CN102354393A - Hyperspectral imaging device data noise elimination method - Google Patents

Abstract

The invention relates to a hyperspectral imaging device data noise elimination method. The hyperspectral imaging device data noise elimination method includes the following steps: (1) obtaining surface reflectance data removed of the affection of atmospheric noise; (2) selecting a uniform area with relatively stable ground object characteristics, and taking statistics on the statistical characteristic quantity of the surface reflectance data of the area; (3) dividing the band of hyperspectral imaging device data into a smooth data information band, a data information distortion-related band and a data information distortion-irrelated band; (4) respectively eliminating the noise of the data information distortion-related band and the data information distortion-irrelated band; (5) utilizing the actually measured surface reflectance to carry out empirical flat field optimal reflectance transformation on the smooth data information band, the noiseless data information distortion-related band and the noiseless data information distortion-irrelated band in order to eliminate the accumulative noise error in the process of hyperspectral imaging device data processing and finally obtain the noiseless surface reflectance data of a hyperspectral imaging device.

Description

A kind of noise remove method of hyperspectral imager data

Technical field

The present invention relates to a kind of noise remove method of hyperspectral imager data, particularly a kind of noise remove method of environment mitigation satellite hyperspectral imager data.

Background technology

The hyperspectral imager data are one " image cubes "; Its spatial image dimension is described face of land two-dimensional space characteristic; Its spectrum dimension discloses the curve of spectrum characteristic of each pixel of image; It obtains abundant atural object spatial information, radiation information and spectral information with nano level spectral resolution, realizes organically blending of remotely-sensed data space dimension and spectrum dimension information thus, has great application prospect aspect the remote sensing quantification.

The view data that hyperspectral imager obtains all is that the two-dimensional projection in the three-dimensional scene shows; Owing to receive the influence of atmosphere radiation transmission, signal acquisition circuit, signal data processing etc.; In the hyperspectral imager data, often there is each noise like; Real radiation information in the hyperspectral imager data has been covered in the existence of noise; Make the signal to noise ratio (S/N ratio), sharpness of hyperspectral imager data reduce, have a strong impact on the hyperspectral imager quality of data and follow-up quantitative remote sensing is used to bring cause adverse effect.At present existing several different methods is carried out noise remove to the hyperspectral imager data; But no matter be based on the spatial domain method removal and also be based on the frequency domain method removal; All be through on the gray value data basis of hyperspectral imager, carrying out noise remove in the hope of reaching good display; And that the genuine property of the clutter reflections rate during to the noise place to go keeps considering is very few, and this quantification to the hyperspectral imager data is used and caused very big uncertainty.

Summary of the invention

Technology of the present invention is dealt with problems and is: overcome the deficiency of prior art, a kind of noise remove method that can fine removal noise can well keep the hyperspectral imager data of atural object real reflectance characteristic again is provided.

Technical solution of the present invention is: a kind of noise remove method of hyperspectral imager data, and step is following:

(1) reading remote sensing satellite hyperspectral imager data, these hyperspectral imager data are carried out pre-service, is apparent spoke brightness data with the hyperspectral imager data-switching; Again apparent spoke brightness data is carried out atmosphere and correct, obtain removing the earth surface reflection rate data of atmospheric noise influence;

(2) according to the revised hyperspectral imager earth surface reflection of above-mentioned atmosphere rate data, choose the metastable homogeneous area of atural object characteristic, add up the statistical characteristic value of these zone earth surface reflection rate data;

(3) according to this range statistics characteristic quantity, in conjunction with surveying reflectivity ρ by zone atural object _Measure(λ), the wave band with the hyperspectral imager data is divided into the data message steadily relevant wave band with the data message distortion of wave band ρ (i), the uncorrelated wave band of data message distortion; Divide foundation as follows:

