Municipal green building design optimization method

Abstract

The application relates to the field of green building design, and specifically discloses a municipal green building design optimization method, which firstly extracts geographic location information from the municipal green building design condition information input by a user, and obtains a bird's-eye view of a target region from a GIS system according to the geographic location information; then further introduces a deep learning-based image processing technology to perform multi-scale feature extraction on the bird's-eye view of the target region, so as to capture the geographic features of the target building region; meanwhile, a natural language processing technology is used to perform semantic analysis on the information input by the user, such as budget limit, functional requirement, environmental requirement and time framework, and the information is used as a design constraint condition to modulate the attention of the geographic features of the target building region, so as to comprehensively consider the geographic features of the building region and the design constraint condition, and intelligently generate a municipal green building design sketch. The application can significantly improve the design efficiency of municipal green buildings, shorten the design cycle, and ensure the innovation and quality of the design results.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of green building design, and more specifically, to a municipal green building design optimization method. BACKGROUND

[0002] With the acceleration of urbanization, municipal buildings, as an important part of urban infrastructure, their design not only to meet the functional and aesthetic needs, but also in environmental protection and sustainability also put forward higher requirements. Municipal green building as an important part of sustainable development, its design optimization has become a research hotspot in the field of urban planning and construction. Green building aims to achieve the harmonious coexistence of man and nature by efficient use of resources, reducing environmental pollution and ecological destruction, and providing healthy and comfortable living and working environment.

[0003] However, the traditional municipal building design method often relies on the experience of designers and hand-drawing, this way not only low efficiency, and in the face of complex design requirements and diverse geographical environment, it is difficult to ensure the quality and innovation of the design. In addition, due to the time-consuming and laborious process of hand-drawing, it needs to go through multiple revisions from the initial concept to the final draft, which increases the project cost and time investment.

[0004] Therefore, an intelligent municipal green building design optimization method is needed. SUMMARY

[0005] In order to solve the above technical problems, the present application is proposed. The embodiments of the present application provide a municipal green building design optimization method, which first extracts the geographical location information from the user input municipal green building design condition information, and accordingly obtains the target area bird's eye view from the GIS system, then further introduces the image processing technology based on deep learning to extract the multi-scale features of the target area bird's eye view, in order to capture the geographical features of the target building area, at the same time, using natural language processing technology to analyze the semantic of the user input information such as budget limit, functional requirement, environmental requirement and time framework, and taking it as the design constraint condition to modulate the attention of the target building area geographical features, in order to comprehensively consider the building area geographical features and design constraint conditions, and intelligently generate the municipal green building design sketch. This can significantly improve the design efficiency of municipal green building, shorten the design cycle, and ensure the innovation and quality of the design results.

[0006] According to one aspect of the present application, a municipal green building design optimization method is provided, which comprises:

[0007] Obtaining the municipal green building design conditions input by the user, the municipal green building design conditions including geographical location, budget limit, functional requirement, environmental requirement and time framework;

[0008] extracting a geographic location from the municipal green building design condition, and extracting a target area bird's eye view from a GIS system based on the geographic location;

[0009] extracting geographic features from the target area bird's eye view to obtain a building area geographic feature multi-scale coding feature map;

[0010] jointly encoding semantics of the municipal green building design condition to obtain a design constraint condition semantic cascade coding vector;

[0011] performing cross-domain joint coding based on fine-grained ablation on the building area geographic feature multi-scale coding feature map and the design constraint condition semantic cascade coding vector to obtain a building area geographic feature-design constraint condition cross-modal ablation coding feature map;

[0012] generating a municipal green building design sketch based on the building area geographic feature-design constraint condition cross-modal ablation coding feature map.

[0013] Preferably, the geographic feature extraction of the target area bird's eye view to obtain the building area geographic feature multi-scale coding feature map comprises: inputting the target area bird's eye view into a building area geographic feature extractor based on a CrossViT model to obtain the building area geographic feature multi-scale coding feature map.

[0014] Preferably, the jointly encoding semantics of the municipal green building design condition to obtain the design constraint condition semantic cascade coding vector comprises: extracting the budget limit, the functional requirement, the environmental requirement and the time framework from the municipal green building design condition, and respectively encoding semantics of the budget limit, the functional requirement, the environmental requirement and the time framework to obtain a budget limit semantic coding vector, a functional requirement semantic coding vector, an environmental requirement semantic coding vector and a time framework semantic coding vector; concatenating the budget limit semantic coding vector, the functional requirement semantic coding vector, the environmental requirement semantic coding vector and the time framework semantic coding vector to obtain the design constraint condition semantic cascade coding vector.

[0015] Preferably, respectively encoding semantics of the budget limit, the functional requirement, the environmental requirement and the time framework to obtain a budget limit semantic coding vector, a functional requirement semantic coding vector, an environmental requirement semantic coding vector and a time framework semantic coding vector comprises: using a semantic encoder based on a Bert model to respectively encode semantics of the budget limit, the functional requirement, the environmental requirement and the time framework to obtain the budget limit semantic coding vector, the functional requirement semantic coding vector, the environmental requirement semantic coding vector and the time framework semantic coding vector.

[0016] Preferably, the cross-domain joint coding based on fine-grained ablation of the building area geographical feature multi-scale coding feature map and the design constraint condition semantic cascading coding vector is performed to obtain a building area geographical feature-design constraint condition cross-modal ablation coding feature map, including: performing feature fine-grained decoupling along the channel dimension on the building area geographical feature multi-scale coding feature map to obtain a set of building area geographical feature multi-scale coding feature matrices; performing correlation strength ablation measurement on the design constraint condition semantic cascading coding vector and each building area geographical feature multi-scale coding feature matrix in the set of building area geographical feature multi-scale coding feature matrices to obtain a set of building area geographical feature-design constraint condition fine-grained ablation factors; based on the set of building area geographical feature-design constraint condition fine-grained ablation factors, performing fine-grained ablation modulation aggregation on the set of building area geographical feature multi-scale coding feature matrices to obtain the building area geographical feature-design constraint condition cross-modal ablation coding feature map.

[0017] Preferably, the correlation strength ablation measurement on the design constraint condition semantic cascading coding vector and each building area geographical feature multi-scale coding feature matrix in the set of building area geographical feature multi-scale coding feature matrices to obtain a set of building area geographical feature-design constraint condition fine-grained ablation factors includes: performing cross-domain query interaction based on an attention mechanism on the design constraint condition semantic cascading coding vector and each building area geographical feature multi-scale coding feature matrix in the set of building area geographical feature multi-scale coding feature matrices to obtain a set of building area geographical feature-design constraint condition cross-domain query interaction feature vectors; inputting each building area geographical feature-design constraint condition cross-domain query interaction feature vector in the set of building area geographical feature-design constraint condition cross-domain query interaction feature vectors into an ablation measurement function respectively to obtain the set of building area geographical feature-design constraint condition fine-grained ablation factors.

