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

High-spectrum remote sensing data vegetation information extraction method

Technical field

The application relates to vegetation information extraction technical field, specifically, relates to a kind of high-spectrum remote sensing data vegetation information extraction method.

Background technology

Traditionally, vegetation Physiology and biochemistry parameter is mainly obtained by Physical Chemistry Experiment.These experiments design according to the definition of vegetation parameter completely, and therefore measuring accuracy is higher.Shortcoming needs to pluck vegetation sample on the spot, wastes time and energy, and have destructiveness to vegetation.This experiment is not suitable for carrying out in large area, and sample can only be used to estimate entirety.In addition, the mankind or can not be not easy to the region arrived, it is completely infeasible for using conventional methods and extracting vegetation information.

Utilizing the estimation of high-spectrum remote sensing data realization to vegetation Physiology and biochemistry parameter, is the active demand of the application industries such as scientific research and precision agriculture such as ecology, agronomy, whole world change.This measuring method has feature fast and easily, and its measured zone even can expand the yardstick of the whole earth to.And it is a kind of nondestructive measurement method, and it can not cause any impact to vegetation growth.

The main method utilizing target in hyperspectral remotely sensed image to carry out vegetation Physiology and biochemistry parameter estimation has the statistical analysis method based on empirical model and the physics method of inversion based on radiative transfer model.Empirical model thinks to have certain correlationship between some feature of Vegetation canopy reflectance spectrum curve and the Physiology and biochemistry parameter of vegetation, is estimated vegetation parameter by the regression prediction equation set up between spectral signature and parameter to be measured.The feature that high-spectral data extracts may be the value of spectral reflectance rate curve in some special wave band, or the value etc. of first derivative spectrum reflectance curve in some special wave band.

By people to the large component analysis of vegetation spectrum and research, propose and multiple spectral band with notable feature is combined, obtain the vegetation index of certain the Physiology and biochemistry sense parameters to vegetation, and set up the empirical equation between vegetation index and Physiology and biochemistry parameter to be measured.It should be noted that the complicacy due to remote sensing of vegetation factor of influence and diversity, a spectrum index with extensive universality be developed and shoulder heavy responsibilities.The principle that the development need of spectral vegetation indexes is followed is, as far as possible insensitive to background interference, and to vegetation sense parameters to be evaluated.

Different vegetation parameters directly can cause the change on spectral reflectivity curve shape, by certain Absorption Characteristics of the curve of spectrum or reflectance signature parametrization, can obtain the index reflecting vegetation Physiology and biochemistry parameter.Most widely used feature is vegetation specific " red limit ", is defined as the wavelength location that the first order derivative maximal value of spectral reflectance rate curve between 680-750nm wavelength is corresponding.The red limit change of characteristic to chlorophyll, nitrogen, phenology etc. of vegetation is very sensitive.

Physical model considers the radiation transmission mechanism of light below Vegetation canopy, and principle is strong.And the restriction of the factors such as time place can not be subject within the scope of the original hypothesis of model, in mechanism, the accurate analysis impact of vegetation physiological parameter on spectral reflectivity, good to the Shandong nation property of noise.But due to the problem that the structure, road radiation transmission process etc. that relate to Vegetation canopy and blade are complicated, model structure meeting more complicated, too much variable may affect the effect of practical application.

Empirical model method thinks some feature of vegetation spectral reflectance rate curve, the reflectivity of such as special wave band or first-derivative reflectance value, red limit feature, vegetation index etc., and has the correlativity in statistical significance between vegetation Physiology and biochemistry parameter.The step of vegetation Physiology and biochemistry parameter is to adopt empirical model method to estimate: first set up the statistical regression model between spectral signature and vegetation parameter, then utilize the ground measured data of sample point and high-spectrum remote sensing data to estimate the parameter in model, finally the precision of model is tested.

Summary of the invention

In view of this, technical problems to be solved in this application there is provided a kind of high-spectrum remote sensing data vegetation information extraction method, utilize the estimation of high-spectrum remote sensing data realization to vegetation Physiology and biochemistry parameter, efficient and convenient, can not cause any impact to vegetation growth, its measured zone even can expand the yardstick of the whole earth to.

