Inversion method and system of grassland nitrogen content

Abstract

The invention discloses a grassland nitrogen content inversion method and system, and relates to the technical field of grassland vegetation quantitative remote sensing. The method comprises the following steps: acquiring measured grassland hyperspectral data and environmental parameter data of a target area, and preprocessing the measured grassland hyperspectral data; according to the preprocessed grassland actually measured hyperspectral data, extracting mapping relation characteristics of hyperspectrum-leaf area indexes in vegetation; extracting a leaf area index-nitrogen content according to the environmental parameter data; and by taking the leaf area index as connection, coupling the extracted leaf area index-nitrogen content with the hyperspectrum-leaf area index, and carrying out inversion on the nitrogen content of the grassland. Through the method provided by the invention, the dynamic change of the nitrogen in the plant growth process can be better captured, and the method has higher stability and transportability so as to quickly and effectively invert the nitrogen content of the grassland, so that the nitrogen of the plant is stably and accurately extracted, and the generalization ability of the nitrogen extraction method is improved.

Description

Technical Field

[0001] The invention relates to the technical field of quantitative remote sensing of grassland vegetation, and in particular to an inversion method and system for grassland nitrogen content. Background Art

[0002] Grassland is an important ecological resource in the world. Its nitrogen content is an important indicator of grassland health and productivity, affecting plant growth, carbon and nitrogen cycles, and the overall function of grassland ecosystems. Accurate monitoring of grassland nitrogen content is crucial for ecological environment management, agricultural production optimization, and grassland ecological restoration.

[0003] PROSAIL is a physical model that combines the leaf reflectance model PROSPECT and the vegetation canopy radiation transfer model SAIL to provide a systematic description of the vegetation reflectance process. The PROSPECT model simulates the leaf-level reflectance and transmission hyperspectral through mesophyll structural parameters and leaf biochemical characteristic parameters, and then simulates the vegetation canopy reflectance hyperspectral data in the SAIL model by combining vegetation canopy structural parameters and soil background parameters. The Decision Support System for Agrotechnology Transfer (DSSAT) brings together more than 40 crop growth models. It simulates crop growth parameters according to the crop growth cycle, combined with crop genetic characteristics, management measures, climate, soil and other parameters. It is widely used in yield forecasting, risk assessment, climate change research and other fields.

[0004] The nitrogen inversion method obtains remote sensing images and ground-based measured data, calculates multiple spectral indices, analyzes the correlation between these indices and nitrogen content, and selects spectral indices that are highly correlated with nitrogen as model input. Then, a regression method is used to construct a nitrogen estimation model, and the accuracy of the model is ensured through parameter optimization and verification. Finally, the optimal model is used to perform spatial and temporal nitrogen inversion of remote sensing data to generate a nitrogen distribution map to provide support for agricultural nitrogen fertilizer management and decision-making.

[0005] However, although this empirical relationship between multispectral features and plant nitrogen content established through regression analysis can quickly estimate plant nitrogen content, its temporal and spatial portability is poor. Although the traditional vegetation radiation transfer physical model has higher temporal and spatial portability than the empirical model, it does not include plant nitrogen as a model parameter. Therefore, it is impossible to extract plant nitrogen stably and accurately through the inversion of the traditional vegetation radiation physical model, which leads to the problem of greater difficulty in plant nitrogen extraction. Summary of the invention

[0006] In order to overcome the problem that the existing grassland nitrogen content detection technology is unable to extract plant nitrogen stably and accurately, thereby making the plant nitrogen extraction work more difficult, the present invention proposes a grassland nitrogen content inversion method and system, which can quickly and effectively invert the grassland nitrogen content, thereby stably and accurately extracting plant nitrogen and improving the generalization ability of the nitrogen extraction method.

[0007] To achieve the purpose of the present invention, the present invention adopts the following technical solutions:

[0008] A method for inverting grassland nitrogen content, the method comprising the following steps:

[0009] Acquire grassland measured data and environmental parameter data of the target area, and preprocess the grassland measured data;

[0010] Based on the pre-processed grassland measured hyperspectral data, the mapping relationship characteristics between hyperspectral and leaf area index in vegetation are extracted;

[0011] Extract leaf area index-nitrogen content based on environmental parameter data;

[0012] Taking the leaf area index as the connection, the extracted leaf area index-nitrogen content was coupled with the leaf area index to invert the grassland nitrogen content.

[0013] In the above technical scheme, by preprocessing the acquired measured hyperspectral data of grassland, the spectral changes caused by scattering during the measurement of grassland nitrogen content can be reduced, especially the spectral differences caused by different scattering levels can be eliminated, thereby standardizing the correlation between the spectrum and the data; and using the extracted leaf area index as a connection, the extracted leaf area index-nitrogen content is coupled with the hyperspectral-leaf area index to invert the grassland nitrogen content, which can better capture the dynamic changes of nitrogen during plant growth, and has higher stability and portability, so as to quickly and effectively invert the grassland nitrogen content, thereby stably and accurately extracting plant nitrogen, and improving the generalization ability of the nitrogen extraction method.

[0014] Furthermore, the process of preprocessing the grassland measured data and environmental parameter data includes:

[0015] The parameter range is divided according to the obtained grassland measured data, and the vegetation canopy radiation transfer model is constructed;

[0016] According to the measured grassland data after parameter range division, the basic parameters of the constructed vegetation canopy radiation transfer model are adjusted, and the spectral reflectance database under different parameters is generated using the vegetation canopy radiation transfer model;

[0017] Based on the spectral reflectance database under different parameters, the vegetation canopy radiation transfer model is used to extract the high spectral reflectance under different parameters;

[0018] According to the high spectral reflectance extracted by the vegetation canopy radiation transfer model, the canopy spectral curve data in the grassland measured data is spectrally corrected.