[0018] Preferably, the cross-domain query interaction based on an attention mechanism on the design constraint condition semantic cascading coding vector and each building area geographical feature multi-scale coding feature matrix in the set of building area geographical feature multi-scale coding feature matrices to obtain a set of building area geographical feature-design constraint condition cross-domain query interaction feature vectors includes: performing linear transformation on the design constraint condition semantic cascading coding vector to obtain a query vector and a value vector; performing linear transformation on the building area geographical feature multi-scale coding feature matrix to obtain a key matrix; inputting the query vector, the value vector and the key matrix into a cross-domain interaction encoder based on an imitation transformer structure to obtain the building area geographical feature-design constraint condition cross-domain query interaction feature vector.

[0019] Preferably, based on the set of building area geographical feature-design constraint condition fine-grained ablation factors, the set of building area geographical feature multi-scale coding feature matrices is subjected to fine-grained ablation modulation aggregation to obtain the building area geographical feature-design constraint condition cross-modal ablation coding feature map, including: inputting the set of building area geographical feature-design constraint condition fine-grained ablation factors into an ablation effect coding module containing a normalization function and a mask function to obtain a set of building area geographical feature-design constraint condition fine-grained ablation weight factors; based on the set of building area geographical feature-design constraint condition fine-grained ablation weight factors, the set of building area geographical feature multi-scale coding feature matrices is subjected to weighted modulation to obtain a set of building area geographical feature-design constraint condition cross-modal ablation coding feature matrices; the set of building area geographical feature-design constraint condition cross-modal ablation coding feature matrices is subjected to feature aggregation along the channel dimension to obtain the building area geographical feature-design constraint condition cross-modal ablation coding feature map.

[0020] Preferably, based on the building area geographical feature-design constraint condition cross-modal ablation coding feature map, a municipal green building design sketch is generated, including: inputting the building area geographical feature-design constraint condition cross-modal ablation coding feature map into a municipal green building design sketch generator based on a large model to obtain the municipal green building design sketch.

[0021] The present application has at least the following technical effects:

[0022] Compared with the prior art, the municipal green building design optimization method provided by the present application first extracts geographical location information from the municipal green building design condition information input by the user, and then obtains the bird's eye view of the target area from the GIS system, and then further introduces the deep learning-based image processing technology to extract multi-scale features from the bird's eye view of the target area to capture the geographical features of the target building area. At the same time, the natural language processing technology is used to analyze the semantic information of the budget limit, functional requirement, environmental requirement and time framework input by the user, and use it as a design constraint condition to modulate the attention of the target building area geographical feature, so as to comprehensively consider the building area geographical feature and the design constraint condition, and intelligently generate a municipal green building design sketch. The present application can significantly improve the design efficiency of municipal green buildings, shorten the design cycle, and ensure the innovation and quality of the design results. BRIEF DESCRIPTION OF DRAWINGS

[0023] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description thereof taken in conjunction with the accompanying drawings in which:

[0024] Figure 1 Flow chart of the municipal green building design optimization method according to an embodiment of the present application.

[0025] Figure 2 Data flow diagram of the municipal green building design optimization method according to an embodiment of the present application.

[0026] Figure 3 Flow chart of sub-step S4 of the municipal green building design optimization method according to an embodiment of the present application.

[0027] Figure 4 Flow chart of sub-step S5 of the municipal green building design optimization method according to an embodiment of the present application.

[0028] Figure 5 Flow chart of sub-step S52 of the municipal green building design optimization method according to an embodiment of the present application. DETAILED DESCRIPTION

[0029] As used in the present application and claims, the indefinite articles "a", "an", and / or "the" are not intended to mean one and only one unless otherwise indicated by the context. Generally, the term "includes" or "including" is intended to mean "comprising" or "comprising." Thus, the method and apparatus of the present application can include a plurality of steps and / or elements, not necessarily depicted, that can include one another in another configuration.

[0030] Although the present application makes various references to certain modules in the system according to embodiments of the present application, however, any number of different modules can be used and run on the user terminal and / or server. The modules are illustrative only, and different aspects of the system and method can use different modules.

[0031] Flow charts are used in the present application to illustrate the operations performed by the system according to embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed in the exact order. Rather, various steps can be processed in reverse order, or at the same time, as desired. Other operations can also be added to, or removed from, these processes, or one or more steps can be removed from these processes.

[0032] Hereinafter, exemplary embodiments according to this application will be described in detail with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of this application, and not all embodiments of this application. It should be understood that this application is not limited to the exemplary embodiments described herein.

[0033] It is worth noting that all data acquisition actions in this application were carried out in compliance with the relevant data protection laws and policies of the country where the application is located, and with the authorization granted by the owner of the relevant device.

[0034] Specifically, Figure 1 This is a flowchart of a municipal green building design optimization method according to an embodiment of this application. Figure 2 This is a schematic diagram of the data flow in the municipal green building design optimization method according to an embodiment of this application. Figure 1 and Figure 2 As shown, the municipal green building design optimization method includes the following steps: S1, obtaining municipal green building design conditions input by the user, the municipal green building design conditions including geographical location, budget constraints, functional requirements, environmental requirements, and time frame; S2, extracting the geographical location from the municipal green building design conditions, and extracting a bird's-eye view of the target area from the GIS system based on the geographical location; S3, extracting geographical features from the bird's-eye view of the target area to obtain a multi-scale coded feature map of the building area's geographical features; S4, performing semantic joint coding on the municipal green building design conditions to obtain a semantic concatenated coding vector of design constraints; S5, performing cross-domain joint coding based on fine-grained ablation on the multi-scale coded feature map of the building area's geographical features and the semantic concatenated coding vector of design constraints to obtain a cross-modal ablation coding feature map of the building area's geographical features and design constraints; S6, generating a municipal green building design sketch based on the cross-modal ablation coding feature map of the building area's geographical features and design constraints.