In order to solve the problems of the technologies described above, the application has following technical scheme:

A kind of high-spectrum remote sensing data vegetation information extraction method, is characterized in that, comprising:

In experimental study region, synchronous or standard synchronously obtains the normative reference value of target in hyperspectral remotely sensed image and vegetation Physiology and biochemistry parameter to be measured;

Dimensionality reduction pre-service is carried out to described high light remote sensing spectrum image, obtains Vegetation canopy spectral reflectance data;

Set up the empirical mathematical model between Vegetation canopy spectral reflectance data and vegetation parameter standard value, obtain the high-spectrum remote-sensing evaluation method of vegetation Physiology and biochemistry parameter;

By described empirical mathematical model, obtained the estimated value of vegetation parameter by the Vegetation canopy spectral reflectance data obtained from Hyperspectral imaging; Sample value is utilized to estimate the parameter in model, the precision of last testing model.

Preferably, wherein, carrying out dimensionality reduction pre-service to described high light remote sensing spectrum image is further:

Standardization is carried out to the data of former each wave band of Hyperspectral imaging X, obtains standardized images matrix X _c;

Normalized matrix X _ccovariance matrix Σ _c;

Ask matrix Σ _ceigenvectors matrix A _c, the rule that wherein proper vector is successively decreased according to eigenwert arranges from left to right;

Utilize the proper vector of trying to achieve to carry out linear transformation to view data and obtain PCA transformation results, computing formula is:

$X_{pca}=A_c^T\·X_c$

Matrix X _pcathe first principal component of the former high spectrum image of middle the first row data representation, the Second principal component, of the former high spectrum image of the second row data representation, by that analogy.

Preferably, wherein, the described empirical mathematical model set up between Vegetation canopy spectral reflectance data and vegetation parameter standard value is further:

Set up the linear statistical regression model between spectral signature and vegetation parameter:

y＝c ₁x+c ₂

The ground measured data of sample point and high-spectrum remote sensing data is utilized to estimate the parameter in model, if observed reading is y _i, regressand value is least error sum-of-squares 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 $\frac{\∂Q}{\∂c_1}=0$ With $\frac{\∂Q}{\∂c_2}=0,$ Obtaining parameter estimation amount is:

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

Test to the precision of model: relation linear between variable x and variable y in linear regression model (LRM), adopt the related coefficient between variable x and variable y to check the conspicuousness of regression equation, expression formula is as follows:

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

Correlation coefficient r is the amount representing variable x and variable y linear relationship level of intimate, and its span is | r|≤1; As r > 0, variable x and variable y correlation; As r < 0, variable x and variable y is negative correlativing relation; And | r| larger expression linear dependence is stronger;

For general regression model, due to the change of stochastic error or independent variable x, the observed reading y of variable y _iincomplete 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, $S_T=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-\stackrel{\&OverBar;}{y})}^2$ Represent total variance quadratic sum, $S_e=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2$ Represent residual sum of squares (RSS), represent that stochastic error is on the impact of regression accuracy, its standardized value is root-mean-square error RMSE, and computing formula is:

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

In formula, $S_R=\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-\stackrel{\&OverBar;}{y}\right)}^2$ For regression sum of square, reflect the degree of scatter of the dependent variable y caused due to independent variable x; Definition coefficient of determination R ², computing formula is:

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

Coefficient of determination R ²span be R ²≤ 1.

Preferably, wherein, described vegetation Physiology and biochemistry parameter comprises leaf area index and chlorophyll content, and described leaf area index utilizes normalization difference vegetation index to calculate, and described chlorophyll content utilizes the red limit feature calculation of reflectance spectrum curve to obtain.

