Green space network optimization method and system for city contraction

Abstract

The invention discloses a green space network optimization method and system oriented to urban contraction, and relates to the technical field of landscape gardens and urban and rural planning, vacant land is identified through a high-resolution remote sensing image, a vacant land identification model is established in combination with a machine learning algorithm, the vacant rate is analyzed, grid unit types are divided by adopting K-means clustering, and an optimal space network is obtained. And constructing a vacant land conversion potential evaluation database and determining an index weight by adopting an objective weighting method. And constructing a green space network based on a graph theory model, and analyzing node connection by applying a gravitation model. Three target preference scenes of life, ecology and production are set, network performance change is simulated through global efficiency to determine an optimization threshold, and finally node importance and conversion priority are divided in combination with network topology structure indexes and potential values. The method can scientifically evaluate the potential of the green space in the urban contraction process, optimizes the spatial layout, improves the overall efficiency of the green space network, and provides scientific support for the planning and management of the green space.

Description

Technical Field

[0001] The invention relates to the technical field of landscape gardening and urban and rural planning, and in particular to a method and system for optimizing a green space network facing urban shrinkage. Background Art

[0002] With the transformation of development patterns and the reshaping of urban functions in resource-depleted cities and population-loss cities, the optimization of green space layout has become an important strategy to cope with urban shrinkage, improve land use efficiency, and improve the ecological environment. In the context of urban shrinkage, the ecological, production, and living functions of green space are particularly critical. Scientifically constructing a green space network can help alleviate the impact of population decline on the decline in demand for public service facilities and enhance the resilience and sustainability of urban ecosystems. However, due to the complex distribution and frequent changes of vacant land in urban shrinking areas, the potential assessment and network optimization of green space layout face many challenges.

[0003] At present, traditional methods for optimizing green space networks mostly rely on qualitative analysis and small-scale experimental research, usually carried out through field surveys, historical data analysis, etc. These methods are limited by the large investment of human and material resources and poor timeliness. It is difficult to dynamically evaluate and systematically optimize large-scale and multi-type vacant land, and it is impossible to fully reveal the complex relationship between green space layout and its ecological, production, and living functions. Summary of the invention

[0004] In order to address the deficiencies mentioned in the above background technology, the purpose of the present invention is to provide a green space network optimization method and system for urban shrinkage, which can evaluate the development potential of vacant space during urban shrinkage, optimize the green space network layout, and provide support for the planning and management of green space.

[0005] In a first aspect, the purpose of the present invention can be achieved by the following technical solution: a method for optimizing a green space network for urban shrinkage, the method comprising the following steps:

[0006] Obtain high-resolution remote sensing images of vacant land, input the high-resolution remote sensing images of vacant land into a pre-trained vacant land recognition model, and output an image of a candidate site for green space development;

[0007] The image of alternative sites for green space development is divided into grids to obtain multiple grid units, and the vacancy rate of each grid unit is obtained. The grid units are divided into types and patterns by combining K-means clustering to obtain a green space supplement strategy framework.

[0008] An objective weighting method is used to calculate the weight of a pre-constructed basic database for evaluating the conversion potential of vacant land, and a comprehensive weight value is obtained; wherein the pre-constructed basic database for evaluating the conversion potential of vacant land is constructed based on the dimensional indicators of life, ecology and production;

[0009] The green space conversion scenarios with different target preferences are pre-set and combined with the comprehensive weight value to obtain the input order. Based on the input order, the green space conversion scenarios with different target preferences are input into the pre-established green space network model to obtain the global optimization efficiency of the network.

[0010] The importance of the nodes is calculated based on the global optimization efficiency of the network and the comprehensive weight value. The conversion priority is obtained based on the importance of the nodes. The types of green spaces suitable for supplementation in different grid units are determined based on the importance of the nodes and the conversion priority combined with the green space supplementation strategy framework. Based on the types of green spaces suitable for supplementation in different grid units, an optimized urban green space network is generated.

[0011] In combination with the first aspect, in certain implementations of the first aspect, the method also includes: the high-resolution remote sensing images of the vacant land rely on Google Earth to obtain high-definition images of the central urban area at intervals of preset years and cut them into tiles, and n tiles covering various types of urban vacant land and non-vacant land are evenly selected to mark vacant land samples.

[0012] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the pre-trained vacant land recognition model uses the DeepLabv3 image segmentation model and Resnet34 as a backbone network for training, and the precision, recall and F1 score are used as indicators for evaluating the recognition accuracy of the model.

[0013] Among them, the pre-trained vacant land recognition model expression is as follows:

[0014] The calculation process of precision, recall and F1 is as follows:

[0015]

[0016] Where TP is the number of correctly predicted positive classes, FP is the number of incorrectly predicted positive classes, and FN is the number of incorrectly predicted negative classes.