[0035] In the above-mentioned municipal green building design optimization method, the step S1, the municipal green building design conditions input by the user are obtained, and the municipal green building design conditions include geographical location, budget limit, function demand, environmental requirement and time framework. It should be understood that the geographical location, budget limit, function demand, environmental requirement and time framework are the key elements of the municipal green building design, which jointly determine the direction and feasibility of the municipal green building design. Among them, the geographical location determines the natural environment and geographical conditions that the building must adapt to; the budget limit determines the economic feasibility of the design, which helps to reasonably plan the allocation of resources; the function demand defines the function layout and use efficiency of the building, which can guide the specific design contents such as space layout, streamline organization and equipment configuration of the building, such as the requirements of schools, hospitals or office buildings for internal facilities and services are different; the environmental requirement involves energy saving and emission reduction target, material selection standard, water resource management and other aspects; the time framework limits the time period of design and construction. By comprehensively obtaining the building design demand information of key aspects such as geographical location, budget limit, function demand, environmental requirement and time framework, the overall requirements and goals of municipal green building design can be more accurately grasped. Then reduce the design deviation caused by information missing or inaccurate, improve the fit degree of the design scheme and user's expectation.

[0036] Specifically, the geographical location is one of the most basic and important information in the municipal green building design. The natural and socio-economic conditions of different geographical regions differ significantly, which puts forward diversified requirements for building design. For example, buildings located in tropical regions need to consider sufficient ventilation and shading measures to cope with high temperature and humidity climate; while in cold regions, it is necessary to strengthen the heat preservation and insulation design to reduce energy consumption. In addition, geographical location also involves local cultural traditions and the availability of building materials and other factors, which have a profound impact on building design. When collecting geographical location information from users, it should be as specific as possible, including but not limited to the name of the city, specific blocks or plots, surrounding environmental features (such as adjacent rivers, parks, etc.), and even more detailed to the specific coordinate range of the plot. Such information helps to fully understand the actual situation of the project site, thereby providing a solid foundation for the next step of work. At the same time, considering the importance of geological conditions to the safety of building structure, it may also be necessary to conduct geological exploration to assess factors such as foundation bearing capacity and underground water level, to ensure the safety and stability of the building.

[0037] Budget constraints are an important consideration for municipal green building design. Reasonable budget planning not only ensures the smooth implementation of the project, but also improves the efficiency of capital use. Therefore, when collecting user input, it is crucial to understand their investment willingness and ability for the entire project and quantify it as a specific amount interval. In addition to the total budget, it is also necessary to subdivide various costs, such as land acquisition, planning and design, construction, equipment procurement, and operation and maintenance, in order to better control costs. In addition, budget constraints play a decisive role in the selection and use of certain technical applications or materials during the architectural design process, such as high-performance energy-saving equipment and environmentally friendly building materials. By obtaining user budget constraint information, we can more targetedly design the scheme to ensure that the functional requirements are met while not exceeding the budget range.

[0038] Functional requirements reflect the basic expectations of users for the use of municipal green buildings, covering space layout, facility equipment, and other aspects. When collecting this information, first, we need to clarify the main service objects of the building and their activity patterns. For example, an office building may focus on the flexibility of office space and intelligent management; a school may focus more on the setting of classrooms, laboratories, and other functional areas, as well as safety performance and other requirements. Second, we need to understand the actual problems faced by users in their daily lives in order to solve them through innovative design concepts, such as creating barrier-free access, setting up mother and baby rooms, and other humanized facilities. In addition to the above-mentioned regular considerations, modern municipal green buildings also emphasize the interaction between people and nature, advocating the creation of open public spaces to promote community exchange and vitality. This means that during the design process, we need to reserve enough green space, outdoor leisure places, and other facilities so that residents can find a quiet and comfortable corner in the busy urban life. In addition, with the development of technology, smart building systems have become part of the functional requirements, such as smart home, smart security, energy management systems, etc. Smart building systems can improve the convenience and safety of buildings, while also meeting the energy-saving and emission-reducing goals of green buildings.

[0039] Environmental requirements reflect the responsibility of municipal green buildings for environmental protection and ecological restoration. With the intensification of global climate change and the frequent occurrence of extreme weather events, how to improve the resistance of buildings and reduce the impact on the external environment has become a pressing problem. Therefore, users need to convey their expectations for energy saving and emission reduction, water resource management, waste disposal, etc. to explore the best practice solution that can meet the current development needs and long-term interests. For example, solar photovoltaic panels can be installed to increase the proportion of renewable energy; rainwater collection systems can be used for irrigation of green plants; or low-carbon emission construction processes and technologies can be used to reduce greenhouse gas emissions during the construction process. In addition, protecting local biodiversity is also an important aspect that cannot be ignored. It is necessary to avoid damaging the original ecological environment as much as possible and actively create small habitats that are conducive to the growth and reproduction of animals and plants. On this basis, the impact of buildings on the surrounding microclimate can also be considered, such as improving local climate conditions through vegetation cover, permeable pavement, etc. to achieve the harmonious coexistence of buildings and environment.

[0040] Finally, the time frame, that is, the construction period of the project. A clear time schedule helps coordinate work progress among all parties and ensures that each link is closely connected to complete the task on time and with high quality. Generally, the time frame will vary depending on the size and complexity of the project, ranging from a few months to several years. Therefore, when collecting user input, it is necessary to confirm the key time nodes of each stage, such as the completion deadline of planning and design, the date of construction permit approval, and the time of main structure topping-off, etc. At the same time, it should be noted that due to the involvement of many uncertain factors in the design of municipal green buildings, such as weather delays and cost overruns caused by rising raw material prices, a certain degree of flexibility should be allowed when making plans to adjust arrangements as appropriate to cope with possible risks and challenges.

[0041] In the above-mentioned municipal green building design optimization method, step S2 extracts the geographic location from the municipal green building design conditions, and extracts the target area bird's-eye view from the GIS system based on the geographic location. It should be understood that the present application considers that the geographic location is one of the important basic information of municipal green building design, which determines the natural environment, surrounding traffic network, infrastructure distribution and relative position relationship with other functional areas of the city and other factors, which has a crucial influence on the layout, orientation, appearance design and integration with the surrounding environment of the building. Based on this, in order to accurately obtain the visual geographic information of the target building area, the present application further utilizes the powerful spatial data management and query function of geographic information system (GIS), according to the geographic position information (such as latitude and longitude coordinates or specific address name) provided by the user, quickly locates and extracts the corresponding target area bird's-eye view in its database, so as to intuitively understand the actual situation of the building site, including the surrounding topography, existing building distribution, road direction, etc. It should be understood that the GIS system can present the geographic information in the real world in the form of maps, images, etc. through digital storage and processing of geographic spatial data, providing intuitive geographic spatial reference for subsequent design work, which helps to fully consider the actual situation of the site at the initial design stage, optimize the layout and form design of the building, improve the coordination between the building and the surrounding environment, and also provides original image data basis for subsequent design sketch generation.