(3.1), it is confirmed as steadily wave band ρ (i) of data message if 2.～5. the following Rule of judgment of hyperspectral imager wave band λ all is true; Otherwise,, it is confirmed as data message distortion wave band ρ (j) if 2.～5. the following Rule of judgment of hyperspectral imager wave band λ has one not for true;

|ρ _Zmax(λ)-ρ _measure(λ)|＜ε ₀ ②

|ρ _Zmin(λ)-ρ _measure(λ)|＜ε ₀ ③

|ρ _Zmean(λ)-ρ _measure(λ)|＜ε ₁ ④

$\mathrm{SNR}\left(\mathrm{\λ}\right)=\frac{{\mathrm{\ρ}}_{\mathrm{Zmean}}\left(\mathrm{\λ}\right)}{\mathrm{\σ}\left(\mathrm{\λ}\right)}>{\mathrm{\ϵ}}_2$ ⑤

Wherein:

ρ _Zmax(λ) the Z section maximal value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Zmin(λ) the Z section minimum value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Zmean(λ) the Z section mean value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Measure(λ) the actual measurement earth surface reflection rate of corresponding statistical regions hyperspectral imager wave band λ;

The mean square deviation of the corresponding statistical regions hyperspectral imager of σ (λ) wave band λ;

The signal to noise ratio (S/N ratio) of the corresponding statistical regions hyperspectral imager of SNR (λ) wave band λ adopts the variance method estimation to calculate;

ε ₀, ε ₁, ε ₂Corresponding error in judgement scope;

(3.2) for data message distortion wave band ρ (j) again through with data message steadily the wave band relevance formula of wave band ρ (i) 6. classify again, if satisfy 6. condition of formula, data message distortion wave band ρ (j) is confirmed as the relevant wave band ρ of data message distortion _Cor(j); Otherwise,, data message distortion wave band is confirmed as the uncorrelated wave band ρ of data message distortion if do not satisfy 6. condition of formula _Irr(j);

$|{\mathrm{Cor}}_{j,i}=\frac{\underset{k}{\overset{N}{\mathrm{\Σ}}}[{\mathrm{\ρ}}_k\left(j\right)-{\mathrm{\ρ}}_{\mathrm{mean}}\left(j\right)][{\mathrm{\ρ}}_k\left(i\right)-{\mathrm{\ρ}}_{\mathrm{mean}}\left(i\right)]}{(N-1){\mathrm{\σ}}_j{\mathrm{\σ}}_i}|>{\mathrm{\ϵ}}_3$ ⑥

Wherein:

The earth surface reflection rate data gray level of the corresponding statistical regions hyperspectral imager of k wave band;

The earth surface reflection rate data gray level sum of the corresponding statistical regions hyperspectral imager of N wave band;

ρ _k(j) the earth surface reflection rate value of the corresponding statistical regions gray level data message distortion wave band j that is k;

ρ _k(i) the steady earth surface reflection rate value of wave band i of the corresponding statistical regions gray level data message that is k;

ρ _Mean(j) the earth surface reflection rate mean value of corresponding statistical regions data message distortion wave band j;

ρ _Mean(i) corresponding statistical regions data message is steadily than the earth surface reflection rate mean value of wave band i;

σ _jThe mean square deviation of corresponding statistical regions data message distortion wave band j;

σ _iCorresponding statistical regions data message is the mean square deviation of wave band i steadily;

Cor _{J, i}It is the steadily wave band related coefficient of wave band i of data message distortion wave band j and data message;

ε ₃Corresponding error in judgement scope;

(4) according to above-mentioned division, respectively the relevant wave band of data message distortion, the uncorrelated wave band of data message distortion are carried out noise remove to the relevant wave band of data message distortion, the uncorrelated wave band of data message distortion;

(5) to data message distortion relevant wave band, the data message distortion uncorrelated wave band of data than steady wave band and the above-mentioned noise remove of process; Utilize actual measurement earth surface reflection rate; Carry out experience flat field territory optimum reflection reflectivity conversion; To eliminate the integrated noise error in the hyperspectral imager data handling procedure, obtain the earth surface reflection rate data that hyperspectral imager is removed noise at last.

The relevant wave band of described data message distortion adopts energy to carry out noise remove than texture variance transplanting algorithm.

The uncorrelated wave band of described data message distortion adopts the histogram matching algorithm to carry out noise remove.

