Urban park site selection decision-making method, system, equipment and medium

Abstract

The invention belongs to the technical field of urban planning, and provides an urban park site selection decision-making method, system, equipment and medium in order to solve the problem that the existing heat island effect relieving technology faces multiple bottlenecks, and multi-dimensional indexes are endowed with combined weights based on a game theory method and subjected to hierarchical normalization processing; according to the conversion heat contribution degrees of different surface feature categories, decomposing the combined weight of the surface temperature to different surface features of the target area; enabling the normalized spatial distribution map corresponding to the surface temperature to correspond to different surface features, and obtaining a thermal contribution normalized distribution map of different surface features in the target area; the above steps are fused to generate a target area site selection suitability distribution map, and screening is performed to obtain an urban park site selection scheme, so that the sensitivity loss of existing'one-step 'weighting on a high heat source area is avoided, and it is ensured that a park site selection decision not only can alleviate a heat island core area in a targeted manner, but also can adapt to the diversity characteristics of an urban spatial pattern.

Description

Technical Field

[0001] The present invention belongs to the technical field related to urban planning, and particularly relates to a method, a system, a device and a medium for making a decision on the location of an urban park. Background Technique

[0002] The statements in this part only provide background technical information related to the present invention, and do not necessarily constitute prior art.

[0003] In the analysis of the formation mechanism of the urban heat island effect, the thermodynamic property differences of surface cover types are considered the core driving factors. There are significant differences in aspects such as heat absorption efficiency, heat conduction rate, and long-wave radiation release ability among different materials of the surface, such as asphalt, concrete, vegetation, etc. For example, high-density building clusters form a significant heat accumulation effect through multiple reflections and retention of solar radiation by their three-dimensional surfaces; while the extensive use of metal and glass materials in industrial areas further enhances the complexity of the local thermal environment. However, existing thermal environment assessment models generally adopt a simplified land cover classification system, roughly dividing surface types into two categories of "hardened" and "natural", and failing to accurately quantify the differences in thermodynamic parameters of different material combinations. This rough treatment leads to systematic biases in heat flux simulation, and it is particularly difficult to analyze the gradual impact of the gradual update of land cover, such as the replacement of traditional pavements with permeable pavements, on the heat island pattern.

[0004] Current urban heat island effect mitigation technologies face multiple bottlenecks. At the data acquisition level, the thermal field inversion method relying on a single remote sensing data source has inherent defects: medium-resolution thermal infrared images, such as those at the 30-meter level, can depict the macroscopic thermal field distribution but cannot identify thermal heterogeneity at the block scale; while high-temporal-resolution data, such as daily-scale data, are severely limited by the spatial resolution and are difficult to support refined thermal environment analysis. In terms of thermodynamic mechanism modeling, existing research mostly uses statistical regression models or machine learning black box models based on the vegetation index NDVI. Although temperature field prediction can be achieved, there is a lack of interpretability of the physical process of surface energy exchange. For example, the vegetation cooling effect is often simplified as a linear function of the leaf area index, ignoring the dynamic regulation of canopy transpiration efficiency by micro-meteorological conditions such as wind speed and humidity. In addition, existing planning models mostly regard the urban heat island effect as a homogeneous background field, only using the heat island intensity index as a constraint condition, and failing to establish a quantitative mapping relationship between land cover type conversion and thermal environment response, resulting in insufficient targeting of cooling measures.

[0005] The limitations of multi-objective decision-making algorithms further restrict the optimization of urban heat island mitigation effectiveness. Traditional spatial multi-criteria decision-making models usually adopt static weight allocation strategies, fixing target parameters such as urban heat island mitigation, service coverage, and economic costs as fixed ratios. However, the differences in urban development stages and regional characteristics require decision-making models to have dynamic response capabilities: in the initial stage of new city construction, urban heat island mitigation may be given priority, while in mature built-up areas, service fairness needs to be emphasized. Existing static models are difficult to adapt to this dynamic demand, resulting in target conflicts in the implementation of planning schemes. In addition, the subjective intervention of humans in the weight allocation process, such as the fuzzy interval setting of the expert scoring method, may introduce systematic biases, marginalizing areas with high heat regulation potential but high development costs in the comprehensive evaluation.

[0006] In recent years, research in related technical fields has focused on multi-source data fusion and mechanism model optimization. Some studies have attempted to construct a land surface temperature downscaling model by combining optical images and thermal infrared data, but its algorithm lacks robustness to cloud cover and aerosol interference, and the computational complexity increases significantly with the increase in data dimensions. In terms of the analysis of thermodynamic processes, some scholars have proposed a method for decomposing surface fluxes based on the energy balance equation to reveal the internal mechanism of heat island formation by separating sensible heat, latent heat, and storage heat fluxes, but the response sensitivity of the model to the gradual change process of land cover still needs to be improved. In the field of decision support, dynamic weight optimization algorithms have begun to be introduced to enhance multi-objective coordination capabilities, but the convergence efficiency and stability of existing methods under complex constraint conditions have not yet reached practical requirements. These technical explorations show that improvements in a single dimension are difficult to fundamentally solve the systematic disconnection between heat island effect analysis and planning response, and there is an urgent need to construct a full-chain technical system from thermodynamic mechanism analysis to spatial decision optimization to provide an operable decision-making path for the scientific layout of parks in urban environments. Summary of the Invention

[0007] To overcome the deficiencies of the above-mentioned prior art, the present invention provides a method, system, device, and medium for urban park site selection decision-making, ensuring that the park site selection decision can both target and mitigate the core area of the heat island and adapt to the diverse characteristics of the urban spatial pattern.

