An object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method

Abstract

The application discloses an object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method and belongs to the technical field of remote sensing information processing. In view of the low FPAR inversion precision and detail loss problem caused by mixed pixels and strong spatial heterogeneity in the broken area of the existing urban landscape, the application obtains homogeneous object patches through object-oriented image segmentation; independent spectral-FPAR linear mapping models are constructed for various objects including vegetation and non-vegetation; a scale consistency constraint equation based on area weighting is established; and a step-by-step strategy of fixing non-vegetation parameters and then solving vegetation parameters is adopted to perform distributed parameter optimization, so that a high-resolution FPAR product is finally generated. The method significantly improves the spatial details and precision of FPAR inversion in the complex urban environment and enhances the stability of model calculation.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing information processing technology, and in particular relates to an object-oriented method for downscaling the photosynthetically active radiation absorption ratio of urban vegetation. Background Technology

[0002] Photosynthetically Active Radiation Absorption Ratio (FPAR) is a key biophysical parameter characterizing vegetation photosynthetic capacity and carbon cycling in terrestrial ecosystems. High-precision FPAR data is crucial for ecological monitoring and carbon sink estimation. Currently, mainstream global FPAR products, such as MODIS FPAR, while possessing high temporal resolution, typically have coarse spatial resolution, typically around 500 meters or 1 kilometer. In highly fragmented and spatially heterogeneous urban environments, such coarse-resolution data struggles to effectively capture subtle vegetation distribution characteristics and physiological differences within built-up areas, parks, and roadside green spaces. Therefore, utilizing higher-resolution satellite imagery, such as Landsat or Sentinel data, combined with coarse-resolution products for downscaling and inversion has become the primary technological approach for acquiring refined FPAR data for urban areas.

[0003] However, while existing downscaling methods are effective in homogeneous natural forests or farmland, they exhibit significant technical limitations when dealing with complex urban surface environments. First, urban surfaces are highly mixed with various land features such as buildings, roads, water bodies, and green spaces, resulting in the vast majority of low-to-medium resolution remote sensing pixels being mixed pixels. Existing techniques typically establish a uniform global or regional average mapping relationship between coarse and fine resolution spectral reflectance and FPAR across the entire study area or a large sliding window. This method, based on macroscopic statistical averaging, cannot adapt to the intense and localized spatial heterogeneity within cities. In fragmented landscape areas where buildings and vegetation are interspersed, it easily produces severe smoothing effects, leading to a significant loss of spatial detail in the inversion results and failing to accurately depict the true FPAR of street trees, small green spaces, and other targets. Second, traditional FPAR calculation and downscaling processes often ignore or simply remove the contributions of non-vegetated areas such as water bodies and buildings. However, non-vegetated features constitute a considerable proportion of urban mixed pixels, possessing their own non-negligible background reflectance signals, while potentially containing weak vegetation signals. Ignoring this contribution leads to a systematic bias in the estimation of the actual vegetation FPAR within a pixel. Furthermore, to describe mixed scenes of multiple land cover types in such complex environments, numerous unknown parameters need to be introduced into the downscaling physical model. Moreover, there are correlations between the spectral characteristics of various land cover types, which easily leads to multicollinearity among variables. This causes the inversion equations to exhibit ill-conditioned characteristics, parameter estimation to be extremely unstable, and the solution process to easily get trapped in local optima, ultimately affecting the robustness and universality of the model.

[0004] The root cause of these problems lies in the fact that existing methods have failed to effectively address the fragmented nature of urban landscapes and the complexity of the internal composition of mixed pixels. The main difficulties in solving these problems lie in how to finely separate the radiation contributions of different land cover types within a pixel, how to establish a mapping model that can adapt to unique local landscape compositions rather than global averages, and how to achieve robust and accurate solution of model parameters under conditions of multivariates and ill-conditioned equations. These challenges restrict the production and application of high-precision urban vegetation FPAR products. Summary of the Invention

[0005] To address the aforementioned technical problems, this invention proposes an object-oriented method for downscaling the photosynthetically active radiation absorption ratio of urban vegetation, thereby resolving the issues present in the prior art.

