Building group automatic layout method and system based on function-morphology interaction

Abstract

The application discloses a kind of based on function-morphology interaction's building group automatic layout method and system, including the following steps: the design boundary data of target block, road network data, ecological environment data and existing building data are collected, and three-dimensional space digital sand table is constructed;Collect the building area of several case blocks, surrounding road attribute, building storey and land function data, construct land function and space form gray comatose matrix;Factor analysis is carried out to obtain function-morphology factor loading matrix file data, generate function-morphology association model based on knowledge graph;Function-morphology dynamic interaction building group generation is carried out using deep deterministic policy gradient algorithm, and building group layout scheme is generated;Input upper planning, output 3D block building group model object set after screening in accordance with the requirements of upper planning;3D block building group model is superimposed on digital sand table by screening, and the final scheme is determined to be output.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of urban planning, and particularly relates to a building group automatic layout method and system based on function-morphology interaction. BACKGROUND

[0002] With the continuous promotion of new urbanization construction, urban development gradually moves from the stage of pursuing incremental expansion to the stage of high-quality development focusing on spatial quality. Urban space is the result of the comprehensive action of two-dimensional function layout and three-dimensional spatial form. As the core components of the block space system, function and form influence and couple each other in space, and jointly create a good spatial order and living environment. Under this background, the contradiction between high-intensity building form and low-density function vitality in urban blocks is increasingly prominent. Simply pursuing the "height" and "breadth" of form can easily ignore the actual needs of function development, resulting in unreasonable production and living space layout, low operation efficiency and lack of spatial vitality, and further causing spatial disorder and imbalance between man and land, and other series of difficulties, hindering the sustainable and healthy development of cities.

[0003] The current common building group model layout research method mainly includes an automatic layout method based on direct nesting of a case library. In terms of research objects, it is limited to the design parameters of existing cases. In terms of research accuracy, due to cost constraints, the number of building group case samples is limited, the recognition accuracy is low, and it is difficult to cover various function scenarios and design positioning. In terms of research conclusions, the direct nesting method is insufficient to solve complex design problems limited by multiple design goals and constraint conditions, which may lead to insufficient matching degree of the scheme and the target block.

[0004] Therefore, it is urgent to solve the above problems. SUMMARY

[0005] The purpose of the application is to provide a building group automatic layout method based on function-morphology interaction. The application can realize high coupling and coordination of function and form, and achieve a more diverse and efficient block building group automatic layout method.

[0006] Technical scheme: To achieve the above purpose, the application discloses a building group automatic layout method based on function-morphology interaction, including the following steps:

[0007] (1) Collecting design boundary data, road network data, ecological environment data and existing building data of the target block, performing data processing and spatial superposition, and constructing a three-dimensional spatial digital sand table;

[0008] (2) Collect the building area, surrounding road attributes, building floor number and land use function data of several case blocks, process to obtain all land use function and spatial form index types of the case block building group, form a function-form case library, uniformly display all land use function and spatial form index types in gray scale, and construct a land use function and spatial form gray co-occurrence matrix;

[0009] (3) Based on the arrangement rules between each pixel point in the gray co-occurrence matrix, factor analysis is performed to obtain a function-form factor loading matrix file data, the factor loading matrix file data is imported into a graph database, the knowledge storage from a text file to a graph database is completed, the graph visualization function of the association relationship between variables and factors is provided, and a function-form association model based on a knowledge graph is generated;

[0010] (4) The deep deterministic policy gradient algorithm is applied to generate a function-form dynamic interaction building group, the function-form factor loading matrix in step (3) is imported and abstracted into the concepts of state, action and reward; the policy model output of the deep deterministic policy gradient algorithm is corresponded to the control signal of the corresponding Grasshopper battery pack, the function-form parameter adjustment cycle of the block building group model is based on the instructions from the deep deterministic policy gradient algorithm, and the state of the entity building group model is updated, the function-form parameters are batched and output, and the building group layout scheme of a commercial block, a commercial block, a residential block, a public service block and a mixed block is generated, i.e. the Rhinoceros model data of the 3D block building group is output;

[0011] (5) The building height, building density and development intensity index threshold range of the target block form in the upper planning are input, the Rhinoceros model data of the 3D block building group is connected to the corresponding input port of the Query Model Objects screening component, the screening conditions are adjusted in the attribute panel of the component, the screening conditions are the building height, building density and development intensity index threshold of the target block form in the upper planning, the defined processing data is run, and the 3D block building group model object set meeting the requirements of the upper planning after screening is output;

[0012] (6) Based on the target block digital sand table constructed in step (1), the 3D block building group model screened is superimposed on the digital sand table; the voice recognition equipment and mixed reality wearable three-dimensional motion capture equipment are carried on the three-dimensional holographic digital sand table to simulate the scheme and display the index, a man-machine interaction instruction library is constructed, and the final scheme is output by comparing and selecting in the actual scene through the mixed reality head-mounted equipment.