[0008] To achieve the above object, the present invention adopts the following technical solutions: In the first aspect, the present invention provides a method for urban park site selection decision-making, including: Obtain multi-temporal remote sensing data of the target area, count the mutual conversion information of ground object categories between the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution degree of different ground object categories; Perform hierarchical normalization processing on multi-dimensional indicators to obtain the normalized spatial distribution maps of each indicator in the target area, and assign combined weights to the relative importance among multi-dimensional indicators based on game theory; decompose the combined weight of the surface temperature to different ground objects in the target area according to the conversion heat contribution degree of different ground object categories to obtain the comprehensive weights of different ground objects in the target area; Correspond the normalized spatial distribution map corresponding to the surface temperature with different ground objects in the target area to obtain the normalized distribution map of heat contribution of different ground objects in the target area; Fuse the comprehensive weights of different ground objects and the normalized distribution maps of heat contribution of corresponding ground objects, as well as the combined weights of each indicator and the normalized spatial distribution maps corresponding to each indicator to generate the suitability distribution map for site selection in the target area, extract the hot spots of suitability, construct an OD cost matrix with the residential area POI as the starting point and the hot spots as the ending point, and screen candidate points according to the time cost to obtain the site selection plan for urban parks.

[0009] In a second aspect, the present invention provides an urban park site selection decision-making system, including: A calculation module configured to: obtain multi-temporal remote sensing data of the target area, count the mutual conversion information of ground object categories among the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution degree of different ground object categories; A weight module configured to: perform hierarchical normalization processing on multi-dimensional indicators to obtain the normalized spatial distribution maps of each indicator in the target area, and assign combined weights to the relative importance among multi-dimensional indicators based on game theory; decompose the combined weight of the surface temperature to different ground objects in the target area according to the conversion heat contribution degree of different ground object categories to obtain the comprehensive weights of different ground objects in the target area; An allocation module configured to: correspond the normalized spatial distribution map corresponding to the surface temperature with different ground objects in the target area to obtain the normalized distribution map of heat contribution of different ground objects in the target area; A site selection module configured to: fuse the comprehensive weights of different ground objects and the normalized distribution maps of heat contribution of corresponding ground objects, as well as the combined weights of each indicator and the normalized spatial distribution maps corresponding to each indicator to generate the suitability distribution map for site selection in the target area, extract the hot spots of suitability, construct an OD cost matrix with the residential area POI as the starting point and the hot spots as the ending point, and screen candidate points according to the time cost to obtain the site selection plan for urban parks.

[0010] In a third aspect, the present invention provides an electronic device, including a memory and a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method described in the first aspect is completed.

[0011] Fourthly, the present invention provides a computer-readable storage medium for storing computer instructions, which, when executed by a processor, implement the method described in the first aspect.

[0012] The above one or more technical solutions have the following beneficial effects: In the present invention, multi-dimensional indicators are subjected to hierarchical normalization processing to obtain a normalized spatial distribution map of each indicator in the target area, and a combined weight is assigned to the relative importance between multi-dimensional indicators based on the game theory method, breaking through the limitations of traditional subjective weighting or single objective weighting; according to the conversion heat contribution degree of different land cover types, the combined weight of the surface temperature is decomposed to different land covers in the target area, and the normalized spatial distribution map corresponding to the surface temperature is corresponded to different land covers in the target area to obtain a normalized heat contribution distribution map of different land covers in the target area, avoiding the sensitivity loss of the existing "one-size-fits-all" weighting to the high heat source area, and ensuring that the park site selection decision can not only target the mitigation of the heat island core area, but also adapt to the diversity characteristics of the urban spatial pattern.

[0013] Advantages of additional aspects of the present invention will be given in part in the following description, become apparent in part from the following description, or be learned through the practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] The specification drawings constituting a part of the present invention are used to provide a further understanding of the present invention. The schematic embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation to the present invention.

[0015] Figure 1 It is a flow chart of the urban park site selection decision method in the first embodiment of the present invention; Figure 2 It is a schematic diagram of the remote sensing data processing process in the first embodiment of the present invention; Figure 3 It is a schematic diagram of the framework of the game theory comprehensive weighting method in the first embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0016] It should be noted that the following detailed description is exemplary and is intended to provide further explanation of the present invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by those of ordinary skill in the technical field to which the present invention belongs.

[0017] It should be noted that the terms used herein are only for describing specific embodiments and are not intended to limit the exemplary embodiments according to the present invention.

[0018] Without conflict, the embodiments in the present invention and the features in the embodiments can be combined with each other.

[0019] Term Explanation: Residential area POI (Point of Interest) refers to geographical elements closely related to residents' lives, such as communities, apartment buildings, residences, etc. These elements are abstracted as point elements in the geographical information system to enrich the map content and provide the geographical information required by users.

[0020] The OD cost matrix (Origin-Destination Cost Matrix) is a network analysis tool used to calculate and store the minimum travel costs between multiple origins and multiple destinations.

