CN103870841A - Polarization SAR (synthetic aperture radar) image classification method based on Freeman decomposition and PSO (particle swarm optimization) - Google Patents

Abstract

The invention discloses a polarization SAR (synthetic aperture radar) image classification method based on Freeman decomposition and PSO (particle swarm optimization), and mainly aims to solve the problems of higher computational complexity and poor classification effect in the prior art. The implementation steps are as follows: (1) inputting a covariance matrix of polarization SAR data; (2) performing Freeman decomposition on the input matrix to obtain three scattering power matrixes of plane scattering, dihedral angle scattering and volume scattering; (3) initially dividing the SAR data according to the three scattering power matrixes; (4) obtaining two threshold values of each category by virtue of the two-dimensional double threshold value Otsu method on the basis of QPSO (quantum-behaved particle swarm optimization); (5) dividing each initially divided category of the polarization SAR data into three categories, thereby dividing the whole polarization SAR data into 9 categories; (6) carrying out Wishart iteration and coloring on the division result of the whole SAR data to obtain a final color classification result image. Compared with the classical classification method, the polarization SAR image classification method based on Freeman decomposition and PSO is more rigorous in dividing the polarization SAR data, the classification result is obvious, and the computational complexity is relatively small.

Description

Based on the Classification of Polarimetric SAR Image method of Freeman decomposition and particle group optimizing

Technical field

The invention belongs to image data processing technology field, specifically a kind of image classification method, the method can be used for the classification to polarization SAR data.

Background technology

Along with the development of Radar Technology, polarization SAR has become the development trend of SAR, polarization SAR can obtain abundanter target information, there is research and using value widely at aspects such as agricultural, forestry, military affairs, geology, hydrology and oceans, as the identification, crops of species grow up supervision, output assessment, terrain classification, sea ice monitoring, Ground Subsidence Monitoring, target detection and marine pollution detection etc.The object of polarization Images Classification is to utilize polarization measurement data airborne or that borne polarization sensor obtains, determines the classification that each pixel is affiliated.Classical polarization SAR sorting technique comprises:

The people such as Cloude have proposed the Unsupervised classification of polarimetric synthetic aperture radar images decomposing based on H/ α target, see Cloude S R, Pottier E.An entropy based classification scheme for land applications of polarimetric SAR[J] .IEEE Trans.Geosci.Remote Sensing.1997, 35 (1): 549-557. the method is mainly to decompose by cloude the feature of obtaining H and two sign polarization data of α, then according to the H/ α plane of H and α composition artificial be divided into 9 regions, remove a region that can not exist in theory, image is divided into 8 classes the most at last.The division that the defect that H/ alpha taxonomy exists is region is too dogmatic, in the time that of a sort data are distributed on the border of two classes or several classes, classifier performance is by variation, another weak point is in the time coexisting several different atural object in same region, can not effectively distinguish.

The people such as Lee have proposed the H/ α-Wishart not supervised classification based on the decomposition of H/ α target and Wishart sorter, see Lee J S, Grunes M R, Ainsworth T L, et a1.Unsupervised classification using polarimetric decomposition and the complex Wishart classifier[J] .IEEE Trans.Geosci.Remote Sensing.1999, 37 (5): 2249-2258. the method is to have increased Wishart iteration on original H/ alpha taxonomy basis, mainly to utilize Wishart sorter to repartition each pixel to 8 classes after H/ α division, thereby effectively raise the precision of classification, but there is again the deficiency of the polarization scattering characteristics that can not well keep all kinds of.

J.S.Lee etc. decompose in having proposed a kind of multipolarization image unsupervised classification algorithm decomposing based on Freeman-Durden based on Freeman, see Lee J S, Grunes M R, Pottier E, et a1.Unsupervised terrain classification preserving polarimetric scattering characteristic[J] .IEEE Trans.Geosci.Remote Sensing.2004, 42 (4): 722-731. the method be mainly by Freeman decompose obtain characterize scatterer scattering properties three features: in-plane scatter power, dihedral angle scattering power and volume scattering power, then according to the size of these three features, polarization data is divided, and initial division is carried out to categories combination, finally recycling Wishart sorter repartitions each pixel.This algorithm combines Freeman scattering model and multiple Wishart distributes, and has the characteristic of the pure property of main scattering mechanism that keeps multipolarization SAR, but in the method due to division and the merging of the multiclass of Freeman in decomposing, thereby computation complexity is higher.

Summary of the invention

The object of the invention is to overcome the deficiency of prior art, on the basis of the above-mentioned multipolarization image unsupervised classification algorithm decomposing based on Freeman-Durden, a kind of Classification of Polarimetric SAR Image method based on Freeman decomposition and particle group optimizing is proposed, to reduce computation complexity and further to improve classifying quality.

