CN103913422A - Rapid monitoring method for aquatic plants in shallow lake based on HJ-CCD images - Google Patents

Abstract

The invention provides a rapid monitoring method for aquatic plants in a shallow lake based on images of a resource environment moonlet No.1 (HJ-CCD). The method comprises the following steps: 1) acquiring HJ-CCD images of a researched region and preprocessing the images; 2) calculating sensitive characteristic indexes of emergent aquatic plants according to the HJ-CCD images of the researched region; 3) subjecting the HJ-CCD images of the researched region to principal component transformation so as to obtain sensitive characteristics of floating-leaved plants; 4) subjecting the HJ-CCD images of the researched region to tasseled cap transformation so as to obtain sensitive characteristic indexes of submerged plants; 5) determining and recognizing thresholds of different aquatic plants according to measured data synchronous to the images of the researched region and bringing forward discrimination conditions; and 6) establishing a decision tree classification model according to the discrimination conditions and generating a shallow lake aquatic plant monitoring result graph. The method can realize rapid, real-time, large-scale and high-precision recognition and monitoring of spatial distribution of aquatic plants of different kinds.

Description

A kind of shallow lake hydrophyte quick monitoring method based on HJ-CCD image

Technical field

The invention belongs to remote sensing application field, relate to the remote sensing quick monitoring method of a kind of shallow lake hydrophyte.

Background technology

Hydrophyte is the important ingredient of lake ecosystem, is also important attemperator and the indicator of lake ecosystem.Real-time, the fast monitored of hydrophyte have directive significance to the assessment of lake ecosystem ecological functions,, reparation and the salvaging of hydrophyte etc. of muddy water state lake ecosystem are had important practical significance meanwhile.

The monitoring method of hydrophyte is still continued to use artificial sample prescription, line-transect investigation method at present, it is wasted time and energy, monitoring range and sample size limited, monitoring difficulty is (the local people of a lot of aquatic plants growths is difficult to arrive) greatly, cannot quick, real-time, large-arealy obtain type and the distribution situation of hydrophyte; Therefore, a kind of shallow lake hydrophyte quick monitoring method based on environmental resource moonlet (HJ-CCD) image is proposed,, carry out shallow lake hydrophyte type and distribution fast, in real time, on a large scale with high-precision identification and monitor imperative.

Summary of the invention

In order to solve traditional hydrophyte monitoring method representativeness, poor in timeliness; Waste time and energy, monitoring range and sample size limited, the problem and shortage such as monitoring difficulty is large, of the present inventionly provide a kind of shallow lake hydrophyte quick monitoring method based on HJ-CCD image, described hydrophyte comprises emergent aquactic plant, floatingleaved plant and submerged plant, and it comprises the following steps:

Step 1, obtain study area HJ-CCD image, and carry out pre-service;

Step 2, according to described study area HJ-CCD image, calculate emergent aquactic plant sensitive features index E VSI;

Step 3, respectively described study area HJ-CCD image is carried out to principal component transform, obtain floatingleaved plant sensitive features index FVSI;

Step 4, described study area HJ-CCD image is carried out to red-tasselled official hat conversion, obtain submerged plant sensitive features index SVSI;

Step 5, according to the measured data of described study area HJ-CCD image synchronous, determine the threshold value of EVSI, FVSI and SVSI, and obtain criterion;

Step 6, set up Decision-Tree Classifier Model according to criterion, generate shallow lake hydrophyte monitoring result, and utilize measured data to verify.

Wherein, the image pre-service in described step 1 comprises radiant correction, geometry correction and study area cutting, and wherein said radiant correction refers to radiation calibration and atmospheric correction.

Wherein, the EVSI computing formula in described step 2 is as follows:

$\mathrm{EVSI}=\frac{R_{\mathrm{nir}}-(R_{\mathrm{blue}}+R_{\mathrm{green}}+R_{\mathrm{red}})}{R_{\mathrm{nir}}+(R_{\mathrm{blue}}+R_{\mathrm{green}}+R_{\mathrm{red}})}$

Wherein, R _bluefor blue wave band, its wavelength coverage is 0.45-0.52 micron; R _greenfor green wave band, its wavelength coverage is 0.52-0.60 micron; R _redfor red wave band, its wavelength coverage is 0.63-0.69 micron; R _nirfor near-infrared band, its wavelength coverage is 0.76-0.90 micron.