[0021] Example 1 This example discloses a method for making a decision on the location of an urban park, including: Obtain multi-temporal remote sensing data of the target area, count the mutual conversion information of land cover types between the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution degree of different land cover types; Perform hierarchical normalization processing on multi-dimensional indicators to obtain the normalized spatial distribution maps of each indicator on the target area, and assign combined weights to the relative importance between multi-dimensional indicators based on game theory; according to the conversion heat contribution degree of different land cover types, decompose the combined weight of the surface temperature to different land covers in the target area to obtain the comprehensive weights of different land covers in the target area; Correspond the normalized spatial distribution map corresponding to the surface temperature with different land covers in the target area to obtain the normalized distribution map of heat contribution of different land covers in the target area; Fuse the comprehensive weights of different land covers and the normalized distribution maps of heat contribution of corresponding land covers, as well as the combined weights of each indicator and the normalized spatial distribution maps corresponding to each indicator to generate the location suitability distribution map of the target area, extract the hot spots of suitability, construct an OD cost matrix with residential area POI as the origin and the hot spots as the destination, and screen candidate points according to time cost to obtain the location plan of the urban park.

[0022] The solution of this example assigns weights to the multi-dimensional indicators of the target area based on game theory, breaking through the limitations of traditional subjective weighting or single objective weighting; dynamically adjusts the weights of the heat island effect of different land cover types according to the conversion heat contribution degree of different land cover types, avoiding the sensitivity loss of high heat source areas in the existing "one-size-fits-all" weighting method, and ensuring that the park location decision can not only target and alleviate the core area of the heat island, but also adapt to the diversity characteristics of the urban spatial pattern.

[0023] The following combines Figures 1 - 2 Make a detailed description of a method for making a decision on the location of an urban park proposed in this example: Step 1: Obtain multi-temporal remote sensing data of the target area, count the mutual conversion information of land cover types between the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution degree of different land cover types.

[0024] Obtain multi-temporal remote sensing data of the target area and perform preprocessing, including: radiometric calibration, atmospheric correction, image registration, etc.

[0025] As an implementation, obtain dual-temporal Sentinel-2 multi-spectral images of the target area in summer, such as from June to August, with cloud cover < 10% and a spatial resolution of 10 meters; synchronously obtain dual-temporal Landsat 8 images of the same area in summer, with cloud cover < 10% and a resolution of 30 meters.

[0026] Use the Sen2Cor tool to perform atmospheric correction on the Sentinel-2 images to obtain surface reflectance data; use the FLAASH model to correct the Landsat 8 images to eliminate aerosol and water vapor interference. The formula is:

[0027] In the formula, is the radiance at the sensor's entrance pupil, DN is the gray value of the original pixel, is the multiplicative coefficient of radiance, is the additive coefficient of radiance, is the solar zenith angle.

[0028] Then, based on the remote sensing data obtained by Sentinel-2, use the ENVI feature matching method to perform geometric fine correction on the remote sensing data obtained by Landsat 8, with the spatial registration error ≤ 1 pixel.

[0029] For example, obtain dual-temporal summer Sentinel-2 remote sensing images and Landsat 8 remote sensing images with cloud cover < 10% for the three cities of Shenyang, Xi'an, and Wuhan, and extract data such as road data, building data, water surface data, and night light data products for the corresponding years of the three cities based on OpenStreetMap. Among them, Shenyang is for 2017 and 2020, Xi'an is for 2019 and 2021, and Wuhan is for 2013 and 2016.

[0030] Step 11: Statistically analyze the mutual conversion information of land cover types between multi-temporal remote sensing data of the target area, calculate the conversion area between land cover types in the target area, and the net change in conversion temperature between land cover types.

[0031] Step 111: Based on the multi-temporal remote sensing data of the target area, respectively obtain the land cover type probability maps corresponding to different temporal remote sensing data through the trained semantic segmentation and change detection classification models.

[0032] By inputting the two registered Sentinel-2 remote sensing images of two periods into the trained semantic segmentation and change detection classification models respectively, four types of probability maps are obtained, namely buildings, bare land, water bodies, and green spaces.

[0033] Using the DeeplabV3+ framework, the semantic segmentation and change detection classification model includes four encoders, namely Block1, Block2, Block3, and Block4, a spatio-temporal feature fusion module, and a decoder.

[0034] Taking the improved ResNet-50 as the Block of the encoder and loading the weights pre-trained on the ImageNet dataset to accelerate model convergence. Fix the weights of the first three convolutional layers of the ResNet-50 network, i.e., Conv1~Conv3, and only fine-tune the deep network, i.e., Conv4 and subsequent modules, to avoid damaging the shallow general features and reduce the risk of overfitting at the same time.

[0035] The improved ResNet-50 is specifically as follows: After the output of the fourth convolutional layer of ResNet-50, i.e., Conv4, add 4 groups of parallel dilated convolutional layers with dilation rates of 6, 12, 18, and 24 respectively to capture the context information of remote sensing data at different scales and obtain multi-scale feature maps; concatenate the multi-scale feature maps with the original features output by Conv4, and then compress the number of channels to 256 through 1×1 convolution to reduce the computational amount; introduce a channel attention module in the skip connection to automatically learn the importance weights of the shallow features of remote sensing data, such as building edges and water body boundaries, and suppress background noises, such as cloud shadows and seasonal changes of vegetation.