For achieving the above object, the present invention includes following steps:

(1) with Lee filtering, the Polarimetric SAR Image of input is done to pre-service;

(2) filtered data are carried out to Freeman decomposition by the method for introducing above, obtain three kinds of scattering power matrix: P _spresentation surface scattering power matrix wherein, P _drepresent dihedral angle scattering power matrix, P _vrepresent volume scattering power matrix;

(3) according to power matrix P _s, P _d, P _vpolarimetric SAR Image data are carried out to initial division:

3a) according to max (P _s, P _v, P _d) value, be three classes by Polarimetric SAR Image data initial division, meet max (P _s, P _v, P _d)=P _spixel be divided into in-plane scatter class, max (P _s, P _d, P _v)=P _dcorresponding pixel is divided into dihedral angle scattering class, max (P _s, P _d, P _v)=P _vcorresponding pixel is divided into volume scattering class;

3b) each class is applied to the dual threshold Otsu method selected threshold based on particle group optimizing, threshold value choose formula as shown in the formula, in the time that following formula is obtained maximal value, vector (u ₁, v ₁), (u ₂, v ₂) be the optimum thresholding vector of two-dimentional dual threshold Otsu

we have just obtained two threshold values of each class so.Adopting said method does Further Division, and view picture Polarimetric SAR Image is divided into 9 classes;

$\mathrm{tr}\left({\mathrm{\σ}}_B\right)=\frac{{\mathrm{\ω}}_1[{\mathrm{\ω}}_0{\mathrm{\μ}}_{\mathrm{Ti}}-{\mathrm{\μ}}_i{\left(u_{1,}{v}_1\right)]}^2+{\mathrm{\ω}}_0[{\mathrm{\ω}}_1{\mathrm{\μ}}_{\mathrm{Ti}}-{\mathrm{\μ}}_i{\left(u_{2,}{v}_2\right)]}^2+[{\mathrm{\ω}}_1{\mathrm{\μ}}_1\left(u_{1,}{v}_1\right)-{\mathrm{\ω}}_0{\mathrm{\μ}}_i{\left(u_{2,}{v}_2\right)]}^2}{{\mathrm{\ω}}_0\×{\mathrm{\ω}}_1\×(1-{\mathrm{\ω}}_0-{\mathrm{\ω}}_1)}$

$\frac{{\mathrm{\ω}}_1[{\mathrm{\ω}}_0{\mathrm{\μ}}_{\mathrm{Ti}}-{\mathrm{\μ}}_i{\left(u_{1,}{v}_1\right)]}^2+{\mathrm{\ω}}_0[{\mathrm{\ω}}_1{\mathrm{\μ}}_{\mathrm{Ti}}-{\mathrm{\μ}}_i{\left(u_{2,}{v}_2\right)]}^2+[{\mathrm{\ω}}_1{\mathrm{\μ}}_1\left(u_{1,}{v}_1\right)-{\mathrm{\ω}}_0{\mathrm{\μ}}_i{\left(u_{2,}{v}_2\right)]}^2}{{\mathrm{\ω}}_0\×{\mathrm{\ω}}_1\×(1-{\mathrm{\ω}}_0-{\mathrm{\ω}}_1)}$

(4) 9 class division results of whole Polarimetric SAR Image data are carried out to multiple Wishart iteration, obtain classification results more accurately;

(5) use red R, green G, tri-color components of blue B as three primary colours, obtain classification results colouring more accurately according to the principle of three primary colours to step (5), obtain final color classification result figure.

Tool of the present invention has the following advantages:

A) the present invention is decomposed into basis with Freeman, extracts three kinds of scattering power matrix P in Polarimetric SAR Image data _s, P _d, P _vand in conjunction with the two-dimentional dual threshold Otsu method of particle group optimizing, Polarimetric SAR Image data are divided, because the distribution of three kinds of scattering powers and the size of same polarization ratio of different scatterers in Polarimetric SAR Image data exist larger difference, therefore can effectively divide in conjunction with scattering power and same polarization comparison polarization data.

B) in the present invention, Polarimetric SAR Image data are directly divided into 9 classes, have avoided division and the merging of the multiclass in Freeman decomposition, more simple in realization, more rigorous, and reduced computation complexity.

C) because Polarimetric SAR Image data do not have strict atural object classification number, so classification number does not have strict defining in the time of classification, the present invention is divided into 9 classes according to concrete data, can find out that by the result of classifying the present invention is in the division result of some zonules, obviously be better than existing Polarimetric SAR Image Data classification and quote the classification results that more two kinds of classical way H/ α, H/ α-Wishart and traditional F reeman decompose, and region consistance is divided better, the edge after zones of different is divided is also more clear.

