Method for inverting chlorophyll fluorescence spectra independent of reflectance training dataset

Abstract

The application discloses a method for inverting chlorophyll fluorescence spectrum without relying on reflectivity training dataset, mainly comprising the following steps: step 1: obtaining a large number of field measurement and chlorophyll fluorescence spectrum based on radiation transfer model simulation, and constituting a chlorophyll fluorescence spectrum training dataset; step 2: performing principal component analysis on the chlorophyll fluorescence spectrum in the training dataset to obtain fluorescence spectrum principal components; step 3: using the expansion formula of high-order Fourier series to reconstruct the real reflectivity without fluorescence information in the inversion model; step 4: inputting the linear combination of the first three fluorescence principal components into the radiation transfer model, using the ordinary least square method to calculate the unknowns in the model, obtaining the weight coefficients of the fluorescence principal components, and finally inverting the chlorophyll fluorescence spectrum. The application does not need any reflectivity training dataset, thereby overcoming the dependence of real reflectivity reconstruction on the training dataset.

Description

Technical Field

[0001] This invention belongs to the technical field of vegetation sunlight-induced chlorophyll fluorescence inversion, and in particular relates to a method for inverting chlorophyll fluorescence spectra without relying on reflectance training datasets. Background Technology

[0002] Over the past decade, the emergence of light-induced chlorophyll fluorescence (fluorescence) has attracted widespread attention from researchers. Fluorescence refers to a spectral signal emitted by plants under sunlight, typically in the range of 650-850 nanometers. In recent years, numerous studies have revealed the enormous potential and research value of fluorescence. Compared to single-band fluorescence, chlorophyll fluorescence spectroscopy provides richer information about plant physiology, offering a more comprehensive understanding of plant functional status. Fluorescence spectroscopy measurements can reflect the efficiency of photosynthesis, the effectiveness of light energy utilization, and the plant's response to various environmental stresses. This makes chlorophyll fluorescence spectroscopy an important tool in plant physiology and ecology research.

[0003] Currently, chlorophyll fluorescence inversion algorithms exist for canopy, airborne, and satellite scales. However, most algorithms can only invert fluorescence information in a single band or a narrow band. Researchers have also developed some chlorophyll fluorescence spectrum inversion algorithms, such as the improved full-spectrum fluorescence reconstruction algorithm (aFSR algorithm) and the full-spectrum fluorescence spectrum fitting method (F-SFM algorithm). These methods can achieve chlorophyll fluorescence spectrum inversion, but they are all limited by a prerequisite: the inversion accuracy is affected by the representativeness of the reflectance training dataset to different scenes in the real world. Due to the complexity and diversity of the real world, it is impossible to always guarantee that the reflectance training dataset is sufficiently representative, and this limitation restricts the generalization of existing algorithms to some extent. Summary of the Invention

[0004] To address the shortcomings of the aforementioned technical problems, this invention provides a method for retrieving chlorophyll fluorescence spectra that does not rely on a reflectance training dataset, thus overcoming the limitation of dependence on a reflectance training dataset.

[0005] The technical solution adopted in this invention is:

[0006] A method for retrieving chlorophyll fluorescence spectra that does not rely on reflectance training datasets includes the following steps:

[0007] Step 1: Obtain a large number of chlorophyll fluorescence spectra from field measurements and simulations based on radiative transfer models to form a chlorophyll fluorescence spectrum training dataset;

[0008] Step 2: Perform principal component analysis on the chlorophyll fluorescence spectra in the training dataset to obtain the principal components of the fluorescence spectra;

[0009] Step 3: Use the expansion of a higher-order Fourier series to reconstruct the true reflectance in the inversion model;

[0010] Step 4: Input the linear combination of the first three fluorescent principal components into the radiative transfer model, calculate the unknowns in the model using the ordinary least squares method, obtain the weight coefficients of each fluorescent principal component, and finally invert to obtain the chlorophyll fluorescence spectrum.

[0011] The method for retrieving chlorophyll fluorescence spectra without relying on reflectance training datasets is characterized in that, in step 1, chlorophyll fluorescence spectra under different vegetation types and atmospheric conditions are measured using a field fluorescence meter, and chlorophyll fluorescence spectra under various canopy parameter conditions are simulated using DART, SCOPE, and FluorWPS models to form a chlorophyll fluorescence spectrum training dataset.

[0012] The method for retrieving chlorophyll fluorescence spectra from a training dataset independent of reflectance is characterized in that, in step 2, principal component analysis (PCA) is used to perform PCA on the chlorophyll fluorescence spectra in the training dataset to extract spectral features and obtain the principal components of the fluorescence spectra. The PCA equation used is:

[0013] P = XW (1)

[0014] In the formula, P represents the principal component matrix, X represents the training dataset for chlorophyll fluorescence spectrum reconstruction, and W is a square matrix composed of feature vectors (Xi, Xj, Xi, Xj, and W is the square matrix composed of feature vectors (Xi, Xj ... T X).

