Urban form iterative generation system and method based on high-frequency interaction space self-learning

Abstract

The application discloses a kind of urban form iterative generation system and method based on high-frequency interaction space self-learning, wherein the system includes, first data acquisition module, for obtaining the target city space form three-dimensional vector model under current state;Second data acquisition module is used to build case city space form three-dimensional vector model database;Graph network module is used to build the associated network data of target city space form three-dimensional vector model and the associated network database of case city space form three-dimensional vector model database;Multi-objective high-frequency iteration module and feasibility scheme extraction and display module.The application realizes the multi-objective oriented generation iteration of city space form, and realizes the high-frequency interaction self-learning of city space form scheme generation, not only beneficial to improve the rationality of city space form scheme, but also avoid the problems of large investment of manpower and material resources in traditional urban design, involve brain judgment, large arbitrariness and small scale.

Description

TECHNICAL FIELD

[0001] The present application belongs to the field of urban design application transformation, and particularly relates to a high-frequency interactive space self-learning-based urban form iterative generation system and method. BACKGROUND

[0002] Urban form generation is one of the core tasks in urban design work, and a good urban form plays a crucial role in urban planning and development. It not only relates to the quality of life of residents and determines the shaping of the city image, but also affects the sustainable development of the city.

[0003] The traditional urban form design method is usually based on historical experience, planning principles and local characteristics, and gradually improves the urban layout through a series of design steps. This results in low efficiency of urban form generation, limited number of urban form generation iterations, and high artificial economic cost. SUMMARY

[0004] The first object of the present application is to provide a high-frequency interactive space self-learning-based urban form iterative generation method that provides high efficiency and can achieve high-frequency iteration under multi-target orientation.

[0005] The second object of the present application is to provide a high-frequency interactive space self-learning-based urban form iterative generation system.

[0006] Technical solution: The disclosed high-frequency interactive space self-learning-based urban form iterative generation method comprises the following steps:

[0007] S1: Obtain a target urban space form three-dimensional vector model under the current state;

[0008] S2: Collect a plurality of case urban space form three-dimensional model data, and construct a case urban space form three-dimensional vector model database;

[0009] S3: Take plots and buildings in the city as nodes, and take the correlation between plots, buildings belonging to the plots, and buildings in the same plot as edges to construct a graph network. From the three dimensions of scale, shape and height, the required indicators in the urban space are screened, the data extracted by quantifying the required indicators are converted into the attributes of the nodes, and the correlation network data of the target urban space form three-dimensional vector model and the correlation network database of the case urban space form three-dimensional vector model database are obtained;

[0010] S4: Based on the correlation network database of the case urban space form three-dimensional vector model database, the required indicators screened from the three dimensions of scale, shape and height are used to construct a plurality of target orientations for urban form iterative generation;

[0011] S5: input the associated network data of the target city spatial form three-dimensional vector model into the Neo4J platform, and the Neo4J platform iterates multiple times under multiple target-oriented reinforcement learning to obtain multiple target-oriented city spatial form schemes and construct a multiple-target-oriented city spatial form associated network scheme set;

[0012] S6: divide the city spatial form associated network scheme set into multiple scheme groups according to multiple target orientations, and cluster the required indexes contained in the schemes in the three scheme groups from the scale, shape and height dimensions to obtain the feasibility schemes corresponding to each group of schemes;

[0013] S7: convert the multiple feasibility schemes into three-dimensional vector spatial form models and output and display.

[0014] Further, the case city spatial form three-dimensional model data in step S2 includes building three-dimensional vector information, road contour line or road red line, plot boundary, and plot plane function division information containing geographic coordinates.

[0015] Further, the associated relationship in step S3 includes distance relationship, orientation relationship and position relationship, and the attributes of the edges of the graph network include distance attribute, orientation attribute and position attribute.

[0016] Further, the distance attribute refers to the distance between the center points of two nodes, the orientation attribute refers to the azimuth angle between two nodes, and the position attribute refers to whether two nodes are apart or connected.

[0017] Further, the required indexes in step S3 refer to: the scale dimension includes plot area, total building area and volume rate; the shape dimension includes shape index, fractal dimension, compactness, enclosure degree and street building ratio; and the height dimension includes building height standard deviation and average height.

[0018] Further, the multiple target orientations in step S3 include central development, group development and axis development.

