Vegetation coverage and management measure remote sensing inversion method combining multi-angle remote sensing data and layered vertical vegetation coverage

Abstract

The invention discloses a vegetation coverage and management measure remote sensing inversion method combining multi-angle remote sensing data and layered vertical vegetation coverage, and the method comprises the following steps: (1), obtaining a multi-angle remote sensing data set, and carrying out the preprocessing of the multi-angle remote sensing data set; (2) selecting an optimal wave band through Borson correlation analysis and XGBoost feature screening; (3) estimating a C factor and a structured vegetation index S-Cs; (4) simulation of leaf reflectivity and transmissivity is carried out, and then vegetation canopy reflectivity is simulated; (5) converting the high spectral reflectivity of the vegetation canopy into equivalent remote sensing reflectivity of a satellite wave band, calculating vegetation indexes, and selecting the vegetation index with the highest correlation to participate in inversion and modeling of a leaf area index LAI; (6) determining an optimal angle combination and evaluating an inversion result of LAI; (7) establishing a novel quantitative coupling relation model between the multi-angle LAI and the SCs and between the SCs and the C factor, and then inverting the C factor; the invention provides a technical means with high practicability for monitoring the dynamic change and expansion process of soil erosion.

Description

Technical Field

[0001] The present invention relates to the technical field of remote sensing image inversion, and particularly relates to a method for remotely sensing inversion of vegetation cover and management measures by combining multi-angle remote sensing data and hierarchical vertical vegetation coverage. Background Art

[0002] Soil erosion, as a surface material migration process that frequently occurs in terrestrial ecosystems and has extensive ecological and environmental effects, not only causes soil nutrient loss, land productivity decline, and food production reduction, but also exacerbates natural disasters such as floods, debris flows, and landslides, and is closely linked to the global biogeochemical cycles (C, N, P, S) and even global climate change. Against the backdrop of the continuous growth of current resource demands and the increasing environmental pressure, soil erosion has become a key factor threatening food security in mountainous areas and hindering the sustainable development of agriculture. Therefore, timely and accurately monitoring and predicting the key factors of soil erosion models is of great theoretical value and practical significance for promoting soil and water conservation research, ensuring national ecological security, and achieving the sustainable development of social economy.

[0003] To obtain accurate quantitative information on soil erosion at the regional scale and scientifically formulate soil and water conservation plans and strategies to address erosion-related issues, domestic and foreign scholars have carried out soil erosion research through technical means such as outdoor runoff plot observations and indoor simulated rainfall experiments. However, these methods are difficult to meet the needs of large-area assessments due to limitations such as limited spatial coverage and long time consumption. In contrast, the model-based quantitative evaluation method has become the preferred approach for realizing large-scale soil erosion quantitative assessment. Currently, the Universal Soil Loss Equation (USLE) and its improved version, the Revised Universal Soil Loss Equation (RUSLE), have been widely combined with GIS and RS technologies and successfully verified and promoted at home and abroad due to their advantages such as simple calculation formulas, low data requirements, and easy parameter acquisition, and have become the most widely used soil erosion quantitative estimation models. Among them, the vegetation cover and management measure factor (C factor) can characterize the influence of surface cover types and management measures on soil erosion and is a key parameter for accurately estimating soil erosion. However, studies have shown that soil erosion is not only affected by vegetation coverage but is also closely related to vegetation types and structures. Under the same vegetation coverage conditions, multi-layer forests are more effective in reducing soil erosion than single-layer forests. However, quantifying the impact of vegetation stratification structure on erosion in space has not been fully practiced. This lack of information limits the accuracy and applicability of the model and poses challenges to the scientific prediction of the soil erosion process. Summary of the Invention

[0004] Objective of the Invention: The objective of the present invention is to provide a remote sensing inversion method for vegetation cover and management measures that combines multi-angle remote sensing data and stratified vertical vegetation coverage. By using physical models and statistical models to extract the stratified vertical vegetation coverage and invert the vegetation cover and management measure (C) factor, the problems existing in the background technology are solved.