The present invention compared with prior art beneficial effect is:

(1) the present invention is according to the hyperspectral imager data characteristics; On the basis of hyperspectral imager earth surface reflection rate data; The concrete analysis hyperspectral imager produces the different reasons of noise; Adopt diverse ways that the hyperspectral imager data noise is removed to the different noise origin causes of formation; Eliminate the various forms noise of a plurality of band images of high spectrum on the whole, thereby effectively guaranteed the removal of noise and the maintenance of atural object characteristic, for the follow-up quantification of hyperspectral imager data is used the raw information of having recovered data well.

(2) the present invention is from the quantification application aspect of hyperspectral imager data; To the different noises of data wave band, utilize distinct methods that the different pieces of information band noise is removed, method is complete, rationally feasible, noise remove is obvious; Information loss is less, has kept the atural object characteristic intactly.

(3) the present invention starts with from the imaging mechanism of hyperspectral imager, after eliminating the atmospheric noise influence, carries out the removal of noise based on hyperspectral imager earth surface reflection rate data, and this makes that the removal of noise is more pointed.

Description of drawings

Fig. 1 is an overview flow chart of the present invention;

Fig. 2 is the inventive method process flow diagram.

Embodiment

Introduce implementation procedure of the present invention in detail below in conjunction with accompanying drawing 1,2, concrete steps are following:

(1) reading remote sensing satellite hyperspectral imager data, these hyperspectral imager data are carried out pre-service, is apparent spoke brightness data with the hyperspectral imager data-switching; Again apparent spoke brightness data is carried out atmosphere and correct, obtain removing the earth surface reflection rate data of atmospheric noise influence;

(1.1) apparent spoke brightness

Read hyperspectral imager DN value (brightness value of each pixel) data, according to the absolute calibration coefficient in the data accompanying document, 1. calculating the spoke brightness of hyperspectral imager entrance pupil place according to formula is apparent spoke brightness:

$L\left(\mathrm{\λ}\right)=\frac{\mathrm{DN}\left(\mathrm{\λ}\right)}{A\left(\mathrm{\λ}\right)}+L_0\left(\mathrm{\λ}\right)$ ①

Wherein: λ is the wave band of hyperspectral imager;

The entrance pupil place spoke brightness of the corresponding hyperspectral imager wave band of L (λ) λ;

The digital DN value of the corresponding hyperspectral imager wave band of DN (λ) λ;

A (λ), L ₀(λ) calibration coefficient of corresponding hyperspectral imager wave band λ.

(1.2) atmosphere is corrected

Atmosphere is corrected the FLAASH atmosphere that adopts among the ENVI and is corrected module; It is the first-selected atmospheric correction model of hyperspectral imager data inversion; Setting according to the user; The FLAASH module forms specific solution to the hyperspectral imager data, accurately eliminates the atmospheric noise influence, obtains the higher earth surface reflection rate data of precision.

Correlation parameter when the parameter of above-mentioned FLAASH module is provided with and can obtains to form images according to hyperspectral imager data accompanying document.Here want to be noted that; At least contain two basic document in the at present general remotely-sensed data bag: one is " remote sensing image data file "; Another is exactly " accompanying document during satellite imagery " (comprising information such as satellite imagery time, regional extent and calibration coefficient), and this partial content is as well known to those skilled in the art.

(2) according to the revised hyperspectral imager earth surface reflection of above-mentioned atmosphere rate data; Choose the metastable homogeneous area of atural object characteristic such as Gobi desert and water field of big area etc., add up the statistical characteristic value of these zone earth surface reflection rate data: the maximum of histogram, variance, covariance, correlativity etc. and this zone Z section, minimum and average statistics;

(3) according to this range statistics characteristic quantity; In conjunction with surveying reflectivity by zone atural object, the following Rule of judgment of foundation is divided into the data message steadily relevant wave band with the data message distortion of wave band, the uncorrelated wave band of data message distortion to the wave band of hyperspectral imager data;