Wherein, the FVSI computing formula in described step 3 is as follows:

FVSI=PC ₂，

Wherein, PC ₂for the Second principal component, after the HJ-CCD image principal component transform of study area.

Wherein, described PC ₂obtaining step is as follows: utilize the study area HJ-CCD image obtaining, carry out principal component transform, the step of described principal component transform is as follows:

Step (a), represent the raw data of multispectral image with the form of matrix, establish the study area HJ-CCD image data matrix obtaining and be X, as follows:

$X=\left[\begin{array}{cccc}{x}_{11}& x_{12}& \·\·\·& x_{1n}\\ x_{21}& x_{22}& \·\·\·& x_{2n}\\ \·& \·& & \·\\ \·& \·& & \·\\ \·& \·& & \·\\ x_{p1}& x_{p2}& \·\·\·& x_{\mathrm{pn}}\end{array}\right]={\left[x_{\mathrm{ik}}\right]}_{p\×n}$

Wherein, n is the wave band number of HJ-CCD image, n=1,2,3,4; Pixel number in the each wave band image of p HJ-CCD image, in matrix, each row vector represents the image of a wave band;

Step (b), based on matrix X, by Y=TX, image is converted the image data matrix Y obtaining after principal component transform, T is transformation matrix, solves transformation matrix T;

Described transformation matrix T is X space covariance ∑ x _ithe transposed matrix of eigenvectors matrix, its solution procedure is as follows:

1), according to raw image data matrix X, the covariance matrix S that obtains X is:

$S=\left[\begin{array}{cccc}{s}_{11}& s_{12}& \·\·\·& s_{1p}\\ s_{21}& s_{22}& \·\·\·& x_{2p}\\ \·& \·& & \·\\ \·& \·& & \·\\ \·& \·& & \·\\ s_{p1}& s_{p2}& \·\·\·& s_{\mathrm{pp}}\end{array}\right]={\left[s_{\mathrm{ij}}\right]}_{p\×p}$

Described S _ijexpression formula is:

$S_{\mathrm{ij}}=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}(x_{\mathrm{ik}}-\stackrel{\‾}{x_i})(x_{\mathrm{ij}}-\stackrel{\‾}{x_j})\stackrel{\‾}{x_i}=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\Σ}}}{x}_{\mathrm{ik}},$ Wherein, k=1,2,3 ... .n

Described is the average of the each wave band of image, and expression formula is as follows:

wherein, i=1,2 ..., n; J=1,2 ..., p

2) eigenvalue λ of covariance S _iwith proper vector U _j, and form transformation matrix T,, solve secular equation (λ _i-S) U=0; Then by eigenvalue λ _iascending arrangement, obtains the unit character vector U of character pair value _j:

U _j=[u _1j?u _2j?...?u _pj] ^T，

Form matrix take each proper vector as row, that is:

$U=\left[\begin{array}{cccc}{U}_1& U_2& ...& U_p\end{array}\right]={\left[u_{\mathrm{ij}}\right]}_{p\×p};$

3) U transpose of a matrix matrix, i.e. U ^tfor required transformation matrix T;

Step (c), bring transformation matrix T into Y=TX, solve the matrix Y after principal component transform:

$Y=\left[\begin{array}{cccc}{u}_{11}& u_{21}& \·\·\·& u_{p1}\\ u_{12}& u_{22}& \·\·\·& u_{p2}\\ \·& \·& & \·\\ \·& \·& & \·\\ \·& \·& & \·\\ u_{1p}& u_{2p}& \·\·\·& u_{\mathrm{pp}}\end{array}\right]X=U^{T}X,$

Step (d), solve PC ₂,

Make Y ₂=PC ₂, the row vector Y of Y matrix ₂=[y _j1y _j2... y _jp] be j major component; Y ₁, Y ₂.Y _pbe the new variables of study area HJ-CCD image after principal component transform, be called as first principal component with this, Second principal component, p major component, reverts to two dimensional image by new variables, obtains p major component image, therefore,

Y ₂=[y ₂₁?y ₂₂?...?y _2p]

By Y ₂revert to two dimensional image, obtain PC ₂image.

Wherein, the SVSI computing formula of described step 4 is as follows:

SVSI=BI-GVI，

Wherein, BI is brightness index, and it is the first component after red-tasselled official hat conversion; GVI is green degree index, and it is the second component after red-tasselled official hat conversion.