[0006] Firstly, to achieve the above objectives, the present invention provides an object-oriented method for downscaling the photosynthetically active radiation absorption ratio of urban vegetation, comprising the following steps:

[0007] S1. Acquire time-matched coarse-resolution FPAR data and high-resolution multispectral remote sensing images within the study area, and perform projection transformation, resampling, and cropping preprocessing.

[0008] S2. Perform object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image, and classify the objects into multiple land cover categories, including vegetation and non-vegetation.

[0009] S3. For each land cover category, establish a linear mapping model between its high-resolution surface reflectance and photosynthetically active radiation absorption ratio, as shown in the following formula:

[0010]

[0011] For FPAR calculated based on fine-scale images; Functional relationship between specific categories and time; In fine-scale images, the first Average reflectance of objects in each wavelength band; It is a constant term in a linear model. These are the coefficients of the k-th band in the linear model. This represents the number of bands in the image that are used in the calculation;

[0012] S4. Based on the area-weighted scale transformation model, construct the constraint relationship between the coarse-resolution FPAR data and the fine-scale photosynthetically active radiation absorption ratio obtained from S3. The formula for the area-weighted scale transformation model is as follows:

[0013]

[0014] It is the FPAR value of the coarse resolution pixel. It represents the proportion of each object category to the area of ​​a MODIS pixel. This is the total number of categories;

[0015] S5. Based on the constraint relationship constructed in S4, a step-by-step strategy is adopted to perform distributed optimization to solve the parameters of the linear mapping model described in S3. First, the parameters of non-vegetation categories are determined; then, based on the fixed parameters of the non-vegetation categories, the parameters of the vegetation categories are solved.

[0016] S6. Using the parameters obtained from S5, calculate the fine-scale photosynthetically active radiation absorption ratio of all objects in the entire study area through the linear mapping model in S3, and generate high-resolution FPAR products.

[0017] Optionally, in S2, the process of performing object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image includes:

[0018] Using a remote sensing trend surface framework, the image spectral features are segmented to generate spectrally homogeneous object patches;

[0019] The land cover categories include at least trees, impermeable surfaces, water bodies, and arable land.

[0020] Optionally, in S4, the process of constructing area-weighted constraints includes:

[0021] The FPAR value of a coarse-resolution pixel is equal to the area-weighted sum of the FPAR values ​​of all fine-scale objects within its spatial range, which are calculated by the linear mapping model in S3.

[0022] Optionally, in S5, the process of using a step-by-step strategy for distributed optimization includes:

[0023] The constraints constructed by S4 are converted into matrix equations, and iterative optimization is performed using the least squares method based on gradient descent. The parameter set with the highest determination coefficient is selected as the final model parameters.

[0024] Optionally, in S5, during the distributed optimization process using a step-by-step strategy, the samples used must satisfy the following conditions:

[0025] In the coarse-resolution pixels corresponding to the sample, the area coverage of the vegetation category is greater than a preset threshold, and the number of samples is three times or more the number of parameters to be determined, wherein the preset threshold is 10%.

[0026] Optionally, in S5, the process of determining the parameters of the non-vegetation category includes:

[0027] We prioritize selecting pure pixel samples containing only a single land cover category for parameter fitting; if there are no pure pixel samples of the corresponding category, we use pixel samples containing mixed land cover categories for parameter fitting, and determine the final parameters based on the minimum root mean square error.

[0028] Secondly, the present invention also provides an object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling system, used to implement an object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method, the system comprising:

[0029] The data acquisition and preprocessing module is used to acquire time-matched coarse-resolution FPAR data and high-resolution multispectral remote sensing images within the study area, and to perform projection transformation, resampling, and cropping preprocessing.

