Method and system for automatic generation of architectural design scenes based on artificial intelligence

Abstract

The invention relates to an artificial intelligence-based automatic generation method and system for architectural design scenes, and relates to the field of architectural design technology. The method comprises: obtaining architectural design scheme data; determining architectural design features according to the architectural design scheme data; generating a three-dimensional architectural model based on the architectural design features; performing rendering optimization on the three-dimensional architectural model based on an artificial intelligence model to generate architectural scene renderings from multiple different perspectives; obtaining effect estimation parameters of the artificial intelligence model for the architectural scene renderings, and generating evaluation results in combination with correction instructions input by a user to correct the effect estimation parameters; and re-rendering and optimizing the three-dimensional architectural model using an artificial intelligence model according to the evaluation results until the evaluation results of the generated architectural scene renderings meet a preset threshold. The implementation of the method can improve the display quality of architectural design.

Description

Technical Field

[0001] The present application relates to the field of architectural design technology, and in particular to an artificial intelligence-based automatic generation method and system for architectural design scenes. Background Art

[0002] With the rapid development of computer graphics and artificial intelligence technology, the field of architectural design has also begun to explore the application of these new technologies to various aspects of the design process to improve design efficiency and quality. Especially in the solution display and interaction stage, traditional manual drawing and rendering methods have been unable to meet the growing demand for visualization. How to use artificial intelligence technology to automatically generate realistic, diverse, and coordinated architectural scene renderings has become a hot topic of common concern in academia and industry.

[0003] At present, some studies have tried to use computer graphics and computer vision methods to assist in the generation of architectural scene renderings. For example, some studies use 3D modeling software such as SketchUp and 3ds Max to generate scene renderings under different perspectives, lighting, and materials through parametric settings and rendering engines based on architectural models; other studies use computer vision technologies such as image synthesis and image translation to automatically replace or synthesize different materials, textures, objects, and other elements on a given scene image to generate a new scene after style transfer.

[0004] However, the scene elements generated by related technologies are inconsistent and unrealistic with the architectural design, and it is difficult to meet the high-quality display requirements of the architectural design. Summary of the invention

[0005] The present application provides an artificial intelligence-based automatic generation method and system for architectural design scenes, which are used to improve the display quality of architectural designs.

[0006] In the first aspect, the present application provides an automatic generation method of architectural design scenes based on artificial intelligence, which is applied to an architectural design system, and the method includes: obtaining architectural design plan data; the architectural design plan data includes architectural floor plans, architectural elevations and architectural sections; determining architectural design features based on the architectural design plan data; architectural design features include spatial layout, material information and lighting conditions; generating a three-dimensional architectural model based on the architectural design features; rendering and optimizing the three-dimensional architectural model based on an artificial intelligence model to generate architectural scene renderings from multiple different perspectives; obtaining effect estimation parameters of the artificial intelligence model for the architectural scene renderings, and generating evaluation results in combination with correction instructions input by the user to correct the effect estimation parameters; the evaluation results include a visual effect realism score and a design intent matching score; based on the evaluation results, using the artificial intelligence model to re-render and optimize the three-dimensional architectural model until the evaluation results of the generated architectural scene renderings meet a preset threshold.

[0007] In the above embodiment, the architectural design system first obtains architectural design scheme data including architectural floor plans, architectural elevations and architectural sections, and determines architectural design features such as spatial layout, material information and lighting conditions. Then, a three-dimensional architectural model is generated based on these architectural design features, and then the three-dimensional architectural model is rendered and optimized using an artificial intelligence model to generate realistic architectural scene renderings from multiple different perspectives. At the same time, the method can also obtain the scene effect estimation parameters given by the artificial intelligence model, and generate evaluation results including visual effect authenticity and design intent matching in combination with the user's correction instructions. Finally, based on the evaluation results, the artificial intelligence model is continuously used to optimize the architectural model and rendering results until the generated architectural scene renderings reach the ideal quality level.

[0008] In combination with some embodiments of the first aspect, in some embodiments, effect estimation parameters of the artificial intelligence model for the building scene renderings are obtained, and combined with the correction instructions input by the user to correct the effect estimation parameters, an evaluation result is generated, specifically including: based on the artificial intelligence model, respectively estimating the visual realism and design intent matching of multiple building scene renderings, and obtaining the initial realism scores and initial intent matching scores corresponding to the multiple building scene renderings as effect estimation parameters; displaying multiple building scene renderings, the corresponding initial realism scores and initial intent matching scores; receiving a correction instruction for the initial realism score and / or initial intent matching score of the selected building scene rendering after the user selects one of the building scene renderings; correcting the initial realism score and / or the initial intent matching score according to the correction instruction to obtain the evaluation result of the corresponding building scene rendering.

[0009] In the above embodiment, after the artificial intelligence model gives the initial estimated score of the building scene rendering, the architectural design system presents the specific scene rendering and the corresponding initial reality score and initial intention matching score to the user through a visual interface. The user can select one of the scenes and make subjective corrections to the score of the scene. The system applies the user's correction instructions to the original score to obtain the final evaluation result of the scene, which serves as the basis for the subsequent optimization rendering of the artificial intelligence model. This human-computer interactive scene evaluation method, on the one hand, uses the scene understanding ability of the artificial intelligence model to give an objective estimate, and on the other hand, allows the user to make fine adjustments based on subjective preferences, thereby improving the evaluation efficiency while taking into account the design intent of the designer, so that the final generated architectural scene is more in line with the design requirements.

[0010] In combination with some embodiments of the first aspect, in some embodiments, a three-dimensional building model is generated based on architectural design features, specifically including: identifying different functional space areas and their positional relationships in the building plan, and dividing the functional zoning layout of different floors according to the identification results; obtaining the building's exterior wall material, appearance shape and vertical door and window opening positions of the building's elevation view to generate the facade; obtaining the floor height, floor thickness and beam-column structure in the building's interior space of the building's section view to generate the internal structure; combining the functional zoning layout, the facade and the internal structure to obtain a complete three-dimensional building geometric model; based on the material information in the architectural design features, applying corresponding material maps to different components of the three-dimensional building geometric model, and based on the lighting conditions in the architectural design features, configuring the light source environment parameters of the three-dimensional building geometric model to obtain the three-dimensional building model.

[0011] In the above embodiment, when the architectural design system generates a three-dimensional building model based on the architectural design features, a series of technical means are used to extract key information from the architectural design drawings and map it to different components of the three-dimensional model. First, the functional zoning layout of each floor is obtained by identifying the plan view; then the material, appearance and window information of the building's exterior wall are obtained from the elevation view to generate a three-dimensional appearance; then the structural parameters such as the floor height, floor slab, beams and columns of the internal space are obtained from the section view to generate the internal structure. Finally, the zoning layout, facade and internal structure are combined into a complete building geometry model, and according to the material and lighting conditions in the design features, the geometry model is given real material mapping and light and shadow effects, and finally a three-dimensional building model that better meets the requirements is obtained, ensuring the quality of the model.

[0012] In combination with some embodiments of the first aspect, in some embodiments, before the step of obtaining the architectural design plan data, the method also includes: determining the building site environmental parameters selected and input by the user; the building site environmental parameters include geographical location, climate conditions, topography and surrounding building distribution; determining the building function attribute parameters selected and input by the user; the building function attribute parameters include building type, usage function, space area and number of floors; determining the architectural style requirement parameters selected and input by the user; the architectural style requirement parameters include exterior style and interior decoration style; inputting the building site environmental parameters, building function attribute parameters and architectural style requirement parameters into a pre-trained architectural design generation model to generate a building plan, a building elevation and a building section as initial design plan data; displaying the initial design plan data for further editing and modification by the user to obtain the final edited and modified architectural design plan data.

[0013] In the above embodiment, the architectural design system first considers the building site environmental parameters, including geographical location, climate conditions, topography and surrounding building distribution, etc. These parameters will affect the orientation, volume and layout of the building; secondly, the building functional attribute parameters, including building type, usage function, space area and number of floors, etc. These parameters will determine the scale, layout and streamline of the building; finally, the architectural style requirement parameters, including exterior style and interior decoration style, etc. These parameters will affect the facade modeling and interior decoration of the building. By inputting the above parameters into the pre-trained architectural design generation model, an initial design scheme that meets various design requirements can be quickly generated, and then after further editing and modification by the user, the perfect architectural design scheme data is finally obtained.

[0014] In combination with some embodiments of the first aspect, in some embodiments, building site environmental parameters, building function attribute parameters and building style requirement parameters are input into a pre-trained building design generation model to generate a building plan, building elevation and building section, specifically including: inputting the building site environmental parameters into the building design generation model to generate a three-dimensional environment model of the building site, and optimizing the coordination between the building layout and the site environment to obtain a preliminary layout of the building plan; the optimization process includes adjusting the building orientation, volume mass and spacing through ventilation environment and sunlight analysis; inputting the building function attribute parameters into the building design generation model to generate the layout relationship, space size and entrance and exit positions of each functional space inside the building to obtain the building plan of each floor of the building; based on the preliminary layout of the building plan and the plan of each floor of the building, the building outline is generated to obtain the building elevation; inputting the building style requirement parameters into the building design generation model to generate internal and external facade designs; based on the building elevation and the internal and external facade designs, a building section is generated.

[0015] In the above embodiment, the architectural design system uses the building site environmental parameters to generate a three-dimensional environmental model, optimizes the coordination between the building layout and the environment, and obtains the preliminary layout of the building plane; then uses the building function attribute parameters to generate the layout relationship, size and entrance and exit positions of each functional space, and improves the floor plan of each floor of the building; then based on the plane layout, the building's outer contour is generated to obtain the elevation drawing; finally, the architectural style requirement parameters are used to generate the internal and external elevation design, and then the building section drawing is generated accordingly. The entire process makes full use of the computing power of artificial intelligence in spatial layout, parameter optimization, etc., and realizes the automated generation and optimization of the site environment, functional requirements, and building plan, elevation and section design.

[0016] In combination with some embodiments of the first aspect, in some embodiments, after the step of generating a three-dimensional building model based on the building design features, the method also includes: performing parametric simulation analysis of building performance on the generated three-dimensional building model according to the requirements of the building design specifications to obtain performance evaluation results; the building performance includes lighting, ventilation, energy saving and fire prevention; displaying the performance evaluation results, and optimizing the parameters of the three-dimensional building model based on the artificial intelligence model; obtaining the three-dimensional building model after parameter optimization, and then rendering to generate a building scene rendering.

