Urban green land carbon sink metering method and system based on multi-source data fusion

Abstract

The invention discloses an urban green land carbon sink metering method and system based on multi-source data fusion, and relates to the technical field of environmental monitoring, and the method comprises the steps: collecting urban green land multi-source data, and carrying out the preprocessing; performing parameter completion on vegetation in the building shielding blind area by adopting a dual-channel generative adversarial network to obtain complete green land leaf area index distribution data; simulating a reflected light path of a building glass curtain wall by using a ray tracing algorithm, obtaining a photosynthetically active radiation correction coefficient received by a vegetation canopy, and identifying urban green land carbon sink distribution data through multi-scale data fusion; and obtaining a carbon sink amount error based on the urban green land carbon sink amount distribution data, performing dynamic correction through a Bayesian optimization algorithm, and generating an urban green land carbon sink amount measurement report. According to the method, the adversarial network is generated through two channels, and the cooperative training mechanism of the U-Net generator and the PatchGAN discriminator is utilized, so that the high-precision spatial continuity reconstruction of the leaf area index of the building shielding blind area is realized.

Description

Technical Field

[0001] The present invention relates to the technical field of environmental monitoring, and particularly to a method and system for measuring the carbon sink amount of urban green spaces based on multi-source data fusion. Background Art

[0002] The measurement of the carbon sink amount of urban green spaces is an important technical direction for urban ecological monitoring and carbon neutrality assessment. The current mainstream methods mainly follow the technical route of combining multi-source remote sensing data fusion with ground observations, including using high-resolution remote sensing images to invert vegetation parameters and establishing a carbon sink estimation model by combining the CO2 flux observation data of flux towers. For the areas blocked by buildings, the existing technologies mostly use the solar trajectory projection method to determine the shadow range and extract the effective vegetation area by combining morphological image processing. In terms of dynamic weight allocation, methods based on time series analysis and spatial clustering are widely used in the spatio-temporal feature modeling of carbon sink amounts, and some studies introduce machine learning algorithms to improve the accuracy of parameter inversion.

[0003] There is room for improvement in the following two aspects of the conventional methods: First, most of the existing weight allocation methods use static or semi-static models, which are difficult to fully characterize the spatio-temporal heterogeneity characteristics of urban vegetation carbon sinks, especially the dynamic response under seasonal changes and urban microclimate effects; Second, the existing carbon sink measurement systems rarely involve the enhancement effect of building glass curtain wall reflected light on vegetation photosynthesis, resulting in systematic biases in the estimation of canopy photosynthetically active radiation. Summary of the Invention

[0004] In view of the above existing problems, the present invention is proposed.

[0005] Therefore, the present invention provides a method for measuring the carbon sink amount of urban green spaces based on multi-source data fusion to solve the problems of insufficient spatio-temporal dynamic weight allocation and imperfect modeling of building reflected light effects in the prior art.

[0006] To solve the above technical problems, the present invention provides the following technical solutions: In a first aspect, the present invention provides a method for measuring the carbon sink amount of urban green spaces based on multi-source data fusion, which includes collecting multi-source data of urban green spaces and performing preprocessing; using a dual-channel generative adversarial network to complete the parameters of the vegetation in the building occlusion blind area and obtain the complete green space leaf area index distribution data; based on the complete green space leaf area index distribution data, using a spatio-temporal perception dynamic weighting network to perform dynamic weight allocation and generate a carbon sink data fusion weight matrix; using a ray tracing algorithm to simulate the reflected light path of the building glass curtain wall, obtaining the correction coefficient of the photosynthetically active radiation received by the vegetation canopy, and identifying the carbon sink amount distribution data of urban green spaces through multi-scale data fusion; obtaining the carbon sink amount error based on the carbon sink amount distribution data of urban green spaces and performing dynamic correction through the Bayesian optimization algorithm to generate a measurement report of the carbon sink amount of urban green spaces.

[0007] As a preferred solution of the method for measuring the carbon sink amount of urban green space based on multi-source data fusion according to the present invention, wherein: the dual-channel generative adversarial network is used to complement the parameters of the vegetation in the building occlusion blind area to obtain the complete green space leaf area index distribution data, and the steps are as follows. The multi-scale morphological segmentation algorithm is combined with the building projection analysis method to identify the building occlusion blind area. The U-Net generator and the PatchGAN discriminator are used to construct a dual-channel generative adversarial network, and Xavier is used to randomly initialize the convolutional layer parameters of the U-Net generator. The first channel is based on the vegetation structure parameters and environmental parameters of the unoccluded area around the blind area, and the U-Net structure generator is used to obtain the leaf area index distribution data of the blind area, so as to complement the parameters of the vegetation in the building occlusion blind area. The second channel is based on the canopy density distribution characteristics of the blind area edge. The PatchGAN discriminator is used to identify the spatial distribution difference between the leaf area index distribution data of the blind area and the canopy density distribution characteristics of the blind area edge, and the convolutional layer parameters of the U-Net structure generator are dynamically optimized through backpropagation. Through the Kriging method, the leaf area index distribution data of the blind area and the initial green space leaf area index distribution data are fused to generate the complete green space leaf area index distribution data.

[0008] As a preferred solution of the method for measuring the carbon sink amount of urban green space based on multi-source data fusion according to the present invention, wherein: the construction process of the spatio-temporal perception dynamic weighting network includes constructing the initial architecture of the spatio-temporal perception dynamic weighting network based on the bidirectional GRU layer, the 3D convolutional layer and the multi-head attention mechanism, and using the spatio-temporal parameter joint training set to initialize the hyperparameters, and finally forming the spatio-temporal perception dynamic weighting network.