[0019] The grassland measured data include leaf area index and vegetation canopy hyperspectral data, and the environmental parameter data include meteorological data, soil data, field management data and field observation data.

[0020] Furthermore, the canopy spectral curve data in the acquired grassland measured data is spectrally corrected, and the expression is:

[0021]

[0022] Among them, R sc (i,j) represents the reflectance value of the ith sample at wavelength j after scattering correction, R(i,j) represents the measured original reflectance value of the ith sample at wavelength j, mean(R(j)) represents the average value of the measured original reflectance value at wavelength j, sd(R(i)) represents the standard deviation of the measured original reflectance value at wavelength j, sd(S(j)) represents the standard deviation of the spectral reflectance value simulated by the vegetation canopy radiation transfer model at j, and mean(S(j)) represents the average value of the spectral reflectance value simulated by the vegetation canopy radiation transfer model at j.

[0023] Furthermore, the constructed vegetation canopy radiation transfer model expression is:

[0024]

[0025] Among them, ρ c represents canopy reflectance, N represents leaf structure parameters, Cab represents chlorophyll content, Cw represents water content, Cm represents dry matter content, LAI represents leaf area index, ALA represents average leaf inclination, Hotspot represents hotspot parameter, Psoil represents soil brightness, rsoli represents soil moisture, θ v represents the observation zenith angle, θ x represents the solar zenith angle, Indicates the relative azimuth between the sun and the observation point.

[0026] In the above technical scheme, spectral correction processing is performed on the canopy spectral curve data in the acquired grassland measured data, which can reduce the spectral changes caused by scattering during the measurement process, especially eliminate the spectral differences caused by different scattering levels, thereby standardizing the correlation between the spectrum and the data.

[0027] Furthermore, the process of extracting the mapping relationship characteristics between the hyperspectral spectrum and the leaf area index in vegetation includes:

[0028] According to the spectral reflectance database under different parameters, the leaf area index under different parameters is extracted using the vegetation canopy radiation transfer model;

[0029] The hyperspectral reflectance and leaf area index extracted by the vegetation canopy radiation transfer model were used as the training set, and the measured grassland data after spectral correction were used as the validation set.

[0030] The training set is used to optimize the leaf area index inversion of the vegetation canopy radiation transfer model, and the trained vegetation canopy radiation transfer model is verified using the verification set. After the verification, the trained vegetation canopy radiation transfer model is obtained.

[0031] The trained vegetation canopy radiation transfer model is used to invert the hyperspectral data in the input grassland measured data and extract the leaf area index of the vegetation.

[0032] Furthermore, the process of optimizing the leaf area index inversion of the vegetation canopy radiation transfer model using the training set includes:

[0033] Based on the hyperspectral reflectance and leaf area index under different parameters in the training set, several rounds of optimization training are set up to traverse different parameter combinations to optimize the parameters of the vegetation canopy radiation transfer model.

[0034] The vegetation canopy radiation transfer model uses the partial least squares regression method to invert the leaf area index of vegetation based on the hyperspectral reflectance and leaf area index under different parameters;

[0035] The loss function is set to compare the leaf area index inversion results under different parameter combinations with the measured results, and the vegetation canopy radiation transfer model corresponding to the parameter combination with the smallest loss function value is taken as the trained vegetation canopy radiation transfer model.

[0036] Furthermore, the process of inverting the leaf area index of vegetation using the partial least squares regression method includes:

[0037] The partial least squares regression method was used to establish the relationship between hyperspectral reflectance and leaf area index, and the main features with the maximum covariance in hyperspectral reflectance and leaf area index were extracted. The expression is:

[0038] X=TP T +E

[0039] y=Tq+f

[0040] The leaf area index of vegetation is inverted based on the main features of the maximum covariance between hyperspectral reflectance and leaf area index.

[0041] The best parameter combination is selected based on the measured leaf area index and the corrected hyperspectral data, and the parameters of the vegetation canopy radiation transfer model are adjusted based on the parameter combination;

[0042] Among them, X represents the hyperspectral reflectance matrix, y represents the leaf area index vector, T represents the principal component matrix, P and q both represent the regression coefficient matrix, and E and f represent the residual matrix and error term, respectively.

[0043] Furthermore, the loss function expression set in the process of parameter optimization of the vegetation canopy radiation transfer model is:

[0044]

[0045] Where n represents the number of data points; y i represents the i-th true value; Y i represents the I-th predicted value; Represents the sum of squared errors between the true value and the predicted value.

[0046] Further, the process of extracting leaf area index-nitrogen content includes:

[0047] Construct an agricultural technology transfer decision support system model, and adjust the agricultural technology transfer decision support system model with different parameter range combinations according to environmental parameter data;

[0048] The agricultural technology transfer decision support system model was used to extract the correlation characteristics between leaf area index and nitrogen content under different parameter range combinations, and the loss function RMSE was used to calculate the loss of the correlation characteristics of each parameter range combination, and the parameter range combination with the smallest loss function RMSE output value was taken as the optimal parameter range combination;

[0049] The optimal parameter range combination was set as the new parameters of the agricultural technology transfer decision support system model for leaf area index-nitrogen content extraction;

[0050] Using the leaf area index as a link, the vegetation canopy radiation transfer model is coupled with the agricultural technology transfer decision support system model, the correlation characteristics between the canopy spectrum and the leaf nitrogen content are extracted, the grassland nitrogen content is inverted, and the inversion results are output.