[0017] In the above embodiment, the architectural design system visualizes the performance evaluation results obtained from the analysis to the designer, intuitively revealing the problems of the current design scheme. Designers can make targeted parameter adjustments to key problem areas based on this, or let the artificial intelligence model automatically optimize the relevant design parameters based on the analysis results. On this basis, the system re-updates the three-dimensional building model and performance evaluation results, and promptly feeds back to the designer. Through this human-computer collaborative performance-based design method, the various performance indicators of the building model are continuously improved, so that the final generated building scene achieves a higher balance in terms of aesthetics and practicality.

[0018] In combination with some embodiments of the first aspect, in some embodiments, the performance evaluation results are displayed, and the parameters of the three-dimensional building model are optimized based on the artificial intelligence model, specifically including: visually displaying the performance analysis results; visually displaying the performance analysis results specifically includes: displaying the lighting analysis results in the form of a room interior brightness cloud map, displaying the ventilation analysis results in the form of animated lines of air flow, and displaying the energy-saving analysis results in the form of an energy consumption indicator radar map; receiving user interaction instructions for problem areas in the visualized performance analysis results, and adjusting relevant parameters of the three-dimensional building model accordingly; or, analyzing the performance evaluation results based on the artificial intelligence model, adjusting relevant parameters; based on the adjustment operation of the relevant parameters, updating the three-dimensional building model, and synchronously updating the visual display of the performance analysis results.

[0019] In the above embodiment, the building design system designs intuitive and easy-to-understand visualization graphics for different performance indicators. For example, the lighting analysis results are displayed in the form of a room interior brightness cloud map, the ventilation analysis results are displayed in the form of animated lines of air flow, and the energy-saving analysis results are displayed in the form of an energy consumption indicator radar map. Designers can directly select areas with performance problems on the visualization interface through interactive operations, adjust relevant building parameters, and view the effects of the adjustments. At the same time, the system's built-in artificial intelligence model can also automatically analyze performance evaluation results, adjust key parameters, and update the three-dimensional building model and the visualization of performance analysis in real time.

[0020] In a second aspect, an embodiment of the present application provides an architectural design system, which includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the architectural design system to perform the method described in the first aspect and any possible implementation method of the first aspect.

[0021] In a third aspect, an embodiment of the present application provides a computer program product comprising instructions, which, when the computer program product is run on an architectural design system, enables the architectural design system to execute the method described in the first aspect and any possible implementation of the first aspect.

[0022] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, comprising instructions. When the instructions are executed on an architectural design system, the architectural design system executes the method described in the first aspect and any possible implementation method of the first aspect.

[0023] It can be understood that the architectural design system provided in the second aspect, the computer program product provided in the third aspect, and the computer storage medium provided in the fourth aspect are all used to execute the method provided in the embodiment of the present application. Therefore, the beneficial effects that can be achieved can refer to the beneficial effects in the corresponding method, and will not be repeated here.

[0024] One or more technical solutions provided in the embodiments of the present application have at least the following technical effects or advantages:

[0025] 1. By taking the architectural design plan data as input, extracting key design features such as spatial layout, material information and lighting conditions, and generating a three-dimensional architectural model based on this method, the three-dimensional modeling process is greatly simplified while ensuring that the architectural model is highly consistent with the original design plan. This effectively solves the problem in the existing technology that designers need to manually build three-dimensional models, which is time-consuming and labor-intensive and difficult to accurately match with the design drawings, thereby realizing the automatic conversion and mapping of architectural plans to three-dimensional models.

[0026] 2. Since the user interaction link is introduced before obtaining the architectural design plan data, allowing designers to independently select and input the building site environment parameters, building function attribute parameters and architectural style requirement parameters, the flexibility and adaptability of architectural design generation are greatly improved, and the problem that the architectural plans generated by artificial intelligence in the existing technology are difficult to meet the diverse design needs is effectively solved, thereby realizing the automatic generation of personalized and customized architectural design plans.

[0027] 3. Since the building performance analysis link based on computer simulation technology is immediately introduced after the three-dimensional building model is generated, and the performance evaluation results of lighting, ventilation, energy saving, fire protection, etc. are visualized, designers can more intuitively understand the performance shortcomings of the current design scheme, thereby greatly improving the efficiency of identifying and locating building performance problems, and effectively solving the problem that traditional design methods are difficult to comprehensively evaluate building performance, resulting in difficulties in subsequent modifications, thereby realizing the pre-evaluation and optimization of building performance. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a flow chart of a method for automatically generating architectural design scenes based on artificial intelligence in an embodiment of the present application;

[0029] Figure 2 is another flowchart of the method for automatically generating architectural design scenes based on artificial intelligence in an embodiment of the present application;

[0030] Figure 3 It is a schematic diagram of a physical device structure of the architectural design system in the embodiment of the present application. DETAILED DESCRIPTION

[0031] The terms used in the following embodiments of the present application are only for the purpose of describing specific embodiments, and are not intended to be used as limitations to the present application. As used in the specification of the present application, the singular expressions "one", "a kind of", "above", "the" and "this" are intended to also include plural expressions, unless there is a clear indication to the contrary in the context. It should also be understood that the term "and / or" used in the present application refers to any or all possible combinations comprising one or more of the listed items.

[0032] In the following, the terms "first" and "second" are used for descriptive purposes only and are not to be understood as suggesting or implying relative importance or implicitly indicating the number of the indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include one or more of the features, and in the description of the embodiments of the present application, unless otherwise specified, "plurality" means two or more.

[0033] For ease of understanding, the application scenarios of the embodiments of the present application are introduced below.

[0034] When a large real estate development company was designing a large commercial complex project, the project site had complex terrain and uneven surrounding environment. At the same time, it also needed to meet the functional requirements of various formats, including commercial, office, hotel and apartment, etc., which put forward high requirements on the spatial layout, appearance style and interior decoration of the building. The design team hopes to quickly generate multiple design schemes that meet various conditions and present them in the form of high-quality renderings for Party A to review and select. However, due to the complex project conditions and heavy design tasks, the traditional design process is difficult to complete all the work in a short time, and the time cost and labor cost are very high. At the same time, the efficiency of manual modeling and rendering by designers cannot meet the needs of rapid generation of schemes, and it is difficult to ensure the uniform quality of renderings of different schemes. Therefore, the project urgently needs a technical solution that can quickly generate high-quality multi-scheme design and display to improve design efficiency and reduce costs.

[0035] In the related art, the three-dimensional visualization of architectural design solutions can be achieved by using manual modeling and rendering. Designers use computer software to manually draw architectural models, add materials, lights and other elements to them, and finally render and output architectural renderings. However, this method is not only time-consuming and labor-intensive, but also difficult to generate multiple solutions in batches, and the realism of the rendering effect is also difficult to ensure. The following introduces a scenario using the automatic generation method of architectural design scenes based on artificial intelligence in the related art.

[0036] The design team tried to use the common 3D architectural design software on the market to improve the design efficiency of the project. The designer imported the design drawings of the project into the software and used the 3D modeling tools of the software to convert the floor plan into a 3D model. However, due to the very irregular project site environment and architectural shape, the modeling components that come with the software are difficult to quickly build a building model that meets the requirements, and the designers need to spend a lot of time manually drawing the building structure. At the same time, it is also a very time-consuming task to implant realistic elements such as material mapping and lighting into the generated architectural model and render and output it, and it is difficult to generate multiple solutions in batches. In addition, since the 3D model is made separately after the design drawings are completed, it is inevitable that there will be deviations from the original design intent. The designer also needs to repeatedly adjust the model in the software to make it match the design drawings. Therefore, the existing design software cannot solve the problems faced by this project well, and the design team needs to seek more intelligent and automated technical means.

[0037] The method for automatically generating architectural design scenes based on artificial intelligence in the embodiment of the present application uses architectural design drawings as input, automatically generates three-dimensional architectural models using a pre-trained deep learning model, and uses artificial intelligence algorithms to optimize scene rendering, thus achieving end-to-end automatic generation from design to renderings. This not only greatly improves the efficiency of architectural design visualization, but also ensures the high degree of realism and consistency of the generated renderings. The following introduces the scenarios using the method for automatically generating architectural design scenes based on artificial intelligence in the present application.

[0038] The design team decided to adopt the artificial intelligence-based automatic generation method of architectural design scenes proposed in this patent to carry out project design and presentation. First, the designer inputs the project's site environment parameters, functional attribute parameters, and style requirement parameters into the system, and uploads the preliminary architectural floor plan, elevation, and section drawings. The system uses the pre-trained architectural design generation model to automatically identify the design points in the drawings, and intelligently generates the skeleton of the three-dimensional architectural model on the basis of optimizing the coordination between the building and the site and the layout relationship of each functional space. Designers only need to partially edit and modify the three-dimensional architectural skeleton model generated by the system to quickly obtain a complete three-dimensional architectural model. Next, the system automatically matches the best material mapping and light and shadow parameters according to the architectural style requirements, and uses artificial intelligence algorithms to render the model, and batch outputs high-quality architectural scene renderings from multiple different perspectives. Based on the system's recommendations, the designer selected several of the most satisfactory renderings for review by Party A. By using this patented technology, the design team completed all the design work in just three days, while if the traditional method was used, it would take at least two weeks. Therefore, this patented method effectively improves the efficiency of design generation and architectural presentation, reduces costs, and meets the requirements of the project.

[0039] It can be seen that the artificial intelligence-based automatic generation method of architectural design scenes in the embodiment of the present application can not only achieve the rapid generation of high-quality architectural scene renderings, but also effectively solve the problems of low efficiency and uneven quality of traditional manual modeling and rendering methods, thereby realizing the intelligence and automation of design performance and improving the overall efficiency and quality of architectural design.

[0040] For ease of understanding, the following describes the process of the method provided by this implementation in combination with the above scenario. Figure 1 , which is a flow chart of a method for automatically generating architectural design scenes based on artificial intelligence in an embodiment of the present application.

[0041] S101. Obtain architectural design plan data.

[0042] Among them, the architectural design plan data includes building floor plans, building elevations and building sections; the building floor plans represent the schematic diagrams of the floor plans of each floor of the building, which are used to show the size and positional relationship of each room, corridor, door, window and other components. For example, a floor plan of an office building will indicate the layout and area of ​​each office, conference room, toilet and other information. The building elevation refers to the projection of the front, side and back facades of the building, showing the three-dimensional effects of the building's appearance, window distribution, exterior wall material, etc. The building section refers to the schematic diagram of the internal structure of the building after it is cut vertically or horizontally, revealing the internal spatial layout, floor height, stair location and other information of the building. It is an important drawing for in-depth understanding of the relationship between the inside and outside of the building.

[0043] Specifically, after receiving the architectural plan drawing file submitted by the architect, the architectural design system will first read and identify the drawing data. The system automatically distinguishes between the building plan, elevation and section through the file name, format, frame, legend and other contents. For different types of drawings, the system uses corresponding algorithms for feature extraction and semantic understanding. For example, for the building plan, the system focuses on extracting elements such as room outlines, dimensioning, door and window symbols; for the elevation, the system focuses on analyzing elements such as wall outlines, window shapes, and exterior wall materials; for the section, the system focuses on information such as elevation, floor thickness, and stair location. Through the extraction and analysis of drawing elements, the system obtains complete architectural design plan data, preparing for the subsequent generation of 3D models.

