Estimation method of net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion

Abstract

The invention discloses a method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectivity inversion, and relates to the technical field of testing and analysis. The invention discloses a method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectivity inversion, and establishes a hyperspectral reflectivity inversion model of the net photosynthetic rate of onion leaves by combining principal component analysis with random forest, and analyzes and studies the accuracy of the hyperspectral reflectivity inversion model constructed by random forest, thereby providing a basis for rapid acquisition and non-destructive monitoring of the net photosynthetic rate (Pn) of onion leaves. The method is an effective means to solve the problem of non-destructive estimation of the net photosynthetic rate of onion tubular leaves.

Description

Technical Field

[0001] The invention relates to the technical field of agricultural product testing and analysis, and in particular to a method for estimating the net photosynthetic rate of onion leaves based on high-spectral reflectance inversion. Background Art

[0002] Onion is one of the most widely grown and consumed vegetables in Northeast my country. It plays an important role in the agricultural industry due to its unique flavor, nutritional value and multiple uses. Onion not only has anti-inflammatory and antioxidant properties, but can also be used for processing. Therefore, it plays an important role in boosting the local economy and increasing farmers' income.

[0003] In agricultural production, effectively obtaining the net photosynthetic rate of crops is crucial to the production management process. However, current photosynthesis meters can only detect the net photosynthetic rate (Pn) of leaf-shaped and needle-shaped leaves, and cannot detect crops with cylindrical leaves, such as onion leaves.

[0004] Without damaging the onion leaves, the photosynthesis meter's leaf chamber cannot clamp the tubular leaves to form a complete seal, resulting in inaccurate gas exchange measurements and thus inability to accurately detect the net photosynthetic rate. Therefore, the current use of photosynthesis meters to measure the net photosynthetic rate (Pn) of onion leaves can only involve splitting the tubular onion leaves with a knife to turn them into flakes, and then using the photosynthesis meter to detect them, which damages the integrity of the onion leaves. In addition, the heterogeneity of stomatal distribution on the surface of onion leaves further complicates the use of standard methods to measure the net photosynthetic rate.

[0005] The above problems highlight the limitations of using photosynthesis meters to measure non-standard plant leaves. Therefore, a new method is needed to achieve rapid and non-destructive detection of the net photosynthetic rate of onion leaves and provide key data for onion growth monitoring. Summary of the invention

[0006] In view of the problems existing in the prior art, the present invention provides an estimation method for the net photosynthetic rate of onion leaves based on inversion of hyperspectral reflectance. Inverting the net photosynthetic rate of crop leaves using hyperspectral reflectance is a fast, non-destructive and effective technical means.

[0007] In order to achieve the above object, the present invention provides the following technical solutions:

[0008] The present invention provides a method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion, comprising the following steps:

[0009] S1. Use a spectrometer to measure the hyperspectral reflectance of onions, cut the onion leaves with a blade, and use a photosynthesis meter to collect the net photosynthetic rate of the onion leaves;

[0010] S2. The principal component analysis method was used to reduce the dimension of onion hyperspectral reflectance, and the CARS algorithm was used to extract the spectral bands in the feature set as the key hyperspectral bands for the determination of onion net photosynthetic rate; the key hyperspectral bands were 454nm~520nm and 676nm~900nm;

[0011] S3. Using the random forest method, a model for estimating the net photosynthetic rate of onion leaves was established based on hyperspectral reflectance.

[0012] S4. Use the determination coefficient R 2 The root mean square error (RMSE) was used to evaluate the prediction performance of the established estimation model.

[0013] Furthermore, the method of obtaining the hyperspectral reflectance of onions by using a spectrometer in step S1 includes:

[0014] Preparation before measurement: The optical fiber receiving end is connected to the cosine corrector and fixed on the operating table to keep it horizontal; an adjustable standard tungsten lamp is used as the light source, and an infrared partition is added at the optical fiber connection point; the light source and the cosine corrector are kept at a fixed distance and set at a 45° angle, and the blade is placed flat just below the angle;

[0015] During measurement: close the operating window of the darkroom, measure the wavelength range of 400~1100 nm, and the step interval is 1 nm;

[0016] After measurement: After removing the outliers of the original hyperspectral reflectance collected, smoothing was performed and 105 samples were randomly collected as the final hyperspectral reflectance spectra of the test samples.

[0017] Furthermore, in step S1, a photosynthesis meter is used to collect the net photosynthetic rate of onion leaves, and the detection time is within 2 minutes.