[0005] Technical Solution: A remote sensing inversion method for vegetation cover and management measures that combines multi-angle remote sensing data and stratified vertical vegetation coverage according to the present invention includes the following steps:

[0006] (1) Obtain the PROBA / CHRIS multi-angle remote sensing data set of the target area at a specific time and perform preprocessing;

[0007] (2) Select the best bands through Pearson correlation analysis and XGBoost feature screening;

[0008] (3) Estimate the C factor and the structured vegetation index S-Cs based on the definition of the C factor in the RUSLE model and the structured vegetation factor index model;

[0009] (4) Based on the PROSAIL model, simulate the leaf reflectance and transmittance, and then simulate the vegetation canopy reflectance;

[0010] (5) Based on the simulated vegetation canopy reflectance, use the spectral response function to convert the vegetation canopy hyperspectral reflectance into the equivalent remote sensing reflectance of the satellite band, calculate the vegetation index, and select the vegetation index with the highest correlation to participate in the inversion and modeling of the leaf area index LAI;

[0011] (6) Based on the obtained LAI and vegetation index, train and validate the constructed single-angle and multi-angle LAI inversion models, determine the best angle combination, and evaluate the inversion result of LAI;

[0012] (7) Based on the multi-angle LAI inversion result and the directional vegetation coverage calculation formula, establish a new quantitative coupling relationship model between multi-angle LAI and S_Cs, and between S_Cs and the C factor, and then invert the C factor.

[0013] Further, step (1) includes the following steps:

[0014] (11) Obtain the multi-angle surface reflectance data according to the PROBA / CHRIS MODE 3 data set released by the European Space Agency;

[0015] (12) For the horizontal and vertical strip noises on the PROBA / CHRIS multi-angle image, use the HDFclean software to fill the missing pixels and remove the noises of the image;

[0016] (13) Using the BEAM software, atmospheric correction of multi-angle data is obtained through a method combining radiative transfer and empirical linearity;

[0017] (14) Using the quadratic polynomial fitting method and bilinear interpolation method, geometric correction is performed on the multi-angle PROBA / CHRIS images after atmospheric correction.

[0018] Further, step (2) includes the following steps:

[0019] (21) Using the ArcGis software to generate random points and obtain the surface reflectance of the corresponding points, and then calculating the correlation coefficient matrix between the CHRIS image bands by means of Pearson correlation analysis. In order to facilitate the selection of the best inversion bands, the CHRIS bands are divided into three parts according to the correlation coefficient matrix;

[0020] (22) Through XGBoost feature screening, calculate the importance of each band in each segment, and select the best inversion bands, namely the blue, red, and near-infrared bands, according to the magnitude of the importance. Use the HDFclean software to determine the central wavelengths of the corresponding bands.

[0021] Further, step (4) includes the following steps:

[0022] (41) Combining the global cost function minimization method and the biochemical component parameters of each tree species in the field sample plots, determine the input parameters of the PROSPECT model, and simulate the leaf reflectance and transmittance;

[0023] (42) Based on the simulation results of leaf reflectance and transmittance, according to the sensitivity analysis results and relevant literature, determine the constants and variable input parameters of the SAIL model, and simulate the leaf canopy reflectance.

[0024] Further, step (5) includes the following steps:

[0025] (51) Based on the leaf canopy reflectance simulated by the PROSAIL model, use the spectral response function to convert the leaf canopy hyperspectral reflectance into the satellite band equivalent remote sensing reflectance;

[0026] (52) Based on the converted satellite band equivalent remote sensing reflectance, construct multiple vegetation indices using the characteristic bands, and further select the vegetation index with the highest correlation through Pearson correlation analysis to participate in the inversion and modeling of LAI.