[0051] In the above technical scheme, the vegetation canopy radiation transfer model (PROSAIL model) inverts the leaf area index (LAI) of vegetation by inputting hyperspectral data; partial least squares regression (PLSR) is a multiple regression method used to establish the relationship between the independent variable (spectral reflectance) and the dependent variable (LAI), which is particularly suitable for the situation where there is multicollinearity between high-dimensional data and independent variables. PLSR simultaneously extracts information from spectral reflectance data (independent variable) and LAI (dependent variable) to find the principal component that can explain the maximum covariance between the two. Environmental parameters are input into the Decision Support System for Agrotechnology Transfer (DSSAT) model to simulate the changes in leaf area index-leaf nitrogen content. The DSSAT model integrates environmental and management data to simulate the crop growth process of forage grass. The model simulation results need to be verified based on actual observations, and the model parameters are adjusted based on the verification results. The coupling of PROSAIL and DSSAT models is achieved by first inverting LAI through remote sensing hyperspectral data using the PROSAIL model, and then simulating the mapping relationship between leaf area index (LAI) and nitrogen content using DSSAT, and then using LAI as a connection to find the relationship between canopy spectrum and leaf nitrogen content. In this way, a method for inverting nitrogen content from spectral data based on the physiological process of plant growth and the physical process of spectral reflectance is implemented, thereby achieving stable and accurate extraction of plant nitrogen and improving the generalization ability of nitrogen extraction methods.

[0052] A grassland nitrogen content inversion system, the system comprising:

[0053] An acquisition module is used to obtain the grassland measured data and environmental parameter data of the target area;

[0054] A processing module, used for preprocessing the grassland measured hyperspectral data;

[0055] A feature extraction module is used to extract the mapping relationship characteristics of the hyperspectral spectrum of vegetation and the leaf area index based on the preprocessed grassland measured hyperspectral data;

[0056] The feature extraction module is also used to extract leaf area index-nitrogen content based on environmental parameter data;

[0057] The grassland nitrogen content inversion module is used to couple the extracted leaf area index-nitrogen content with the hyperspectral-leaf area index based on the leaf area index to invert the grassland nitrogen content.

[0058] Compared with the prior art, the present invention has the following beneficial effects:

[0059] The present invention proposes a grassland nitrogen content inversion method and system, which can reduce the spectral changes caused by scattering during the measurement of grassland nitrogen content by preprocessing the acquired grassland measured data, especially eliminate the spectral differences caused by different scattering levels, so as to standardize the correlation between the spectrum and the data; and use the extracted leaf area index as a connection, couple the extracted leaf area index-nitrogen content with the hyperspectral-leaf area index, and invert the grassland nitrogen content, which can better capture the dynamic changes of nitrogen in the process of plant growth, and has higher stability and portability, so as to quickly and effectively invert the grassland nitrogen content, thereby stably and accurately extracting plant nitrogen, and improving the generalization ability of the nitrogen extraction method. BRIEF DESCRIPTION OF THE DRAWINGS

[0060] Figure 1 A flowchart of the steps of a method for inverting grassland nitrogen content provided in an embodiment of the present application;

[0061] Figure 2 A specific flow chart for grassland nitrogen content inversion provided in the embodiment of the present application;

[0062] Figure 3 A schematic diagram of LAI inversion based on PLSR canopy spectrum provided in an embodiment of the present application;

[0063] Figure 4 Schematic diagram of simulated spectra generated under different LAI provided in the embodiments of the present application;

[0064] Figure 5 A schematic diagram of leaf area index inversion based on the PROSAIL model provided in an embodiment of the present application;

[0065] Figure 6 A schematic diagram of the corresponding relationship between LAI and LeafN% in non-gramineous plants and gramineous plants simulated by DSSAT provided in an embodiment of the present application;

[0066] Figure 7 A schematic diagram of the inversion results of the DSSAT and PROSAIL coupling model N provided in the embodiment of the present application;

[0067] Figure 8 A schematic diagram of the structure of a grassland nitrogen content inversion system provided in an embodiment of the present application. DETAILED DESCRIPTION

[0068] In order to facilitate the understanding of the present invention, the present invention will be described more fully below with reference to the relevant drawings. Preferred embodiments of the present invention are provided in the drawings. However, the present invention can be implemented in many different forms and is not limited to the embodiments described herein. On the contrary, the purpose of providing these embodiments is to make the understanding of the disclosure of the present invention more thorough and comprehensive.

[0069] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as those commonly understood by those skilled in the art of the present invention. The terms used herein in the specification of the present invention are only for the purpose of describing specific embodiments and are not intended to limit the present invention. The term "and / or" used herein includes any and all combinations of one or more related listed items.

[0070] Embodiment 1:

[0071] This embodiment provides a method for inverting grassland nitrogen content, see Figure 1 , the method comprises the following steps:

[0072] Step S1: obtaining grassland measured data and environmental parameter data of a target area, and preprocessing the grassland measured data;

[0073] Step S2: extracting the mapping relationship characteristics between the hyperspectral spectrum and the leaf area index of the vegetation according to the preprocessed grassland measured hyperspectral data;

[0074] Step S3: extracting leaf area index-nitrogen content according to environmental parameter data;

[0075] Step S4: Using the leaf area index as a link, the extracted leaf area index-nitrogen content is coupled with the leaf area index to invert the grassland nitrogen content.

[0076] In step S1, the process of preprocessing the grassland measured data and environmental parameter data includes:

[0077] The parameter range is divided according to the obtained grassland measured data, and the vegetation canopy radiation transfer model is constructed;

[0078] According to the measured grassland data after parameter range division, the basic parameters of the constructed vegetation canopy radiation transfer model are adjusted, and the spectral reflectance database under different parameters is generated using the vegetation canopy radiation transfer model;

[0079] Based on the spectral reflectance database under different parameters, the vegetation canopy radiation transfer model is used to extract the high spectral reflectance under different parameters;

[0080] According to the high spectral reflectance extracted by the vegetation canopy radiation transfer model, the canopy spectral curve data in the grassland measured data is spectrally corrected.

