CN105352895A - Hyperspectral remote sensing data vegetation information extraction method - Google Patents

Abstract

The invention discloses a hyperspectral remote sensing data vegetation information extraction method. The method comprises in an experiment research zone, synchronously or quasi-synchronously acquiring a hyperspectral remote sensing image and reference standard values of physiological and biochemical parameters of vegetation to be detected, carrying out dimensionality reduction pre-treatment on the hyperspectral remote sensing image to obtain vegetation canopy spectral reflectance data, building an empirical mathematic model of the vegetation canopy spectral reflectance data and the vegetation parameter standard values to obtain a hyperspectral remote sensing estimation method of the vegetation physiological and biochemical parameters, through the empirical mathematic model, acquiring a vegetation parameter estimated value by the vegetation canopy spectral reflectance data obtained by the hyperspectral remote sensing image, carrying out estimation on the parameters of the model by the sample value and carrying out examination on a model precision. Through the hyperspectral remote sensing data, vegetation physiological and biochemical parameter estimation is realized so that the method is convenient and fast, does not influence vegetation growth, is suitable for large-area popularization and has a measurement area of the whole earth range.

Description

Hyperspectral remote sensing data vegetation information extraction method

Technical Field

The application relates to the technical field of vegetation information extraction, in particular to a method for extracting vegetation information from hyperspectral remote sensing data.

Background

Traditionally, the physiological and biochemical parameters of vegetation are mainly obtained through physicochemical experiments. The experiments are completely designed according to the definition of vegetation parameters, so the measurement precision is high. Disadvantages are the need to pick vegetation samples in the field, time and labor consuming, and destructive to vegetation. This experiment is not suitable for large area and only can be estimated using the sample as a whole. In addition, for areas that are not or not easily reachable by humans, it is simply not feasible to extract vegetation information using traditional methods.

The method for estimating physiological and biochemical parameters of vegetation by utilizing hyperspectral remote sensing data is an urgent need of scientific research such as ecology, agriculture, global change and the like and application industries such as precision agriculture and the like. The measuring method has the characteristics of rapidness and convenience, and the measuring area can even be enlarged to the size of the whole earth. And it is a non-destructive measurement that does not have any effect on vegetation growth.

The main methods for estimating the physiological and biochemical parameters of vegetation by utilizing the hyperspectral remote sensing images include a statistical analysis method based on an empirical model and a physical inversion method based on a radiation transmission model. The empirical model considers that certain characteristics of the reflectivity spectrum curve of the vegetation canopy and physiological and biochemical parameters of the vegetation have certain correlation, and the vegetation parameters are estimated by establishing a statistical regression equation between the spectrum characteristics and the parameters to be measured. The characteristic of the hyperspectral data extraction may be the value of a spectral reflectance curve on some special wave bands, or the value of a first derivative spectral reflectance curve on some special wave bands, and the like.

Through a large amount of analysis and research on vegetation spectra, the method provides that a plurality of spectral bands with remarkable characteristics are combined to obtain a vegetation index sensitive to a certain physiological and biochemical parameter of vegetation, and an empirical equation between the vegetation index and the physiological and biochemical parameter to be measured is established. It should be noted that due to the complexity and diversity of the vegetation remote sensing influencing factors, a spectral index with wide universality is developed. The development of the spectral vegetation index is required to follow the principle that it is as insensitive as possible to background interference and sensitive to the vegetation parameters to be estimated.

Different vegetation parameters can directly cause the change on the curve shape of the spectral reflectivity, and certain absorption characteristics or reflection characteristics of the spectral curve are parameterized, so that indexes reflecting the physiological and biochemical parameters of the vegetation can be obtained. The most widely used feature is the "red edge" specific to vegetation, defined as the wavelength position corresponding to the maximum of the first derivative of the spectral reflectance curve between 680 and 750nm wavelengths. The red-edge character of vegetation is sensitive to changes in chlorophyll, nitrogen, phenology, etc.

The physical model considers the radiation transmission mechanism of light below the vegetation canopy, and is strong in principle. And the influence of the vegetation physiological parameters on the spectral reflectivity is accurately analyzed in mechanism, and the robustness to noise is good. However, due to the problems of complicated structures of vegetation canopies and blades, radiation transmission process and the like, the model structure is complicated, and excessive variables may affect the effect of practical application.

The empirical model method considers that some characteristics of the vegetation spectral reflectivity curve, such as the reflectivity of a special wave band or a first derivative reflectivity value, a red edge characteristic, a vegetation index and the like, have statistical relevance with the physiological and biochemical parameters of the vegetation. The method for estimating the physiological and biochemical parameters of the vegetation by adopting an empirical model method comprises the following steps: firstly, a statistical regression model between spectral characteristics and vegetation parameters is established, then parameters in the model are estimated by utilizing ground measured data and hyperspectral remote sensing data of sample points, and finally the precision of the model is tested.