[0027] Further, step (6) includes the following steps:

[0028] (61) Based on the obtained vegetation indices and the corresponding LAI, construct single-angle and multi-angle inversion datasets. Use the vegetation indices as independent variables and LAI as the dependent variable, and then construct extreme gradient boosting inversion models for single-angle and multi-angle data.

[0029] (62) Select three evaluation indicators, namely Nash-Sutcliffe efficiency coefficient NSE, coefficient of determination R2, and root mean square error RMSE, to evaluate the model accuracy and determine the best angle combination for LAI inversion, and perform LAI inversion.

[0030] Further, step (7) includes the following steps:

[0031] (71) Based on the multi-angle LAI inversion results and the estimated structural vegetation index, improve the calculation formula of directional vegetation coverage, and invert the stratified vertical vegetation coverage.

[0032] (72) Based on the previously estimated structural vegetation index and C factor value, construct a regression relationship between the measured structural vegetation index and the measured C value through the XGBoost model, and further use this regression relationship and remote sensing images to invert the C factor.

[0033] A remote sensing inversion system for vegetation cover and management measures combining multi-angle remote sensing data and stratified vertical vegetation coverage according to the present invention includes:

[0034] Preprocessing module: used to obtain the PROBA / CHRIS multi-angle remote sensing dataset of the target area at a specific time and perform preprocessing.

[0035] Optimal band module: used to select the optimal band through Pearson correlation analysis and XGBoost feature screening.

[0036] RUSLE module: used to estimate the C factor and the structural vegetation index S-Cs based on the definition of the C factor in the RUSLE model and the structural vegetation factor index model.

[0037] PROSAIL module: used to simulate the leaf reflectance and transmittance based on the PROSAIL model, and then simulate the vegetation canopy reflectance.

[0038] Inversion and modeling module: used to convert the vegetation canopy hyperspectral reflectance into satellite band equivalent remote sensing reflectance using the spectral response function based on the simulated vegetation canopy reflectance, calculate the vegetation index, and select the vegetation index with the highest correlation to participate in the inversion and modeling of the leaf area index LAI.

[0039] Training module: used to train and validate the constructed single-angle and multi-angle LAI inversion models based on the obtained LAI and vegetation indices, determine the best angle combination, and evaluate the LAI inversion results.

[0040] C factor module: Based on the multi-angle LAI inversion results and the directional vegetation coverage calculation formula, a new quantitative coupling relationship model between multi-angle LAI and S_Cs, as well as between S_Cs and C factor, is established to invert the C factor.

[0041] An electronic device according to the present invention includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements a remote sensing inversion method for vegetation cover and management measures combining multi-angle remote sensing data and hierarchical vertical vegetation coverage according to any one of the above.

[0042] A storage medium according to the present invention stores a computer program. When the computer program is executed by a processor, it implements a remote sensing inversion method for vegetation cover and management measures combining multi-angle remote sensing data and hierarchical vertical vegetation coverage according to any one of the above.

[0043] Beneficial effects: Compared with the prior art, the present invention has the following remarkable advantages: (1) It improves the limitation that the traditional vegetation coverage inversion C factor only reflects the horizontal vegetation coverage, and comprehensively considers the comprehensive impact of different vertical vegetation layers on soil erosion; (2) By combining multi-angle remote sensing images and hierarchical vertical vegetation coverage, the C factor reflecting the vertical vegetation characteristics can be accurately and efficiently obtained, providing a highly practical technical means for monitoring the dynamic changes and expansion process of soil erosion. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] Figure 1 is a flowchart of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0045] The technical solution of the present invention will be further described below with reference to the drawings.