[0013] Optionally, in step (1), based on the GIS geographic information platform, the design boundary data, road network data, ecological environment data and existing building data of the target block are collected, data processing and spatial superposition are performed in the ArcGis platform, and a three-dimensional spatial digital sand table is constructed; the ecological environment data includes mountain data, water data, ecological sensitive area data and open space data, and the existing building data includes longitude and latitude coordinates, building bottom area and building layer information.

[0014] Optionally, in step (2), the building area, building layer number and surrounding road attribute are obtained by format conversion after being retrieved from the Open Street Map platform in the target block range; the land use function data is calculated by the POI data through the TF-IDF algorithm, and the calculation formula is:

[0015]

[0016] Wherein, v is the number of block unit, w is a certain POI type, k represents the number of cases blocks containing t v , S v,w represents the area of w type land in the v block unit, n v,w is the frequency of w type format in the v block unit, D represents the total amount of all POIs, |j:t v ∈d w | represents the number containing v; there is a corresponding relationship between the POI type and the land use function.

[0017] Optionally, in step (2), the spatial form index type includes three aspects of quantization index, namely, scale, landmark and diversity; wherein the scale index includes total area, total base area, building density, volume rate and average height; the landmark index includes maximum height and maximum base area; the diversity index includes height stagger and base area stagger; the land use function includes production type service function, life type service function and public welfare type service function.

[0018] Optionally, in step (2), the specific steps of obtaining all land use functions and spatial form index types of the case block building group are: based on the spatial position coordinates in the ArcGIS platform, the building group spatial form data obtained from the OSM platform is spatially linked with the case block unit data obtained by road network buffer processing to obtain all building base contour and building layer information in the case block, that is, all buildings in the same neighborhood share a neighborhood ID; then, according to the index calculation formula, the scale index, landmark index and diversity index of each block spatial form are calculated.

[0019] Optionally, in step (2), the gray level co-occurrence matrix of the land use function and the spatial form is constructed by using the imread function in MATLAB to input the spatial form data and the land use function data of the case block building group reclassified by the natural break method; the gray level co-occurrence matrix is calculated by the graycomatrix function, and the correlation C of the gray level co-occurrence matrix is calculated according to the formula,

[0020]

[0021] wherein N is the number of gray levels segmented by the natural break method, P(p, q) is the value of each data of the case block building group in the gray level co-occurrence matrix, p and q are the gray values of the land use function of the case block building group and each data of the case block building group, respectively; the correlation degree of the pixels in the image and their adjacent pixels is measured from the gray level co-occurrence matrix by the graycoprops function, reflecting the regularity and consistency of the texture; based on the frequency of the pixel pairs with specific gray value combinations in specific direction and distance data, the correlation characteristics between each spatial form data and land use function are quantitatively calculated.

[0022] Optionally, in step (3), the factor loading matrix is constructed based on the spatial form factors and the land use function type factors of the case block, and the spatial form data and the land use function data are derived as factor data and variable data, respectively; a FactorAnalysis object is created by using the scikit-learn library of Python, and the number of factors to be extracted is specified as 9 and the number of original variables is specified as 1; the spatial form factor data of the case block is imported as a NumPy array as the data set to be analyzed, the fit_transform method is used to fit the factor data, and the components_attribute is used to obtain the factor loading matrix of the three types of function industries, i.e., production service function, living service function and public service function; wherein the size of the load indicates the correlation strength between the building function industry variable and each factor, and the symbol indicates the positive or negative relationship between them; the correlation relationship between the factors in step (3) is created by Neo4jDesktop according to the correlation structure of the factor loading matrix, and the corresponding factor nodes, variable nodes and relationship networks are created; the visualization of the factor loading matrix is designed by using the Neo4j Browser graphical visualization tool, the Cypher query is run to create a graphical view, and the visualization settings are adjusted to render and export the graphical data set.