[0030] The object segmentation and classification module is used to perform object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image and classify the objects into multiple land cover categories, including vegetation and non-vegetation.

[0031] The fine-scale modeling module is used to establish a linear mapping model between high-resolution surface reflectance and photosynthetically active radiation absorption ratio for each land cover category.

[0032] The scale transformation modeling module is used to construct an area-weighted constraint relationship between the coarse-resolution FPAR data and the fine-scale photosynthetically active radiation absorption ratio output by the fine-scale modeling module, based on the principle of scale consistency.

[0033] The parameter optimization solution module is used to perform distributed optimization of the parameters of the linear mapping model in the fine-scale modeling module based on the constraint relationship constructed by the scale transformation modeling module, using a step-by-step strategy. Specifically, the parameters of non-vegetation categories are determined first; then, based on the fixed parameters of the non-vegetation categories, the parameters of the vegetation categories are solved.

[0034] The product generation module is used to calculate the fine-scale photosynthetically active radiation absorption ratio of all objects in the entire study area using the parameters obtained by the parameter optimization solution module and the linear mapping model in the fine-scale modeling module, and generate high-resolution FPAR products.

[0035] Thirdly, the present invention also provides a computer terminal device, comprising:

[0036] One or more processors;

[0037] A memory, coupled to the processor, for storing one or more programs;

[0038] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method in the first aspect above.

[0039] Fourthly, the present invention also provides a computer-readable storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the steps of the object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method described in the first aspect above are implemented.

[0040] Fifthly, the present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method described in the first aspect above.

[0041] Compared with the prior art, the present invention has the following advantages and technical effects:

[0042] This invention provides an object-oriented method for downscaling the photosynthetically active radiation absorption ratio (FPAR) of urban vegetation, effectively overcoming the limitations of existing technologies in complex urban environments. Through object-oriented image segmentation and classification, this invention can finely characterize fragmented urban landscape units, establishing independent spectral-FPAR mapping relationships for various vegetation and non-vegetation features. Combined with area-weighted scale consistency constraints, it significantly improves the spatial detail and overall accuracy of FPAR inversion. This invention explicitly models and separates the radiation contribution of non-vegetation features in mixed pixels, effectively eliminating the resulting systematic estimation bias. Furthermore, a step-by-step strategy of fixing background parameters before solving for vegetation parameters, along with an iterative optimization algorithm for distributed parameter estimation, enhances the model's robustness and stability in the face of multivariate ill-conditioned equations, ultimately achieving reliable generation of high-precision, high-resolution urban vegetation FPAR products. Attached Figure Description

[0043] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0044] Figure 1 This is an overall flowchart of the object-oriented urban vegetation FPAR downscaling method according to an embodiment of the present invention. Detailed Implementation

[0045] It should be noted that, unless otherwise specified, the embodiments and features described in the present invention can be combined with each other. The present invention will now be described in detail with reference to the accompanying drawings and embodiments.

[0046] It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions, and although a logical order is shown in the flowchart, in some cases the steps shown or described may be executed in a different order than that shown here.

[0047] Example 1

[0048] like Figure 1 As shown, this embodiment provides an object-oriented method for downscaling the photosynthetically active radiation absorption ratio of urban vegetation, including:

[0049] S1. Acquire time-matched coarse-resolution FPAR data and high-resolution multispectral remote sensing images within the study area, and perform projection transformation, resampling, and cropping preprocessing.

[0050] S2. Perform object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image, and classify the objects into multiple land cover categories, including vegetation and non-vegetation.