Wherein, brightness index BI and green degree index GVI ask calculation step as follows:

Carry out red-tasselled official hat conversion by following formula:

y=Cx+α，

α, for representing constant offset, is the negative value for avoiding occurring in conversion process; X is the pixel vector of the HJ-CCD multispectral image before conversion; Y is the pixel vector of the multispectral image after HJ-CCD conversion; C is transformation matrix, as follows:

$C=\left[\begin{array}{cccc}0.433& 0.632& 0.586& 0.264\\ -0.290& -0.562& -0.600& 0.491\\ -0.829& 0.522& -0.039& 0.194\\ 0.223& 0.012& 0.543& 0.810\end{array}\right]$

Therefore, can obtain:

BI=0.433R _b+0.632R _g+0.586R _r+0.264R _Nir；

GVI=-0.290R _b+(-0.562R _g)+(-0.600R _r)+0.491R _Nir。

Wherein, described step 6 comprises the following steps:

Step 6.1, in the time that study area pixel meets EVSI>a and D<500m, be identified as emergent aquactic plant;

Step 6.2, in the pixel that does not meet step 6.1, differentiate, in the time of FVSI<b, be identified as floatingleaved plant;

Step 6.3, in the pixel that does not meet step 6.1 and 6.2 conditionals, differentiate, in the time of SVSI<c, be identified as submerged plant;

The pixel of step 6.4, the discontented condition that is enough to three steps is identified as water body;

Wherein, the distance that D is offshore; A is the threshold value of EVSI identification emergent aquactic plant, and b is the threshold value of FVSI identification floatingleaved plant, and c is the threshold value of SVSI identification submerged plant; A, b and c are by obtaining with the measured data training of image synchronous.

Wherein, hydrophyte monitoring accuracy utilizes overall accuracy and Kappa coefficient to evaluate, and its computing formula is as follows:

Overall classification accuracy $\left(\mathrm{OA}\right)=\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_{\mathrm{kk}}/\mathrm{\&beta;},$

Wherein, for the total sample number of correctly being classified on image with actual sample, β is total sample number;

$\mathrm{kappa}=\frac{N\underset{i}{\overset{r}{\mathrm{\&Sigma;}}}{A}_{\mathrm{ii}}-\mathrm{\&Sigma;}(A_{i+}+...A_{+i})}{N^2-\mathrm{\&Sigma;}(A_{i+}+...A_{+i})}$

Wherein, r is total columns, i.e. total classification number in Error Matrix; A _iithat i in Error Matrix is capable, j lists pixel quantity, i.e. the number of correct classification; A _i+and A _{+ i}respectively total pixel quantity of i and i row; N is total pixel quantity for accuracy evaluation.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of technical solution of the present invention;

Fig. 2 is the Hongchehu Lake hydrophyte monitoring result on May 13rd, 2013 in embodiment of the method for the present invention;

Fig. 3 is the aquatic plant monitoring result in the Taihu Lake on July 11st, 2013 in embodiment of the method for the present invention;

Fig. 4 is the aquatic plant monitoring result in the Taihu Lake on September 26th, 2013 in embodiment of the method for the present invention.

Concrete embodiment

Below in conjunction with drawings and Examples, the present invention is further described.A shallow lake hydrophyte quick monitoring method based on HJ-CCD image, take the large-scale shallow water lake (mean depth is as 1.9 meters) of Chinese Typical Representative---instance analysis is carried out as example in Taihu Lake and Hongchehu Lake.Wherein, Taihu Lake longitude and latitude scope is 119.55-120.34E, 30.55-31.32N; ; Hongchehu Lake longitude and latitude scope is 118.18-118.52E, north latitude 33.22～33.37N.Show through investigation and data, east, Taihu Lake and Hongchehu Lake are distributed with large-area hydrophyte, are mainly following three major types: emergent aquactic plant, floatingleaved plant and submerged plant.At present, because mankind's activity is in recent years violent, lake aquatic plant and Wetland Environment cause serious interference and destruction, utilize hydrophyte distributed areas and the area of remote sensing technology to Taihu Lake and Hongchehu Lake to carry out monitoring real-time, ecological restoration of lakes and aquatic plant salvaging are had to important directive significance.Take Taihu Lake and Hongchehu Lake as study area, carry out instance analysis below:

Step 1, according to actual measurement sampling point acquisition time, from the data product inquiry of China Resource Satellite Applied Center order website ( http: // 218.247.138.121/DSSPlatform/index.html) order the environmental resource moonlet image data (HJ-CCD) of downloading with the 11 days July in 2013 in actual measurement sampling point covering Taihu Lake of synchronizeing and 13 days Mays in 2013 in covering Hongchehu Lake.

