Urban green view factor spatial distribution simulation method based on green facility environment characteristic framework

Abstract

The present application relates to the technical field of data processing, and more particularly to a city green view rate spatial distribution simulation method based on a green facility environment feature framework, comprising: acquiring city-wide street view images, satellite remote sensing images, road network distribution data and land use type grids, marking vegetation coverage areas and generating an initial green view rate grid layer, performing semantic segmentation on the street view images to calculate the proportion of vegetation pixels to generate green view rate observation values of road network areas. When constructing a multi-scale environment feature framework, the road network weight parameters and the land use type category coding parameters are merged to form an environment parameter index. Then, based on the framework, a spatial interpolator is constructed, and the observation values and the environment parameter index are jointly input to complete the green view rate calculation of non-road areas, and the global interpolation result is generated. The present application breaks through the spatial limitations of existing street view sampling, realizes global coverage using public geographic data, reduces the implementation threshold of small and medium-sized cities, and provides a scientific decision basis for heat island effect analysis and green land planning.

Description

Technical Field

[0001] This invention relates to the field of data processing technology, and in particular to a method for simulating the spatial distribution of urban green visibility based on a framework of green facility environmental characteristics. Background Technology

[0002] Urban green view rate refers to a comprehensive indicator of the proportion of natural vegetation visible within the pedestrian's field of vision. Its spatial distribution is characterized by significant unevenness, mainly due to land use limitations caused by high building density in urban core areas, as well as differences in land use types. In central business districts and transportation hubs, high-intensity development compresses green space resources, resulting in generally low green view rates; while residential areas and suburbs have relatively high green view rates due to more reserved green space in planning, and peripheral areas can even reach peak levels due to abundant ecological land.

[0003] Existing simulations of urban green view rate spatial distribution based on the environmental characteristics of green facilities suffer from technical pain points such as limited coverage and high data acquisition barriers, hindering precise assessment and scientific decision-making. Specifically, relying on street view imagery sampling points requires commercial APIs or street view vehicle deployment, which is costly and limited in implementation, especially in old urban areas or remote areas where imagery is lacking, creating sampling blind spots. View analysis methods require high-precision 3D data such as DEM and vegetation height, and acquisition requires advanced equipment such as LiDAR, which is difficult to obtain in small and medium-sized cities due to insufficient resources, resulting in low model extrapolation accuracy. For example, in the application in the central urban area of ​​Xi'an, existing methods can only generate sampling points along the road network. Green view rate data is lacking in non-road areas such as inside parks or residential areas, affecting the formulation of heat island effect mitigation strategies, failing to meet the needs of green equity assessment across the entire region, and hindering the scientific decision-making of urban planning departments to optimize green space layout. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides a method for simulating the spatial distribution of urban green view rate based on a framework of green facility environmental characteristics. This invention solves the technical problems that arise from the numerous limitations and difficulties in obtaining street view images, as well as the high data requirements and acquisition difficulties of existing methods, which result in difficulties in obtaining green view rate in some urban areas, limited coverage, and inability to support detailed assessment and scientific decision-making.

[0005] To solve the above-mentioned technical problems, the specific contents of the present invention are as follows:

[0006] The present invention provides a method for simulating the spatial distribution of urban green view rate based on a framework of green facility environmental characteristics, comprising:

[0007] Step 1: Obtain the geospatial database of the entire city. The geospatial database includes satellite remote sensing imagery, road network distribution data, land use type raster, urban administrative division vector data, and digital elevation model data.

[0008] Step 2: Process satellite remote sensing images using the normalized vegetation index algorithm to mark vegetation coverage areas; generate street view sampling points along the road network at preset intervals, acquire multi-directional street view images of each sampling point, perform semantic segmentation on the street view images to calculate the proportion of vegetation pixels, and generate green view rate observation values ​​for the road network area.

[0009] Step 3: Perform density quantization processing on the road network distribution data to generate road network weight parameters. At the same time, convert the land use type raster into category coding parameters. Merge the road network weight parameters and the category coding parameters to generate an environmental parameter index. Embed the environmental parameter index into the spatial structure hierarchy to construct a multi-scale environmental feature framework.

[0010] Step 4: Construct a spatial interpolator based on the multi-scale environmental feature framework, and input the observed green visibility rate and the environmental parameter index into the spatial interpolator to complete the numerical calculation of the green visibility rate in non-road areas under the framework constraints, and generate global interpolation results;

[0011] Step 5: Perform rasterization processing on the global interpolation results to generate a green view rate raster representing the distribution of green view rate in non-road areas. Spatially overlay the green view rate raster with the urban administrative division vector data to output a spatial distribution map of urban green view rate covering the entire urban area.

[0012] Furthermore, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework of the present invention includes the following in step 1:

[0013] The road network distribution data uses open street map data as the basis for road network density quantification.

[0014] The land use type grid is generated by performing a vectorization transformation on the government's publicly available urban planning map. Industrial land is assigned a code value of 1 to identify the clustering characteristics of production facilities, residential land is assigned a code value of 2 to characterize the distribution characteristics of residences, and green space is assigned a code value of 3 to mark the vegetation coverage space. The code values ​​are used to constrain the differences in land use type zoning in the environmental parameter index.

[0015] Furthermore, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework described in this invention further includes, in step 1:

[0016] Sensor radiometric calibration is performed on satellite remote sensing images, atmospheric correction is performed using dark pixel subtraction, the target spatial resolution is determined, and all images are resampled to that resolution using the nearest neighbor method.

[0017] Furthermore, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework described in this invention includes the following in step 2:

[0018] The normalized vegetation index pixel value calculated by the normalized vegetation index algorithm is compared with a preset first threshold.

[0019] Normalized vegetation index pixel values ​​that are higher than the preset first threshold are marked as vegetation cover pixels, and a vegetation distribution binary raster is generated to identify the spatial location of vegetation.