Preferably, wherein, the method utilizing normalization difference vegetation index to calculate described leaf area index is:

Normalization difference vegetation index NDVI is obtained by following formulae discovery:

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

Wherein, R _nirand R _redrepresent the value of spectral reflectivity near infrared and ruddiness place respectively;

For MODIS data, the empirical estimating model of leaf area index is:

LAI＝0.3775·exp(2.4293·NDVI)；

For ASTER data, the empirical estimating model of leaf area index is:

LAI＝0.3773·exp(2.4317·NDVI)。

Preferably, wherein, the method for chlorophyll content described in the red limit feature calculation of reflectance spectrum curve is utilized to be:

Adopt two some calculated line equations that linear extrapolation I utilizes wavelength to be 680nm and 694nm in long-red-wave side, the straight-line equation being set in long-red-wave side is: FDR=m ₁λ+c ₁; Two the some calculated line equations utilizing wavelength to be 724nm and 760nm near infrared side, the straight-line equation being set near infrared side is: FDR=m ₂λ+c ₂;

Show that red limit wavelength equation is:

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

Described linear extrapolation I is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-1111.01+1.63REP, (R ²=0.75).

Preferably, wherein, the method for chlorophyll content described in the red limit feature calculation of reflectance spectrum curve is utilized to be:

Adopt two some calculated line equations that linear extrapolation II utilizes wavelength to be 680nm and 694nm in long-red-wave side, the straight-line equation being set in long-red-wave side is: FDR=m ₁λ+c ₁; Two the some calculated line equations utilizing wavelength to be 732nm and 760nm near infrared side, the straight-line equation being set near infrared side is: FDR=m ₂λ+c ₂;

Show that red limit wavelength equation is:

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

Described linear extrapolation II is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-866.41+1.28REP, (R ²=0.70).

Compared with prior art, the method and system described in the application, reaches following effect:

The first, high-spectrum remote sensing data vegetation information extraction method provided by the present invention, is solved traditional vegetation Physiology and biochemistry parameter and is obtained by Physical Chemistry Experiment, waste time and energy, have destructiveness to vegetation, be not suitable for the problem of carrying out in large area.

Second, high-spectrum remote sensing data vegetation information extraction method provided by the present invention, utilizes the estimation of high-spectrum remote sensing data realization to vegetation Physiology and biochemistry parameter, efficient and convenient, can not cause any impact to vegetation growth, its measured zone even can expand the yardstick of the whole earth to.

Accompanying drawing explanation

Accompanying drawing described herein is used to provide further understanding of the present application, and form a application's part, the schematic description and description of the application, for explaining the application, does not form the improper restriction to the application.In the accompanying drawings:

Fig. 1 is described a kind of method high-spectrum remote sensing data vegetation information extraction method flow diagram of the present invention;

Fig. 2 is the schematic diagram that the present invention adopts linear extrapolation determination Red edge position.

Embodiment

As employed some vocabulary to censure specific components in the middle of instructions and claim.Those skilled in the art should understand, and hardware manufacturer may call same assembly with different noun.This specification and claims are not used as with the difference of title the mode distinguishing assembly, but are used as the criterion of differentiation with assembly difference functionally." comprising " as mentioned in the middle of instructions and claim is in the whole text an open language, therefore should be construed to " comprise but be not limited to "." roughly " refer to that in receivable error range, those skilled in the art can solve the technical problem within the scope of certain error, reach described technique effect substantially.In addition, " couple " word and comprise directly any and indirectly electric property coupling means at this.Therefore, if describe a first device in literary composition to be coupled to one second device, then represent described first device and directly can be electrically coupled to described second device, or be indirectly electrically coupled to described second device by other devices or the means that couple.Instructions subsequent descriptions is implement the better embodiment of the application, and right described description is for the purpose of the rule that the application is described, and is not used to the scope limiting the application.The protection domain of the application is when being as the criterion depending on the claims person of defining.

Embodiment 1

Specific embodiment for high-spectrum remote sensing data vegetation information extraction method a kind of described in the application shown in Figure 1, described in the present embodiment, method comprises the following steps:

Step 101, in experimental study region, synchronous or standard synchronously obtains the normative reference value of target in hyperspectral remotely sensed image and vegetation Physiology and biochemistry parameter to be measured;

Step 102, dimensionality reduction pre-service is carried out to described high light remote sensing spectrum image, obtain Vegetation canopy spectral reflectance data;

Step 103, set up empirical mathematical model between Vegetation canopy spectral reflectance data and vegetation parameter standard value, obtain the high-spectrum remote-sensing evaluation method of vegetation Physiology and biochemistry parameter;

Step 104, by described empirical mathematical model, obtained the estimated value of vegetation parameter by the Vegetation canopy spectral reflectance data obtained from Hyperspectral imaging; Sample value is utilized to estimate the parameter in model, the precision of last testing model.