Disclosure of Invention

In view of this, the technical problem to be solved by the present application is to provide a method for extracting vegetation information from hyperspectral remote sensing data, which is fast and convenient to use, and can not affect the growth of vegetation, and the measurement area can even be expanded to the scale of the whole earth.

In order to solve the technical problem, the following technical scheme is adopted:

a hyperspectral remote sensing data vegetation information extraction method is characterized by comprising the following steps:

synchronously or quasi-synchronously acquiring a hyperspectral remote sensing image and a reference standard value of a physiological and biochemical parameter of vegetation to be measured in an experimental research area;

performing dimensionality reduction pretreatment on the high light remote sensing spectrum image to obtain spectral reflectivity data of a vegetation canopy;

establishing an empirical mathematical model between the spectral reflectance data of the vegetation canopy and the standard value of the vegetation parameter to obtain a hyperspectral remote sensing estimation method of the physiological and biochemical parameters of the vegetation;

obtaining an estimated value of a vegetation parameter through the empirical mathematical model and the spectral reflectance data of the vegetation canopy obtained from the hyperspectral image; and estimating parameters in the model by using the sample values, and finally checking the precision of the model.

Preferably, the performing the dimensionality reduction preprocessing on the highlight remote sensing spectrum image further comprises:

standardizing the data of each wave band of the original hyperspectral image X to obtain a standardized image matrix X_c；

Calculating a normalized matrix X_cOf the covariance matrix sigma_c；

Matrix sigma_cCharacteristic vector matrix A of_cThe eigenvectors are arranged from left to right according to the decreasing rule of the eigenvalues;

and performing linear transformation on the image data by using the obtained characteristic vector to obtain a PCA transformation result, wherein the calculation formula is as follows:

$X_{pca}=A_c^T\·X_c$

matrix X_pcaThe data in the first row represents the first principal component of the original hyperspectral image, the data in the second row represents the second principal component of the original hyperspectral image, and so on.

Preferably, the establishing an empirical mathematical model between the spectral reflectance data of the vegetation canopy and the standard value of the vegetation parameter is further:

establishing a linear statistical regression model between the spectral characteristics and the vegetation parameters:

y＝c₁x+c₂

estimating parameters in the model by using ground measured data and hyperspectral remote sensing data of sample points, and setting an observed value as y_iThe regression value isThe sum of squared errors criterion is expressed as:

$Q({\hat{c}}_1,{\hat{c}}_2)=\underset{c_1,c_2}{\min }\underset{i=1}{\overset{n}{\Σ}}{(y_i-{\hat{y}}_i)}^2,({\hat{y}}_i=c_1{x}_i+c_2)$

order to $\frac{\∂Q}{\∂c_1}=0$ And $\frac{\∂Q}{\∂c_2}=0,$ the parameter estimates are obtained as:

$\left\{\begin{array}{c}{c}_1=\frac{\underset{i=1}{\overset{n}{\Σ}}(x_i-\stackrel{\‾}{x})(y_i-\stackrel{\‾}{y})}{\underset{i=1}{\overset{n}{\Σ}}{(x_i-\stackrel{\‾}{x})}^2}\\ c_2=\stackrel{\‾}{y}-c_1\stackrel{\‾}{x}\end{array}\right.$

the accuracy of the model is checked: in the linear regression model, a variable x and a variable y have a linear relation, and the significance of the regression equation is checked by adopting a correlation coefficient between the variable x and the variable y, wherein the expression is as follows:

$r=\frac{\underset{i=1}{\overset{n}{\Σ}}\left(x_i-\stackrel{\‾}{x}\right)\left(y_i-\stackrel{\‾}{y}\right)}{\sqrt{\underset{i=1}{\overset{n}{\Σ}}{\left(x_i-\stackrel{\‾}{x}\right)}^2\underset{i=1}{\overset{n}{\Σ}}{\left(y_i-\stackrel{\‾}{y}\right)}^2}}$

the correlation coefficient r is a quantity representing the degree of closeness of the linear relation between the variable x and the variable y, and the value range of the correlation coefficient r is that | r | < 1; when r is more than 0, the variable x and the variable y are in positive correlation; when r is less than 0, the variable x and the variable y are in a negative correlation relationship; and greater | r | indicates greater linear correlation;

for a general regression model, the observed value y of the variable y is due to random error or variation of the independent variable x_iAre not exactly the same:

$\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-\stackrel{\&OverBar;}{y})}^2=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2+\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-\stackrel{\&OverBar;}{y}\right)}^2$

wherein, $S_T=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-\stackrel{\&OverBar;}{y})}^2$ which represents the sum of the squares of the total variation, $S_e=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2$ the sum of the squares of the residuals is expressed, the influence of the random error on the regression accuracy is expressed, the normalized value is the root mean square error RMSE, and the calculation formula is as follows:

$RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-y_i\right)}^2}{n}}$

in the formula, $S_R=\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-\stackrel{\&OverBar;}{y}\right)}^2$ is a regression sum of squares, reflecting the degree of dispersion of the dependent variable y due to the independent variable x; definition decision coefficient R²The calculation formula is as follows:

$R^2=\frac{S_R}{S_e}=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{({\hat{y}}_i-\stackrel{\&OverBar;}{y})}^2}{\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2}$

determining the coefficient R²Has a value range of R²≤1。

Preferably, the physiological and biochemical parameters of the vegetation comprise a leaf area index and a chlorophyll content, the leaf area index is calculated by using a normalized difference vegetation index, and the chlorophyll content is calculated by using a red edge characteristic of a reflectivity spectral curve.

Preferably, the method for calculating the leaf area index by using the normalized difference vegetation index comprises the following steps:

the normalized difference vegetation index NDVI is calculated by the following formula:

$NDVI=\frac{R_{nir}-R_{red}}{R_{nir}+R_{red}}$

wherein R is_nirAnd R_redRespectively representing the values of the spectral reflectivity at the near infrared and red light positions;

for the MODIS data, the empirical estimation model of the leaf area index is:

LAI＝0.3775·exp(2.4293·NDVI)；

for ASTER data, the empirical estimation model of the leaf area index is:

LAI＝0.3773·exp(2.4317·NDVI)。

preferably, the method for calculating the chlorophyll content by using the red edge characteristic of the reflectance spectrum curve comprises the following steps:

calculating the straight line square by linear extrapolation method I on the long-wave red light side by using two points with the wavelengths of 680nm and 694nmAnd (3) setting a linear equation positioned on the long-wave red light side as follows: FDR ═ m₁λ+c₁(ii) a Calculating a linear equation on the near infrared side by using two points with wavelengths of 724nm and 760nm, and setting the linear equation on the near infrared side as follows: FDR ═ m₂λ+c₂；

The calculation formula of the red edge wavelength is as follows:

$REP=\frac{-(c_1-c_2)}{m_1-m_2}$

the regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting the linear extrapolation method I as follows: CC ═ 1111.01+ 1.63. REP (R)²＝0.75)。

Preferably, the method for calculating the chlorophyll content by using the red edge characteristic of the reflectance spectrum curve comprises the following steps:

calculating a linear equation by using two points with the wavelengths of 680nm and 694nm on the long-wave red light side by adopting a linear extrapolation method II, and setting the linear equation on the long-wave red light side as follows: FDR ═ m₁λ+c₁(ii) a Calculating a linear equation on the near infrared side by using two points with wavelengths of 732nm and 760nm, and setting the linear equation on the near infrared side as follows: FDR ═ m₂λ+c₂；

The calculation formula of the red edge wavelength is as follows:

$REP=\frac{-(c_1-c_2)}{m_1-m_2}$

the regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting the linear extrapolation method II as follows: CC ═ 866.41+ 1.28. REP, (R)²＝0.70)。

Compared with the prior art, the method and the system have the advantages that:

firstly, the method for extracting the vegetation information from the hyperspectral remote sensing data solves the problems that the traditional vegetation physiological and biochemical parameters are obtained through a physicochemical experiment, time and labor are wasted, the vegetation is destructive, and the method is not suitable for being developed in a large area.

Secondly, the hyperspectral remote sensing data vegetation information extraction method provided by the invention realizes estimation of physiological and biochemical parameters of vegetation by using the hyperspectral remote sensing data, is fast and convenient, does not cause any influence on vegetation growth, and can even expand the measurement area to the scale of the whole earth.

Drawings

The accompanying drawings, which are included to provide a further understanding of the application and are incorporated in and constitute a part of this application, illustrate embodiment(s) of the application and together with the description serve to explain the application and not to limit the application. In the drawings:

FIG. 1 is a flow chart of a method for extracting vegetation information from hyperspectral remote sensing data according to the method of the invention;

FIG. 2 is a schematic diagram of the present invention employing linear extrapolation to determine the position of a red edge.