[0020] Based on open street map data processed by topology inspection, a road network vector boundary is generated. The road network vector boundary is used to define the spatial range of vegetation clipping. The road network vector boundary serves as the study area. Using a vector boundary clipping tool in a preset geographic information system, spatial clipping is performed on the binary raster of vegetation distribution according to the road network vector boundary. Vegetation marker data outside the road network coverage area is removed, and vegetation coverage information inside the road boundary is extracted. The road network vector boundary provides a basis for defining the spatial range of the road area, supports the spatial positioning of subsequent street view sampling point deployment, and is used for the construction of a multi-scale environmental feature framework.

[0021] The preset first threshold is determined by calculating the normalized vegetation index value of each pixel using the normalized vegetation index algorithm on the satellite remote sensing image obtained in step 1, and performing K-means cluster analysis on all normalized vegetation index values. The number of clusters is set to 2 to correspond to the two land cover categories of vegetation and non-vegetation, which is used to distinguish between the two types of land cover. When the normalized vegetation index pixel value is higher than the preset first threshold, it is marked as a vegetation-covered pixel, and when the normalized vegetation index pixel value is lower than the preset first threshold, it is marked as a non-vegetation pixel.

[0022] The preset first threshold is determined by the following method: K-means clustering analysis is performed on the normalized vegetation index (NDI) values ​​of all pixels in the satellite remote sensing image obtained in step 1, with the number of clusters set to 2. A machine learning algorithm is used to automatically distinguish between vegetation and non-vegetation features, and the optimal threshold is determined based on the cluster centers. Healthy, dense vegetation has significantly higher reflectance in the near-infrared band than in the visible red band, resulting in a generally higher NDI value. Conversely, the NDI values ​​of non-vegetation features such as bare soil, water bodies, and buildings are generally lower. This K-means clustering method can adapt to the differences in spectral characteristics between vegetation and non-vegetation in different regions and at different times, overcoming the limitations of applying a fixed threshold in different spatial units. The specific value of the preset first threshold can also be dynamically optimized and adjusted by calling a historical vegetation sample database based on the regional vegetation type distribution characteristics. Combining this with the K-means clustering results enhances the accuracy and robustness of vegetation identification in complex urban environments, thereby achieving reliable differentiation between vegetation and non-vegetation areas.

[0023] Furthermore, the urban green view rate spatial distribution simulation method based on the green facility environmental feature framework described in this invention, in step 3, constructs a multi-scale environmental feature framework including:

[0024] Slope analysis is performed on the digital elevation model data obtained in step 1 to generate a terrain relief factor; the terrain relief factor is then aligned and superimposed onto the land use type raster according to the coordinate space.

[0025] The superimposed composite parameters are used as the core constraint terms of the environmental parameter index.

[0026] Furthermore, in the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework described in this invention, step 4 includes:

[0027] Use the green visibility observations generated in step 2 as input points for the Kriging interpolation algorithm;

[0028] The terrain relief factor and the land use type raster are used as spatial covariates;

[0029] Kriging interpolation is performed to generate a green visibility prediction surface for non-road areas, with a resolution of 30m×30m grid.

[0030] Furthermore, in the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework described in this invention, step 4 includes:

[0031] Obtain publicly available building survey data from urban planning departments;

[0032] Extract the building density values ​​of old urban areas from the building census data to generate a building density weighting factor;

[0033] The building density weighting factor is used as an additional covariate input to the spatial interpolator to adjust the numerical calculation results of the green view rate in old urban areas.

[0034] Furthermore, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework described in this invention further includes step 6, which includes:

[0035] Collect temperature monitoring data from the weather station;

[0036] Inverse distance weighted interpolation is performed on the temperature monitoring data to generate heat map data;

[0037] The heat map data is overlaid with the urban green visibility spatial distribution map output in step 5;

[0038] Based on the overlay results, the correlation coefficient between green view rate and temperature for each administrative region is calculated, and a spatial coupling analysis report on urban heat island effect and green view rate is generated.

[0039] Furthermore, in the urban green view rate spatial distribution simulation method based on the environmental characteristics framework of green facilities described in this invention, step 6 further includes:

[0040] Obtain the daily average temperature data monitored by the weather station within a preset time period, set the search radius parameter of the inverse distance weighted interpolation algorithm to 5 kilometers, and perform raster interpolation to output a temperature raster layer with a 1km×1km grid.

[0041] Furthermore, in the urban green view rate spatial distribution simulation method based on the environmental characteristics framework of green facilities described in this invention, step 6 further includes:

[0042] Read the urban green view rate spatial distribution map output in step 5, and determine the preset second threshold by performing K-means cluster analysis on the green view rate values ​​of all grid cells in the green view rate value grid generated in step 5. Set the number of clusters to 2 to distinguish between low green view rate and high green view rate areas, and identify grid cells in all administrative regions whose green view rate is lower than the preset second threshold.

[0043] Areas with continuous low green visibility are marked as green spaces to be optimized. The location data and population density data of the green spaces to be optimized are associated to generate a green space planning decision suggestion table with priority ranking.

[0044] Beneficial effects of this invention;

[0045] This invention replaces street view image sampling with full coverage of satellite remote sensing imagery, avoiding limitations imposed by commercial API calls and street view vehicle deployment, and solving the problem of data blind spots in areas outside roads. It constructs a multi-scale environmental feature framework integrating road network weights and land use type parameters, and combines this with the Kriging spatial interpolation algorithm to achieve accurate prediction of green visibility in non-road areas with the support of publicly available data. It generates a full-area distribution map to support heat island effect analysis and green space planning decisions, lowering the implementation threshold for small and medium-sized cities, improving the ability to assess the fairness of green exposure, and providing urban planning departments with a comprehensive scientific decision-making basis. Attached Figure Description

[0046] To more clearly illustrate the technical solution of the present invention, the drawings used in the embodiments will be briefly introduced below. Obviously, for those skilled in the art, other drawings can be obtained based on the drawings without creative effort.

[0047] Figure 1 A flowchart illustrating the spatial distribution simulation method of urban green visibility rate based on the environmental characteristics framework of green facilities, provided in an embodiment of the present invention.

[0048] Figure 2 This is a street view sampling point distribution map of the method of the present invention.

[0049] Figure 3 This is a spatial distribution map of the environmental characteristic framework indicators of green facilities in an embodiment of the method of the present invention.