[0017] The high-resolution remote sensing images of vacant land are input into the pre-trained vacant land identification model, and the vacancy time of different vacant lands is obtained through overlay analysis. The permanent and temporary vacant lands are determined, and the permanent vacant lands are used as alternative sites for green space development, finally obtaining the image of alternative sites for green space development.

[0018] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the calculation process of obtaining the vacancy rate of each grid unit is as follows:

[0019]

[0020] In the formula, R i is the vacancy rate of the i-th grid unit, V i is the total area of ​​vacant land in the i-th grid unit, S i is the total area of ​​the i-th grid cell.

[0021] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the process of dividing the grid units into types and modes respectively in combination with K-means clustering includes:

[0022] The grid cells are divided into four types: no vacancy, low vacancy, medium vacancy, and high vacancy through K-means clustering;

[0023] Taking the preset years as intervals, the final value of the vacancy rate is subtracted from the initial value to obtain the values ​​of each stage and the overall change. Combined with the overall change value, each unit is divided into three modes: contraction, stability and growth.

[0024] Among them, in the units of the contraction mode, priority should be given to adding production and ecological green spaces; in the units of the stability mode, priority should be given to adding living and ecological green spaces; in the units of the growth mode, priority should be given to adding living green spaces.

[0025] In combination with the first aspect, in certain implementations of the first aspect, the method further includes: the dimensional indicators of life, ecology, and production respectively include:

[0026] The indicators of the living dimension are selected from four indicators: population density, park competition effect, adjacent facility support effect and slow-moving network connection effect, which respectively evaluate the intensity of demand, the impact of park competition on demand, the facility support effect and the connection between the plot and the slow-moving network; the indicators of the ecological dimension are selected from four indicators: gyration radius, fractal dimension index, patch importance index and connectivity comprehensive index, to quantify the area characteristics, shape complexity, contribution to network connectivity and overall connectivity of the plot; the indicators of the production dimension are selected from four indicators: slope, sunshine conditions, transportation convenience and irrigation convenience, which respectively analyze the slope suitability of the plot, the proportion of the effective sunshine area and the distance from the traffic roads and water systems;

[0027] Then, the objective weighting method is used to generate the comprehensive weights of each indicator and obtain the comprehensive potential value of each dimension. The calculation process is as follows:

[0028]

[0029] Where: P j is the comprehensive potential value of the jth dimension; W j is the comprehensive weight of the jth indicator; I j is the normalized value of the jth indicator.

[0030] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the pre-set green space conversion scenarios with different target preferences include: life potential priority: according to the evaluation results of the life potential of the node, connecting to the green space network from high to low; ecological potential priority: according to the evaluation results of the ecological potential of the node, connecting to the green space network from high to low; production potential priority: according to the evaluation results of the production potential of the node, connecting to the green space network from high to low;

[0031] The construction process of the pre-built green space network model:

[0032] Select plots that have been idle for at least a preset number of years, and perform node connection simulations based on existing parks to determine the best optimized network and its conversion sequence;

[0033] Obtain urban road data, including crawling open street map urban road network data through Python's OSMnx package, adding three-level resistance points to model the time cost of connectivity, applying the travel OD cost matrix to analyze the time cost of node connectivity, using the gravity model to analyze the interaction between nodes, and creating node connections;

[0034] Using the urban road network as the basis, the green space network model is constructed by simplifying the existing urban parks and vacant land into network nodes and taking the shortest path and time cost as the edge weights.

[0035] The interaction between nodes is analyzed through the gravity model, and the connection length threshold is set according to the size of the node area. The connection rule between nodes is based on the area of ​​the two nodes and the distance relationship between them. If the length exceeds the set threshold, no node connection will be created.

[0036] The green space network model calculation process is as follows:

[0037] D(i,j)=Cost(O i ,D j )

[0038] In the formula, D(i,j) represents the i To the end point D j The time cost, Cost(O i ,D j ) is a function of the computational cost.

[0039] In combination with the first aspect, in some implementations of the first aspect, the method further includes: using a network global efficiency index to measure changes in overall network performance during the connection process, and relying on Python's network analysis library Network X to perform programming to implement node connection simulation, and obtain a network optimization performance curve and a key optimization threshold, wherein the key optimization threshold is calculated as follows:

[0040] Calculation process of network global optimization efficiency:

[0041]

[0042] In the formula, E(G) is the global efficiency of the network, n is the number of nodes in the network, i and j are any two nodes in the network, and d ij is the shortest cost path between node i and node j.

[0043] With the number of connected nodes as the horizontal axis and the network global efficiency improvement rate as the vertical axis, the performance change curve of the network during the connection process is fitted by polynomials;

[0044] The first-order derivative of zero and the first-order derivative below the average value are used as the key thresholds for improving the performance of the urban green space network.