(3.1), it is confirmed as steadily wave band ρ (i) of data message if 2.～5. the following Rule of judgment of hyperspectral imager wave band λ all is true; Otherwise,, it is confirmed as data message distortion wave band ρ (j) if 2.～5. the following Rule of judgment of hyperspectral imager wave band λ has one not for true;

|ρ _Zmax(λ)-ρ _measure(λ)|＜ε ₀②

|ρ _Zmin(λ)-ρ _measure(λ)|＜ε ₀③

|ρ _Zmean(λ)-ρ _measure(λ)|＜ε ₁④

$\mathrm{SNR}\left(\mathrm{\λ}\right)=\frac{{\mathrm{\ρ}}_{\mathrm{Zmean}}\left(\mathrm{\λ}\right)}{\mathrm{\σ}\left(\mathrm{\λ}\right)}>{\mathrm{\ϵ}}_2$ ⑤

Wherein:

ρ _Zmax(λ) the Z section maximal value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Zmin(λ) the Z section minimum value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Zmean(λ) the Z section mean value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Measure(λ) the actual measurement earth surface reflection rate of corresponding statistical regions hyperspectral imager wave band λ;

The mean square deviation of the corresponding statistical regions hyperspectral imager of σ (λ) wave band λ;

The signal to noise ratio (S/N ratio) of the corresponding statistical regions hyperspectral imager of SNR (λ) wave band λ adopts the variance method estimation to calculate;

ε ₀, ε ₁, ε ₂Corresponding error in judgement scope;

(3.2) for data message distortion wave band ρ (j) again through with data message steadily the wave band relevance formula of wave band ρ (i) 6. classify again, if satisfy 6. condition of formula, data message distortion wave band ρ (j) is confirmed as the relevant wave band ρ of data message distortion _Cor(j); Otherwise,, data message distortion wave band is confirmed as the uncorrelated wave band ρ of data message distortion if do not satisfy 6. condition of formula _Irr(j);

$|{\mathrm{Cor}}_{j,i}=\frac{\underset{k}{\overset{N}{\mathrm{\Σ}}}[{\mathrm{\ρ}}_k\left(j\right)-{\mathrm{\ρ}}_{\mathrm{mean}}\left(j\right)][{\mathrm{\ρ}}_k\left(i\right)-{\mathrm{\ρ}}_{\mathrm{mean}}\left(i\right)]}{(N-1){\mathrm{\σ}}_j{\mathrm{\σ}}_i}|>{\mathrm{\ϵ}}_3$ ⑥

Wherein:

The earth surface reflection rate data gray level of the corresponding statistical regions hyperspectral imager of k wave band;

The earth surface reflection rate data gray level sum of the corresponding statistical regions hyperspectral imager of N wave band;

ρ _k(j) the earth surface reflection rate value of the corresponding statistical regions gray level data message distortion wave band j that is k;

ρ _k(i) the steady earth surface reflection rate value of wave band i of the corresponding statistical regions gray level data message that is k;

ρ _Mean(j) the earth surface reflection rate mean value of corresponding statistical regions data message distortion wave band j;

ρ _Mean(i) corresponding statistical regions data message is steadily than the earth surface reflection rate mean value of wave band i;

σ _jThe mean square deviation of corresponding statistical regions data message distortion wave band j;

σ _iCorresponding statistical regions data message is the mean square deviation of wave band i steadily;

Cor _{J, i}It is the steadily wave band related coefficient of wave band i of data message distortion wave band j and data message;

ε ₃Corresponding error in judgement scope.

Above-mentioned: ε ₀, ε ₁, ε ₂And ε ₃Error range, accept or reject accordingly according to reckoner's accuracy requirement is different, for as well known to those skilled in the art.