[0050] Figure 4 This is a scatter plot showing the actual green visibility rate and the predicted value from the model simulation in an embodiment of the method of the present invention.

[0051] Figure 5 This is a simulation result diagram of green visibility rate in an embodiment of the method of the present invention. Detailed Implementation

[0052] To make the technical solution of the present invention clearer, the present invention will be clearly and completely described below with reference to specific embodiments and corresponding drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention. The present invention provided by various embodiments will be described in detail below with reference to the accompanying drawings. To better understand the purpose of the present invention, the present invention will be described in further detail below.

[0053] Please see Figures 1 to 5 The present invention provides a method for simulating the spatial distribution of urban green visibility rate based on a framework of green facility environmental characteristics, comprising:

[0054] Step 1: Obtain the geospatial database of the entire city. The geospatial database includes satellite remote sensing imagery, road network distribution data, land use type raster, urban administrative division vector data, and digital elevation model data.

[0055] Step 2: Process satellite remote sensing images using the normalized vegetation index algorithm to mark vegetation coverage areas; generate street view sampling points along the road network at preset intervals, acquire multi-directional street view images of each sampling point, perform semantic segmentation on the street view images to calculate the proportion of vegetation pixels, and generate green view rate observation values ​​for the road network area.

[0056] Step 3: Perform density quantization processing on the road network distribution data to generate road network weight parameters. At the same time, convert the land use type raster into category coding parameters. Merge the road network weight parameters and the category coding parameters to generate an environmental parameter index. Embed the environmental parameter index into the spatial structure hierarchy to construct a multi-scale environmental feature framework.

[0057] Step 4: Construct a spatial interpolator based on the multi-scale environmental feature framework, and input the green view rate observation value and the environmental parameter index into the spatial interpolator. The green view rate observation value (i.e. the green view rate observation value of the road network area) completes the numerical calculation of the green view rate of the non-road area under the framework constraints, and generates the global interpolation result.

[0058] Step 5: Perform rasterization processing on the global interpolation results to generate a green view rate raster representing the distribution of green view rate in non-road areas. Spatially overlay the green view rate raster with the urban administrative division vector data to output a spatial distribution map of urban green view rate covering the entire urban area.

[0059] In the urban green view rate spatial distribution simulation method based on the environmental characteristic framework of green facilities, step 1 involves acquiring a city-wide geospatial database. This database includes satellite remote sensing imagery, road network distribution data, land use type raster, urban administrative division vector data, and digital elevation model data. The urban administrative division vector data is generated through digital acquisition and topological checking of publicly available administrative boundary maps, including spatial boundaries and attribute information of multi-level administrative units such as cities, districts, and streets, used for spatial unit division and zoning summarization of subsequent green view rate statistical results. The digital elevation model data is generated through the analysis of aerial survey stereo image pairs or satellite stereo imagery, employing elevation anomaly removal and elevation benchmark unification processing to form a regular grid digital terrain surface, used to extract terrain relief factors to participate in the construction of the environmental characteristic framework. The road network distribution data uses open street map data as the quantitative basis, and the land use type raster is generated through vectorization conversion of publicly available government urban planning maps. Industrial land is assigned a code value of one, residential land a code value of two, and green space a code value of three to match the category coding parameter requirements. Satellite remote sensing imagery requires sensor radiometric calibration. Atmospheric correction is performed using dark pixel subtraction, and after determining the target spatial resolution, all images are resampled to that resolution using the nearest neighbor method. This step provides standardized input data for subsequent processing, supporting data consistency throughout the entire process.

[0060] Step 2 involves using the Normalized Difference Vegetation Index (NDI) algorithm to process the NDI pixel values ​​of satellite remote sensing images. By comparing these values ​​with a preset first threshold, NDI pixel values ​​exceeding this threshold are marked as vegetation cover pixels. This marking operation serves the spatial positioning of vegetation areas. Subsequently, street view sampling points are generated along the road network at preset intervals. Street view image data from multiple directions are acquired at each sampling point. Semantic segmentation is performed on the street view images to calculate the proportion of vegetation pixels, generating green view rate observation values ​​for the road network area. The preset first threshold is automatically determined using a machine learning clustering algorithm. This step ensures accurate generation of observation values, providing reliable input for subsequent spatial interpolation and supporting the scientific basis for simulating the spatial distribution of urban green view rate.

[0061] Step 3 involves integrating environmental parameters. Density quantization is performed on the road network distribution data to generate road network weight parameters, while land use type rasters are converted into category coding parameters. The road network weight parameters and category coding parameters are merged to generate an environmental parameter index. This index is then embedded into the spatial structure hierarchy to construct a multi-scale environmental feature framework. This multi-scale framework integrates the road network weight parameters, category coding parameters, and topographic relief factors. The topographic relief factors are aligned in coordinate space and superimposed onto the land use type rasters to form composite parameters. These composite parameters serve as the core constraint terms of the environmental parameter index. The multi-scale environmental feature framework strengthens the spatial correlation of environmental features, supporting subsequent spatial interpolation calculations.

[0062] Step 4 constructs a spatial interpolator based on a multi-scale environmental feature framework. The green view rate values ​​of the road network area extracted in Step 2 are used as input observation points, and environmental parameter indices are used as spatial covariates. The spatial interpolator performs calculations under the framework constraints, completing the prediction of green view rate values ​​in non-road areas and generating global interpolation results. The interpolation algorithm uses the Kriging method, and the output predicted surface resolution meets the requirements of standard grids. Optionally, publicly available building survey data from urban planning departments can be introduced to extract building density values ​​and generate weighting factors as additional covariates to optimize the calculation results for old urban areas. This step achieves full-area green view rate coverage, solving the data blind spot problem in non-road areas.

