Greenbelt cooling effect evaluation method and system based on remote sensing image and streetscape image

Abstract

The invention discloses a greenbelt cooling effect evaluation method and system based on a remote sensing image and a streetscape image, and the method comprises the steps: obtaining the data of the streetscape image, carrying out the denoising and cutting, recognizing, extracting and calculating the pixel coverage rate of arbors, grass trees and shrubs, and constructing the greening feature indexes of the streetscape image; acquiring an urban satellite remote sensing image, performing radiometric calibration, correction and identification, and constructing a greening characteristic index of the remote sensing image; acquiring a multispectral remote sensing image of the city, performing cloud detection, and obtaining the surface temperature of the city through surface temperature inversion; and constructing a random forest model and performing iterative training based on greening characteristic indexes of the surface temperature, the streetscape image and the remote sensing image of the city, and evaluating the cooling effect by using the fitted random forest model. The method gives full play to the advantages of wide coverage range of the remote sensing image and rich details of the street scene image, makes up for the limitation of a single data source, and achieves the comprehensive and accurate evaluation of the cooling effect of the green land.

Description

Technical Field

[0001] The present invention belongs to the technical field of urban ecological environment monitoring and assessment, and particularly relates to a method and system for evaluating the cooling effect of green spaces based on remote sensing images and street view images. Background Technique

[0002] With the acceleration of the urbanization process, the urban heat island effect has become increasingly serious, having many negative impacts on residents' lives and the ecological environment. As an important factor in alleviating the heat island effect, the accurate evaluation of the cooling effect of urban green spaces is of great significance for urban planning and ecological construction. Urban green spaces are lands partially or fully covered by artificial or natural vegetation within the urban area. As an indispensable part of the urban ecosystem, urban green spaces not only provide urban residents with spaces for relaxation, play, and enjoying life, but also play a key role in reducing the urban heat, optimizing the urban thermal environment, and regulating the local microclimate of the city. The essence of the cooling effect of green spaces is the evaluation of the impact of green space landscapes on the temperature within the region. Previous methods for evaluating the cooling effect of green spaces either relied on a single data source, such as only using remote sensing images and being unable to obtain the fine greening characteristics at the street level, or adopted on-site monitoring, which had problems such as low efficiency and limited coverage. Although street view images can reflect the micro greening conditions of streets, it is difficult to grasp the impact of the overall layout of urban green spaces on cooling from a macro perspective when used alone. Although remote sensing images can conduct macro monitoring, they are insufficient in depicting micro greening details. Therefore, how to integrate remote sensing images and street view images to achieve the accurate evaluation of the cooling effect of green spaces has become a key issue in urban ecological research.

[0003] Street view images are visual data obtained by map service providers through professional street view vehicles collecting data along the urban road system in all directions. They completely record the physical space form of urban streets in the form of high-definition pictures from multiple angles and aspects. As a new type of geospatial data, street view images have significant technical advantages: in terms of spatial coverage, their sampling points are densely distributed in urban road networks at all levels, and the visual data of adjacent sampling points can be seamlessly connected to construct a complete urban street space expression system; in terms of information expression, street view images truly and meticulously present the actual state characteristics of the urban physical space from the first-person perspective, providing rich visual materials for urban landscape research. In recent years, with the rapid development of artificial intelligence technologies such as deep learning and computer vision, the semantic information mining ability of street view images has been significantly improved. It can not only accurately identify and extract semantic targets in the scene, but also quantitatively analyze and express the characteristics of the built environment, providing strong technical support for urban research. Remote sensing images are image data captured by sensors carried by remote sensing platforms such as satellites, airplanes or drones along specific areas or routes. Each remote sensing image usually represents the surface features of a wide area and can express different surface information through images of multiple bands. Remote sensing images can comprehensively cover large geographical areas and accurately present various physical features of the surface through high-resolution data. Compared with street view images, remote sensing images have a wider coverage range and high-density spatial data, and can be seamlessly stitched to form a complete geospatial information. In terms of the expressed content, remote sensing images can not only reflect the natural landscape of the surface, but also reveal the distribution and changes of artificial landscapes, with extremely high data acquisition efficiency and timeliness. With the support of advanced data processing and analysis technologies, remote sensing images can provide accurate geographical information and strong data support for multiple fields such as environmental monitoring, land use planning, and resource management. Through these technologies, remote sensing images can accurately extract feature information in the scene and efficiently understand complex geographical phenomena. Therefore, by using remote sensing images and street view images, comprehensive green space information can be extracted from the horizontal and vertical perspectives to evaluate the cooling effect of urban green spaces, helping planners to formulate more accurate and feasible urban green space planning schemes. Summary of the Invention