[0009] As a preferred solution of the method for measuring the carbon sink amount of urban green space based on multi-source data fusion according to the present invention, wherein: the steps of generating the carbon sink data fusion weight matrix are as follows. The spatio-temporal convolution network is used to extract the spatio-temporal features of each pixel in the complete green space leaf area index distribution data. The spatio-temporal perception dynamic weighting network is used to dynamically assign weights to the spatio-temporal feature matrix of each pixel, and non-linear interaction is performed through feature concatenation and gating mechanism to generate the carbon sink data fusion weight matrix.

[0010] As a preferred solution of the method for measuring the carbon sink amount of urban green space based on multi-source data fusion according to the present invention, wherein: the steps of using the ray tracing algorithm to simulate the reflected light path of the building glass curtain wall to obtain the photosynthetically active radiation correction coefficient received by the vegetation canopy are as follows. Based on the carbon sink data fusion weight matrix, the Monte Carlo ray tracing algorithm is used to simulate the reflected light path of the building glass curtain wall, and a photosynthetically active radiation increment distribution map is generated; Perform a pixel-by-pixel ratio operation on the photosynthetically active radiation increment distribution map and the canopy photosynthetically active radiation distribution data obtained by inverting the urban vegetation spatial image to generate a photosynthetically active radiation correction coefficient.

[0011] As a preferred solution of the urban green space carbon sink volume measurement method based on multi-source data fusion according to the present invention, wherein: the steps of identifying the urban green space carbon sink volume distribution data through multi-scale data fusion are as follows. Extract convolution features from the photosynthetically active radiation correction coefficient and the complete green space leaf area index distribution data, and fuse them through a gated feature modulation network to generate a light-corrected leaf area index distribution data; Perform Gaussian pyramid decomposition on the light-corrected leaf area index distribution data, obtain the spatial feature sub-bands at each scale, and perform multi-scale fusion through ridge regression to generate the urban green space carbon sink volume distribution data.

[0012] As a preferred solution of the urban green space carbon sink volume measurement method based on multi-source data fusion according to the present invention, wherein: the steps of obtaining the carbon sink volume error and dynamically correcting it through the Bayesian optimization algorithm to generate an urban green space carbon sink volume measurement report are as follows. Adopt the Bootstrap resampling method to analyze the carbon sink volume error of each pixel; Take the carbon sink data fusion weight matrix as the spatial prior, and combine multi-source auxiliary data to construct a Gaussian process regression model to generate a prior distribution of the spatial correlation of the carbon sink volume; Take the prior distribution of the spatial correlation of the carbon sink volume as the spatial constraint, and dynamically correct the carbon sink volume error of each pixel in the urban green space carbon sink volume distribution data through the Bayesian optimization algorithm; Through Markov chain Monte Carlo, perform posterior probability sampling on the corrected urban green space carbon sink volume distribution data to generate a grid-type carbon sink volume probability distribution map; Through the spatial integration method driven by probability density, perform spatial integration on the mean values of the pixels in the grid-type carbon sink volume probability distribution map to calculate the total carbon sink volume of the urban green space and generate an urban green space carbon sink volume measurement report.

[0013] In a second aspect, the present invention provides an urban green space carbon sink volume measurement system based on multi-source data fusion, including a data acquisition module for collecting multi-source data of urban green space and performing preprocessing; A parameter completion module for using a dual-channel generative adversarial network to complete the parameters of the vegetation in the building occlusion blind area to obtain the complete green space leaf area index distribution data; A weight allocation module, which is used to perform dynamic weight allocation by using a spatio-temporal perception dynamic weighting network based on the complete green space leaf area index distribution data, and generate a carbon sink data fusion weight matrix; A carbon sink amount identification module, which is used to simulate the reflected light path of the building glass curtain wall by using the ray tracing algorithm, obtain the photosynthetically active radiation correction coefficient received by the vegetation canopy, and identify the urban green space carbon sink amount distribution data through multi-scale data fusion; A carbon sink amount distribution generation module, which is used to obtain the carbon sink amount error based on the urban green space carbon sink amount distribution data, and perform dynamic correction through the Bayesian optimization algorithm to generate an urban green space carbon sink amount measurement report.

[0014] In a third aspect, the present invention provides a computer device, including a memory and a processor, where the memory stores a computer program, and: when the computer program is executed by the processor, any step of the method for measuring the urban green space carbon sink amount based on multi-source data fusion as described in the first aspect of the present invention is implemented.

[0015] In a fourth aspect, the present invention provides a computer-readable storage medium, on which a computer program is stored, and: when the computer program is executed by the processor, any step of the method for measuring the urban green space carbon sink amount based on multi-source data fusion as described in the first aspect of the present invention is implemented.

[0016] The beneficial effects of the present invention are as follows: Through the dual-channel generative adversarial network, by using the collaborative training mechanism of the U-Net generator and the PatchGAN discriminator, combined with the three-stage supervision of the cycle consistency loss function, the high-precision spatial continuity reconstruction of the leaf area index in the building occlusion blind area is realized. Through the ray tracing algorithm, the reflected light path of the building glass curtain wall is simulated by using Monte Carlo, and combined with the multi-scale fusion of the gated feature modulation network, the enhancement effect of the reflected light on the photosynthetically active radiation of the vegetation is accurately quantified, providing a key correction basis for the calculation of the carbon sink amount in the complex urban environment. Description of the Drawings

[0017] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings required for the description of the embodiments. Obviously, the following drawings are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0018] Figure 1 It is a flowchart of the method for measuring the urban green space carbon sink amount based on multi-source data fusion.