(4) according to above-mentioned division, adopt diverse ways that hyperspectral imager earth surface reflection rate data message distortion wave band is carried out noise remove respectively to hyperspectral imager data message distortion wave band ρ (j);

(4.1) the relevant wave band of data message distortion adopts energy to transplant algorithm than texture variance

With the relevant wave band ρ of data message distortion _Cor(j) regard the target wave band as, will relevant wave band ρ with the data message distortion _Cor(j) data message steadily wave band ρ (i) is regarded the source wave band as, then has energy than condition than function formula to be:

$d_{l,m}=\frac{{\mathrm{\ρ}}_{\mathrm{co}{r}_{l,m}}\left(j\right)}{{\mathrm{\ρ}}_{l,m}\left(i\right)}$ ⑦

Wherein:

L, the line number and the columns of the corresponding statistical regions data of m;

ρ _{L, m}(i) corresponding statistical regions data message is steadily than the earth surface reflection rate value of the line number l columns m of wave band i;

ρ _{Corl, m}(j) the earth surface reflection rate value of the line number l columns m of the relevant wave band j of corresponding statistical regions data message distortion;

d _{L, m}The energy of corresponding statistical regions earth surface reflection rate number of data lines l columns m is than condition ratio;

So, when the earth surface reflection rate of the steady wave band ρ (i) of source wave band data information is:

$\mathrm{\ρ}\left(i\right)=\left[\begin{array}{ccccc}{\mathrm{\ρ}}_{\mathrm{0,0}}\left(i\right)& {\mathrm{\ρ}}_{\mathrm{0,1}}\left(i\right)& {\mathrm{\ρ}}_{\mathrm{0,2}}\left(i\right)...& & {\mathrm{\ρ}}_{0,m}\left(i\right)\\ {\mathrm{\ρ}}_{\mathrm{1,0}}\left(i\right)& {\mathrm{\ρ}}_{\mathrm{1,1}}\left(i\right)& {\mathrm{\ρ}}_{\mathrm{1,2}}\left(i\right)...& & {\mathrm{\ρ}}_{1,m}\left(i\right)\\ {\mathrm{\ρ}}_{\mathrm{2,0}}\left(i\right)& {\mathrm{\ρ}}_{\mathrm{2,1}}\left(i\right)& {\mathrm{\ρ}}_{\mathrm{2,2}}\left(i\right)...& & {\mathrm{\ρ}}_{2,m}\left(i\right)\\ .& .& .& & .\\ .& .& .& & .\\ .& .& .& ...& .\\ {\mathrm{\ρ}}_{l,0}\left(i\right)& {\mathrm{\ρ}}_{l,1}\left(i\right)& {\mathrm{\ρ}}_{l,2}\left(i\right)...& & {\mathrm{\ρ}}_{l,m}\left(i\right)\end{array}\right]$ ⑧

Then have, transplant algorithm through energy than texture variance and remove the relevant wave band ρ of object wave segment data information distortion _Cor(j) the earth surface reflection rate ρ behind the noise _E-σ(j) be:

${\mathrm{\ρ}}_{E-\mathrm{\σ}}\left(i\right)=\left[\begin{array}{ccccc}{d_{\mathrm{0,0}}\mathrm{\ρ}}_{\mathrm{0,0}}\left(i\right)& {d_{\mathrm{0,1}}\mathrm{\ρ}}_{\mathrm{0,1}}\left(i\right)& d_{\mathrm{0,2}}{\mathrm{\ρ}}_{\mathrm{0,2}}\left(i\right)...& & d_{0,m}{\mathrm{\ρ}}_{0,m}\left(i\right)\\ {d_{\mathrm{1,0}}\mathrm{\ρ}}_{\mathrm{1,0}}\left(i\right)& d_{\mathrm{1,0}}{\mathrm{\ρ}}_{\mathrm{1,1}}\left(i\right)& {d_{\mathrm{1,0}}\mathrm{\ρ}}_{\mathrm{1,2}}\left(i\right)...& & d_{\mathrm{1,0}}{\mathrm{\ρ}}_{1,m}\left(i\right)\\ d_{\mathrm{2,0}}{\mathrm{\ρ}}_{\mathrm{2,0}}\left(i\right)& {d_{\mathrm{2,0}}\mathrm{\ρ}}_{\mathrm{2,1}}\left(i\right)& {d_{\mathrm{2,0}}\mathrm{\ρ}}_{\mathrm{2,2}}\left(i\right)...& & d_{\mathrm{2,0}}{\mathrm{\ρ}}_{2,m}\left(i\right)\\ .& .& .& & .\\ .& .& .& & .\\ .& .& .& ...& .\\ d_{l,0}{\mathrm{\ρ}}_{l,0}\left(i\right)& d_{l,0}{\mathrm{\ρ}}_{l,1}\left(i\right)& d_{l,0}{\mathrm{\ρ}}_{l,2}\left(i\right)...& & d_{l,0}{\mathrm{\ρ}}_{l,m}\left(i\right)\end{array}\right]$ ⑨