[0023] Optionally, in the deep deterministic policy gradient algorithm in step (4), the state is the current spatial form feature and land use function type of the building group in the block, the action is the next step design decision made according to the current state, and the action is realized by modifying the form attribute and function attribute of the building in the block; the reward is a method for evaluating the function-form matching degree of the building in the block by using a reward function; based on the factor load values of the corresponding function industry and the arrangement rules between the pixel points of the gray level co-occurrence matrix, the corresponding reward function is set to evaluate the advantages and disadvantages of the action.

[0024] Optionally, the function-form parameter adjustment cycle in step (4) first inputs the multi-agent and the environment chassis to obtain the initial state S0; at the tth step, the spatial agent is randomly assigned with actions At and Bt to obtain the state St; the reward function is used to calculate the income of St, and the evaluation value Qt of a series of action strategies in the current round is output; the current sequence (At, Bt, St, Qt) is stored in the experience pool D, and the experience replay and fixed Q network mechanism are combined to reduce process fitting and optimize training to obtain a series of action strategies with maximum Q value.

[0025] Based on the same inventive concept, the application discloses an automatic building group layout system based on function-form interaction, comprising:

[0026] A digital sand table construction module is used to collect design boundary data, road network data, ecological environment data and existing building data of a target block, perform data processing and spatial superposition, and construct a three-dimensional spatial digital sand table.

[0027] A gray level co-occurrence matrix construction module is used to collect building area, surrounding road attribute, building layer and land use function data of a plurality of case blocks, process the data to obtain all land use functions and spatial form index types of the building group of the case blocks, form a function-form case library, uniformly display all land use functions and spatial form index types in gray levels, and construct a land use function and spatial form gray level co-occurrence matrix.

[0028] A function-form association model generation module is used to perform factor analysis based on the arrangement rules between the pixel points of the gray level co-occurrence matrix to obtain a function-form factor loading matrix file data, import the factor loading matrix file data into a graph database, complete the knowledge storage from a text file to the graph database, provide a graph visualization function for the association relationship between variables and factors, and generate a function-form association model based on a knowledge graph.

[0029] The building group layout scheme generation module is used for building group generation of function-morphology dynamic interaction by applying a deep deterministic policy gradient algorithm, importing a function-morphology factor load matrix, and abstracting the matrix into the concepts of state, action and reward; the strategy model output of the deep deterministic policy gradient algorithm is corresponded to the control signal of the corresponding Grasshopper battery group, the function-morphology parameter adjustment cycle of the block building group model is based, instructions from the deep deterministic policy gradient algorithm are received, and the state of the entity building group model is updated, the function-morphology parameters are batched and output, and the building group layout schemes of the commercial block, the commercial block, the residential block, the public service block and the mixed block are generated, that is, the Rhinoceros model data of the 3D block building group is output;

[0030] The building group screening module is used for inputting the building height, building density and development intensity index threshold range of the target block morphology in the upper planning, connecting the Rhinoceros model data of the 3D block building group to the corresponding input port of the QueryModelObjects screening component, adjusting the screening conditions in the attribute panel of the component, the screening conditions being the building height, building density and development intensity index threshold of the target block morphology in the upper planning, running the defined processing data, and outputting the 3D block building group model object set meeting the requirements of the upper planning after screening;

[0031] The display output module is used for superimposing the 3D block building group model screened through the digital sand table of the target block on the digital sand table; the voice recognition equipment and the mixed reality wearable three-dimensional motion capture equipment are carried on the three-dimensional space holographic digital sand table to simulate the scheme and display the index, a man-machine interaction instruction library is constructed, the final scheme is determined through the mixed reality head-mounted equipment in the actual scene comparison and selection, and the final scheme is output.

[0032] Advantages: Compared with the prior art, the building group layout scheme generation method has the following advantages: the building group layout scheme generation method can improve the design efficiency, meet the high coupling and coordination of function and morphology, and effectively make up for the shortcomings of the traditional research method; the building group layout scheme generation method applies the function and morphology dual dimensions, constructs the function-morphology gray level co-occurrence matrix and the function-morphology correlation graph, avoids the limitation of single dimension of function or morphology, strengthens the coupling relationship between the function and the morphology of the building group, and makes the layout scheme more scientific and reliable; the building group layout scheme generation method applies the deep deterministic policy gradient algorithm, combines the Grasshopper battery group to execute parameter input, and realizes the building group generation of function-morphology dynamic interaction; the building group layout scheme generation method simulates different building group scheme design decisions and evaluates the effects, processes the nonlinear and multidimensional conditional constraints in the function-morphology layout of the scheme, pays attention to the differentiated characteristics of the site itself, enhances the pertinence and usability of the generated scheme, and thus creates innovative and efficient building group layout solutions. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 A flow chart of the present application;

[0034] Figure 2 A layout scheme iteration mechanism diagram in the present application;

[0035] Figure 3 A structural schematic diagram of the building group automatic layout system in the present application. DETAILED DESCRIPTION

[0036] The technical solutions of the present application are further described below in combination with the drawings.