[0004] The main purpose of the present invention is to overcome the deficiencies of the prior art and provide a method and system for evaluating the cooling effect of green spaces based on remote sensing images and street view images.

[0005] To achieve the above purpose, the present invention adopts the following technical solutions: In one aspect of the present invention, a method for evaluating the cooling effect of green spaces based on remote sensing images and street view images is provided, including the following steps: Obtain street view image data and perform denoising and cropping; Use a semantic segmentation algorithm to identify street view image data, extract and calculate the pixel coverage rates of arbors, herbs, and shrubs, and construct greening feature indicators for street view images; Obtain urban satellite remote sensing images and perform radiometric calibration and correction. Identify the remote sensing images through a semantic segmentation algorithm, and construct greening feature indicators for the remote sensing images; Obtain urban multispectral remote sensing images and perform cloud detection, and obtain the surface temperature of the city through surface temperature inversion; Based on the surface temperature of the city, and the greening feature indicators of street view images and remote sensing images, construct a random forest model and perform iterative training, and use the fitted random forest model to evaluate the cooling effect.

[0006] As a preferred technical solution, the obtaining of street view image data and performing denoising and cropping are specifically as follows: Obtain urban road data and perform road merging and topological inspection to obtain processed urban road data; Set sampling points at a set distance interval on the processed urban road data, obtain street view image data at the sampling points and perform denoising and cropping. Among them, one street view picture is collected for each of the four directions of up, down, left, and right at each sampling point.

[0007] As a preferred technical solution, the using of a semantic segmentation algorithm to identify street view image data, extract and calculate the pixel coverage rates of arbors, herbs, and shrubs, and construct greening feature indicators for street view images is specifically as follows: Use a pre-trained Deeplabv3+ object semantic segmentation model to identify street view image data, extract the place objects, and calculate the total pixel proportion of the pixels of arbors, herbs, and shrubs in the street view image data, and construct greening feature indicators for street view images.

[0008] As a preferred technical solution, the obtaining of urban satellite remote sensing images and performing radiometric calibration and correction, and identifying the remote sensing images through a semantic segmentation algorithm, and constructing greening feature indicators for the remote sensing images is specifically as follows: Download urban satellite remote sensing images from a satellite remote sensing platform, and perform operations of radiometric calibration, atmospheric correction, and geometric correction; Use an HRNet-W48 image classification model to separate land use of urban satellite remote sensing images and extract urban green spaces; According to the composition and spatial distribution of urban green spaces, construct greening feature indicators for the remote sensing images. The greening feature indicators for the remote sensing images include green space landscape composition indicators and spatial configuration indicators, specifically: Landscape composition indicators: AREA = a ij ; Among them,AREA is the green area, a ij is the i area of the j th patch within the PERIM = p ij ; Among them, PERIM is the green perimeter, p ij is the i perimeter of the j th patch within the ; Among them, FRAC is an index reflecting the complexity of the patch shape, and FRAC takes values between 1 and 2, and the closer it is to 2, the more complex the shape; ; Among them, PLAND represents the relative proportion of a certain patch type in the entire landscape area, n i is the i number of patches within the A th research scope, NP = n i ; Among them, n i is the i number of patches within the ; Among them, PD is an index reflecting the density of the patches; ; Among them, CA represents the total area of all green patches; ; Among them, LPI represents the proportion of the largest patch in a certain patch type in the entire landscape area; Spatial configuration index: ; Among them, LSI represents the standardized measurement of the landscape shape index; ; Among them, COHESION represents the connection degree of similar type patches in the landscape,​m Indicates the total number of research ranges, Z representing the total number of grids; COHESION A higher value indicates relatively concentrated patches, while COHESION a lower value indicates dispersed patches.