(4.2) the uncorrelated wave band of data message distortion adopts the histogram matching method

A) to the uncorrelated wave band ρ of data message distortion _Irr(j) carry out statistics with histogram, the calculating probability density fonction:

$s_{k-\mathrm{irr}}\left(j\right)=\underset{k}{\overset{N}{\mathrm{\Σ}}}{p}_k\left({\mathrm{\ρ}}_{\mathrm{irr}}\left(j\right)\right)$ ⑩

Wherein:

p _k(ρ _Irr(j)) the uncorrelated wave band ρ of corresponding statistical regions data message distortion _Irr(j) frequency values that gray level is k appears;

s _K-irr(j) the uncorrelated wave band ρ of corresponding statistical regions data message distortion _Irr(j) probability density function;

Other parameters are the same.

B), set up the probability density function t of coupling according to the wave band response function of statistical regions actual measurement earth surface reflection and hyperspectral imager data _K-irr(j):

$t_{k-\mathrm{irr}}\left(j\right)=\underset{k}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\ρ}}_k\left({\mathrm{\ρ}}_{\mathrm{measure}}\right)$

C) set up the original probability density function s of the uncorrelated wave band of data message distortion _K-irr(j) with matching probability density function t _K-irr(j) mapping table is with the uncorrelated wave band ρ of data message distortion _Irr(j) original histogram can be eliminated the noise in the uncorrelated wave band of data message distortion, the quantity of information of recovering damage through the histogram of look-up table correspondence mappings to coupling.

(5) to data message distortion relevant wave band, the data message distortion uncorrelated wave band of data than steady wave band and the above-mentioned noise remove of process; Utilize actual measurement earth surface reflection rate; Carry out experience flat field territory optimum reflection reflectivity conversion (EFFORT); To eliminate the integrated noise error in the hyperspectral imager data handling procedure, obtain the earth surface reflection rate data that hyperspectral imager is removed noise at last.

Through actual measurement contrast, Ben Fafa can not only remove the noise of hyperspectral imager data well, and recovers well and keep atural object real surface reflectivity Characteristics.

The unspecified part of the present invention belongs to general knowledge as well known to those skilled in the art.

Claims (3)

1. the noise remove method of hyperspectral imager data is characterized in that step is following:

(1) reading remote sensing satellite hyperspectral imager data, these hyperspectral imager data are carried out pre-service, is apparent spoke brightness data with the hyperspectral imager data-switching; Again apparent spoke brightness data is carried out atmosphere and correct, obtain removing the earth surface reflection rate data of atmospheric noise influence;

(2) according to the revised hyperspectral imager earth surface reflection of above-mentioned atmosphere rate data, choose the metastable homogeneous area of atural object characteristic, add up the statistical characteristic value of these zone earth surface reflection rate data;

(3) according to this range statistics characteristic quantity, in conjunction with surveying reflectivity ρ by zone atural object _Measure(λ), the wave band with the hyperspectral imager data is divided into the data message steadily relevant wave band with the data message distortion of wave band ρ (i), the uncorrelated wave band of data message distortion; Divide foundation as follows:

(3.1), it is confirmed as steadily wave band ρ (i) of data message if 2.～5. the following Rule of judgment of hyperspectral imager wave band λ all is true; Otherwise,, it is confirmed as data message distortion wave band ρ (j) if 2.～5. the following Rule of judgment of hyperspectral imager wave band λ has one not for true;

|ρ _Zmax(λ)-ρ _measure(λ|＜ε ₀ ②

|ρ _Zmin(λ)-ρ _measure(λ)|＜ε ₀ ③

|ρ _Zmean(λ)-ρ _measure(λ)|＜ε ₁ ④

$\mathrm{SNR}\left(\mathrm{\λ}\right)=\frac{{\mathrm{\ρ}}_{\mathrm{Zmean}}\left(\mathrm{\λ}\right)}{\mathrm{\σ}\left(\mathrm{\λ}\right)}>{\mathrm{\ϵ}}_2$ ⑤