[0081] The grassland measured data include leaf area index and vegetation canopy hyperspectral data, and the environmental parameter data include meteorological data, soil data, field management data and field observation data.

[0082] Specifically, the canopy spectral curve data in the acquired grassland measured data is spectrally corrected, and the expression is:

[0083]

[0084] Among them, R sc (i,j) represents the reflectance value of the ith sample at wavelength j after scattering correction, R(i,j) represents the measured original reflectance value of the ith sample at wavelength j, mean(R(j)) represents the average value of the measured original reflectance value at wavelength j, sd(R(i)) represents the standard deviation of the measured original reflectance value at wavelength j, sd(S(j)) represents the standard deviation of the spectral reflectance value simulated by the vegetation canopy radiation transfer model at j, and mean(S(j)) represents the average value of the spectral reflectance value simulated by the vegetation canopy radiation transfer model at j.

[0085] The constructed vegetation canopy radiation transfer model (PROSAIL model) couples the leaf optical property model PROSPECT and the canopy reflectance model SAIL. The PROSPECT model simulates the reflectance and transmittance of vegetation leaves in the spectral range of 400nm to 2500nm. The PROSPECT model has two types of input parameters: the first type is the structural parameters of the leaves; the second type is the biochemical content parameters of the leaves, including chlorophyll content, water content and dry matter content. The SAIL model solves the scattering and absorption of four upward and downward radiation quantities to simulate the reflectance of the vegetation canopy. The SAIL model has two types of input parameters: the first type is the leaf reflectance and transmittance simulated by the PROSPECT model; the second type includes parameters such as leaf area index, average leaf inclination, hot spot parameters, solar zenith angle, observation zenith angle and relative azimuth. The expression of the constructed vegetation canopy radiation transfer model is:

[0086]

[0087] Among them, ρ c represents canopy reflectance, N represents leaf structure parameters, Cab represents chlorophyll content, Cw represents water content, Cm represents dry matter content, LAI represents leaf area index, ALA represents average leaf inclination, Hotspos represents hot spot parameters, Psoil represents soil brightness, rsoil represents soil moisture, θ v represents the observation zenith angle, θ x represents the solar zenith angle, Indicates the relative azimuth between the sun and the observation point.

[0088] It can be understood that spectral correction processing of the canopy spectral curve data in the acquired grassland measured data can reduce the spectral changes caused by scattering during the measurement process, especially eliminate the spectral differences caused by different scattering levels, thereby standardizing the correlation between the spectrum and the data, and reducing the influence of environmental factors on the canopy spectral curve measured in the field.

[0089] In step S2, see Figure 2 ,The process of extracting the mapping relationship features of the vegetation medium and high spectrum - leaf area index includes:

[0090] S21: Based on the spectral reflectance database under different parameters, the vegetation canopy radiation transfer model is used to extract the leaf area index under different parameters;

[0091] S22: The hyperspectral reflectance and leaf area index extracted by the vegetation canopy radiation transfer model are used as the training set, and the spectrally corrected grassland measured data are used as the validation set;

[0092] S23: Optimize the leaf area index inversion of the vegetation canopy radiation transfer model using the training set, and verify the trained vegetation canopy radiation transfer model using the verification set. After the verification is passed, the trained vegetation canopy radiation transfer model is obtained;

[0093] S24: Use the trained vegetation canopy radiation transfer model to invert the hyperspectral data in the input grassland measured data and extract the leaf area index of the vegetation.

[0094] Specifically, the range of PROSAIL model parameters is determined by measured data and references. After setting the parameters, the PROSAIL model is used to generate a spectral reflectance database under different parameters. The PROSAIL model has many parameters, and the generated spectral curve is hyperspectral data, which can express complex reflectance spectrum characteristics. In order to simplify the LAI inversion process based on the PROSAIL model, the present invention applies partial least squares regression to the spectral database generated based on the PROSAIL model. In this way, the inversion of LAI relies on the physical model and has the simple and fast characteristics of the empirical model. For specific observation data, the key to ensuring the accuracy of this inversion model is to determine the appropriate PROSAIL parameter range based on the measured data, among which the determination of the blade structure parameter N is particularly critical.

[0095] In step S23, the process of optimizing the leaf area index inversion of the vegetation canopy radiation transfer model using the training set includes:

[0096] Based on the hyperspectral spectral reflectance and leaf area index under different parameters in the training set, several rounds of optimization training are set up to traverse different parameter combinations to optimize the parameters of the vegetation canopy radiation transfer model;

[0097] The vegetation canopy radiation transfer model uses the partial least squares regression method to invert the leaf area index of vegetation based on the hyperspectral reflectance and leaf area index under different parameters;

[0098] The loss function is set to compare the leaf area index inversion results under different parameter combinations with the measured results, and the vegetation canopy radiation transfer model corresponding to the parameter combination with the smallest loss function value is taken as the trained vegetation canopy radiation transfer model.

[0099] Furthermore, the process of inverting the leaf area index of vegetation using the partial least squares regression method includes:

[0100] The partial least squares regression method was used to establish the relationship between hyperspectral reflectance and leaf area index, and the main features with the maximum covariance in hyperspectral reflectance and leaf area index were extracted. The expression is:

[0101] X=TP T +E

[0102] y=Tq+f

[0103] The leaf area index of vegetation is inverted based on the main features of the maximum covariance between hyperspectral reflectance and leaf area index.