[0009] As a preferred technical solution, obtaining the urban multi-spectral remote sensing image and performing cloud detection, and obtaining the surface temperature of the city through surface temperature inversion, specifically: Obtaining TOA data and SR data, performing cloud detection and masking, and obtaining ASTER GED data and TCWV data; Based on the ASTER GED data, calculating the Normalized Difference Vegetation Index NDVI , as follows: ; where, NIR is the reflectance in the near-infrared band, Red is the reflectance in the red band; Based on the Normalized Difference Vegetation Index NDVI , calculating the vegetation coverage FVC , as follows: ; where, NDVI bare and NDVI veg are the Normalized Difference Vegetation Index NDVI values for pixels of bare land and vegetation respectively; Based on the vegetation coverage FVC calculating the effective emissivity, as follows: ; where, is the effective emissivity, and are the emissivities of vegetation and bare land respectively, and b is the given spectral band; ; where, LST is the surface temperature, Tb is the TOA brightness temperature of the thermal infrared radiation TIR channel, is the surface emissivity of the same channel; A i , B i and C i are algorithm coefficients determined by linear regression of radiative transfer simulation; Based on the SMW algorithm combined with TCWV and surface emissivity data, performing surface temperature inversion, specifically: Select endmembers representing different ground objects from remote sensing images, construct a spectral mixture model, calculate the endmember coefficients of each pixel, that is, the proportion of ground objects, and then output the inversion result after result verification to finally obtain the urban surface temperature distribution data.

[0010] As a preferred technical solution, based on the greening characteristic indexes of the urban surface temperature, street view images and remote sensing images, a random forest model is constructed and iteratively trained, and the cooling effect is evaluated by using the fitted random forest model. Specifically: Standardize the greening characteristic indexes of the green space samples and the inverted surface temperature extracted from all remote sensing images and street view images, divide the study area into study units of a set size, each study unit is a sample, and divide all samples into training samples and verification samples according to a set ratio; Input the greening characteristic indexes and surface temperature in the training samples into the random forest model, and perform iterative training. Optimize the hyperparameters through grid search to find the optimal parameters and generate the optimally fitted random forest model; S53: Input the greening characteristic indexes and surface temperature in the verification samples into the fitted random forest model to generate the surface temperature of the corresponding green space.

[0011] Another aspect of the present invention also provides a green space cooling effect evaluation system based on remote sensing images and street view images, which is applied to the above-mentioned green space cooling effect evaluation method based on remote sensing images and street view images, and includes a data acquisition and preprocessing module, a feature extraction and processing module, a model construction and training module, and an evaluation and result analysis module; The data acquisition and preprocessing module is used to obtain street view image data and perform denoising and cropping; The feature extraction and processing module is used to identify the street view image data by using the Deeplabv3+ semantic segmentation algorithm, extract and calculate the pixel coverage rates of trees, herbs and shrubs, and construct the greening characteristic indexes of the street view images; obtain urban satellite remote sensing images and perform radiometric calibration and correction, identify the remote sensing images by using the HRNet-W48 semantic segmentation algorithm, and construct the greening characteristic indexes of the remote sensing images; obtain urban multi-spectral remote sensing images and perform cloud detection, and obtain the urban surface temperature through surface temperature inversion; The model construction and training module is used to construct a random forest model and perform iterative training according to the greening characteristic indexes of the urban surface temperature, street view images and remote sensing images; The evaluation and result analysis module uses the fitted random forest model to evaluate the cooling effect.

[0012] Another aspect of the present invention further provides a storage medium storing a program which, when executed by a processor, implements the above-mentioned method for evaluating the cooling effect of green spaces based on remote sensing images and street view images.