[0045] In combination with the first aspect, in some implementations of the first aspect, the method further includes: the importance of the node is obtained by standardizing the node's degree centrality, betweenness degree, proximity degree, and potential value and performing equal weighted superposition;

[0046] The importance of each node in the three priority scenarios is superimposed, and K-means clustering is used to divide the nodes into three levels: high, medium, and low, corresponding to the high, medium, and low conversion priorities, as follows:

[0047]

[0048] S i Indicates the importance of the node; D i , B i , P i ,V i They represent the degree centrality weight, betweenness centrality, proximity degree and potential value of node i respectively, Max() is the maximum value and Min() is the minimum value.

[0049] In a second aspect, in order to achieve the above-mentioned purpose, the present invention discloses a green space network optimization system for urban shrinkage, comprising:

[0050] An image processing module is used to obtain high-resolution remote sensing images of vacant land, input the high-resolution remote sensing images of vacant land into a pre-trained vacant land recognition model, and output an image of a candidate site for green space development;

[0051] The grid division module is used to divide the green space development alternative site image into grids to obtain multiple grid units, obtain the vacancy rate of each grid unit, and combine K-means clustering to divide the grid units into types and patterns to obtain the green space supplement strategy framework;

[0052] A weight calculation module is used to calculate the weight of a pre-constructed vacant land conversion potential evaluation basic database using an objective weighting method to obtain a comprehensive weight value; wherein the pre-constructed vacant land conversion potential evaluation basic database is constructed based on dimensional indicators of life, ecology and production;

[0053] The model optimization module is used to input the green space conversion scenarios with different target preferences into the pre-established green space network model based on the input sequence and obtain the global optimization efficiency of the network by combining the pre-set green space conversion scenarios with different target preferences and the comprehensive weight value;

[0054] The network output module is used to calculate the importance of nodes based on the global optimization efficiency of the network and the comprehensive weight value, obtain the conversion priority based on the importance of the nodes, determine the types of green spaces suitable for supplementation in different grid units based on the importance of the nodes and the conversion priority combined with the green space supplementation strategy framework, and generate an optimized urban green space network based on the types of green spaces suitable for supplementation in different grid units.

[0055] Beneficial effects of the present invention:

[0056] The present invention can evaluate the development potential of vacant space during urban shrinkage, optimize the layout of green space networks, and provide support for the planning and management of green spaces. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0058] Figure 1 It is a schematic flow chart of the method of the present invention;

[0059] Figure 2 It is a schematic diagram of the supplementary types of urban green space of the present invention;

[0060] Figure 3 It is a schematic diagram of the optimization of the urban green space network of the present invention;

[0061] Figure 4 It is a schematic diagram of the system structure of the present invention. DETAILED DESCRIPTION

[0062] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] Embodiment 1:

[0064] like Figure 1 As shown, a green space network optimization method for urban shrinkage includes the following steps:

[0065] Obtain high-resolution remote sensing images of vacant land, input the high-resolution remote sensing images of vacant land into a pre-trained vacant land recognition model, and output an image of a candidate site for green space development;

[0066] High-resolution remote sensing images of vacant land are obtained by using Google Earth at preset intervals of five years for the central urban area and cut into tiles. Then n tiles covering various types of urban vacant land and non-vacant land are evenly selected to mark the vacant land samples.

[0067] Specifically, Google Earth was used to collect high-resolution remote sensing images of a central urban area in 2014, 2019, and 2024, with a resolution of about 1.19 meters. After the images were cut into 2 km × 2 km tiles, 5 tiles covering different types of vacant land and non-vacant land were selected for annotation to enhance the recognition ability and robustness of the model. After the annotation was completed, the images were cut into 224 × 224 pixel tiles. In order to further improve the accuracy of vacant land recognition, three common image segmentation models were compared and analyzed: U-Net, FCN, and DeepLabv3. U-Net is suitable for small sample data and can provide good fine-grained segmentation effects, but the segmentation accuracy is low in complex backgrounds; FCN is suitable for large-scale data processing, but it has deficiencies in detail segmentation and may require additional post-processing steps; DeepLabv3 can effectively capture a wide range of contextual information through the dilated convolution mechanism, with high accuracy, especially when dealing with complex backgrounds and multi-scale features. Finally, the combination of DeepLabv3 and ResNet34 performed best in terms of precision, recall, and F1 score, with precision, recall, and F1 score all exceeding 80%, so we adopted the DeepLabv3 model and used ResNet34 as the backbone network for model training and verification.

