CN101893550A - Semi-analytical method for realizing inversion of water body chlorophyll alpha concentration - Google Patents

Abstract

The invention relates to a semi-analytical method for realizing inversion of water body chlorophyll alpha concentration, which comprises the steps of: (1) reading the data of chlorophyll alpha, water body spectroscopic data and high-spectrum remote sensing data; (2) utilizing the optical characteristics of the water body containing the chlorophyll alpha to establish semi-analytical model of the chlorophyll alpha concentration; (3) on the basis of the data in the step (1), adopting a method of enumerating by band and linear iteration, and calculating the best wave band and model parameter of the semi-analytical algorithm; and (4) on the basis of the steps (1) and (3), extracting the space distribution information of the chlorophyll alpha concentration from high-spectrum remote sensing image.

Description

A kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration

Technical field

The present invention relates to a kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration, belong to the theoretical and application technical research field of water colour remote sensing.

Background technology

Water is Source of life, is the indispensable basic substance of global biosphere.Water quality condition is directly connected to the survival and development of human daily life and physical environment.Yet along with the development of economy and society, water pollution problems is serious day by day, and water quality monitoring and management role are arduous day by day.The core content of water quality monitoring is monitoring and assay water resource quality and change with time thereof, for national and governments at all levels' rational exploitation and utilization, management and protection water resource provide scientific basis.Therefore, obtain water quality spatial distribution state and variation tendency thereof in real time and effectively, have crucial meaning for improving water resource quality and water environment situation.

Traditional water quality space distribution information is by on-the-spot water sampling, and water analysis obtains in the laboratory.This water quality information has only by arranging that in the waters a large amount of measuring points obtains.Utmost point expense human and material resources and the work of time are not only in the water quality monitoring in big zone, and are difficult to satisfy the demand of real-time water quality condition tracking observation.Remote sensing technology is for addressing the above problem the approach that provides new.Water colour remote sensing is mainly used in the untouchable detection of (mainly comprising chlorophyll a, no life suspension and yellow substance etc.) of water quality factor concentration.The ultimate principle of water colour remote sensing is to utilize the spectral signature of water quality factor, set up the relational model (empirical model, half point are analysed model and analytical model) between water concentration and the spectral signature parameter, and utilize this model from remote sensing image, to extract the water concentration space distribution information.The chlorophyll-a concentration remote-sensing inversion is the important research content of water colour remote sensing.Generally, the bandwidth of the feature Wave crest and wave trough of chlorophyll a is narrower, some have the information of indicative significance in order to obtain this, and the waveband width that needs to select is preferably less than 5nm, and this certainly will be unfavorable for the application of broadband satellite data aspect the dynamic monitoring of water body chlorophyll alpha concentration space distribution.

High-spectrum remote-sensing has characteristics such as spectral resolution height, spectral information be abundant.These characteristics play crucial effects for improving chlorophyll-a concentration inverting level.In research in the past, mainly lay particular emphasis on the design of " narrow wave band " inverse model based on the chlorophyll-a concentration inversion theory of high spectrum, and " narrow wave band " model is relative less with the applied research that target in hyperspectral remotely sensed image combines.

Summary of the invention

The object of the present invention is to provide a kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration, be in view of the not high present Research of the precision of remote-sensing inversion chlorophyll-a concentration, from the water body chlorophyll alpha concentration optical characteristics, make up a kind of semi-analytical method of high-precision inversion of water body chlorophyll alpha concentration, and then helped improving the chlorophyll-a concentration remote sensing inversion accuracy.

A kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration of the present invention, it specifically comprises the steps:

(1) reads in chlorophyll-a concentration data, water body spectroscopic data and high-spectrum remote sensing data.

(2) utilize the optical property that contains the chlorophyll a water body, make up the chlorophyll-a concentration half point and analyse model.

