CN104931950A - Target decomposition method based on model for fully-polarized synthetic aperture radar - Google Patents

Abstract

The invention relates to a target decomposition method based on a model for a fully-polarized synthetic aperture radar. The target decomposition method comprises the steps of: acquiring fully-polarized data of a self-correlation matric form; getting volume scattering component power based on a nonnegative eigenvalue decomposition method under the theoretical situation of non-reflective symmetry; calculating a remainder self-correlation matrix according to the obtained volume scattering component power, and carrying out eigenvalue decomposition on the obtained remainder self-correlation matrix to obtain a characteristic vector with eigenvalue being zero and two characteristic vectors with eigenvalue being non-zero; conducting orientation angle compensation and helical angle compensation on the two characteristic vectors with eigenvalue being non-zero in the remainder self-correlation matrix; calculating a new remainder self-correlation matrix on the basis of the characteristic vectors after orientation angle compensation and helical angle compensation; and getting surface scattering power and double scattering power based on the obtained new remainder self-correlation matrix according to a method of getting surface scattering power and double scattering power through three-component decomposition. The target decomposition method provided by the invention has the advantages of no loss of information and no occurrence of negative power.

Description

A kind of goal decomposition method based on model of polarimetric synthetic aperture radar

Technical field

The present invention relates to the Image Information Processing field of polarimetric synthetic aperture radar, particularly polarimetric synthetic aperture radar goal decomposition and the goal decomposition field based on model.

Background technology

Goal decomposition has important effect in the application such as target identification, target classification, geophysical parameters estimation of polarimetric synthetic aperture radar.Goal decomposition based on model is the study hotspot in goal decomposition field.Initiative work based on the goal decomposition of model is that the three-component that Freeman and Durden proposes decomposes (see list of references 1:A.Freeman and S.L.Durden, " A three-component scattering model for polarimetric SARdata; " IEEE Trans.Geosci.Remote Sens., vol.36, no.3, pp.963-973, May 1998.).The back scattering of three-component point stem-butts cutting off target is decomposed into three, is respectively area scattering component, two scattering components and volume scattering component.Three-component decomposes due to its explicit physical meaning, model is simple and be used widely.Think that all cross polarization power is from volume scattering component because three-component decomposes, and when scattering component and two scattering components are subject to the modulation of the angle of orientation and produce a part of cross polarization power face to face, there will be volume scattering and crossed the problem estimated, and cause area scattering component or two scattering components to occur the phenomenon of negative power further.In addition, three-component decomposes hypothetical target and is in reflective symmetry state, and decomposition result comprises five degree of freedom, and autocorrelation matrix has nine degree of freedom, so lost four degree of freedom in decomposable process, causes information dropout.

Four components decompose can alleviate the phenomenon that estimation is crossed in volume scattering, and decomposition result adds one degree of freedom.The angle of orientation of Compensation Objectives also can alleviate the phenomenon that estimation is crossed in volume scattering, and the degree of freedom of autocorrelation matrix can be made to be reduced to eight by nine.The non-negative Eigenvalues Decomposition method that the people such as van Zyl propose, obtain available maximum volume scattering power, avoid the appearance of negative power (see list of references 2:J.J.van Zyl simultaneously, M.Arii, andY.Kim, " Model-based decomposition of polarimetric SAR covariance matricesconstrained for nonnegative eigenvalues; " IEEE Trans.Geosci.Remote Sens., vol.49, no.9, pp.1104 – 1113, Sep.2011).But non-negative Eigenvalues Decomposition method, still based on the hypothesis of reflective symmetry, still can cause the problem of information dropout.Under the first-class people of Cui is generalized to non-reflective symmetrical supposed situation the non-negative Eigenvalues Decomposition under reflective symmetry hypothesis, avoid the problem of information dropout (see list of references 3:Y.Cui, Y.Yamaguchi, J.Yang, H.Kobayashi, S.Park, and G.Singh, " On complete model-based decomposition ofpolarimetric SAR coherency matrix data, " IEEE Trans.Geosci.Remote Sens., vol.52, no.4, pp.1991 – 2001, Apr.2014.), but its area scattering power and two scattering powers are not asked for by scattering model, but asked for by eigenwert, and area scattering power or two scattering powers may be zero.Although the angle of orientation of Compensation Objectives also can alleviate the phenomenon that estimation is crossed in volume scattering, but this method only employs an angle of orientation, the orientation situation of all scatterers in the resolution element in fact representated by autocorrelation matrix all cannot represent by an angle of orientation.Although the method that the first-class people of Cui proposes also compensate for the angle of orientation of two eigenvectors respectively, and the helix angle of eigenvector does not compensate further.In fact, the helix angle of target also can contribute a part of cross polarization.Actual target all can show certain helicity, although the helix angle of most of target may be smaller.The modulation of the angle of orientation and helix angle is not considered when the three-component that Freeman and Durden proposes decomposes opposite scattering component and two scattering component modelings, if so the method using three-component to decompose asks for area scattering component power and two scattering component power, be necessary the impact first compensating the angle of orientation and helix angle.