[0037] Example 1

[0038] As shown in the drawings, an automatic building group layout method based on function-morphology interaction includes the following steps: Figure 1

[0039] (1) Collect the basic chassis data of the target block;

[0040] Based on the GIS geographic information platform, collect the design boundary data, road network data, ecological environment data and existing building data of the target block, process and spatially superimpose the data in the ArcGis platform, and construct a three-dimensional spatial digital sand table;

[0041] The design boundary data of step (1) is provided by the local planning department, the road network data is obtained by format conversion of FME (Feature Manipulation Engine) after the target block range is exported from the Open Street Map (OSM) platform, the ecological environment data is obtained from the local ecological environment bureau, the ecological environment data includes mountain data, water data, ecological sensitive area data and open space data, the existing building data is extracted from the Bing Maps open data set, and the existing building data includes longitude and latitude coordinates, building floor area and building floor number information;

[0042] (2) Construct the land use function and spatial morphology gray level co-occurrence matrix;

[0043] Collect the building area, surrounding road attributes, building number and land use function data of a number of case blocks, store the data using a database management system (DBMS), process to obtain all land use function and spatial morphology index types of the case block building group, form a function-morphology case library, uniformly display all land use function and spatial morphology index types in the function-morphology case library in gray scale, and construct the land use function and spatial morphology gray level co-occurrence matrix (GLCM);

[0044] ​The building footprint, building layers, and surrounding road attributes in step (2) are retrieved from the target block range exported by the Open Street Map platform and then converted into a format by FME (Feature Manipulation Engine); the land use function data is obtained from the POI data obtained by the Baidu Map open platform, and the TF-IDF algorithm is used to calculate the land use function data, and the calculation formula is as follows:

[0045]

[0046] wherein v is the number of a block unit, w is a certain POI type, k represents the number of cases of the block unit v containing the POI type t v , S v,w represents the area of the w type land use in the v block unit, n v,w is the frequency of the w type format in the v block unit, and D represents the total amount of all POIs, |j:t v ∈d w | represents the number containing v; the corresponding relationship between the POI type and the land use function is shown in Table 1.

[0047] Table 1: Corresponding relationship between POI type and land use function

[0048]

[0049] In step (2), the spatial form index type includes three aspects of quantification indexes, namely, scale, landmark, and diversity; the scale index includes total area, total base area, building density, volume rate, and average height; the landmark index includes maximum height and maximum base area; the diversity index includes height stagger and base area stagger; the land use function includes production type service function, life type service function, and public welfare type service function.

[0050] In the ArcGIS platform, based on the spatial position coordinates, the building group spatial form data obtained from the OSM platform is spatially linked with the case block unit data obtained by the road network buffer processing, to obtain all building base contour and building layer information in the case block, that is, all buildings in the same neighborhood share a neighborhood ID; then, according to the index calculation formula shown in Table 2, the field calculator is used to calculate the scale index, landmark index, and diversity index of each block spatial form.

[0051] Table 2: Index calculation formula

[0052]

[0053]

[0054] In step (2), the land use function and space form gray level co-occurrence matrix is constructed, the imread function in MATLAB (Matrix Laboratory) is used to input the case block building group space form data including total area, total base area, building density, volume rate, average height, maximum height, maximum base area, height stagger, base area stagger, and land use function data including production type service function, life type service function, and public welfare type service function, which are reclassified by the natural break method; the gray level co-occurrence matrix is calculated through the graycomatrix function, and the correlation C of the gray level co-occurrence matrix is calculated according to the formula,

[0055]

[0056] wherein N is the number of gray levels segmented by the natural break method, P(p, q) is the value of each item of data of the case block building group in the gray level co-occurrence matrix, p and q are the gray values of the land use function of the case block building group and each item of data of the case block building group respectively; the correlation degree of the pixels in the image and their adjacent pixels is measured from the gray level co-occurrence matrix through the graycoprops function, which reflects the regularity and consistency of the texture; based on the frequency of the pixel pairs with specific gray value combinations in specific direction and distance data, the correlation characteristics between each item of space form data and land use function are quantitatively calculated;

[0057] (3) Generating a function-form correlation model based on a knowledge graph

[0058] Based on the arrangement rules between each pixel point in the gray level co-occurrence matrix, the factor loading matrix file data is obtained by executing factor analysis through the FactorAnalysis of the Python library, so as to reveal the potential connection and structure between the function-form indicators of the block building group; the factor loading matrix file data in the csv format is imported into the Neo4j graph database, the knowledge storage from the text file to the graph database is completed, the graphical visualization function of the correlation relationship between the variables and factors is provided, the construction of the knowledge graph of the function-form related factors of the block building group is realized, and the function-form correlation model based on the knowledge graph is generated.