[0019] Further, the central development, group development and axis development refer to:

[0020] Extract the data of the required indexes of each case city spatial form three-dimensional vector model from the scale, shape and height dimensions from the associated network database of the case city spatial form three-dimensional vector model database in step S3;

[0021] Normalize all the extracted data of the required indexes to obtain a normalized data set x', and the formula for normalization processing is as follows:

[0022]

[0023] wherein x ij denotes the value of the i-th case city on the j-th required index;

[0024] The weight w j corresponding to each required index is calculated, and the calculation formula is as follows:

[0025]

[0026] wherein y ij denotes the proportion of the i-th case city under the j-th required index, e j is the entropy value of the j-th required index, w j is the weight of the j-th required index;

[0027] According to the weight, the normalized required index quantification extraction data is calculated, the data of the required index quantification extraction after calculation is extracted, the land plots in the top 10% in terms of size, height and shape are obtained, adjacent land plots are clustered to obtain a cluster, and a cluster with a land plot number greater than 15 is defined as a core cluster.

[0028] If the spatial form three-dimensional vector model has only one core cluster, it is judged as central development.

[0029] If the spatial form three-dimensional vector model has multiple non-adjacent core clusters, it is judged as group development.

[0030] If the spatial form three-dimensional vector model has multiple adjacent core clusters, it is judged as axial development.

[0031] The graph network of the spatial form three-dimensional vector model of central development, group development and axial development is calculated respectively.

[0032] Further, the manner of obtaining multiple target-oriented urban spatial form schemes in step S5 is as follows:

[0033] The Neo4J platform is iterated multiple times under a certain target-oriented reinforcement learning, and a new spatial form association network is generated each time. The spatial form association network generated in the n-th iteration is A-NET n A-NET n is interacted and compared with the A-NET n-1 generated in the last iteration, and the unreasonable land plot part in A-NET n is restored to the state in A-NET n-1 ; the iteration stopping condition is set, and after the iteration stops, multiple urban spatial form schemes of the target orientation are obtained.

[0034] Further, the unreasonable land block refers to a land block in which the change in the A-NETn is more than 20%.

[0035] Based on the same inventive concept, the application also discloses a city form iterative generation system based on high-frequency interactive space self-learning, comprising,

[0036] A first data acquisition module is configured to acquire a target city space form three-dimensional vector model in a current state;

[0037] A second data acquisition module is configured to acquire a plurality of case city space form three-dimensional model data and construct a case city space form three-dimensional vector model database;

[0038] A graph network module is configured to construct a graph network of each city space form vector model in the target city space form three-dimensional vector model and the case city space form three-dimensional vector model database, obtain associated network data of the target city space form three-dimensional vector model and associated network database of the case city space form three-dimensional vector model database; wherein a land block and a building in a city are nodes of the graph network, and an associated relationship between land blocks, between a land block and a building belonging to the land block, and between buildings in the same land block is an edge of the graph network; required indexes in the city space are selected from three dimensions of scale, shape and height; and data extracted by quantizing the required indexes is converted into attributes of the nodes;

[0039] A multi-target high-frequency iteration module is configured to perform multiple iterations on the associated network data of the target city space form three-dimensional vector model under multiple target-oriented reinforcement learning, and obtain a multi-target-oriented city space form associated network scheme set; wherein the multiple target orientations are constructed based on the associated network database of the case city space form three-dimensional vector model database and required indexes selected from three dimensions of scale, shape and height;

[0040] A feasibility scheme extraction and display module is configured to divide the city space form associated network scheme set into multiple scheme groups under the multi-target orientation, respectively cluster schemes in the three scheme groups from required indexes contained in three dimensions of scale, shape and height, obtain a feasibility scheme corresponding to each group of schemes, and convert the multiple feasibility schemes into a three-dimensional vector space form model and output and display.

[0041] Beneficial effects: compared with the prior art, the present application has the following obvious advantages: the present application combines scale, shape and height three-dimensional indexes to construct three target orientations of central development, group development and axis development of city space form generation, realizes the multi-target oriented generation iteration of city space form, and realizes the high-frequency interactive self-learning of city space form scheme generation through the target-oriented reinforcement learning of the present application, which not only helps to improve the rationality of city space form scheme, but also avoids the problems of large investment of manpower and material resources, large randomness and small scale of traditional city design. BRIEF DESCRIPTION OF DRAWINGS

[0042] Figure 1 is a flowchart of the method of the present application;

[0043] Figure 2 is a graph network of the target city space form three-dimensional vector model of the present application;

[0044] Figure 3 is a city central development schematic diagram of the present application;

[0045] Figure 4 is a city group development schematic diagram of the present application;

[0046] Figure 5 is a city axis development schematic diagram of the present application;

[0047] Figure 6 is a plot city design guide translation diagram of the feasible scheme of the present application.