[0019] Figure 2 It is a schematic diagram of the system for measuring the urban green space carbon sink amount based on multi-source data fusion.

[0020] Figure 3 Flow chart for complementing parameters of building occlusion blind areas

[0021] Figure 4 Flow chart for ray tracing and multi-scale fusion Specific implementation manners

[0022] To make the above objects, features and advantages of the present invention more obvious and understandable, the specific implementation manners of the present invention will be described in detail below with reference to the accompanying drawings of the specification.

[0023] Many specific details are set forth in the following description in order to fully understand the present invention, but the present invention may also be implemented in other ways different from those described herein. Those skilled in the art can make similar extensions without departing from the connotation of the present invention. Therefore, the present invention is not limited by the specific embodiments disclosed below.

[0024] Secondly, the so-called "one embodiment" or "embodiment" herein refers to a specific feature, structure or characteristic that may be included in at least one implementation manner of the present invention. The appearances of "in one embodiment" in different places in this specification do not all refer to the same embodiment, nor are they separate or alternative embodiments that exclude each other with other embodiments.

[0025] Refer to Figures 1 to 4 , which is an embodiment of the present invention. This embodiment provides a method for measuring the carbon sink amount of urban green spaces based on multi-source data fusion, including the following steps: S1. Collect multi-source data of urban green spaces and perform preprocessing; The multi-source data of urban green spaces includes vegetation structure parameters, environmental parameters, spatial image data of urban vegetation, and building three-dimensional data; It should be noted that the vegetation structure parameters are obtained by lidar scanning to obtain vegetation height, canopy density, and leaf area index; the environmental parameters are from the solar radiation intensity, air temperature and humidity, and carbon dioxide concentration recorded by the meteorological station; the spatial image data of urban vegetation is used to invert the vegetation index by high-resolution multi-spectral remote sensing images; the building three-dimensional data is obtained based on the oblique photogrammetry technology.

[0026] The preprocessing includes data cleaning, format standardization, missing value filling, and spatial registration.

[0027] Furthermore, data cleaning eliminates outliers in the vegetation structure parameters through an outlier detection algorithm; format standardization uniformly converts the environmental parameters into the CSV format under the WGS84 coordinate system; missing value filling uses the Kriging interpolation method to complement the cloud occlusion areas in the spatial image data of urban vegetation; spatial registration uses a feature point matching algorithm to align the spatial resolutions of the building three-dimensional data and the spatial image data of urban vegetation.

[0028] S2. Use a dual-channel generative adversarial network to complement the parameters of the vegetation in the building occlusion blind area and obtain the complete green leaf area index distribution data; Based on the spatial image data of urban vegetation and the 3D building data, use a multi-scale morphological segmentation algorithm combined with the building projection analysis method to identify the building occlusion blind area; Furthermore, based on the spatial image data of urban vegetation and the 3D building data, first perform a solar trajectory simulation on the 3D building data through the solar trajectory projection analysis method to identify the building shadow areas at different times. Use a multi-scale morphological segmentation algorithm to extract the vegetation coverage area based on the spatial image data of urban vegetation, and combine it with the building projection analysis method to perform a spatial overlay analysis of the building shadow area and the vegetation coverage area to identify the vegetation areas completely blocked by the building. Determine the precise boundary of the occlusion blind area through an edge detection algorithm, and finally generate a spatial distribution vector map of the building occlusion blind area, including attribute information such as the blind area location, area, and surrounding environmental characteristics.

[0029] Use a U-Net generator and a PatchGAN discriminator to construct a dual-channel generative adversarial network; Furthermore, the U-Net generator includes an encoder and a decoder structure. The encoder extracts multi-scale features through 5 layers of downsampling convolution, and the decoder realizes feature reconstruction through skip connections and transposed convolutions. The PatchGAN discriminator adopts a fully convolutional network structure and locally discriminates the structural parameters and environmental parameters of the unoccluded vegetation around the blind area through a sliding window method. Finally, a dual-channel generative adversarial network is formed.

[0030] Based on the 3D building data and the vegetation height data, train the dual-channel generative adversarial network through a cyclic consistency loss function; Furthermore, in each iteration, the dual-channel generative adversarial network receives the 3D building data as input. The U-Net generator outputs the predicted vegetation parameters, and the PatchGAN discriminator discriminates the authenticity of the leaf area index distribution data in the blind area. The cyclic consistency loss function includes three parts: the adversarial loss identifies the distribution difference between the generated data and the real data through the discriminator output; the cyclic reconstruction loss re-enters the leaf area index distribution data in the blind area into the U-Net generator to obtain the reconstruction error, and optimizes the dual-channel generative adversarial network according to the reconstruction error.

[0031] Based on the spatial image data of urban vegetation, use a vegetation index inversion algorithm to extract the distribution characteristics of the canopy density at the edge of the blind area; Further, first extract the edge features of the vegetation index image through the Canny edge detection algorithm, and obtain a single-pixel-width vegetation boundary line by combining morphological thinning processing; establish a buffer zone with a width of 3-5 pixels centered on the boundary of the building occlusion blind area, and statistically analyze the spatial distribution characteristics of the vegetation index in the buffer zone, including the mean value, coefficient of variation, and texture feature parameters; use the Gaussian filtering algorithm to smooth the edge features and generate the canopy density distribution characteristics of the blind area edge with continuous gradient characteristics. The canopy density distribution characteristics of the blind area edge include quantitative parameters such as the spatial position, density gradient, and morphological characteristics of the vegetation edge.