First the similarities and differences of the ultimate constituent of material and the difference of derivative spectomstry reflectance curve is the physical basis of carrying out terrain analysis and classification based on target in hyperspectral remotely sensed image, so need or standard synchronous in experimental study region synchronously to obtain the normative reference value of target in hyperspectral remotely sensed image and vegetation Physiology and biochemistry parameter to be measured.The wave band quantity of target in hyperspectral remotely sensed image is many, wave band interval is close, and observation data has redundancy to a certain degree, and in addition, due to the problem of remote sesing detector sole mass, the image of some wave band may have larger noise.Therefore, dimensionality reduction pre-service is carried out to Hyperspectral imaging, Vegetation canopy spectral reflectance data can be obtained.

In said method, carrying out dimensionality reduction Preprocessing Algorithm to high light remote sensing spectrum image is principal component analysis (PCA), and its basic ideas are: utilize Karhunen-Loeve transformation, and the variance of usage data carrys out descriptor amount, attempts to make the data after converting to successively decrease distribution according to quantity of information.It mainly adopts linear projection method raw data to be projected in new coordinate space, the quantity of information that first principal component comprises in new coordinate space is maximum, Second principal component, uncorrelated with first principal component data and in residual components quantity of information maximum, by that analogy.For the high spectrum image with dozens of wave band, general front four, five major components just can comprise the quantity of information of image more than 90%, and what information higher composition there is no.Therefore can replace view picture high spectrum image by front several major component, realize the object of dimensionality reduction.Concrete grammar is as follows:

Standardization is carried out to the data of former each wave band of Hyperspectral imaging X, obtains standardized images matrix X _c;

Normalized matrix X _ccovariance matrix Σ _c;

Ask matrix Σ _ceigenvectors matrix A _c, the rule that wherein proper vector is successively decreased according to eigenwert arranges from left to right;

Utilize the proper vector of trying to achieve to carry out linear transformation to view data and obtain PCA transformation results, computing formula is:

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

Matrix X _pcathe first principal component of the former high spectrum image of middle the first row data representation, the Second principal component, of the former high spectrum image of the second row data representation, by that analogy.

In said method, the empirical mathematical model set up between Vegetation canopy spectral reflectance data and vegetation parameter standard value is further:

(1) statistical regression model between spectral signature and vegetation parameter is set up:

y＝c ₁x+c ₂

(2) the ground measured data of sample point and high-spectrum remote sensing data is utilized to estimate the parameter in model, if observed reading is y _i, regressand value is least error sum-of-squares 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 $\frac{\&part;Q}{\&part;c_1}=0$ With $\frac{\&part;Q}{\&part;c_2}=0,$ Obtaining parameter estimation amount is:

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

(3) test to the precision of model: relation linear between variable x and variable y in linear regression model (LRM), adopt the related coefficient between variable x and variable y to check the conspicuousness of regression equation, expression formula is as follows:

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

Correlation coefficient r is the amount representing variable x and variable y linear relationship level of intimate, and its span is | r|≤1; As r > 0, variable x and variable y correlation; As r < 0, variable x and variable y is negative correlativing relation; And | r| larger expression linear dependence is stronger;

For general regression model, due to the change of stochastic error or independent variable x, the observed reading y of variable y _iincomplete 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, $S_T=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-\stackrel{\&OverBar;}{y})}^2$ Represent total variance quadratic sum, $S_e=\underset{i=1}{\overset{n}{\&Sigma;}}{(y_i-{\hat{y}}_i)}^2$ Represent residual sum of squares (RSS), represent that stochastic error is on the impact of regression accuracy, its standardized value is root-mean-square error RMSE, and computing formula is:

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

In formula, for regression sum of square, reflect the degree of scatter of the dependent variable y caused due to independent variable x; Due to S _t=S _e+ S _r, work as S _tafter given, S _rlarger then S _eless, the impact of variable x on variable y is more remarkable.And S _rless then S _elarger, variable x is more not remarkable on the impact of variable y.Therefore, coefficient of determination R is defined ², computing formula is:

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

Coefficient of determination R ²reflect the conspicuousness that variable x affects variable y in the regression model selected, its span is R ²≤ 1.R ²larger expression causes the change of variable y more remarkable by the change of variable x, otherwise represents that the change of variable x does not have anything to affect on variable y.