[0013] Compared with the prior art, the present invention has the following advantages and beneficial effects: (1) The present invention integrates remote sensing image and street view image data, and by means of deep learning and machine learning algorithms, realizes a comprehensive and accurate evaluation of the cooling effect of green spaces. From multi-source data acquisition and preprocessing, to greening feature extraction, evaluation index construction, and model establishment and training, a complete evaluation system is formed. Compared with the traditional method for evaluating the cooling effect of green spaces, the present invention gives full play to the advantages of wide coverage of remote sensing images and rich details of street view images, and makes up for the limitations of a single data source. On the one hand, remote sensing images can obtain the overall distribution and characteristics of urban green spaces from a macroscopic level, providing a basis for analyzing the regional differences in the cooling effect of green spaces; on the other hand, street view images can obtain fine greening information at the street level, accurately reflecting the impact of green spaces on the surrounding environment at the micro scale. During the evaluation process, the application of semantic segmentation models and machine learning algorithms improves the accuracy of greening feature extraction and the accuracy of the evaluation model. By establishing a comprehensive evaluation index system, the landscape composition and spatial configuration characteristics of green spaces are comprehensively considered, making the evaluation results more scientific and reliable.

[0014] (2) The evaluation method and system of the present invention have a high degree of automation, greatly improving the evaluation efficiency and saving labor, material and time costs. This technical solution can not only provide a scientific basis for urban green space planning and optimization, help create a more livable urban thermal environment, but also be applied to urban ecological environment monitoring and urban sustainable development research. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 is a flowchart of the method for evaluating the cooling effect of green spaces based on remote sensing images and street view images according to an embodiment of the present invention; Figure 2 is a schematic structural diagram of the system for evaluating the cooling effect of green spaces based on remote sensing images and street view images according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0017] In order to enable those skilled in the art of the present technology to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative efforts shall fall within the protection scope of the present application.

[0018] Example: As Figure 1 shown, this embodiment provides a method for evaluating the cooling effect of green spaces based on remote sensing images and street view images, including the following steps: S1: Obtain urban road network data and perform preprocessing, and set sampling points to obtain street view image data and perform preprocessing; S11: Process the urban road data through professional geographic information system software, perform preprocessing on the urban road data, and the preprocessing includes road merging and topological checking to obtain preprocessed urban road data; S12: Set sampling points every 100 meters on the preprocessed urban road data through professional geographic information system software. Use web crawler technology combined with the API of map service providers to obtain street view image data around the sampling points. During the process of obtaining pictures, set parameters such as the shooting position and pitch angle of the pictures, and obtain one street view picture for each of the four directions (up, down, left, and right) at a single sampling point; finally, perform preprocessing such as denoising and cropping on the street view images.

[0019] S2: Use the deep learning image semantic segmentation algorithm to identify street view pictures, extract trees, grasses, and shrubs in the street view images, calculate their pixel coverage rates respectively, and construct street view greening feature indicators; S21: Based on street view pictures, use the image semantic segmentation technology in deep learning to construct the greening features of street trees, shrubs, and grasses. The object semantic segmentation algorithm of the image refers to precisely classifying each pixel point of the street view picture, so as to identify the objects in the picture and extract information such as the position and type of the objects.

[0020] In this embodiment, the Deeplabv3+ model framework is used to identify the sampled street view pictures. And calculate the total pixel proportion of the pixels of trees, grasses, and shrubs in the street view image; S22: In this embodiment, the Deeplabv3+ deep model pre-trained using the ADE20K dataset is used to identify the sampled street view pictures, extract the place objects in the street view pictures, calculate the pixel coverage rates of trees, grasses, and shrubs (that is, the ratio of the pixels of this type to the total pixels of this picture), and construct the street greening features described above.