[0032] Use Xavier to randomly initialize the convolutional layer parameters of the U-Net generator; Further, Xavier automatically determines the reasonable distribution range of the weight parameters according to the number of input and output channels of each layer of the network to ensure that the activation values of each layer maintain stable numerical characteristics. The weight parameters are randomly generated from a normal distribution within a specific range, and the bias term is uniformly set to zero. This initialization strategy can effectively maintain the stability of gradient propagation during network training, avoid the phenomenon of numerical non-convergence in the initial stage of training, and provide a good initial state for subsequent network parameter optimization.

[0033] It should be noted that the cycle consistency loss function refers to a three-stage supervision mechanism composed of constructing a bidirectional generation path (such as forward generation and reverse reconstruction) through an adversarial training framework, identifying the differences between the reconstructed data (such as the surrounding vegetation parameter data generated by reverse reconstruction) and the original input data (such as the true values of the surrounding vegetation parameters obtained by on-site survey), and combining spatial gradient constraints (such as the detection of the consistency of the vegetation canopy edge based on the Sobel operator).

[0034] The first channel is based on the vegetation structure parameters and environmental parameters of the unobstructed vegetation around the blind area, and obtains the distribution data of the leaf area index of the blind area through the U-Net structure generator to complete the parameter filling of the vegetation in the building occlusion blind area; Further, the vegetation structure parameters and environmental parameters of the unobstructed vegetation around the blind area are subjected to multi-level feature extraction through the encoder part of the U-Net structure generator. The encoder gradually captures the hierarchical features of the vegetation structure parameters of the unobstructed vegetation around the blind area through five convolutional operations; the decoder part realizes feature reconstruction through transposed convolution and skip connections, fuses the multi-scale features extracted by the encoder with the spatial information of the corresponding levels, and finally outputs the distribution data of the leaf area index of the building occlusion blind area; the completed leaf area index distribution data maintains spatial continuity with the vegetation parameters around the blind area to ensure the ecological rationality of the reconstruction results of the vegetation parameters in the building occlusion blind area.

[0035] The second channel, based on the canopy density distribution characteristics at the edge of the blind area, uses a PatchGAN discriminator to identify the spatial distribution differences between the blind area leaf area index distribution data and the canopy density distribution characteristics at the edge of the blind area, and dynamically optimizes the convolution layer parameters of the U-Net structure generator through backpropagation; Furthermore, the discriminator performs local feature matching on the blind area leaf area index distribution data output by the U-Net structure generator through a sliding window method; the PatchGAN discriminator uses a five-layer convolution structure to gradually extract spatial features, identify the spatial distribution differences between the blind area leaf area index distribution data and the canopy density distribution characteristics at the edge of the blind area; and quantifies the degree of spatial consistency through an adversarial loss function. The spatial distribution differences are transmitted to the U-Net structure generator via the backpropagation algorithm to dynamically adjust the weight parameters of each convolution layer of the generator; the optimization process focuses on correcting the leaf area index distribution in the transition zone between the blind area and the surrounding area.

[0036] It should be noted that the convolution layer parameters of the U-Net structure generator include the downsampling convolution kernel weights of each level of the encoder, the transposed convolution kernel weights of each level of the decoder, the scaling and translation parameters of the batch normalization layer, and the skip connection weight coefficients between each level.

[0037] Through Kriging method, the blind area leaf area index distribution data and the initial green space leaf area index distribution data are fused to generate the complete green space leaf area index distribution data.

[0038] Furthermore, first, the spatial distribution characteristics of the blind area leaf area index distribution data and the initial green space leaf area index distribution data are analyzed through Kriging interpolation method to determine the correlation degree between different positions; the interpolation parameters are set including the minimum influence distance and the maximum search range, the blind area leaf area index distribution data is used as the area to be supplemented, and the initial green space leaf area index distribution data is used as the reference area; the optimal estimated value of each position is calculated through a spatial weighted average method, and the weight is determined by the data point spacing and distribution density; an incremental region expansion strategy is adopted to calculate gradually from the blind area boundary outwards to ensure the smooth connection of the transition region; the finally generated complete green space leaf area index distribution data realizes the natural transition between the blind area and the surrounding area.

[0039] It should be noted that the initial green space leaf area index distribution data is obtained through statistical analysis by analyzing vegetation indices (such as NDVI, EVI) from the direct band reflectance of multispectral satellite remote sensing images.

[0040] S3. Based on the complete green space leaf area index distribution data, a spatio-temporal perception dynamic weighted network is used for dynamic weight allocation, and a carbon sink data fusion weight matrix is generated; The spatio-temporal features of each pixel in the complete green space leaf area index distribution data are extracted through a spatio-temporal convolutional network (STCN); It should be noted that a pixel refers to the smallest spatial unit in the raster image of the complete green space leaf area index distribution data; the spatio-temporal characteristics reflect the dynamic change rules of the complete green space leaf area index distribution data in the time dimension (such as seasonal change trends) and the spatial dimension (such as urban regional differences).