Summary of the invention

The object of the invention is to a kind of goal decomposition method based on model proposing polarimetric synthetic aperture radar broad sense, successfully can solve information dropout, area scattering power or two scattering powers may for negative problem etc.

To achieve these goals, the invention provides a kind of goal decomposition method based on model of polarimetric synthetic aperture radar, comprising:

Step 1), be obtained from the full polarimetric SAR data of correlation matrix form;

Step 2), ask for volume scattering component power based on the method for the non-negative Eigenvalues Decomposition under non-reflective symmetrical supposed situation;

Step 3), according to step 2) the volume scattering component power that obtains calculates residue autocorrelation matrix, for obtained residue autocorrelation matrix does Eigenvalues Decomposition, obtaining an eigenwert is the eigenvector of zero, and the eigenvector of two eigenwert non-zeros;

Step 4), angle of orientation compensation is carried out to two eigenvectors of eigenwert non-zero in residue autocorrelation matrix and helix angle compensates;

Step 5), by step 4) obtain through the angle of orientation compensate and helix angle compensation after eigenvector calculate new residue autocorrelation matrix;

Step 6), based on step 5) the new residue autocorrelation matrix that obtains, decompose according to three-component ask for area scattering power and two scattering powers method to ask for area scattering power and two scattering powers.

In technique scheme, the step 2 described) in, make T-P _vt _vdeterminant be zero, solve | T-P _vt _v| the cubic equation represented by=0, using root minimum for cubic equation as the required volume scattering component power P got _vvalue; Wherein, T is autocorrelation matrix, P _vt _vfor volume scattering component, P _vfor volume scattering component power, T _vfor the model of volume scattering component, $T_V=\left[\begin{array}{ccc}1/2& 0& 0\\ 0& 1/4& 0\\ 0& 0& 1/4\end{array}\right].$

In technique scheme, the step 3 described) in, for obtained residue autocorrelation matrix does Eigenvalues Decomposition with reference to following formula:

$T_{remainder}=\underset{i=1}{\overset{3}{\Σ}}{\mathrm{\λ}}_i{u}_i{u}_i^{*T}$

Wherein, T _remainderfor residue autocorrelation matrix, λ _ifor eigenwert, u _ifor eigenvector, subscript * T is complex-conjugate transpose operational character.

In technique scheme, described step 4) comprise further:

Step 4-1), suppose two eigenvector u of eigenwert non-zero ₁, u ₂corresponding scattering matrix is S ₁, S ₂, utilize the common diagonalization of Kennaugh-Huynen to scattering matrix to carry out modeling to scattering matrix, modeling formula is as follows:

S _i＝R _θ(θ _i)R _τ(τ _i)S _diR _τ(τ _i)R _θ(-θ _i),i＝1,2.

Wherein, θ _ifor the angle of orientation of scattering matrix, τ _ifor the helix angle of scattering matrix, S _difor diagonal matrix; S _didiagonal element be altogether eigenwert x _i1and x _i2; Matrix R _θ(θ _i) and R _τ(τ _i) being respectively the angle of orientation and unitary transformation matrix corresponding to helix angle, its form is respectively $R_{\mathrm{\θ}}\left({\mathrm{\θ}}_i\right)=\left[\begin{array}{cc}{\mathrm{cos\θ}}_i& -{\mathrm{sin\θ}}_i\\ {\mathrm{sin\θ}}_i& {\mathrm{cos\θ}}_i\end{array}\right]$ With $R_{\mathrm{\τ}}\left({\mathrm{\τ}}_i\right)=\left[\begin{array}{cc}{\mathrm{cos\τ}}_i& -{\mathrm{jsin\τ}}_i\\ -{\mathrm{jsin\τ}}_i& {\mathrm{cos\τ}}_i\end{array}\right];$

Scattering matrix is write as the vector form under Pauli base, is obtained characteristic of correspondence vector u _i:

$u_i=\left[\begin{array}{c}\frac{x_{i1}+x_{i2}}{2}\cos 2{\mathrm{\τ}}_i\\ \frac{x_{i1}-x_{i2}}{2}\cos 2{\mathrm{\θ}}_i+j\frac{x_{i1}+x_{i2}}{2}\sin 2{\mathrm{\τ}}_i\sin 2{\mathrm{\θ}}_i\\ \frac{x_{i1}-x_{i2}}{2}\sin 2{\mathrm{\θ}}_i-j\frac{x_{i1}+x_{i2}}{2}\sin 2{\mathrm{\τ}}_i\cos 2{\mathrm{\θ}}_i\end{array}\right];$