[0044] In some embodiments, obtaining architectural design scheme data can be achieved through the following ways: Optionally, the architect uploads the drawn architectural drawing file directly to the system; the system automatically identifies the drawing type and extracts relevant design elements. Optionally, the architect draws architectural drawings on the graphical interface of the system; the system obtains the drawing data in real time and stores it as architectural design scheme data. Optionally, the system is seamlessly integrated with mainstream architectural design software, and the architect completes the scheme design in third-party software; the system calls the software interface to directly obtain standardized architectural design scheme data. It is understandable that the system can also use other ways to obtain architectural design scheme data, such as through the text description of the design scheme, sketch scans, etc., and the required data can be obtained after certain conversion and recognition, which is not limited here.

[0045] S102. Determine architectural design features based on architectural design plan data.

[0046] Among them, spatial layout refers to the plane distribution and three-dimensional combination relationship of various functional spaces in the building, such as the size, location, orientation, entrances and exits, and public space forms of the rooms, which reflects the use function and spatial order of the building. For example, the key point of the spatial layout of a residential design is to reasonably organize the living room, bedroom, kitchen, bathroom and other spaces to form a comfortable and convenient living environment. Material information refers to the type, color, and texture characteristics of the materials used in various parts of the building, such as exterior wall paint, glass curtain walls, stone veneers, wooden floors, etc., which reflects the style and texture of the building. Lighting conditions are used to indicate the lighting conditions of various surfaces of the building during the day, such as orientation, window-to-wall ratio, shading measures, etc., which affect the lighting performance and visual effects of the building.

[0047] Specifically, after obtaining the architectural design scheme data, the architectural design system begins to extract and analyze the architectural design features. For the architectural plan, the system identifies the room boundary lines, calculates the area, length and width of each room, and analyzes the relative position relationship and entrance and exit orientation of the room to form the spatial layout characteristics of each floor of the building in the brain. For the elevation and section drawings, the system extracts key information such as the building's height, number of floors, floor height, window position and size, respectively, to supplement and improve the vertical spatial characteristics of the building. At the same time, the system also analyzes the material symbols and text descriptions marked in the plan, elevation and section drawings to obtain information such as the type and color of the building's surface material. In addition, the system will refer to the building's orientation, window-to-wall ratio, external shading and other design points to analyze the building's lighting and ventilation conditions. After a comprehensive interpretation of the drawings, the architectural design system has completed the transformation of architectural design features from two-dimensional drawings to spatial, material, and light and shadow. This feature information is the key basis for the generation of three-dimensional models.

[0048] In some embodiments, the system can determine the architectural design features by the following methods: Optionally, the system pre-trains the architectural drawing parsing model based on deep learning technology; the acquired architectural design scheme data is input into the model to automatically extract features such as spatial layout, material, and lighting. Optionally, the system has a built-in architectural design feature extraction rule library, and uses a rule-based feature recognition method to match the drawing elements one by one to extract key feature information. Optionally, the system interface guides the architectural designer to manually mark key design features, such as marking the location and size of the main rooms, specifying the material and color of key parts, etc.; the system records the marking information to form an architectural design feature library. It is understandable that the system can also use other methods to determine the architectural design features, such as combining the above methods, first automatically extracting, then manually checking and supplementing, to form a more comprehensive and accurate set of architectural design features, which is not limited here.

[0049] S103: Generate a three-dimensional building model based on the building design features.

[0050] Among them, architectural design features refer to the data set that characterizes the key attributes of a building extracted from architectural design drawings, including architectural floor plan, facade modeling, spatial scale, material information, etc. A three-dimensional architectural model refers to a digital model built in a computer that is highly consistent with the appearance and structure of a real building. It can present the three-dimensional image of the building through three-dimensional geometry, material mapping, etc. For example, for a modern-style museum, its architectural design features may include irregular plane outlines, large-span glass curtain wall facades, tall and open atrium spaces, etc. By extracting these feature information, the corresponding geometric model and effect performance can be generated in the three-dimensional modeling software.

[0051] Specifically, once the system obtains the complete architectural design drawings, it can start to automatically generate the 3D model of the building. The system first applies computer vision algorithms to perform semantic analysis on the drawings, extract key design elements in the drawings, such as the outline of the exterior wall, the location of doors and windows, and the arrangement of column grids, and convert them into quantifiable feature vectors. Then, based on the pre-built 3D architectural model database, the system uses deep learning algorithms such as generative adversarial networks to learn the mapping relationship between architectural design features and 3D models. Based on the extracted design features, the system retrieves the 3D model components with the highest matching degree from the model library, such as monomer geometry, component models, etc., and assembles each component into a complete architectural model according to the relative position, proportion and other information indicated in the drawings. During the assembly process, the system also considers the semantic logic of the architectural design, such as the spatial relationship between load-bearing walls and beams and columns, and the nesting of doors and windows in the wall, to ensure the structural correctness of the generated model. Through the above process, the system can quickly convert 2D design drawings into 3D models, providing a geometric basis for subsequent scene rendering.

[0052] In some embodiments, the system can realize the automatic conversion from architectural design features to three-dimensional models in a variety of ways: Optionally, the system adopts a rule-based procedural modeling method, understands the design drawings as a series of parameterized modeling rules, and generates a three-dimensional model through logical reasoning, such as identifying the wall as a rectangular plane perpendicular to the ground, and understanding the sloping roof as an inclined triangular plane, etc.; Optionally, the system applies a data-driven end-to-end deep learning method, using a large number of drawing-model pairs as training data, and directly learning the mapping relationship from two-dimensional to three-dimensional through the encoder-decoder network structure to achieve one-click generation of the building model; Optionally, the system introduces a human-computer collaborative modeling mechanism, and allows users to interactively adjust model details, such as door and window styles, curtain wall partitions, etc., based on the automatically generated preliminary model, and finally obtains a building model that meets the design intent and engineering specifications. It is understandable that the system can also use other three-dimensional reconstruction technologies to achieve the generation of building models, such as SFM, SLAM, etc., with the aid of building point cloud data obtained by oblique photography, laser scanning, etc. to assist the modeling process, which is not limited here.

[0053] S104. Rendering and optimizing the three-dimensional building model based on the artificial intelligence model to generate building scene renderings from multiple different perspectives.

[0054] Among them, the artificial intelligence model refers to a neural network model trained by applying a deep learning algorithm, which is used to automatically optimize the rendering parameters of the three-dimensional model, such as material editing, light and shadow settings, and camera layout. Rendering optimization means adding realistic material maps to the three-dimensional building model, setting realistic lighting conditions, and selecting the best viewing angle to generate high-quality architectural scene renderings. The rendering refers to a static view that can intuitively reflect the effect of the architectural design by converting the building model scene into a lifelike image through a rendering engine. For example, a residential exterior rendering creates a realistic architectural scene atmosphere by applying real brick and stone materials, simulating natural lighting, and adding rich vegetation and characters.

[0055] Specifically, once the system generates an accurate three-dimensional building model, it begins to enter the rendering optimization phase. The system loads a pre-trained rendering optimization artificial intelligence model, which takes the three-dimensional building model as input and outputs the optimal rendering parameter configuration plan. Based on the model reasoning results, the system automatically matches the corresponding material maps for the surfaces of each component of the building model to ensure that the color, scale, and distribution of the texture are highly close to the real material; at the same time, it adds ambient light and main light sources to the outdoor scenes, adjusts the light source parameters, and simulates real light and shadow effects; in addition, the system will also consider the characteristics of the building and intelligently set the best camera perspective, depth of field and other parameters. Through rendering optimization, the system can automatically batch generate architectural scene renderings that reflect the key points of architectural design and have outstanding visual effects, greatly improving the performance quality and efficiency of architectural design.

[0056] In some embodiments, the system can achieve AI-based rendering optimization in a variety of ways: Optionally, the system builds a convolutional neural network model, takes architectural design features and three-dimensional model data as input, outputs key rendering parameters, and achieves end-to-end full-process optimization; Optionally, the system builds a GAN generative adversarial network, in which the generator is responsible for generating renderings based on the three-dimensional model and rendering parameters, and the discriminator is responsible for evaluating the authenticity of the generated image, and through adversarial learning between the two, a high-quality rendering effect is obtained; Optionally, the system adopts a reinforcement learning model, regards the setting of rendering parameters as a series of decision-making actions, takes the visual score of the generated image as a reward, and learns the rendering strategy through reward feedback to achieve adaptive optimization. It is understandable that the system can also adopt other AI-based rendering optimization technologies, such as introducing attention mechanisms, transfer learning, etc., to further improve the quality of rendering generation, which is not limited here.

[0057] S105. Obtain effect estimation parameters of the artificial intelligence model for the building scene rendering, and generate evaluation results in combination with correction instructions input by the user to correct the effect estimation parameters.

[0058] Among them, the effect estimation parameters refer to a series of quantitative indicators obtained by the artificial intelligence model to evaluate the generated architectural scene renderings, which are used to reflect the visual quality and design compliance of the image. The visual effect realism score indicates the degree to which the rendering is close to the real-world scene in terms of material texture, light and shadow details, and overall atmosphere. The higher the score, the more realistic the effect. The design intent matching score refers to the degree to which the rendering reflects the key design points and design style of the original design plan. The higher the score, the more the visualization conforms to the designer's creative ideas. Correction instructions refer to users making adjustment suggestions to certain elements of the renderings based on their own aesthetic preferences and design experience, guiding the generation results to be closer to the design requirements. For example, designers can modify the rendering parameters to make the building materials richer and the lighting brighter, so that the renderings can more fully express the design intent.

[0059] Specifically, after the system completes the generation of AI-based architectural scene rendering, it will also simultaneously evaluate the effect of the generated image. The system uses another pre-trained image quality assessment model to simulate the subjective perception of the human eye on the visual quality of the image, and gives quantitative scores for the renderings in terms of realism, completeness, and aesthetics. At the same time, the evaluation model can also compare the visual consistency between the original design drawings and the generated renderings, judge the degree of expression of the generated images for the design points, and give a quantitative score for the design compliance. After the evaluation model generates preliminary effect estimation parameters, the system also provides a human-computer interaction interface for designers to refer to the estimation results and input subjective correction opinions, such as adjusting the brightness of the material, optimizing the scene composition, etc. The system integrates machine estimation and expert correction, iteratively updates the evaluation model, and finally generates a comprehensive evaluation result that takes into account both objective indicators and subjective aesthetics.