[0059] In step (3), the factor loading matrix is constructed based on the case block's spatial form factors and land use function type factors, and the spatial form data and land use function data are derived as factor data and variable data, respectively; a FactorAnalysis object is created using the scikit-learn library of Python, and the number of factors to be extracted is specified as 9 (n_components parameter) and the number of original variables is specified as 1 (n_features parameter); the spatial form factor data of the case block is imported as a NumPy array as the data set to be analyzed, the fit_transform method is used to fit the factor data, and the components_attribute is used to obtain the factor loading matrix of the three major function industries of production service function, life service function and public service function; the size of the load indicates the correlation strength of the building function industry variable and each factor, and the sign indicates the positive or negative relationship between them;

[0060] The correlation between factors in step (3) is based on the correlation structure of the factor loading matrix, and the corresponding factor nodes, variable nodes and relationship networks are created through Neo4jDesktop; the visualization of the factor loading matrix is designed using the Neo4j Browser graphical visualization tool, the Cypher query is run to create a graphical view, and the visualization settings are adjusted to render and export the graphical data set;

[0061] (4) Iterative layout scheme based on reinforcement learning algorithm

[0062] The DDPG (Deep Deterministic Policy Gradient) algorithm is applied to generate the building group based on the dynamic interaction of function-form, and the function-form factor loading matrix of the block building group in step (3) is imported and abstracted into the concepts of state, action and reward;

[0063] In step (4), in the DDPG (Deep Deterministic Policy Gradient) algorithm, the state (State) is the current spatial form characteristics and land use function type of the block building group, the action (Action) is the next design decision made according to the current state, which is realized by modifying the form attributes and function attributes of the block building; the reward (Reward) is a method for evaluating the function-form matching degree of the block building by using a reward function; based on the factor load values of the corresponding function industries and the arrangement rules between the pixel points of the gray level co-occurrence matrix, the corresponding reward function is set to evaluate the advantages and disadvantages of the action;

[0064] The policy model output of the DDPG deep deterministic policy gradient algorithm is corresponded to the control signal of the established corresponding Grasshopper battery pack, based on the function-morphology parameter adjustment cycle of the block building model constructed in Rhino, receives instructions from the function-morphology DDPG deep deterministic policy gradient descent algorithm, and updates the state of the entity building group model, and outputs the function-morphology parameters in batches to generate the building group layout scheme of the commercial block, the commercial block, the residential block, the public service block and the mixed block, that is, the Rhinoceros model data of the 3D block building group is output;

[0065] The function-morphology parameter adjustment cycle in step (4) first inputs the multi-agent and the environment chassis to obtain the initialization state S0; at the tth step, the spatial agent is randomly assigned an action A t and B t , to obtain the state S t ; the reward function is calculated for S t , and the evaluation value Q t of a series of action strategies in the current round is output; the current sequence (A t , B t , S t , Q t ) is stored in the experience pool D, combined with experience replay and fixed Q network mechanism to reduce process fitting, and a series of action strategies with maximum Q value are obtained by optimization training, as shown in Figure 2 ;

[0066] (5) Screening building group layout scheme

[0067] The building height, building density and development intensity index threshold range of the target block form in the upper planning are input; the Rhinoceros model data of the 3D block building group is connected to the corresponding input port of the Query Model Objects screening component, the screening conditions are adjusted in the attribute panel of the component, the screening conditions are the building height, building density and development intensity index threshold of the target block form in the upper planning, the defined processing data is run, and the 3D block building group model object set meeting the requirements of the upper planning after screening is output.

[0068] (6) Scheme display, optimization and output

[0069] Based on the target block digital sand table constructed in step (1), the 3D model of the screened block building group scheme is superimposed on the digital sand table; the voice recognition device and the mixed reality wearable three-dimensional motion capture device are carried on the three-dimensional holographic digital sand table to simulate the scheme and display the index, a human-computer interaction instruction library is constructed, and the final scheme is output by comparing and selecting in the actual scene through the mixed reality head-mounted device.