[0042] Specifically, first of all, it is necessary to ensure that the geographic location information is accurately extracted from the municipal green building design conditions provided by the user. This includes but is not limited to the name of the city, the specific block or plot, the surrounding environmental features (such as rivers, parks, etc.), and even more detailed to the specific coordinate range of the plot. In order to ensure the accuracy of the information, all the files and materials submitted by the user can be reviewed carefully to filter out the information related to the geographic location; at the same time, face-to-face or remote meetings can be held with the user to discuss the details of the geographic location in depth, clarify any ambiguities, and ensure a clear understanding of the project site. Then, online map services (such as Google Maps, Bing Maps) or specialized geographic information system software (such as ArcGIS Online) can be used to quickly locate the specific location according to the address or description provided by the user, and mark it out for further analysis. Considering the importance of geological conditions to the safety of building structures, in some cases, professional geological survey may also be needed to assess factors such as foundation bearing capacity and underground water level to ensure the safety and stability of the building. For some special projects, more background information may also need to be collected, such as historical relic protection range, natural protection zone boundary, etc., to avoid potential legal risks.

[0043] After determining the geographical location of the project, the next step is to use a Geographic Information System (GIS) to extract an aerial view of the area. GIS is a powerful tool that integrates spatial data from different sources and provides a visual interface for users to intuitively view and analyze this information. It is crucial to choose the appropriate GIS platform, as there are many mature GIS platforms available on the market, such as Esri's ArcGIS, the open-source QGIS platform, and others. Choose the most suitable tool based on project requirements and personal preferences. Most GIS platforms come with commonly used basic map layers, such as satellite imagery, street maps, and others. If the default map layers of the platform cannot meet the accuracy requirements, you can access third-party API interfaces (such as Google EarthEngine, Microsoft Bing Maps API) to obtain higher-resolution image data. You can also contact relevant agencies to apply for high-precision DEM (Digital Elevation Model), DOM (Digital Orthophoto Map), and other data sets as supplements.

[0044] Based on the previously extracted geographical location information, a reasonable rectangle or polygon boundary is drawn on the GIS platform to define the scope of the target area. This boundary should cover all possible factors that may affect architectural design as much as possible, while not being too broad to increase unnecessary computational burden. Overlay various thematic layers on the base map, such as land use types, vegetation distribution, transportation networks, public facility layout, and others. In addition, buildings located near transportation hubs need to be particularly considered for human flow evacuation problems. Modern GIS technology not only presents two-dimensional plane images, but also constructs realistic three-dimensional scenes. By processing DEM data, the terrain undulations of the target area can be obtained; combined with building contour lines and height information, a complete three-dimensional model of the city can be created, making the computer more intuitively feel the spatial relationship of the building area. The final step is to export the processed data as a high-quality aerial view that meets the design requirements. Most GIS platforms support multiple formats for output, such as JPEG, PNG, TIFF, and others. Adjust the image resolution, color mode, and other parameters according to actual needs to ensure that the final result is both beautiful and practical. During the export process, attention should be paid to protecting personal privacy and sensitive information to avoid unnecessary disclosure.

[0045] In the above-mentioned municipal green building design optimization method, the step S3 is to perform geographic feature extraction on the target area bird's-eye view to obtain a building area geographic feature multi-scale coding feature map. In a specific example of the present application, the step S3 includes: inputting the target area bird's-eye view into a building area geographic feature extractor based on a CrossViT model to obtain the building area geographic feature multi-scale coding feature map. Specifically, considering that the target area bird's-eye view contains rich geographic information of different scales, such as overall topography, building distribution, local vegetation coverage, etc. Therefore, in order to comprehensively capture the geographic features of the target building area, the present application adopts a CrossViT model to perform multi-scale feature extraction on the target area bird's-eye view. The CrossViT model is a cross-attention multi-scale visual Transformer model for image feature extraction, which can effectively process multi-scale visual information in images. Specifically, the CrossViT model is based on a double-branch Transformer architecture, which processes the input image through parallel branches of different scales. In each branch, the target area bird's-eye view is first divided into a series of image blocks, which are converted into feature sequences after linear embedding and input into a multi-layer Transformer encoder. The Transformer encoder models the semantic association between image block features through self-attention mechanisms, thereby capturing feature information of the target area bird's-eye view at different scales. Finally, through a cross-scale feature fusion module, the geographic features extracted by different scale branches are fused to obtain a building area geographic feature multi-scale coding feature map containing multi-scale geographic feature information. In this way, rich geographic feature information from macro topography to micro building texture can be effectively obtained, so as to better understand the site characteristics and integrate them into the building design, realizing the organic combination of green buildings and geographic environment.

[0046] In the above-mentioned municipal green building design optimization method, the step S4 is to perform semantic joint coding on the municipal green building design conditions to obtain a design constraint condition semantic cascade coding vector. Wherein, Figure 3 The flowchart of the sub-step S4 of the municipal green building design optimization method according to the embodiment of the present application. As Figure 3As shown, the step S4 includes steps of: S41, extracting the budget limit, the function requirement, the environment requirement and the time framework from the municipal green building design conditions, and respectively performing semantic coding on the budget limit, the function requirement, the environment requirement and the time framework to obtain a budget limit semantic coding vector, a function requirement semantic coding vector, an environment requirement semantic coding vector and a time framework semantic coding vector; S42, cascading the budget limit semantic coding vector, the function requirement semantic coding vector, the environment requirement semantic coding vector and the time framework semantic coding vector to obtain the design constraint condition semantic cascading coding vector.

[0047] Specifically, in one specific example of the present application, the step S41 includes using a semantic encoder based on a Bert model to respectively perform semantic coding on the budget limit, the function requirement, the environment requirement and the time framework to obtain the budget limit semantic coding vector, the function requirement semantic coding vector, the environment requirement semantic coding vector and the time framework semantic coding vector. It should be understood that, since the budget limit, the function requirement, the environment requirement and the time framework and the like information are unstructured data input in the form of text, it is difficult for the computer to directly understand their internal meanings and mutual relationships. Therefore, the present application further performs semantic coding on the budget limit, the function requirement, the environment requirement and the time framework to convert the text information into a vector form that can be processed by the computer, so as to adjust and optimize the design scheme based on the quantified constraint conditions in the subsequent design process, and ensure that the design scheme meets the user's function and environment requirements, and is realized within the budget and time limit. It should be known by those skilled in the art that the Bert model uses its bidirectional Transformer architecture to deeply understand the context semantics of the text, captures the complex relationships between words through multi-layer neural networks, and generates high-dimensional vector representations rich in information. In the coding task of the municipal green building design conditions, the Bert model is used to perform semantic feature extraction and vectorization processing on the unstructured text descriptions of the budget limit, the function requirement, the environment requirement and the time framework input by the user, respectively, to ensure that each design condition is converted into a computable vector form and retains the core meaning of the original text, thereby providing an accurate and rich semantic basis for subsequent analysis.