Wherein:

ρ _Zmax(λ) the Z section maximal value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Zmin(λ) the Z section minimum value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Zmean(λ) the Z section mean value of corresponding statistical regions hyperspectral imager wave band λ;

ρ _Measure(λ) the actual measurement earth surface reflection rate of corresponding statistical regions hyperspectral imager wave band λ;

The mean square deviation of the corresponding statistical regions hyperspectral imager of σ (λ) wave band λ;

The signal to noise ratio (S/N ratio) of the corresponding statistical regions hyperspectral imager of SNR (λ) wave band λ adopts the variance method estimation to calculate;

ε ₀, ε ₁, ε ₂Corresponding error in judgement scope;

(3.2) for data message distortion wave band ρ (j) again through with data message steadily the wave band relevance formula of wave band ρ (i) 6. classify again, if satisfy 6. condition of formula, data message distortion wave band ρ (j) is confirmed as the relevant wave band ρ of data message distortion _Cor(j); Otherwise,, data message distortion wave band is confirmed as the uncorrelated wave band ρ of data message distortion if do not satisfy 6. condition of formula _Irr(j);

$|{\mathrm{Cor}}_{j,i}=\frac{\underset{k}{\overset{N}{\mathrm{\Σ}}}[{\mathrm{\ρ}}_k\left(j\right)-{\mathrm{\ρ}}_{\mathrm{mean}}\left(j\right)][{\mathrm{\ρ}}_k\left(i\right)-{\mathrm{\ρ}}_{\mathrm{mean}}\left(i\right)]}{(N-1){\mathrm{\σ}}_j{\mathrm{\σ}}_i}|>{\mathrm{\ϵ}}_3$ ⑥

Wherein:

The earth surface reflection rate data gray level of the corresponding statistical regions hyperspectral imager of k wave band;

The earth surface reflection rate data gray level sum of the corresponding statistical regions hyperspectral imager of N wave band;

ρ _k(j) the earth surface reflection rate value of the corresponding statistical regions gray level data message distortion wave band j that is k;

ρ _k(i) the steady earth surface reflection rate value of wave band i of the corresponding statistical regions gray level data message that is k;

ρ _Mean(j) the earth surface reflection rate mean value of corresponding statistical regions data message distortion wave band j;

ρ _Mean(i) corresponding statistical regions data message is steadily than the earth surface reflection rate mean value of wave band i;

σ _jThe mean square deviation of corresponding statistical regions data message distortion wave band j;

σ _iCorresponding statistical regions data message is the mean square deviation of wave band i steadily;

Cor _{J, i}It is the steadily wave band related coefficient of wave band i of data message distortion wave band j and data message;

ε ₃Corresponding error in judgement scope;

(4) according to above-mentioned division, respectively the relevant wave band of data message distortion, the uncorrelated wave band of data message distortion are carried out noise remove to the relevant wave band of data message distortion, the uncorrelated wave band of data message distortion;

(5) to data message distortion relevant wave band, the data message distortion uncorrelated wave band of data than steady wave band and the above-mentioned noise remove of process; Utilize actual measurement earth surface reflection rate; Carry out experience flat field territory optimum reflection reflectivity conversion; To eliminate the integrated noise error in the hyperspectral imager data handling procedure, obtain the earth surface reflection rate data that hyperspectral imager is removed noise at last.

2. the noise remove method of a kind of hyperspectral imager data according to claim 1 is characterized in that: the relevant wave band of described data message distortion adopts energy to carry out noise remove than texture variance transplanting algorithm.

3. the noise remove method of a kind of hyperspectral imager data according to claim 1 is characterized in that: the uncorrelated wave band of described data message distortion adopts the histogram matching algorithm to carry out noise remove.

Images