[0015] The method for retrieving chlorophyll fluorescence spectra without relying on reflectance training datasets is characterized in that, in step 3, the true reflectance in the inversion model is reconstructed using the expansion of a higher-order Fourier series. The original radiative transfer equation is:

[0016]

[0017] In the formula, L represents radiance, E represents irradiance, R represents true reflectance excluding fluorescence contribution, F represents fluorescence, and λ represents wavelength. The true reflectance reconstruction equation used is:

[0018]

[0019] In the formula, R represents the true reflectivity, λ represents the wavelength, a0 is the intercept, n is the order of the Fourier series expansion, and a i and b i T1 is a coefficient, and T2 is the wavelength period.

[0020] The method for retrieving chlorophyll fluorescence spectra without relying on reflectance training datasets, in step 4, involves inputting a linear combination of the first three principal fluorescence components into a radiative transfer model. The unknowns in the model are calculated using ordinary least squares to obtain the weight coefficients of each principal fluorescence component, and finally, the chlorophyll fluorescence spectrum is retrieved. The complete radiative transfer equation used is as follows:

[0021]

[0022] In the formula, L represents radiance, E represents irradiance, and n f ω represents the number of fluorescent principal components. j Main component ψ j The weighting coefficients, where λ represents the wavelength, a0 is the intercept, n is the order of the Fourier series expansion, and a i and b i Let ω be the weighting coefficient, and T2 be the wavelength period. The weighting coefficient ω is solved using the least squares method. j Then, the chlorophyll fluorescence spectrum can be obtained by inversion.

[0023] This invention primarily develops a method for retrieving chlorophyll fluorescence spectra that does not rely on a reflectance training dataset, thus eliminating the limitation of such a dataset. First, a large number of measured and simulated chlorophyll fluorescence spectra under various conditions are acquired to form a fluorescence spectrum training dataset. Then, principal component analysis is used to decompose the dataset into eigenvalues ​​to extract the corresponding principal components of the fluorescence spectrum, which are then used to reconstruct the chlorophyll fluorescence spectrum. The core of this method lies in treating the true reflectance as a function of wavelength and using a higher-order Fourier series expansion to reconstruct the true reflectance. Furthermore, the measured radiance is considered as a superposition of vegetation reflectance and chlorophyll fluorescence spectral signals, constructing a radiative transfer equation and solving it using ordinary least squares, ultimately retrieving the chlorophyll fluorescence spectrum.

[0024] This invention offers the following advantages: it provides a method for retrieving chlorophyll fluorescence spectra without relying on a reflectance training dataset. This method can retrieve chlorophyll fluorescence spectra without a reflectance training dataset, unlike existing methods which require principal components of the reflectance spectrum. This method reduces the requirements for principal component types in chlorophyll fluorescence spectrum retrieval, is simple in principle, and is easier to promote and apply. Attached Figure Description

[0025] Figure 1 These are the first three principal components of the fluorescence spectrum;

[0026] Figure 2 The actual reflectance and the reconstructed reflectance;

[0027] Figure 3 The results are from the diurnal variation in chlorophyll fluorescence spectra.

[0028] Figure 4 The RRMSE results of chlorophyll fluorescence spectrum inversion as a function of wavelength; Detailed Implementation

[0029] The present invention will be described in detail below with reference to specific embodiments.

[0030] Step 1: Simulate chlorophyll fluorescence spectra under various parameter conditions using the DART, SCOPE, and FluorWPS models; measure chlorophyll fluorescence spectra of various vegetation types; combine the simulated data and measured data to form a fluorescence spectrum training dataset.

[0031] Step 2: Use principal component analysis (PCA) to extract features from the fluorescence spectroscopy training dataset to obtain the principal components of the fluorescence spectra. The specific formula for PCA used is as follows:

[0032] P = XW (1)

[0033] In the formula, P represents the principal component matrix, X represents the training dataset for chlorophyll fluorescence spectrum reconstruction, and W is a square matrix composed of feature vectors (Xi, Xj, Xi, Xj, and W is the square matrix composed of feature vectors (Xi, Xj ... T X).

[0034] Step 3: Reconstruct the true reflectance in the inversion model using the expansion of a higher-order Fourier series. The period of the wavelength is 800-650=150nm. The order of the Fourier series expansion was set from 50 to 500. It was found that the reflectance reconstruction result was stable in the region when the order was 350. Therefore, the order of the expansion was set to 350 for reflectance reconstruction. The formula used is as follows:

[0035]

[0036] In the formula, R represents the true reflectivity, λ represents the wavelength, a0 is the intercept, n is set to 350, and a i and b i The weighting factor is T, which is set to 150.