[0060] In some embodiments, the system can obtain effect prediction parameters and generate evaluation results through the following ways: Optionally, the system builds a comparative learning neural network, takes the generated image and the real image as input, and predicts the visual realism of the generated image through self-supervised visual similarity learning; Optionally, the system constructs a twin neural network, takes the original design drawing and the generated image as input, respectively, and judges the degree of match between the design scheme and the rendering by comparing the similarity of the embedded vectors; Optionally, the system designs a joint loss function, comprehensively considers factors such as image quality score and user feedback correction, and realizes the deep integration of the evaluation model and user intention through end-to-end back propagation. It is understandable that the system can also use other machine learning methods to carry out rendering evaluation, such as transfer learning, self-learning, etc., and dynamically optimize the evaluation model based on user feedback, which is not limited here.

[0061] S106. According to the evaluation result, the three-dimensional building model is re-rendered and optimized using the artificial intelligence model until the evaluation result of the generated building scene rendering meets the preset threshold.

[0062] Among them, the preset threshold refers to the quantitative score line of the architectural scene rendering in the two dimensions of visual realism and design matching, which reflects the quality standard of architectural visualization and the bottom line of acceptance of designers. When the evaluation result exceeds the threshold, it means that the generated rendering has reached a high level in visual effect and design expression, and can be used for review and reporting of architectural plans. The setting of the threshold needs to strike a balance between effect quality and generation efficiency. It can generally be determined based on the score distribution of a large amount of architectural rendering data and combined with the subjective judgment of the designer. For example, the average score of visual realism and design matching can reach 8 points (out of 10 points) as the passing line for rendering generation.

[0063] Specifically, after the system generates the architectural scene rendering, it will obtain the comprehensive image evaluation results in real time. The evaluation results are compared with the pre-set thresholds. If it is found that some aspects are not up to standard, the rendering optimization iteration process is triggered. The system calls the rendering optimization artificial intelligence model again, takes the current rendering as input, and performs secondary rendering on its deficiencies. For example, if it is found that the material texture of the rendering is not rich enough, the system will search for more material mapping materials, optimize the material mapping scheme through deep learning algorithms, and make the surface texture of the building more varied; if it is found that the position of the main building is not prominent enough, the system will automatically optimize the scene composition through the camera perspective intelligent recommendation algorithm to highlight the key design elements. The rendering optimization model focuses on the weak links in the evaluation, carries out targeted rendering parameter adjustments, and takes into account the overall coordination. After several rounds of iterations, the generated renderings can reach a level that satisfies the designer in all indicators.

[0064] In some embodiments, the system can achieve closed-loop optimization of architectural scene renderings through the following approaches: Optionally, the system constructs a rendering optimization model based on reinforcement learning, takes the evaluation results as the reward function, and uses algorithms such as policy gradient to enable the agent to learn the adjustment strategy of rendering parameters to achieve adaptive improvement of rendering quality; Optionally, the system adopts a multi-agent collaborative optimization framework, and each agent is responsible for optimizing different rendering parameters. Through communication and learning between agents, task decomposition and joint optimization are achieved to improve the convergence speed; Optionally, the system introduces meta-learning ideas, and through small sample learning and rapid adaptation, the rendering optimization model has the ability to generalize and optimize for new design scenarios, thereby enhancing the robustness of the system. It is understandable that the system can also use other iterative optimization methods to improve the quality of rendering generation, such as human-machine hybrid enhanced intelligence, active learning, etc., and make full use of the professional experience of designers to guide the optimization process through human-machine collaboration, which is not limited here.

[0065] In the above embodiments, the process of generating and rendering three-dimensional building models using the technical solution of the present application is mainly described. In practical applications, the present application can also integrate various building performance analysis tools, and immediately carry out lighting, ventilation, energy saving and other performance simulation analysis after the building model is generated, and visualize the analysis results for reference by designers. Designers can optimize and adjust the key performance parameters of the building based on this, and the system will update the three-dimensional model and renderings in real time, thereby achieving closed-loop optimization of building performance. This function expands the application scenarios of the technical solution of the present application and further improves the scientificity and practicality of architectural design. The scenarios of this embodiment are supplemented below.

[0066] After the last round of design report, Party A was satisfied with the overall design effect, but proposed some modification opinions on the local space layout, hoping to further optimize the lighting and ventilation effects of some areas, and asked the design team to provide optimized renderings. The design team input Party A's feedback into the system, and used the computer simulation analysis technology integrated in the system to evaluate the lighting and ventilation performance of the building. Through the visual analysis results, the designer quickly found the problem area and adjusted the parameters such as the window-to-wall ratio and external shading device. The system generates an updated three-dimensional building model in real time, re-renders it, and feeds back the optimized renderings to Party A for confirmation. Party A approved the optimized design, and the project successfully passed the scheme review and entered the construction drawing design stage. In the subsequent design process, the design team made full use of the system for scheme optimization and design performance, and integrated management of the entire process of design and construction, effectively ensuring the design quality and construction progress of the project, and won high praise from customers. It can be seen that this patented technology can not only cope with complex design problems and improve the efficiency of design generation and presentation, but also has good scalability and can integrate various analysis and management tools to fully support the design, construction and operation and maintenance of construction projects.

[0067] After combining the above scenarios, the following is a more detailed description of the process of the method provided by this implementation. Figure 2 , which is another flow chart of the method for automatically generating architectural design scenes based on artificial intelligence in an embodiment of the present application.

[0068] S201. Obtain architectural design plan data.

[0069] Referring to step S101 , the architectural design system acquires architectural design solution data.

[0070] In some embodiments, the architectural design system will first determine the building site environmental parameters selected and input by the user; the building site environmental parameters include geographical location, climate conditions, topography and surrounding building distribution; determine the building function attribute parameters selected and input by the user; the building function attribute parameters include building type, usage function, space area and number of floors; determine the architectural style requirement parameters selected and input by the user; the architectural style requirement parameters include exterior style and interior decoration style; input the building site environmental parameters, building function attribute parameters and architectural style requirement parameters into a pre-trained architectural design generation model to generate a building plan, building elevation and building section as initial design plan data; display the initial design plan data for further editing and modification by the user to obtain the final edited and modified architectural design plan data.

[0071] Among them, the building site environmental parameters refer to a series of indicators that describe the characteristics of the environment in which the building is located, including geographical location, climatic conditions, topography and distribution of surrounding buildings. Geographical location refers to the longitude and latitude coordinates of the building, which determines the macro-environmental background of the building. Climate conditions refer to the changing laws of meteorological elements such as temperature, humidity, wind direction, etc. at the location of the building, which affect the thermal performance and indoor comfort of the building. Topography refers to the natural geographical features such as elevation fluctuations and slope changes of the building site, which restricts the layout and landscape design of the building. The distribution of surrounding buildings reflects the current status of the built environment of the site, and it is necessary to consider the coordination relationship between the new building and it. Building function attribute parameters are used to indicate the use nature and scale requirements of the building, including building type, use function, space area and number of floors. Among them, building type indicates the use classification of the building, such as residential, office, commercial, etc.; use function further refines the name and location relationship of each functional space; space area refers to the use area and size requirements of each room; the number of floors determines the height and vertical space organization of the building. Architectural style requirement parameters refer to the architectural aesthetic style expected by the owner or designer, including exterior style and interior decoration style. The exterior style is related to the overall image and artistic expression of the building, such as modern, classical, postmodern and other schools; the interior decoration style is reflected in the choice of space atmosphere and decorative materials, such as simple, luxurious, industrial style, etc. For example, a residential community located in a cold northern region, with a large site height difference and close to a commercial district, the owner requires a new Chinese style, a three-bedroom and two-living room apartment, and a total number of floors not exceeding 6.

[0072] Specifically, when the user starts a new architectural design task, the system first needs to guide the user to enter a series of design condition parameters through the interactive interface. For the building site, the user needs to select a specific location on the map, and the system automatically obtains its geographical coordinates, weather information of the city, etc.; then the user describes the terrain undulations, road red lines, etc. of the site by uploading CAD drawings or drawing online, and marks the location and height of existing buildings in the surrounding area. For the building function, the user needs to select the building type of this project from the drop-down menu, and list the key attributes such as the name, area, and number of floors of each functional space by filling in the form. For the architectural style, the system provides a variety of preset style templates for users to choose from, and users can also upload style pictures as customized design references. After completing the input of all parameters, the system passes these structured data to the background architectural design generation model. The model is a deep learning network trained with a large number of excellent cases, which can establish a mapping relationship between design conditions and design plans. The model comprehensively considers the restrictions and guidance of each parameter, iteratively optimizes and generates a set of preliminary design plans, including conventional design drawings such as building floor plans, elevations, and sections. It is worth noting that the scheme is not unique, but based on the design knowledge learned in the model training, it provides a design idea for reference and further modification. Therefore, the system will display the generated initial scheme drawings to the user and provide online editing drawing tools, allowing users to adjust and deepen the scheme according to their own design experience and intentions. After multiple rounds of modifications by the user, the system records the finalized architectural drawings as the basis for subsequent design optimization and visualization.

[0073] In some embodiments, the architectural design system can realize the generation of initial design solutions from design conditions in a variety of ways: Optionally, the architectural design generation model of the system adopts the network structure of GAN (Generative Adversarial Network), learns the intrinsic distribution law of architectural design drawings through game training of generators and discriminators, and then randomly samples according to the input condition parameters to generate diversified design solutions; Optionally, the system uses the idea of ​​reinforcement learning to regard the generation of design solutions as a sequential decision-making process, and each decision step outputs a design action (such as drawing walls, doors and windows, etc.) through a neural network, and optimizes the action sequence according to reward functions such as design specifications and user feedback, and finally obtains a feasible design solution; Optionally, the system can also use the strategy of transfer learning to pre-train a backbone network that understands the semantics of drawings on a large amount of image data, and then fine-tune the network weights for specific design tasks using a small number of cases, quickly adapt to new design conditions, and generate customized design results. It is understandable that the system can also integrate other machine learning algorithms, such as support vector machines, decision trees, etc., to enhance the performance and generalization ability of the architectural design generation model from different angles, which is not limited here.

[0074] In some embodiments, the architectural design system inputs the building site environmental parameters into the architectural design generation model, generates a three-dimensional environmental model of the building site, and optimizes the coordination between the building layout and the site environment to obtain a preliminary layout of the building floor plan; the optimization process includes adjusting the building orientation, volume mass and spacing through ventilation environment and sunlight analysis; inputting the building functional attribute parameters into the architectural design generation model to generate the layout relationship, space dimensions and entrance and exit positions of each functional space inside the building to obtain the building floor plan of each floor of the building; based on the preliminary layout of the building floor plan and the floor plans of each floor of the building, the building outline is generated to obtain the building facade drawing; the architectural style requirement parameters are input into the architectural design generation model to generate the internal and external facade design; based on the building facade drawing and the internal and external facade design, a building section drawing is generated.