[0068] Table 1 Comparative analysis of the accuracy of vacant land identification

[0069] Model Precision Recall F1 score (F1) U-Net 78％ 76％ 77％ FCN 74％ 72％ 73％ DeepLabv3 82％ 80％ 81％

[0070] By training a qualified model, the images of the central urban area of ​​a city in three years were predicted to generate a vacant land distribution map, and the vacancy time of each plot was calculated through overlay analysis. The results showed that long-term vacant land (vacancy time ≥ 5 years) accounted for 7.17% of the total land area. These plots are less likely to be converted to other uses and are given priority as potential land for green space optimization.

[0071] The image of alternative sites for green space development is divided into grids to obtain multiple grid units, and the vacancy rate of each grid unit is obtained. The grid units are divided into types and patterns by combining K-means clustering to obtain a green space supplement strategy framework.

[0072] The pre-trained vacant land recognition model uses the DeepLabv3 image segmentation model and Resnet34 as the backbone network for training. Precision, recall, and F1 are used as indicators to evaluate the recognition accuracy of the model.

[0073] Among them, the pre-trained vacant land recognition model expression is as follows:

[0074] The calculation process of precision, recall and F1 is as follows:

[0075]

[0076] Where TP is the number of correctly predicted positive classes, FP is the number of incorrectly predicted positive classes, and FN is the number of incorrectly predicted negative classes.

[0077] The high-resolution remote sensing images of vacant land are input into the pre-trained vacant land recognition model, and the vacancy time of different vacant lands is obtained through overlay analysis. The permanent and temporary vacant lands are determined, and the permanent vacant land is used as an alternative site for green space development, finally obtaining an image of the alternative site for green space development.

[0078] The calculation process of obtaining the vacancy rate of each grid unit is as follows:

[0079]

[0080] In the formula, R i is the vacant land ratio of the i-th grid unit, i.e., the vacancy rate, V i is the total area of ​​vacant land in the i-th grid unit, S i is the total area of ​​the i-th grid cell.

[0081] The process of dividing the grid units into types and modes respectively by combining K-means clustering includes:

[0082] K-means clustering was used to divide grid cells into four types: no vacancy, low vacancy, medium vacancy, and high vacancy, and the values ​​were assigned as 0, 1, 2, and 3 respectively;

[0083] Taking the preset years as intervals, the final value of the vacancy rate is subtracted from the initial value to obtain the values ​​of each stage and the overall change. Combined with the overall change value, each unit is divided into three modes: contraction, stability and growth.

[0084] Among them, in the units of the contraction mode, priority should be given to adding production and ecological green spaces; in the units of the stability mode, priority should be given to adding living and ecological green spaces; in the units of the growth mode, priority should be given to adding living green spaces.

[0085] Specifically, the study area was divided into 1 km × 1 km grid cells, and after calculating the vacancy rate, the grids were divided into four types: “no vacancy”, “low vacancy”, “medium vacancy”, and “high vacancy” through K-means clustering. Figure 2Shown is a schematic diagram of the urban green space supplementation type of the present invention. The results show that the proportion of high vacancy grids has increased from 0.34% to 7.22% in the past decade, and is mainly distributed in urban fringe areas. By analyzing the trend of vacancy rate changes from 2014 to 2024, the grid units are divided into "contraction mode" (vacancy rate increases), "stable mode" (vacancy rate remains basically unchanged) and "growth mode" (vacancy rate decreases). The results show that the proportion of contraction mode units in the central urban area of ​​a certain city is as high as 57.03%, while the growth mode units account for only 19.26%. Combined with the local development model, a green space supplementation strategy is set, in which contraction mode units give priority to supplementing production and ecological spaces, stable mode units give priority to supplementing living and ecological spaces, and growth mode units give priority to supplementing living spaces to meet the development needs of different regions. As shown in Table 2 below;

[0086] Table 2 Definition of local urban development model

[0087]

[0088] An objective weighting method is used to calculate the weight of a pre-constructed basic database for evaluating the conversion potential of vacant land, and a comprehensive weight value is obtained; wherein the pre-constructed basic database for evaluating the conversion potential of vacant land is constructed based on the dimensional indicators of life, ecology and production;

[0089] The respectively include:

[0090] As shown in Table 3 below, the indicators of the living dimension are selected from four indicators: population density, park competition effect, adjacent facility support effect, and slow-moving network connection effect, which respectively evaluate the intensity of demand, the impact of park competition on demand, the facility support effect, and the connection between the plot and the slow-moving network; the indicators of the ecological dimension are selected from four indicators: gyration radius, fractal dimension index, patch importance index, and connectivity comprehensive index, to quantify the area characteristics, shape complexity, contribution to network connectivity, and overall connectivity of the plot; the indicators of the production dimension are selected from four indicators: slope, sunshine conditions, transportation convenience, and irrigation convenience, which respectively analyze the slope suitability of the plot, the proportion of the effective sunshine area, and the distance from the traffic roads and water systems;

[0091] In order to overcome the possible deviation caused by a single method, the CRITIC method and the entropy weight method are combined to generate a comprehensive weight value through linear weighted fusion. The CRITIC method evaluates the comparison strength by calculating the correlation between indicators. The indicator with a larger weight indicates that its information is more independent in multidimensional data. The weight calculation formula is:

[0092]

[0093] In the formula, Critic jis the comparison strength weight of the jth indicator; C j is the comparison strength of the jth indicator.