[0048] Figure 7 is a schematic diagram of the system of the present application. DETAILED DESCRIPTION

[0049] The technical solutions of the present application will be further described below in combination with the drawings.

[0050] Example 1

[0051] The city form iterative generation method based on high-frequency interactive space self-learning according to the present application, as shown in Figure 1 includes the following steps:

[0052] S1: Obtain a 3D vector model of the target city's spatial morphology in the current state. Preferably, the 3D vector model of the target city's spatial morphology is preprocessed. Preprocessing includes unifying the dimensions of the 3D vector model, retaining the model data, cleaning the model data (removing excess data, deleting duplicate data, and supplementing missing data), and including the height of buildings, area of ​​buildings, coordinates of building vertices, width of roads, coordinates of road centerlines, and coordinates of land parcel vertices in the target city.

[0053] S2: Collect 3D model data of urban spatial forms from multiple case studies and construct a database of 3D vector models of urban spatial forms from these case studies. The 3D model data of urban spatial forms from these case studies refers to 3D vector information of buildings containing geographic coordinates, road outlines or road red lines, plot boundaries, and plot functional zoning information.

[0054] S3: Using plots and buildings within the city as nodes, and the relationships between plots, between plots and buildings belonging to the same plot, and between buildings within the same plot as edges, construct a graph network. Filter the required indicators in the urban space from three dimensions: scale, shape, and height. Quantify the data extracted from the required indicators and convert them into node attributes. Obtain the associated network data of the target city spatial morphology 3D vector model and the associated network database of the case city spatial morphology 3D vector model database.

[0055] A graph network is constructed using land parcels and buildings as nodes, and the relationships between land parcels, between a land parcel and a building belonging to that land parcel, and between buildings within the same land parcel as edges. These relationships include distance relationships, orientation relationships, and location relationships.

[0056] The process involves quantifying and extracting indicators from all 3D vector information in the target city space, then filtering the required indicators based on three dimensions: scale, shape, and height. The data extracted from these indicators is then converted into node attributes. Similarly, the process involves quantifying and extracting the relationships within the 3D vector model of the target city's spatial morphology, converting this data into edge attributes. Finally, the nodes and their attributes, as well as the edges and their attributes, are stored in a graph network to obtain the network data of the 3D vector model of the target city's spatial morphology. Figure 2 As shown, buildings and land parcels are treated as nodes in a graph network. Data extracted from the required indicators (scale, shape, and height) is quantified and converted into node attributes. The relationships between land parcels, between a land parcel and buildings belonging to that land parcel, and between buildings belonging to the same land parcel are represented as edges. These edges have attributes including distance, orientation, and location. Preferably, the distance attribute refers to the distance between the center points of two nodes, the orientation attribute refers to the azimuth angle between two nodes, and the location attribute refers to whether two nodes are adjacent or separated.

[0057] With the city as an independent main body, the three-dimensional vector information of all single case cities in the case city spatial form three-dimensional vector model database is quantified and extracted one by one. The required indexes are screened from three dimensions of scale, shape and height, the data of the required index quantification extraction is converted into the attributes of nodes, the correlation of the single case city spatial form three-dimensional vector model is quantified and extracted, the data of the correlation quantification extraction is converted into the attributes of edges, and the nodes and their attributes, edges and their attributes are saved to the graph network to obtain the correlation network data of the single case city spatial form three-dimensional vector model. That is, the buildings and land plots are nodes of the graph network, the data of the required indexes screened from three dimensions of scale, shape and height after quantification extraction is converted into the attributes of nodes, and the correlation between land plots, land plots and buildings, and buildings and buildings is an edge, and its attributes include distance attribute, direction attribute and position attribute. The correlation network data of all single case city spatial form three-dimensional vector models constitutes the correlation network database of the case city spatial form three-dimensional vector model database.

[0058] In this embodiment, the spatial information of the three-dimensional model is digitized and converted by spatial computing technology, and the structured knowledge integration and storage of the spatial information of the three-dimensional model are realized by graph technology, which ensures the integrity and standardization of the data and improves the accuracy and efficiency of the calculation.

[0059] In this embodiment, the specific description of the required indexes screened from three dimensions of scale, shape and height, and the formula of the required indexes quantification extraction after screening are shown in Table 1.

[0060] Table 1

[0061]

[0062]

[0063] S4: Based on the correlation network database of the case city spatial form three-dimensional vector model database, the required indexes screened from three dimensions of scale, shape and height are used to construct multiple target orientations of iterative generation of urban form.