[0048] Specifically, the step S42 cascades the budget limit semantic coding vector, the function requirement semantic coding vector, the environment requirement semantic coding vector and the time framework semantic coding vector to obtain the design constraint condition semantic cascaded coding vector. It can be understood that, since each design constraint condition (budget limit, function requirement, environment requirement and time framework) is an integral whole that is interrelated and interdependent in the actual architectural design process. Therefore, the present application further cascades and fuses the budget limit semantic coding vector, the function requirement semantic coding vector, the environment requirement semantic coding vector and the time framework semantic coding vector to integrate the scattered constraint information and generate the design constraint condition semantic cascaded coding vector, so that the subsequent design process can comprehensively consider all design constraint conditions and the interaction between each design constraint condition, so as to better balance the requirements and restrictions in all aspects and realize the comprehensive optimization design of the municipal green building.

[0049] In the above-mentioned municipal green building design optimization method, the step S5 performs cross-domain joint coding based on fine-grained ablation on the building area geographic feature multi-scale coding feature map and the design constraint condition semantic cascaded coding vector to obtain a building area geographic feature-design constraint condition cross-modal ablation coding feature map. In particular, considering that the geographic features of the building area and the design constraint conditions come from different modalities (images and texts), both of which jointly affect the final design result of the building, but directly combining information from different modalities may be difficult to fully explore the deep-level correlation and interaction therebetween. Based on this, the present application proposes a cross-domain joint coding method based on fine-grained ablation, which performs fine-grained ablation analysis on the geographic features of the building area and the design constraint conditions to reveal the internal relationship and interaction therebetween, thereby focusing on the key geographic feature area that has a close relationship with the design constraint conditions, obtaining a building area geographic feature-design constraint condition cross-modal ablation coding feature map, so as to provide more accurate and useful information for subsequent design sketch generation, so that it can make full use of the geographical advantages of the site and meet the requirements of various design constraint conditions. Wherein, Figure 4 The flowchart of the sub-step S5 of the municipal green building design optimization method according to the embodiment of the present application. As shown in FIG. 4, the step S5 performs cross-domain joint coding based on fine-grained ablation on the building area geographic feature multi-scale coding feature map and the design constraint condition semantic cascaded coding vector to obtain a building area geographic feature-design constraint condition cross-modal ablation coding feature map. Figure 4As shown, step S5 includes the following steps: S51, performing fine-grained feature decoupling along the channel dimension on the multi-scale coding feature map of the building area geographic features to obtain a set of multi-scale coding feature matrices of the building area geographic features; S52, performing association strength ablation measurement on each multi-scale coding feature matrix of the building area geographic features in the set of design constraint semantic concatenated coding vectors and multi-scale coding feature matrices of the building area geographic features to obtain a set of fine-grained ablation factors of the building area geographic features-design constraints; S53, performing fine-grained ablation modulation aggregation on the set of multi-scale coding feature matrices of the building area geographic features based on the set of fine-grained ablation factors of the building area geographic features-design constraints to obtain a cross-modal ablation coding feature map of the building area geographic features-design constraints.

[0050] Specifically, step S51 is expressed by the formula as follows:

[0051]

[0052] in, This indicates feature decoupling. A multi-scale encoded feature map representing the geographical features of a building area. , and The first, second, and third elements in the set of multi-scale encoded feature matrices representing the geographic features of a building area are respectively represented. Multi-scale encoded feature matrix of geographical features of a building area The number of channels for the multi-scale encoded feature map of the geographical features of the building area.

[0053] Specifically, by decoupling the multi-scale encoded feature map of the building area's geographic features along the channel dimension, a finer-grained representation of geographic features is obtained, providing a more detailed feature basis for subsequent ablation analysis.

[0054] Specifically, in step S52, the association strength ablation metric is performed on each of the multi-scale coding feature matrices of the building area geographic features in the set of the design constraint semantic concatenation coding vector and the multi-scale coding feature matrix of the building area geographic features to obtain a set of fine-grained ablation factors of building area geographic features-design constraints. Figure 5 This is a flowchart of sub-step S52 of the municipal green building design optimization method according to an embodiment of this application. Figure 5As shown, the step S52 includes steps of: S521, performing attention mechanism based cross-domain query interaction on each building area geographic feature multi-scale coding feature matrix in the set of the design constraint condition semantic cascade coding vector and the building area geographic feature multi-scale coding feature matrix to obtain a set of building area geographic feature-design constraint condition cross-domain query interaction feature vectors; S522, inputting each building area geographic feature-design constraint condition cross-domain query interaction feature vector in the set of the building area geographic feature-design constraint condition cross-domain query interaction feature vector into an ablation measure function respectively to obtain a set of building area geographic feature-design constraint condition fine-grained ablation factors.

[0055] More specifically, the step S521 includes: first, performing linear transformation on the design constraint condition semantic cascade coding vector to obtain a query vector and a value vector, which is expressed by a formula as:

[0056]

[0057]

[0058] wherein, represents the design constraint condition semantic cascade coding vector, and represent a query embedding matrix and a value embedding matrix respectively, and represent a query embedding bias term and a value embedding bias term respectively, represents a matrix multiplication operation, and represent the key matrix corresponding to the query vector and the value vector respectively.

[0059] Then, performing linear transformation on the building area geographic feature multi-scale coding feature matrix to obtain a key matrix, which is expressed by a formula as:

[0060]

[0061] wherein, represents the building area geographic feature multi-scale coding feature matrix, represents a key embedding matrix, represents a key embedding bias term, represents the key matrix corresponding to the building area geographic feature multi-scale coding feature matrix. Finally, inputting the query vector, the value vector and the key matrix into a cross-domain interaction encoder based on an imitation converter structure to obtain the building area geographic feature-design constraint condition cross-domain query interaction feature vector, which is expressed by a formula as:

[0062]

[0063] ​

[0064] wherein, is a normalized exponential function, denotes the transpose of a vector, denotes the and the between the building area geographical feature-design constraint condition cross-domain query interaction feature vector.

[0065] Specifically, based on the design constraint condition semantic cascade coding vector, a query vector and a value vector are constructed, based on each building area geographical feature multi-scale coding feature matrix in the set of building area geographical feature multi-scale coding feature matrix, a key matrix is constructed, and the correlation between the building area geographical feature and the design constraint condition is captured through the cross-domain query interaction based on the converter structure. In this process, the converter structure can effectively process the information interaction between different domains, strengthen the significant correlation features between the building area geographical feature and the design constraint condition through the attention mechanism, and suppress irrelevant information interference, so as to obtain the correlation interaction feature representation between the building area geographical feature and the design constraint condition.