[0046] As Figure 1 shown, an embodiment of the present invention provides a remote sensing inversion method for vegetation cover and management measures combining multi-angle remote sensing data and hierarchical vertical vegetation coverage, including the following steps:

[0047] In the embodiment of the present invention, Zijin Mountain in Nanjing, Jiangsu Province is taken as the research area to verify the effectiveness of the remote sensing inversion method of vegetation cover and management measures combining multi-angle remote sensing data and hierarchical vertical vegetation coverage: Based on multi-angle remote sensing data, relying on the MATLAB and Python computing platforms, the present invention combines the radiative transfer model and the statistical model to construct an extreme gradient boosting tree LAI inversion model based on the radiative transfer model and multi-angle PROBA / CHRIS remote sensing data. On this basis, a new type of quantitative coupling relationship model between multi-angle LAI and S-Cs, and between S-Cs and C factor is established, providing a new method for the remote sensing quantitative inversion of C factor. This method provides new theoretical and technical support for the quantitative and precise inversion of vegetation parameters on the basis of making full use of multi-angle remote sensing data.

[0048] As Figure 1 shown, in step P1, a PROBA / CHRIS multi-angle remote sensing data set is obtained and preprocessed.

[0049] P1.1. Select the PROBA / CHRIS MODE 3 data set (spatial resolution of 17m) released by the European Space Agency (ESA). This data set provides observation information at five different angles (0°, 36°, -36°, 55°, -55°). To ensure data quality, images with lower cloud cover are preferentially selected;

[0050] P1.2. For the noise of horizontal and vertical stripes existing in the imaging of PROBA / CHRIS data, use the HDFclean software provided by the European Space Agency to denoise the image and fill the missing pixels;

[0051] P1.3. Use the BEAM software provided by the European Space Agency to perform atmospheric correction on the multi-angle data;

[0052] P1.4. Taking the calibrated TM image as the reference base map, combine the quadratic polynomial fitting and bilinear interpolation methods to finely calibrate the atmospherically corrected image. By selecting ground control points in the study area, strictly control the geometric correction error within 0.5 pixels to perform geometric correction on the atmospherically corrected PROBA / CHRIS image.

[0053] In step P2, based on the preprocessed multi-angle remote sensing data set, the best bands are selected through Pearson correlation analysis and XGBoost feature screening.

[0054] P2.1. Using the ArcGIS software, extract the surface reflectance of multiple points by generating random points and the function of extracting multiple values to points. Further, calculate the correlation coefficient matrix between 18 bands of the CHRIS image by means of Pearson correlation analysis. To facilitate the selection of the best bands, divide the CHRIS bands into three parts according to the correlation coefficient matrix.

[0055] P2.2. Through XGBoost feature screening, calculate the importance of each principal component to the bands in each segment, select the best inversion bands (blue, red, near-infrared bands) according to the importance, and determine the central wavelength with the help of the HDFclean software.

[0056] In step P3, estimate the C factor and the structured vegetation index (S-Cs) based on the definition of the C factor in the RUSLE model and the structured vegetation factor index model, specifically including:

[0057] P3.1. The specific formula for calculating the C factor in the RUSLE model is:

[0058] C = PLU × CC × SC × SR × SM

[0059] where C is the vegetation cover and management practice factor, PLU is the previous land use factor, CC is the canopy cover factor, SC is the surface cover factor, SR is the surface roughness, and SM is the soil moisture;

[0060] CC = 1 - Fc × exp[-0.1 × H]

[0061] where F c is the ratio of the canopy cover to the land area, and H is the average tree height;

[0062] SC = exp[-b × S p ×(0.24 / R u ) 0.08

[0063] where b and R u are the random roughness, and S p is the surface cover percentage (%);

[0064] SR = exp[-0.66 × (R u - 0.24)]

[0065] P3.2. Calculate the structured vegetation index (S-Cs) according to the structured vegetation factor index model (Table 2):

[0066] Table 1 Vegetation Index and Its Expression

[0067] ​

[0068] In step P4, based on the PROSAIL model, the leaf reflectance and transmittance are simulated, and then the vegetation canopy reflectance is simulated.