[0018] Furthermore, the specific process of step S2 includes:

[0019] First, the data is standardized so that the mean of each feature is 0 and the standard deviation is 1. The standardized formula is:

[0020] ,

[0021] in, It is a feature The mean of is its standard deviation;

[0022] After normalization, calculate the covariance matrix of the dataset:

[0023] C ij ，

[0024] Among them, C ij Represents feature X i and X j covariance between, n is the number of samples, X ki and X kj are the Kth sample X i and X j Eigenvalues, and The eigenvalues ​​X are i and X j The mean of .

[0025] According to the size of the eigenvalue, the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to form the basis of the new space; then the original data matrix X is projected onto the selected eigenvector matrix V to obtain the reduced-dimensional data, which is expressed as:

[0026] Y=X·V,

[0027] Among them, Y is the data matrix after dimensionality reduction, and each column represents a principal component.

[0028] Furthermore, the establishment process of the step S3 model includes: first, by repeatedly sampling the original data set, the random forest generates multiple new data subsets, each subset is used to independently construct a decision tree; then, each split point of the decision tree is constructed, and the best segmentation feature is selected from a randomly selected feature subset; when making predictions, the final decision in the classification task is determined by the voting results of the majority of trees, and the regression task takes the average of the prediction results of each tree.

[0029] Furthermore, in step S4, the coefficient R is determined 2 The calculation formula is as follows:

[0030] ,

[0031] Among them, y act is the measured value, is the average value of the measured value; pred is the predicted value; n is the sample size; the coefficient of determination R² reflects the degree of fit of the model to the observed data. The closer its value is to 1, the better the fit.

[0032] Furthermore, the calculation formula of the root mean square error RMSE in step S4 is as follows:

[0033] ,

[0034] Among them, y act is the measured value; ypred is the predicted value; n is the sample size; RMSE is the square root of the mean of the sum of squares of the differences between the predicted value and the actual value, reflecting the size of the prediction error.

[0035] Compared with the prior art, the present invention has the following beneficial effects:

[0036] The method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion provided by the present invention combines principal component analysis with random forest to establish a hyperspectral reflectance inversion model for the net photosynthetic rate of onion leaves, and analyzes and studies the accuracy of the hyperspectral reflectance inversion model constructed by random forest, thereby providing a basis for rapid acquisition and non-destructive monitoring of the net photosynthetic rate (Pn) of onion leaves. The method is an effective means to solve the problem of non-destructive estimation of the net photosynthetic rate of onion tubular leaves. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in the present invention. For ordinary technicians in this field, other drawings can also be obtained based on these drawings.

[0038] Figure 1 The original spectrum of the net photosynthetic rate of onion leaves provided by the embodiment of the present invention.

[0039] Figure 2 A three-dimensional point diagram consisting of the first three principal components of the method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion provided by an embodiment of the present invention.

[0040] Figure 3 The cumulative credibility of the first 10 principal components of the estimation method for the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion provided by the embodiment of the present invention.

[0041] Figure 4 Characteristic band screening of a method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion provided in an embodiment of the present invention.

[0042] Figure 5 This is a prediction model result diagram of the method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0043] In order to enable those skilled in the art to better understand the technical solution of the present invention, the present invention will be further described in detail below in conjunction with the accompanying drawings. The following embodiments are only specific embodiments exemplified in this application, which are used to illustrate the technical solution of the present application, rather than to limit it, and the protection scope of the present application is not limited thereto. In the description of the present invention, it should be noted that the orientation or positional relationship indicated by the terms "inside", "outside", etc. is based on the orientation or positional relationship shown in the accompanying drawings, which is only for the convenience of describing the present invention and simplifying the description, rather than indicating or implying that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore cannot be understood as a limitation on the present invention.

[0044] The present invention provides a method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion, comprising the following steps:

[0045] S1. Use a spectrometer to measure the hyperspectral reflectance of onions, cut the onion leaves with a blade, and use a photosynthesis meter to collect the net photosynthetic rate of the onion leaves.

[0046] The data sources of this method are the onion hyperspectral reflectance and the net photosynthetic rate of tillering onion leaves. Specifically, the method of obtaining the onion hyperspectral reflectance by using a spectrometer includes:

[0047] Preparation before measurement: The optical fiber receiving end is connected to the cosine corrector and fixed on the operating table to keep it horizontal; an adjustable standard tungsten lamp is used as the light source, and an infrared baffle is added at the optical fiber connection point to reduce the light intensity of the infrared band and prevent excessive light from burning the blade; a fixed distance is maintained between the light source and the cosine corrector, and an angle of 45° is set, and the blade is placed flat directly below the angle;

[0048] During measurement: close the operating window of the darkroom to avoid interference from external light sources, ensuring more accurate measurement of hyperspectral reflectance. The measurement wavelength range is 400-1100 nm, with a step interval of 1 nm.