[0104] The best parameter combination is selected based on the measured leaf area index and the corrected hyperspectral data, and the parameters of the vegetation canopy radiation transfer model are adjusted based on the parameter combination;

[0105] Among them, X represents the hyperspectral reflectance matrix, y represents the leaf area index vector, T represents the principal component matrix, P and q both represent the regression coefficient matrix, and E and f represent the residual matrix and error term, respectively.

[0106] Specifically, partial least squares regression (PLSR) is a multivariate regression method used to establish the relationship between the independent variable (spectral reflectance) and the dependent variable (LAI), which is particularly suitable for situations where there is multicollinearity between high-dimensional data and independent variables. PLSR extracts information from spectral reflectance data (independent variable) and LAI (dependent variable) at the same time to find the principal components that can explain the maximum covariance between the two. These principal components represent the main information in the spectrum and are closely related to the change of LAI.

[0107] The core formula of PLSR can be expressed as:

[0108] X=TP T +E (3)

[0109] y=Tq+f(4)

[0110] Among them, X is the spectral reflectance matrix, y is the LAI value vector, T is the principal component matrix, P and q are the regression coefficient matrices, and E and f are the residual matrix and error term.

[0111] After the training is completed, for a new spectral reflectance data X new , we can use the trained PLSR model to predict its corresponding LAI value:

[0112] LAI pred =T new q(5)

[0113] Among them, T new is the principal component extracted from the new spectral data.

[0114] Furthermore, the loss function expression set in the process of parameter optimization of the vegetation canopy radiation transfer model is:

[0115]

[0116] Where n represents the number of data points; y i represents the i-th true value; Y i represents the I-th predicted value; Represents the sum of squared errors between the true value and the predicted value.

[0117] Specifically, the specific model construction process is to use the spectral data simulated by PROSAIL and the input LAI as training data, and use partial least squares regression (PLSR) to establish the relationship between spectral features and LAI. In order to ensure the accuracy of the model, the measured spectral data is used to input the inverted LAI, and the root mean square error (RMSE) with the measured LAI is calculated. By traversing the blade structure N in the input parameters of the PROSAIL model, the most suitable parameter range of N is found according to the accuracy RMSE of the inversion result, and finally the LAI inversion process based on the PROSAIL model and PLSR is completed. By inputting hyperspectral data into the model, the inverted LAI value can be output.

[0118]

[0119] Where: n is the number of data points; y i is the i-th true value; Y i is the I-th predicted value; It is the sum of the squared errors between the true value and the predicted value. RMSE takes the mean of the squared errors and squares them to reflect the average difference between the model's predicted value and the true value. The smaller the value, the more accurate the model's prediction.

[0120] The main function of this model is to obtain a LAI model based on hyperspectral features. The LAI of vegetation is inverted by inputting hyperspectral data, and the inverted LAI is used as an input parameter of the DSSAT model to simulate the change of nitrogen content. In this way, the PROSAIL model and the DSSAT model are coupled to achieve the inversion from hyperspectral data to nitrogen content.

[0121] In step S3, see Figure 2 , the process of extracting leaf area index-nitrogen content includes:

[0122] Construct an agricultural technology transfer decision support system model, and adjust the agricultural technology transfer decision support system model with different parameter range combinations according to environmental parameter data;

[0123] The agricultural technology transfer decision support system model was used to extract the correlation characteristics between leaf area index and nitrogen content under different parameter range combinations, and the loss function RMSE was used to calculate the loss of the correlation characteristics of each parameter range combination, and the parameter range combination with the smallest loss function RMSE output value was taken as the optimal parameter range combination;

[0124] The optimal parameter range combination was set as the new parameters of the agricultural technology transfer decision support system model for leaf area index-nitrogen content extraction;

[0125] Using the leaf area index as a link, the vegetation canopy radiation transfer model is coupled with the agricultural technology transfer decision support system model, the correlation characteristics between the canopy spectrum and the leaf nitrogen content are extracted, the grassland nitrogen content is inverted, and the inversion results are output.

[0126] Specifically, the main purpose of the DSSAT model is to aggregate various types of crop models, standardize the input and output modes of the model, and realize the simulation requirements under different growth conditions after inputting the soil, climate, crop variety parameters and field management data of the corresponding test. DSSAT is not a universal model, and different models are developed for different crops. In order to simulate the growth of forage, the present invention uses the Forages-Alfalfa model and the Bermuda Grass model in DSSAT to simulate the growth process of non-grass and grass forage.

[0127] The input data of the DSSAT model mainly include meteorological data, soil data, management data and observation data. Specifically: (1) Meteorological data: The minimum data set must include precipitation, maximum temperature, minimum temperature, and solar radiation to drive crop growth simulation. First, the collected data needs to be organized into a format that DSSAT can recognize and input using the meteorological input module Weather Data. The operations that need to be performed when importing data include: establishing a meteorological station, naming, and inputting longitude and latitude. Finally, the "WTH" file format is generated for model operation. (2) Soil data: Soil properties are imported and stored in the model internal file using the soil input module soildata. Basic soil property information includes Clay%, Silt%, Stones%, Organic carbon%, pH in water, Cation exchange capacity cmol / kg, Totalnitrogen%, and finally the "SOL" file format is generated for model operation. (3) Field management data: Management data includes multiple parameters such as planting density, irrigation, fertilization, and harvesting. They are input and edited through the Xbuild interface of the Crop Management Data module. (4) Field observation data: After the field experiment is created, File A (End of Season) and File T (Time Course) are prepared through the Experimental Data module according to the time series data of leaf area index (LAI) and leaf nitrogen content (LeafN%) observed during the experiment, and input into the model to adjust the crop parameters in the DSSAT model locally (crop parameter files include .CUL, .ECO and .SPE) until the simulated LAI and Leaf N% are close to the observed LAI and Leaf N%.