[0070] In terms of research objects, the application is not limited to existing cases and can dynamically adjust design parameters according to specific needs and environmental feedback; in terms of research accuracy, the application is based on a reinforcement learning algorithm and generates a model covering various functional scenarios and design orientations through continuous iteration and optimization design; in terms of research conclusions, the application has dynamic adaptability and optimization capability and can propose deeper design rules for complex design problems limited by multiple design objectives and constraint conditions.

[0071] Embodiment 2

[0072] As Figure 3 shown, based on the same inventive concept, the application discloses an automatic building group layout system based on function-morphology interaction, comprising:

[0073] A digital sand table construction module is used to collect design boundary data, road network data, ecological environment data and existing building data of a target block, perform data processing and spatial superposition, and construct a three-dimensional spatial digital sand table; the digital sand table construction module specifically executes the specific steps in step (1) in embodiment 1;

[0074] A gray level co-occurrence matrix construction module is used to collect building area, surrounding road attribute, building floor number and land function data of a plurality of case blocks, process the data to obtain all land function and spatial morphology index types of the building groups of the case blocks, form a function-morphology case library, uniformly display all land function and spatial morphology index types in gray level classification, and construct a land function and spatial morphology gray level co-occurrence matrix; the gray level co-occurrence matrix construction module specifically executes the specific steps in step (2) in embodiment 1;

[0075] A function-morphology association model generation module is used to perform factor analysis based on the arrangement rules between the pixel points of the gray level co-occurrence matrix to obtain a function-morphology factor loading matrix file data, import the factor loading matrix file data into a graph database, complete the knowledge storage from a text file to a graph database, provide a graph visualization function for the association relationship between variables and factors, and generate a function-morphology association model based on a knowledge graph; the function-morphology association model generation module specifically executes the specific steps in step (3) in embodiment 1;

[0076] The building group layout scheme generation module is configured to generate the building group by applying the deep deterministic policy gradient algorithm for function-morphology dynamic interaction, import a function-morphology factor loading matrix, abstract the function-morphology factor loading matrix into the concepts of state, action and reward, output the control signals of the corresponding Grasshopper battery pack from the policy model of the deep deterministic policy gradient algorithm, receive the instructions from the deep deterministic policy gradient algorithm based on the function-morphology parameter adjustment cycle of the block building group model, and update the state of the entity building group model, batch output the function-morphology parameters, and generate the building group layout schemes of the commercial block, the commercial block, the residential block, the public service block and the mixed block, that is, output the Rhinoceros model data of the 3D block building group. The building group layout scheme generation module specifically performs the specific steps in step (4) in Embodiment 1.

[0077] The building group screening module is configured to input the building height, building density and development intensity index threshold range of the target block morphology in the upper planning, connect the Rhinoceros model data of the 3D block building group to the corresponding input port of the QueryModelObjects screening component, adjust the screening conditions in the attribute panel of the component, the screening conditions are the building height, building density and development intensity index threshold of the target block morphology in the upper planning, run the defined processing data, and output the 3D block building group model object set that meets the requirements of the upper planning after screening.

[0078] The display output module is configured to superimpose the 3D block building group model screened according to the target block digital sand table on the digital sand table. The voice recognition device and the mixed reality wearable three-dimensional motion capture device are carried on the three-dimensional holographic digital sand table to simulate the scheme and display the index, a man-machine interaction instruction library is constructed, the final scheme is determined by comparing and selecting in the actual scene through the mixed reality head-mounted device, and the final scheme is output.