[0063] Step 5 performs rasterization on the global interpolation results to generate green view rate distribution density raster data. The green view rate raster is spatially overlaid with urban administrative division vector data to output a spatial distribution map covering the entire city. Subsequent extended processing includes collecting temperature monitoring data from meteorological stations, performing inverse distance weighted interpolation to generate heat map data, overlaying the green view rate distribution map, calculating the correlation coefficient between administrative region green view rate and temperature, and generating an urban heat island effect analysis report. Raster cells with green view rates below a preset second threshold are identified, and continuous low green view rate areas are marked as green spaces to be optimized. Location data and population density data are correlated to generate a priority-ranked green space planning decision suggestion table. This step outputs visualization results, directly supporting urban planning decisions.

[0064] Specifically, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework of the present invention includes the following in step 1:

[0065] The road network distribution data uses open street map data as the basis for road network density quantification.

[0066] The land use type grid is generated by performing a vectorization transformation on the government's publicly available urban planning map. Industrial land is assigned a code value of 1 to identify the clustering characteristics of production facilities, residential land is assigned a code value of 2 to characterize the distribution characteristics of residences, and green space is assigned a code value of 3 to mark the vegetation coverage space. The code values ​​are used to constrain the differences in land use type zoning in the environmental parameter index.

[0067] In road network distribution data processing, open street map data needs to undergo topology checks to correct dangling nodes and overlapping line segments, generating topology-clean road network vector data. Topology checks include identifying isolated road segments not connected to other road endpoints and eliminating duplicate road line features to ensure spatial continuity of the road network. Based on the topology-clean road network, road level filtering is performed, selecting arterial and secondary roads as density quantification objects, excluding side roads and alleys to control the calculation scale. Road levels are determined based on the original attribute fields of the open street map: Level 1 roads correspond to urban expressways and arterial roads, and Level 2 roads correspond to regional arterial roads.

[0068] Land use type raster conversion utilizes a pre-defined GIS raster-to-vector tool to process publicly available urban planning maps, generating land use data with vector polygon feature structures. During vectorization, pixel value aggregation rules are set to merge adjacent raster cells of the same land use type into a single polygon. Industrial land is assigned an integer code value one corresponding to areas with clustered production facilities; residential land is assigned an integer code value two representing residential distribution characteristics; and green space is assigned a code value three identifying vegetation cover space. Attribute assignment is completed in batches using a field calculator, and the coded values ​​are stored in the land use type field as category identifiers.

[0069] The coded values ​​implement land use type zoning differences constraints within a multi-scale environmental feature framework. During the construction of the environmental parameter index, the industrial land coded value one is spatially weighted and overlaid with road network weight parameters to strengthen the traffic impact factor of industrial areas; the residential land coded value two is associated with building density weight factors to quantify the development intensity characteristics of residential areas; and the green space coded value three is coupled with topographic relief factors to reflect the topographic adaptability of ecological space. Spatial overlay analysis is performed using a preset geographic information system overlay tool, adjusting the environmental parameter index weight coefficients based on the zoning differences of the coded values ​​to form differentiated spatial constraint rules.

[0070] The land use type zoning difference constraint mechanism is reflected in a three-level logic: industrial zones focus on the impact of road network transportation efficiency, residential zones emphasize the building shading effect, and ecological zones focus on the synergistic relationship between topography and vegetation. Difference constraints are implemented through spatial algebra operations, with the coded value serving as the core variable in the conditional expression, controlling the contribution weight of environmental parameters in different zones. Figure 3 The distribution map of environmental feature framework indicators visualizes the spatial differentiation characteristics of encoded values, providing evidence for the partitioning constraint mechanism.

[0071] Specifically, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework of the present invention further includes, in step 1:

[0072] Sensor radiometric calibration is performed on satellite remote sensing images, atmospheric correction is performed using dark pixel subtraction, the target spatial resolution is determined, and all images are resampled to that resolution using the nearest neighbor method.

[0073] Satellite remote sensing image preprocessing includes sensor radiometric calibration, which uses a radiative transfer model to convert raw digital quantization values ​​into top-level atmospheric reflectance. Dark pixel subtraction atmospheric correction selects low-reflectance near-infrared regions in the image as dark targets, eliminating aerosol influences through atmospheric scattering parameter estimation. Resampling processing, based on target spatial resolution requirements, employs the nearest neighbor method to maintain pixel spectral characteristics and avoid vegetation index distortion caused by interpolation. This workflow addresses the critical impact of remote sensing data quality on initial green view rate calculations.

[0074] Specifically, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework of the present invention includes step 2 as follows:

[0075] The normalized vegetation index pixel value calculated by the normalized vegetation index algorithm is compared with a preset first threshold.

[0076] Normalized vegetation index pixel values ​​that are higher than the preset first threshold are marked as vegetation cover pixels, and a vegetation distribution binary raster is generated to identify the spatial location of vegetation.

[0077] Based on open street map data processed through topology inspection, a road network vector boundary is generated. This road network vector boundary is used to define the spatial range of vegetation clipping. As the study area, the road network vector boundary serves as the boundary of the study area. Using a vector boundary clipping tool in a pre-defined geographic information system, spatial clipping is performed on the binary raster of vegetation distribution according to the road network vector boundary. Vegetation marker data outside the road network coverage area is removed, and vegetation coverage information inside the road boundary is extracted. The road network vector boundary provides a basis for defining the spatial range of the road area, supports the spatial positioning of subsequent street view sampling point deployment, and is used for the construction of a multi-scale environmental feature framework.

[0078] The preset first threshold is determined by calculating the normalized vegetation index value of each pixel using the normalized vegetation index algorithm on the satellite remote sensing image obtained in step 1, and performing K-means cluster analysis on all normalized vegetation index values. The number of clusters is set to 2 to correspond to the two land cover categories of vegetation and non-vegetation, which is used to distinguish between the two types of land cover. When the normalized vegetation index pixel value is higher than the preset first threshold, it is marked as a vegetation-covered pixel, and when the normalized vegetation index pixel value is lower than the preset first threshold, it is marked as a non-vegetation pixel.