[0128] The DSSAT model integrates environmental and management data to simulate the growth process of forage crops. The model simulation results need to be verified based on actual observations, and the model parameters are adjusted based on the verification results. The parameter adjustment process is to set different crop parameter range combinations, traverse each combination, and calculate the RMSE (RMSE) between the simulated values ​​of LAI and Leaf N% and the observed values. LAI and RMSE Leaf N％ ), select RMSE LAI and RMSE Leaf N％ The crop parameters with the smallest sum are taken as the optimal parameters.

[0129] In step S4, see Figure 2The coupling of PROSAIL and DSSAT models is that PROSAIL model first inverts LAI through remote sensing spectral data, and then DSSAT simulates the mapping relationship between leaf area index (LAI) and nitrogen content, and then uses LAI as a connection to find the relationship between canopy spectrum and leaf nitrogen content. In this way, a method of inverting nitrogen content from spectral data based on the physiological process of plant growth and the physical process of spectral reflectance is realized.

[0130] In the above technical scheme, the vegetation canopy radiation transfer model (PROSAIL model) inverts the leaf area index (LAI) of vegetation by inputting hyperspectral data, and the inverted LAI is used as the input parameter of the agricultural technology transfer decision support system model (DSSAT) to simulate the change of nitrogen content; partial least squares regression (PLSR) is a multivariate regression method used to establish the relationship between the independent variable (spectral reflectance) and the dependent variable (LAI), which is particularly suitable for the situation where there is multicollinearity between high-dimensional data and independent variables. PLSR extracts information from spectral reflectance data (independent variable) and LAI (dependent variable) at the same time to find the principal component that can explain the maximum covariance between the two. The DSSAT model integrates environmental and management data to simulate the crop growth process of forage. The model simulation results need to be based on actual observation results. The results were verified and the model parameters were adjusted according to the verification results. The coupling of PROSAIL and DSSAT models was achieved by firstly inverting LAI through remote sensing spectral data by PROSAIL model, and then simulating the mapping relationship between leaf area index (LAI) and nitrogen content by DSSAT, and then using LAI as the connection to find the relationship between canopy spectrum and leaf nitrogen content. In this way, a method for inverting nitrogen content from spectral data based on the physiological process of plant growth and the physical process of spectral reflectance was realized, so as to stably and accurately extract plant nitrogen and improve the generalization ability of nitrogen extraction methods.

[0131] In this embodiment, by preprocessing the acquired grassland measured data, the spectral changes caused by scattering during the measurement of grassland nitrogen content can be reduced, especially the spectral differences caused by different scattering levels can be eliminated, thereby standardizing the correlation between the spectrum and the data; and using the extracted leaf area index as a connection, the extracted leaf area index-nitrogen content is coupled with the leaf area index to invert the grassland nitrogen content, which can better capture the dynamic changes of nitrogen during plant growth, and has higher stability and portability, so as to quickly and effectively invert the grassland nitrogen content, thereby stably and accurately extracting plant nitrogen, and reducing the difficulty of plant nitrogen extraction.

[0132] Compared with other technologies, the present invention adopts the coupling of PROSAIL and DSSAT models instead of relying solely on spectral empirical models. The mapping relationship between LAI and nitrogen content generated by the DSSAT model is beneficial to improving the problem of poor spatiotemporal portability of spectral inversion nitrogen. Although the coupling complexity of the DSSAT and PROSAIL models is relatively high, compared with traditional empirical models, it can more accurately capture the dynamic process of vegetation growth and is suitable for large-scale remote sensing monitoring. This coupling method not only provides an effective remote sensing data assimilation framework, but also can supplement the inference of the vegetation state of grassland during the period without remote sensing data, reduce the limitation of the temporal resolution of remote sensing data, and provide a complete spatiotemporal dynamic distribution of grassland vegetation characteristics in large areas. It is of great significance for long-term monitoring and evolution analysis of grassland.

[0133] Embodiment 2:

[0134] This embodiment provides a grassland nitrogen content inversion method based on the first embodiment, and provides corresponding experimental data proof, as follows:

[0135] The leaf nitrogen content inversion effect based on the DSSAT and PROSAIL coupling model is close to the spectral empirical model of the canopy. The research results are conducive to capturing the dynamic changes of nitrogen during plant growth and have higher stability and portability.

[0136] See also Figure 3 , Partial least squares regression modeling of Leaf N% was performed at the canopy level, and the results were as follows: PLSR with ncomp=3 was the best, modeling: n=10, r=0.870, RMSE=0.489%; verification: n=4, r=0.592, RMSE=0.787%.

[0137] The DSSAT and PROSAIL models were coupled through LAI. Based on the measured canopy spectrum, the spectral database constructed by PROSAIL was used to invert the LAI, and then the nitrogen value corresponding to the original canopy spectrum was obtained through the mapping relationship between LAI and leaf nitrogen simulated by the DSSAT model. The nitrogen inversion evaluation results of the coupled model were r = 0.744 and RMSE = 0.777%.

[0138] Taking the measured grassland in the suburban area of ​​Guangzhou as an example, the first step of model coupling is to obtain accurate observation data to provide accurate input for the DSSAT and PROSAIL models. These data include: (1) leaf and canopy spectral data, which are collected by handheld ground spectrometers; (2) forage. Vegetation physical, chemical and physiological data, including leaf area index (LAI), chlorophyll content, carotene content, leaf nitrogen content, etc. These data are obtained through laboratory measurements to ensure that they can match the spectral data; (3). Environmental data: including meteorological data, including precipitation, maximum temperature, minimum temperature, solar radiation (such as temperature, rainfall, light intensity, etc.), which can be provided by the European Medium-Term Weather Centre; (4) Soil data, including Clay%, Silt%, Stones%, Organic carbon%, pH in water, Cationexchange capacity cmol / kg, Total nitrogen%, which can be provided by the World Soil Database (HWSD) (such as soil nitrogen content, humidity, etc.); and (5) Management measures (such as irrigation, etc.), such as irrigation and harvesting data, which need to be recorded according to actual conditions during the experiment and serve as input parameters for the DSSAT model.