And then ask for the angle of orientation θ of eigenvector _i:

${\mathrm{\θ}}_i=\frac{1}{2}{\tan }^{-1}\[\frac{\mathrm{Re}\[u_i\left(3\right)/u_i\left(1\right)\]}{\mathrm{Re}\[u_i\left(2\right)/u_i\left(1\right)\]}\];$

Wherein, Re is the operational character asking for real;

Step 4-2), angle of orientation compensation is carried out to two eigenvectors of eigenwert non-zero:

u′ _i＝R _θu _i

Wherein, u ' _ithe eigenvector after the angle of orientation compensates, R _θit is angle of orientation rotation matrix; R _θform be:

$R_{\mathrm{\θ}}=\left[\begin{array}{ccc}1& 0& 0\\ 0& cos2{\mathrm{\θ}}_i& sin2{\mathrm{\θ}}_i\\ 0& -sin2{\mathrm{\θ}}_i& cos2{\mathrm{\θ}}_i\end{array}\right];$

Step 4-3), asked for the helix angle τ of eigenvector by the eigenvector after the angle of orientation compensates _i:

${\mathrm{\τ}}_i=\frac{1}{2}{\tan }^{-1}\left(\frac{{\mathrm{ju}}_i^{\′}\left(3\right)}{u_i^{\′}\left(1\right)}\right);$

Step 4-4), to through the angle of orientation compensate after eigenvector carry out helix angle compensation:

u″ _i＝U _τu′ _i；

Wherein, u " _ithe eigenvector after carrying out angle of orientation compensation and helix angle compensation, U _τhelix angle unitary transformation matrix, U _τform be $U_{\mathrm{\τ}}=\left[\begin{array}{ccc}cos2\mathrm{\τ}& 0& jsin2\mathrm{\τ}\\ 0& 1& 0\\ j\sin 2\mathrm{\τ}& 0& \cos 2\mathrm{\τ}\end{array}\right].$

In technique scheme, described step 6) comprise further:

By step 5) expression formula of new residue autocorrelation matrix that obtains is:

$T_{remainder}^{\′}=\underset{i=1}{\overset{2}{\Σ}}{\mathrm{\λ}}_i{u}_i^{\′\′}{u}_i^{\′\′*T}=\left[\begin{array}{ccc}{T}_{11}^{\′}& T_{12}^{\′}& 0\\ T_{12}^{\′*}& T_{22}^{\′}& 0\\ 0& 0& 0\end{array}\right];$

Wherein, T ' ₁₁t ' _remainderthe first row of matrix, the element of first row, T ' ₁₂t ' _remainderthe first row of matrix, the element of secondary series, T ' ₂₂t ' _remaindersecond row of matrix, the element of secondary series; Element T ' ₁₂* be element T ' ₁₂complex conjugate;

The model of area scattering component is: $T_S=\frac{1}{1+|\mathrm{\&beta;}{|}^2}\left[\begin{array}{ccc}1& \mathrm{\&beta;}& 0\\ \mathrm{\&beta;}*& |\mathrm{\&beta;}{|}^2& 0\\ 0& 0& 0\end{array}\right],$ | β | < 1; β is the parameter in model;

The model of two scattering components is: $T_D=\frac{1}{1+|\mathrm{\&alpha;}{|}^2}\left[\begin{array}{ccc}|\mathrm{\&alpha;}{|}^2& \mathrm{\&alpha;}& 0\\ {\mathrm{\&alpha;}}^{*}& 1& 0\\ 0& 0& 0\end{array}\right],$ | α | < 1; α is the parameter in model;

When asking for area scattering power and two scattering powers, if T ' ₁₁> T ' ₂₂, suppose α=0, can β=T ' be obtained ₁₂/ T ' ₁₁, area scattering power two scattering powers if T ' ₁₁≤ T ' ₂₂, suppose β=0, can α=T ' be obtained ₁₂/ T ' ₂₂, area scattering power two scattering powers $P_D=T_{22}^{\&prime;}+\frac{|T_{12}^{\&prime;}{|}^2}{T_{11}^{\&prime;}}.$

The invention has the advantages that: method of the present invention uses the non-negative Eigenvalues Decomposition under non-reflective symmetric condition to ask for volume scattering component can ensure that residue autocorrelation matrix is positive semidefinite matrix, ensure that residue autocorrelation matrix has physical significance.All compensate for two angles of orientation and two helix angles when carrying out angle of orientation compensation and helix angle compensation to residue autocorrelation matrix, the model of the form of the residue autocorrelation matrix after the compensation obtained and area scattering and two scatterings is completely the same.The all scatterers in a resolution element cannot be characterized by an angle of orientation or helix angle, present invention uses two angles of orientation and two helix angles can alleviate this problem.In addition, decompose and obtain nine degree of freedom, there is no information dropout.