[0066] More specifically, the step S522 is expressed by the formula as follows:

[0067]

[0068] wherein, denotes an ablation measure function, denotes a maximum value function, and denote the feature mean and the feature variance of the , respectively, denotes a regularization term, denotes the corresponding building area geographical feature-design constraint condition fine-grained ablation factor.

[0069] Specifically, the present application further introduces an ablation metric function to evaluate and quantify the interaction features between the building area geographical features and the design constraints. The ablation metric function is similar to the ablation analysis in experimental design, which is used to determine the impact on the final interaction effect after removing a certain specific relationship, and helps to identify the degree of influence of design constraint information on building area geographical features. For example, if a certain vegetation coverage area is highly related to environmental requirements, special consideration should be given to the protection and use of vegetation in this area during the design process to ensure that the design meets the requirements of environmental sustainability. Conversely, if the geographical features of a certain area are not closely related to the design constraints, more flexibility can be given during the design to accommodate other more important constraints. Through this fine-grained ablation analysis, geographical advantages of the site can be more accurately identified and utilized, while ensuring that the building design scheme meets all design constraints while achieving environmental sustainability and economic rationality.

[0070] Specifically, the step S53 includes: first, inputting the set of building area geographical feature-design constraint fine-grained ablation factors into the ablation effect encoding module containing the normalization function and the mask function to obtain the set of building area geographical feature-design constraint fine-grained ablation weight factors, which is expressed by the formula:

[0071]

[0072] wherein, represents the exponential function with e as the base, represents the set of building area geographical feature-design constraint fine-grained ablation factors, the corresponding normalized building area geographical feature-design constraint fine-grained ablation factor, is a mask function, is a gating mask threshold, is the set of building area geographical feature-design constraint fine-grained ablation factors, the corresponding building area geographical feature-design constraint fine-grained ablation weight factor.

[0073] Specifically, in order to standardize the weights and exclude irrelevant items, the present application further normalizes and masks the generated set of building area geographical feature-design constraint fine-grained ablation factors. Here, the normalization function is used to convert all ablation factors to a standard range (such as between 0 and 1) for direct comparison with each other; while the mask function is used to increase the ablation factors of building area geographical features that have significant association with design constraints, while reducing the ablation factors of building area geographical features that are not related or have low correlation with design constraints, thereby focusing resources and attention on the geographical features that are most closely related to design constraints.

[0074] Then, based on the set of building region geographical feature-design constraint condition fine-grained ablation weight factors, the set of building region geographical feature multi-scale coding feature matrices is weighted and modulated to obtain a set of building region geographical feature-design constraint condition cross-modal ablation coding feature matrices; finally, the set of building region geographical feature-design constraint condition cross-modal ablation coding feature matrices is aggregated along the channel dimension to obtain a building region geographical feature-design constraint condition cross-modal ablation coding feature map, which is expressed by a formula as:

[0075]

[0076] wherein, and are the corresponding building region geographical feature-design constraint condition fine-grained ablation weight factors of the , the and the , indicates a building region geographical feature-design constraint condition cross-modal ablation coding feature map.

[0077] Specifically, based on the generated set of building region geographical feature-design constraint condition fine-grained ablation weight factors, the original set of building region geographical feature multi-scale coding feature matrices is fine-grained ablation modulated and aggregated along the channel dimension, so as to strengthen the building region geographical features highly related to the design constraint conditions, while also retaining the diversity of the building region geographical features, thereby providing a more rich and accurate feature representation for subsequent design sketch generation.

[0078] ​In the above-mentioned municipal green building design optimization method, the step S6 generates a municipal green building design sketch based on the building area geographical feature-design constraint condition cross-modal ablation encoding feature map. In a specific example of the present application, the step S6 includes inputting the building area geographical feature-design constraint condition cross-modal ablation encoding feature map into a large model-based municipal green building design sketch generator to obtain the municipal green building design sketch. Specifically, in order to convert the building area geographical feature-design constraint condition cross-modal ablation encoding feature map from an abstract feature representation to an intuitive building design sketch, so that designers and users can quickly understand and evaluate the general outline and layout of the design scheme, the present application further utilizes the powerful generation capability of the large model to generate a high-quality municipal green building design sketch that meets the geographical features and design constraint conditions according to the rich information contained in the building area geographical feature-design constraint condition cross-modal ablation encoding feature map, combined with its learning experience on large-scale building design data, to provide a visual basis and reference for subsequent detailed design and scheme optimization. In an embodiment of the present application, the large model is a generative adversarial network (GANs). As known to those skilled in the art, a generative adversarial network (GANs) consists of a generator and a discriminator, which co-evolve through adversarial training. In the generation of municipal green building design sketches, the generative adversarial network converts the building area geographical feature-design constraint condition cross-modal ablation encoding feature map into a high-quality design sketch through the adversarial training mechanism of its generator and discriminator. The generator learns to create sketches that meet the design requirements, while the discriminator evaluates the similarity of these generated sketches to real samples, both competing and optimizing each other, and ultimately generating high-quality design sketches that meet various constraint conditions, significantly improving design efficiency.

[0079] Specifically, in the case where the building area geographical feature multi-scale encoding feature map and the design constraint condition semantic cascade encoding vector respectively represent the building area geographical image multi-scale semantic encoding features of the target area bird's eye view and the semantic cascade features of the design constraint condition, after joint encoding based on fine-grained ablation across domains, the building area geographical feature-design constraint condition cross-modal ablation encoding feature map also has significant interactive optimization distribution space structure differences due to cross-modal semantic fine-grained ablation differences, affecting the convergence consistency of the large model-based municipal green building design sketch generator, thereby affecting the image quality of the municipal green building design sketch obtained by the large model-based municipal green building design sketch generator.

[0080] Therefore, in a preferred example of the present application, the building area geographical feature-design constraint condition cross-modal ablation encoding feature map is optimized, and the optimization process includes:

[0081] First, the sum of absolute values of all feature values of the building area geographical feature-design constraint condition cross-modal ablation encoding feature map is calculated, and the square root of the sum of squares is obtained to obtain a first building area geographical feature-design constraint condition cross-modal ablation encoding space structure value and a second building area geographical feature-design constraint condition cross-modal ablation encoding space structure value, which is expressed by a formula as follows:

[0082]

[0083]

[0084] wherein, denotes the first building area geographical feature-design constraint condition cross-modal ablation encoding space structure value, denotes the i-th feature value in the set of all feature values of the building area geographical feature-design constraint condition cross-modal ablation encoding feature map, denotes the second building area geographical feature-design constraint condition cross-modal ablation encoding space structure value.