[0069] The PROSAIL model integrates the leaf optical property model PROSPECT and the canopy reflectance model SAIL. By comprehensively considering various key factors, including the non-Lambertian reflection characteristics of the soil, the specular reflection effect of the leaves, the hot spot effect of the vegetation canopy, and the leaf inclination distribution characteristics, a more accurate simulation framework is constructed. The model adopts a two-layer canopy structure and further incorporates the heterogeneity of the canopy in the horizontal and vertical directions, so as to accurately characterize the reflection characteristics of a homogeneous vegetation canopy. Its complete physical basis and comprehensive consideration enable the model to exhibit excellent accuracy and reliability in the simulation of reflection characteristics, providing strong theoretical support for the inversion of vegetation remote sensing parameters.

[0070] P4.1. Combine the global cost function minimization method and input the biochemical component parameters of each tree species in the field plots of the study area measured in the laboratory, including leaf area, leaf equivalent water thickness, dry matter content, chlorophyll, and carotenoids, etc., into the PROSPECT model to simulate the leaf reflectance and transmittance;

[0071] The expression of the PROSPECT model is:

[0072] (ρ 1 ,τ 1 ) = PROSPECT(N, C ab , C w , C m )

[0073] p 1 is the leaf reflectance; τ 1 is the leaf transmittance; N is the leaf internal structure parameter, C ab is the leaf chlorophyll content (μg / cm 2 ), C w is the leaf equivalent water thickness (cm), C m is the leaf dry matter content (mg / cm 2 );

[0074] P4.2. Parametric sensitivity analysis is an important tool widely used to evaluate the sensitivity of biochemical component parameters to the change of canopy reflectance and the wavelength range of its influence. This process helps to identify key parameters and optimize model efficiency. Sensitivity analysis usually includes qualitative and quantitative methods. Qualitative analysis systematically changes the value of the target parameter while keeping other parameters fixed, generates a series of simulated results of canopy reflectance, and evaluates the strength of parameter sensitivity through the change trend of the reflectance curve; while quantitative analysis focuses on the change range of the parameter near a specific reference value and quantitatively measures the response degree of the model output canopy reflectance to the parameter change.

[0075] Based on the results of sensitivity analysis and relevant literature, the constants and variable input parameters of the SAIL model were determined. At the same time, the biochemical component parameters, multi-angle remote sensing data, LAI, and the leaf reflectance and transmittance simulated by the PROSPECT model were used as inputs to the SAIL model, and the simulation of leaf canopy reflectance was further completed. The expression of the SAIL model is:

[0076]

[0077] where p c is the canopy spectral reflectance, LAI is the leaf area index, ALA is the average leaf angle (°), S L is the hot spot parameter, Diff is the diffuse reflectance coefficient, θ v and θ s are the viewing zenith angle and the solar zenith angle respectively, is the relative azimuth angle between the sun and the observation

[0078] In step P5, based on the simulated leaf canopy reflectance, using the spectral response function, the leaf canopy hyperspectral reflectance is converted into the equivalent remote sensing reflectance of the satellite band, and the vegetation index is calculated.

[0079] P5.1. Based on the leaf canopy reflectance simulated by the PROSAIL model, using the spectral response function, the leaf canopy hyperspectral reflectance is converted into the equivalent remote sensing reflectance of the satellite band, and the expression is as follows:

[0080]

[0081] where R rs (band i ) is the equivalent remote sensing reflectance of the i - band of the satellite, λ 1 and λ 2 are the wavelength range of this band, R rs (λ) is the hyperspectral remote sensing reflectance, and SRF(λ) is the spectral response rate at the wavelength of λ

[0082] P5.2. According to relevant research, eight vegetation indices were selected, namely the ratio vegetation index, the normalized vegetation index, the perpendicular vegetation index, the difference vegetation index, the soil-adjusted vegetation index, the enhanced vegetation index, the modified soil-adjusted vegetation index, and the canopy structure-insensitive vegetation index. Based on the converted equivalent remote sensing reflectance of satellite bands, characteristic bands were used to construct various vegetation indices, and then the vegetation index with the highest correlation was selected through Pearson correlation analysis to participate in the inversion and modeling of LAI.