Accompanying drawing explanation

Fig. 1 is the process flow diagram of the goal decomposition method based on model of polarimetric synthetic aperture radar of the present invention;

Fig. 2 is the schematic diagram of the decomposition result utilizing method of the present invention to obtain.

Embodiment

Now the invention will be further described by reference to the accompanying drawings.

After method of the present invention utilizes the non-negative Eigenvalues Decomposition under non-reflective symmetry hypothesis to obtain volume scattering power, Eigenvalues Decomposition is carried out to residue autocorrelation matrix, the angle of orientation of difference complementary characteristics vector and helix angle, then the eigenvector linear combination after compensation together, new residue autocorrelation matrix after being compensated, finally asks for area scattering power and two scattering powers based on new residue autocorrelation matrix according to the method that three-component decomposes.The problem of all scatterers in a resolution element cannot be represented in order to solve an angle of orientation and helix angle, method of the present invention all employs two angles of orientation and helix angle when compensate for residual autocorrelation matrix, and after decomposition, the degree of freedom of autocorrelation matrix is not also lost.With reference to figure 1, the concrete steps of the inventive method are as follows:

Step 1), be obtained from the full polarimetric SAR data of correlation matrix T-shaped formula.

The autocorrelation matrix T that this step obtains can be expressed as volume scattering component P _vt _vwith residue autocorrelation matrix T _remainderand form, its expression formula is as follows:

T＝P _VT _V+T _remainder(1)

Wherein, autocorrelation matrix T is remained _remaindercomprise area scattering component and two scattering components; What the model of volume scattering component adopted is that three-component decomposes the model used $T_V=\left[\begin{array}{ccc}1/2& 0& 0\\ 0& 1/4& 0\\ 0& 0& 1/4\end{array}\right].$

Step 2), ask for volume scattering component power P based on the method for the non-negative Eigenvalues Decomposition under non-reflective symmetrical supposed situation _v.

Asking volume scattering component power P _vtime, make T-P _vt _vdeterminant be zero, solve | T-P _vt _v| the cubic equation represented by=0, the minimum root of this cubic equation be exactly required by the volume scattering component power P that gets _vvalue.

Step 3), calculate residue autocorrelation matrix, for obtained residue autocorrelation matrix does Eigenvalues Decomposition.

In this step, from autocorrelation matrix T, deduct volume scattering component P _vt _v, residue autocorrelation matrix T can be obtained _remainder=T-T _vp _v.

To residue autocorrelation matrix T _remainderthe formula carrying out Eigenvalues Decomposition is:

$T_{remainder}=\underset{i=1}{\overset{3}{\&Sigma;}}{\mathrm{\&lambda;}}_i{u}_i{u}_i^{*T}---\left(2\right)$

Wherein, λ _ifor eigenwert, u _ifor eigenvector, subscript * T is complex-conjugate transpose operational character.Due to T-P _vt _vdeterminant be zero, so residue autocorrelation matrix T _remainderan eigenwert is had to be zero.Suppose λ ₁> λ ₂> λ ₃, so λ ₃=0, only need to eigenvector u ₁, u ₂carry out processing, and remain autocorrelation matrix T _remaindercan be rewritten into $T_{remainder}=\underset{i=1}{\overset{2}{\&Sigma;}}{\mathrm{\&lambda;}}_i{u}_i{u}_i^{*T}.$

Step 4), two eigenvectors of eigenwert non-zero in residue autocorrelation matrix are compensated.

When compensating two eigenvectors of eigenwert non-zero, first asking for the angle of orientation and the helix angle of two eigenvectors, then compensating the angle of orientation and the helix angle of these two eigenvectors.Specifically, this step specifically comprises following sub-step:

Step 4-1), suppose eigenvector u ₁, u ₂corresponding scattering matrix is S ₁, S ₂.The common diagonalization of Kennaugh-Huynen to scattering matrix is utilized to carry out modeling to scattering matrix

S _i＝R _θ(θ _i)R _τ(τ _i)S _diR _τ(τ _i)R _θ(-θ _i),i＝1,2. (3)