[0041] Based on the bidirectional GRU layer, 3D convolutional layer, and multi-head attention mechanism, the initial architecture of the spatio-temporal perception dynamic weighting network is constructed, and the spatio-temporal parameter joint training set is used to initialize the hyperparameters, and finally the spatio-temporal perception dynamic weighting network is formed; Furthermore, the bidirectional GRU layer is responsible for capturing the dynamic change rules of the complete green space leaf area index distribution data in the time dimension, and extracting temporal features through the forward and backward processing directions; the 3D convolutional layer processes the three-dimensional data block composed of the spatial dimension and the time dimension, and uses a 3×3×3 convolutional kernel to extract local spatio-temporal features; the multi-head attention mechanism analyzes the correlation between different spatio-temporal features and calculates the dependence weight between features; the spatio-temporal parameter joint training set includes temporal leaf area index data, vegetation classification data, and multi-source auxiliary data, which are used to initialize the learning rate, batch size, and number of training epochs of the network; finally, the spatio-temporal perception dynamic weighting network is formed.

[0042] The spatio-temporal parameter joint training set includes temporal leaf area index data, spatial reference data (vegetation classification data), and multi-source auxiliary data (CO2 flux, soil carbon content, and meteorological data); It should be noted that the acquisition process of the spatio-temporal parameter joint training set is as follows: high-resolution remote sensing images of urban green spaces are collected, and temporal leaf area index data is generated through an inversion algorithm; lidar point cloud data of the same area is synchronously obtained through lidar, and a point cloud classification algorithm based on random forest is used to distinguish vegetation and non-vegetation ground objects to generate spatial reference data, that is, vegetation classification data; combined with the CO2 flux data recorded by the ground fixed observation station, soil carbon content data, and meteorological data provided by the meteorological station, multi-source auxiliary data is formed. All data is processed through spatio-temporal alignment to ensure that the temporal leaf area index data, spatial reference data, and multi-source auxiliary data are completely consistent in timestamp (UTC standard time) and spatial coordinates (WGS84 coordinate system), and finally the spatio-temporal parameter joint training set is formed.

[0043] Using the spatio-temporal perception dynamic weighting network, dynamic weight allocation is performed on the spatio-temporal feature matrix of each pixel, and non-linear interaction is performed through feature concatenation and gating mechanism to generate a carbon sink data fusion weight matrix.

[0044] Furthermore, the spatio-temporal feature matrix of each pixel in the complete green space leaf area index distribution data is first input into a bidirectional GRU layer to extract the forward and backward dependencies in the time series; the 3D convolutional layer performs a local receptive field scan on the spatio-temporal feature matrix to capture the feature change patterns of the spatial neighborhood and adjacent time; the multi-head attention mechanism calculates the correlation weights between different spatio-temporal features, focusing on the feature interaction of the seasonal changes of vegetation growth and the impact of urban microclimate; the feature concatenation operation concatenates the temporal features output by the bidirectional GRU layer, the spatial features extracted by the 3D convolutional layer, and the attention weights; the gating mechanism dynamically adjusts the contribution weights of multi-source features through a feature selection gate controlled by the Sigmoid function; the final output carbon sink data fusion weight matrix precisely quantifies the relative importance of the spatio-temporal feature matrix to the vegetation carbon sink capacity.

[0045] It should be noted that the spatio-temporal feature matrix is a spatio-temporal dimensional feature tensor extracted by integrating multi-temporal remote sensing data (such as NDVI time series) and spatial auxiliary data (such as meteorology, terrain) using the attention mechanism.

[0046] S4. Use the ray tracing algorithm to simulate the reflected light path of the building glass curtain wall, obtain the photosynthetically active radiation correction coefficient received by the vegetation canopy, and identify the distribution data of urban green space carbon sink amount through multi-scale data fusion; Based on the carbon sink data fusion weight matrix, use the Monte Carlo ray tracing algorithm to simulate the reflected light path of the building glass curtain wall and generate a photosynthetically active radiation increment distribution map; Furthermore, the glass curtain wall surface in the building three-dimensional data is divided into multiple reflection units according to the material properties, and each unit sets specific reflectivity and scattering parameters based on the optical property data of building materials; the Monte Carlo ray tracing algorithm emits a large number of randomly sampled rays from the sun position, and when the rays encounter the glass curtain wall surface, the reflection path is calculated according to the Fresnel reflection law; the intersection position of the reflected rays and the vegetation canopy is sampled according to the importance of the carbon sink data fusion weight matrix, and more ray tracing calculations are received in the high-weight area; the energy contribution of each reflected ray is quantified according to the incident angle, distance attenuation, and curtain wall reflectivity, and the reflected light energy distribution received by each position of the vegetation canopy is cumulatively generated; the final output photosynthetically active radiation increment distribution map records the radiation enhancement amount caused by the building reflected light in the form of a grid.

[0047] Perform a pixel-by-pixel ratio operation on the photosynthetically active radiation increment distribution map and the canopy photosynthetically active radiation distribution data retrieved from the urban vegetation spatial image to generate a photosynthetically active radiation correction coefficient, and the expression is: ; where is the photosynthetically active radiation correction coefficient of pixel , is the pixel of the photosynthetically active radiation increment distribution image value (building reflected light weight coefficient), is the pixel of the photosynthetically active radiation distribution in the canopy value (vegetation inversion value); It should be noted that the dimension is unified through standardization and normalization processing before calculation; Perform convolution feature extraction on the photosynthetically active radiation correction coefficient and the complete green space leaf area index distribution data, and fuse them through a gated feature modulation network to generate the light-corrected leaf area index distribution data; Furthermore, first use a dual-branch convolutional neural network to extract the spatial distribution characteristics of the photosynthetically active radiation correction coefficient (such as: the radiation hotspot distribution formed by building reflected light) and the hierarchical characteristics of the complete green space leaf area index distribution data (such as: local canopy structure); then, based on the gated unit of the Sigmoid function, use the spatial distribution characteristics as the control signal to dynamically adjust the response intensity of the hierarchical characteristics; finally, perform spatial reconstruction on the modulated hierarchical characteristics through a transposed convolutional layer to fuse and generate the light-corrected leaf area index distribution data.