[0051] S3. For each land cover category, establish a linear mapping model between its high-resolution surface reflectance and photosynthetically active radiation absorption ratio, as shown in the following formula:

[0052]

[0053] For FPAR calculated based on fine-scale images; Functional relationship between specific categories and time; In fine-scale images, the first Average reflectance of objects in each wavelength band; It is a constant term in a linear model. These are the coefficients of the k-th band in the linear model. This represents the number of bands in the image that are used in the calculation;

[0054] S4. Based on the area-weighted scale transformation model, construct the constraint relationship between the coarse-resolution FPAR data and the fine-scale photosynthetically active radiation absorption ratio obtained from S3. The formula for the area-weighted scale transformation model is as follows:

[0055]

[0056] It is the FPAR value of the coarse resolution pixel. It represents the proportion of each object category to the area of ​​a MODIS pixel. This is the total number of categories;

[0057] S5. Based on the constraint relationship constructed in S4, a step-by-step strategy is adopted to perform distributed optimization to solve the parameters of the linear mapping model described in S3. First, the parameters of non-vegetation categories are determined; then, based on the fixed parameters of the non-vegetation categories, the parameters of the vegetation categories are solved.

[0058] S6. Using the parameters obtained from S5, calculate the fine-scale photosynthetically active radiation absorption ratio of all objects in the entire study area through the linear mapping model in S3, and generate high-resolution FPAR products.

[0059] As one implementation method in this embodiment, in S2, the process of performing object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image includes:

[0060] Using a remote sensing trend surface framework, the image spectral features are segmented to generate spectrally homogeneous object patches;

[0061] The land cover categories include at least trees, impermeable surfaces, water bodies, and arable land.

[0062] As one implementation method in this embodiment, in S4, the process of constructing the area-weighted constraint relationship includes:

[0063] The FPAR value of a coarse-resolution pixel is equal to the area-weighted sum of the FPAR values ​​of all fine-scale objects within its spatial range, which are calculated by the linear mapping model in S3.

[0064] As one implementation method in this embodiment, in S5, the process of performing distributed optimization solution using a step-by-step strategy includes:

[0065] The constraints constructed by S4 are converted into matrix equations, and iterative optimization is performed using the least squares method based on gradient descent. The parameter set with the highest determination coefficient is selected as the final model parameters.

[0066] As one implementation method in this embodiment, in S5, during the distributed optimization solution process using a step-by-step strategy, the samples used satisfy the following conditions:

[0067] In the coarse-resolution pixels corresponding to the sample, the area coverage of the vegetation category is greater than a preset threshold, and the number of samples is three times or more the number of parameters to be determined, wherein the preset threshold is 10%.

[0068] As one implementation method in this embodiment, in S5, the process of determining the parameters of the non-vegetation category includes:

[0069] We prioritize selecting pure pixel samples containing only a single land cover category for parameter fitting; if there are no pure pixel samples of the corresponding category, we use pixel samples containing mixed land cover categories for parameter fitting, and determine the final parameters based on the minimum root mean square error.

[0070] Specifically, this invention provides an object-oriented method for downscaling and inverting the photosynthetically active radiation absorption ratio (FPAR) of urban vegetation. Addressing the high-precision FPAR calculation problem caused by strong spatial heterogeneity and landscape fragmentation in urban areas, and resulting in mixed pixels, this method proposes a downscaling solution. To address the uncertainties and information gaps in the scale transformation process, this method is based on the Remote Sensing Trend Surface (RSTS) framework. It employs an object-oriented modeling approach to construct independent spectral-FPAR mapping relationships for multiple land cover patches and establishes an area-weighted scale consistency constraint equation. Combined with a distributed parameter estimation strategy, this improves the cross-scale parameter instability problem under complex land cover confusion, achieving the production of high-precision, high-resolution FPAR products. The specific scheme includes:

[0071] Step 1: Acquire multi-source remote sensing data of similar acquisition time within the study area, including coarse-resolution FPAR products (as reference data, such as MODIS FPAR) and high-resolution multispectral imagery (as input data, such as Landsat 8 OLI). Perform preprocessing on all data, such as projection transformation, resampling, and cropping, to ensure spatial reference consistency.