[0049] After measurement: After removing the outliers of the original hyperspectral reflectance collected, smoothing was performed and 105 samples were randomly collected as the final hyperspectral reflectance spectra of the test samples.

[0050] Figure 1 This is the original spectrum of the estimation method of the net photosynthetic rate of onion leaves based on the inversion of hyperspectral reflectance. It can be observed from the figure that in the visible light band of 400~700 nm, the spectral characteristics of plant leaves are the key characterization of their photosynthetic efficiency and physiological activities. This method analyzes the hyperspectral reflectance of onion leaves and reveals their absorption and reflection behaviors at different wavelengths, which is of great significance for optimizing photosynthesis.

[0051] In the blue light region (400-500 nm), the reflectivity is significantly lower, indicating that the light energy in this band is largely absorbed by chlorophyll a and b. The high absorption rate of chlorophyll to blue light is essential for plant photosynthesis and its growth and development. Blue light also affects the morphological construction of plants, such as promoting stem elongation and leaf expansion. In the green light region (500-600 nm), the reflectivity increases slightly. This phenomenon indicates that green light is less absorbed by chlorophyll, causing most of the green light to be reflected, which is one of the reasons why onion leaves are green. Although the direct absorption of green light is not as good as that of blue or red light, it can penetrate deeper into the leaves, activate the photosynthesis of the underlying chloroplasts, and contribute to the overall light energy utilization efficiency. In the red light region (600-700 nm), the reflectivity drops again, showing a low reflectivity pattern similar to that of the blue light region. This characteristic of red light indicates that it is a very effective light region in photosynthesis, mainly because chlorophyll has a strong absorption of red light, which contributes to the efficient conversion of energy and the conduct of photosynthesis.

[0052] These data show that by accurately measuring and analyzing the hyperspectral reflectance of different light bands, we can gain a deeper understanding of the photosynthetic dynamics and physiological responses of onions. Onion leaves effectively absorb light energy within these wavelength ranges for photosynthesis. In the near-infrared region of 700-1100 nm, the hyperspectral reflectance increases significantly, indicating that the leaf pigments reflect green light to a certain extent and strongly absorb blue and red light.

[0053] The method of using photosynthesis meter to collect the net photosynthetic rate of onion leaves is as follows: after testing the hyperspectral reflectance data, cut the onion leaves with a blade, put them flat in the leaf chamber, and record the accurate net photosynthetic rate value of the onion leaves after the value stabilizes. The detection time is within 2 minutes, and the photosynthetic capacity of the leaves will not change within 2 minutes. The test sample is consistent with the hyperspectral reflectance test sample.

[0054] S2. The principal component analysis (PCA) method was used to reduce the dimension of onion hyperspectral reflectance, and the CARS algorithm was used to extract the spectral bands in the feature set as the key hyperspectral bands for the determination of onion net photosynthetic rate; the key hyperspectral bands were 454nm~520nm and 676nm~900nm.

[0055] The core of principal component analysis is to find the main change direction of the data by decomposing the covariance matrix through eigenvalues, and determine the importance of these directions based on the size of the eigenvalues. By retaining the most important principal components, the information of the original data can be retained as much as possible while reducing the data dimension.

[0056] Specifically, the specific process of using principal component analysis (PCA) to reduce the dimensionality of onion hyperspectral reflectance includes:

[0057] Since principal component analysis (PCA) is very sensitive to the scale of the data, before performing PCA, the data is first standardized so that the mean of each feature is 0 and the standard deviation is 1. The standardized formula is:

[0058] ,

[0059] in, It is a feature The mean of is its standard deviation;

[0060] After normalization, calculate the covariance matrix of the dataset:

[0061] C ij ，

[0062] Among them, C ij Represents feature X i and X j covariance between, n is the number of samples, X ki and X kj are the Kth sample X i and X j Eigenvalues, and The eigenvalues ​​X are i and X j The mean of .

[0063] According to the size of the eigenvalue, the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to form the basis of the new space, and the size of the eigenvalue determines the importance of each basis;

[0064] Then project the original data matrix X onto the selected eigenvector matrix V to obtain the reduced-dimensional data, expressed as:

[0065] Y=X·V,

[0066] Among them, Y is the data matrix after dimensionality reduction, and each column represents a principal component.