Step 2, utilize ENVI software respectively two scape images to be carried out to atmospheric correction (radiation calibration and atmospheric correction) and geometry correction, then utilize Taihu Lake and vector border, Hongchehu Lake to cut out corresponding study area image from image;

Step 3, obtain the EVSI image of two scape images according to the computing formula of EVSI;

Step 4, the study area image on May 13rd, 2013, on July 11st, 2013 and on September 26th, 2013 is carried out respectively to principal component transform, and according to formula F VSI=PC ₂, obtain respectively its FVSI image;

Step 5, on May 13rd, 2013, on July 11st, 2013 and on September 26th, 2013 study area image carry out respectively red-tasselled official hat conversion, obtain brightness index BI and the green degree index GVI of study area, and according to formula S VSI=BI-GVI, obtain respectively its SVSI image;

Step 6, utilize EVSI, FVSI and SVSI, build the discriminant of the different hydrophyte of identification, as follows:

Condition 1, in the time that study area pixel meets EVSI>a and D<500m, be identified as emergent aquactic plant;

Condition 2, in 1 the pixel of not satisfying condition, in the time of FVSI<b, be identified as floatingleaved plant;

Condition 3, not meeting in 1 and 2 pixel, in the time of SVSI<c, be identified as submerged plant;

Condition 4, the discontented pixel that is enough to a condition in step are identified as water body.

Wherein, the distance that D is offshore; A is the threshold value of EVSI identification emergent aquactic plant, and b is the threshold value of FVSI identification floatingleaved plant, and c is the threshold value of SVSI identification floatingleaved plant; A, b and c are by obtaining with the measured data training of image synchronous; Table 1 is Hongchehu Lake and each characteristic threshold value of on July 11st, 2013 and the aquatic plant monitoring in Taihu Lake on September 26 in 2013 on May 13rd, 2013.

The each characteristic threshold value in table 1 Hongchehu Lake and Taihu Lake

Step 7, in ArcGIS, classification results is published picture, as Fig. 3, and according to actual measurement sampling point and the classification results in July 11 and August 16, calculate classification overall accuracy and Kappa coefficient, classification of assessment result, as table 2.

Table 2 is precision evaluation tables of the aquatic plant monitoring result in Taihu Lake obtained of the method described in the embodiment of the present invention

Table 2 is precision evaluation tables of the aquatic plant monitoring result in Taihu Lake obtained of the method described in the embodiment of the present invention, contrast by monitoring result and measured result is known, the inventive method to the monitoring accuracy of shallow lake (Taihu Lake and Hongchehu Lake) hydrophyte all higher than 80%, Kappa coefficient is greater than 0.8, meets major applications demand.In addition, Fig. 2 is the Hongchehu Lake hydrophyte monitoring result schematic diagram that the method described in the embodiment of the present invention is obtained; Fig. 3 and Fig. 4 are the aquatic plant monitoring result schematic diagrams in Taihu Lake that the method described in the embodiment of the present invention is obtained.Can find out, the advantage that the present invention compares existing classic method is: utilize limited actual measurement sampling point and remote sensing image, can obtain fast and accurately distribution pattern and the distributed areas of study area hydrophyte.

Although illustrated and described embodiments of the invention, those having ordinary skill in the art will appreciate that: in the situation that not departing from principle of the present invention and aim, can carry out multiple variation, modification, replacement and modification to these embodiment, scope of the present invention is limited by claim and equivalent thereof.

Claims (9)

1. the shallow lake hydrophyte quick monitoring method based on HJ-CCD image, described hydrophyte comprises emergent aquactic plant, floatingleaved plant and submerged plant, it comprises the following steps:

Step 1, obtain study area HJ-CCD image, and carry out pre-service;

Step 2, according to described study area HJ-CCD image, calculate emergent aquactic plant sensitive features index E VSI;

Step 3, respectively described study area HJ-CCD image is carried out to principal component transform, obtain floatingleaved plant sensitive features index FVSI;

Step 4, described study area HJ-CCD image is carried out to red-tasselled official hat conversion, obtain submerged plant sensitive features index SVSI;

Step 5, according to the measured data of described study area HJ-CCD image synchronous, determine the threshold value of EVSI, FVSI and SVSI, and obtain criterion;

Step 6, set up Decision-Tree Classifier Model according to criterion, generate shallow lake hydrophyte monitoring result, and utilize measured data to verify.

