corresponding three-dimensional building model of the historical and cultural block to be built through the parameterized control model; perform preliminary screening of the three-dimensional building models of the historical and cultural block to be built, eliminate three-dimensional building models that violate boundary parameters, use the remaining three-dimensional building models of the historical and cultural block to be built as candidate models, and use the candidate models that meet the evaluation requirements as the new building models of the historical and cultural block to be built.

8. The intelligent generation system for new building models of historical and cultural blocks according to claim 7, characterized in that, The appearance gene extraction unit includes: The dominant morphological gene extraction unit is used to input the fused image of individual historical building data into the display characteristic extraction model and output the dominant morphological genes of the components contained in the individual historical building. The recessive quantization gene extraction unit is used to: segment the point cloud subset of the components contained in the historical building unit data, calculate the geometric quantization parameters of the components contained in the historical building unit; extract the component regions contained in the historical building unit data, convert the constructed regions to the HSV color space, extract the mean and standard deviation of hue, saturation, and brightness from the HSV color space, and use the mean and standard deviation of hue, saturation, and brightness of the components as the color quantization parameters of the components contained in the historical building unit. The implicit quantification genes that construct components.

Images