[0021] S3: Obtain high-resolution urban remote sensing images and perform preprocessing, and use the deep learning image semantic segmentation algorithm to identify the remote sensing images and construct the greening feature indicators of the remote sensing images; S31: Download high-resolution remote sensing images from the satellite remote sensing platform, and use professional geographic information system software to perform preprocessing operations such as radiometric calibration, atmospheric correction, and geometric correction on the remote sensing images to improve data quality and ensure the accuracy of the images; S32: Based on high-resolution remote sensing images, the image classification technology in deep learning is used to separate land uses in the remote sensing images and extract urban green spaces. HRNet-W48 adopts a repetitive multi-scale fusion strategy. The low-resolution representation of similar feature layers is used to assist in enhancing the representation ability of high-resolution, thereby improving the classification effect. In this embodiment, the HRNet-W48 model framework is used to identify and extract green spaces in the remote sensing images; S33: According to the composition and spatial distribution of green spaces, a greening feature index of the remote sensing image is constructed. The greening feature index of the remote sensing image consists of two parts: a green space landscape composition index and a spatial configuration index. The calculation formula is as follows: Landscape composition index: AREA = a ij ; Among them, AREA is the green space area, a ij is the i th j patch area in the PERIM = p ij ; Among them, PERIM is the green space perimeter, p ij is the i th j patch perimeter in the ; Among them, FRAC is an index reflecting the complexity of the patch shape, and FRAC takes values between 1 and 2, and the closer it is to 2, the more complex the shape; ; Among them, PLAND represents the relative proportion of a certain patch type in the entire landscape area, n i is the i th A patch number in the study area, NP = n i ; Among them, n i is the i th ; Among them, PD is an index reflecting the patch density; ; Among them, CA represents the total area of all green space patches; ; Among them, LPI represents the proportion of the largest patch in a certain patch type occupying the entire landscape area; Spatial configuration index: ; Among them, LSI represents the standardized measurement of the landscape shape index; ; Among them, COHESION represents the degree of connection of patches of similar types in the landscape, m represents the total number of research scopes, Z represents the total number of grids; COHESION A higher value indicates that the patches are relatively concentrated, while COHESION a lower value indicates that the patches are dispersed.

[0022] S4: Obtain the urban multi - spectral remote sensing image and perform pre - processing, and obtain the surface temperature of the city through surface temperature inversion; S41: Load the TOA data and SR data in the Landsat database on the GEE platform and perform pre - processing, perform cloud detection and masking to remove the interference of clouds on the inversion process, and load the ASTER GED data and TCWV data; among them, the TOA data refers to the data measured at the top of the atmosphere after the surface reflection or radiation passes through the atmosphere, and these data have not been atmospherically corrected, so they contain the influence of the atmosphere on radiation; the SR data is the surface reflectance data after atmospheric correction, representing the optical characteristics of the ground surface itself without atmospheric interference; the ASTER GED data is the global emissivity database obtained by the ASTER satellite sensor, containing the emissivity information of different substances on the surface; the TCWV data refers to the total amount of water vapor in the atmosphere, usually in millimeters, which represents the water vapor content in the entire atmospheric column from the ground to the top of the atmosphere.

[0023] S42: Calculate the vegetation index and emissivity. Based on the surface reflectance data of Landsat, calculate the normalized difference vegetation index NDVI , based on NDVI data, calculate the vegetation coverage FVC , and use the ASTER GED data set to calculate the emissivity of bare land. The calculation formulas are as follows: ; Among them NIR is the reflectance of the near - infrared band, Redis the reflectance in the red light band; ; NDVI is the Normalized Difference Vegetation Index, NDVI bare and NDVI veg are the NDVI values of the pixel for bare land and vegetation respectively; ; where is the effective emissivity, and are the emissivities of vegetation and bare land respectively, and b is the given spectral band; ; where Tb is the TOA brightness temperature of the TIR (Thermal Infrared Radiation) channel, is the surface emissivity of the same channel. The algorithm coefficients A i , B i and C i are determined by linear regression of radiative transfer simulation; S43: Based on the SMW algorithm combined with TCWV and surface emissivity data, perform surface temperature inversion, and finally accurate urban surface temperature distribution data can be obtained.

[0024] S5: Based on the greening characteristic indicators of street view images and remote sensing images, construct a random forest model and perform iterative training to evaluate the cooling effect.

[0025] S51: Extract the greening characteristic indicators and the inverted surface temperature of the green space samples from all remote sensing images and street view images, perform standardization processing, divide the study area into 390m * 390m study units, each study unit is a sample, and divide all samples into training samples and validation samples in a ratio of 7:3; S52: Input the greening characteristic indicators and surface temperature in the training samples into the random forest model, and perform iterative training. Optimize the hyperparameters through grid search to find the optimal parameters and generate an optimally fitted random forest model; S53: Input the greening characteristic indicators and surface temperature in the validation samples into the fitted random forest model to generate the corresponding surface temperature of the green space, reflecting the strength of the cooling effect.