[0083] Table 2 Vegetation Indices and Their Expressions

[0084]

[0085] In the formula, ρ BLUE , ρ RED , ρ NIR are the reflectances of the blue, red, and near-infrared bands respectively

[0086] In step P6, based on the obtained LAI and vegetation indices, the constructed single-angle and multi-angle LAI inversion models are trained and verified to determine the best angle combination and evaluate the inversion results of LAI.

[0087] The Extreme Gradient Boosting Tree model (XGBoost) is an enhanced algorithm based on supervised gradient boosting. XGBoost reduces the residual between the actual value and the predicted value by continuously forming new decision trees to fit the residuals of the previous predictions. The XGBoost algorithm is developed on the basis of the boosting ensemble learning method and incorporates the advantages of the bagging ensemble learning method in the evolution process. By customizing the loss function through the gradient boosting framework, the ability of the algorithm to solve general problems is improved, and at the same time, more controllable parameters are introduced to optimize the algorithm for the problem scenario. In addition, the details of the engineering implementation are optimized to ensure the stability of the algorithm when processing large-scale data efficiently. The XGBoost algorithm contains 22 parameters such as n_estimators, learning_rate, subsample, colsample_bytree, max_depth, gamma, and silent.

[0088] P6.1. Take the vegetation index as the independent variable and LAI as the dependent variable and input them into the XGBoost model. Select 70% of the samples as the modeling set and 30% of the samples as the validation set in each angle combination. Use the method of cross-validation combined with grid search to determine the best parameters in the model inversion;

[0089] Table 3 Angle Combination Modes

[0090]

[0091] P6.2. Select the root mean square error (RMSE), mean absolute percentage error (MAPE), and square correlation coefficient (R 2 ) as three evaluation indicators to evaluate the model accuracy and determine the best angle combination for LAI inversion.

[0092]

[0093]

[0094] Where x' i is the LAI inversion value; x i is the measured LAI value; n is the number of samples

[0095] As Figure 1 shown, in step P7, based on the multi-angle LAI inversion results and the directional vegetation coverage calculation formula, a new type of quantitative coupling relationship model between multi-angle LAI and S_Cs, and between S_Cs and C factor is established, and then the C factor is inverted.

[0096] P7.1. Incorporate the multi-angle idea into the directional vegetation coverage estimation formula to establish a new type of quantitative coupling relationship model between multi-angle LAI and S_Cs:

[0097] F cover (θ) = 1 - e -LAI / (2*cosθ)

[0098] Where θ is the solar zenith angle of incidence

[0099] Take X 0 , X 36 , X -36 , X 55 , X -55 as independent variables, and take 1 - F cover (θ) as the dependent variable and input it into the XGBoost model. Construct single-feature and multi-feature data sets. Select 70% of the samples in each angle combination as the modeling set and 30% of the samples as the validation set. Use the method of cross-validation combined with grid search to determine the best parameters in the model inversion. Finally, determine the best angle combination and evaluate the inversion results of the structured vegetation index S-C s according to the evaluation indicators output by the model

[0100]

[0101] Y = 1 - F cover (θ)

[0102] Where we take the structured vegetation index S-C s as F cover\((\theta)\);

[0103] P7.2. On the basis of obtaining the vertical coverage rate of vegetation layers, using the measured S-C s and C factors, taking S-C s as the independent variable and the C factor as the dependent variable for regression modeling, and applying this regression relationship to the image.