[0139] 1) Operation of PROSAIL model

[0140] See also Figure 4 The parameter range of the model is determined according to the measured values ​​and the empirical values ​​provided in the references, as follows: ① Leaf structure parameter N: 1-3.5, increment 0.5; ② Chlorophyll Cab: 10-80ug / cm 2 , increment 2ug / cm 2 ; ③ Carotene Car: 2~24ug / cm 2 , increment 2ug / cm 2 ④ The water content Cw value is set to 0.014 g / cm 2 ⑤ The dry matter content Cm is set to 0.0125 g / cm 2 . After testing, for this grassland sample, the optimal leaf structure parameter range is 1.5 to 2.5. The LAI range is adjusted to 0.25 to 9.75 based on the measured LAI data, with an increment of 0.5. According to the measured soil spectrum curve, the soil spectrum parameters are set to 55% wet soil + 45% dry soil, and the parameters are input to generate a simulated spectrum. Figure 2 is the simulated spectral reflectance under different LAI.

[0141] According to formula (1), the measured spectrum is corrected, and then the PROSAIL model parameters are continuously adjusted to reduce the deviation between the predicted value and the actual observed value. The final modeling and verification results are as follows ( Figure 5 ), the gray dots are the LAI modeling and prediction results based on the PROSAIL spectral database. As LAI increases, the data is concentrated in the area with larger values, so the logarithm of the values ​​is taken to reduce the deviation of the model. The model prediction ability of PLSR modeling with the first principal component is the best. Among them, r = 0.969, RMSE = 1.29% for the modeling part; r = 0.877, RMSE = 0.83% for the validation part, and the model stability and prediction ability are both within the acceptable range.

[0142] 2)DSSAT model operation

[0143] See also Figure 6 , by setting up the weather, soil, and management module files in the DSSAT model for simulation, the forage models Forages-Alfalfa and Bermuda Grass are used to simulate the relationship between LAI and leaf nitrogen content (LeafN%) of C3 non-grass vegetation (left picture) and C4 grass vegetation (right picture). Among them, weather data is provided by the European Medium-Term Weather Center, soil data is provided by the World Soil Database (HWSD), and management data mainly includes irrigation records. By assimilating the measured LAI and nitrogen with the DSSAT model simulation data, debugging parameters such as plant sowing density, plant variety and ecology, the mapping relationship between LAI and LeafN% is obtained, such as Figure 6 .

[0144] 3) Model coupling

[0145] See also Figure 7 , DSSAT is coupled to the PROSAIL model through LAI. According to the measured canopy spectrum, the spectral database constructed by PROSAIL is used to invert the LAI, and then the nitrogen value corresponding to the original canopy spectrum is obtained through the mapping relationship between LAI and leaf nitrogen simulated by the DSSAT model. The nitrogen inversion evaluation results of the coupled model are r=0.744 and RMSE=0.777%.

[0146] Embodiment three:

[0147] This embodiment provides a grassland nitrogen content inversion system, see Figure 8 , the system comprising:

[0148] An acquisition module is used to obtain the grassland measured data and environmental parameter data of the target area;

[0149] A processing module, used for preprocessing the grassland measured data;

[0150] The feature extraction module is used to extract the mapping relationship characteristics of the vegetation medium and high spectrum-leaf area index based on the pre-processed grassland measured data;

[0151] The feature extraction module is also used to extract leaf area index-nitrogen content based on environmental parameter data;

[0152] The grassland nitrogen content inversion module is used to couple the extracted leaf area index-nitrogen content with the leaf area index based on the leaf area index to invert the grassland nitrogen content.

[0153] The above descriptions are merely embodiments of the present invention and are not intended to limit the patent scope of the present invention. Any equivalent structure or equivalent process transformation made using the contents of the present invention specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present invention.

Claims

1. A method for inverting grassland nitrogen content, characterized in that: The method comprises the following steps: Acquire grassland measured data and environmental parameter data of the target area, and preprocess the grassland measured data; Based on the pre-processed grassland measured data, the mapping relationship characteristics of vegetation medium and high spectrum-leaf area index are extracted; Extract leaf area index-nitrogen content based on environmental parameter data; Taking leaf area index as the connection, the extracted leaf area index-nitrogen content was coupled with the hyperspectral-leaf area index to invert the grassland nitrogen content.

2. The inversion method for grassland nitrogen content according to claim 1, characterized in that: The process of preprocessing the grassland measured hyperspectral data includes: The parameter range is divided according to the obtained grassland measured data, and the vegetation canopy radiation transfer model is constructed; According to the measured grassland data after parameter range division, the basic parameters of the constructed vegetation canopy radiation transfer model are adjusted, and the spectral reflectance database under different parameters is generated using the vegetation canopy radiation transfer model; Based on the spectral reflectance database under different parameters, the vegetation canopy radiation transfer model is used to extract the high spectral reflectance under different parameters; According to the high spectral reflectance extracted by the vegetation canopy radiation transfer model, the canopy spectral curve data in the grassland measured data is spectrally corrected. The grassland measured data include leaf area index and vegetation canopy hyperspectral data, and the environmental parameter data include meteorological data, soil data, field management data and field observation data.