[0085] Secondly, the total number of feature values of all feature values of the building area geographical feature-design constraint condition cross-modal ablation encoding feature map is determined, and for each feature value of the building area geographical feature-design constraint condition cross-modal ablation encoding feature map, the first building area geographical feature-design constraint condition cross-modal ablation encoding space structure value is calculated minus the product of the feature value and the total number of feature values to obtain a first building area geographical feature-design constraint condition cross-modal ablation encoding long-range dependence value, which is expressed by a formula as follows:

[0086]

[0087] wherein, denotes the corresponding first building area geographical feature-design constraint condition cross-modal ablation encoding long-range dependence value, denotes the total number of feature values of all feature values of the building area geographical feature-design constraint condition cross-modal ablation encoding feature map.

[0088] Then, the product of the square root of the total number of feature values and the feature value is calculated minus the second building area geographical feature-design constraint condition cross-modal ablation encoding space structure value to obtain a second building area geographical feature-design constraint condition cross-modal ablation encoding long-range dependence value, which is expressed by a formula as follows:

[0089]

[0090] wherein, denotes the corresponding first building area geographical feature-design constraint condition cross-modal ablation encoding long-range dependence value, ​​The corresponding second building area geographical feature-design constraint condition cross-modal ablation coding long-range dependency value;

[0091] Then, the first building area geographical feature-design constraint condition cross-modal ablation coding long-range dependency value is calculated as an exponential value of a natural constant, and the reciprocal of the second building area geographical feature-design constraint condition cross-modal ablation coding long-range dependency value is weighted and summed to obtain an optimized feature value corresponding to each feature value, which is expressed by the formula:

[0092]

[0093] Among them, The first building area geographical feature-design constraint condition cross-modal ablation coding long-range dependency value, The corresponding optimized feature value, , Respectively represent different weight parameters, Indicates the reciprocal, Indicates the exponential value with the natural constant as the base;

[0094] Finally, the optimized feature values form an optimized building area geographical feature-design constraint condition cross-modal ablation coding feature map.

[0095] Specifically, the lack of spatial structure of the feature set of the building area geographical feature-design constraint condition cross-modal ablation coding feature map in the high-dimensional space leads to inconsistent convergence of the generator based on the implicit inference of spatial structure information by the feature-based weight, and the present application establishes a long-distance feature dependency relationship based on the overall feature scale of the building area geographical feature-design constraint condition cross-modal ablation coding feature map relative to the spatial structure representation of the building area geographical feature-design constraint condition cross-modal ablation coding feature map, to establish the local connectivity of the features of the building area geographical feature-design constraint condition cross-modal ablation coding feature map, and to capture the spatial ambiguity information of the object feature value through the unstructured feature value point prediction of the building area geographical feature-design constraint condition cross-modal ablation coding feature map, thereby improving the spatial inductive bias perception ability of the feature set of the building area geographical feature-design constraint condition cross-modal ablation coding feature map, improving the convergence consistency of the generator, and improving the image quality of the municipal green building design sketch generated by the building area geographical feature-design constraint condition cross-modal ablation coding feature map through the municipal green building design sketch generator based on a large model.

[0096] In summary, the municipal green building design optimization method based on the embodiments of the present application is illustrated, which firstly extracts geographic location information from the municipal green building design condition information input by the user, and obtains the bird's eye view of the target area from the GIS system accordingly, then further introduces the deep learning-based image processing technology to perform multi-scale feature extraction on the bird's eye view of the target area, so as to capture the geographic features of the target building area, at the same time, the natural language processing technology is used to perform semantic analysis on the information such as budget limit, functional requirement, environmental requirement and time framework input by the user, and the geographic features of the target building area are modulated by attention based on the design constraint conditions, so as to comprehensively consider the geographic features of the building area and the design constraint conditions, and intelligently generate the municipal green building design sketch. In this way, the design efficiency of the municipal green building can be significantly improved, the design cycle can be shortened, and the innovation and quality of the design results can be ensured.

[0097] The basic principles of the present application are described above in combination with specific embodiments, but it should be noted that the advantages, advantages, effects and the like mentioned in the present application are only examples and not limitations, and these advantages, advantages, effects and the like cannot be considered as necessary for each embodiment of the present application. In addition, the specific details of the above embodiments are only for the purpose of example and for the purpose of understanding, and the above details do not limit the present application to the above specific details.

[0098] In the above embodiments, the description of each embodiment has its own emphasis, and the parts not described or recorded in a certain embodiment can be referred to the related description of other embodiments. In the several embodiments provided by the present application, it should be understood that the disclosed system and method can be implemented in other ways. For example, the system embodiments described above are only schematic, for example, the unit division is only a logical function division, and there can be other division ways in actual implementation. The units described as separated components can or can not be physically separated, and the components displayed as units can be or can not be physical units, that is, they can be located in one place or distributed on a network. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiments.

[0099] It is obvious for those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and the present application can be implemented in other specific forms without departing from the spirit or essential characteristics of the present application. Therefore, the embodiments should be considered as exemplary and non-limiting, and the scope of the present application is defined by the appended claims rather than the above description, and all changes falling within the meaning and scope of the equivalent elements of the claims are intended to be included in the present application. Any reference signs in the claims should not be considered as limiting the claims.

[0100] Finally, it should be noted that the above description has been given for the purpose of illustration and description. Furthermore, the above examples are only used to illustrate the technical solutions of the present application and not to limit the present application. Although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present application can be modified or equivalently replaced without departing from the spirit and scope of the present application.

Claims

1. A municipal green building design optimization method, characterized by, The method comprises the following steps: obtaining municipal green building design conditions input by a user, the municipal green building design conditions comprising a geographic location, a budget limit, a functional requirement, an environmental requirement, and a time frame; extracting the geographic location from the municipal green building design conditions and extracting a target area bird's-eye view from a GIS system based on the geographic location; performing geographic feature extraction on the target area bird's-eye view to obtain a building area geographic feature multi-scale coding feature map; performing semantic joint coding on the municipal green building design conditions to obtain a design constraint condition semantic cascade coding vector; performing fine-grained ablation-based cross-domain joint coding on the building area geographic feature multi-scale coding feature map and the design constraint condition semantic cascade coding vector to obtain a building area geographic feature-design constraint condition cross-modal ablation coding feature map; generating a municipal green building design sketch based on the building area geographic feature-design constraint condition cross-modal ablation coding feature map; optimizing the building area geographic feature-design constraint condition cross-modal ablation coding feature map, the optimization process comprising: calculating the sum of absolute values and the square root of the sum of squares of all feature values of the building area geographic feature-design constraint condition cross-modal ablation coding feature map to obtain a first building area geographic feature-design constraint condition cross-modal ablation coding space structure value and a second building area geographic feature-design constraint condition cross-modal ablation coding space structure value; determining the total number of feature values of all feature values of the building area geographic feature-design constraint condition cross-modal ablation coding feature map, and for each feature value of the building area geographic feature-design constraint condition cross-modal ablation coding feature map, calculating the first building area geographic feature-design constraint condition cross-modal ablation coding space structure value minus the product of the feature value and the total number of feature values to obtain a first building area geographic feature-design constraint condition cross-modal ablation coding long-range dependence value; calculating the square root of the total number of feature values multiplied by the product of the feature values minus the second building area geographic feature-design constraint condition cross-modal ablation coding space structure value to obtain a second building area geographic feature-design constraint condition cross-modal ablation coding long-range dependence value; performing weighted summation on the first building area geographic feature-design constraint condition cross-modal ablation coding long-range dependence value as an exponential value calculated by taking the exponential of a natural constant and the reciprocal of the second building area geographic feature-design constraint condition cross-modal ablation coding long-range dependence value to obtain an optimized feature value corresponding to each feature value; composing the optimized feature values into an optimized building area geographic feature-design constraint condition cross-modal ablation coding feature map.