[0072] Step 2: Object-Oriented Image Segmentation and Land Feature Classification. To address the mixed pixel problem caused by fragmented urban landscapes, this step utilizes the Remote Sensing Trend Surface (RSTS) framework to establish a mapping relationship between coarse-resolution FPAR products and high-resolution surface reflectance. First, an object-oriented approach is used to segment the high-resolution multispectral satellite imagery at multiple scales, generating object patches with spectral homogeneity. The segmented objects are then classified into different land feature categories, primarily including: trees, impermeable surfaces, water bodies, and cultivated land.

[0073] Step 3: Construct a fine-scale photosynthetically active radiation absorptive ratio (FPAR) mapping model. For each homogeneous object, including buildings, water bodies, etc., establish a linear functional relationship between high-resolution surface reflectance and FPAR. For a specific time period (t) and a specific land cover category (c), construct the following linear model:

[0074]

[0075] For FPAR calculated based on fine-scale images; Functional relationship between specific categories and time; In fine-scale images, the first Average reflectance of objects in each wavelength band; It is a constant term in a linear model. These are the coefficients of the k-th band in the linear model. This represents the number of bands in the image that are used in the calculation.

[0076] Step 4: Construct an area-weighted scale transformation model. Based on the principle of scale consistency, establish the correlation between coarse-resolution FPAR and fine-scale FPAR, that is, the FPAR value of a coarse-resolution pixel is equal to the area-weighted average of the FPAR values ​​of all fine-scale objects within its coverage area:

[0077]

[0078] It is the FPAR value of the coarse resolution pixel. It represents the proportion of each object category to the area of ​​a MODIS pixel. This represents the total number of categories.

[0079] Substituting the linear model from step 3 into the above equation, we obtain the final downscaling simultaneous equations to be solved:

[0080]

[0081] Step 5: Construct the matrix equation for parameter solving. Considering the heterogeneity among city pixels, unlike the traditional method of averaging similar features across the entire study area, this invention uses a single MODIS pixel as the calculation unit to calculate the average reflectance of each feature within the object range of each MODIS pixel. To utilize batch samples for solving, the function model is converted into matrix form:

[0082]

[0083] yes The matrix represents the FPAR sample values ​​of the L coarse-resolution pixels involved in the calculation; for A matrix representing the area proportion of each land cover category in each sample; yes The coefficient matrix to be determined; yes The three-dimensional tensor represents the fine-scale reflectivity matrix.

[0084] Step 6: Distributed parameter optimization solution. A combination of step-by-step strategy and iterative optimization is used to solve for the coefficient matrix. To overcome the ambiguity problem caused by mixed pixels, only coarse-resolution pixels with a tree coverage greater than 10% were selected as modeling samples, ensuring that the sample size was at least three times the number of parameters to be determined. In the parameter solution stage, the least squares method based on gradient descent was used for parameter fitting. To avoid local optima, 20 independent optimization runs were performed on each parameter set, and the final determination coefficients (...) were selected. The highest set of parameters is used as the final model configuration.

[0085] Step 601: Fix Background Values. Using "clean pixel" samples containing only a single land cover type, prioritize determining the model parameters for non-tree features (impermeable surfaces, water bodies, and farmland). Select the results with the smallest root mean square error (RMSE) as the fixed parameters for these land cover features.

[0086] Step 602: Solve for vegetation parameters. Based on the fixed non-tree parameters, solve for the undetermined parameters of trees using mixed pixel samples.

[0087] Step 7: Generate high-resolution FPAR products. Utilize the optimal parameter matrix obtained in Step 6. Substitute the values ​​into the linear model from step 3, calculate the high-resolution FPAR values ​​for the entire study area for each object, and finally synthesize a complete fine distribution map of urban vegetation FPAR.