[0067] Figure 2This is a three-dimensional dot map composed of the first three principal components of the estimation method of the net photosynthetic rate of onion leaves based on the inversion of hyperspectral reflectance. After the training set samples were reduced in dimension by PCA, the first three principal components PC1, PC2, and PC3 were used as x, y, and z coordinates to establish a three-dimensional distribution map of 105 samples. Among them, the variance contribution rates of PC1, PC2, and PC3 to the spectral matrix were 2.8%, 11.3%, and 22.5%, respectively, and the cumulative contribution reached 36.6%. According to the principles of statistics, in the process of model building, if too few principal components are selected, the information of some variables will be lost. If too many principal components are selected, the noise information in the spectrum will be introduced into the model, resulting in a fitting phenomenon and a decrease in the accuracy of model prediction. Therefore, this method selects a smaller number of principal components when the cumulative credibility does not change much.

[0068] After the spectrum is subjected to PCA dimension reduction, the cumulative credibility of the first 10 principal components is combined with the scatter plot and the cumulative credibility of the principal components as follows Figure 3 As shown. This figure reflects the cumulative contribution of each principal component to the total population variation explanation. Among them, PC1 (principal component 1): the component that explains the largest variance in the data set, that is, the direction of the largest change in the data. PC2 (principal component 2): explains the direction of the second largest variance, and is orthogonal to PC1 (no correlation). PC3 (principal component 3): explains the third largest variance, and so on... PC10 (principal component 10): explains the component with less variance. The cumulative credibility of the first 9 principal components is 92.9%, which almost fully explains the variation in the data. Therefore, the present invention selects the first 9 principal components for subsequent calculation and analysis.

[0069] After PCA dimension reduction, the present invention uses the competitive adaptive reweighting (CARS) algorithm to extract the spectral bands in the feature set. The feature band screening is as follows: Figure 4 As shown. The black line shown in the figure represents the original reflectance spectrum, and the red circle represents the key characteristic wavelengths selected by the CARS algorithm. These wavelengths are mainly concentrated in specific areas, such as the green light region and the near-infrared region. The spectral characteristics of these regions are closely related to the absorption of chlorophyll and the photosynthetic activity of plants. The screened characteristic wavelengths can effectively reflect the photosynthetic activity of plants, indicating that the CARS algorithm is effective in extracting spectral features related to plant photosynthesis. Finally, the present invention selects 454nm~520nm and 676nm~900nm as the key hyperspectral bands for determining the net photosynthetic rate of onions.

[0070] S3. Using the random forest method, a model for inversion and estimation of the net photosynthetic rate of onion leaves was established based on hyperspectral reflectance.

[0071] Specifically, the process of establishing the model in step S3 includes: first, by repeatedly sampling the original data set, the random forest generates multiple new data subsets, each of which is used to independently build a decision tree; then, each split point of the decision tree is constructed, and the best segmentation feature is selected from a randomly selected feature subset, which further enhances the diversity and robustness of the model; when making predictions, the final decision in the classification task is determined by the voting results of the majority of trees, and the regression task takes the average of the prediction results of each tree. The construction of each tree is independent of other trees, so that the entire model can more effectively perform data classification or regression analysis.

[0072] The invention uses 105 samples, obtains the first 9 principal components by PCA dimension reduction as input variables, and establishes an onion leaf net photosynthetic rate inversion estimation model based on random forest in MATLAB.

[0073] S4. Use the determination coefficient R 2 The root mean square error (RMSE) was used to evaluate the prediction performance of the established estimation model.

[0074] The R2 indicator reflects the model's ability to explain the variation in observed values, while the RMSE measures the average deviation between the model's predicted values ​​and the actual observed values. Together, the two provide comprehensive information about the accuracy and reliability of the model.

[0075] Coefficient of determination R 2 The calculation formula is as follows:

[0076] ,

[0077] Among them, y act is the measured value, is the average value of the measured value; pred is the predicted value; n is the sample size; the coefficient of determination R² reflects the degree of fit of the model to the observed data. The closer its value is to 1, the better the fit.

[0078] The calculation formula of root mean square error RMSE is as follows:

[0079] ,

[0080] Among them, y act is the measured value; y pred is the predicted value; n is the sample size; RMSE is the square root of the mean of the sum of squares of the differences between the predicted value and the actual value, reflecting the size of the prediction error.

[0081] In order to evaluate the accuracy of the model, 1204 eigenvalues ​​of the 454nm~520nm and 676nm~900nm bands were input into the prediction model established by the method of the present invention. After calculation, the determination coefficient R of the prediction model of the present invention was2 The R² value of 0.94 indicates that the model can explain 94% of the data variation, showing a very high prediction accuracy. Figure 5 This figure shows the relationship between the predicted results of the estimation model of the net photosynthetic rate of onion leaves from spectral inversion and the actual measured values. Figure 5 The horizontal axis represents the true value, and the vertical axis represents the predicted value. The distribution of data points shows the corresponding relationship between the model predicted value and the true value. Figure 5 It can be observed that the data points are closely arranged around the diagonal line, indicating a high consistency between the predicted values ​​and the true values.