2. method according to claim 1, the image pre-service in described step 1 comprises radiant correction, geometry correction and study area cutting, wherein said radiant correction refers to radiation calibration and atmospheric correction.

3. method according to claim 1, the EVSI computing formula in described step 2 is as follows:

$\mathrm{EVSI}=\frac{R_{\mathrm{nir}}-(R_{\mathrm{blue}}+R_{\mathrm{green}}+R_{\mathrm{red}})}{R_{\mathrm{nir}}+(R_{\mathrm{blue}}+R_{\mathrm{green}}+R_{\mathrm{red}})}$

Wherein, R _bluefor blue wave band, its wavelength coverage is 0.45-0.52 micron; R _greenfor green wave band, its wavelength coverage is 0.52-0.60 micron; R _redfor red wave band, its wavelength coverage is 0.63-0.69 micron; R _nirfor near-infrared band, its wavelength coverage is 0.76-0.90 micron.

4. method according to claim 1, the FVSI computing formula in described step 3 is as follows:

FVSI=PC ₂，

Wherein, PC ₂for the Second principal component, after the HJ-CCD image principal component transform of study area.

5. method according to claim 4, described PC ₂obtaining step is as follows: utilize the study area HJ-CCD image obtaining, carry out principal component transform, the step of described principal component transform is as follows:

Step (a), represent the raw data of multispectral image with the form of matrix, establish the study area HJ-CCD image data matrix obtaining and be X, as follows:

$X=\left[\begin{array}{cccc}{x}_{11}& x_{12}& \&CenterDot;\&CenterDot;\&CenterDot;& x_{1n}\\ x_{21}& x_{22}& \&CenterDot;\&CenterDot;\&CenterDot;& x_{2n}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ x_{p1}& x_{p2}& \&CenterDot;\&CenterDot;\&CenterDot;& x_{\mathrm{pn}}\end{array}\right]={\left[x_{\mathrm{ik}}\right]}_{p\&times;n}$

Wherein, n is the wave band number of HJ-CCD image, n=1,2,3,4; Pixel number in the each wave band image of p HJ-CCD image, in matrix, each row vector represents the image of a wave band;

Step (b), based on matrix X, by Y=TX, image is converted the image data matrix Y obtaining after principal component transform, T is transformation matrix, solves transformation matrix T;

Described transformation matrix T is X space covariance ∑ x _ithe transposed matrix of eigenvectors matrix, its solution procedure is as follows:

1), according to raw image data matrix X, the covariance matrix S that obtains X is:

$S=\left[\begin{array}{cccc}{s}_{11}& s_{12}& \&CenterDot;\&CenterDot;\&CenterDot;& s_{1p}\\ s_{21}& s_{22}& \&CenterDot;\&CenterDot;\&CenterDot;& x_{2p}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ s_{p1}& s_{p2}& \&CenterDot;\&CenterDot;\&CenterDot;& s_{\mathrm{pp}}\end{array}\right]={\left[s_{\mathrm{ij}}\right]}_{p\&times;p}$

Described S _ijexpression formula is:

$S_{\mathrm{ij}}=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}(x_{\mathrm{ik}}-\stackrel{\&OverBar;}{x_i})(x_{\mathrm{ij}}-\stackrel{\&OverBar;}{x_j})\stackrel{\&OverBar;}{x_i}=\frac{1}{n}\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{x}_{\mathrm{ik}},$ Wherein, k=1,2,3 ... .n

Described is the average of the each wave band of image, and expression formula is as follows:

wherein, i=1,2 ..., n; J=1,2 ..., p

2) eigenvalue λ of covariance S _iwith proper vector U _j, and form transformation matrix T,, solve secular equation (λ _i-S) U=0; Then by eigenvalue λ _iascending arrangement, obtains the unit character vector U of character pair value _j:

U _j=[u _1j?u _2j?...?u _pj] ^T，

Form matrix take each proper vector as row, that is:

$U=\left[\begin{array}{cccc}{U}_1& U_2& ...& U_p\end{array}\right]={\left[u_{\mathrm{ij}}\right]}_{p\&times;p};$