[0075] Among them, the building site environmental parameters refer to the data describing the characteristics of the environment in which the building is located, including topography, climate characteristics, surrounding building conditions, etc., which are used to represent the relationship between the building and the site. The three-dimensional environmental model of the building site represents a three-dimensional digital description of the site environment, including elements such as terrain, landforms, and vegetation, and is usually generated by technologies such as digital elevation model (DEM) and city information model (CIM). The coordination between the building layout and the site environment refers to the design effect that the spatial form and scale ratio of the building echo each other with the natural and humanistic attributes of the site, and is harmonious and unified. For example, the orientation of the building considers welcoming favorable wind directions, and the shape considers conforming to the terrain trend. The preliminary layout of the building plan indicates the spatial arrangement relationship and scale ratio of the functional zones of the building, which is the design result of the conceptual scheme stage. Ventilation environment analysis is used to simulate the wind field distribution law around the building and optimize the comfort of the wind environment. Sunlight analysis is used to simulate the impact of solar radiation on the surface and internal space of the building at a specific geographical location and time, and optimize the lighting and thermal environment. Building volume refers to the three-dimensional scale characteristics of the building, including length, width, and height, which is the key parameter for controlling the scale of the building. Building quality refers to the properties of the components and materials that make up the building, such as area and volume, and is the basis for calculating the engineering quantity. Building spacing refers to the distance between adjacent buildings, which is related to the density, ventilation environment, fire safety and other indicators of the building complex. Building function attribute parameters refer to the use attributes of each component space of the building, such as the name, area, floor height, orientation, comfort requirements, etc. of the room. The floor plan of each floor of the building represents the functional layout design of the corresponding standard floor, including the specific shape, dimensioning, door and window positions and other graphic contents of each room. The building outline refers to the three-dimensional body characteristics of the building, which is composed of the plane shape and spatial form of the building. The building elevation represents the two-dimensional projection of the building's elevation in all directions, focusing on the building's height, number of floors, window form, exterior wall material and other facade elements. The architectural style requirement parameters are used to describe the architectural style characteristics expected by the owner or designer, such as modern, classical, simple, etc. The internal and external facade design represents the decoration design of the internal and external walls of the building, including the form and material of facade components such as doors, windows, railings, and curtain walls. The building section represents the internal structure after the building or site is cut longitudinally or transversely, and is often used to express the internal space combination, floor elevation, vertical transportation, etc. of the building.

[0076] Specifically, after receiving the architectural design task input by the user, the architectural design system first identifies and extracts the environmental parameters of the building site, such as the surrounding terrain, roads, vegetation, and weather, and inputs them into the architectural design generation model, which generates a three-dimensional environmental model of the building site. Then, the system uses built-in environmental simulation analysis tools, such as ventilation environment and sunshine analysis, to evaluate the impact of building orientation, volume, spacing and other parameters on indoor and outdoor comfort under different building layout schemes, and iteratively optimizes the best building layout that is coordinated with the site environment, and outputs the architectural floor plan of the conceptual stage. Next, the system identifies the qualitative and quantitative requirements for each functional space of the building input by the user, converts them into architectural functional attribute parameters, and inputs them into the design generation model. The model generates the layout relationship, spatial dimensions and entrance and exit locations of each functional space that meet the building use requirements by solving the spatial relationship constraint problem, and outputs the floor plan of each floor of the building. Based on the generated floor plans of each floor, the system uses parametric modeling technology to automatically generate the three-dimensional outline of the building, extract the elevation projections of the building in all directions, and obtain the building elevation. Finally, the system matches the architectural style requirement parameters entered by the user, such as modern, classical, decorative, etc., retrieves similar instances from the massive building facade case library collected in advance, and transfers the style characteristics of the instances to generate interior and exterior facade design plans that match the building facade and style positioning. Combined with the facade design, it generates a cross-section drawing that details the interior space of the building, completing the preliminary design process of the architectural plan.

[0077] In some embodiments, the architectural design system can realize the generative design from site environment to architectural scheme in a variety of ways: Optionally, the system adopts a rule-based parametric design method, predefines a series of parametric logic rules describing the relationship between architectural form and site, and after inputting the architectural site parameters, quickly generates a three-dimensional architectural model that meets the rule constraints by calculating the functional relationship between the parameters; Optionally, the system adopts a data-driven end-to-end deep learning method, uses a large number of excellent architectural design cases as training data, trains a mapping relationship model between architectural images and site environment, inputs the building site environment image, and the model directly outputs the architectural plan, elevation and section images that match the environment; Optionally, the system adopts a cognitive reasoning method that integrates expert knowledge and machine learning, formalizes the architectural design process into a series of knowledge rules and decision sequences, simulates the designer's step-by-step decision-making process through a knowledge reasoning engine, and interacts and iterates with the environmental analysis module to generate an architectural scheme that meets both the design specifications and the site characteristics. It can be understood that other computational design methods can also be used to realize the intelligent generation of architectural three-dimensional models, such as evolutionary optimization algorithms, morphological grammars, topological optimization, etc. This system is not limited by architectural generation technology means and is not limited here.

[0078] S202. Determine architectural design features based on architectural design plan data.

[0079] Referring to step S102 , the architectural design system determines architectural design features.

[0080] S203, identifying different functional space areas and their positional relationships in the building plan, and dividing the functional zoning layouts of different floors according to the identification results.

[0081] Among them, the architectural floor plan refers to a pattern that shows the internal space layout of a building from a bird's-eye view, including room boundaries, door and window locations, coordinate dimensions and other information. Functional space areas represent the division of rooms in the architectural plane, such as living room, bedroom, kitchen, bathroom, etc., reflecting the usage attributes of each room. Positional relationship refers to the topological or metric relationship between functional spaces, such as spatial adjacency and accessibility. Functional zoning layout is used to represent the combination of several functional spaces into areas to form spatial configuration plans for different floors in the horizontal and vertical directions. For example, in a residential floor plan, the living room, dining room, and kitchen are usually arranged in one area to form a public activity area; the bedroom and bathroom are combined to form a private rest area; the public area and the private area are relatively independent and interconnected in space, reflecting a good functional zoning idea.

[0082] Specifically, once the system obtains the floor plan of each floor of the building, it can start to identify the functional zoning layout. The system first uses computer vision methods to parse the floor plan into computable semantic data through image segmentation, target detection and other technologies, and extract geometric elements such as room boundary boxes, door and window positions. On this basis, the system uses rule-based or machine learning classification methods to judge the functional attributes of each room and identify different functional spaces in the plane according to the room shape characteristics, area size, number of doors and windows, equipment symbols, etc. Then, the system constructs a room topology diagram by analyzing the boundaries of adjacent rooms, the relationship between entrances and exits, etc., and explores the connection patterns between functional spaces. Finally, the system uses algorithms such as spatial clustering and graph segmentation to divide closely related functional spaces into the same area, while considering factors such as the independence, compactness, and visibility of each area to generate a reasonable functional zoning layout. According to the number of floors of the building and the elevation information of each floor, the system expands the zoning layout to the vertical direction to form a complete three-dimensional functional layout.

[0083] In some embodiments, the system can realize the functional zoning layout of the building plan in a variety of ways: Optionally, the system adopts a semantic-based logical reasoning method to abstract the plan layout into a series of symbolic representations of rooms and relationships, and uses the knowledge of the spatial layout field to construct reasoning rules, such as the living room should be close to the entrance, the bedroom should be far away from the kitchen, etc., and the optimal zoning scheme is obtained through symbolic calculation; Optionally, the system applies a data-driven deep learning method, using a large number of excellent graphic design cases as training data, and learns the mapping relationship from the plan to the functional partition through an end-to-end neural network to realize the automatic generation of the partition layout; Optionally, the system introduces natural language processing technology to convert textual knowledge such as architectural design specifications and case analysis reports into computable constraints to guide the functional zoning reasoning and optimization process. It is understandable that the system can also use other spatial division methods to obtain the functional zoning layout of the building, such as force-guided layout based on physical simulation, particle swarm optimization based on swarm intelligence, etc., to ensure the rationality of the function while improving the richness of the layout, which is not limited here.

[0084] S204, obtaining the building exterior wall material, exterior shape and vertical door and window opening positions of the building elevation drawing to generate the exterior facade.

[0085] Among them, the building elevation drawing refers to a drawing that shows the vertical view of the building's orthographic projection, showing the material texture, line shape, door and window openings and other design elements of the building's exterior. The exterior wall material refers to the actual material of the building's surface, such as bricks, paint, glass, metal plates, etc., reflecting the color and texture of the facade. The exterior shape refers to the outline lines, decorative patterns, shadow changes, etc. of the facade, which shape the overall visual image of the building. The location of the door and window openings reflects the virtual-real contrast, density rhythm, etc. of the facade. The facade is used to represent a specific orientation view of the building formed by integrating the above design elements. For example, the main entrance facade of a theater building uses a large area of ​​glass curtain wall material, combined with simple and elegant line shape, and a striking identification entrance is set up, giving people a transparent and flowing modern feeling.

[0086] Specifically, while the system identifies the building's floor plan, it also obtains the corresponding facade information. First, the system uses an image segmentation algorithm to divide the facade into different parts such as exterior walls, doors and windows, and roofs, and extracts the geometric boundaries of each part. Then, the system uses a material recognition technology based on texture classification to determine the exterior wall material information corresponding to each part based on the color and texture characteristics of the image. For the lines and patterns of the facade, the system obtains the geometric modeling information related to the facade through methods such as contour extraction and shape detection. At the same time, the system restores the proportional dimensions of the facade and calculates the position coordinates of the door and window openings. Through the comprehensive use of information such as materials, shapes, and openings, the system can generate a building facade model in three-dimensional space. It is worth noting that the design information extracted by the system from different views such as plane, elevation, and section needs to be verified with each other to ensure the integrity and consistency of the model. For example, if the system learns from the facade that the exterior wall is made of red brick material, it is necessary to mark the outer wall as the corresponding material in the floor plan.

[0087] In some embodiments, the system can achieve three-dimensional reconstruction of building facades in a variety of ways: optionally, the system adopts a procedural modeling method to predefine a series of facade composition rules, such as parametric templates such as exterior wall partition patterns and window arrangements, and then maps the facade information obtained from the drawings to relevant parameters to automatically construct a three-dimensional facade that meets the design intent; optionally, the system uses parametric design tools to formalize the manual modeling process as a set of parameters, constraints, and rules, and generates and optimizes the three-dimensional facade by adjusting parameter values ​​and solving constraints; optionally, the system integrates multiple modeling methods such as visual programming and BIM modeling, flexibly organizes the modeling process in a visual environment, intuitively controls design parameters, and realizes interactive facade modeling and optimization.