[0064] The required indexes screened from three dimensions of scale, shape and height are used to determine the weight of each required index by entropy weight method, and three target orientations of central development, group development and axis development of urban spatial form generation are constructed based on the weight combination of each required index.

[0065] (1) Extract the data of the required indexes screened from three dimensions of scale, shape and height of each case city spatial form three-dimensional vector model in the correlation network database of the case city spatial form three-dimensional vector model database in step S3.

[0066] (2) The quantified extraction data of all required indicators contained in the three dimensions of scale, shape, and height are divided by weight using the entropy weight method.

[0067] First, the normalized data set x' is obtained by normalizing the quantified extraction data of all required indicators, and the formula for normalization is as follows:

[0068]

[0069] In the formula, x ij represents the value of the i-th case city on the j-th required indicator;

[0070] The weight w j corresponding to each required indicator is calculated, and the formula is as follows:

[0071]

[0072] In the formula, y ij represents the proportion of the i-th case city under the j-th required indicator, e j is the entropy value of the j-th required indicator, and w j is the weight of the j-th required indicator. The obtained weight is used to measure the proportion of the required indicators contained in the three dimensions of scale, shape, and height, and to explore the landform characteristics, so as to finally obtain three groups of data: scale group, shape group, and height group.

[0073] According to the weight, the normalized quantified extraction data of the required indicators is calculated, and the land plots with the top 10% in scale, height, and shape dimensions in the calculated quantified extraction data of the required indicators are extracted. The adjacent land plots are clustered to obtain clustering clusters, and the clustering cluster with more than 15 land plots is defined as a core clustering cluster.

[0074] (3) If a spatial form three-dimensional vector model has only one core clustering cluster, it is judged as central development, as shown in Figure 3 ; if a spatial form three-dimensional vector model has multiple non-adjacent core clustering clusters, it is judged as group development, as shown in Figure 4 ; and if a spatial form three-dimensional vector model has multiple adjacent core clustering clusters, it is judged as axis type, as shown in Figure 5 .

[0075] (4) The required indicator quantization transformed data and the graph network composed of the associated relationships of the scale, shape, and height dimensions of the spatial form three-dimensional vector model of central development, group development, and axis type development are calculated respectively.

[0076] S5: input the associated network data of the target city spatial form three-dimensional vector model into the Neo4J platform, and the Neo4J platform iterates multiple times under multiple target-oriented reinforcement learning to obtain multiple target-oriented city spatial form schemes and construct a multi-target-oriented city spatial form associated network scheme set. That is, high-frequency interactive space self-learning is realized.

[0077] Based on the associated network data of the target city spatial form three-dimensional vector model in step S3, the Neo4J platform is iterated multiple times through target-oriented reinforcement learning, and a new spatial form associated network is generated each time. The spatial form associated network generated in the nth iteration is A-NETn. n A-NETn n is compared with the A-NETn n-1 generated in the last iteration, and the unreasonable plot part in A-NETn n is restored to the state in A-NETn n-1 , so as to form a new city spatial form associated network through continuous iteration. The spatial form associated network formed through each interactive iteration is compared with the cases in the associated network database of the case city spatial form three-dimensional vector model database. When the required indicators of the three dimensions of scale, form and height are all less than 10% of the difference of a certain case city, the iteration is stopped, otherwise the iteration is continued until the iteration condition is met.

[0078] Multiple rounds of high-frequency iteration of city spatial form associated network are carried out under the reinforcement learning of central development, group development and axis development, respectively, to obtain city spatial form schemes oriented to central development, group development and axis development, respectively. All the obtained city spatial form schemes constitute a multi-target-oriented city spatial form associated network scheme set.

[0079] The unreasonable plot is identified according to the following: a plot whose change is more than 20% compared with the associated network data of the target city spatial form three-dimensional vector model is defined as an unreasonable plot.

[0080] S6: According to the multi-target orientation, the urban space form related network scheme set is divided into multiple scheme groups, and the schemes in the three scheme groups are clustered from the required indexes contained in the three dimensions of scale, form and height to obtain the feasibility schemes corresponding to each group of schemes. The multi-target oriented urban space form related network scheme set of the target city is divided into three scheme groups according to the central development, group development and axis development, and the schemes in the three scheme groups are clustered from the required indexes contained in the three dimensions of scale, form and height. The schemes obtained after clustering are the feasibility schemes corresponding to each group of schemes. Preferably, the clustering mode is selected as K-means clustering.