[0026] As Figure 2 shown, in another embodiment of the present application, a green space cooling effect evaluation system based on remote sensing images and street view images is provided. The system includes the following modules: Data acquisition and preprocessing module 10: responsible for collecting remote sensing image data, street view image data and related geographic information data in the study area. Preprocessing the collected data, including radiation correction, atmospheric correction, geometric correction of remote sensing images, denoising and cropping of street view images, and organization and format conversion of geographic information data.

[0027] Feature extraction and processing module 20: Use deep learning models to process street view images and remote sensing images to extract green features. For street view images, use the Deeplabv3+ semantic segmentation model to identify green types and calculate pixel coverage; for remote sensing images, use the HRNet-W48 model to classify land use, combine relevant indicators to refine green space features, and invert surface temperature data.

[0028] Model building and training module 30: Integrate multi-source data feature sets to build a green space cooling effect evaluation model based on the random forest model. Use cross-validation, grid search and other technologies to optimize model parameters, improve model performance, and train and evaluate the model.

[0029] Evaluation and result analysis module 40: Input the multi-source data of the area to be evaluated into the trained model to obtain the evaluation results of the green space cooling effect. Visualize the evaluation results on the map, analyze the differences in the green space cooling effects in different regions and their relationship with the green space landscape characteristics, and provide decision support for urban green space planning.

[0030] Furthermore, in the data acquisition and preprocessing module, the urban road data is processed by a geographic information system professional software, and the urban road data is preprocessed, and the preprocessing includes merging and topological checking to obtain the preprocessed urban road data; the street view image data is processed by a geographic information system professional software, and the street view pictures are preprocessed, and the preprocessing includes setting sampling points, street view picture downloading, picture checking, picture cutting and fusion, etc., to obtain the preprocessed street view image data; the remote sensing image data is processed by a geographic information system professional software, and the preprocessing includes radiation calibration, atmospheric correction and geometric correction, etc., to obtain the preprocessed remote sensing image data.

[0031] Furthermore, in the feature extraction and processing module, the steps of constructing street greening feature indicators are as follows: (1) Model preparation: download the required files of the Deeplabv3+ model from the open source platform and run the test files that come with the model to test whether the model is built successfully; (2) Traverse all streets and street view images collected at sampling points on all streets, cut the street view images into a specified size (the size of the street view images in this patent is 512*512), and use this as the model input; (3) Run the Deeplabv3+ model and wait for the output of the result image; the result image is the result of semantic segmentation of the street view image; (4) Using geographic information system professional software, identify the object category of each pixel in the above result image, and use statistical tools to count the frequency of trees, shrubs and grasses, and calculate the proportion of the frequency of each type of vegetation; the pixel coverage rate of the above different vegetation is the street greening characteristic index.

[0032] The steps to construct greening characteristic indicators of remote sensing images are: (1) Model preparation: download the required files of the HRNet-W48 model from the open source platform and run the test files that come with the model to test whether the model is built successfully; (2) Traverse all remote sensing images, cut them into a specified size (the size of the street view image in this patent is 512*512), and use them as model input; (3) Run the HRNet-W48 model and wait for the output of the results; the result image is the result of semantic segmentation of the remote sensing image; (4) Using geographic information system professional software to identify the category of each pixel in the above result image, and using statistical tools to merge adjacent green patches, using landscape index calculation tools to calculate the indexes of green landscape composition and spatial configuration; the above indexes are the remote sensing image greening characteristic indexes.

[0033] It should be noted here that the system provided in the above embodiment is only illustrated by the division of the above functional modules. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure is divided into different functional modules to complete all or part of the functions described above. The system is a green space cooling effect evaluation method based on remote sensing images and street view images applied to the above embodiment.