[0048] Perform Gaussian pyramid decomposition on the light-corrected leaf area index distribution data, obtain the spatial feature sub-bands at each scale, and perform multi-scale fusion through ridge regression to generate the urban green space carbon sink amount distribution data.

[0049] Furthermore, first adopt the Gaussian pyramid decomposition algorithm to generate feature sub-bands with 5 scale levels through successive downsampling, and each sub-band retains the vegetation feature information with different spatial frequencies; during the decomposition process, use a Gaussian kernel with a standard deviation of 1.6 to smooth the light-corrected leaf area index distribution data, and use bilinear interpolation between adjacent scales to maintain geometric consistency; register the corresponding scales of each scale feature sub-band with the photosynthetically active radiation increment distribution map to ensure accurate spatial position correspondence; set the regularization parameter λ = 0.5 in the ridge regression fusion stage, calculate the fusion weights of each scale feature through least squares optimization, and finally generate the urban green space carbon sink amount distribution data.

[0050] S5. Obtain the carbon sink amount error based on the urban green space carbon sink amount distribution data, and perform dynamic correction through the Bayesian optimization algorithm to generate the urban green space carbon sink amount measurement report.

[0051] Based on the urban green space carbon sink amount distribution data, combined with meteorological data and the light-corrected leaf area index distribution data, use the Bootstrap resampling method to analyze the carbon sink amount error of each pixel; Furthermore, based on the urban green space carbon sink distribution data, combined with the temperature, radiation, and precipitation parameters in the meteorological data, as well as the light-corrected leaf area index distribution data, a spatial dataset is constructed; the Bootstrap resampling method is used to conduct multiple random samplings for each pixel location to generate the probability distribution of the carbon sink values; by analyzing the characteristics of the probability distribution, the error range of the carbon sink estimate value for each pixel is calculated; the final output result is the carbon sink error recorded separately for each pixel.

[0052] Taking the carbon sink data fusion weight matrix as a spatial prior, a Gaussian process regression model is constructed in combination with multivariate auxiliary data to generate the prior distribution of the spatial correlation of the carbon sink amount.

[0053] Furthermore, the carbon sink data fusion weight matrix is used as a spatial weight constraint condition and is spatially overlaid with the vegetation type, soil properties, and terrain parameters in the multivariate auxiliary data; the kernel function adopts a linear combination of the radial basis function and the carbon sink data fusion weight matrix, and the length scale parameter is set to 30 meters according to the average size of the vegetation patches; the mean function is initialized by the linear regression result of the multivariate auxiliary data, the covariance matrix is determined through spatial autocorrelation analysis, and the conjugate gradient method is used for training to maximize the marginal likelihood function; finally, the prior distribution of the spatial correlation of the carbon sink amount is generated.

[0054] Taking the prior distribution of the spatial correlation of the carbon sink amount as a spatial constraint, the carbon sink error of each pixel in the urban green space carbon sink distribution data is dynamically corrected through the Bayesian optimization algorithm; Furthermore, through the Bayesian optimization algorithm, the error is gradually corrected using the information of neighboring pixels; the correction amplitude is adjusted according to the carbon sink weight in each iteration, and finally, the carbon sink error of each pixel in the urban green space carbon sink distribution data is corrected; among them, the spatial constraint refers to the condition that restricts the data distribution using the known spatial correlation law. For example, in the analysis of vegetation carbon sinks, the carbon sink amounts of adjacent pixels usually have similarity, and this spatial autocorrelation characteristic constitutes a spatial constraint.

[0055] It should be noted that the spatial constraint refers to using the correlation of geospatial data (such as the similarity of adjacent area attributes). For example, in the correction of the carbon sink amount, it is forcibly required that the difference between the corrected pixel value and the weighted average value of its surrounding pixels does not exceed the standard range (such as ±15%) to maintain spatial continuity.

[0056] Through Markov chain Monte Carlo, posterior probability sampling is performed on the corrected urban green space carbon sink distribution data to generate a grid-type carbon sink probability distribution map.

[0057] Furthermore, taking the corrected urban green space carbon sink distribution data as the initial state, a Markov chain with spatial correlation as the transition rule is constructed (for example, taking the corrected carbon sink pixel value as the state node, when setting the transition probability matrix, making the carbon sink value transition probability between adjacent pixels positively correlated with their spatial weights, so as to simulate the spatial dynamic evolution process of carbon sink through Markov chain Monte Carlo (MCMC)), and the sampling step size is set; each iteration is based on the Metropolis-Hastings criterion, taking the prior distribution of carbon sink spatial correlation as the proposal distribution to obtain the acceptance probability of the new state; during the sampling process, the spatial autocorrelation characteristics between pixels are mainly maintained (such as: the carbon sink of adjacent vegetation pixels shows a similar change trend), and intensive sampling is carried out on high-error areas; finally, a grid-type carbon sink probability distribution map is generated.

[0058] Through the spatial integration method driven by probability density, the mean value of each pixel in the grid-type carbon sink probability distribution map is spatially integrated to calculate the total urban green space carbon sink, and the expression is: ; Among them, is the total urban green space carbon sink, is the total number of pixels participating in the calculation of the total carbon sink, is the th pixel's carbon sink mean value, is the dynamic weight coefficient driven by probability density (the value range is: 0.3 - 0.7), is the th pixel's carbon sink standard deviation, is a very small constant, is the th pixel's area.