[0081] S7: The feasibility schemes corresponding to the central development, group development and axis development are converted into three-dimensional vector space form models and displayed.

[0082] Preferably, the feasibility schemes are displayed on the sand table and human-computer interaction is realized.

[0083] A holographic projection sand table is established to interactively display the high-frequency iterative generation schemes of urban space form under multi-target orientation. The required equipment includes a three-dimensional physical sand table, a digital holographic projector and a gesture recognizer.

[0084] The interactive display process specifically includes using a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, a sand table to build a three-dimensional digital sand table, and a handheld controller to display the feasible schemes in step S7 in a virtual scene. VR glasses and virtual reality gloves are used to realize scene interaction of the model. Among them, the handheld controller is used to realize in-depth experience of the model; the virtual reality gloves are used to realize selection, scaling and modification of the model; and the VR glasses are used to realize interactive display of the model, that is, the model can be operated together with the virtual reality gloves.

[0085] A certain area of Chuzhou City is taken as a target city, and three-dimensional model data of urban space form of Hangzhou City, Nanjing City, Wuxi City, Suzhou City and Shanghai City are collected to build a three-dimensional vector model database of case city space form, and the high-frequency interactive space self-learning based urban form iterative generation method of the application is implemented. The feasibility schemes finally obtained are displayed in the plot urban design guide translation map of the feasible schemes together with the three-dimensional vector model of the target city space form in the current state, as shown in Figure 6 .

[0086] Example 2

[0087] A high-frequency interactive space self-learning based urban form iterative generation system is provided, as shown in Figure 7As shown, it comprises a first data acquisition module, a second data acquisition module, a graph network module, a multi-target high-frequency iteration module, and a feasible scheme extraction and display module.

[0088] The first data acquisition module is used to acquire the target urban spatial form three-dimensional vector model under the current state. Preferably, the target urban spatial form three-dimensional vector model is preprocessed, which includes unifying the dimensions of the target urban spatial form three-dimensional vector model, retaining the model data of the target urban spatial form three-dimensional vector model, cleaning the model data, removing the remaining data in the model data, deleting the repeated data in the model data, and supplementing the missing data in the model data. The model data includes the height of the building in the target city, the area of the building, the coordinates of each vertex of the building, the width of the road, the coordinates of the center line of the road, and the coordinates of each vertex of the plot.

[0089] The second data acquisition module is used to acquire a plurality of case city spatial form three-dimensional model data and construct a case city spatial form three-dimensional vector model database. The case city spatial form three-dimensional model data refers to the three-dimensional vector information of the building containing geographic coordinates, the road contour line or road red line, the plot boundary, and the plot plane function division information.

[0090] The graph network module is used to construct a graph network of each city spatial form vector model in the target city spatial form three-dimensional vector model and the case city spatial form three-dimensional vector model database, obtain the associated network data of the target city spatial form three-dimensional vector model and the associated network database of the case city spatial form three-dimensional vector model database; wherein the plots and buildings in the city are nodes of the graph network, the associated relationships between plots, between plots and buildings belonging to the plots, and between buildings in the same plot are edges of the graph network, and the required indexes in the city space are selected from three dimensions of scale, shape, and height. The data extracted by quantifying the required indexes is converted into the attributes of the nodes. The required indexes include plot area, total building area, and volume rate in the scale dimension, shape index, fractal dimension, compactness, enclosure degree, and street building ratio in the shape dimension, and building height standard deviation and average height in the height dimension. The specific description of the required indexes, as well as the formula for quantifying the required indexes and Table 1 in Embodiment 1 are consistent.

[0091] All three-dimensional vector information in the target city space is quantitatively extracted, and then the required indicators are screened from the three dimensions of scale, shape and height, and the data of the required indicators are converted into the attributes of the nodes; the correlation relationship of the target city space form three-dimensional vector model is quantitatively extracted, and the data of the correlation relationship are converted into the attributes of the edges. The nodes and their attributes, the edges and their attributes are saved to the graph network to obtain the correlation network data of the target city space form three-dimensional vector model. That is, the buildings and plots are nodes of the graph network, the data of the required indicators screened from the three dimensions of scale, shape and height are converted into the attributes of the nodes, and the correlation relationship between the plots, the plots and the buildings, and the buildings is the edge, and the attributes include the distance attribute, the azimuth attribute and the position attribute. Preferably, the distance attribute refers to the distance between the center points of two nodes, the azimuth attribute refers to the azimuth angle between two nodes, and the position attribute refers to whether two nodes are apart or connected.