3. The inversion method for grassland nitrogen content according to claim 2, characterized in that: The canopy spectral curve data in the grassland measured data is spectrally corrected, and the expression is: Among them, R sc (i,j) represents the reflectance value of the ith sample at wavelength j after scattering correction, R(i,j) represents the measured original reflectance value of the ith sample at wavelength j, mean(R(j)) represents the average value of the measured original reflectance value at wavelength j, sd(R(i)) represents the standard deviation of the measured original reflectance value at wavelength j, sd(S(j)) represents the standard deviation of the spectral reflectance value simulated by the vegetation canopy radiation transfer model at j, and mean(S(j)) represents the average value of the spectral reflectance value simulated by the vegetation canopy radiation transfer model at j.

4. The inversion method for grassland nitrogen content according to claim 2, characterized in that: The constructed vegetation canopy radiation transfer model expression is: Among them, ρ c represents canopy reflectance, N represents leaf structure parameters, Cab represents chlorophyll content, Cw represents water content, Cm represents dry matter content, LAI represents leaf area index, ALA represents average leaf inclination, Hotspot represents hotspot parameter, Psoil represents soil brightness, rsoil represents soil moisture, θ v represents the observation zenith angle, θ x represents the solar zenith angle, Indicates the relative azimuth between the sun and the observation point.

5. The inversion method for grassland nitrogen content according to claim 4, characterized in that: The process of extracting the mapping relationship characteristics between vegetation hyperspectral and leaf area index includes: According to the spectral reflectance database under different parameters, the leaf area index under different parameters is extracted using the vegetation canopy radiation transfer model; The hyperspectral reflectance and leaf area index extracted by the vegetation canopy radiation transfer model were used as the training set, and the measured grassland data after spectral correction were used as the validation set. The training set is used to optimize the leaf area index inversion of the vegetation canopy radiation transfer model, and the trained vegetation canopy radiation transfer model is verified using the verification set. After the verification, the trained vegetation canopy radiation transfer model is obtained. The trained vegetation canopy radiation transfer model is used to invert the hyperspectral data in the input grassland measured data and extract the leaf area index of the vegetation.

6. The inversion method of grassland nitrogen content according to claim 5, characterized in that: The process of optimizing the leaf area index inversion of the vegetation canopy radiation transfer model using the training set includes: Based on the hyperspectral reflectance and leaf area index under different parameters in the training set, several rounds of optimization training are set up to traverse different parameter combinations to optimize the parameters of the vegetation canopy radiation transfer model. The vegetation canopy radiation transfer model uses the partial least squares regression method to invert the leaf area index of vegetation based on the hyperspectral reflectance and leaf area index under different parameters; The loss function is set to compare the leaf area index inversion results under different parameter combinations with the measured results, and the vegetation canopy radiation transfer model corresponding to the parameter combination with the smallest loss function value is taken as the trained vegetation canopy radiation transfer model.

7. The inversion method for grassland nitrogen content according to claim 6, characterized in that: The process of inverting the leaf area index of vegetation using the partial least squares regression method includes: The partial least squares regression method was used to establish the relationship between hyperspectral reflectance and leaf area index, and the main features with the maximum covariance in hyperspectral reflectance and leaf area index were extracted. The expression is: X=TP T +E y=Tq+f The leaf area index of vegetation is inverted based on the main features of the maximum covariance between hyperspectral reflectance and leaf area index. The best parameter combination is selected based on the measured leaf area index and the corrected hyperspectral data, and the parameters of the vegetation canopy radiation transfer model are adjusted based on the parameter combination; Among them, X represents the hyperspectral reflectance matrix, y represents the leaf area index vector, T represents the principal component matrix, P and q both represent the regression coefficient matrix, and E and f represent the residual matrix and error term, respectively.

8. The inversion method for grassland nitrogen content according to claim 6, characterized in that: The loss function expression set in the process of parameter optimization of the vegetation canopy radiation transfer model is: Where n represents the number of data points; y i represents the i-th true value; Y i represents the i-th predicted value; Represents the sum of squared errors between the true value and the predicted value.

9. The inversion method for grassland nitrogen content according to any one of claims 1 to 8, characterized in that: The process of extracting LAI-Nitrogen content includes: Construct an agricultural technology transfer decision support system model, and adjust the agricultural technology transfer decision support system model with different parameter range combinations according to environmental parameter data; The agricultural technology transfer decision support system model was used to extract the correlation characteristics between leaf area index and nitrogen content under different parameter range combinations, and the loss function RMSE was used to calculate the loss of the correlation characteristics of each parameter range combination, and the parameter range combination with the smallest loss function RMSE output value was taken as the optimal parameter range combination; The optimal parameter range combination was set as the new parameters of the agricultural technology transfer decision support system model for leaf area index-nitrogen content extraction; Using the leaf area index as a link, the vegetation canopy radiation transfer model is coupled with the agricultural technology transfer decision support system model, the correlation characteristics between the canopy spectrum and the leaf nitrogen content are extracted, the grassland nitrogen content is inverted, and the inversion results are output.

10. A grassland nitrogen content inversion system, characterized in that: The system comprises: An acquisition module is used to obtain the grassland measured data and environmental parameter data of the target area; A processing module, used for preprocessing the grassland measured hyperspectral data; The feature extraction module is used to extract the mapping relationship characteristics of the vegetation medium and high spectrum-leaf area index based on the pre-processed grassland measured data; The feature extraction module is also used to extract leaf area index-nitrogen content based on environmental parameter data; The grassland nitrogen content inversion module is used to couple the extracted leaf area index-nitrogen content with the hyperspectral-leaf area index based on the leaf area index to invert the grassland nitrogen content.

Images