Typical vegetation Physiology and biochemistry parameter comprises leaf area index and chlorophyll content.Leaf area index is calculated, if R by normalization difference vegetation index _nirand R _redrepresent the value of spectral reflectivity near infrared and ruddiness place respectively, the computing formula of normalization difference vegetation index (NDVI) is:

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

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

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

Chlorophyll content described in the red limit feature calculation utilizing reflectance spectrum curve, the steps include:

Linear extrapolation can follow the tracks of the slope change near 700nm and 725nm that peak value occurs, alleviates the instability of vegetation physiological parameter because bimodal problem causes and Red edge position relation.Extrapolated two straight lines in the both sides of the method to the spectral reflectivity first order derivative curve near red limit, and the position on red limit is determined by the intersection point calculating two straight lines, as shown in Figure 2.

Article 1, straight line is in long-red-wave side, if its straight-line equation is: FDR=m ₁λ+c ₁.

Article 2 straight line is near infrared side, if its straight-line equation is: FDR=m ₂λ+c ₂.

Determine that the method for these two straight lines has two kinds.Two some calculated line equations that linear extrapolation I utilizes wavelength to be 680nm and 694nm in long-red-wave side, two the some calculated line equations utilizing wavelength to be 724nm and 760nm near infrared side.Two some calculated line equations that linear extrapolation II utilizes wavelength to be 680nm and 694nm in long-red-wave side equally, and two the some calculated line equations utilizing wavelength to be 732nm and 760nm near infrared side.

Article two, straight line has identical λ and FDR value in point of intersection, can obtain red limit wavelength equation to be thus:

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

Described linear extrapolation I is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-1111.01+1.63REP, (R ²=0.75).

Described linear extrapolation II is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-866.41+1.28REP, (R ²=0.70).

Embodiment 2

A kind of Application Example of the present invention is below provided.

High-spectrum remote sensing data vegetation information extraction method comprises:

(1) first high spectrum image X is read;

(2) principal component analysis dimensionality reduction is carried out to image: first normalized matrix X _cand covariance matrix Σ _c, ask matrix Σ _ceigenvectors matrix A _c, obtain analytic transformation result:

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

(3) vegetation spectral reflectivity is obtained by the image after dimensionality reduction;

(4) the linear statistical regression model between spectral signature and vegetation parameter is set up:

y＝c ₁x+c ₂

If observed reading is y _i, regressand value is least error sum-of-squares 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 $\frac{\&part;Q}{\&part;c_1}=0$ With $\frac{\&part;Q}{\&part;c_2}=0,$ Obtaining parameter estimation amount is:

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

(5) precision of model is tested:

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

Coefficient of determination R ²reflect the conspicuousness that in regression model, variable x affects variable y, its span is R ²≤ 1.R ²larger expression causes the change of variable y more remarkable by the change of variable x.

(6) typical vegetation Physiology and biochemistry parameter is calculated:

Wherein the computing method of leaf area index are:

If R _nirand R _redrepresent the value of spectral reflectivity near infrared and ruddiness place respectively, normalization difference vegetation index NDVI is:

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

For MODIS data, the empirical estimating model of leaf area index is:

LAI＝0.3775·exp(2.4293·NDVI)；

For ASTER data, the empirical estimating model of leaf area index is:

LAI＝0.3773·exp(2.4317·NDVI)。

Being calculated as of vegetation chlorophyll content:

Extrapolated two straight lines in both sides of the spectral reflectivity first order derivative curve near red limit, Article 1 straight line is in long-red-wave side, if its straight-line equation is: FDR=m ₁λ+c ₁; Article 2 straight line is near infrared side, if its straight-line equation is: FDR=m ₂λ+c ₂.

Red limit wavelength equation is:

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

Linear extrapolation I is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-1111.01+1.63REP, (R ²=0.75).