[0092] With the city as an independent main body, all three-dimensional vector information of a single case city in the case city space form three-dimensional vector model database is quantitatively extracted, and the required indicators are screened from the three dimensions of scale, shape and height, and the data of the required indicators are converted into the attributes of the nodes, and the correlation relationship of the single case city space form three-dimensional vector model is quantitatively extracted, and the data of the correlation relationship are converted into the attributes of the edges. The nodes and their attributes, the edges and their attributes are saved to the graph network to obtain the correlation network data of the single case city space form three-dimensional vector model. That is, the buildings and plots are nodes of the graph network, the data of the required indicators screened from the three dimensions of scale, shape and height are converted into the attributes of the nodes, and the correlation relationship between the plots, the plots and the buildings, and the buildings is the edge, and the attributes include the distance attribute, the azimuth attribute and the position attribute. All the correlation network data of the single case city space form three-dimensional vector model constitute the correlation network database of the case city space form three-dimensional vector model database.

[0093] A multi-target high-frequency iteration module is used to perform multiple iterations on the correlation network data of the target city space form three-dimensional vector model under multiple target-oriented reinforcement learning to obtain a multi-target-oriented city space form correlation network scheme set. The multiple target orientations are constructed based on the correlation network database of the case city space form three-dimensional vector model database and the required indicators screened from the three dimensions of scale, shape and height. That is, high-frequency interactive space self-learning is realized.

[0094] The required indicators screened from the three dimensions of scale, shape and height are used to determine the weight of each required indicator by using the entropy weight method, and three target orientations of central development, group development and axis development of city space form generation are constructed based on the weight combination of each required indicator.

[0095] (1) Extract the required index quantization extraction data of each case city spatial form three-dimensional vector model in the associated network database of the case city spatial form three-dimensional vector model database in the extraction graph network module from the three dimensions of scale, shape and height.

[0096] (2) The quantization extraction data of all required indexes contained in the three dimensions of scale, shape and height are divided by weight using entropy weight method.

[0097] First, the normalized data set x' is obtained by normalizing all the quantization extraction data of the required indexes respectively, and the formula of the normalization processing is as follows:

[0098]

[0099] In the formula, x ij refers to the value of the i-th case city on the j-th required index;

[0100] The weight w j corresponding to each required index is calculated, and the calculation formula is as follows:

[0101]

[0102] In the formula, y ij refers to the proportion of the i-th case city under the j-th required index, e j is the entropy value of the j-th required index, and w j is the weight of the j-th required index. The obtained weight is used to measure the proportion of the required indexes contained in the three dimensions of scale, shape and height, and to explore the plot form characteristics, so as to finally obtain three groups of data of scale group, shape group and height group.

[0103] According to the weight calculation of the normalized required index quantization extraction data, the plots with the top 10% of scale, height and shape dimensions in the calculated required index quantization extraction data are extracted, and the adjacent plots are clustered to obtain clustering clusters. The clustering cluster with the number of plots in the clustering cluster greater than 15 is defined as the core clustering cluster.

[0104] (3) If a spatial form three-dimensional vector model has only one core clustering cluster, it is judged as central development; if a spatial form three-dimensional vector model has multiple non-adjacent core clustering clusters, it is judged as group development; if a spatial form three-dimensional vector model has multiple adjacent core clustering clusters, it is judged as axis type.

[0105] (4) The required index quantization conversion data and the graph network composed of the associated relationship of the three dimensions of scale, shape and height of the spatial form three-dimensional vector model of central development, group development and axis type are calculated respectively.

[0106] The associated network data of the target city spatial form three-dimensional vector model is input into the Neo4J platform, and the Neo4J platform is iterated multiple times under multiple target-oriented reinforcement learning to obtain multiple target-oriented city spatial form schemes and construct a multi-target-oriented city spatial form associated network scheme set.

[0107] Based on the associated network data of the target city spatial form three-dimensional vector model of the graph network module, the Neo4J platform is iterated multiple times through target-oriented reinforcement learning, and a new spatial form associated network is generated each time. The spatial form associated network generated in the nth iteration is A-NETn. n A-NETn n is compared with the A-NETn n-1 generated in the last iteration, and the unreasonable plot part in A-NETn n is restored to the state in A-NETn n-1 , so as to continuously iterate to form a new city spatial form associated network. The spatial form associated network formed by each interactive iteration is compared with the cases in the associated network database of the case city spatial form three-dimensional vector model database. When the required indicators in the three dimensions of scale, form, and height are less than 10% different from a certain case city, the iteration is stopped, otherwise the iteration is continued until the iteration condition is met.