[0094] The entropy weight law determines the weight based on the discreteness of each indicator. The greater the discreteness, the greater the weight of the indicator. The weight formula is:

[0095]

[0096] In the formula, Entropy j is the discrete weight of the jth index, E j is the comparison strength of the jth indicator, which is calculated based on the correlation between this indicator and other indicators.

[0097] Then, the weights of the CRITIC method and the entropy weight method are combined, and the contrast intensity weights calculated by the CRITIC method and the discreteness weights calculated by the entropy weight method are linearly weighted and fused to generate a comprehensive weight value. The calculation process is as follows:

[0098] W j =αEntropy j +(1-α)Critic j

[0099] Where W j The comprehensive weight of the jth indicator, α is the linear weighting coefficient, α∈[0,1], which is used to balance the weight contribution of the CRITIC method and the entropy weight method.

[0100] Finally, the comprehensive potential value of each dimension is generated. The calculation process is as follows:

[0101]

[0102] Where: P j is the comprehensive potential value of the jth dimension, W j is the comprehensive weight of the jth indicator, I j is the normalized value of the jth indicator.

[0103] At the same time, the Getis-Ord Gi* statistical method was used to analyze the spatial aggregation effect of potential values. The results showed that the aggregation effect of ecological potential hotspots was weak overall, with a small number of hotspots distributed in the center and northwest of the urban area, production potential hotspots mainly located in the northwest of the urban area, and cold spots distributed in the northeast; living potential hotspots were mostly distributed in the central urban area, and cold spots were scattered on the edge of the urban area.

[0104] Table 3 Construction of indicator system

[0105]

[0106] The green space conversion scenarios with different target preferences are pre-set and combined with the comprehensive weight value to obtain the input order. Based on the input order, the green space conversion scenarios with different target preferences are input into the pre-established green space network model to obtain the global optimization efficiency of the network.

[0107] The pre-set green space conversion scenarios with different target preferences include: life potential priority: according to the evaluation results of the life potential of the node, the nodes are connected to the green space network from high to low; ecological potential priority: according to the evaluation results of the ecological potential of the node, the nodes are connected to the green space network from high to low; production potential priority: according to the evaluation results of the production potential of the node, the nodes are connected to the green space network from high to low;

[0108] The construction process of the pre-built green space network model:

[0109] Select plots that have been idle for at least five years, and conduct node connection simulations based on existing parks in 2024 to determine the best optimized network and its conversion sequence;

[0110] Use Python's OSMnx package to crawl open street map urban road network data, add three-level resistance points to model the time cost of connectivity, apply the travel OD cost matrix to analyze the time cost of node connectivity, use the gravity model to analyze the interaction between nodes, and create node connections;

[0111] Using the urban road network as the basis, the green space network model is constructed by simplifying the existing urban parks and vacant land into network nodes and taking the shortest path and time cost as the edge weights.

[0112] The interaction between nodes is analyzed through the gravity model, and the connection length threshold is set according to the size of the node area. The connection rule between nodes is based on the area of ​​the two nodes and the distance relationship between them. If the length exceeds the set threshold, no node connection will be created.

[0113] Specifically, the green space network is constructed based on graph theory, and the existing parks and potential land are simplified into network nodes. The shortest path time cost is used as the edge weight, and the road data extracted by the OSMnx tool are combined to add three levels of resistance points, namely, expressway intersections (113s), main road intersections (63s), and secondary road intersections (43s), for path optimization. The node connection rule sets the connection length threshold according to the size of the node area to ensure the scientificity and flexibility of the network connection. The results show that the node distribution is relatively random in the ecological priority scenario, the nodes in the production priority scenario are concentrated in the marginal area, and the nodes in the life priority scenario are concentrated in the central area of ​​the city.

[0114] The green space network model is as follows:

[0115] D(i,j)=Cost(O i ,D j )

[0116] In the formula, D(i,j) represents the i To the end point D j The time cost, Cost(O i ,D j ) is a function of the computational cost.