[0036] The steps of the spatio-temporal feature fusion module are specifically as follows: Difference feature extraction: For the dual-temporal feature maps output by the encoder, calculate the absolute difference pixel by pixel to generate an initial difference feature map. After the difference feature map is spatially smoothed by a 3×3 convolutional kernel, an initial change feature with a channel dimension of 256 is output; Temporal modeling: Use a bidirectional ConvLSTM unit to model the initial change features, capture the progressive process of ground object changes, such as vegetation degradation, and mutation events, such as building demolition, to obtain a temporal feature map. The hidden layer dimension of the ConvLSTM is set to 512, the time step is set to 2, and the convolutional kernel size is 3×3. The forward and backward LSTMs respectively learn the temporal dependence relationships, and finally concatenate the bidirectional hidden states into a 1024-dimensional feature; Attention-weighted fusion: Dynamically calibrate the fused temporal features based on the channel attention mechanism. Specifically, first, compress the spatial dimension of the temporal feature map through global average pooling to obtain channel-level statistics; then use two fully connected layers to generate a channel weight vector. Multiply the weight vector with the initial change features channel by channel to strengthen the response intensity of the key feature channels and suppress the interference of non-significant features. The weighted feature map is compressed to 256 channels through 1×1 convolution to reduce the computational complexity, and a residual connection is made with the original features output by the encoder, i.e., Block4, to retain the spatial context information of the unchanged regions.

[0037] The decoder steps are specifically as follows: The decoder adopts a four-level upsampling structure to gradually restore the spatial resolution: In the first level, the input features (H / 32×W / 32) are upsampled to H / 16×W / 16 through a 3×3 transposed convolution (stride 2, padding 1). After outputting 512 channels, the 1024-dimensional features of encoder Block3 are fused, and the channels are aligned and added through a 1×1 convolution; in the second level, the transposed convolution operation is repeated to output 256-channel features of H / 8×W / 8, and the Block2 features are fused; in the third level, it is upsampled to 128 channels of H / 4×W / 4, and the shallow texture features of Block1 are fused; in the fourth level, it is finally restored to 64-channel features of the original resolution H×W, and bilinear interpolation is superimposed to refine the local features to ensure the integrity of small-scale features.

[0038] Binary mask generation: The 64-dimensional features output by the decoder are compressed to 4 channels, namely buildings, bare land, water bodies, and green spaces, through a 1×1 convolution, and a probability map is generated through the Sigmoid function and processed using a dynamic threshold segmentation strategy. The dynamic threshold segmentation strategy is specifically as follows: The thresholds for buildings and bare land are set to 0.6 to reduce misclassification of vegetation, and the thresholds for water bodies and green spaces are set to 0.4 to improve sensitivity; in the post-processing stage, salt-and-pepper noise is removed through a 5×5 morphological opening operation.

[0039] Confidence estimation. The confidence is jointly calculated by information entropy and spatial consistency, and the formula is as follows:

[0040] In the formula, H is the entropy value, which measures the uncertainty of the classification result. The lower the entropy value, the higher the classification certainty; represents the probability that a pixel belongs to the i-th type of feature (buildings, bare land, water bodies, green spaces).

[0041] When training the semantic segmentation and change detection classification model, the cross-entropy loss function is used to handle the class imbalance problem, and higher weights are assigned to minority classes such as water bodies. For example, the weight of water bodies is 1.5, the weight of bare land is 1.2, and the weight of green spaces is 0.9. The Dice loss is used to optimize the overlap rate of the segmented regions and improve the boundary accuracy. The weight ratio with the cross-entropy loss is 4:6. The Adam optimizer is used, with β1 = 0.9, β2 = 0.999, the initial learning rate is 0.0001, and the cosine annealing strategy is used to dynamically adjust the learning rate, which decays to 10% of the initial value every 50 epochs.

[0042] Step 112: Compare and mark pixel by pixel the pixels with inconsistent classifications in the classification results of remote sensing data in different phases to generate an original difference map, and calculate the conversion area between the feature classes in the target area according to the original difference map.

[0043] Specifically, compare and mark pixel by pixel the pixels with inconsistent categories in the two-phase classification results to generate an original difference map; construct a 4×4 land cover conversion matrix, count the mutual conversions between the four land cover categories, such as "bare land → building", "green land → bare land", etc., and calculate the area and proportion of each land cover category.

[0044] An example of the land cover conversion matrix is:

[0045] In the table, M i,j represents the number of pixels converted from the previous-phase category i to the next-phase category j.

[0046] Step 12: Calculate the conversion heat contribution degree of different land cover categories based on the net change in conversion temperature between land cover categories and the conversion area between land cover analogies.

[0047] Based on the thermal infrared bands (Band10 - 11) of Landsat8, perform land surface temperature inversion, and use the radiative transfer equation RTE to invert the land surface temperature LST. The formula is as follows:

[0048] In the formula, .

[0049] Extract the average daily temperature (T 1 , T 2 ) of the two scenes of images from the meteorological data network, and calculate the background temperature difference: .

[0050] Based on the land cover conversion matrix, calculate the average value of temperature change for each conversion type: ； Eliminate meteorological interference and obtain the net heat effect value. The net temperature change of land cover conversion is: , and calculate the net temperature change of the four land cover types respectively.