[0104] An embodiment of the present invention also provides a remote sensing inversion system for vegetation cover and management measures combining multi-angle remote sensing data and stratified vertical vegetation coverage, including:

[0105] Preprocessing module: used to obtain the multi-angle remote sensing data set of PROBA / CHRIS at a specific time in the target area and perform preprocessing;

[0106] Optimal band module: used to select the optimal band through Polson correlation analysis and XGBoost feature screening;

[0107] RUSLE module: used to estimate the C factor and the structured vegetation index S-Cs based on the definition of the C factor in the RUSLE model and the structured vegetation factor index model;

[0108] PROSAIL module: used to simulate the leaf reflectance and transmittance based on the PROSAIL model, and then simulate the vegetation canopy reflectance;

[0109] Inversion and modeling module: used to convert the hyperspectral reflectance of the vegetation canopy into the equivalent remote sensing reflectance of the satellite band using the spectral response function based on the simulated vegetation canopy reflectance, calculate the vegetation index, and select the vegetation index with the highest correlation to participate in the inversion and modeling of the leaf area index LAI;

[0110] Training module: used to train and validate the constructed single-angle and multi-angle LAI inversion models based on the obtained LAI and vegetation index, determine the optimal angle combination, and evaluate the inversion result of LAI;

[0111] C factor module: used to establish a new type of quantitative coupling relationship model between multi-angle LAI and S_Cs, and between S_Cs and the C factor based on the multi-angle LAI inversion result and the directional vegetation coverage calculation formula, and then invert the C factor.

[0112] An embodiment of the present invention also provides an electronic device, including a memory, a processor, and a computer program stored on the memory and executable on the processor. When the computer program is loaded into the processor, it implements a remote sensing inversion method for vegetation cover and management measures combining multi-angle remote sensing data and stratified vertical vegetation coverage according to any one of the above.

[0113] An embodiment of the present invention also provides a storage medium storing a computer program, which when executed by a processor implements any one of the remote sensing inversion methods of vegetation cover and management measures combining multi-angle remote sensing data and hierarchical vertical vegetation coverage.

Claims

1. A remote sensing inversion method for vegetation cover and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage, characterized in that: The following steps are involved: (1) Obtain the PROBA / CHRIS multi-angle remote sensing dataset of the target area at a specific time and perform preprocessing; (2) Select the best bands through Poulsen correlation analysis and XGBoost feature screening; (3) Based on the definition of C factor in RUSLE model and the structured vegetation factor index model, estimate C factor and structured vegetation index S-Cs; (4) Based on the PROSAIL model, the reflectivity and transmittance of leaves are simulated, and then the reflectivity of vegetation canopy is simulated; (5) Based on the simulated vegetation canopy reflectance, the spectral response function is used to convert the vegetation canopy hyperspectral reflectance into the equivalent remote sensing reflectance of the satellite band, and the vegetation index is calculated. The vegetation index with the highest correlation is selected to participate in the inversion and modeling of the leaf area index LAI; (6) Based on the obtained LAI and vegetation index, single-angle and multi-angle LAI inversion models are constructed and trained and verified to determine the optimal angle combination and evaluate the LAI inversion results; (7) Based on the multi-angle LAI inversion results and the directional vegetation coverage calculation formula, a new quantitative coupling relationship model between multi-angle LAI and S_Cs, and between S_Cs and C factor was established to invert the C factor.

2. The remote sensing inversion method for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage according to claim 1, characterized in that: Step (1) comprises the following steps: (11) Obtain multi-angle surface reflectivity data based on the PROBA / CHRIS MODE 3 dataset released by ESA; (12) To address the horizontal and vertical stripe noise on PROBA / CHRIS multi-angle images, HDFclean software was used to fill missing pixels and remove noise from the images; (13) Using BEAM software, the atmospheric correction of multi-angle data is obtained by combining radiation transfer and empirical linear methods; (14) The PROBA / CHRIS multi-angle images after atmospheric correction are geometrically corrected using quadratic polynomial fitting and bilinear interpolation methods.

3. The remote sensing inversion method for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage according to claim 1, characterized in that: Step (2) comprises the following steps: (21) ArcGis software was used to generate random points and obtain the surface reflectance of the corresponding points. The correlation coefficient matrix between the CHRIS image bands was calculated using Pearson correlation analysis. In order to facilitate the selection of the best inversion band, the CHRIS band was divided into three parts according to the correlation coefficient matrix; (22) Through XGBoost feature screening, the importance of each band in each segment is calculated, and the best inversion bands, namely the blue, red, and near-infrared bands, are selected according to the importance. The central wavelength of the corresponding band is determined using HDFclean software.