[0059] Based on the carbon sink error, the total urban green space carbon sink, the grid-type carbon sink probability distribution map, and the urban green space carbon sink distribution data, an urban green space carbon sink measurement report is generated.

[0060] This embodiment also provides an urban green space carbon sink measurement system based on multi-source data fusion, including: a data acquisition module for acquiring and preprocessing multi-source data of urban green space; a parameter completion module for using a dual-channel generative adversarial network to complete the parameters of the vegetation in the building occlusion blind area and obtain the complete green space leaf area index distribution data; a weight allocation module for dynamically allocating weights based on the complete green space leaf area index distribution data by using a spatio-temporal perception dynamic weighting network and generating a carbon sink data fusion weight matrix; A carbon sink quantity identification module, which is used to simulate the reflected light path of a building glass curtain wall using a ray tracing algorithm, obtain the correction coefficient of photosynthetically active radiation received by the vegetation canopy, and identify the distribution data of urban green space carbon sink quantity through multi-scale data fusion; A carbon sink quantity distribution generation module, which is used to obtain the carbon sink quantity error based on the urban green space carbon sink quantity distribution data, and perform dynamic correction through a Bayesian optimization algorithm to generate an urban green space carbon sink quantity measurement report.

[0061] This embodiment also provides a computer device, which is applicable to the situation of the urban green space carbon sink quantity measurement method based on multi-source data fusion, including: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the urban green space carbon sink quantity measurement method based on multi-source data fusion proposed in the above embodiment.

[0062] This computer device can be a terminal. The computer device includes a processor, a memory, a communication interface, a display screen, and an input device connected through a system bus. Among them, the processor of this computer device is used to provide computing and control capabilities. The memory of this computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of this computer device is used to communicate with an external terminal in a wired or wireless manner. The wireless manner can be achieved through WIFI, a carrier network, NFC (Near Field Communication), or other technologies. The display screen of this computer device can be a liquid crystal display screen or an electronic ink display screen. The input device of this computer device can be a touch layer covered on the display screen, or a button, a trackball, or a touchpad set on the outer shell of the computer device, or an external keyboard, a touchpad, or a mouse, etc.

[0063] This embodiment also provides a storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the method for measuring the carbon sink amount of urban green space based on multi-source data fusion proposed in the above embodiment; the storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (Static Random Access Memory, abbreviated as SRAM), electrically erasable programmable read-only memory (Electrically Erasable Programmable Read-Only Memory, abbreviated as EEPROM), erasable programmable read-only memory (Erasable Programmable Read Only Memory, abbreviated as EPROM), programmable read-only memory (Programmable Red-Only Memory, abbreviated as PROM), read-only memory (Read-Only Memory, abbreviated as ROM), magnetic memory, flash memory, magnetic disk or optical disc.

[0064] In summary, the present invention realizes the high-precision spatial continuity reconstruction of the leaf area index in the building occlusion blind area through: a dual-channel generative adversarial network, using the collaborative training mechanism of the U-Net generator and the PatchGAN discriminator, combined with the three-stage supervision of the cycle consistency loss function. Through the ray tracing algorithm, the Monte Carlo simulation is used to trace the reflected light path of the building glass curtain wall, combined with the multi-scale fusion of the gated feature modulation network, and the enhancement effect of the reflected light on the photosynthetically active radiation of vegetation is accurately quantified, providing a key correction basis for the calculation of the carbon sink amount in the complex urban environment.

[0065] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to the preferred embodiments, those of ordinary skill in the art should understand that the technical solutions of the present invention can be modified or equivalently replaced without departing from the spirit and scope of the technical solutions of the present invention, and they should all be covered by the scope of the claims of the present invention.