[0051] The formula for calculating the conversion heat contribution degree of land cover categories is:

[0052] Among them, is the heat contribution degree of land cover category i , is the net temperature change when land cover category j changes to land cover category i , is the conversion area when land cover category j changes to land cover category i .

[0053] For example, on August 15, 2017 in Shenyang , August 18, 2020 , .

[0054] Step 2: Perform hierarchical normalization processing on multi-dimensional indicators to obtain the normalized spatial distribution maps of each indicator in the target area, and assign combined weights to the relative importance among multi-dimensional indicators based on game theory methods; according to the conversion heat contribution degrees of different land cover types, decompose the combined weights of surface temperature to different land covers in the target area to obtain the comprehensive weights of different land covers in the target area; correspond the normalized spatial distribution map corresponding to the surface temperature with different land covers in the target area to obtain the heat contribution normalized distribution map of different land covers in the target area.

[0055] Step 21: Classify the multi-dimensional indicators respectively using the natural breaks method, and perform normalization processing on the classification results to obtain the normalized spatial distribution maps of each indicator in the target area.

[0056] Classify the multi-dimensional indicators for urban park site selection, namely surface temperature, NDVI, traffic accessibility, and night light index, respectively using the natural breaks method. For example: the surface temperature is divided into 5 levels according to the severity of the heat island, 1 = very unsuitable → 5 = very suitable; NDVI is divided into 5 levels according to the vegetation scarcity degree, 1 = very suitable → 5 = very unsuitable; traffic accessibility is divided into 5 levels according to the numerical value, 1 = very unsuitable → 5 = very suitable; the night light index is divided into 5 levels according to the numerical value, 1 = very unsuitable → 5 = very suitable.

[0057] Normalize the classification results to 0-1 scores. For example, level 1 = 0.2, level 2 = 0.4,..., level 5 = 1.0.

[0058] Step 22: Use the analytic hierarchy process to determine the subjective weights of multi-dimensional indicators, use the independent weight coefficient method to determine the objective weights of multi-dimensional indicators, and take the minimization of the difference between subjective weights and objective weights as the goal, and use the Nash equilibrium solution to obtain the combined weights of multi-dimensional indicators.

[0059] As Figure 3 shown, using the analytic hierarchy process AHP, through experts' pairwise comparison of the relative importance between indicators, constructing a judgment matrix by quantitative scoring based on the 1-9 scale method, obtaining the subjective weights of the four indicators, and passing the consistency test, CR < 0.1; Using the independent weight coefficient method IWCM, based on the information entropy to calculate the index dispersion to determine the objective weights of multi-dimensional indicators, the formula is:

[0060] In the formula, the j objective weight of the th jThe information entropy of the item index is the coefficient of variation The sum of the coefficients of variation of all indicators

[0061] With the goal of minimizing the difference between subjective weights and objective weights, the combined weights of multi-dimensional indicators are obtained by solving the Nash equilibrium. The formula is as follows

[0062]

[0063] Among them is the subjective weight is the objective weight , is the optimal coefficient is the combined weight

[0064] The analytic hierarchy process is used to determine the subjective weights of multi-dimensional indicators, and the independent weight coefficient method is combined to quantify the objective weights. The collaborative relationship between the two types of weights is optimized through game theory, so that the weight assignment of the four core indicators of land surface temperature, vegetation coverage, accessibility, and night light index can adapt to the spatial heterogeneity characteristics of different regions

[0065] Step 23: Decompose the comprehensive weight of the land surface temperature to each land use type, including buildings, bare land, water bodies, and green spaces, to achieve spatially heterogeneous weighting. Specifically According to the heat contribution degree of the four types of land cover (buildings, bare land, water bodies, and green spaces) calculated in step 1 The comprehensive weight of the land surface temperature is proportionally distributed to each land use type

[0066] In the formula is the comprehensive weight of the four types of land cover after decomposition, and the subscript takes , is the combined weight of the land surface temperature is the heat contribution degree of the four types of land cover i is from 1 to 4

[0067] Step 24: Split the normalized distribution map of the land surface temperature obtained in step 21 into four types of land cover according to the four types of probability maps at the later time phase obtained in step 111, that is, use the spatial distribution of the four types of land cover at the later time phase to clip the normalized distribution map of the land surface temperature to obtain the normalized distribution map of the heat contribution corresponding to the four types of land cover

[0068] Step 3: Integrate the comprehensive weights of different land features and the normalized distribution maps of the thermal contributions of corresponding land features, as well as the combined weights of each index and the normalized spatial distribution maps corresponding to each index, to generate a suitability distribution map for the target area site selection. Extract the hotspots of suitability, construct an OD cost matrix with residential area POIs as the starting point and the hotspots as the ending point, and screen candidate points according to the time cost to obtain the urban park site selection plan.

[0069] Overlay the surface temperature, vegetation scarcity index, night light index, and resident accessibility index of four types of land features to generate a suitability distribution map for site selection, and screen candidate points in combination with road network topological constraints and non-hardened surface protection rules; through the analysis of the potential for heat island mitigation and the blind areas of public service coverage, output the optimal park site selection plan.