[0088] Based on this, the present invention provides an object-oriented method for downscaling the photosynthetically active radiation absorption ratio of urban vegetation, the beneficial effects of which include the following:

[0089] This invention significantly improves the accuracy and spatial detail representation of FPAR inversion in highly fragmented urban landscapes. This advantage primarily stems from steps 4 and 5. Existing techniques typically establish a uniform mapping by averaging across the entire study area or a large sliding window. This macroscopic averaging is ill-suited to the severe spatial heterogeneity of urban areas. This invention reduces the computational unit to a single coarse-resolution pixel (e.g., a MODIS pixel) and calculates a weighted average of all fine-scale land cover objects within that pixel. This localized computational approach accurately captures the unique landscape composition within each pixel, effectively improving upon the smoothing effect and detail loss issues that often occur with traditional methods in highly fragmented urban environments.

[0090] This invention effectively eliminates the mixed-cell error caused by the obfuscation of non-vegetation features, improving model accuracy. This advantage mainly stems from steps 2, 3, and 6. Existing technologies often fail to consider the FPAR contribution of non-vegetation areas such as water bodies and buildings, or simply exclude them. However, in urban mixed-cell data, the background reflection of these features cannot be ignored. This invention constructs independent linear mapping models not only for trees but also for impermeable surfaces, water bodies, and other non-vegetation features. By explicitly including the contribution of these non-vegetation features in the model and separating them from the total signal, this invention can more accurately restore the true FPAR value of vegetation, significantly reducing the systematic bias caused by mixed-cell data.

[0091] This enhances the robustness of model parameter calculation and reduces the occurrence of local optima. This advantage mainly stems from step 6. Downscaling models for multiple land cover types involve numerous unknown parameters, and direct solution can easily lead to ill-conditioned equations. This invention adopts a step-by-step strategy: first, non-vegetation background parameters are fixed using pixels covering only a single category, and then vegetation parameters are solved in mixed pixels, reducing the dimensionality of the unknowns. Simultaneously, multiple iterative optimizations using gradient descent-based least squares ensure that the final parameters are globally optimal. This strategy enables the algorithm to maintain the stability and reliability of its output results even when facing complex and ever-changing urban environmental data.

[0092] Example 2

[0093] In this embodiment, a computer terminal device is provided, including:

[0094] One or more processors;

[0095] A memory, coupled to the processor, for storing one or more programs;

[0096] When the one or more programs are executed by the one or more processors, the one or more processors implement the steps of the above-described object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method.

[0097] In this embodiment, a computer-readable storage medium is also provided, on which a computer program is stored. When the computer program is executed by a processor, it implements the steps of the above-described object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method.

[0098] In this embodiment, an electronic device is also provided, including a memory and a processor. The memory stores a computer program, and the processor is configured to run the computer program to perform the steps of the above-described object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method.

[0099] In this embodiment, a computer program product is also provided, including a computer program that, when executed by a processor, implements the steps of the above-described object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling method.

[0100] The aforementioned program can run on a processor or be stored in memory (or a computer-readable medium). Computer-readable media includes both permanent and non-permanent, removable and non-removable media, and information storage can be achieved by any method or technology. Information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase-change memory (PRAM), static random-access memory (SRAM), dynamic random-access memory (DRAM), other types of random-access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0101] These computer programs may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes can be implemented by different modules for different steps.

[0102] This embodiment provides such a device or system. The system, referred to as an object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling system, includes:

[0103] The data acquisition and preprocessing module is used to acquire time-matched coarse-resolution FPAR data and high-resolution multispectral remote sensing images within the study area, and to perform projection transformation, resampling, and cropping preprocessing.

[0104] The object segmentation and classification module is used to perform object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image and classify the objects into multiple land cover categories, including vegetation and non-vegetation.

[0105] The fine-scale modeling module is used to establish a linear mapping model between high-resolution surface reflectance and photosynthetically active radiation absorption ratio for each land cover category.

[0106] The scale transformation modeling module is used to construct an area-weighted constraint relationship between the coarse-resolution FPAR data and the fine-scale photosynthetically active radiation absorption ratio output by the fine-scale modeling module, based on the principle of scale consistency.