Wherein, θ _ifor the angle of orientation of scattering matrix, τ _ifor the helix angle of scattering matrix, S _difor diagonal matrix.S _didiagonal element be altogether eigenwert x _i1and x _i2.Matrix R _θ(θ _i) and R _τ(τ _i) being respectively the angle of orientation and unitary transformation matrix corresponding to helix angle, its form is respectively $R_{\mathrm{\&theta;}}\left({\mathrm{\&theta;}}_i\right)=\left[\begin{array}{cc}{\mathrm{cos\&theta;}}_i& -{\mathrm{sin\&theta;}}_i\\ {\mathrm{sin\&theta;}}_i& {\mathrm{cos\&theta;}}_i\end{array}\right]$ With $R_{\mathrm{\&tau;}}\left({\mathrm{\&tau;}}_i\right)=\left[\begin{array}{cc}{\mathrm{cos\&tau;}}_i& -{\mathrm{jsin\&tau;}}_i\\ -{\mathrm{jsin\&tau;}}_i& {\mathrm{cos\&tau;}}_i\end{array}\right].$ Scattering matrix is write as the vector form under Pauli base, characteristic of correspondence vector can be obtained

$u_i=\left[\begin{array}{c}\frac{x_{i1}+x_{i2}}{2}cos2{\mathrm{\&tau;}}_i\\ \frac{x_{i1}-x_{i2}}{2}cos2{\mathrm{\&theta;}}_i+j\frac{x_{i1}+x_{i2}}{2}sin2{\mathrm{\&tau;}}_{i}sin2{\mathrm{\&theta;}}_i\\ \frac{x_{i1}-x_{i2}}{2}sin2{\mathrm{\&theta;}}_i-j\frac{x_{i1}+x_{i2}}{2}sin2{\mathrm{\&tau;}}_{i}cos2{\mathrm{\&theta;}}_i\end{array}\right]---\left(4\right)$

Utilize formula (4) can be easy to ask for the angle of orientation θ of eigenvector _i

${\mathrm{\&theta;}}_i=\frac{1}{2}{\tan }^{-1}\&lsqb;\frac{\mathrm{Re}\&lsqb;u_i\left(3\right)/u_i\left(1\right)\&rsqb;}{\mathrm{Re}\&lsqb;u_i\left(2\right)/u_i\left(1\right)\&rsqb;}\&rsqb;---\left(5\right)$

Wherein, Re is the operational character asking for real.The span of the angle of orientation that formula (5) is asked is [-45 °, 45 °].

Step 4-2), angle of orientation compensation is carried out to two eigenvectors:

u′ _i＝R _θu _i(6)

Wherein, u ' _ithe eigenvector after the angle of orientation compensates, R _θit is angle of orientation rotation matrix.R _θform be: $R_{\mathrm{\&theta;}}=\left[\begin{array}{ccc}1& 0& 0\\ 0& cos2{\mathrm{\&theta;}}_i& sin2{\mathrm{\&theta;}}_i\\ 0& -sin2{\mathrm{\&theta;}}_i& cos2{\mathrm{\&theta;}}_i\end{array}\right].$

Step 4-3), asked for the helix angle τ of eigenvector by the eigenvector after the angle of orientation compensates _i:

${\mathrm{\&tau;}}_i=\frac{1}{2}{\tan }^{-1}\left(\frac{{\mathrm{ju}}_i^{\&prime;}\left(3\right)}{u_i^{\&prime;}\left(1\right)}\right)---\left(7\right)$

Step 4-4), to through the angle of orientation compensate after eigenvector carry out helix angle compensation:

u″ _i＝U _τu′ _i(8)

Wherein, u " _ithe eigenvector after carrying out angle of orientation compensation and helix angle compensation, U _τhelix angle unitary transformation matrix, U _τform be $U_{\mathrm{\&tau;}}=\left[\begin{array}{ccc}cos2\mathrm{\&tau;}& 0& jsin2\mathrm{\&tau;}\\ 0& 1& 0\\ j\sin 2\mathrm{\&tau;}& 0& \cos 2\mathrm{\&tau;}\end{array}\right].$

The eigenvector after the angle of orientation compensates and helix angle compensates utilizing formula (8) to obtain has following form:

$u_i^{\&prime;\&prime;}=\left[\begin{array}{c}\frac{x_{i1}+x_{i2}}{2}\\ \frac{x_{i1}-x_{i2}}{2}\\ 0\end{array}\right]---\left(9\right)$

Its 3rd element is zero, shows that its cross polar component is all compensated.

Step 5), by step 4) obtain through the angle of orientation compensate and helix angle compensation after eigenvector calculate new residue autocorrelation matrix; The expression formula of described residue autocorrelation matrix is newly:

$T_{remainder}^{\&prime;}=\underset{i=1}{\overset{2}{\&Sigma;}}{\mathrm{\&lambda;}}_i{u}_i^{\&prime;\&prime;}{u}_i^{\&prime;\&prime;*T}=\left[\begin{array}{ccc}{T}_{11}^{\&prime;}& T_{12}^{\&prime;}& 0\\ T_{12}^{\&prime;*}& T_{22}^{\&prime;}& 0\\ 0& 0& 0\end{array}\right]---\left(10\right)$

Wherein, T ' ₁₁t ' _remainderthe first row of matrix, the element of first row, other three element definitions are similar.Element T ' ₁₂* be element T ' ₁₂complex conjugate.