[0117] The study set up three scenarios: ecological potential priority, production potential priority, and life potential priority, to simulate the optimization process of green space network layout with life, ecology, and production value priority. In each scenario, nodes are connected to the network in order from high to low according to the potential value to form an optimized layout plan with different target preferences. The ecological priority scenario maximizes the ecological function of urban green space, the production priority scenario improves the economic benefits of land in marginal areas, and the life priority scenario improves the leisure accessibility of residents in the central urban area.

[0118] Calculation process of network global optimization efficiency:

[0119]

[0120] In the formula, E(G) is the global efficiency of the network, n is the number of nodes in the network, i and j are any two nodes in the network, and d ij is the shortest cost path between node i and node j.

[0121] The importance of the nodes is calculated based on the global optimization efficiency of the network and the comprehensive weight value. The conversion priority is obtained based on the importance of the nodes. The types of green spaces suitable for supplementation in different grid units are determined based on the importance of the nodes and the conversion priority combined with the green space supplementation strategy framework. Based on the types of green spaces suitable for supplementation in different grid units, an optimized urban green space network is generated.

[0122] Specifically, the least square method is used to fit the performance change curve, and the highest threshold of network performance is determined when the first-order derivative is zero, and the most cost-effective threshold is determined when the first-order derivative is lower than the average. Figure 3 The figure shows the schematic diagram of the urban green space network optimization of the present invention. The results show that the most cost-effective nodes in different scenarios are 216 (ecological priority), 192 (production priority) and 312 (life priority), while the highest performance nodes are 756 (ecological priority), 773 (production priority) and 819 (life priority).

[0123] The importance of the node is obtained by standardizing the node's degree centrality, betweenness degree, proximity degree and potential value and superimposing them with equal weights;

[0124] The importance of each node in the three priority scenarios is superimposed, and K-means clustering is used to divide the nodes into three levels: high, medium, and low, corresponding to the high, medium, and low conversion priorities, as follows:

[0125]

[0126] S i Indicates the importance of the node; D i , B i , P i ,V i They represent the degree centrality weight, betweenness centrality, proximity degree and potential value of node i respectively, Max() is the maximum value and Min() is the minimum value.

[0127] Embodiment 2: In the second aspect, as Figure 4 As shown, in order to achieve the above-mentioned purpose, the present invention discloses a green space network optimization system for urban shrinkage, comprising:

[0128] The image processing module 11 is used to obtain high-resolution remote sensing images of vacant land, input the high-resolution remote sensing images of vacant land into a pre-trained vacant land recognition model, and output an image of a candidate site for green space development;

[0129] A grid division module 12 is used to divide the green space development alternative site image into grids to obtain a plurality of grid units, obtain the vacancy rate of each grid unit, and divide the grid units into types and modes respectively in combination with K-means clustering to obtain a green space supplement strategy framework;

[0130] The weight calculation module 13 is used to calculate the weight of the pre-constructed vacant land conversion potential evaluation basic database by using an objective weighting method to obtain a comprehensive weight value; wherein the pre-constructed vacant land conversion potential evaluation basic database is constructed based on the dimensional indicators of life, ecology and production;

[0131] The model optimization module 14 is used to obtain an input sequence by combining the pre-set green space conversion scenarios with different target preferences with the comprehensive weight value, and input the green space conversion scenarios with different target preferences into the pre-established green space network model based on the input sequence to obtain the global optimization efficiency of the network;

[0132] The network output module 15 is used to calculate the importance of the node based on the global optimization efficiency of the network and the comprehensive weight value, obtain the conversion priority based on the importance of the node, determine the type of green space suitable for supplementation in different grid units based on the importance of the node and the conversion priority combined with the green space supplementation strategy framework, and generate an optimized urban green space network based on the type of green space suitable for supplementation in different grid units.

[0133] Based on the same inventive concept, the present invention also provides a computer device, which includes: one or more processors, and a memory for storing one or more computer programs; the program includes program instructions, and the processor is used to execute the program instructions stored in the memory. The processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA) or other programmable logic devices, discrete gates or transistor logic devices, discrete hardware components, etc. It is the computing core and control core of the terminal, which is used to implement one or more instructions, specifically for loading and executing one or more instructions in a computer storage medium to implement the above method.

[0134] It needs to be further explained that, based on the same inventive concept, the present invention also provides a computer storage medium, on which a computer program is stored, and the computer program is executed by a processor to execute the above method. The storage medium can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to, an electrical, magnetic, optical, electrical, magnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples (non-exhaustive list) of computer-readable storage media include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present invention, a computer-readable storage medium can be any tangible medium containing or storing a program, which can be used by an instruction execution system, device or device or used in combination with it.

[0135] In the description of this specification, the description with reference to the terms "one embodiment", "example", "specific example", etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present disclosure. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described can be combined in any one or more embodiments or examples in a suitable manner.