Detailed Description

As used in the specification and in the claims, certain terms are used to refer to particular components. As one skilled in the art will appreciate, manufacturers may refer to a component by different names. This specification and claims do not intend to distinguish between components that differ in name but not function. In the following description and in the claims, the terms "include" and "comprise" are used in an open-ended fashion, and thus should be interpreted to mean "include, but not limited to. "substantially" means within an acceptable error range, and a person skilled in the art can solve the technical problem within a certain error range to substantially achieve the technical effect. Furthermore, the term "coupled" is intended to encompass any direct or indirect electrical coupling. Thus, if a first device couples to a second device, that connection may be through a direct electrical coupling or through an indirect electrical coupling via other devices and couplings. The description which follows is a preferred embodiment of the present application, but is made for the purpose of illustrating the general principles of the application and not for the purpose of limiting the scope of the application. The protection scope of the present application shall be subject to the definitions of the appended claims.

Example 1

Referring to fig. 1, a specific embodiment of the method for extracting vegetation information from hyperspectral remote sensing data according to the present application is shown, and in this embodiment, the method includes the following steps:

101, synchronously or quasi-synchronously acquiring a hyperspectral remote sensing image and a reference standard value of a physiological and biochemical parameter of vegetation to be measured in an experimental research area;

102, performing dimensionality reduction pretreatment on the high-light remote sensing spectrum image to obtain spectral reflectivity data of a vegetation canopy;

103, establishing an empirical mathematical model between the spectral reflectance data of the vegetation canopy and the standard value of the vegetation parameter to obtain a hyperspectral remote sensing estimation method of the physiological and biochemical parameters of the vegetation;

104, obtaining an estimated value of a vegetation parameter through the spectral reflectance data of the vegetation canopy obtained from the hyperspectral image through the empirical mathematical model; and estimating parameters in the model by using the sample values, and finally checking the precision of the model.

The difference of the spectral reflectance curves caused by the difference of the basic composition components of the substances is a physical basis for analyzing and classifying the ground objects based on the hyperspectral remote sensing images, so that the hyperspectral remote sensing images and the reference standard values of the physiological and biochemical parameters of the vegetation to be measured need to be synchronously or quasi-synchronously obtained in an experimental research area. The hyperspectral remote sensing image has the advantages of large number of wave bands, close interval of the wave bands, certain redundancy of observation data, and large noise of images of certain wave bands due to the quality of the remote sensing detector. Therefore, the hyperspectral image is subjected to dimensionality reduction preprocessing, and the spectral reflectivity data of the vegetation canopy can be obtained.

In the method, the algorithm for performing dimensionality reduction preprocessing on the high-light remote sensing spectrum image is a principal component analysis method, and the basic idea is as follows: with the K-L transform, the variance of the data is used to describe the information volume, in an attempt to make the transformed data distribute in decreasing amounts according to the information volume. The method mainly comprises the steps of projecting original data into a new coordinate space by adopting a linear projection method, wherein the information content of a first principal component in the new coordinate space is maximum, a second principal component is irrelevant to the data of the first principal component, the information content of the second principal component in the rest components is maximum, and the like. For hyperspectral images with tens of bands, the first four and five principal components can generally contain more than 90% of the information content of the image, while higher components have substantially no information. Therefore, the first few principal components can be used for replacing the whole hyperspectral image, and the purpose of reducing the dimension is achieved. The specific method comprises the following steps:

standardizing the data of each wave band of the original hyperspectral image X to obtain a standardized image matrix X_c；

Calculating a normalized matrix X_cOf the covariance matrix sigma_c；

Matrix sigma_cCharacteristic vector matrix A of_cThe eigenvectors are arranged from left to right according to the decreasing rule of the eigenvalues;

and performing linear transformation on the image data by using the obtained characteristic vector to obtain a PCA transformation result, wherein the calculation formula is as follows:

$X_{pca}=A_c^T\&CenterDot;X_c$

matrix X_pcaThe data in the first row represents the first principal component of the original hyperspectral image, the data in the second row represents the second principal component of the original hyperspectral image, and so on.