[0037] Step 4: Invert the chlorophyll fluorescence spectrum using the final canopy radiative transfer model, setting the number of principal components of the fluorescence spectrum to 3. The radiative transfer equation is as follows:

[0038]

[0039] In the formula, L represents radiance, E represents irradiance, and ω j Main component ψ j The weighting coefficients, where R represents the true reflectance, λ represents the wavelength, a0 is the intercept, and ai and b i These are the weighting coefficients. Solve for the weighting coefficients ω. j Then, the chlorophyll fluorescence spectrum can be obtained by inversion.

[0040] The first three principal components extracted from the chlorophyll fluorescence spectrum training dataset are as follows: Figure 1 As shown, the reflectance reconstruction results are as follows: Figure 2 As shown, the diurnal variation of the retrieved chlorophyll fluorescence spectrum is as follows: Figure 3 As shown, the RRMSE of the retrieved chlorophyll fluorescence spectrum varies with wavelength as follows: Figure 4 As shown. From Figure 2-4 As can be seen, the chlorophyll fluorescence spectroscopy inversion method developed in this paper can accurately reconstruct the chlorophyll fluorescence spectrum curve. The diurnal variation results of the inverted chlorophyll fluorescence spectrum show a clear diurnal variation trend of increasing in the morning and decreasing in the afternoon, which is very consistent with existing diurnal variation observation patterns. Furthermore, the fluorescence spectrum of the inverted results exhibits a clear bimodal characteristic, and the transition of the fluorescence spectrum over continuous time is very smooth. Compared with the actual chlorophyll fluorescence spectrum data, the average relative root mean square error (RRMSE) of the inversion results is only 1.961%, which directly reflects the reliability of this method. The formula for calculating RRMSE is as follows:

[0041]

[0042] In the formula, SIF r,i SIF represents the fluorescence value obtained from the inversion. t,i This represents the actual fluorescence value.

[0043] It should be understood that those skilled in the art can make improvements or modifications based on the above description, and all such improvements and modifications should fall within the protection scope of the appended claims.

Claims

1. A method for retrieving chlorophyll fluorescence spectra without relying on reflectance training datasets, characterized in that, Includes the following steps: Step 1: Obtain a large number of chlorophyll fluorescence spectra from field measurements and simulations based on radiative transfer models to form a chlorophyll fluorescence spectrum training dataset; Step 2: Perform principal component analysis on the chlorophyll fluorescence spectra in the training dataset to obtain the principal components of the fluorescence spectra; Step 3: Reconstruct the true reflectance in the inversion model using the expansion of a higher-order Fourier series: The original radiative transfer equation is: (2) In the formula, Represents radiance, Represents irradiance. Represents the true reflectance without fluorescence contribution. Represents fluorescence, The wavelength is represented; the equation used to reconstruct the true reflectance is: (3) In the formula, Represents true reflectance. Represents wavelength, It is the intercept. It is the order of the expansion of the Fourier series. and These are the weighting coefficients. The wavelength period; Step 4: Input the linear combination of the first three principal fluorescence components into the radiative transfer model, calculate the unknowns in the model using the ordinary least squares method, obtain the weighting coefficients of each principal fluorescence component, and finally invert to obtain the chlorophyll fluorescence spectrum; the complete radiative transfer equation used is as follows: (4) In the formula, Represents radiance, Represents irradiance. The number of fluorescent principal components. Main component The weighting coefficients, Indicates wavelength. It is the intercept. It is the order of the expansion of the Fourier series. and These are the weighting coefficients. The wavelength period is used; the weighting coefficients are solved using the least squares method. Then, the chlorophyll fluorescence spectrum can be obtained by inversion.

2. The method for retrieving chlorophyll fluorescence spectra without relying on reflectance training datasets as described in claim 1, characterized in that, In step 1, chlorophyll fluorescence spectra under different vegetation types and atmospheric conditions are measured using a field fluorescence meter, and chlorophyll fluorescence spectra under various canopy parameter conditions are simulated using the DART model, SCOPE model and FluorWPS model to form a chlorophyll fluorescence spectrum training dataset.

3. The method for retrieving chlorophyll fluorescence spectra without relying on reflectance training datasets as described in claim 1, characterized in that, In step 2, principal component analysis (PCA) is used to perform PCA on the chlorophyll fluorescence spectra in the training dataset to extract spectral features and obtain the principal components of the fluorescence spectra; the PCA equation used is: (1) In the formula, Represents the principal component matrix. The training dataset representing the reconstruction of chlorophyll fluorescence spectra. It is a square matrix composed of eigenvectors ( ).

Images