[0070] Specifically, integrate 7 indicators including the surface temperature, NDVI, night light index, and accessibility of four types of land features, and perform weighted overlay to output the suitability spatial distribution map of the target area with a resolution of 30 meters. The formula is:

[0071] In the formula, represents the comprehensive weight of the thermal contribution of the building land type, represents the comprehensive weight of the thermal contribution of the bare land type, represents the comprehensive weight of the thermal contribution of the green land type, represents the comprehensive weight of the thermal contribution of the water body land type; represents the combined weight of the index NDVI; represents the combined weight of the index night light index; represents the combined weight of the index accessibility; is the normalized spatial distribution map of the thermal contribution of the building land type, is the normalized spatial distribution map of the thermal contribution of the bare land type, is the normalized spatial distribution map of the thermal contribution of the green land type, is the normalized spatial distribution map of the thermal contribution of the water body land type, is the NDVI normalized spatial distribution map, is the night light index normalized spatial distribution map, is the accessibility normalized spatial distribution map; S represents the suitability spatial distribution map for urban park site selection.

[0072] Use Getis-Ord Gi* hotspot analysis to extract the significant regions of heat island intensity and vegetation scarcity, with Z-score > 2.58 and confidence level > 99%.

[0073] Exclude existing building areas and permanent water bodies based on OpenStreetMap building and water surface vector data.

[0074] Construct an OD cost matrix with residential area POIs as the starting points and hotspots as the ending points. Use an isochrone such as a 15-minute walk as the time cost, and overlay water body blockage and terrain impedance to generate an accessibility distribution map. Specifically: Data preparation and processing: Obtain residential area POI point data, vector boundaries of suitable hotspots, road network data, and high-precision water system layers; Establish traffic rules: Use the transportation road network as the path and set water body areas as non-traversable areas; Generate cost rasters: Integrate road network distribution and water body blockage factors to construct comprehensive traffic cost raster data; Calculate the shortest path: Use a path planning algorithm, with residential areas as the starting points and suitable hotspots as the ending points, iteratively calculate the minimum cumulative travel time, and generate an accessibility distribution map; Extract isochrones: Segment the accessibility raster based on a 15-minute walk threshold to generate a continuous isochrone area; Construct an OD matrix: Spatially connect residential area POIs and suitable hotspots, establish a start-end accessibility time relationship table, and output the candidate point locations and path topological structures.

[0075] Screen candidate points that simultaneously meet the conditions of "suitability > set value such as 0.8", "accessibility time < set time such as 15 minutes", and "proportion of non-hardened surfaces such as bare land / green land > 70%"; Sort in descending order according to the suitability score and output the coordinates and attribute tables of the top-ranked candidate points such as the top 15.

[0076] In this embodiment, by fusing Sentinel-2 high-resolution multispectral data and the thermal infrared band of Landsat 8, a sub-pixel geometric registration framework is constructed, overcoming the problems of insufficient resolution of traditional single data sources and rough land cover classification. When using the radiative transfer equation to invert the land surface temperature, meteorological parameters are innovatively introduced for dynamic correction to strip the interference of background air temperature fluctuations on the attribution of the urban heat island effect, and achieve accurate mapping of land cover types and temperature changes. Through a semantic segmentation and change detection classification model, combined with multi-scale dilated convolution and attention mechanisms to enhance feature extraction capabilities, the recognition accuracy of the boundaries of key land covers such as buildings and bare land is significantly improved. A bi-temporal land cover conversion matrix is constructed to analyze the difference in thermal contribution, providing a spatially explicit quantitative basis for tracing the source of the urban heat island effect.

[0077] The method for dynamically assigning weights to multi-dimensional indicators for urban park location selection based on game theory coupling breaks through the limitations of traditional subjective weight assignment or single objective weight assignment. By integrating the analytic hierarchy process and the independent weight coefficient method through the Nash equilibrium theory, the weight distribution of core indicators such as temperature and vegetation index is both scientific and adaptable. An innovative weight decomposition mechanism for land surface temperature is proposed, and the influence weights of types such as buildings and green spaces in the global evaluation are dynamically adjusted according to the heat contribution degree of different land type conversions, avoiding the sensitivity loss of high heat source areas caused by "one-size-fits-all" weight assignment. This embodiment realizes the local optimization of weight parameters through spatial heterogeneity analysis, ensuring that the park location selection decision can target and alleviate the core area of the heat island and adapt to the diversity characteristics of the urban spatial pattern.

[0078] The park location selection decision-making method based on multi-constraint network analysis in this embodiment constructs a multi-objective collaborative optimization framework of "heat environment governance - ecological service supply and demand - resident accessibility", uses spatial statistical methods to identify suitable hot spots for park location selection, and combines road network topology analysis and walking isochrone modeling. Through multi-dimensional indicator weighted overlay and spatial exclusion rules such as non-hardened surface protection and avoidance of existing construction, the optimal solution set that simultaneously meets the heat island mitigation potential, ecological restoration needs, and public service fairness is screened out. This method breaks through the limitations of traditional single-objective optimization and realizes the deep coupling of ecological benefits and people's livelihood needs through network analysis technology, providing an operable decision-making path for the scientific layout of parks in high-density urban environments.