In the method, the establishing of the empirical mathematical model between the spectral reflectance data of the vegetation canopy and the standard value of the vegetation parameter further comprises the following steps:

(1) establishing a statistical regression model between the spectral characteristics and the vegetation parameters:

y＝c₁x+c₂

(2) estimating parameters in the model by using ground measured data and hyperspectral remote sensing data of sample points, and setting an observed value as y_iThe regression value isThe sum of squared errors criterion is expressed as:

$Q({\hat{c}}_1,{\hat{c}}_2)=\underset{c_1,c_2}{\min }\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2,({\hat{y}}_i=c_1{x}_i+c_2)$

order to $\frac{\&part;Q}{\&part;c_1}=0$ And $\frac{\&part;Q}{\&part;c_2}=0,$ the parameter estimates are obtained as:

$\left\{\begin{array}{c}{c}_1=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\underset{i=1}{\overset{n}{\&Sigma;}}{(x_i-\stackrel{\&OverBar;}{x})}^2}\\ c_2=\stackrel{\&OverBar;}{y}-c_1\stackrel{\&OverBar;}{x}\end{array}\right.$

(3) the accuracy of the model is checked: in the linear regression model, a variable x and a variable y have a linear relation, and the significance of the regression equation is checked by adopting a correlation coefficient between the variable x and the variable y, wherein the expression is as follows:

$r=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}\left(x_i-\stackrel{\&OverBar;}{x}\right)\left(y_i-\stackrel{\&OverBar;}{y}\right)}{\sqrt{\underset{i=1}{\overset{n}{\&Sigma;}}{\left(x_i-\stackrel{\&OverBar;}{x}\right)}^2\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-\stackrel{\&OverBar;}{y}\right)}^2}}$

the correlation coefficient r is a quantity representing the degree of closeness of the linear relation between the variable x and the variable y, and the value range of the correlation coefficient r is that | r | < 1; when r is more than 0, the variable x and the variable y are in positive correlation; when r is less than 0, the variable x and the variable y are in a negative correlation relationship; and greater | r | indicates greater linear correlation;

for a general regression model, the observed value y of the variable y is due to random error or variation of the independent variable x_iAre not exactly the same:

$\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-\stackrel{\&OverBar;}{y})}^2=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2+\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-\stackrel{\&OverBar;}{y}\right)}^2$

wherein, $S_T=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-\stackrel{\&OverBar;}{y})}^2$ which represents the sum of the squares of the total variation, $S_e=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2$ the sum of the squares of the residuals is expressed, the influence of the random error on the regression accuracy is expressed, the normalized value is the root mean square error RMSE, and the calculation formula is as follows:

$RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-y_i\right)}^2}{n}}$

in the formula,is a regression sum of squares, reflecting the degree of dispersion of the dependent variable y due to the independent variable x; due to S_T＝S_e+S_RWhen S is_TAfter being given, S_RThe larger the size of S_eThe smaller the influence of the variable x on the variable y. And S_RThe smaller the size S_eThe larger the variable x, the less significant the effect of the variable y. Therefore, the determination coefficient R is defined²The calculation formula is as follows:

$R^2=\frac{S_R}{S_e}=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{({\hat{y}}_i-\stackrel{\&OverBar;}{y})}^2}{\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2}$

determining the coefficient R²Reflecting the significance of the influence of the variable x on the variable y in the selected regression model, wherein the value range is R²≤1。R²A larger value indicates that the change of the variable y is more significant due to the change of the variable x, whereas it indicates that the change of the variable x has little influence on the variable y.

Typical physiological and biochemical parameters of vegetation include leaf area index and chlorophyll content. The leaf area index is calculated from the normalized difference vegetation index, and R is set_nirAnd R_redRespectively representing the values of the spectral reflectivity at the near infrared and the red light, and the calculation formula of the Normalized Difference Vegetation Index (NDVI) is as follows:

$NDVI=\frac{R_{nir}-R_{red}}{R_{nir}+R_{red}}$

for the MODIS data, the empirical estimation model of the leaf area index is: LAI 0.3775 exp (2.4293 NDVI);

for ASTER data, the empirical estimation model of the leaf area index is: LAI is 0.3773 exp (2.4317 NDVI).

Calculating the chlorophyll content by using the red edge characteristics of the reflectance spectrum curve, which comprises the following steps:

the linear extrapolation method can track the slope changes of 700nm and 725nm near the peak value, and relieve the instability of the relationship between the vegetation physiological parameters and the red edge position caused by the double peak problem. The method extrapolates two straight lines at two sides of a first derivative curve of the spectral reflectivity near the red edge, and determines the position of the red edge by calculating the intersection point of the two straight lines, as shown in fig. 2.

The first straight line is arranged on the long-wave red light side, and the equation of the straight line is set as follows: FDR ═ m₁λ+c₁。

The second straight line is on the near infrared side, and the equation of the straight line is set as follows: FDR ═ m₂λ+c₂。

There are two methods of determining these two lines. The linear extrapolation method I calculates a straight-line equation using two points with wavelengths of 680nm and 694nm on the long-wave red side, and calculates a straight-line equation using two points with wavelengths of 724nm and 760nm on the near-infrared side. Linear extrapolation II calculates the equation of a straight line on the long-wave red side, again using two points with wavelengths of 680nm and 694nm, and on the near-infrared side, using two points with wavelengths of 732nm and 760 nm.