2. The municipal green building design optimization method of claim 1, wherein, performing geographic feature extraction on the target area bird's-eye view to obtain a building area geographic feature multi-scale coding feature map, comprising: inputting the target area bird's-eye view into a building area geographic feature extractor based on a CrossViT model to obtain the building area geographic feature multi-scale coding feature map.

3. The municipal green building design optimization method of claim 2, wherein, performing semantic joint coding on the municipal green building design conditions to obtain a design constraint condition semantic cascade coding vector, comprising: extracting the budget limit, the function requirement, the environment requirement and the time framework from the municipal green building design condition, and respectively performing semantic coding on the budget limit, the function requirement, the environment requirement and the time framework to obtain a budget limit semantic coding vector, a function requirement semantic coding vector, an environment requirement semantic coding vector and a time framework semantic coding vector; concatenating the budget limit semantic coding vector, the function requirement semantic coding vector, the environment requirement semantic coding vector and the time framework semantic coding vector to obtain the design constraint condition semantic concatenated coding vector.

4. The municipal green building design optimization method of claim 3, wherein, respectively performing semantic coding on the budget limit, the function requirement, the environment requirement and the time framework to obtain a budget limit semantic coding vector, a function requirement semantic coding vector, an environment requirement semantic coding vector and a time framework semantic coding vector, comprising: respectively performing semantic coding on the budget limit, the function requirement, the environment requirement and the time framework to obtain a budget limit semantic coding vector, a function requirement semantic coding vector, an environment requirement semantic coding vector and a time framework semantic coding vector using a semantic encoder based on a Bert model.

5. The municipal green building design optimization method of claim 4, wherein, performing fine-grained ablation-based cross-domain joint coding on the building area geographic feature multi-scale coding feature graph and the design constraint condition semantic concatenated coding vector to obtain a building area geographic feature-design constraint condition cross-modal ablation coding feature graph, comprising: performing feature fine-grained decoupling along the channel dimension on the building area geographic feature multi-scale coding feature graph to obtain a set of building area geographic feature multi-scale coding feature matrices; performing correlation strength ablation measurement on the design constraint condition semantic concatenated coding vector and each building area geographic feature multi-scale coding feature matrix in the set of building area geographic feature multi-scale coding feature matrices to obtain a set of building area geographic feature-design constraint condition fine-grained ablation factors; based on the set of building area geographic feature-design constraint condition fine-grained ablation factors, performing fine-grained ablation modulation aggregation on the set of building area geographic feature multi-scale coding feature matrices to obtain the building area geographic feature-design constraint condition cross-modal ablation coding feature graph.

6. The municipal green building design optimization method of claim 5, wherein, performing correlation strength ablation measurement on the design constraint condition semantic concatenated coding vector and each building area geographic feature multi-scale coding feature matrix in the set of building area geographic feature multi-scale coding feature matrices to obtain a set of building area geographic feature-design constraint condition fine-grained ablation factors, comprising: performing attention mechanism-based cross-domain query interaction on the design constraint condition semantic concatenated coding vector and each building area geographic feature multi-scale coding feature matrix in the set of building area geographic feature multi-scale coding feature matrices to obtain a set of building area geographic feature-design constraint condition cross-domain query interaction feature vectors; Input each of the building region geographic feature-design constraint cross-domain query interaction feature vectors in the set of the building region geographic feature-design constraint cross-domain query interaction feature vectors into the ablation measure function respectively to obtain the set of the building region geographic feature-design constraint fine-grained ablation factors.

7. The municipal green building design optimization method of claim 6, wherein, Perform attention mechanism-based cross-domain query interaction on the design constraint condition semantic cascade coding vector and each of the building region geographic feature multi-scale coding feature matrices in the set of the building region geographic feature multi-scale coding feature matrices to obtain a set of building region geographic feature-design constraint cross-domain query interaction feature vectors, including: Perform linear transformation on the design constraint condition semantic cascade coding vector to obtain a query vector and a value vector; Perform linear transformation on the building region geographic feature multi-scale coding feature matrix to obtain a key matrix; Input the query vector, the value vector, and the key matrix into the cross-domain interaction encoder based on the imitation transformer structure to obtain the building region geographic feature-design constraint cross-domain query interaction feature vector.

8. The municipal green building design optimization method of claim 7, wherein, Based on the set of the building region geographic feature-design constraint fine-grained ablation factors, perform fine-grained ablation modulation aggregation on the set of the building region geographic feature multi-scale coding feature matrices to obtain the building region geographic feature-design constraint cross-modal ablation coding feature map, including: Input the set of the building region geographic feature-design constraint fine-grained ablation factors into the ablation effect coding module containing the normalization function and the mask function to obtain a set of building region geographic feature-design constraint fine-grained ablation weight factors; Based on the set of the building region geographic feature-design constraint fine-grained ablation weight factors, perform weighted modulation on the set of the building region geographic feature multi-scale coding feature matrices to obtain a set of building region geographic feature-design constraint cross-modal ablation coding feature matrices; Perform feature aggregation along the channel dimension on the set of the building region geographic feature-design constraint cross-modal ablation coding feature matrices to obtain the building region geographic feature-design constraint cross-modal ablation coding feature map.

9. The municipal green building design optimization method of claim 8, wherein, Based on the building region geographic feature-design constraint cross-modal ablation coding feature map, generate a municipal green building design sketch, including: Input the building region geographic feature-design constraint cross-modal ablation coding feature map into the municipal green building design sketch generator based on the large model to obtain the municipal green building design sketch.

Images