[0079] Embodiment 2 The purpose of this embodiment is to provide an urban park location selection decision-making system, including: A calculation module, which is configured to: obtain multi-temporal remote sensing data of the target area, count the mutual conversion information of land cover types between the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution degree of different land cover types; A weight module, which is configured to: perform hierarchical normalization processing on multi-dimensional indicators to obtain the normalized spatial distribution map of each indicator on the target area, and assign combined weights to the relative importance between multi-dimensional indicators based on game theory methods; according to the conversion heat contribution degree of different land cover types, decompose the combined weight of land surface temperature to different land cover types in the target area to obtain the comprehensive weights of different land cover types in the target area; An allocation module, which is configured to: correspond the normalized spatial distribution map corresponding to the land surface temperature to different land cover types in the target area to obtain the normalized distribution map of heat contribution of different land cover types in the target area; A site selection module, which is configured to: fuse the comprehensive weights of different features, the normalized distribution maps of the thermal contributions of the corresponding features, the combined weights of each index, and the normalized spatial distribution maps corresponding to each index to generate a site selection suitability distribution map for the target area, extract the hot spots of suitability, construct an OD cost matrix with the residential area POI as the starting point and the hot spots as the end point, and screen candidate points according to the time cost to obtain a site selection plan for urban parks.

[0080] In more embodiments, there is also provided: An electronic device includes a memory, a processor, and computer instructions stored on the memory and running on the processor. When the computer instructions are run by the processor, the method described in Embodiment 1 is completed. For the sake of brevity, it will not be elaborated here.

[0081] It should be understood that in this embodiment, the processor may be a central processing unit CPU, and the processor may also be other general-purpose processors, digital signal processors DSP, application-specific integrated circuits ASIC, off-the-shelf programmable gate arrays FPGA, or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0082] The memory may include a read-only memory and a random access memory, and provide instructions and data to the processor. A part of the memory may also include a non-volatile random access memory. For example, the memory may also store information about the device type.

[0083] A computer-readable storage medium is used to store computer instructions. When the computer instructions are executed by the processor, the method described in Embodiment 1 is completed.

[0084] The method in Embodiment 1 can be directly implemented by a hardware processor, or by a combination of hardware and software modules in the processor. The software modules may be located in mature storage media in the art such as random access memory, flash memory, read-only memory, programmable read-only memory, or electrically erasable programmable memory, registers, etc. This storage medium is located in the memory, and the processor reads the information in the memory and combines its hardware to complete the steps of the above method. To avoid repetition, it will not be described in detail here.

[0085] Those of ordinary skill in the art can realize that the units and algorithm steps of each example described in combination with this embodiment can be implemented by electronic hardware or a combination of computer software and electronic hardware. Whether these functions are executed in a hardware or software manner depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods for each specific application to implement the described functions, but this implementation should not be considered to exceed the scope of this application.

[0086] Although the specific embodiments of the present invention have been described above in conjunction with the accompanying drawings, it is not a limitation on the protection scope of the present invention. Those skilled in the art should understand that various modifications or deformations that can be made without creative efforts on the basis of the technical solutions of the present invention are still within the protection scope of the present invention.

Claims

1. A method for urban park site selection decision-making, characterized in that: include: Obtain multi-temporal remote sensing data of the target area, count the mutual conversion information of the object categories between the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution of different object categories; The multi-dimensional indicators are graded and normalized to obtain the normalized spatial distribution map of each indicator in the target area, and the relative importance of the multi-dimensional indicators is assigned a combined weight based on the game theory method; according to the conversion heat contribution of different types of land features, the combined weight of the surface temperature is decomposed to different land features in the target area to obtain the comprehensive weight of different land features in the target area; The normalized spatial distribution map corresponding to the surface temperature is matched with different objects in the target area to obtain the normalized distribution map of heat contribution of different objects in the target area; The comprehensive weights of different landforms and the normalized distribution map of thermal contribution of the corresponding landforms, as well as the combined weights of each indicator and the normalized spatial distribution map corresponding to each indicator are integrated to generate a distribution map of the suitability of the target area. The suitability hotspot areas are extracted, and an OD cost matrix is ​​constructed with the residential POI as the starting point and the hotspot area as the end point. The candidate points are screened according to the time cost to obtain the urban park site selection plan.

2. A method for urban park site selection decision-making as claimed in claim 1, characterized in that: The mutual conversion information of the object categories between the multi-temporal remote sensing data of the target area is counted, and the conversion heat contribution of different object categories is calculated, specifically: Statistics are collected on the conversion information between the multi-temporal remote sensing data of the target area, and the conversion area between the target area and the net change in temperature between the conversion of the target area and the target area; Based on the net change in conversion temperature between feature categories and the conversion area between feature analogies, the conversion heat contribution of different feature categories is calculated.

3. A method for urban park site selection decision-making as claimed in claim 2, characterized in that: The conversion information of the object categories between the multi-temporal remote sensing data of the target area is counted, and the conversion area between the object categories in the target area is calculated, specifically: Based on the multi-temporal remote sensing data of the target area, the probability map of the ground object category corresponding to the remote sensing data of different temporal phases is obtained through the trained semantic segmentation and change detection classification models respectively; Compare and mark the pixels with inconsistent analogy in the classification results of remote sensing data of different phases pixel by pixel to generate the original difference map; The conversion area between the object categories in the target area is calculated according to the original difference map.