[0079] The preset first threshold is determined by the following method: K-means clustering analysis is performed on the normalized vegetation index (NDI) values ​​of all pixels in the satellite remote sensing image obtained in step 1, with the number of clusters set to 2. A machine learning algorithm is used to automatically distinguish between vegetation and non-vegetation features, and the optimal threshold is determined based on the cluster centers. Healthy, dense vegetation has significantly higher reflectance in the near-infrared band than in the visible red band, resulting in a generally higher NDI value. Conversely, the NDI values ​​of non-vegetation features such as bare soil, water bodies, and buildings are generally lower. This K-means clustering method can adapt to the differences in spectral characteristics between vegetation and non-vegetation in different regions and at different times, overcoming the limitations of applying a fixed threshold in different spatial units. The specific value of the preset first threshold can also be dynamically optimized and adjusted by calling a historical vegetation sample database based on the regional vegetation type distribution characteristics. Combining this with the K-means clustering results enhances the accuracy and robustness of vegetation identification in complex urban environments, thereby achieving reliable differentiation between vegetation and non-vegetation areas.

[0080] In step 2, the satellite remote sensing image is processed using the Normalized Difference Vegetation Index (NDVI) algorithm to calculate the vegetation index value corresponding to each pixel. The NDVI algorithm calculates the difference between near-infrared and red light reflectance by dividing it by the sum of the near-infrared and red light reflectances. The formula is: NDVI = (NIR - Red) / (NIR + Red), where NIR represents near-infrared reflectance and Red represents red light reflectance. The algorithm generates a continuous numerical layer of Normalized Difference Vegetation Index (NDVI). The algorithm calculates a NDVI for each pixel in the satellite remote sensing image using the above formula, thus generating a continuous numerical layer of NDVI. The NDVI pixel value of each pixel in the layer is logically compared with a preset first threshold, and a raster calculator tool is used to traverse all pixel units. NDVI pixel values ​​higher than the preset first threshold are classified as vegetation cover pixels, generating a binary raster mask to form a binary raster layer of vegetation distribution. The vegetation distribution binary raster stores vegetation spatial location information in Boolean data type, where a cell value of 1 indicates a vegetation-covered area and a cell value of 0 indicates a non-vegetated area.

[0081] Spatial clipping of the vegetation distribution binary raster is performed based on the road network vector boundary. Using a pre-defined vector boundary clipping tool in the geographic information system, vegetation marker data outside the road network coverage area is removed. Vegetation cover information within the clipped road area is extracted to obtain the spatial distribution characteristics of vegetation along the road network. This operation is used to quantify the vegetation cover rate of road sections, providing a spatial reference for subsequent street view sampling point deployment.

[0082] The preset first threshold is automatically determined based on data distribution characteristics using K-means clustering analysis. Broadleaf trees exhibit significantly higher reflectance in the near-infrared band than bare soil, while coniferous shrubs show moderate reflectance. The clustering analysis method can adapt to the identification needs of different vegetation types, while excluding highly reflective features such as water reflections and asphalt pavements. The dynamic threshold optimization mechanism, determined by combining historical vegetation sample databases with K-means clustering results, enhances the robustness of vegetation identification in complex urban environments.

[0083] Specifically, the urban green view rate spatial distribution simulation method based on the green facility environmental feature framework described in this invention includes the following steps in step 3: constructing a multi-scale environmental feature framework.

[0084] Perform slope analysis on the digital elevation model data obtained in step 1 to generate terrain relief factors;

[0085] The terrain relief factor is aligned and superimposed onto the land use type raster according to the coordinate space;

[0086] The superimposed composite parameters are used as the core constraint terms of the environmental parameter index.

[0087] The topographic relief factor is generated by calculating the rate of change of surface slope using digital elevation model data. Spatial alignment is achieved through geographic coordinate system unification and pixel size matching, with land use type rasters and slope rasters being weighted and overlaid using a raster calculator. The composite parameter construction employs a multi-dimensional feature fusion strategy, with road network weight parameters, land use type coding parameters, and the topographic relief factor forming a three-dimensional environmental feature tensor. This framework enables multi-scale representation of spatial heterogeneity, supporting the selection of covariates in subsequent interpolation algorithms.

[0088] Specifically, in the urban green view rate spatial distribution simulation method based on the environmental characteristics framework of green facilities described in this invention, step 4 includes:

[0089] Use the green visibility observations generated in step 2 as input points for the Kriging interpolation algorithm;

[0090] The terrain relief factor and the land use type raster are used as spatial covariates;

[0091] Kriging interpolation is performed to generate a green visibility prediction surface for non-road areas, with a resolution of 30m×30m grid.

[0092] The Kriging interpolation algorithm uses a semi-variogram to model spatial autocorrelation, with road network green view rate observation points as the known sample point set. Topographic relief factor is used as a continuous covariate in the variogram modeling, and land type raster is used as a categorical covariate in the local estimation. The predicted surface is generated using a standard Kriging spatial prediction equation, with the grid resolution consistent with the spatial resolution of satellite remote sensing imagery. This design ensures the spatial continuity between the predicted results and the observed data for non-road areas.

[0093] Specifically, in the urban green view rate spatial distribution simulation method based on the environmental characteristics framework of green facilities described in this invention, step 4 includes:

[0094] Obtain publicly available building survey data from urban planning departments;

[0095] Extract the building density values ​​of old urban areas from the building census data to generate a building density weighting factor;

[0096] The building density weighting factor is used as an additional covariate input to the spatial interpolator to adjust the numerical calculation results of the green view rate in old urban areas.

[0097] In building survey data processing, the ratio of building footprint area to land area is extracted to generate a building density raster. Identification of old urban areas employs a comparative analysis of historical urban planning maps and current land use, with building density weighting factors used as spatial heterogeneity moderating terms input into the Kriging equations. Additional covariates are embedded into the interpolation model through linear combination to enhance the accuracy of green view rate prediction in high-building-density areas.

[0098] Specifically, the urban green view rate spatial distribution simulation method based on the green facility environmental characteristic framework of the present invention further includes step 6, which includes:

[0099] Collect temperature monitoring data from the weather station;

[0100] Inverse distance weighted interpolation is performed on the temperature monitoring data to generate heat map data;

[0101] The heat map data is overlaid with the urban green visibility spatial distribution map output in step 5;

[0102] Based on the overlay results, the correlation coefficient between green view rate and temperature for each administrative region is calculated, and a spatial coupling analysis report on urban heat island effect and green view rate is generated.