[0082] The above-described embodiments are only specific implementation methods of the present application, which are used to illustrate the technical solutions of the present application, rather than to limit them. The protection scope of the present application is not limited thereto. Although the present application is described in detail with reference to the aforementioned embodiments, ordinary technicians in the field should understand that any technician familiar with the technical field can still modify the technical solutions recorded in the aforementioned embodiments within the technical scope disclosed in the present application, or can easily think of changes, or make equivalent replacements for some of the technical specialties therein; and these modifications, changes or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application. They should all be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion, characterized in that: The following steps are involved: S1. Use a spectrometer to measure the hyperspectral reflectance of onions, cut the onion leaves with a blade, and use a photosynthesis meter to collect the net photosynthetic rate of the onion leaves; S2. The principal component analysis method was used to select the first 9 principal components to reduce the dimension of the onion hyperspectral reflectance, and the CARS algorithm was used to extract the spectral bands in the feature set as the key hyperspectral bands for the determination of the net photosynthetic rate of onions; the key hyperspectral bands were 454nm-520nm and 676nm-900nm; S3. Using the random forest method, a model for inversion and estimation of the net photosynthetic rate of onion leaves was established based on hyperspectral reflectance. S4. Use the determination coefficient R 2 The root mean square error (RMSE) was used to evaluate the prediction performance of the established estimation model.

2. The method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion according to claim 1, characterized in that: Step S1: The method of using a spectrometer to measure and obtain the hyperspectral reflectance of onions includes: Preparation before measurement: The optical fiber receiving end is connected to the cosine corrector and fixed on the operating table to keep it horizontal; an adjustable standard tungsten lamp is used as the light source, and an infrared partition is added at the optical fiber connection point; the light source and the cosine corrector are kept at a fixed distance and set at a 45° angle, and the blade is placed flat just below the angle; During measurement: close the operating window of the darkroom, measure the wavelength range of 400-1100nm, and the step interval is 1nm; After measurement: After removing the outliers of the original hyperspectral reflectance collected, smoothing was performed and 105 samples were randomly collected as the final hyperspectral reflectance spectra of the test samples.

3. The method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion according to claim 1, characterized in that: Step S1 uses a photosynthesis meter to collect the net photosynthetic rate of onion leaves, and the detection time is within 2 minutes.

4. The method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion according to claim 1, characterized in that: The specific process of step S2 includes: First, the data is standardized so that the mean of each feature is 0 and the standard deviation is 1. i The standardized formula is: Among them, X i normal Yes X i The original value of It is feature X i The mean value, σ i is its standard deviation; After normalization, calculate the covariance matrix of the dataset: Among them, C ij Represents feature X i and X j covariance between, n is the number of samples, X ki and X kj are the Kth sample X i and X j Eigenvalues, and The eigenvalues ​​X are i and X j The mean of According to the size of the eigenvalue, the eigenvectors corresponding to the first k largest eigenvalues ​​are selected to form the basis of the new space; then the original data matrix X is projected onto the selected eigenvector matrix V to obtain the reduced-dimensional data, which is expressed as: Y=X·V Among them, Y is the data matrix after dimensionality reduction, and each column represents a principal component.

5. The method for estimating the net photosynthetic rate of onion leaves based on inversion of hyperspectral reflectance according to claim 1, characterized in that: The process of establishing the model in step S3 includes: first, by repeatedly sampling the original data set, the random forest generates multiple new data subsets, each subset is used to independently build a decision tree; then, each split point of the decision tree is constructed, and the best segmentation feature is selected from a randomly selected feature subset; when making predictions, the final decision in the classification task is determined by the voting results of the majority of trees, and the regression task takes the average of the prediction results of each tree.

6. The method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion according to claim 1, characterized in that: In step S4, the coefficient R is determined 2 The calculation formula is as follows: Among them, y act is the measured value, is the average value of the measured value; pred is the predicted value; n is the sample size; the determination coefficient R 2 It reflects the degree of fit of the model to the observed data. The closer its value is to 1, the better the fit.

7. The method for estimating the net photosynthetic rate of onion leaves based on hyperspectral reflectance inversion according to claim 1, characterized in that: The calculation formula of the root mean square error RMSE in step S4 is as follows: Among them, y act is the measured value; y pred is the predicted value; n is the sample size; RMSE is the square root of the mean of the sum of squares of the differences between the predicted value and the actual value, reflecting the size of the prediction error.

Images