Claims

1. A functional-morphological interaction based automatic layout method for building complexes, characterized by, Comprise the following steps: (1) Collect the design boundary data, road network data, ecological environment data and existing building data of the target block, process and spatially superimpose the data, and construct a three-dimensional spatial digital sand table; (2) Collect the building area, surrounding road attributes, building layers and land use function data of several case blocks, process to obtain all land use function and spatial form index types of case block building groups, form a function-form case library, and uniformly display all land use function and spatial form index types in gray scale, and construct a land use function and spatial form gray co-occurrence matrix; (3) Based on the arrangement rules between each pixel point in the gray co-occurrence matrix, factor analysis is carried out to obtain a function-form factor loading matrix file data, which is imported into a graph database to complete the knowledge storage from a text file to a graph database, provide a graphical visualization function for the association between variables and factors, and generate a function-form association model based on a knowledge graph; (4) Apply deep deterministic policy gradient algorithm to generate building groups with dynamic interaction of function-form, import the function-form factor loading matrix in step (3) and abstract it into the concepts of state, action and reward; The output of the policy model of the deep deterministic policy gradient algorithm corresponds to the control signal of the corresponding Grasshopper battery pack, and the function-form parameter adjustment cycle of the block building group model is based on the instructions from the deep deterministic policy gradient algorithm and updates the state of the entity building group model, and outputs the function-form parameters in batches, respectively generating building group layout schemes of commercial blocks, commercial blocks, residential blocks, public service blocks and mixed blocks, that is, outputting Rhinoceros model data of 3D block building groups; (5) Input the building height, building density and development intensity index threshold range of the target block form in the upper planning, connect the Rhinoceros model data of the 3D block building group to the corresponding input port of the Query Model Objects filtering component, adjust the filtering conditions in the component's property panel, and the filtering conditions are the building height, building density and development intensity index threshold of the target block form in the upper planning, run the defined processing data, and output the 3D block building group model object set that meets the requirements of the upper planning after filtering; (6) Based on the target block digital sand table constructed in step (1), superimpose the 3D block building group model filtered on the digital sand table; Load voice recognition equipment, mixed reality wearable three-dimensional motion capture equipment on the three-dimensional holographic digital sand table to simulate the scheme and display the indicators, construct a human-computer interaction instruction library, and output the final scheme through mixed reality head-mounted equipment in the actual scene comparison and selection.

2. The functional-morphological interaction based automatic layout method of a building group according to claim 1, characterized in that: The step (1) is based on the GIS geographic information platform, collects design boundary data, road network data, ecological environment data and existing building data of the target block, processes the data and spatially superimposes in the ArcGis platform, and constructs a three-dimensional spatial digital sand table; the ecological environment data includes mountain data, water data, ecological sensitive area data and open space data, and the existing building data includes longitude and latitude coordinates, building base area and building layer information.

3. The method of claim 1, wherein: The building area, building layer, and surrounding road attribute are obtained by format conversion after being retrieved from the target block range through the Open Street Map platform in the step (2); and the land use function data is calculated by the POI data through the TF-IDF algorithm, and the calculation formula is: where v is the number of the block unit, w is a certain type of POI, k represents the number of cases of blocks containing the t v POI to be retrieved, S v,w represents the area of the w type of land in the v block unit, n v,w is the frequency of the w type of format in the vth block unit, D represents the total amount of all POIs, |j:t v ∈d w | indicates the number containing v; there is a corresponding relationship between the POI type and the land function.

4. The functional-morphological interaction based automatic layout method of a building group according to claim 3, characterized in that: The spatial form index type in the step (2) includes three aspects of quantization indexes of scale, landmark and diversity; wherein the scale index includes total area, total base area, building density, volume rate and average height; the landmark index includes maximum height and maximum base area; the diversity index includes height stagger and base area stagger; the land use function includes production type service function, life type service function and public welfare type service function.

5. The functional-morphological interaction based automatic layout method of a building group according to claim 4, characterized in that: The specific steps of obtaining all land use functions and spatial form index types of the case block building group in the step (2) are as follows: based on the spatial position coordinates in the ArcGIS platform, the building group spatial form data obtained from the OSM platform is spatially linked with the case block unit data obtained by road network buffer processing to obtain all building base contours and building layer information in the case block, that is, all buildings in the same neighborhood share a neighborhood ID; and then according to the index calculation formula, the scale index, landmark index and diversity index of each block spatial form are calculated.

6. The functional-morphological interaction based automatic layout method of a building group according to claim 5, characterized in that: In the step (2), the land use function and spatial form gray level co-occurrence matrix is constructed by using the imread function in MATLAB to input the case block building group spatial form data and land use function data reclassified by the natural break point method; the gray level co-occurrence matrix is calculated by the graycomatrix function, and the calculation formula of the correlation C of the gray level co-occurrence matrix is, Wherein, N is the number of gray levels segmented by the natural break point method, P(p,q) is the value of each data of each case block building group in the gray level co-occurrence matrix, p and q are the gray values of the land use function of the case block building group and each data of each case block building group, respectively; the correlation degree of the pixels and their adjacent pixels in the image is measured from the gray level co-occurrence matrix by the graycoprops function, which reflects the regularity and consistency of the texture; based on the frequency of the specific gray value combination of the pixels in the specific direction and distance data, the correlation characteristics between each spatial form data and land use function are quantitatively calculated.