[0088] S205. Obtain the floor height, floor thickness, and beam-column structure of the building cross-section in the building interior space, and generate the internal structure.

[0089] Among them, the building section drawing refers to a drawing that shows the internal space composition along a certain vertical section of the building, reflecting the height relationship and structural construction method of the various layers of the building. The floor height refers to the vertical distance between the ground of two adjacent floors, which determines the sense of scale of the space. The floor thickness refers to the height difference between the upper and lower surfaces of the floor structure, which affects the overall height of the building. The beam-column structure refers to the load-bearing components of the building, such as beams, slabs, columns, etc., which transmit and disperse the load of the building. The internal structure is used to represent the internal spatial structural relationship and entity composition of the building. For example, in the section drawing of a split-level residential unit, the living room has a higher floor height and the bedroom has a lower floor height, and the two are connected by a floor slab; the floor slab adopts a reinforced concrete structure, and the thickness meets the requirements of the specification; the beams and columns above the living room can be cantilevered to form a rich spatial change.

[0090] Specifically, after the system obtains the section information of the building, it can disassemble it and extract the key elements of the internal structure. The system first identifies the floor lines, column symbols, etc. in the section view, and obtains the height data of each layer of space through image processing. For the floor slab, the system obtains the thickness value of the floor slab by analyzing the cross-sectional performance of the floor slab. For beams and columns, the system uses symbol recognition technology to extract their position and size information. On the basis of obtaining various types of data, the system comprehensively considers the existing information such as the functional zoning and facade of the building, and gradually constructs the internal structure in three-dimensional space. The system generates the corresponding floor plan according to the elevation of each floor; then embeds the floor slab, beams and columns and other components into the plane to form a complete floor structure; finally, assemble each floor vertically upward, and arrange the traffic core such as elevators and stairs accordingly, and finally form a three-dimensional framework inside the building. During the construction process, the system fully considers the architectural design specifications, such as the thickness and span of the floor slab must meet the structural force requirements, and the arrangement of beams and columns must meet the space use requirements, etc., to ensure the rationality of the internal structure.

[0091] In some embodiments, the system can realize the generation of the internal structure of the building in a variety of ways: Optionally, the system adopts a parametric modeling method, defines the floor slab, beams and columns as preset family files, dynamically adjusts the parameters of the family by reading the profile information, and automatically arranges the components in the space based on the constraints, so as to quickly build the internal structure of the building; Optionally, the system uses three-dimensional reconstruction technology, such as SFM, SLAM, etc., to understand the profile as a cross-sectional observation of the building under a specific perspective, restore the three-dimensional shape through the principle of multi-view geometry, and consider the semantic information to distinguish and assemble the components; Optionally, the system integrates rule reasoning and deep learning technology, on the one hand, uses engineering specification knowledge to impose constraints on structural construction, and on the other hand, learns the internal logic of building construction through end-to-end neural networks to realize the reasoning generation of internal structure. It is understandable that the system can also apply other three-dimensional construction methods, such as programmatic generation, 3D matching assembly, etc., to improve the efficiency of internal structure generation while ensuring the rationality of building construction, which will not be elaborated here.

[0092] S206. Combine the functional zoning layout, the exterior facade and the interior structure to obtain a complete three-dimensional building geometric model.

[0093] Functional zoning layout, facade and internal structure are the three key components of the system pre-generation: among them, the functional zoning layout determines the plane division and vertical organization of the building's internal space; the facade reflects the building's external shape and material texture characteristics; the internal structure reflects the building's skeleton and the logic of the space. The three-dimensional architectural geometric model refers to the organic combination of the above three parts, integrated into a unified digital three-dimensional model, which fully expresses the spatial form and composition relationship of the building. For example, a modern art gallery has its internal space divided into three functional areas: exhibition, public service, and office. The facade highlights the design concept through special-shaped curved surface shapes and transparent glass curtain walls. The interior uses steel structure large-span beam-slab floor slabs. Finally, the three are integrated to form an architectural image with great artistic appeal.

[0094] Specifically, after the system completes the extraction of geometric information of each part of the building and separate modeling, it begins to integrate and assemble the model. First, the system combines the floor plan layout with the vertical floor parameters, and translates, rotates and splices them in three-dimensional space to form the overall spatial outline of the building. Then, the system coordinates the facade model with the spatial outline, so that the facade outline coincides with the plane boundary, and hollows out the door and window openings in the facade model, so that the attributes such as the exterior wall material and detailed decoration are assigned to the model surface. Next, the system fills the floor slab, beam and column and other component models inside the spatial outline, so that each floor plane has a reasonable load-bearing structure, and embeds the vertical traffic models such as elevators and stairs into it, and finally establishes a complete internal structure of the building. It is worth noting that during the assembly process, the system automatically detects the geometric relationship between the components, such as the consistency of the internal and external wall contours, the permeability of the door and window openings, etc., and adjusts them in time when problems are found to ensure the accuracy and logic of the model. After the system's meticulous assembly, a building information model with precise geometry, clear structure and complete semantics is constructed.

[0095] In some embodiments, the system can achieve the integration of building components in a variety of ways: optionally, the system converts each part model into a unified scene tree representation based on a three-dimensional space geometry engine, and records the assembly logic of the model through structured information such as parent-child relationships and transformation matrices to achieve efficient scene organization and management; optionally, the system adopts a semantic BIM modeling method to map each professional model into an IFC standard component object, achieves model integration through a unified data structure, and adds attribute information such as quality parameters and related documents to enhance the intelligence of the model.

[0096] S207. Based on the material information in the architectural design features, corresponding material maps are applied to different components of the three-dimensional architectural geometric model. Based on the lighting conditions in the architectural design features, light source environment parameters of the three-dimensional architectural geometric model are configured to obtain a three-dimensional architectural model.

[0097] Among them, material information refers to the type, color, texture and other performance attributes of various decorative materials marked in the architectural design drawings, such as wood, stone, paint, etc. Material mapping uses computer graphics technology to map the surface texture and concave and convex texture of the material to the surface of each component of the three-dimensional model to simulate the rendering effect of the real material. Lighting conditions represent the sunlight conditions of the building environment, such as the direction and intensity of the light source. The light source environment parameters are used to simulate the physical phenomena of light reflection, refraction, absorption on the building surface, and control the rendering elements such as the light and dark changes and shadow generation on the building surface. The three-dimensional building model refers to the digital model of the building rendered with realistic materials and lighting parameters, which highly realistically reproduces the spatial form and real materials of the building, and provides a three-dimensional scene foundation for subsequent design evaluation, scheme display and other links. For example, in an interior space design, the floor is paved with wooden floors, the walls are painted with latex paint and decorated with wallpaper, and warm-toned lighting and natural lighting are used to create a warm and comfortable atmosphere.

[0098] Specifically, after the system builds the three-dimensional geometric model of the building, it enters the rendering stage of the material and lighting environment. For the material information, the system first parses it into a standardized material description, including material type, color RGB value, texture pattern, etc. Then, the system uses UV mapping, texture coordinates and other technologies to establish the mapping relationship between the surface mesh of the three-dimensional model and the two-dimensional material map. At the same time, the system also considers the physical properties of the surface such as convexity and reflectivity to select a suitable shader model for the material. Finally, the system applies the generated material to various parts of the building model, such as curtain walls, roofs, exterior walls, etc., and simulates a realistic and credible material effect through real-time preview and repeated adjustments. For the lighting environment, the system first obtains the longitude and latitude coordinates of the building location, determines the changing laws of the angle and intensity of sunlight, and simulates the dynamic changes of natural light within a day; on the other hand, the system reasonably sets artificial lighting sources according to the lighting design of each space in the building, and coordinates the combined effects of natural light and artificial light. After the light source is arranged, the system calculates the propagation path of light on the building surface, simulates the interaction between light and material, and takes into account phenomena such as light scattering and color penetration, and finally renders lifelike light and shadow effects and a rich and delicate sense of space atmosphere. It is worth mentioning that the system also supports multiple modes such as dynamic roaming and interactive operation, allowing designers to review the model details in all aspects and gain an immersive spatial experience.

[0099] In some embodiments, the system can achieve material and lighting rendering of building models in a variety of ways: Optionally, the system is based on GPU hardware-accelerated ray tracing technology, by simulating the physical interaction between light and material surface, real-time calculation of surface light and shade changes, light refraction, etc., to generate a highly realistic rendering picture, and support dynamic adjustment of material and light source parameters; Optionally, the system adopts a global illumination model with pre-calculated radiance, and achieves realistic soft lighting and rich color performance through offline baking of scenes, photon mapping and other methods, while saving rendering computing resources; Optionally, the system uses cloud rendering, distributed computing and other technologies to decompose complex lighting material rendering tasks to multiple computers for parallel processing, significantly improving the rendering speed and effect, while facilitating cross-platform collaboration. It is understandable that the system can also integrate other graphics drawing and post-processing algorithms, such as halo effects, to further enhance the artistic expression and visual impact of the rendered scene, which will not be elaborated here.

[0100] In some embodiments, the architectural design system will perform parametric simulation analysis of the building performance of the generated three-dimensional building model according to the requirements of the architectural design specifications to obtain performance evaluation results; the building performance includes lighting, ventilation, energy saving and fire prevention; the performance evaluation results are displayed, and the parameters of the three-dimensional building model are optimized based on the artificial intelligence model; the three-dimensional building model after parameter optimization is obtained, and then rendered to generate the building scene renderings.

[0101] Specifically, after the system generates a preliminary three-dimensional model of the building, it automatically performs a series of performance evaluations on it. The first is the lighting performance. Based on the geographical latitude of the building, the system calculates the solar altitude angle and azimuth in different seasons and at different times. Combined with the window opening information, window lintel height, window sill depth, etc. of the three-dimensional model, the system simulates the propagation of light in the room through the ray tracing algorithm, obtains quantitative indicators such as the daylighting factor and sunshine duration of each room, and compares them with national standards to diagnose the problem areas of insufficient lighting. The second is ventilation performance. The system reads the local perennial dominant wind direction, wind speed and other meteorological information, and uses the computational fluid dynamics (CFD) method to simulate the flow law of indoor and outdoor air under the action of wind, and obtains the ventilation and air exchange times and wind pressure distribution data of each area to evaluate the natural ventilation effect of the building. The third is energy-saving performance. Based on the thermal insulation performance and airtightness performance of the building envelope structure, as well as the local heating and cooling design parameters, the system uses the energy consumption simulation engine to calculate the building's heating, cooling, lighting, hot water and other energy consumption indicators throughout the year, and gives optimization suggestions for energy-saving design. Finally, the fire performance, the system calculates and verifies the building's fire zoning, safe evacuation, fire resistance, etc., to ensure that the building design meets the requirements of fire protection regulations. After completing the performance evaluation, the system will present the quantitative results in the form of visual charts, marking the compliance of key indicators. At the same time, the artificial intelligence model in the background of the system will automatically adjust the building parameters that affect performance according to the evaluation results, such as window-to-wall ratio, shape coefficient, shading measures, etc. After multiple rounds of iterative optimization, the building design intent is taken into account while ensuring the building performance. Finally, the system updates the three-dimensional building model and submits it to the rendering engine for light and shadow material rendering to obtain realistic building scene renderings, allowing designers and owners to experience the design space in person.