Accompanying drawing explanation

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

Fig. 2 is the image of two groups of original polarization SAR data using of the present invention;

Fig. 3 is with the present invention and existing H/ α and the classification simulation result figure of H/ α-Wishart sorting technique to San Francisco Bay data;

Fig. 4 is simulation result figure Flevoland data being divided by the sorting technique that the present invention and existing H/ α-Wishart and traditional F reeman decompose.

Embodiment

With reference to Fig. 1, specific implementation step of the present invention is as follows:

Step 1, carries out Freeman decomposition to a width Polarimetric SAR Image, obtains scattering power matrix P _s, P _d, P _v, wherein, P _spresentation surface scattering power matrix, P _drepresent dihedral angle scattering power matrix, P _vrepresent volume scattering power matrix.

Concrete steps are as follows:

1a) each pixel of described Polarimetric SAR Image is 3 × 3 polarization covariance matrix C that contain 9 elements;

$C=\left[\begin{array}{ccc}\⟨|S_{\mathrm{HH}}{|}^2\⟩& \sqrt{2}\⟨S_{\mathrm{HH}}{S}_{\mathrm{HV}}^{*}\⟩& \⟨S_{\mathrm{HH}}{S}_{\mathrm{VV}}^{*}\⟩\\ \sqrt{2}\⟨S_{\mathrm{HV}}{S}_{\mathrm{HH}}^{*}\⟩& 2\⟨|S_{\mathrm{HV}}{|}^2\⟩& \sqrt{2}\⟨S_{\mathrm{HV}}{S}_{\mathrm{VV}}^{*}\⟩\\ \⟨S_{\mathrm{VV}}{S}_{\mathrm{HH}}^{*}\⟩& \sqrt{2}\⟨S_{\mathrm{VV}}{S}_{\mathrm{HV}}^{*}\⟩& \⟨|S_{\mathrm{VV}}{|}^2\⟩\end{array}\right]---1)$

Wherein, H represents horizontal polarization, and V represents vertical polarization, S _hHexpression level to transmitting and level to reception echo data, S _vVrepresent vertically to transmitting and the vertical echo data to receiving, S _hVexpression level to transmitting vertically to the echo data receiving, ^*represent the conjugation of these data, < > represents by looking number average, || represent to get the mould value of this number; 1b) covariance matrix C is resolved into following expression:

C＝＜C＞ _s+＜C＞ _d+＜C＞ _v 2）

Wherein, $\&lang;C{\&rang;}_s=f_s\left[\begin{array}{ccc}|\mathrm{\&beta;}{|}^2& 0& \mathrm{\&beta;}\\ 0& 0& 0\\ {\mathrm{\&beta;}}^{*}& 0& 1\end{array}\right],\&lang;C{\&rang;}_d=f_d\left[\begin{array}{ccc}|\mathrm{\&alpha;}{|}^2& 0& \mathrm{\&alpha;}\\ 0& 0& 0\\ {\mathrm{\&alpha;}}^{*}& 0& 1\end{array}\right],\&lang;C{\&rang;}_V={\mathrm{<figure-callout\; id="1"\; label="f\; v"\; filenames="HDA0000475185500000011.png"\; state="\{\{state\}\}">f</figure-callout>}}_v\left[\begin{array}{ccc}1& 0& 1/3\\ 0& 2/3& 0\\ 1/3& 0& 1\end{array}\right],$

< C > _sfor the covariance matrix of surface scattering component, < C > _dfor the covariance matrix of dihedral angle scattering component, ' < C > _vfor the covariance matrix of volume scattering component, f _sfor the coefficient of dissociation of in-plane scatter component, f _dfor the coefficient of dissociation of dihedral angle scattering component, f _vfor the coefficient of dissociation of volume scattering component, β is that horizontal emission level receives the back scattering reflection coefficient ratio that receives back scattering emission ratio vertical with Vertical Launch, and α is defined as α=R _ghr _vh/ R _gvr _vv, R _ghand R _gvrepresent respectively level and the vertical reflection coefficient on earth's surface, R _vhand R _vvrepresent level and the vertical reflection coefficient of vertical body of wall, ^*represent the conjugation of these data, < > represents by looking number average, || represent to get the mould value of this number, therefore C can be expressed as again:

C＝＜C＞ _s+＜C＞ _d+＜C＞ _v

$=f_s\left[\begin{array}{ccc}|\mathrm{\&beta;}{|}^2& 0& \mathrm{\&beta;}\\ 0& 0& 0\\ {\mathrm{\&beta;}}^{*}& 0& 1\end{array}\right]+f_d\left[\begin{array}{ccc}|a{|}^2& 0& \mathrm{\&alpha;}\\ 0& 0& 0\\ {\mathrm{\&alpha;}}^{*}& 0& 1\end{array}\right]+f_v\left[\begin{array}{ccc}1& 0& 1/3\\ 0& 2/3& 0\\ 1/3& 0& 1\end{array}\right]---3)$