[0136] The above shows and describes the basic principles, main features and advantages of the present disclosure. Those skilled in the art should understand that the present disclosure is not limited by the above embodiments, and the above embodiments and descriptions are only for explaining the principles of the present disclosure. Without departing from the spirit and scope of the present disclosure, the present disclosure may have various changes and improvements, and these changes and improvements fall within the scope of the present disclosure to be protected.

Claims

1. A green space network optimization method for urban shrinkage, characterized in that: The method comprises the following steps: Obtain high-resolution remote sensing images of vacant land, input the high-resolution remote sensing images of vacant land into a pre-trained vacant land recognition model, and output an image of a candidate site for green space development; The image of alternative sites for green space development is divided into grids to obtain multiple grid units, and the vacancy rate of each grid unit is obtained. The grid units are divided into types and patterns by combining K-means clustering to obtain a green space supplement strategy framework. An objective weighting method is used to calculate the weight of a pre-constructed basic database for evaluating the conversion potential of vacant land, and a comprehensive weight value is obtained; wherein the pre-constructed basic database for evaluating the conversion potential of vacant land is constructed based on the dimensional indicators of life, ecology and production; The green space conversion scenarios with different target preferences are pre-set and combined with the comprehensive weight value to obtain the input order. Based on the input order, the green space conversion scenarios with different target preferences are input into the pre-established green space network model to obtain the global optimization efficiency of the network. The importance of the nodes is calculated based on the global optimization efficiency of the network and the comprehensive weight value. The conversion priority is obtained based on the importance of the nodes. The types of green spaces suitable for supplementation in different grid units are determined based on the importance of the nodes and the conversion priority combined with the green space supplementation strategy framework. Based on the types of green spaces suitable for supplementation in different grid units, an optimized urban green space network is generated.

2. The green space network optimization method for urban shrinkage according to claim 1 is characterized in that: The high-resolution remote sensing images of the vacant land are obtained by relying on Google Earth at preset intervals of time for the central urban area and cut into tiles. N tiles covering various types of urban vacant land and non-vacant land are evenly selected to mark the vacant land samples.

3. The green space network optimization method for urban shrinkage according to claim 1 is characterized in that: The pre-trained vacant land recognition model uses the DeepLabv3 image segmentation model and Resnet34 as the backbone network for training. Precision, recall and F1 score are used as indicators to evaluate the recognition accuracy of the model. Among them, the pre-trained vacant land recognition model expression is as follows: The calculation process of precision, recall and F1 is as follows: Where TP is the number of correctly predicted positive classes, FP is the number of incorrectly predicted positive classes, and FN is the number of incorrectly predicted negative classes; The high-resolution remote sensing images of vacant land are input into the pre-trained vacant land identification model, and the vacancy time of different vacant lands is obtained through overlay analysis. The permanent and temporary vacant lands are determined, and the permanent vacant lands are used as alternative sites for green space development, finally obtaining the image of alternative sites for green space development.

4. The green space network optimization method for urban shrinkage according to claim 1 is characterized in that: The calculation process of obtaining the vacancy rate of each grid unit is as follows: In the formula, R i is the vacancy rate of the i-th grid unit, V i is the total area of ​​vacant land in the i-th grid unit, S i is the total area of ​​the i-th grid cell.

5. The green space network optimization method for urban shrinkage according to claim 1 is characterized in that: The process of dividing the grid units into types and modes respectively by combining K-means clustering includes: The grid cells are divided into four types: no vacancy, low vacancy, medium vacancy, and high vacancy through K-means clustering; Taking the preset years as intervals, the final value of the vacancy rate is subtracted from the initial value to obtain the values ​​of each stage and the overall change. Combined with the overall change value, each unit is divided into three modes: contraction, stability and growth. Among them, in the units of the contraction mode, priority should be given to adding production and ecological green spaces; in the units of the stability mode, priority should be given to adding living and ecological green spaces; in the units of the growth mode, priority should be given to adding living green spaces.

6. The green space network optimization method for urban shrinkage according to claim 1 is characterized in that: The indicators of life, ecology and production dimensions include: The indicators of the living dimension are selected from four indicators: population density, park competition effect, adjacent facility support effect and slow-moving network connection effect, which respectively evaluate the intensity of demand, the impact of park competition on demand, the facility support effect and the connection between the plot and the slow-moving network; the indicators of the ecological dimension are selected from four indicators: gyration radius, fractal dimension index, patch importance index and connectivity comprehensive index, to quantify the area characteristics, shape complexity, contribution to network connectivity and overall connectivity of the plot; the indicators of the production dimension are selected from four indicators: slope, sunshine conditions, transportation convenience and irrigation convenience, which respectively analyze the slope suitability of the plot, the proportion of the effective sunshine area and the distance from the traffic roads and water systems; Then, the objective weighting method is used to generate the comprehensive weights of each indicator and obtain the comprehensive potential value of each dimension. The calculation process is as follows: Where P j is the comprehensive potential value of the jth dimension, W j is the comprehensive weight of the jth indicator, I j is the normalized value of the j-th indicator.