[0108] Multiple rounds of city spatial form associated network high-frequency iteration are performed under reinforcement learning of central development, group development, and axis development, respectively, to obtain city spatial form schemes oriented to central development, group development, and axis development, respectively. All the obtained city spatial form schemes constitute a multi-target-oriented city spatial form associated network scheme set.

[0109] The unreasonable plots are identified according to the following criteria: The graph network for generating a new spatial form associated network A-NETn through target-oriented reinforcement learning is constructed, and the plots with a change of more than 20% compared with the associated network data of the target city spatial form three-dimensional vector model are defined as unreasonable plots.

[0110] The feasibility scheme extraction and display module is used to divide the multi-target-oriented city spatial form associated network scheme set into multiple scheme groups, and to cluster the required indicators in the three dimensions of scale, form, and height for the schemes in the three scheme groups to obtain the corresponding feasibility schemes of each group of schemes. The multiple feasibility schemes are converted into three-dimensional vector spatial form models and output for display.

[0111] According to the multi-objective orientation, the urban space form related network scheme set is divided into multiple scheme groups, and the schemes in the three scheme groups are clustered from the required indexes contained in the three dimensions of scale, form and height to obtain the feasibility schemes corresponding to each group of schemes. The multi-objective oriented urban space form related network scheme set of the target city is divided into three scheme groups according to the central development, group development and axis development, and the schemes in the three scheme groups are clustered from the required indexes contained in the three dimensions of scale, form and height. The schemes obtained after clustering are the feasibility schemes corresponding to each group of schemes. Preferably, the clustering mode is selected as K-means clustering.

[0112] The feasibility schemes corresponding to the central development, group development and axis development are converted into three-dimensional vector space form model outputs and displayed.

[0113] Preferably, the feasibility schemes are displayed in a sand table and human-computer interaction is realized.

[0114] A holographic projection sand table is established to interactively display the high-frequency iterative generation schemes of urban space form under multi-objective orientation. The required equipment includes a three-dimensional physical sand table, a digital holographic projector and a gesture recognizer.

[0115] The interactive display process specifically includes using a projector with a resolution of 4K or above, a depth sensor, a motion sensor, a touch sensor, a sand table to build a three-dimensional digital sand table, and a handheld controller to display the feasible schemes in step S7 in a virtual scene. VR glasses and virtual reality gloves are used to realize scene interaction of the model. Among them, the handheld controller is used to realize in-depth experience of the model; the virtual reality gloves are used to realize selection, scaling and modification of the model; the VR glasses are used to realize interactive display of the model, that is, the model can be operated together with the virtual reality gloves.

Claims

1. A method for generating urban form iteration based on high-frequency interaction space self-learning, characterized in that: The method comprises the following steps: S1: obtaining a target city space form three-dimensional vector model in the current state; S2: collecting a plurality of case city space form three-dimensional model data, and constructing a case city space form three-dimensional vector model database; S3: constructing a graph network with land plots and buildings in the city as nodes, and the correlation between land plots, buildings belonging to the land plots, and buildings in the same land plot as edges, filtering required indexes in the city from three dimensions of scale, shape, and height, and converting the data extracted by quantifying the required indexes into attributes of the nodes; obtaining the correlation network data of the target city space form three-dimensional vector model and the correlation network database of the case city space form three-dimensional vector model database; S4: based on the correlation network database of the case city space form three-dimensional vector model database, filtering the required indexes from three dimensions of scale, shape, and height, and constructing a plurality of target orientations of city form iterative generation; S5: inputting the correlation network data of the target city space form three-dimensional vector model into a Neo4J platform, and performing multiple iterations on the target city space form three-dimensional vector model under reinforcement learning of the plurality of target orientations respectively to obtain city space form schemes of the plurality of target orientations, and constructing a city space form correlation network scheme set of the plurality of target orientations; S6: dividing the city space form correlation network scheme set into three scheme groups according to the plurality of target orientations, clustering the required indexes included in the schemes in the three scheme groups from three dimensions of scale, shape, and height respectively to obtain a feasible scheme corresponding to each group of schemes; S7: converting the plurality of feasible schemes into three-dimensional vector space form models and outputting and displaying the three-dimensional vector space form models. 2.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 1, characterized in that: The case city space form three-dimensional model data in step S2 refers to three-dimensional vector information of buildings containing geographic coordinates, road contour lines or road red lines, land plot boundaries, and land plot plane function division information. 3.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 1, characterized in that: The correlation in step S3 includes distance relationship, orientation relationship, and position relationship, and the attributes of the edges of the graph network include distance attribute, orientation attribute, and position attribute. 4.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 3, characterized in that: The distance attribute refers to the distance between the center points of two nodes, the orientation attribute refers to the azimuth angle between two nodes, and the position attribute refers to whether two nodes are separated or connected. 5.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 1, characterized in that: The required indexes in step S3 refer to land plot area, total building area, and volume rate in the scale dimension, shape index, fractal dimension, compactness, enclosure degree, and street building ratio in the shape dimension, and building height standard deviation and average height in the height dimension. 6.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 1, characterized in that: The plurality of target orientations in step S4 include central development, group development, and axis development.