3) U transpose of a matrix matrix, i.e. U ^tfor required transformation matrix T;

Step (c), bring transformation matrix T into Y=TX, solve the matrix Y after principal component transform:

$Y=\left[\begin{array}{cccc}{u}_{11}& u_{21}& \&CenterDot;\&CenterDot;\&CenterDot;& u_{p1}\\ u_{12}& u_{22}& \&CenterDot;\&CenterDot;\&CenterDot;& u_{p2}\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ \&CenterDot;& \&CenterDot;& & \&CenterDot;\\ u_{1p}& u_{2p}& \&CenterDot;\&CenterDot;\&CenterDot;& u_{\mathrm{pp}}\end{array}\right]X=U^{T}X,$

Step (d), solve PC ₂

Make Y ₂=PC ₂, the row vector Y of Y matrix ₂=[y _j1y _j2... y _jp] be j major component; Y ₁, Y ₂.Y _pbe the new variables of study area HJ-CCD image after principal component transform, be called as first principal component with this, Second principal component, p major component, reverts to two dimensional image by new variables, obtains p major component image, therefore,

Y ₂=[y ₂₁?y ₂₂?...?y _2p]

By Y ₂revert to two dimensional image, obtain PC ₂image.

6. method according to claim 1, is characterized in that, the SVSI computing formula of described step 4 is as follows:

SVSI=BI-GVI，

Wherein, BI is brightness index, and it is the first component after red-tasselled official hat conversion; GVI is green degree index, and it is the second component after red-tasselled official hat conversion.

7. method according to claim 6, is characterized in that, brightness index BI and green degree index GVI ask calculation step as follows:

Carry out red-tasselled official hat conversion by following formula:

y=Cx+α，

α, for representing constant offset, is the negative value for avoiding occurring in conversion process; X is the pixel vector of the HJ-CCD multispectral image before conversion; Y is the pixel vector of the multispectral image after HJ-CCD conversion; C is transformation matrix, as follows:

$C=\left[\begin{array}{cccc}0.433& 0.632& 0.586& 0.264\\ -0.290& -0.562& -0.600& 0.491\\ -0.829& 0.522& -0.039& 0.194\\ 0.223& 0.012& 0.543& 0.810\end{array}\right]$

Therefore, can obtain:

BI=0.433R _b+0.632R _g+0.586R _r+0.264R _Nir；

GVI=-0.290R _b+(-0.562R _g)+(-0.600R _r)+0.491R _Nir。

8. method according to claim 1, described step 6 comprises the following steps:

Step 6.1, in the time that study area pixel meets EVSI>a and D<500m, be identified as emergent aquactic plant;

Step 6.2, in the pixel that does not meet step 6.1, differentiate, in the time of FVSI<b, be identified as floatingleaved plant;

Step 6.3, in the pixel that does not meet step 6.1 and 6.2 conditionals, differentiate, in the time of SVSI<c, be identified as submerged plant;

The pixel of step 6.4, the discontented condition that is enough to three steps is identified as water body;

Wherein, the distance that D is offshore; A is the threshold value of EVSI identification emergent aquactic plant, and b is the threshold value of FVSI identification floatingleaved plant, and c is the threshold value of SVSI identification submerged plant; A, b and c are by obtaining with the measured data training of image synchronous.

9. method according to claim 1, hydrophyte monitoring accuracy utilizes overall accuracy and Kappa coefficient to evaluate, and its computing formula is as follows:

Overall classification accuracy $\left(\mathrm{OA}\right)=\underset{k=1}{\overset{n}{\mathrm{\&Sigma;}}}{\mathrm{\&beta;}}_{\mathrm{kk}}/\mathrm{\&beta;},$

Wherein, for the total sample number of correctly being classified on image with actual sample, β is total sample number;

$\mathrm{kappa}=\frac{N\underset{i}{\overset{r}{\mathrm{\&Sigma;}}}{A}_{\mathrm{ii}}-\mathrm{\&Sigma;}(A_{i+}+...A_{+i})}{N^2-\mathrm{\&Sigma;}(A_{i+}+...A_{+i})}$

Wherein, r is total columns, i.e. total classification number in Error Matrix; A _iithat i in Error Matrix is capable, j lists pixel quantity, i.e. the number of correct classification; A _i+and A _{+ i}respectively total pixel quantity of i and i row; N is total pixel quantity for accuracy evaluation.