Step 6), based on step 5) the new residue autocorrelation matrix T ' that obtains _remainder, decompose according to three-component and ask for area scattering power with pair method of time scattering powers to ask for area scattering power P _swith two scattering power P _d.

Wherein, the model of area scattering component is: $T_S=\frac{1}{1+|\mathrm{\&beta;}{|}^2}\left[\begin{array}{ccc}1& \mathrm{\&beta;}& 0\\ {\mathrm{\&beta;}}^{*}& |\mathrm{\&beta;}{|}^2& 0\\ 0& 0& 0\end{array}\right],$ | β | < 1; β is the parameter in model;

The model of two scattering components is: $T_D=\frac{1}{1+|\mathrm{\&alpha;}{|}^2}\left[\begin{array}{ccc}|\mathrm{\&alpha;}{|}^2& \mathrm{\&alpha;}& 0\\ {\mathrm{\&alpha;}}^{*}& 1& 0\\ 0& 0& 0\end{array}\right],$ | α | < 1; α is the parameter in model.

If T ' ₁₁> T ' ₂₂, suppose α=0, can β=T ' be obtained ₁₂/ T ' ₁₁, area scattering power two scattering powers if T ' ₁₁≤ T ' ₂₂, suppose β=0, can α=T ' be obtained ₁₂/ T ' ₂₂, area scattering power $P_S=T_{11}^{\&prime;}-\frac{|T_{12}^{\&prime;}{|}^2}{T_{11}^{\&prime;}},$ Two scattering powers $P_D=T_{22}^{\&prime;}+\frac{|T_{12}^{\&prime;}{|}^2}{T_{11}^{\&prime;}}.$

So far, volume scattering power P _v, area scattering power P _s, two scattering power P _dall obtained.Although in the methods of the invention, area scattering power P _swith two scattering power P _ddo not asked for by eigenwert, but the area scattering component P utilizing the present invention to obtain _swith the power P of two scattering components _djust be.

In the decomposable process of the inventive method, obtain nine parameter P altogether _s, P _d, P _v, α, β, θ ₁, θ ₂, τ ₁, τ ₂, because autocorrelation matrix T also has nine degree of freedom, so decomposable process does not have degree of freedom to reduce, there is no information dropout.In addition, three the scattering power P obtained _s, P _d, P _vvalue be all greater than zero, do not have the situation of negative power and zero energy to occur.To residue autocorrelation matrix T ' _remaindercompensate for two angles of orientation and two helix angles, compensate more complete.

Fully polarization synthetic aperture radar data below in conjunction with reality verifies the validity of the inventive method.The data used are that the spaceborne polarimetric synthetic aperture radar RADARSAT-2 of Canadian space office is to the imaging data of San Francisco, State of California, US city near zone.Data acquisition time is on April 9th, 2008, and raw data is the form of multiple scattering matrix, and data are looked process to obtain autocorrelation matrix data by six.Because former data scene is too large, the part that we only choose this scene processes.There are sea area, construction zone, park areas in scene and have the mountain area of vegetative coverage.Wherein, the back scattering of sea area based on area scattering, the back scattering of construction zone based on two scatterings, park areas and have the back scattering in district of vegetative coverage based on volume scattering.Fig. 2 is the decomposition result utilizing method of the present invention to obtain.In figure, area scattering is that the region of main scattering represents by grey codes, and two scatterings are the region black coded representation of main scattering, and volume scattering is that the region of main scattering represents by white color-coded.As seen from Figure 2, sea area presents grey, and construction zone presents black, park areas and have the mountain area of vegetative coverage to present white, shows that these atural objects are correctly validated.In order to further illustrate the validity of the inventive method, have chosen three pieces of typical atural objects, and the average power ratio of three of Statistics decomposition result kinds of scattering components, result is as shown in table 1.The three pieces of atural objects chosen are respectively water area, park areas and construction zone.The average power of the area scattering component of sea area is 94.38% as can be seen from Table 1, and area scattering component accounts for definitely main status.The average power ratio of park areas volume scattering component is the highest, tallies with the actual situation.The average power ratio of two scattering components of construction zone is up to 57.17%, and be secondly that the average power ratio of area scattering component is 32.28%, the average power ratio of volume scattering component is minimum is 10.55%.In addition, there is not the situation of negative power and zero energy in the power of three kinds of scattering components of inspection decomposition result.