7. The green space network optimization method for urban shrinkage according to claim 1 is characterized in that: The pre-set green space conversion scenarios with different target preferences include: life potential priority: according to the evaluation results of the life potential of the node, the nodes are connected to the green space network from high to low; ecological potential priority: according to the evaluation results of the ecological potential of the node, the nodes are connected to the green space network from high to low; production potential priority: according to the evaluation results of the production potential of the node, the nodes are connected to the green space network from high to low; The construction process of the pre-built green space network model: Select plots that have been idle for at least a preset number of years, and perform node connection simulations based on existing parks to determine the best optimized network and its conversion sequence; Obtain urban road data, including crawling open street map urban road network data through Python's OSMnx package, adding three-level resistance points to model the time cost of connectivity, applying the travel OD cost matrix to analyze the time cost of node connectivity, using the gravity model to analyze the interaction between nodes, and creating node connections; Using the urban road network as the basis, the green space network model is constructed by simplifying the existing urban parks and vacant land into network nodes and taking the shortest path and time cost as the edge weights. The interaction between nodes is analyzed through the gravity model, and the connection length threshold is set according to the size of the node area. The connection rule between nodes is based on the area of ​​the two nodes and the distance relationship. If the length exceeds the set threshold, no node connection will be created. The green space network model calculation process is as follows: D(i,j)=Cost(O i ,D j ) In the formula, D(i,j) represents the i To the end point D j The time cost, Cost(O i ,D j ) is a function of the computational cost.

8. The green space network optimization method for urban shrinkage according to claim 7, characterized in that: The network global efficiency index is used to measure the changes in the overall network performance during the connection process, and the node connection simulation is implemented by programming based on the Python network analysis library Network X to obtain the performance curve of network optimization and the key optimization threshold. The key optimization threshold is calculated as follows: Calculation process of network global optimization efficiency: In the formula, E(G) is the global efficiency of the network, n is the number of nodes in the network, i and j are any two nodes in the network, and d ij is the shortest cost path between node i and node j; With the number of connected nodes as the horizontal axis and the network global efficiency improvement rate as the vertical axis, the performance change curve of the network during the connection process is fitted by polynomials; The first-order derivative of zero and the first-order derivative below the average value are used as the key thresholds for improving the performance of the urban green space network.

9. The green space network optimization method for urban shrinkage according to claim 1, characterized in that: The importance of the node is obtained by standardizing the node's degree centrality, betweenness degree, proximity degree and potential value and superimposing them with equal weights; The importance of each node in the three priority scenarios is superimposed, and K-means clustering is used to divide the nodes into three levels: high, medium, and low, corresponding to the high, medium, and low conversion priorities, as follows: S i Indicates the importance of the node; D i , B i , P i ,V i They represent the degree centrality weight, betweenness centrality, proximity degree and potential value of node i respectively, Max() is the maximum value and Min() is the minimum value.

10. A green space network optimization system for urban shrinkage, which adopts a green space network optimization method for urban shrinkage according to any one of claims 1 to 9, characterized in that: include: An image processing module is used to obtain high-resolution remote sensing images of vacant land, input the high-resolution remote sensing images of vacant land into a pre-trained vacant land recognition model, and output an image of a candidate site for green space development; The grid division module is used to divide the green space development alternative site image into grids to obtain multiple grid units, obtain the vacancy rate of each grid unit, and combine K-means clustering to divide the grid units into types and patterns to obtain the green space supplement strategy framework; A weight calculation module is used to calculate the weight of a pre-constructed vacant land conversion potential evaluation basic database using an objective weighting method to obtain a comprehensive weight value; wherein the pre-constructed vacant land conversion potential evaluation basic database is constructed based on dimensional indicators of life, ecology and production; The model optimization module is used to input the green space conversion scenarios with different target preferences into the pre-established green space network model based on the input sequence and obtain the global optimization efficiency of the network by combining the pre-set green space conversion scenarios with different target preferences and the comprehensive weight value; The network output module is used to calculate the importance of nodes based on the global optimization efficiency of the network and the comprehensive weight value, obtain the conversion priority based on the importance of the nodes, determine the types of green spaces suitable for supplementation in different grid units based on the importance of the nodes and the conversion priority combined with the green space supplementation strategy framework, and generate an optimized urban green space network based on the types of green spaces suitable for supplementation in different grid units.

Images