[0107] The parameter optimization solution module is used to perform distributed optimization of the parameters of the linear mapping model in the fine-scale modeling module based on the constraint relationship constructed by the scale transformation modeling module, using a step-by-step strategy. Specifically, the parameters of non-vegetation categories are determined first; then, based on the fixed parameters of the non-vegetation categories, the parameters of the vegetation categories are solved.

[0108] The product generation module is used to calculate the fine-scale photosynthetically active radiation absorption ratio of all objects in the entire study area using the parameters obtained by the parameter optimization solution module and the linear mapping model in the fine-scale modeling module, and generate high-resolution FPAR products.

[0109] As one implementation method in this embodiment, the object segmentation and classification module includes:

[0110] The image segmentation unit is used to perform multi-scale segmentation based on the spectral characteristics of the high-resolution multispectral remote sensing image using a remote sensing trend surface framework, thereby generating spectrally homogeneous object patches.

[0111] A land cover classification unit is used to classify the object patch into land cover categories that include at least trees, impermeable surfaces, water bodies, and arable land.

[0112] As one implementation method in this embodiment, the scale transformation modeling module includes:

[0113] The relation construction unit is used to construct the value of a cell in the coarse-resolution FPAR data into an area-weighted sum of the FPAR values ​​of all fine-scale objects within its spatial range, wherein the FPAR values ​​of the fine-scale objects are calculated by the fine-scale modeling module.

[0114] As one implementation method in this embodiment, the parameter optimization solution module includes:

[0115] The equation construction unit is used to convert the constraint relationships constructed by the scale transformation modeling module into matrix equation form;

[0116] The iterative optimization unit is used to iteratively optimize the matrix equation using the least squares method based on gradient descent, and select the parameter set with the highest determination coefficient as the final model parameters.

[0117] As one implementation method in this embodiment, the parameter optimization solution module further includes:

[0118] The sample screening unit is used to screen coarse-resolution pixels with a vegetation category area coverage greater than a preset threshold and ensure that the number of samples is three times or more the number of parameters to be determined as modeling samples, wherein the preset threshold is 10%.

[0119] As one implementation method in this embodiment, the process for determining non-vegetation category parameters in the parameter optimization solution module includes:

[0120] The clean sample selection unit is used to select clean pixels containing only a single land cover category as samples;

[0121] The background parameter determination unit is used to perform parameter fitting using pixel samples containing mixed land cover categories when there are no pure pixel samples of the corresponding category, and to determine the final parameters of non-vegetation categories based on the criterion of minimizing the root mean square error.

[0122] The system or apparatus is used to implement the functions of the methods in the above embodiments. Each module in the system or apparatus corresponds to each step in the method, as has been described in the method and will not be repeated here.

[0123] The above implementation method solves the problem of downscaling of the photosynthetically active radiation absorption ratio of urban vegetation in related technologies, thereby ensuring that the problems existing in the prior art are resolved.

[0124] The above are merely preferred embodiments of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. An object-oriented method for downscaling the photosynthetically active radiation absorption ratio of urban vegetation, characterized in that, Includes the following steps: S1. Acquire time-matched coarse-resolution FPAR data and high-resolution multispectral remote sensing images within the study area, and perform projection transformation, resampling, and cropping preprocessing. S2. Perform object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image, and classify the objects into multiple land cover categories, including vegetation and non-vegetation. S3. For each land cover category, establish a linear mapping model between its high-resolution surface reflectance and photosynthetically active radiation absorption ratio, as shown in the following formula: For FPAR calculated based on fine-scale images; Functional relationship between specific categories and time; In fine-scale images, the first Average reflectance of objects in each wavelength band; It is a constant term in a linear model. These are the coefficients of the k-th band in the linear model. This represents the number of bands in the image that are used in the calculation; S4. Based on the area-weighted scale transformation model, construct the constraint relationship between the coarse-resolution FPAR data and the fine-scale photosynthetically active radiation absorption ratio obtained from S3. The formula for the area-weighted scale transformation model is as follows: It is the FPAR value of the coarse resolution pixel. It represents the proportion of each object category to the area of ​​a MODIS pixel. This is the total number of categories; S5. Based on the constraint relationship constructed in S4, a step-by-step strategy is adopted to perform distributed optimization to solve the parameters of the linear mapping model described in S3. First, the parameters of non-vegetation categories are determined; then, based on the fixed parameters of the non-vegetation categories, the parameters of the vegetation categories are solved. S6. Using the parameters obtained from S5, calculate the fine-scale photosynthetically active radiation absorption ratio of all objects in the entire study area through the linear mapping model in S3, and generate high-resolution FPAR products.