[0102] In some embodiments, the architectural design system can realize parametric simulation analysis of building performance in a variety of ways: Optionally, the system adopts mature building performance simulation engines such as EnergyPlus and Radiance, which are internationally used, converts the BIM model into the input file format required by the engine through the interface, and calls the API interface provided by the engine to realize batch calculation of performance indicators such as lighting and energy consumption; Optionally, the system develops its own lightweight building performance evaluation algorithm library, abstracts the relationship between building design parameters and performance indicators, responds to the adjustment of design parameters in real time, and supports seamless integration and interface interaction with design software; Optionally, the system also builds a distributed building performance simulation platform based on cloud computing, and users can upload three-dimensional models to the cloud, call the simulation service interface provided by the platform, select the performance type to be analyzed, and submit concurrent simulation tasks. The cloud platform is responsible for managing computing resources and task scheduling, and users only need to wait for the result feedback. It can be understood that the simulation analysis of building performance is a complex system engineering, which needs to consider many influencing factors and boundary conditions. The system can reasonably select different simulation analysis technical solutions according to actual needs to take into account factors such as analysis accuracy, calculation efficiency and ease of use, which are not limited here.

[0103] In some embodiments, the building design system will visualize the performance analysis results; the visualization of the performance analysis results specifically includes: displaying the lighting analysis results in the form of a room interior brightness cloud map, displaying the ventilation analysis results in the form of animated lines of air flow, and displaying the energy-saving analysis results in the form of an energy consumption indicator radar map; receiving user interaction instructions for problem areas in the visualized performance analysis results, and adjusting relevant parameters of the three-dimensional building model accordingly; or, analyzing the performance evaluation results based on the artificial intelligence model, adjusting relevant parameters; based on the adjustment operation of the relevant parameters, updating the three-dimensional building model, and synchronously updating the visualization of the performance analysis results.

[0104] Specifically, the brightness cloud map refers to an image that uses pseudo-color coding to represent the distribution of lighting brightness levels at various locations inside the room, which directly reflects the uniformity and adequacy of the room's lighting. The animated lines of air flow are used to describe the flow path and velocity field of air inside and outside the building. The warmer the line color, the higher the wind speed. The animated flow direction represents the wind direction, which directly expresses the rationality of the building's ventilation organization. The energy consumption index radar map is used to compare the performance of the building in terms of energy consumption items such as heating, cooling, lighting, and hot water with the compliance with national standards. The dimensions of each indicator are normalized and presented in a unified polygonal area. The smaller the area, the higher the overall energy efficiency level. User interaction instructions refer to the operation of the user observing and judging the visualization results through the human-computer interface, using input devices such as the mouse and keyboard to select the area with performance problems in the figure, and manually adjust the parameters of related components, such as dragging windows and changing wall materials. Artificial intelligence model generally refers to the use of machine learning algorithms to automatically learn and summarize the relationship between architectural parameters such as spatial layout, material selection, and appearance design and performance indicators such as lighting, ventilation, and energy saving from a large number of excellent architectural design case data, and to assist and optimize the calculation model of building performance. A three-dimensional building model refers to a collection of three-dimensional geometric and semantic information that digitally describes the spatial form, component composition, material properties, etc. of a building. It is the premise and basis for conducting building performance analysis and visualization.

[0105] Specifically, after generating a three-dimensional model of a building, the architectural design system will analyze and evaluate the building's performance indicators such as lighting, ventilation, and energy saving based on the physical environment simulation engine, and present the analysis results to the user in an intuitive visual form. Among them, the lighting analysis results are represented by a cloud map of the brightness distribution at various locations inside the room. The warmer the color, the more sufficient the lighting; the ventilation analysis results are represented by animated lines of air flow. The warmer the line color, the higher the wind speed, and the animation direction represents the wind direction; the energy saving analysis results are represented by a radar map of energy consumption indicators. The dimensions of energy consumption indicators such as heating, cooling, lighting, and hot water are normalized and presented in a unified polygonal area. The smaller the area, the higher the comprehensive energy efficiency. Users can select areas with problems in the performance analysis results through the human-computer interaction interface and give adjustment instructions, such as stretching windows to improve lighting, adjusting exterior wall materials to improve insulation, etc. After receiving user instructions, the system will automatically modify the relevant parameters of the building's three-dimensional model, run the performance analysis again, and update the visualization results in real time until the user's needs are met. In addition, the system can also build in expert experience and machine learning algorithms, summarize performance optimization strategies based on massive cases, and automatically adjust parameters of the building model to reduce the user's interaction burden and improve the efficiency of performance optimization. As the building 3D model is updated, the corresponding performance analysis results will also be updated synchronously to ensure the consistency of the visualization display with the current design plan.

[0106] In some embodiments, the building design system can realize the analysis and optimization of building performance in a variety of ways: Optionally, the system calls mature engineering simulation software, such as Radiance, EnergyPlus, CFD, etc., maps the material, space, and equipment parameters of the three-dimensional model as the input items of the software, and conducts batch special simulation analysis of light, heat, wind, sound, etc., and then converts the numerical value of the analysis results into a visual effect diagram; Optionally, the system exerts the regression prediction ability of the artificial intelligence algorithm, and based on the accumulated design case data, trains the mapping relationship between the building model parameters and the performance indicators, realizes the rapid estimation of performance and the sensitivity analysis of the optimization direction, and is used to guide the automatic tuning of the parametric model; Optionally, the system integrates the industry's common design specifications and evaluation standards, develops a performance evaluation knowledge base and rule base, and checks the building model item by item through the knowledge reasoning engine, diagnoses the performance shortcomings that do not meet the requirements of the specifications, and gives optimization suggestions under parameter constraints. It is understandable that other technical means can also be used to realize building performance analysis and optimization, such as complex system multi-objective optimization, swarm intelligence algorithm, reinforcement learning, etc., and this system does not limit this.

[0107] S208. Rendering and optimizing the three-dimensional building model based on the artificial intelligence model to generate building scene renderings from multiple different perspectives.

[0108] Referring to step S104, the architectural design system generates a plurality of architectural scene renderings from different perspectives.

[0109] S209: Obtain effect estimation parameters of the artificial intelligence model for the building scene rendering, and generate evaluation results in combination with correction instructions input by the user to correct the effect estimation parameters.

[0110] Referring to step S105 , the architectural design system generates an evaluation result.

[0111] In some embodiments, the architectural design system will estimate the visual realism and design intent matching of multiple architectural scene renderings based on an artificial intelligence model, and obtain initial realism scores and initial intent matching scores corresponding to the multiple architectural scene renderings as effect estimation parameters; display multiple architectural scene renderings, corresponding initial realism scores and initial intent matching scores; after receiving a user's selection of one of the architectural scene renderings, provide a correction instruction for the initial realism score and / or initial intent matching score of the selected architectural scene rendering; correct the initial realism score and / or initial intent matching score according to the correction instruction to obtain an evaluation result of the corresponding architectural scene rendering.

[0112] Specifically, for a series of architectural scene renderings generated by the system, before submitting them to designers and owners for review, the system first uses a pre-trained computer vision model to conduct a preliminary evaluation and scoring of the visual quality and design value of the image. Among them, visual realism refers to the drawing effect of the rendered image. Through the image quality assessment algorithm, the image clarity, color richness, light and dark contrast, texture details and other angles are used to quantify the real feeling brought to the human eye by the image, and an objective realism score is obtained. The design intent matching refers to whether the rendering can accurately convey the core concept of the design. The system uses image semantic segmentation and scene understanding technology to extract and understand the content elements of the image, such as spatial composition, material texture, light and shadow atmosphere, etc., and then compares the extracted feature representation with the design description text, calculates the semantic similarity, and finally obtains the matching degree score of the rendering and the design intent. The system uses image realism and intent matching as the initial screening criteria to automatically filter out rendering schemes with lower scores. On this basis, the system submits the remaining better renderings and corresponding scores to designers and owners, who make subjective judgments based on their professional experience and personal preferences. Designers and owners can not only choose the most satisfactory solution, but also make corrections to the reality and matching scores given by the system. For example, the system may evaluate the reality score of an artistic non-realistic rendering as too low due to the lack of human aesthetic cognition. Designers can use this to improve the reality score of the image to make it consistent with the designer's subjective evaluation. For another example, a rendering may focus too much on the performance of the building itself and ignore the rendering of the environmental background, resulting in the overall image being inconsistent with the scene atmosphere described by the design intent. The owner can use this to lower the intention matching score of the image. The system collects feedback from designers and owners on the scoring of each rendering, and updates the scoring labels in the image database. These annotated data can be used to subsequently optimize the system evaluation model to make the scoring results closer to the real ideas of designers and owners.

[0113] In some embodiments, the architectural design system can achieve visual evaluation and intent matching of architectural renderings in a variety of ways: Optionally, the system adopts an end-to-end image aesthetic evaluation model based on deep learning, trains neural networks on large-scale high-quality image data sets, learns human preferences and judgment criteria for beauty, and then migrates them to the visual evaluation task of architectural renderings to achieve automatic prediction of authenticity scores; Optionally, the system applies image caption automatic generation technology to allow the AI ​​model to generate corresponding text descriptions based on the image content, and then compares the generated text with the design instructions, calculates their similarity in the semantic space, and obtains the intent matching score; Optionally, the system can also introduce a crowdsourcing manual scoring mechanism, publish renderings to designer and owner communities, collect users' subjective scores on multiple dimensions such as authenticity, creativity, and artistry, and then remove noise and abnormal scores through clustering and combined sorting algorithms to obtain relatively objective and fair evaluation results. It is understandable that visual realism and the degree of match between design intent are themselves relatively subjective and difficult to quantify evaluation criteria. Although existing artificial intelligence technology can give reasonable evaluation predictions, it is still difficult to completely replace the judgment of the human eye. Therefore, in practical applications, the system often needs to combine automatic machine scoring with manual scoring to complement each other's advantages in order to obtain a reliable and convincing evaluation plan, which is not limited here.

[0114] S210. According to the evaluation result, the three-dimensional building model is re-rendered and optimized using the artificial intelligence model until the evaluation result of the generated building scene rendering meets the preset threshold.