In the visible light wave range scope, suppose to exist two groups of wave band (λ ₁And λ ₂, λ ₃And λ ₄), the absorption coefficient (a of solvable organic matter _CDOM) and the absorption coefficient (a of suspension _Tripton) there is a following approximation relation:

$\frac{a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_1\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_1\right)}{b\left({\mathrm{\λ}}_1\right)}\≈k_1\frac{a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_2\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_2\right)}{b\left({\mathrm{\λ}}_2\right)}$

$\frac{a_{\mathrm{chla}}\left({\mathrm{\λ}}_3\right)+a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_3\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_3\right)}{b\left({\mathrm{\λ}}_3\right)}\≈k_2\frac{a_{\mathrm{chla}}\left({\mathrm{\λ}}_4\right)+a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_4\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_4\right)}{b\left({\mathrm{\λ}}_4\right)}$

$\frac{[R^{-1}\left({\mathrm{\λ}}_1\right)-k_1{R}^{-1}\left({\mathrm{\λ}}_2\right)]}{[R^{-1}\left({\mathrm{\λ}}_3\right)-k_2{R}^1\left({\mathrm{\λ}}_4\right)]}=\frac{a_{\mathrm{chla}}\left({\mathrm{\λ}}_1\right)-k_1{a}_{\mathrm{chla}}\left({\mathrm{\λ}}_2\right)+a_w\left({\mathrm{\λ}}_1\right)-k_1{a}_w\left({\mathrm{\λ}}_2\right)}{a_w\left({\mathrm{\λ}}_4\right)-k_2{a}_w\left({\mathrm{\λ}}_3\right)}$

$P_{\mathrm{Chla}}=\frac{R^{-1}\left({\mathrm{\λ}}_1\right)-k_1{R}^{-1}\left({\mathrm{\λ}}_2\right)}{R^{-1}\left({\mathrm{\λ}}_3\right)-k_2{R}^{-1}\left({\mathrm{\λ}}_4\right)}$

Chla＝a+bP _Chla

In the formula, λ ₁, λ ₂, λ ₃And λ ₄Be four different wavelength; k ₁And k ₂Model parameter; R is a reflectivity; Chla is a chlorophyll-a concentration; PChla is that half point is analysed algorithm remote sensing parameter; A and b are the empirical relationship parameter between remote sensing parameter and the chlorophyll-a concentration.

(3) on the data basis of step (1), adopt by wave band and enumerate method with linear iteration, calculate half point and analyse the best band and the model parameter of algorithm.

The present invention has adopted half point to analyse the predicted value (C of model _{Pred, chla, i}) and measured value (C _{Mea, chla, i}) between the standard deviation (STE) of deviation and relative error (RE) thereof as the standard of model quality.Standard deviation involved in the present invention and relative error expression formula are as follows:

$\mathrm{STE}=\sqrt{\frac{\mathrm{\Σ}{(C_{\mathrm{pred},\mathrm{chla},i}-C_{\mathrm{mea},\mathrm{chla},i})}^2}{N}}$

$\mathrm{RE}=\frac{\mathrm{STE}}{\mathrm{MEAN}\left(C_{\mathrm{mea},\mathrm{chla}}\right)}\×100\%$

Linear iteration algorithm involved in the present invention is as follows: known experimental data (x _i, y _i), i=1,2 ... m, contain nonlinear parameter b among the inverting function y=f (x) ₁, b ₂... b _n, be designated as y=f (b ₁, b ₂... b _n).Make vectorial b=(b ₁, b ₂... b _n), then y=f (x, b) its residual sum of squares (RSS) is:

$S\left(b\right)=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{[y_i-f(x_i,b)]}^2$

(x b) exists with y=f

The Taylor of place launches, and omits high-order term:

$f(x_i,b)=f(x_i,b_1^0,b_2^0,\·\·\·,b_n^0)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{{\∂f}_i}{{\∂b}_j}(b_j-b_j^0)$

Comprehensive above-mentioned two formulas, it is as follows to get iterative formula in conjunction with principle of least square method:

$A=\left[\begin{array}{cccc}\frac{{\∂f}_1}{\∂b_1}& \frac{{\∂f}_1}{{\∂b}_2}& \·\·\·& \frac{{\∂f}_1}{{\∂b}_n}\\ \frac{{\∂f}_2}{{\∂b}_1}& \frac{{\∂f}_2}{{\∂b}_2}& \·\·\·& \frac{{\∂f}_2}{{\∂b}_n}\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \frac{{\∂f}_m}{{\∂b}_1}& \frac{{\∂f}_m}{{\∂b}_2}& \·\·\·& \frac{{\∂f}_m}{{\∂b}_n}\end{array}\right]$

$A^{T}A(b_j-b_j^0)=A^T(y-f)$

(4) on the basis of step (1) and (3), from target in hyperspectral remotely sensed image, extract the chlorophyll-a concentration space distribution information.

A kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration of the present invention, its advantage and effect are: the inventive method can be extracted the space distribution information of chlorophyll-a concentration with degree of precision from target in hyperspectral remotely sensed image.

Description of drawings

Fig. 1 is that the experiment centre space of points of the present invention distributes.

Fig. 2 for of the present invention based on October 28th, 2003 the spectrum data unified half point analyse inverse model.

Fig. 3 for of the present invention based on August 19th, 2004 the spectrum data unified half point analyse inverse model.

Embodiment

For a kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration that the present invention relates to better is described, this research to be chlorophyll-a concentration data, spectral reflectance data and on August 19th, 2004 that on October 27th, 2003,28 days and water quality experiment on August 19th, 2004 are measured testing synchronous Hyperion Hyperspectral imaging data with water quality, the half point of this research analysed algorithm analyze and verify.

The specific implementation step is as follows:

(1) reads in chlorophyll-a concentration data, spectral reflectance data and on the August 19th, 2004 and the synchronous Hyperion Hyperspectral imaging data of water quality experiment that on October 27th, 2003,28 days and on August 19th, 2004, the water quality experiment was measured, wherein, on August 19th, 2004 water quality to test the pairing curve of spectrum be directly from extracting through the Hyperion image of atmospheric correction.The experiment position as shown in Figure 1.

(2) utilize the optical property that contains the chlorophyll a water body, make up the chlorophyll-a concentration half point and analyse model.

$\mathrm{Chla}=a\frac{R^{-1}\left({\mathrm{\λ}}_1\right)-k_1{R}^{-1}\left({\mathrm{\λ}}_2\right)}{R^{-1}\left({\mathrm{\λ}}_3\right)-k_2{R}^{-1}\left({\mathrm{\λ}}_4\right)}+b$

In the formula, λ _iIt is the wavelength of i wave band; R is a reflectivity; Chla is a chlorophyll-a concentration; A, b, K ₁And K ₂For model parameter gets by linear iteration calculating.

(3) on the data basis of step (1), adopt by wave band and enumerate method with linear iteration, calculate the best band λ that half point is analysed algorithm _i, λ ₂, λ ₃, λ ₄And and model parameter K ₁And K ₂(the related computing formula of this step is seen step 3 part of summary of the invention).Analysis result and model related coefficient, error and relative error are as shown in Figures 2 and 3.

(4) on the basis of step (1) and (3), adopt the remote sensing parameter shown in the step (2), utilize on August 19th, 2004 experimental data to make up unified half point as shown in Figure 3 and analyse algorithm, and on this basis, in conjunction with synchronous Hyperion image data, utilize above-mentioned half point to analyse model and from target in hyperspectral remotely sensed image, extract the chlorophyll-a concentration space distribution information.

Claims (4)

1. semi-analytical method of realizing inversion of water body chlorophyll alpha concentration, its step is as follows:

(1) reads in chlorophyll-a concentration data, water body spectroscopic data and high-spectrum remote sensing data;

(2) utilize the optical property that contains the chlorophyll a water body, make up the chlorophyll-a concentration half point and analyse model;

(3) on the data basis of step (1), adopt by wave band and enumerate method with linear iteration, calculate half point and analyse the best band and the model parameter of algorithm;

(4) on the basis of step (1) and (3), from target in hyperspectral remotely sensed image, extract the chlorophyll-a concentration space distribution information.

2. a kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration according to claim 1 is characterized in that: it is as follows that the chlorophyll-a concentration half point described in the step (2) is analysed the model rule:

In the visible light wave range scope, suppose to exist two groups of wave band (λ ₁And λ ₂, λ ₃And λ ₄), the absorption coefficient (a of solvable organic matter _CDOM) and the absorption coefficient (a of suspension _Tripton) there is a following approximation relation:

$\frac{a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_1\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_1\right)}{b\left({\mathrm{\λ}}_1\right)}\≈k_1\frac{a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_2\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_2\right)}{b\left({\mathrm{\λ}}_2\right)}$

$\frac{a_{\mathrm{chla}}\left({\mathrm{\λ}}_3\right)+a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_3\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_3\right)}{b\left({\mathrm{\λ}}_3\right)}\≈k_2\frac{a_{\mathrm{chla}}\left({\mathrm{\λ}}_4\right)+a_{\mathrm{CDOM}}\left({\mathrm{\λ}}_4\right)+a_{\mathrm{tripton}}\left({\mathrm{\λ}}_4\right)}{b\left({\mathrm{\λ}}_4\right)}$