[0034] In another embodiment of the present application, a storage medium is provided, storing a program, and when the program is executed by a processor, a method for evaluating the cooling effect of green space based on remote sensing images and street view images is implemented, specifically: S1: Obtain street view image data and perform denoising and cropping; S2: Use the Deeplabv3+ semantic segmentation algorithm to identify street view image data, extract and calculate the pixel coverage of trees, grasses and shrubs, and construct green feature indicators of street view images; S3: Obtain urban satellite remote sensing images and perform radiometric calibration and correction, identify remote sensing images through the HRNet-W48 semantic segmentation algorithm, and construct greening feature indicators of remote sensing images; S4: Obtain multi-spectral remote sensing images of the city and perform cloud detection, and obtain the surface temperature of the city through surface temperature inversion; S5: Based on the surface temperature of the city, the greening characteristic indexes of street view images and remote sensing images, construct a random forest model and perform iterative training, and use the fitted random forest model to evaluate the cooling effect.

[0035] It should be understood that each part of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, any one or a combination of the following techniques well known in the art can be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0036] The above embodiments are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. Any other changes, modifications, substitutions, combinations, and simplifications made without departing from the spirit and principle of the present invention shall be equivalent replacement methods and are all included in the protection scope of the present invention.

Claims

1. A method for evaluating the cooling effect of green spaces based on remote sensing images and street view images, characterized in that, It includes the following steps: Obtain street view image data and perform denoising and cropping; Use semantic segmentation algorithm to identify the street view image data, extract and calculate the pixel coverage rates of arbors, herbs and shrubs, and construct the greening feature index of the street view image; Obtain urban satellite remote sensing images and perform radiometric calibration and correction, identify the remote sensing images through semantic segmentation algorithm, and construct the greening feature index of the remote sensing images; Obtain urban multispectral remote sensing images and perform cloud detection, and obtain the surface temperature of the city through surface temperature inversion; Based on the surface temperature of the city, and the greening feature indexes of the street view images and remote sensing images, construct a random forest model and perform iterative training, and use the fitted random forest model to evaluate the cooling effect.

2. The method for evaluating the cooling effect of green spaces based on remote sensing images and street view images according to claim 1, characterized in that, The obtaining of the street view image data and performing denoising and cropping is specifically as follows: Obtain urban road data and perform road merging and topological checking to obtain the processed urban road data; Set sampling points at a set distance interval on the processed urban road data, obtain the street view image data at the sampling points and perform denoising and cropping. Among them, one street view picture is collected for each of the four directions of up, down, left and right at each sampling point.

3. The green space cooling effect evaluation method based on remote sensing images and street view images according to claim 1, wherein The using of the semantic segmentation algorithm to identify the street view image data, extract and calculate the pixel coverage rates of arbors, herbs and shrubs, and construct the greening feature index of the street view image is specifically as follows: Use the pre-trained Deeplabv3+ object semantic segmentation model to identify the street view image data, extract the place objects, and calculate the total pixel proportion of the pixels of arbors, herbs and shrubs in the street view image data, and construct the greening feature index of the street view image.

4. The green space cooling effect evaluation method based on remote sensing images and street view images according to claim 1, wherein, The obtaining of the urban satellite remote sensing images and performing radiometric calibration and correction, identifying the remote sensing images through semantic segmentation algorithm, and constructing the greening feature index of the remote sensing images is specifically as follows: Download urban satellite remote sensing images from the satellite remote sensing platform, and perform operations of radiometric calibration, atmospheric correction and geometric correction; Use the HRNet-W48 image classification model to separate land use of the urban satellite remote sensing images and extract urban green spaces; According to the composition and spatial distribution of the urban green spaces, construct the greening feature index of the remote sensing images. The greening feature index of the remote sensing images includes a green space landscape composition index and a spatial configuration index, specifically: Landscape composition index: AREA = a ij ； Among them, AREA is the green area, a ij is the i th area of the j th patch within the PERIM = p ij ； Among them, PERIM is the perimeter of the green space, p ij is the i th perimeter of the j th patch within the th research scope; ； Among them, FRAC is an index reflecting the complexity of the plaque shape, and FRAC takes values between 1 and 2, and the closer it is to 2, the more complex the shape indicates; ； Among them, PLAND represents the relative proportion of a certain patch type in the entire landscape area, n i is the i number of patches within the A total area of the study area; NP = n i ； Among them, n i is the i number of patches within the research scope; ； Among them, PD is an index reflecting the density of plaques; ； Among them, CA represents the total area of all green space patches; ； Among them, LPI represents the proportion of the largest patch in a certain patch type occupying the entire landscape area; Spatial configuration index: ； Among them, LSI represents the standardized measurement of the landscape shape index; ； Among them, COHESION represents the connection degree of similar type patches in the landscape, m represents the total number of research ranges, and Z represents the total number of grids; a higher COHESION indicates that the patches are relatively concentrated, while a lower COHESION indicates that the patches are dispersed.