[0103] Temperature monitoring data collection requires screening meteorological stations that comply with the installation specifications of the World Meteorological Organization, excluding data from stations in areas with abnormal heat sources. Data preprocessing involves performing time-series integrity checks and using linear interpolation to fill in missing records, forming a continuous-time-scale temperature dataset. This step ensures the spatial representativeness and temporal continuity of the basic data, providing qualified input for thermal distribution modeling.

[0104] Inverse distance weighted interpolation uses a power-law decay function to calculate spatial influence weights and sets a dynamic search radius to adapt to the non-uniform distribution characteristics of meteorological stations. During interpolation, a weighted average temperature value of neighboring stations is calculated pixel-by-pixel to generate raster data of the spatial temperature distribution. This raster data is then transformed into heatmap data through color mapping, establishing a correspondence between temperature value ranges and color gradients. This process transforms discrete station data into a continuous spatial surface, forming a visualized temperature distribution base.

[0105] The overlay of heatmap data with the spatial distribution map of urban green view rate employs pre-defined geographic information system raster algebra operations. Coordinate system transformation and pixel size matching are performed to ensure spatial overlay capability between the temperature and green view rate rasteres. During the overlay process, data association is performed on corresponding pixels within each administrative unit, generating a multidimensional attribute table including temperature and green view rate values. This operation establishes spatial coupling relationships among environmental parameters, supporting subsequent quantitative analysis.

[0106] Within each administrative unit, the Pearson product-moment correlation coefficient algorithm was used to calculate the linear correlation between green view rate and temperature. The algorithm determines the correlation coefficient by the ratio of covariance to standard deviation, with the result ranging from -1 to +1. The correlation coefficient calculation results, combined with significance tests, generated a spatial coupling analysis report including statistical significance levels. The analysis report quantitatively characterizes the mitigation effect of green view rate distribution on the urban heat island effect, providing a basis for climate adaptation planning.

[0107] Specifically, in the urban green view rate spatial distribution simulation method based on the environmental characteristics framework of green facilities described in this invention, step 6 further includes:

[0108] Obtain the daily average temperature data monitored by the weather station within a preset time period, set the search radius parameter of the inverse distance weighted interpolation algorithm to 5 kilometers, and perform raster interpolation to output a temperature raster layer with a 1km×1km grid.

[0109] Acquiring daily average temperature data requires screening meteorological monitoring stations that comply with the installation specifications of the World Meteorological Organization, excluding data from stations located in areas with abnormal heat sources or in environments with obstructed views. The raw temperature monitoring data within a preset time period undergoes time-series integrity verification, and short-term missing records are filled using a sliding window mean. Data standardization eliminates sensor system errors from different stations, forming a daily average temperature dataset with a consistent time scale. This process establishes a qualified data foundation for temperature spatial interpolation.

[0110] The inverse distance weighted interpolation algorithm employs a dynamic search radius mechanism, with the radius parameter adaptively adjusted based on the spatial distribution density of meteorological stations. The search range is narrowed in densely populated areas to preserve local characteristics, while the search range is expanded in sparsely populated areas to enhance data continuity. Weight calculation uses an exponential decay function, and the distance decay coefficient is correlated with the spatial representativeness of the stations. Parameter configuration is batch-processed using a pre-defined GIS script tool.

[0111] The spatial interpolation process calculates a weighted average temperature of neighboring stations pixel by pixel, with the weight value inversely proportional to the distance from the station to the target pixel. The interpolation calculation employs a parallel computing architecture to process large-scale raster data, optimizing computational efficiency through memory partitioning. The output results undergo topology consistency checks to eliminate outliers at interpolation boundaries.

[0112] The temperature raster layer is generated using standard geographic grid cells, with the grid coordinate system consistent with the city's basic geographic data. Cell values ​​are stored using a floating-point data structure to preserve decimal precision for temperature values. The raster layer includes a spatial reference information file, containing metadata attributes such as coordinate system, cell size, and data range. This output can be directly input into a spatial overlay analysis module, supporting the assessment of thermal environmental effects.

[0113] Specifically, in the urban green view rate spatial distribution simulation method based on the environmental characteristics framework of green facilities described in this invention, step 6 further includes:

[0114] Read the urban green view rate spatial distribution map output in step 5, and determine the preset second threshold by performing K-means cluster analysis on the green view rate values ​​of all grid cells in the green view rate value grid generated in step 5. Set the number of clusters to 2 to distinguish between low green view rate and high green view rate areas, and identify grid cells in all administrative regions whose green view rate is lower than the preset second threshold.

[0115] The preset second threshold is determined by the following method: K-means clustering analysis is performed on the green view rate values ​​of all spatial units in the urban green view rate spatial distribution map output in step 5. The number of clusters is set to 2. A machine learning algorithm automatically distinguishes between low and high green view rate areas. The optimal threshold is determined based on the cluster centers. Low green view rate areas typically correspond to densely built-up areas or areas with insufficient vegetation cover, with generally low green view rate values. High green view rate areas correspond to parks, green spaces, or densely vegetated areas, with relatively high green view rate values. This K-means clustering method can adapt to the spatial heterogeneity of green view rate distribution under different administrative divisions, overcoming the limitations of applying fixed thresholds in different urban areas. The specific value of the preset second threshold can also be dynamically optimized and adjusted by calling the historical green space planning database, based on regional greening standards and population density characteristics. Combining the K-means clustering results enhances the scientific rigor and relevance of urban green space planning decisions, thereby achieving accurate identification and optimization of low green view rate areas.

[0116] Areas with continuous low green visibility are marked as green spaces to be optimized. The location data and population density data of the green spaces to be optimized are associated to generate a green space planning decision suggestion table with priority ranking.

[0117] After reading the urban green view rate spatial distribution map output in step 5, the first step is to identify areas with low green view rate. A preset second threshold is set to distinguish areas with insufficient green view rate; this threshold is dynamically determined based on urban greening standards or historical experience values. The identification process uses a preset geographic information system raster analysis tool to traverse all raster cells within each administrative jurisdiction unit, filtering cells with green view rate values ​​lower than the preset second threshold. The raster cell filtering results are stored as a binary mask layer, identifying the locations of potential greening defects. This step automates the detection of urban green view rate distribution defects, providing target area input for subsequent optimization.