4. A method for urban park site selection decision-making as claimed in claim 3, characterized in that: The semantic segmentation and change detection classification model includes an encoder, a spatiotemporal feature fusion module and a decoder. The encoder adopts an improved ResNet-50 network. The improved ResNet-50 network is specifically as follows: multiple groups of parallel hole convolutions are added after the fourth convolution layer of the ResNet-50 network to capture context information of different scales of remote sensing data and obtain a multi-scale feature map; the multi-scale feature map is spliced ​​with the output features of the fourth convolution layer to obtain a spliced ​​feature; channel attention is introduced in the jump connection to learn the shallow features of the remote sensing data; Based on the spatiotemporal feature fusion module, the initial change features are calculated for the dual-phase feature map output by the encoder, and the initial change features are modeled using a bidirectional ConvLSTM unit to obtain a time series feature map, and the time series feature map is dynamically calibrated using a channel attention mechanism.

5. The method for urban park site selection decision-making according to claim 1, characterized in that: The multi-dimensional indicators are graded and normalized to obtain the normalized spatial distribution map of each indicator in the target area. Based on the game theory method, the relative importance of the multi-dimensional indicators is assigned a combined weight, which is as follows: The multi-dimensional indicators of the target area are graded using the natural breakpoint method, and the classification results are normalized to obtain the normalized spatial distribution map of each indicator in the target area; The hierarchical analysis method is used to determine the subjective weights of multidimensional indicators, and the independent weight coefficient method is used to determine the objective weights of multidimensional indicators. With the goal of minimizing the difference between subjective weights and objective weights, the Nash equilibrium solution is used to obtain the combined weights of multidimensional indicators.

6. A method for urban park site selection decision-making as claimed in claim 5, characterized in that: The comprehensive weights of different landforms and the normalized distribution map of heat contribution of the corresponding landforms, as well as the combined weights of each indicator and the normalized spatial distribution map corresponding to each indicator are integrated to generate the distribution map of site suitability of the target area, specifically: ， in, Represents the comprehensive weight of the building ground heat contribution, represents the comprehensive weight of the heat contribution of bare land, Represents the comprehensive weight of the heat contribution of green land, represents the comprehensive weight of the thermal contribution of water bodies; The combined weight of the representative indicator NDVI; The combined weight of the night light index of the representative indicator; represents the combined weight of accessibility; Normalized spatial distribution map of building ground heat contribution, Normalized spatial distribution map of bare ground heat contribution, Normalized spatial distribution map of green space land heat contribution, Normalized spatial distribution map of ground heat contribution of water bodies, is the normalized spatial distribution map of NDVI, is the normalized spatial distribution map of night light index, is the normalized spatial distribution map of accessibility; S represents the spatial distribution map of urban park site suitability.

7. The method for urban park site selection decision-making according to claim 1, characterized in that: According to the suitability distribution map of the target area, the suitability hotspot area is extracted, and the OD cost matrix is ​​constructed with the residential area POI as the starting point and the hotspot area as the end point. The candidate points are screened according to the time cost to obtain the urban park site selection plan, which is as follows: Getis-Ord Gi* hotspot analysis was used to identify areas with significant heat island intensity and vegetation deficiency. Exclude existing built-up areas and permanent water bodies based on OSM building vector data; Taking the residential area as the starting point and the urban park hotspot area as the end point, an OD cost matrix is ​​constructed to calculate the walking cost from the residential area to the hotspot area; Screen candidate points whose suitability and reachability time meet the preset conditions and whose surfaces are not hardened; The candidate points are ranked according to their heat island mitigation potential and coverage of residents' service blind spots to obtain the urban park site selection plan.

8. An urban park site selection decision system, characterized in that: include: A calculation module is configured to: obtain multi-temporal remote sensing data of a target area, count the mutual conversion information of ground object categories between the multi-temporal remote sensing data of the target area, and calculate the conversion heat contribution of different ground object categories; The weight module is configured to: perform hierarchical normalization processing on the multi-dimensional indicators to obtain the normalized spatial distribution map of each indicator in the target area, and assign combined weights to the relative importance of the multi-dimensional indicators based on the game theory method; decompose the combined weights of the surface temperature to different ground objects in the target area according to the conversion heat contribution of different ground object categories, and obtain the comprehensive weights of different ground objects in the target area; An allocation module is configured to: correspond the normalized spatial distribution map corresponding to the surface temperature to different land objects in the target area to obtain a normalized distribution map of heat contribution of different land objects in the target area; The site selection module is configured to: fuse the comprehensive weights of different landforms and the normalized distribution map of thermal contributions of the corresponding landforms, as well as the combined weights of each indicator and the normalized spatial distribution map corresponding to each indicator, generate a site selection suitability distribution map for the target area, extract suitable hot spots, construct an OD cost matrix with residential POI as the starting point and the hot spots as the end point, and screen candidate points according to time cost to obtain an urban park site selection plan.

9. An electronic device, characterized in that: The method comprises a memory and a processor and computer instructions stored in the memory and executed on the processor, wherein when the computer instructions are executed by the processor, the method according to any one of claims 1 to 7 is completed.

10. A computer-readable storage medium, characterized in that: Used to store computer instructions, which, when executed by a processor, complete the method described in any one of claims 1 to 7.

Images