$\frac{[R^{-1}\left({\mathrm{\λ}}_1\right)-k_1{R}^{-1}\left({\mathrm{\λ}}_2\right)]}{[R^{-1}\left({\mathrm{\λ}}_3\right)-k_2{R}^1\left({\mathrm{\λ}}_4\right)]}=\frac{a_{\mathrm{chla}}\left({\mathrm{\λ}}_1\right)-k_1{a}_{\mathrm{chla}}\left({\mathrm{\λ}}_2\right)+a_w\left({\mathrm{\λ}}_1\right)-k_1{a}_w\left({\mathrm{\λ}}_2\right)}{a_w\left({\mathrm{\λ}}_4\right)-k_2{a}_w\left({\mathrm{\λ}}_3\right)}$

$P_{\mathrm{Chla}}=\frac{R^{-1}\left({\mathrm{\λ}}_1\right)-k_1{R}^{-1}\left({\mathrm{\λ}}_2\right)}{R^{-1}\left({\mathrm{\λ}}_3\right)-k_2{R}^{-1}\left({\mathrm{\λ}}_4\right)}$

Chla＝a+b+P _Chla

In the formula, λ ₁, λ ₂, λ ₃And λ ₄Be four different wavelength; k ₁And k ₂Model parameter; R is a reflectivity; Chla is a chlorophyll-a concentration; P _ChlaFor half point is analysed algorithm remote sensing parameter; A and b are the empirical relationship parameter between remote sensing parameter and the chlorophyll-a concentration.

3. a kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration according to claim 1 is characterized in that: the optimum band selection standard described in the step (3) is as follows:

The present invention has adopted half point to analyse the predicted value (C of model _{Pred, chla, i}) and measured value (C _{Mea, chla, i}) between the standard deviation (STE) of deviation and relative error (RE) thereof as the standard of model quality; Described standard deviation and relative error expression formula are as follows:

$\mathrm{STE}=\sqrt{\frac{\mathrm{\Σ}{(C_{\mathrm{pred},\mathrm{chla},i}-C_{\mathrm{mea},\mathrm{chla},i})}^2}{N}}$

$\mathrm{RE}=\frac{\mathrm{STE}}{\mathrm{MEAN}\left(C_{\mathrm{mea},\mathrm{chla}}\right)}\×100\%$

4. a kind of semi-analytical method of realizing inversion of water body chlorophyll alpha concentration according to claim 1 is characterized in that: the nonlinear iteration algorithm described in the step (3) is as follows:

Known experimental data (x _i, y _i), i=1,2 ... m, contain nonlinear parameter b among the inverting function y=f (x) ₁, b ₂... b _n, be designated as y=f (b ₁, b ₂... b _n).Make vectorial b=(b ₁, b ₂... b _n), then y=f (x, b) its residual sum of squares (RSS) is:

$S\left(b\right)=\underset{i=1}{\overset{m}{\mathrm{\Σ}}}{[y_i-f(x_i,b)]}^2$

(x b) exists with y=f The Taylor of place launches, and omits high-order term:

$f(x_i,b)=f(x_i,b_1^0,b_2^0,\·\·\·,b_n^0)+\underset{i=1}{\overset{n}{\mathrm{\Σ}}}\frac{{\∂f}_i}{{\∂b}_j}(b_j-b_j^0)$

Comprehensive above-mentioned two formulas, it is as follows to get iterative formula in conjunction with principle of least square method:

$A=\left[\begin{array}{cccc}\frac{{\∂f}_1}{\∂b_1}& \frac{{\∂f}_1}{{\∂b}_2}& \·\·\·& \frac{{\∂f}_1}{{\∂b}_n}\\ \frac{{\∂f}_2}{{\∂b}_1}& \frac{{\∂f}_2}{{\∂b}_2}& \·\·\·& \frac{{\∂f}_2}{{\∂b}_n}\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \·& \·& \·& \·\\ \frac{{\∂f}_m}{{\∂b}_1}& \frac{{\∂f}_m}{{\∂b}_2}& \·\·\·& \frac{{\∂f}_m}{{\∂b}_n}\end{array}\right]$

$A^{T}A(b_j-b_j^0)=A^T(y-f)$

Images