7. The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 6, characterized in that: The central development, group development, and axis development refer to: extracting the quantified data of the required indexes of each case city space form three-dimensional vector model in the correlation network database of the case city space form three-dimensional vector model database in step S3 from three dimensions of scale, shape, and height; performing normalization processing on the extracted quantified data of all the required indexes respectively to obtain a normalized data set x´, and the formula of the normalization processing is as follows: ； wherein x ij denotes the value of the i-th case city on the j-th required indicator; The weight w corresponding to each required index is calculated j The calculation formula is as follows: ； In the formula, y ij is the proportion of the i-th case city under the j-th required index, e j is the entropy value of the j-th required index, w j is the weight of the j-th required index; According to the weight, the data extracted by the normalized required index quantization is calculated, the land plots with the scale, height and shape dimensions in the top 10% of the calculated required index quantized extracted data are extracted, adjacent land plots are clustered to obtain a cluster, and a cluster with a number of land plots greater than 15 in the cluster is defined as a core cluster. If the spatial form three-dimensional vector model has only one core cluster, it is judged as central development. If the spatial form three-dimensional vector model has multiple non-adjacent core clusters, it is judged as group development. If the spatial form three-dimensional vector model has multiple adjacent core clusters, it is judged as axis development. The graph network of the spatial form three-dimensional vector model of central development, group development and axis development is calculated respectively. 8.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 1, characterized in that: The way to obtain multiple target-oriented urban spatial form schemes in step S5 is as follows: The Neo4J platform iterates multiple times under a goal-oriented reinforcement learning, and a new spatial form association network is generated each time. The spatial form association network generated in the nth iteration is A-NET n A-NET n is interactively compared with the A-NET n-1 generated in the last iteration, and unreasonable land portions in A-NET n are restored to the state in A-NET n-1 . Set the iteration stop condition, and obtain the multiple target-oriented urban spatial form schemes after the iteration stops. 9.The urban form iterative generation method based on high-frequency interaction space self-learning according to claim 8, characterized in that: The unreasonable land plots refer to the land plots with a change of more than 20% in the A-NETn.

10. A city form iterative generation system based on high-frequency interaction space self-learning, characterized in that: Comprise, The first data acquisition module is used to obtain the target urban spatial form three-dimensional vector model under the current state. The second data acquisition module is used to collect multiple case urban spatial form three-dimensional model data and construct a case urban spatial form three-dimensional vector model database. The graph network module is used to construct the graph network of each urban spatial form vector model in the target urban spatial form three-dimensional vector model and the case urban spatial form three-dimensional vector model database, obtain the correlation network data of the target urban spatial form three-dimensional vector model and the correlation network database of the case urban spatial form three-dimensional vector model database; wherein the land plots and buildings in the city are nodes of the graph network, the correlation relationship between the land plots, the land plots and the buildings belonging to the land plots, and the buildings in the same land plot is the edge of the graph network, and the required indexes in the city space are selected from the scale, shape and height dimensions, and the data extracted by the required index quantization is converted into the attributes of the nodes. The multi-target high-frequency iteration module is used to perform multiple iterations on the correlation network data of the target urban spatial form three-dimensional vector model under multiple target-oriented reinforcement learning, and obtain a set of multiple target-oriented urban spatial form correlation network schemes; wherein the multiple target orientations are constructed based on the correlation network database of the case urban spatial form three-dimensional vector model database and the required indexes selected from the scale, shape and height dimensions; The feasibility scheme extraction and display module is used to divide the set of urban spatial form correlation network schemes into three scheme groups according to the multiple target orientations, cluster the required indexes contained in the scale, shape and height dimensions in the schemes in the three scheme groups, obtain the corresponding feasibility schemes of each group of schemes, convert the multiple feasibility schemes into three-dimensional vector spatial form models, and output and display.

Images