[0118] After identification, low-green-view-rate raster units require spatial continuity processing. A morphological dilation algorithm is applied to connect adjacent low-green-view-rate units, eliminating isolated noise points. The boundaries of continuous areas are extracted using a boundary tracing algorithm, generating closed polygon vector data. These continuous areas are marked as green spaces to be optimized, and their spatial boundary data is stored in association with an attribute table. This process ensures that the spatial units to be optimized possess spatial integrity and operability, supporting unit division for planning decisions.

[0119] Spatial overlay analysis was used to correlate location data and population density data of green spaces to be optimized. Location data was extracted from the geometric center coordinates of the vector boundary of the space to be optimized, and population density data was generated as a continuous surface using kernel density estimation. Population density values ​​were assigned to the spatial units to be optimized through spatial join operations, forming a composite attribute table including location coordinates, green visibility value, and population density. This correlation establishes a spatial correspondence between greening demand and population distribution, quantifying the urgency of optimization in different areas.

[0120] The generation of a green space planning decision recommendation form is based on a weighted scoring mechanism. The attribute table of the spatial units to be optimized is input into a priority ranking model, which assigns different weights to the green view gap and population density values. The ranking results are output as a structured table, including area identification, optimization priority level, and spatial location reference information. The decision recommendation form can be directly imported into urban planning management software to guide resource allocation for green space expansion or renovation projects. This step achieves a closed-loop transformation from data analysis to action recommendations, improving the scientific rigor and relevance of urban green space planning.

[0121] The specific implementation of this invention is as follows:

[0122] In the spatially comprehensive observation extraction stage, the city's entire geospatial database is first acquired, including road network distribution data provided by open street maps. Based on this road network distribution data, street view sampling points are generated at preset intervals using a point generation tool along the road network. These sampling points are evenly distributed along the road network and serve as the base locations for subsequent green view rate observations. Subsequently, the point generation tool along the road network is used to generate street view sampling points at 100-meter intervals along the simplified road network; these points serve as the base locations for subsequent green view rate observations. Moving to the street view image acquisition and green view rate calculation stage, the latitude and longitude coordinates of all street view sampling points in the WGS-84 coordinate system are acquired. After converting to the target coordinate system, street view images are acquired at four directions (0°, 90°, 180°, and 270°) for each sampling point. Then, the Normalized Difference Vegetation Index (NDVI) algorithm is applied to directly identify vegetation pixels in the street view images. The green view rate of each street view sampling point is calculated as the average of the green view rates of the four street view images for that point. The calculation formula is as follows:

[0123] ;

[0124] in, Indicates the street view sampling point number The number of vegetation pixels in the street view image captured in the direction. GVI represents the total number of pixels in the street view image captured in the i-th direction of the street view sampling point, and GVI is the green visibility rate. This process generates green visibility rate observations for the road area, and uses the generated green visibility rate observations as input point data for the spatial interpolation algorithm for subsequent calculation of green visibility rate values ​​in non-road areas.

[0125] In the stage of constructing the simulation model based on the environmental feature framework of green facilities, environmental parameter indexes are integrated: A fishing net tool is used to divide the city into grids, and the proportion of road network length within each grid is statistically analyzed to generate road network weight parameters. Simultaneously, land use type grids (industrial land assigned code value 1, residential land assigned code value 2, and green space assigned code value 3) are converted into category coding parameters. After merging these parameters, they are embedded into the spatial structure hierarchy to construct a multi-scale environmental feature framework. In addition, terrain undulation factors or building density weight factors can be optionally added as core constraints to enhance robustness. Based on this framework, street view sampling point observations are input into the multi-scale environmental feature framework, extrapolating to generate the green view rate distribution of non-road areas, and outputting a global result with a resolution of 70m × 70m grids. The framework parameters are optimized through training.

[0126] In the urban spatial unit division and result output stage, a fishing net creation tool is used to divide the city into uniform grids. The interpolation results across the entire area are rasterized to generate raster data representing the density of green view rate distribution. The green view rate raster is then spatially overlaid with the city's administrative division vector data to output a spatial distribution map covering the entire city. In the application example, the central urban area of ​​Xi'an is selected as the experimental subject. The research scope is determined based on road network segmentation, the street view sampling point interval is set to 100 meters, and the grid resolution is adjusted to 70 meters × 70 meters. After simulating the spatial distribution of green view rate using this method, the heat island effect can be further analyzed or a green space planning decision-making suggestion table can be generated. However, experimental details are presented in text form to avoid visual output. The entire process is efficient and low-cost, solving the problem of blind spots in street view data and supporting urban planning decisions.

[0127] This invention addresses the technical problem of insufficient street view image coverage by constructing a multi-scale environmental feature framework and a spatial interpolation mechanism. Specifically, it establishes a spatial correlation between street view measurements in road areas and environmental feature parameters, and extrapolates the green view rate distribution in non-road areas under the constraints of the environmental framework. This breaks through the dependence of existing methods on full-area street view sampling, avoids high-cost street view vehicle deployment or commercial application interface calls, and significantly reduces the data acquisition threshold and implementation difficulty.

[0128] In the data processing stage, the initial green view rate raster layer is clipped based on the road network vector boundary, and the green view rate values ​​of the road areas are extracted as observations. This step utilizes publicly available resources such as open street map data to simplify the road network structure, achieve representativeness of observation points, and control costs. An environmental parameter index is constructed that integrates road network weight parameters and land use category coding parameters, embedding a multi-scale environmental feature framework to provide structured constraints for subsequent calculations. The framework includes optional elements such as topographic relief factors, enhancing spatial representation capabilities, avoiding the need for high-precision 3D data acquisition such as LiDAR, and adapting to the resource status of small and medium-sized cities.