2. The method according to claim 1, characterized in that, In S2, the process of performing object-oriented multi-scale segmentation on high-resolution multispectral remote sensing images includes: Using a remote sensing trend surface framework, the image spectral features are segmented to generate spectrally homogeneous object patches; The land cover categories include at least trees, impermeable surfaces, water bodies, and arable land.

3. The method according to claim 1, characterized in that, In S4, the process of constructing area-weighted constraints includes: The FPAR value of a coarse-resolution pixel is equal to the area-weighted sum of the FPAR values ​​of all fine-scale objects within its spatial range, which are calculated by the linear mapping model in S3.

4. The method according to claim 1, characterized in that, In S5, the process of using a step-by-step strategy for distributed optimization includes: The constraints constructed by S4 are converted into matrix equations, and iterative optimization is performed using the least squares method based on gradient descent. The parameter set with the highest determination coefficient is selected as the final model parameters.

5. The method according to claim 4, characterized in that, In S5, during the distributed optimization process using a step-by-step strategy, the samples used satisfy the following conditions: In the coarse-resolution pixels corresponding to the sample, the area coverage of the vegetation category is greater than a preset threshold, and the number of samples is three times or more the number of parameters to be determined, wherein the preset threshold is 10%.

6. The method according to claim 1, characterized in that, In S5, the process of determining the parameters for non-vegetation categories includes: We prioritize selecting pure pixel samples containing only a single land cover category for parameter fitting; if there are no pure pixel samples of the corresponding category, we use pixel samples containing mixed land cover categories for parameter fitting, and determine the final parameters based on the minimum root mean square error.

7. An object-oriented urban vegetation photosynthetically active radiation absorption ratio downscaling system, characterized in that, The system for implementing the method of any one of claims 1-6 comprises: The data acquisition and preprocessing module is used to acquire time-matched coarse-resolution FPAR data and high-resolution multispectral remote sensing images within the study area, and to perform projection transformation, resampling, and cropping preprocessing. The object segmentation and classification module is used to perform object-oriented multi-scale segmentation on the high-resolution multispectral remote sensing image and classify the objects into multiple land cover categories, including vegetation and non-vegetation. The fine-scale modeling module is used to establish a linear mapping model between high-resolution surface reflectance and photosynthetically active radiation absorption ratio for each land cover category. The scale transformation modeling module is used to construct an area-weighted constraint relationship between the coarse-resolution FPAR data and the fine-scale photosynthetically active radiation absorption ratio output by the fine-scale modeling module, based on the principle of scale consistency. The parameter optimization solution module is used to perform distributed optimization of the parameters of the linear mapping model in the fine-scale modeling module based on the constraint relationship constructed by the scale transformation modeling module, using a step-by-step strategy. Specifically, the parameters of non-vegetation categories are determined first; then, based on the fixed parameters of the non-vegetation categories, the parameters of the vegetation categories are solved. The product generation module is used to calculate the fine-scale photosynthetically active radiation absorption ratio of all objects in the entire study area using the parameters obtained by the parameter optimization solution module and the linear mapping model in the fine-scale modeling module, and generate high-resolution FPAR products.

8. A computer terminal device, characterized in that, include: One or more processors; A memory, coupled to the processor, for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the steps of the method as described in any one of claims 1-6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the steps of the method as described in any one of claims 1-6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the steps of the method according to any one of claims 1-6.

Images