The two straight lines have the same lambda and FDR values at the intersection point, and the red edge wavelength calculation formula can be obtained by the following steps:

$REP=\frac{-(c_1-c_2)}{m_1-m_2}$

the regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting the linear extrapolation method I as follows: CC ═ 1111.01+ 1.63. REP (R)²＝0.75)。

The regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting the linear extrapolation method II as follows: CC ═ 866.41+ 1.28. REP, (R)²＝0.70)。

Example 2

An application example of the present invention is provided below.

The method for extracting the vegetation information from the hyperspectral remote sensing data comprises the following steps:

(1) firstly, reading a hyperspectral image X;

(2) performing principal component analysis and dimensionality reduction on the image: first, a normalized matrix X is calculated_cAnd its covariance matrix sigma_cFind matrix Σ_cCharacteristic vector matrix A of_cAnd obtaining an analysis transformation result:

$X_{pca}=A_c^T\&CenterDot;X_c$

(3) obtaining the spectral reflectivity of vegetation from the image after dimensionality reduction;

(4) establishing a linear statistical regression model between the spectral characteristics and the vegetation parameters:

y＝c₁x+c₂

let the observed value be y_iThe regression value isThe sum of squared errors criterion is expressed as:

$Q({\hat{c}}_1,{\hat{c}}_2)=\underset{c_1,c_2}{\min }\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2,({\hat{y}}_i=c_1{x}_i+c_2)$

order to $\frac{\&part;Q}{\&part;c_1}=0$ And $\frac{\&part;Q}{\&part;c_2}=0,$ the parameter estimates are obtained as:

$\left\{\begin{array}{c}{c}_1=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}(x_i-\stackrel{\&OverBar;}{x})(y_i-\stackrel{\&OverBar;}{y})}{\underset{i=1}{\overset{n}{\&Sigma;}}{(x_i-\stackrel{\&OverBar;}{x})}^2}\\ c_2=\stackrel{\&OverBar;}{y}-c_1\stackrel{\&OverBar;}{x}\end{array}\right.$

(5) the accuracy of the model is checked:

$R^2=\frac{S_R}{S_e}=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{({\hat{y}}_i-\stackrel{\&OverBar;}{y})}^2}{\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2}$

determining the coefficient R²Reflects the significance of the influence of the variable x on the variable y in the regression model, and the value range is R²≤1。R²A larger value indicates a more significant change in the variable y caused by a change in the variable x.

(6) Calculating the typical vegetation physiological and biochemical parameters:

the calculation method of the leaf area index comprises the following steps:

let R_nirAnd R_redRespectively representing the values of the spectral reflectivity at the near infrared and red light positions, and the normalized difference vegetation index NDVI is as follows:

$NDVI=\frac{R_{nir}-R_{red}}{R_{nir}+R_{red}}$

for the MODIS data, the empirical estimation model of the leaf area index is:

LAI＝0.3775·exp(2.4293·NDVI)；

for ASTER data, the empirical estimation model of the leaf area index is:

LAI＝0.3773·exp(2.4317·NDVI)。

the calculation of the chlorophyll content of the vegetation is as follows:

two straight lines are pushed out from two sides of the first derivative curve of the spectral reflectivity near the red edge, the first straight line is on the long-wave red light side, and the equation of the straight line is as follows: FDR ═ m₁λ+c₁(ii) a The second straight line is arranged at the near infrared sideThe linear equation is as follows: FDR ═ m₂λ+c₂。

The red edge wavelength calculation formula is:

$REP=\frac{-(c_1-c_2)}{m_1-m_2}$

the regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting a linear extrapolation method I as follows: CC ═ 1111.01+ 1.63. REP (R)²＝0.75)。

The regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting a linear extrapolation method II as follows: CC ═ 866.41+ 1.28. REP, (R)²＝0.70)。

The hyperspectral remote sensing data vegetation information extraction method provided by the invention comprises the steps of firstly carrying out principal component analysis dimensionality reduction processing on a hyperspectral image to obtain vegetation canopy spectral reflectance data, then establishing a linear regression empirical mathematical model between the vegetation canopy spectral reflectance data and a vegetation parameter standard value to obtain a hyperspectral remote sensing estimation method of physiological and biochemical parameters, and finally calculating a leaf area index according to a normalized difference vegetation index and estimating the content of vegetation chlorophyll by using the red edge characteristic of a reflectance spectral curve.