7. The functional-morphological interaction based automatic layout method of a building group according to claim 1, characterized in that: The step (3) constructs a factor loading matrix based on the case block spatial form factor and land use function type factor, respectively derives the spatial form data and land use function data as factor data and variable data; through the scikit-learn library of Python, a FactorAnalysis object is created, and the number of factors to be extracted is specified as 9 and the number of original variables is specified as 1; the spatial form factor data of the case block is imported as a NumPy array as a data set to be analyzed, the factor data is fitted using the fit_transform method, and the factor loading matrix of the three major function industries of production service function, life service function and public service function is obtained using the components_ attribute; wherein, the size of the load represents the correlation strength of the building function industry variable and each factor, and the symbol represents the positive or negative relationship between them; the correlation relationship between the factors in the step (3) is created through Neo4j Desktop according to the correlation structure of the factor loading matrix, and the corresponding factor nodes, variable nodes and relationship networks are created; the visualization of the factor loading matrix is designed using the Neo4j Browser graphical visualization tool, the Cypher query is run to create a graphical view, and the visualization settings are adjusted to render and export the graphical data set.

8. The functional-morphological interaction based automatic layout method of a building group according to claim 1, characterized in that: In the step (4), in the deep deterministic policy gradient algorithm, the state is the current spatial form feature and land use function type of the block building group, the action is the next step design decision made according to the current state, and the action is realized by modifying the form attribute and function attribute of the block building; the reward is a method for evaluating the function-form matching degree of the block building through a reward function; based on the factor load value of the corresponding function industry and the arrangement rule between each pixel point of the gray level co-occurrence matrix, the corresponding reward function is set to evaluate the advantages and disadvantages of the action.

9. The functional-morphological interaction based automatic layout method of a building group according to claim 8, characterized in that: The function-form parameter adjustment cycle in the step (4) first inputs the multi-agent and the environment chassis to obtain the initial state S0; at the tth step, the spatial agent is randomly assigned with actions Bt and Bt to obtain the state St; the reward function is used to calculate the income of St, and the evaluation value Qt of a series of action strategies in the current round is output; the current sequence (At, Bt, St, Qt) is stored in the experience pool D, and the experience replay and fixed Q network mechanism are combined to reduce process fitting and optimize training to obtain a series of action strategies with maximum Q value.

10. A functional-morphological interaction based automatic layout system for a building complex, characterized by, It comprises: A digital sand table construction module for collecting design boundary data, road network data, ecological environment data and existing building data of a target block, performing data processing and spatial superposition, and constructing a three-dimensional spatial digital sand table; A gray level co-occurrence matrix construction module for collecting building area, surrounding road attribute, building layer and land use function data of a plurality of case blocks, processing to obtain all land use functions and spatial form index types of the case block building group, forming a function-form case library, uniformly grading all land use functions and spatial form index types, and constructing a land use function and spatial form gray level co-occurrence matrix; The function-morphology association model generation module is configured to perform factor analysis based on the arrangement rules between pixels in the gray level co-occurrence matrix to obtain a function-morphology factor loading matrix file data, import the factor loading matrix file data into a graph database, complete knowledge storage from a text file to the graph database, provide a graphical visualization function for the association relationship between variables and factors, and generate a function-morphology association model based on a knowledge graph. The building group layout scheme generation module is configured to apply a deep deterministic policy gradient algorithm to generate a building group based on function-morphology dynamic interaction, import the function-morphology factor loading matrix, and abstract the function-morphology factor loading matrix into concepts of states, actions, and rewards. The function-morphology parameter adjustment cycle based on the building group model is configured to receive instructions from the deep deterministic policy gradient algorithm and update the state of the entity building group model, output the function-morphology parameters in batches, and generate building group layout schemes of a commercial street block, a commercial street block, a residential street block, a public service street block, and a mixed street block, that is, output Rhinoceros model data of a 3D street building group. The building group screening module is configured to input threshold ranges of building height, building density, and development intensity indicators for a target street form in high-level planning, connect the Rhinoceros model data of the 3D street building group to corresponding input ports of a Query Model Objects screening component, adjust screening conditions in a property panel of the component, run defined processing data, and output a 3D street building group model object set that meets requirements of high-level planning after screening. The display output module is configured to superimpose the 3D street building group model selected through screening on a digital sand table according to a target street digital sand table. The voice recognition device, the mixed reality wearable three-dimensional motion capture device, the scheme simulation, and the index display are configured to be carried on the three-dimensional holographic digital sand table, a human-computer interaction instruction library is constructed, and a final scheme is output through comparison and selection in an actual scene by using the mixed reality head-mounted device.

Images