[0115] Referring to step S106 , the architectural design system will repeatedly perform rendering optimization.

[0116] In the embodiments of the present application, due to the use of artificial intelligence-driven architectural design generation technology, the machine learning model automatically understands and extracts the key points of the design drawings, and maps them to the generation and rendering of the three-dimensional model, so it can maximize the consistency between the design intent and the three-dimensional performance, effectively solve the problem of easy errors in manual modeling and the mismatch between the renderings and the design drawings, and thus achieve seamless connection and high-fidelity conversion from design drawings to renderings. At the same time, the present application also introduces user interaction-guided parametric generation and human-computer collaborative optimization mechanisms, which, while giving full play to the advantages of artificial intelligence generation efficiency, well retains the designer's control and decision-making power over the design key points, and provides a new model of human-computer interactive architectural design that is both flexible and convenient.

[0117] The following describes the architectural design system in the embodiment of the present invention from the perspective of hardware processing. Figure 3 , which is a schematic diagram of a physical device structure of the architectural design system in an embodiment of the present application.

[0118] It should be noted that Figure 3 The structure of the architectural design system shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present invention.

[0119] like Figure 3 As shown, the architectural design system includes a central processing unit (CPU) 301, which can perform various appropriate actions and processes according to the program stored in the read-only memory (ROM) 302 or the program loaded from the storage part 308 to the random access memory (RAM) 303, such as executing the method described in the above embodiment. In the RAM 303, various programs and data required for system operation are also stored. The CPU 301, the ROM 302 and the RAM 303 are connected to each other through the bus 304. The input / output (I / O) interface 305 is also connected to the bus 304.

[0120] The following components are connected to the I / O interface 305: an input section 306 including an audio input device, a button switch, etc.; an output section 307 including a liquid crystal display (LCD) and an audio output device, an indicator light, etc.; a storage section 308 including a hard disk, etc.; and a communication section 309 including a network interface card such as a LAN (Local Area Network) card, a modem, etc. The communication section 309 performs communication processing via a network such as the Internet. A drive 310 is also connected to the I / O interface 305 as needed. A removable medium 311, such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, etc., is installed on the drive 310 as needed so that a computer program read therefrom is installed into the storage section 308 as needed.

[0121] In particular, according to an embodiment of the present invention, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program includes a computer program for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication part 309, and / or installed from a removable medium 311. When the computer program is executed by the central processing unit (CPU) 301, various functions defined in the present invention are performed.

[0122] It should be noted that specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM), a flash memory, an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device.

[0123] The flowcharts and block diagrams in the accompanying drawings illustrate the possible architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. Each box in the flowchart or block diagram may represent a module, a program segment, or a part of a code, and the above-mentioned module, program segment, or a part of a code contains one or more executable instructions for implementing the specified logical functions. It should also be noted that in some alternative implementations, the functions marked in the box may also occur in an order different from that marked in the accompanying drawings.

[0124] Specifically, the architectural design system of this embodiment includes a processor and a memory, and a computer program is stored in the memory. When the computer program is executed by the processor, the method for automatically generating architectural design scenes based on artificial intelligence provided in the above embodiment is implemented.

[0125] As another aspect, the present invention further provides a computer-readable storage medium, which may be included in the architectural design system described in the above embodiment; or may exist independently without being assembled into the architectural design system. The above storage medium carries one or more computer programs, and when the above one or more computer programs are executed by a processor of the architectural design system, the architectural design system implements the method for automatically generating architectural design scenes based on artificial intelligence provided in the above embodiment.

[0126] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

[0127] As used in the above embodiments, the term "when..." may be interpreted to mean "if..." or "after..." or "in response to determining..." or "in response to detecting...", depending on the context. Similarly, the phrases "upon determining..." or "if (the stated condition or event) is detected" may be interpreted to mean "if determining..." or "in response to determining..." or "upon detecting (the stated condition or event)" or "in response to detecting (the stated condition or event)", depending on the context.

[0128] Those skilled in the art can understand that to implement all or part of the processes in the above-mentioned embodiments, the processes can be completed by computer programs to instruct related hardware, and the programs can be stored in computer-readable storage media. When the programs are executed, they can include the processes of the above-mentioned method embodiments. The aforementioned storage media include: ROM or random access memory RAM, magnetic disk or optical disk and other media that can store program codes.

Claims

1. An automatic generation method of architectural design scenes based on artificial intelligence, characterized in that: Applied to a building design system, the method comprises: Acquire architectural design plan data; the architectural design plan data includes a building plan, a building elevation, and a building section; Determining architectural design features according to the architectural design scheme data; the architectural design features include space layout, material information and lighting conditions; Based on the architectural design features, generating a three-dimensional architectural model; Rendering and optimizing the three-dimensional building model based on an artificial intelligence model to generate architectural scene renderings at multiple different perspectives; the rendering optimization steps include matching material maps, setting light source parameters, and setting camera perspectives; the artificial intelligence model includes a convolutional neural network model, a generative adversarial network, or a reinforcement learning model; Obtaining effect prediction parameters of the artificial intelligence model for the architectural scene rendering, and generating an evaluation result in combination with a correction instruction input by a user to correct the effect prediction parameters; the evaluation result includes a visual effect reality score and a design intent matching score; the effect prediction parameters are generated by a contrastive learning neural network, a twin neural network or a joint loss function; According to the evaluation result, the three-dimensional building model is re-rendered and optimized using the artificial intelligence model until the evaluation result of the generated building scene rendering meets a preset threshold.

2. The method according to claim 1, characterized in that The obtaining of effect estimation parameters of the artificial intelligence model for the building scene rendering, and combining with a correction instruction input by a user to correct the effect estimation parameters, generating an evaluation result, specifically includes: Based on the artificial intelligence model, the visual realism and the design intention matching degree are respectively estimated for a plurality of architectural scene renderings, and the initial realism scores and the initial intention matching scores corresponding to the plurality of architectural scene renderings are obtained as effect estimation parameters; Displaying the plurality of architectural scene renderings, the corresponding initial reality scores and the initial intention matching scores; After receiving a user selecting one of the building scene renderings, a correction instruction for an initial reality score and / or an initial intention matching score of the selected building scene rendering; The initial reality score and / or the initial intention matching score are corrected according to the correction instruction to obtain an evaluation result of the corresponding building scene rendering.

3. The method according to claim 1, characterized in that: The generating of the three-dimensional building model based on the building design features specifically includes: Identify different functional space areas and their positional relationships in the building plan, and divide the functional zoning layout of different floors according to the identification results; Obtain the building exterior wall material, exterior shape and vertical door and window opening positions of the building elevation drawing to generate the exterior facade; Obtain the floor height, floor thickness and beam-column structure of the building section in the interior space of the building to generate the internal structure; Combining the functional zoning layout, the exterior facade and the interior structure to obtain a complete three-dimensional building geometric model; Based on the material information in the architectural design features, corresponding material maps are applied to different components of the three-dimensional architectural geometric model. At the same time, based on the lighting conditions in the architectural design features, the light source environment parameters of the three-dimensional architectural geometric model are configured to obtain a three-dimensional architectural model.

4. The method according to claim 1, characterized in that Before the step of obtaining the architectural design solution data, the method further includes: Determine the construction site environmental parameters selected and input by the user; the construction site environmental parameters include geographical location, climate conditions, topography and surrounding building distribution; Determine the building function attribute parameters selected and input by the user; the building function attribute parameters include building type, usage function, space area and number of floors; Determine the architectural style requirement parameters selected and input by the user; the architectural style requirement parameters include the exterior style and the interior decoration style; Inputting the building site environment parameters, building function attribute parameters and building style requirement parameters into a pre-trained building design generation model to generate a building plan, a building elevation and a building section as initial design scheme data; The initial design scheme data is displayed for the user to further edit and modify to obtain the final edited and modified architectural design scheme data.

5. The method according to claim 4, characterized in that The step of inputting the building site environment parameters, building function attribute parameters and building style requirement parameters into a pre-trained building design generation model to generate a building plan, building elevation and building section diagram specifically includes: Inputting the building site environmental parameters into the building design generation model, generating a three-dimensional environmental model of the building site, and optimizing the coordination between the building layout and the site environment to obtain a preliminary layout of the building plan; the optimization process includes adjusting the building orientation, volume mass and spacing through ventilation environment and sunlight analysis; Inputting the building function attribute parameters into the building design generation model, generating the layout relationship, space dimensions and entrance and exit positions of each functional space inside the building, and obtaining the building plan of each floor of the building; Based on the preliminary layout of the building plan and the floor plans of each floor of the building, a building outline is generated to obtain a building elevation; Inputting the architectural style requirement parameters into the architectural design generation model to generate interior and exterior facade designs; A building section drawing is generated based on the building elevation drawing and the interior and exterior elevation designs.

6. The method according to claim 1, characterized in that After the step of generating a three-dimensional building model based on the building design features, the method further comprises: According to the requirements of the building design specifications, the generated three-dimensional building model is subjected to parametric simulation analysis of building performance to obtain performance evaluation results; the building performance includes lighting, ventilation, energy saving and fire prevention; Displaying the performance evaluation results, and optimizing parameters of the three-dimensional building model based on the artificial intelligence model; Obtain the three-dimensional building model after parameter optimization, and then render it to generate the building scene rendering.

7. The method according to claim 6, characterized in that The displaying of the performance evaluation results and optimizing the parameters of the three-dimensional building model based on the artificial intelligence model specifically includes: Visually display the performance analysis results; the visual display of the performance analysis results specifically includes: displaying the lighting analysis results in the form of a room interior brightness cloud map, displaying the ventilation analysis results in the form of animated lines of air flow, and displaying the energy-saving analysis results in the form of an energy consumption index radar map; Receiving interactive instructions from users regarding problematic areas in the visualized performance analysis results, and adjusting relevant parameters of the three-dimensional building model accordingly; or, analyzing the performance evaluation results based on an artificial intelligence model, and adjusting relevant parameters; Based on the adjustment of relevant parameters, the 3D building model is updated, and the visual display of the performance analysis results is updated synchronously.

8. A building design system, characterized in that: The architectural design system includes: one or more processors and a memory; the memory is coupled to the one or more processors, the memory is used to store computer program code, the computer program code includes computer instructions, and the one or more processors call the computer instructions to enable the architectural design system to execute the method described in any one of claims 1-7.

9. A computer-readable storage medium comprising instructions, characterized in that: When the instructions are executed on an architectural design system, the architectural design system is caused to execute the method according to any one of claims 1 to 7.

10. A computer program product, characterized in that When the computer program product is run on an architectural design system, the architectural design system is enabled to execute the method according to any one of claims 1 to 7.

Images