	Sea area	Park areas	Construction zone
				P S	94.38％	38.30％	32.28％
P D	2.64％	14.79％	57.17％
				P V	2.98％	46.91％	10.55％

Table 1

It should be noted last that, above embodiment is only in order to illustrate technical scheme of the present invention and unrestricted.Although with reference to embodiment to invention has been detailed description, those of ordinary skill in the art is to be understood that, modify to technical scheme of the present invention or equivalent replacement, do not depart from the spirit and scope of technical solution of the present invention, it all should be encompassed in the middle of right of the present invention.

Claims (5)

1. the goal decomposition method based on model of polarimetric synthetic aperture radar, comprising:

Step 1), be obtained from the full polarimetric SAR data of correlation matrix form;

Step 2), ask for volume scattering component power based on the method for the non-negative Eigenvalues Decomposition under non-reflective symmetrical supposed situation;

Step 3), according to step 2) the volume scattering component power that obtains calculates residue autocorrelation matrix, for obtained residue autocorrelation matrix does Eigenvalues Decomposition, obtaining an eigenwert is the eigenvector of zero, and the eigenvector of two eigenwert non-zeros;

Step 4), angle of orientation compensation is carried out to two eigenvectors of eigenwert non-zero in residue autocorrelation matrix and helix angle compensates;

Step 5), by step 4) obtain through the angle of orientation compensate and helix angle compensation after eigenvector calculate new residue autocorrelation matrix;

Step 6), based on step 5) the new residue autocorrelation matrix that obtains, decompose according to three-component ask for area scattering power and two scattering powers method to ask for area scattering power and two scattering powers.

2. the goal decomposition method based on model of polarimetric synthetic aperture radar according to claim 1, is characterized in that, the step 2 described) in, make T-P _vt _vdeterminant be zero, solve | T-P _vt _v| the cubic equation represented by=0, using root minimum for cubic equation as the required volume scattering component power P got _vvalue; Wherein, T is autocorrelation matrix, P _vt _vfor volume scattering component, P _vfor volume scattering component power, T _vfor the model of volume scattering component, $T_V=\left[\begin{array}{ccc}1/2& 0& 0\\ 0& 1/4& 0\\ 0& 0& 1/4\end{array}\right].$

3. the goal decomposition method based on model of polarimetric synthetic aperture radar according to claim 1, is characterized in that, the step 3 described) in, for obtained residue autocorrelation matrix does Eigenvalues Decomposition with reference to following formula:

$T_{remainder}=\underset{i=1}{\overset{3}{\&Sigma;}}{\mathrm{\&lambda;}}_i{u}_i{u}_i^{*T}$

Wherein, T _remainderfor residue autocorrelation matrix, λ _ifor eigenwert, u _ifor eigenvector, subscript * T is complex-conjugate transpose operational character.

4. the goal decomposition method based on model of polarimetric synthetic aperture radar according to claim 1, is characterized in that, described step 4) comprise further:

Step 4-1), suppose two eigenvector u of eigenwert non-zero ₁, u ₂corresponding scattering matrix is S ₁, S ₂, utilize the common diagonalization of Kennaugh-Huynen to scattering matrix to carry out modeling to scattering matrix, modeling formula is as follows:

S _i＝Rθ(θ _i)R _τ(τ _i)S _diR _τ(τ _i)R _θ(-θ _i),i＝1,2.

Wherein, θ _ifor the angle of orientation of scattering matrix, τ _ifor the helix angle of scattering matrix, S _difor diagonal matrix; S _didiagonal element be altogether eigenwert x _i1and x _i2; Matrix R _θ(θ _i) and R _τ(τ _i) being respectively the angle of orientation and unitary transformation matrix corresponding to helix angle, its form is respectively $R_{\mathrm{\&theta;}}\left({\mathrm{\&theta;}}_i\right)=\left[\begin{array}{cc}{\mathrm{cos\&theta;}}_i& -{\mathrm{sin\&theta;}}_i\\ {\mathrm{sin\&theta;}}_i& {\mathrm{cos\&theta;}}_i\end{array}\right]$ With $R_{\mathrm{\&tau;}}\left({\mathrm{\&tau;}}_i\right)=\left[\begin{array}{cc}{\mathrm{cos\&tau;}}_i& -{\mathrm{jsin\&tau;}}_i\\ -{\mathrm{jsin\&tau;}}_i& {\mathrm{cos\&tau;}}_i\end{array}\right];$

Scattering matrix is write as the vector form under Pauli base, is obtained characteristic of correspondence vector u _i:

$u_i=\left[\begin{array}{c}\frac{x_{i1}+x_{i2}}{2}cos2{\mathrm{\&tau;}}_i\\ \frac{x_{i1}-x_{i2}}{2}cos2{\mathrm{\&theta;}}_i+j\frac{x_{i1}+x_{i2}}{2}sin2{\mathrm{\&tau;}}_{i}sin2{\mathrm{\&theta;}}_i\\ \frac{x_{i1}-x_{i2}}{2}sin2{\mathrm{\&theta;}}_i-j\frac{x_{i1}+x_{i2}}{2}sin2{\mathrm{\&tau;}}_{i}cos2{\mathrm{\&theta;}}_i\end{array}\right];$

And then ask for the angle of orientation θ of eigenvector _i:

${\mathrm{\&theta;}}_i=\frac{1}{2}{\tan }^{-1}\&lsqb;\frac{\mathrm{Re}\&lsqb;u_i\left(3\right)/u_i\left(1\right)\&rsqb;}{\mathrm{Re}\&lsqb;u_1\left(2\right)/u_i\left(1\right)\&rsqb;}\&rsqb;$

Wherein, Re is the operational character asking for real;

Step 4-2), angle of orientation compensation is carried out to two eigenvectors of eigenwert non-zero:

u′ _i＝R _θu _i

Wherein, u ' _ithe eigenvector after the angle of orientation compensates, R _θit is angle of orientation rotation matrix; R _θform be:

$R_{\mathrm{\&theta;}}=\left[\begin{array}{ccc}1& 0& 0\\ 0& cos2{\mathrm{\&theta;}}_i& sin2{\mathrm{\&theta;}}_i\\ 0& -sin2{\mathrm{\&theta;}}_i& cos2{\mathrm{\&theta;}}_i\end{array}\right];$

Step 4-3), asked for the helix angle τ of eigenvector by the eigenvector after the angle of orientation compensates _i:

${\mathrm{\&tau;}}_i=\frac{1}{2}{\tan }^{-1}\left(\frac{{\mathrm{ju}}_i^{\&prime;}\left(3\right)}{u_i^{\&prime;}\left(1\right)}\right);$

Step 4-4), to through the angle of orientation compensate after eigenvector carry out helix angle compensation:

u″ _i＝U _τu′ _i

Wherein, u " _ithe eigenvector after carrying out angle of orientation compensation and helix angle compensation, U _τhelix angle unitary transformation matrix, U _τform be $U_{\mathrm{\&tau;}}=\left[\begin{array}{ccc}cos2\mathrm{\&tau;}& 0& jsin2\mathrm{\&tau;}\\ 0& 1& 0\\ j\sin 2\mathrm{\&tau;}& 0& \cos 2\mathrm{\&tau;}\end{array}\right].$

5. the goal decomposition method based on model of polarimetric synthetic aperture radar according to claim 1, is characterized in that, described step 6) comprise further:

By step 5) expression formula of new residue autocorrelation matrix that obtains is:

$T_{\mathrm{remainder}}^{\&prime;}=\underset{i=l}{\overset{2}{\mathrm{\&Sigma;}}}{y}_i{u}_i^{\&prime;\&prime;}{u}_i^{\&prime;\&prime;*T}=\left[\begin{array}{ccc}{T}_{11}^{\&prime;}& T_{12}^{\&prime;}& 0\\ T_{12}^{\&prime;}& T_{22}^{\&prime;}& 0\\ 0& 0& 0\end{array}\right]$

Wherein, T ' ₁₁t ' _remainderthe first row of matrix, the element of first row, T ' ₁₂t ' _remainderthe first row of matrix, the element of secondary series, T ' ₂₂t ' _remaindersecond row of matrix, the element of secondary series; Element be element T ' ₁₂complex conjugate;

The model of area scattering component is: $T_S=\frac{1}{1+|\mathrm{\&beta;}{|}^2}\left[\begin{array}{ccc}1& \mathrm{\&beta;}& 0\\ \mathrm{\&beta;}*& |\mathrm{\&beta;}{|}^2& 0\\ 0& 0& 0\end{array}\right],|$ | β | < 1; β is the parameter in model;

The model of two scattering components is: $T_D=\frac{1}{1+|\mathrm{\&alpha;}{|}^2}\left[\begin{array}{ccc}|\mathrm{\&alpha;}{|}^2& \mathrm{\&alpha;}& 0\\ \mathrm{\&alpha;}*& 1& 0\\ 0& 0& 0\end{array}\right],$ | α | < 1; α is the parameter in model;

When asking for area scattering power and two scattering powers, if T ' ₁₁> T ' ₂₂, suppose α=0, can β=T ' be obtained ₁₂/ T ' ₁₁, area scattering power two scattering powers if T ' ₁₁≤ T ' ₂₂, suppose β=0, can α=T ' be obtained ₁₂/ T ' ₂₂, area scattering power two scattering powers

$P_D=T_{22}^{\&prime;}+{\frac{\left|T_{12}^{\&prime;}\right|}{T_{11}^{\&prime;}}}^2.$