5. The method for evaluating the cooling effect of green spaces based on remote sensing images and street view images according to claim 1, wherein The obtaining of the urban multispectral remote sensing images and performing cloud detection, and obtaining the surface temperature of the city through surface temperature inversion is specifically as follows: Based on ASTER GED data, calculate the Normalized Difference Vegetation Index NDVI , as follows: ； Among them, NIR is the reflectance in the near-infrared band, Red is the reflectance in the red light band; Based on the Normalized Difference Vegetation Index NDVI , calculate the vegetation coverage FVC , as follows: ； Among them, NDVI bare and NDVI veg are the normalized difference vegetation index values of pixels with bare land and vegetation, respectively NDVI value. Based on vegetation coverage FVC Calculate the effective emissivity using the following formula: ； wherein, is the effective emissivity, and are the emissivities of vegetation and bare land respectively, and b is the given spectral band; ； Among them, LST is the surface temperature, Tb is the TOA brightness temperature of the thermal infrared radiation TIR channel, is the surface emissivity of the same channel; A i , B i and C i are algorithm coefficients determined by linear regression of radiative transfer simulation; Based on the SMW algorithm combined with TCWV and surface emissivity data, perform surface temperature inversion, specifically: Select endmembers representing different ground objects from remote sensing images, construct a spectral mixture model, calculate the endmember coefficients of each pixel, that is, the proportion of ground objects, and then output the inversion result after result verification to finally obtain the urban surface temperature distribution data.

6. The method for evaluating the cooling effect of green spaces based on remote sensing images and street view images according to claim 1, wherein Based on the greening characteristic indicators of the urban surface temperature, street view images and remote sensing images, construct a random forest model and perform iterative training, and use the well-fitted random forest model to evaluate the cooling effect, specifically: Standardize the greening characteristic indicators and the inverted surface temperature of the green space samples extracted from all remote sensing images and street view images, divide the study area into study units of a set size, each study unit is a sample, and divide all samples into training samples and validation samples according to a set ratio; In the training samples, input the greening characteristic indicators and the surface temperature into the random forest model, and perform iterative training. Optimize the hyperparameters through grid search to find the optimal parameters and generate an optimally fitted random forest model; In the validation samples, input the greening characteristic indicators and the surface temperature into the well-fitted random forest model to generate the corresponding surface temperature of the green space.

7. The green space cooling effect evaluation system based on remote sensing images and street view images is characterized in that Applied to the green space cooling effect evaluation method based on remote sensing images and street view images described in any one of claims 1-6, including a data acquisition and preprocessing module, a feature extraction and processing module, a model construction and training module, and an evaluation and result analysis module; The data acquisition and preprocessing module is used to obtain street view image data and perform denoising and cropping; The feature extraction and processing module is used to use the Deeplabv3+ semantic segmentation algorithm to identify the street view image data, extract and calculate the pixel coverage rates of trees, herbs and shrubs, and construct the greening characteristic indicators of the street view image; obtain urban satellite remote sensing images and perform radiometric calibration and correction, identify the remote sensing images through the HRNet-W48 semantic segmentation algorithm, and construct the greening characteristic indicators of the remote sensing images; obtain urban multispectral remote sensing images and perform cloud detection, and obtain the urban surface temperature through surface temperature inversion; The model construction and training module is used to construct a random forest model and perform iterative training according to the greening characteristic indicators of the urban surface temperature, street view images and remote sensing images; The evaluation and result analysis module uses the well-fitted random forest model to evaluate the cooling effect.

8. A storage medium stores a program, characterized in that: When the program is executed by a processor, it implements the green space cooling effect evaluation method based on remote sensing images and street view images described in any one of claims 1-6.

Images