According to the embodiments, the application has the following beneficial effects:

firstly, the method for extracting the vegetation information from the hyperspectral remote sensing data solves the problems that the traditional vegetation physiological and biochemical parameters are obtained through a physicochemical experiment, time and labor are wasted, the vegetation is destructive, and the method is not suitable for being developed in a large area.

Secondly, the hyperspectral remote sensing data vegetation information extraction method provided by the invention realizes estimation of physiological and biochemical parameters of vegetation by using the hyperspectral remote sensing data, is fast and convenient, does not cause any influence on vegetation growth, and can even expand the measurement area to the scale of the whole earth.

As will be appreciated by one skilled in the art, embodiments of the present application may be provided as a method, apparatus, or computer program product. Accordingly, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment or an embodiment combining software and hardware aspects. Furthermore, the present application may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, and the like) having computer-usable program code embodied therein.

The foregoing description shows and describes several preferred embodiments of the present application, but as aforementioned, it is to be understood that the application is not limited to the forms disclosed herein, but is not to be construed as excluding other embodiments and is capable of use in various other combinations, modifications, and environments and is capable of changes within the scope of the inventive concept as expressed herein, commensurate with the above teachings, or the skill or knowledge of the relevant art. And that modifications and variations may be effected by those skilled in the art without departing from the spirit and scope of the application, which is to be protected by the claims appended hereto.

Claims (7)

1. A hyperspectral remote sensing data vegetation information extraction method is characterized by comprising the following steps:

synchronously or quasi-synchronously acquiring a hyperspectral remote sensing image and a reference standard value of a physiological and biochemical parameter of vegetation to be measured in an experimental research area;

performing dimensionality reduction pretreatment on the high light remote sensing spectrum image to obtain spectral reflectivity data of a vegetation canopy;

establishing an empirical mathematical model between the spectral reflectance data of the vegetation canopy and the standard value of the vegetation parameter to obtain a hyperspectral remote sensing estimation method of the physiological and biochemical parameters of the vegetation;

obtaining an estimated value of a vegetation parameter through the empirical mathematical model and the spectral reflectance data of the vegetation canopy obtained from the hyperspectral image; and estimating parameters in the model by using the sample values, and finally checking the precision of the model.

2. The method for extracting vegetation information according to claim 1,

the dimensionality reduction pretreatment is further carried out on the high-light remote sensing spectrum image:

standardizing the data of each wave band of the original hyperspectral image X to obtain a standardized image matrix X_c；

Calculating a normalized matrix X_cOf the covariance matrix sigma_c；

Matrix sigma_cCharacteristic vector matrix A of_cThe eigenvectors are arranged from left to right according to the decreasing rule of the eigenvalues;

and performing linear transformation on the image data by using the obtained characteristic vector to obtain a PCA transformation result, wherein the calculation formula is as follows:

$X_{pca}=A_c^T\&CenterDot;X_c$

matrix X_pcaThe data in the first row represents the first principal component of the original hyperspectral image, the data in the second row represents the second principal component of the original hyperspectral image, and so on.

3. The method for extracting vegetation information according to claim 1,

the establishing of the empirical mathematical model between the spectral reflectance data of the vegetation canopy and the standard value of the vegetation parameter further comprises the following steps:

establishing a linear statistical regression model between the spectral characteristics and the vegetation parameters:

y＝c₁x+c₂

estimating parameters in the model by using ground measured data and hyperspectral remote sensing data of sample points, and setting an observed value as y_iThe regression value isThe sum of squared errors criterion is expressed as:

$Q({\hat{c}}_1,{\hat{c}}_2)=\underset{c_1,c_2}{min}\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2,({\hat{y}}_i=c_1{x}_i+c_2)$

order to $\frac{\&part;Q}{\&part;c_1}=0$ And $\frac{\&part;Q}{\&part;c_2}=0,$ the parameter estimates are obtained as:

$\left\{\begin{array}{c}{c}_1=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}\left(x_i-\stackrel{\&OverBar;}{x}\right)\left(y_i-\stackrel{\&OverBar;}{y}\right)}{\underset{i=1}{\overset{n}{\&Sigma;}}{\left(x_i-\stackrel{\&OverBar;}{x}\right)}^2}\\ c_2=\stackrel{\&OverBar;}{y}-c_1\stackrel{\&OverBar;}{x}\end{array}\right.$

the accuracy of the model is checked: in the linear regression model, a variable x and a variable y have a linear relation, and the significance of the regression equation is checked by adopting a correlation coefficient between the variable x and the variable y, wherein the expression is as follows:

$r=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}\left(x_i-\stackrel{\&OverBar;}{x}\right)\left(y_i-\stackrel{\&OverBar;}{y}\right)}{\sqrt{\underset{i=1}{\overset{n}{\&Sigma;}}{\left(x_i-\stackrel{\&OverBar;}{x}\right)}^2\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-\stackrel{\&OverBar;}{y}\right)}^2}}$

the correlation coefficient r is a quantity representing the degree of closeness of the linear relation between the variable x and the variable y, and the value range of the correlation coefficient r is that | r | < 1; when r is more than 0, the variable x and the variable y are in positive correlation; when r is less than 0, the variable x and the variable y are in a negative correlation relationship; and greater | r | indicates greater linear correlation;

for a general regression model, the observed value y of the variable y is due to random error or variation of the independent variable x_iAre not exactly the same:

$\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-\stackrel{\&OverBar;}{y}\right)}^2=\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-{\hat{y}}_i\right)}^2+\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-\stackrel{\&OverBar;}{y}\right)}^2$

wherein, $S_T=\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-\stackrel{\&OverBar;}{y}\right)}^2$ which represents the sum of the squares of the total variation, $S_e=\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-{\hat{y}}_i\right)}^2$ the sum of the squares of the residuals is expressed, the influence of the random error on the regression accuracy is expressed, the normalized value is the root mean square error RMSE, and the calculation formula is as follows: $RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-y_i\right)}^2}{n}}$

in the formula,is a regression sum of squares, reflecting the degree of dispersion of the dependent variable y due to the independent variable x; definition decision coefficient R²The calculation formula is as follows:

$R^2=\frac{S_R}{S_e}=\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-\stackrel{\&OverBar;}{y}\right)}^2}{\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-{\hat{y}}_i\right)}^2}$

determining the coefficient R²Has a value range of R²≤1。

4. The method for extracting vegetation information according to claim 1,

the vegetation physiological and biochemical parameters comprise a leaf area index and a chlorophyll content, the leaf area index is obtained by utilizing normalized difference vegetation index calculation, and the chlorophyll content is obtained by utilizing red edge characteristics of a reflectivity spectral curve.

5. The method for extracting vegetation information according to claim 4, wherein the vegetation information is selected from the group consisting of vegetation information,

the method for calculating the leaf area index by utilizing the normalized difference vegetation index comprises the following steps:

the normalized difference vegetation index NDVI is calculated by the following formula:

$NDVI=\frac{R_{nir}-R_{red}}{R_{nir}+R_{red}}$

wherein R is_nirAnd R_redRespectively representing the values of the spectral reflectivity at the near infrared and red light positions;

for the MODIS data, the empirical estimation model of the leaf area index is:

LAI＝0.3775·exp(2.4293·NDVI)；

for ASTER data, the empirical estimation model of the leaf area index is:

LAI＝0.3773·exp(2.4317·NDVI)。

6. the method for extracting vegetation information according to claim 4, wherein the vegetation information is selected from the group consisting of vegetation information,

the method for calculating the chlorophyll content by using the red edge characteristics of the reflectance spectrum curve comprises the following steps:

calculating a linear equation by using two points with wavelengths of 680nm and 694nm on the long-wave red light side by adopting a linear extrapolation method I, and setting the linear equation on the long-wave red light side as follows: FDR ═ m₁λ+c₁(ii) a Calculating a linear equation on the near infrared side by using two points with wavelengths of 724nm and 760nm, and setting the linear equation on the near infrared side as follows: FDR ═ m₂λ+c₂；

The calculation formula of the red edge wavelength is as follows:

$REP=\frac{-\left(c_1-c_2\right)}{m_1-m_2}$

the regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting the linear extrapolation method I as follows: CC ═ 1111.01+ 1.63. REP (R)²＝0.75)。

7. The method for extracting vegetation information according to claim 4, wherein the vegetation information is selected from the group consisting of vegetation information,

the method for calculating the chlorophyll content by using the red edge characteristics of the reflectance spectrum curve comprises the following steps:

calculating a linear equation by using two points with the wavelengths of 680nm and 694nm on the long-wave red light side by adopting a linear extrapolation method II, and setting the linear equation on the long-wave red light side as follows: FDR ═ m₁λ+c₁(ii) a Calculating a linear equation on the near infrared side by using two points with wavelengths of 732nm and 760nm, and setting the linear equation on the near infrared side as follows: FDR ═ m₂λ+c₂；

The calculation formula of the red edge wavelength is as follows:

$REP=\frac{-\left(c_1-c_2\right)}{m_1-m_2}$

the regression equation and the precision between the red edge wavelength REP and the chlorophyll content CC are calculated by adopting the linear extrapolation method II as follows: CC ═ 866.41+ 1.28. REP, (R)²＝0.70)。