4. The remote sensing inversion method for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage according to claim 1, characterized in that: Step (4) comprises the following steps: (41) Combining the global cost function minimization method and the biochemical component parameters of each tree species in the field sample, the input parameters of the PROSPECT model were determined to simulate the leaf reflectance and transmittance; (42) Based on the simulation results of leaf reflectance and transmittance, the constant and variable input parameters of the SAIL model were determined according to the sensitivity analysis results and relevant literature to simulate the leaf canopy reflectance.

5. The remote sensing inversion method for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage according to claim 1, characterized in that: Step (5) comprises the following steps: (51) Based on the leaf canopy reflectance simulated by the PROSAIL model, the leaf canopy hyperspectral reflectance is converted into equivalent remote sensing reflectance in the satellite band using the spectral response function; (52) Based on the converted satellite band equivalent remote sensing reflectance, multiple vegetation indices were constructed using characteristic bands, and the vegetation index with the highest correlation was further selected through Pearson correlation analysis to participate in the inversion and modeling of LAI.

6. The remote sensing inversion method for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage according to claim 1, characterized in that: Step (6) comprises the following steps: (61) Based on the obtained vegetation index and the corresponding LAI, single-angle and multi-angle inversion datasets were constructed, with the vegetation index as the independent variable and the LAI as the dependent variable, and then the extreme boosting tree inversion model of single-angle and multi-angle data was constructed; (62) Three evaluation indicators, namely the root mean square error (RMSE), mean absolute percentage error (MAPE), and square correlation coefficient (R2), were selected to evaluate the model accuracy and determine the optimal angle combination for inverting LAI for LAI inversion.

7. The remote sensing inversion method for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage according to claim 1, characterized in that: Step (7) comprises the following steps: (71) Based on the multi-angle LAI inversion results and the estimated structured vegetation index, the directional vegetation coverage calculation formula is improved to invert the layered vertical vegetation coverage; (72) Based on the previously estimated structured vegetation index and C factor values, the regression relationship between the measured structured vegetation index and the measured C value was constructed through the XGBoost model, and the C factor was further inverted using this regression relationship and remote sensing images.

8. A remote sensing inversion system for vegetation coverage and management measures combining multi-angle remote sensing data and layered vertical vegetation coverage, characterized in that: include: Preprocessing module: used to obtain PROBA / CHRIS multi-angle remote sensing data sets at a specific time in the target area and perform preprocessing; Best band module: used to select the best band through Paulsen correlation analysis and XGBoost feature screening; RUSLE module: used to estimate the C factor and structured vegetation index S-Cs based on the definition of C factor and structured vegetation factor index model in the RUSLE model; PROSAIL module: used to simulate leaf reflectance and transmittance based on the PROSAIL model, and then simulate the reflectance of vegetation canopy; Inversion and modeling module: used to obtain the vegetation canopy reflectance based on simulation, convert the vegetation canopy hyperspectral reflectance into equivalent remote sensing reflectance in satellite bands using spectral response function, calculate vegetation index, and select the vegetation index with the highest correlation to participate in the inversion and modeling of leaf area index LAI; Training module: used to train and verify the constructed single-angle and multi-angle LAI inversion models based on the obtained LAI and vegetation index, determine the best angle combination, and evaluate the LAI inversion results; C factor module: It is used to establish a new quantitative coupling relationship model between multi-angle LAI and S_Cs, and between S_Cs and C factor based on the multi-angle LAI inversion results and the directional vegetation cover calculation formula, and then invert the C factor.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that: When the computer program is loaded into a processor, a method for designing an intelligent online evaluation system for personalized services according to any one of claims 1 to 7 is implemented.

10. A storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, a method for designing an intelligent online evaluation system for personalized services according to any one of claims 1 to 7 is implemented.

Images