Linear extrapolation II is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-866.41+1.28REP, (R ²=0.70).

High-spectrum remote sensing data vegetation information extraction method provided by the present invention, first principal component analysis dimension-reduction treatment is carried out to Hyperspectral imaging, obtain Vegetation canopy spectral reflectance data, then the linear regression empirical mathematical model between Vegetation canopy spectral reflectance data and vegetation parameter standard value is set up, obtain the high-spectrum remote-sensing evaluation method of Physiology and biochemistry parameter, finally calculated leaf area index by normalization difference vegetation index, utilized the red limit feature of reflectance spectrum curve to estimate vegetation chlorophyll content.

Known by above each embodiment, the beneficial effect that the application exists is:

The first, high-spectrum remote sensing data vegetation information extraction method provided by the present invention, is solved traditional vegetation Physiology and biochemistry parameter and is obtained by Physical Chemistry Experiment, waste time and energy, have destructiveness to vegetation, be not suitable for the problem of carrying out in large area.

Second, high-spectrum remote sensing data vegetation information extraction method provided by the present invention, utilizes the estimation of high-spectrum remote sensing data realization to vegetation Physiology and biochemistry parameter, efficient and convenient, can not cause any impact to vegetation growth, its measured zone even can expand the yardstick of the whole earth to.

Those skilled in the art should understand, the embodiment of the application can be provided as method, device or computer program.Therefore, the application can adopt the form of complete hardware embodiment, completely software implementation or the embodiment in conjunction with software and hardware aspect.And the application can adopt in one or more form wherein including the upper computer program implemented of computer-usable storage medium (including but not limited to magnetic disk memory, CD-ROM, optical memory etc.) of computer usable program code.

Above-mentioned explanation illustrate and describes some preferred embodiments of the application, but as previously mentioned, be to be understood that the application is not limited to the form disclosed by this paper, should not regard the eliminating to other embodiments as, and can be used for other combinations various, amendment and environment, and can in invention contemplated scope described herein, changed by the technology of above-mentioned instruction or association area or knowledge.And the change that those skilled in the art carry out and change do not depart from the spirit and scope of the application, then all should in the protection domain of the application's claims.

Claims (7)

1. a high-spectrum remote sensing data vegetation information extraction method, is characterized in that, comprising:

In experimental study region, synchronous or standard synchronously obtains the normative reference value of target in hyperspectral remotely sensed image and vegetation Physiology and biochemistry parameter to be measured;

Dimensionality reduction pre-service is carried out to described high light remote sensing spectrum image, obtains Vegetation canopy spectral reflectance data;

Set up the empirical mathematical model between Vegetation canopy spectral reflectance data and vegetation parameter standard value, obtain the high-spectrum remote-sensing evaluation method of vegetation Physiology and biochemistry parameter;

By described empirical mathematical model, obtained the estimated value of vegetation parameter by the Vegetation canopy spectral reflectance data obtained from Hyperspectral imaging; Sample value is utilized to estimate the parameter in model, the precision of last testing model.

2. high-spectrum remote sensing data vegetation information extraction method according to claim 1, is characterized in that,

Carrying out dimensionality reduction pre-service to described high light remote sensing spectrum image is further:

Standardization is carried out to the data of former each wave band of Hyperspectral imaging X, obtains standardized images matrix X _c;

Normalized matrix X _ccovariance matrix Σ _c;

Ask matrix Σ _ceigenvectors matrix A _c, the rule that wherein proper vector is successively decreased according to eigenwert arranges from left to right;

Utilize the proper vector of trying to achieve to carry out linear transformation to view data and obtain PCA transformation results, computing formula is:

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

Matrix X _pcathe first principal component of the former high spectrum image of middle the first row data representation, the Second principal component, of the former high spectrum image of the second row data representation, by that analogy.