Claims

1. A method for measuring the carbon sink of urban green space based on multi-source data fusion, characterized in that: Including Collecting multi-source data of urban green spaces and preprocessing them; the multi-source data of urban green spaces include vegetation structure parameters, environmental parameters, spatial image data of urban vegetation, and 3D building data Using a dual-channel generative adversarial network to complete the parameter complementation of vegetation in the building occlusion blind area and obtain the complete green space leaf area index distribution data Based on the complete green space leaf area index distribution data, using a spatio-temporal perception dynamic weighting network to perform dynamic weight allocation and generate a carbon sink data fusion weight matrix Using a ray tracing algorithm to simulate the reflected light path of building glass curtain walls, obtaining the photosynthetically active radiation correction coefficient received by the vegetation canopy, and identifying the urban green space carbon sink amount distribution data through multi-scale data fusion Based on the urban green space carbon sink amount distribution data, obtaining the carbon sink amount error and dynamically correcting it through a Bayesian optimization algorithm to generate an urban green space carbon sink amount measurement report 2. The method for measuring the carbon sink amount of urban green space based on multi-source data fusion according to claim 1, wherein: The steps of using a dual-channel generative adversarial network to complete the parameter complementation of vegetation in the building occlusion blind area and obtain the complete green space leaf area index distribution data are as follows Using a multi-scale morphological segmentation algorithm combined with a building projection analysis method to identify the building occlusion blind area Using a U-Net generator and a PatchGAN discriminator to construct a dual-channel generative adversarial network, and randomly initializing the convolutional layer parameters of the U-Net generator using Xavier The first channel, based on the vegetation structure parameters and environmental parameters of the unoccluded area around the blind area, uses the U-Net structure generator to obtain the leaf area index distribution data of the blind area and complete the parameter complementation of the vegetation in the building occlusion blind area The second channel, based on the canopy density distribution characteristics of the blind area edge, uses the PatchGAN discriminator to identify the spatial distribution difference between the leaf area index distribution data of the blind area and the canopy density distribution characteristics of the blind area edge, and dynamically optimizes the convolutional layer parameters of the U-Net structure generator through backpropagation Through Kriging method, fusing the leaf area index distribution data of the blind area and the initial green space leaf area index distribution data to generate the complete green space leaf area index distribution data 3. The method for measuring the urban green space carbon sink amount based on multi-source data fusion according to claim 1, wherein: The construction process of the spatio-temporal perception dynamic weighting network includes constructing an initial architecture of the spatio-temporal perception dynamic weighting network based on a bidirectional GRU layer, a 3D convolutional layer, and a multi-head attention mechanism, and initializing the hyperparameters using a spatio-temporal parameter joint training set, and finally forming the spatio-temporal perception dynamic weighting network 4. The method for measuring the urban green space carbon sink amount based on multi-source data fusion according to claim 3, wherein: The steps of generating the carbon sink data fusion weight matrix are as follows Extracting the spatio-temporal features of each pixel in the complete green space leaf area index distribution data through a spatio-temporal convolutional network Using a spatio-temporal perception dynamic weighting network to perform dynamic weight allocation on the spatio-temporal feature matrix of each pixel, and performing non-linear interaction through feature concatenation and gating mechanism to generate a carbon sink data fusion weight matrix 5. The method for measuring the urban green space carbon sink amount based on multi-source data fusion according to claim 1, wherein: The steps of using a ray tracing algorithm to simulate the reflected light path of building glass curtain walls and obtaining the photosynthetically active radiation correction coefficient received by the vegetation canopy are as follows Based on the carbon sink data fusion weight matrix, using the Monte Carlo ray tracing algorithm to simulate the reflected light path of building glass curtain walls and generating a photosynthetically active radiation increment distribution map Perform pixel-by-pixel ratio operation on the photosynthetically active radiation increment distribution map and the canopy photosynthetically active radiation distribution data obtained by inverting the spatial image of urban vegetation to generate a photosynthetically active radiation correction coefficient.

6. The method for measuring the urban green space carbon sink amount based on multi-source data fusion according to claim 5, wherein: The steps of identifying the urban green space carbon sink amount distribution data through multi-scale data fusion are as follows. Extract convolution features from the photosynthetically active radiation correction coefficient and the complete green space leaf area index distribution data, and fuse them through a gated feature modulation network to generate a light-corrected leaf area index distribution data. Perform Gaussian pyramid decomposition on the light-corrected leaf area index distribution data to obtain spatial feature sub-bands at each scale, and perform multi-scale fusion through ridge regression to generate urban green space carbon sink amount distribution data.

7. The method for measuring the carbon sink volume of urban green space based on multi-source data fusion according to claim 5, wherein: The steps of obtaining the carbon sink amount error and dynamically correcting it through the Bayesian optimization algorithm to generate an urban green space carbon sink amount measurement report are as follows. Use the Bootstrap resampling method to calculate the carbon sink amount error of each pixel. Take the carbon sink data fusion weight matrix as a spatial prior, and combine multi-source auxiliary data to construct a Gaussian process regression model to generate a prior distribution of the spatial correlation of the carbon sink amount. Take the prior distribution of the spatial correlation of the carbon sink amount as a spatial constraint, and dynamically correct the carbon sink amount error of each pixel in the urban green space carbon sink amount distribution data through the Bayesian optimization algorithm. Through Markov chain Monte Carlo, perform posterior probability sampling on the corrected urban green space carbon sink amount distribution data to generate a raster-type carbon sink amount probability distribution map. Through the spatial integration method driven by probability density, perform spatial integration on the mean values of each pixel in the raster-type carbon sink amount probability distribution map to calculate the total carbon sink amount of urban green space and generate an urban green space carbon sink amount measurement report.

8. A measurement system for urban green space carbon sink based on multi-source data fusion, based on the measurement method for urban green space carbon sink based on multi-source data fusion according to any one of claims 1 to 7, characterized in that: Including A data acquisition module for collecting multi-source data of urban green space and performing preprocessing. A parameter completion module for using a dual-channel generative adversarial network to complete the parameters of the vegetation in the building occlusion blind area and obtain the complete green space leaf area index distribution data. A weight allocation module for dynamically allocating weights based on the complete green space leaf area index distribution data using a spatio-temporal perception dynamic weighting network and generating a carbon sink data fusion weight matrix. A carbon sink amount identification module for using a ray tracing algorithm to simulate the reflected light path of building glass curtain walls, obtaining the photosynthetically active radiation correction coefficient received by the vegetation canopy, and identifying the urban green space carbon sink amount distribution data through multi-scale data fusion. A carbon sink amount distribution generation module for obtaining the carbon sink amount error based on the urban green space carbon sink amount distribution data and dynamically correcting it through the Bayesian optimization algorithm to generate an urban green space carbon sink amount measurement report.

9. A computer device, comprising a memory and a processor, the memory storing a computer program, characterized in that: When the processor executes the computer program, it implements the steps of the method for measuring the urban green space carbon sink amount based on multi-source data fusion according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by the processor, it implements the steps of the method for measuring the urban green space carbon sink amount based on multi-source data fusion according to any one of claims 1 to 7.

Images