[0129] The spatial interpolator, designed based on the Kriging algorithm, takes road area observations and environmental parameter indices as input and performs numerical prediction of green view rate in non-road areas under framework constraints. The interpolation process uses environmental features as spatial covariates and extrapolates blind zone data to generate global interpolation results, achieving full coverage of the spatial distribution of urban green view rate.

Claims

1. A method for simulating the spatial distribution of urban green view rate based on a framework of green facility environmental characteristics, characterized in that, include: Step 1: Obtain the geospatial database of the entire city. The geospatial database includes satellite remote sensing imagery, road network distribution data, land use type raster, urban administrative division vector data, and digital elevation model data. Step 2: Process satellite remote sensing images using the normalized vegetation index algorithm to mark vegetation coverage areas; generate street view sampling points along the road network at preset intervals, acquire multi-directional street view images of each sampling point, perform semantic segmentation on the street view images to calculate the proportion of vegetation pixels, and generate green view rate observation values ​​for the road network area. Step 3: Perform density quantization processing on the road network distribution data to generate road network weight parameters. At the same time, convert the land use type raster into category coding parameters. Merge the road network weight parameters and the category coding parameters to generate an environmental parameter index. Embed the environmental parameter index into the spatial structure hierarchy to construct a multi-scale environmental feature framework. Step 4: Construct a spatial interpolator based on the multi-scale environmental feature framework, and input the observed green visibility rate and the environmental parameter index into the spatial interpolator to complete the numerical calculation of the green visibility rate in non-road areas under the framework constraints, and generate global interpolation results; Step 5: Perform rasterization processing on the global interpolation results to generate a green view rate value raster representing the spatial distribution of green view rate in non-road areas. Spatially overlay the green view rate value raster with the urban administrative division vector data to output a spatial distribution map of urban green view rate covering the entire urban area. The construction of the multi-scale environmental feature framework in step 3 includes: Perform slope analysis on the digital elevation model data obtained in step 1 to generate terrain relief factors; The terrain relief factor is aligned and superimposed onto the land use type raster according to the coordinate space; The superimposed composite parameters are used as the core constraint terms of the environmental parameter index; Step 4 includes: Use the green visibility observations generated in step 2 as input points for the Kriging interpolation algorithm; The terrain relief factor and the land use type raster are used as spatial covariates; Kriging interpolation is performed to generate a green visibility prediction surface for non-road areas, with a resolution of 30m×30m grid.

2. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 1, characterized in that, Step 1 includes: The road network distribution data uses open street map data as the basis for road network density quantification. The land use type grid is generated by performing a vectorization transformation on the government's publicly available urban planning map. Industrial land is assigned a code value of 1 to identify the clustering characteristics of production facilities, residential land is assigned a code value of 2 to characterize the distribution characteristics of residences, and green space is assigned a code value of 3 to mark the vegetation coverage space. The code values ​​are used to constrain the differences in land use type zoning in the environmental parameter index.

3. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 2, characterized in that, Step 1 also includes: Sensor radiometric calibration is performed on satellite remote sensing images, atmospheric correction is performed using dark pixel subtraction, the target spatial resolution is determined, and all images are resampled to that resolution using the nearest neighbor method.

4. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 3, characterized in that, Step 2 includes: The normalized vegetation index pixel value calculated by the normalized vegetation index algorithm is compared with a preset first threshold. Normalized vegetation index pixel values ​​that are higher than the preset first threshold are marked as vegetation cover pixels, and a vegetation distribution binary raster is generated to identify the spatial location of vegetation. A road network vector boundary is generated based on open street map data processed by topology inspection. The road network vector boundary is used to define the spatial range of vegetation clipping. Using a vector boundary clipping tool in a preset geographic information system, spatial clipping is performed on the binary raster of vegetation distribution according to the road network vector boundary. Vegetation marker data outside the road network coverage area is removed, and vegetation coverage information inside the road boundary is extracted. The road network vector boundary provides a basis for defining the spatial range of road areas and is used for the spatial positioning of street view sampling points. The preset first threshold is determined by calculating the normalized vegetation index (NVI) value of each pixel using the normalized vegetation index algorithm on the satellite remote sensing image obtained in step 1, and then performing K-means cluster analysis on all NVI values. The number of clusters is set to 2 to correspond to the two land cover categories of vegetation and non-vegetation, which is used to distinguish between the two types of land cover. When the NVI pixel value is higher than the preset first threshold, it is marked as a vegetation-covered pixel, and when the NVI pixel value is lower than the preset first threshold, it is marked as a non-vegetation pixel.

5. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 4, characterized in that, Step 4 includes: Obtain publicly available building survey data from urban planning departments; Extract the building density values ​​of old urban areas from the building census data to generate a building density weighting factor; The building density weighting factor is used as an additional covariate input to the spatial interpolator to adjust the numerical calculation results of the green view rate in old urban areas.

6. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 5, characterized in that, It also includes step 6, which includes: Collect temperature monitoring data from the weather station; Inverse distance weighted interpolation is performed on the temperature monitoring data to generate heat map data; The heat map data is overlaid with the urban green visibility spatial distribution map output in step 5; Based on the overlay results, the correlation coefficient between green view rate and temperature for each administrative region is calculated, and a spatial coupling analysis report on urban heat island effect and green view rate is generated.

7. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 6, characterized in that, Step 6 also includes: Obtain the daily average temperature data monitored by the weather station within a preset time period, set the search radius parameter of the inverse distance weighted interpolation algorithm to 5 kilometers, and perform raster interpolation to output a temperature raster layer with a 1km×1km grid.

8. The method for simulating the spatial distribution of urban green view rate based on the environmental characteristics framework of green facilities according to claim 7, characterized in that, Step 6 also includes: Read the urban green view rate spatial distribution map output in step 5, and determine the preset second threshold by performing K-means cluster analysis on the green view rate values ​​of all grid cells in the green view rate value grid generated in step 5. Set the number of clusters to 2 to distinguish between low green view rate and high green view rate areas, and identify grid cells in all administrative regions whose green view rate is lower than the preset second threshold. Areas with continuous low green visibility are marked as green spaces to be optimized. The location data and population density data of the green spaces to be optimized are associated to generate a green space planning decision suggestion table with priority ranking.

Images