3. high-spectrum remote sensing data vegetation information extraction method according to claim 1, is characterized in that,

The described empirical mathematical model set up between Vegetation canopy spectral reflectance data and vegetation parameter standard value is further:

Set up the linear statistical regression model between spectral signature and vegetation parameter:

y＝c ₁x+c ₂

The ground measured data of sample point and high-spectrum remote sensing data is utilized to estimate the parameter in model, if observed reading is y _i, regressand value is least error sum-of-squares 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 $\frac{\&part;Q}{\&part;c_1}=0$ With $\frac{\&part;Q}{\&part;c_2}=0,$ Obtaining parameter estimation amount is:

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

Test to the precision of model: relation linear between variable x and variable y in linear regression model (LRM), adopt the related coefficient between variable x and variable y to check the conspicuousness of regression equation, expression formula is as follows:

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

Correlation coefficient r is the amount representing variable x and variable y linear relationship level of intimate, and its span is | r|≤1; As r > 0, variable x and variable y correlation; As r < 0, variable x and variable y is negative correlativing relation; And | r| larger expression linear dependence is stronger;

For general regression model, due to the change of stochastic error or independent variable x, the observed reading y of variable y _iincomplete 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, $S_T=\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-\stackrel{\&OverBar;}{y}\right)}^2$ Represent total variance quadratic sum, $S_e=\underset{i=1}{\overset{n}{\&Sigma;}}{\left(y_i-{\hat{y}}_i\right)}^2$ Represent residual sum of squares (RSS), represent that stochastic error is on the impact of regression accuracy, its standardized value is root-mean-square error RMSE, and computing formula is: $RMSE=\sqrt{\frac{\underset{i=1}{\overset{n}{\&Sigma;}}{\left({\hat{y}}_i-y_i\right)}^2}{n}}$

In formula, for regression sum of square, reflect the degree of scatter of the dependent variable y caused due to independent variable x; Definition coefficient of determination R ², computing formula is:

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

Coefficient of determination R ²span be R ²≤ 1.

4. high-spectrum remote sensing data vegetation information extraction method according to claim 1, is characterized in that,

Described vegetation Physiology and biochemistry parameter comprises leaf area index and chlorophyll content, and described leaf area index utilizes normalization difference vegetation index to calculate, and described chlorophyll content utilizes the red limit feature calculation of reflectance spectrum curve to obtain.

5. high-spectrum remote sensing data vegetation information extraction method according to claim 4, is characterized in that,

The method utilizing normalization difference vegetation index to calculate described leaf area index is:

Normalization difference vegetation index NDVI is obtained by following formulae discovery:

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

Wherein, R _nirand R _redrepresent the value of spectral reflectivity near infrared and ruddiness place respectively;

For MODIS data, the empirical estimating model of leaf area index is:

LAI＝0.3775·exp(2.4293·NDVI)；

For ASTER data, the empirical estimating model of leaf area index is:

LAI＝0.3773·exp(2.4317·NDVI)。

6. high-spectrum remote sensing data vegetation information extraction method according to claim 4, is characterized in that,

The method of chlorophyll content described in the red limit feature calculation of reflectance spectrum curve is utilized to be:

Adopt two some calculated line equations that linear extrapolation I utilizes wavelength to be 680nm and 694nm in long-red-wave side, the straight-line equation being set in long-red-wave side is: FDR=m ₁λ+c ₁; Two the some calculated line equations utilizing wavelength to be 724nm and 760nm near infrared side, the straight-line equation being set near infrared side is: FDR=m ₂λ+c ₂;

Show that red limit wavelength equation is:

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

Described linear extrapolation I is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-1111.01+1.63REP, (R ²=0.75).

7. high-spectrum remote sensing data vegetation information extraction method according to claim 4, is characterized in that,

The method of chlorophyll content described in the red limit feature calculation of reflectance spectrum curve is utilized to be:

Adopt two some calculated line equations that linear extrapolation II utilizes wavelength to be 680nm and 694nm in long-red-wave side, the straight-line equation being set in long-red-wave side is: FDR=m ₁λ+c ₁; Two the some calculated line equations utilizing wavelength to be 732nm and 760nm near infrared side, the straight-line equation being set near infrared side is: FDR=m ₂λ+c ₂;

Show that red limit wavelength equation is:

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

Described linear extrapolation II is adopted to calculate regression equation between the long REP of red side wave and chlorophyll content CC and precision is: CC=-866.41+1.28REP, (R ²=0.70).