1c) by formula 2) in entry of a matrix element and formula 1) element of middle covariance matrix C is corresponding, obtain one and there are five unknown number f _s, f _v, f _d, α, the system of equations of β and four equations is as follows:

$\left\{\begin{array}{c}\&lang;|S_{\mathrm{HH}}{|}^2\&rang;=f_s|\mathrm{\&beta;}{|}^2+f_d|\mathrm{\&alpha;}{|}^2+f_v\\ \&lang;|S_{\mathrm{VV}}{|}^2\&rang;=f_s+f_d+f_v\\ \&lang;S_{\mathrm{HH}}{S}_{\mathrm{VV}}^{*}\&rang;=f_s\mathrm{\&beta;}+f_d\mathrm{\&alpha;}+f_v/3\\ \&lang;|S_{\mathrm{HV}}{|}^2\&rang;=f_v/3\end{array}\right.---4)$

1d) in calculating pixel point covariance matrix C

value and judge positive and negative, if

make α=-1, if

make β=1, after the value of given α or β, remaining 4 unknown numbers are according to formula 4) solve, wherein Re () represents to get real part;

1e) according to the f solving _s, f _v, f _d, α, β, according to formula 5) and solve scattering power matrix Ps, Pd, Pv.

$P_v=\frac{{8f}_v}{3},P_d=f_d(1+|\mathrm{\&alpha;}{|}^2),P_s(1+|\mathrm{\&beta;}{|}^2)---5)$

Wherein, P _vrepresent volume scattering power matrix, P _drepresent dihedral angle scattering power matrix, P _spresentation surface scattering power matrix.

Document Freeman A and Durden S.A three-component scattering model for polarimetric SAR data.IEEE Transactions on Geoscience and Remote Sensing1998,36 (3): 963-973 are shown in Freeman decomposition].

Step 2, according to power matrix P _s, P _d, P _vdescribed Polarimetric SAR Image data are carried out to initial division.

Detailed process is as follows:

2a) according to max (P _s, P _d, P _v) value, be three classes by Polarimetric SAR Image data initial division, by max (P _s, P _d, P _v)=P _scorresponding pixel points be divided into in-plane scatter class, by max (P _s, P _d, P _v)=P _dcorresponding pixel points be divided into dihedral angle scattering class, by max (P _s, P _d, P _v)=P _vcorresponding pixel points is divided into volume scattering class;

2b) to step 2a) each class of obtaining of classification apply two-dimentional dual threshold maximum between-cluster variance (Otsu) the method selected threshold based on particle group optimizing, threshold value choose formula as shown in the formula, in the time that following formula is obtained maximal value, vectorial (u ₁, v ₁), (u ₂, v ₂) be the optimum thresholding vector of two-dimentional dual threshold Otsu obtain two threshold values of each class.Adopting said method does Further Division, and each class is divided into 3 classes, and view picture Polarimetric SAR Image is divided into 9 classes;

$\mathrm{tr}\left({\mathrm{\&sigma;}}_B\right)=\frac{{\mathrm{\&omega;}}_1[{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{1,}{v}_1\right)]}^2+{\mathrm{\&omega;}}_0[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2+[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_1\left(u_{1,}{v}_1\right)-{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2}{{\mathrm{\&omega;}}_0\&times;{\mathrm{\&omega;}}_1\&times;(1-{\mathrm{\&omega;}}_0-{\mathrm{\&omega;}}_1)}$

$\frac{{\mathrm{\&omega;}}_1[{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{1,}{v}_1\right)]}^2+{\mathrm{\&omega;}}_0[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2+[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_1\left(u_{1,}{v}_1\right)-{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2}{{\mathrm{\&omega;}}_0\&times;{\mathrm{\&omega;}}_1\&times;(1-{\mathrm{\&omega;}}_0-{\mathrm{\&omega;}}_1)}$

Wherein, ω ₀=ω ₀(u ₁, v ₁), ω ₁=ω ₁(u ₁, v ₁, u ₂, v ₂), ω ₂=ω ₂(u ₂, v ₂) be the probability that each several part occurs, μ ₀, μ ₁, μ ₂the average of all kinds of gray scale condition probability of background and target, μ _i(u ₁, v ₁), μ _i(u ₂, v ₂), μ _j(u ₁, v ₁), μ _j(u ₂, v ₂), μ _i(L, L), μ _j(L, L), μ _ti, μ _tjit is the average of background and all kinds of gray scales of target.

Step 3, carries out multiple Wishart iteration to 9 class division results of whole Polarimetric SAR Image data, obtains more accurate classification results.

Utilize this alternative manner as follows to 9 class division results of whole Polarimetric SAR Image data being carried out to the step of iteration:

3a) the 9 class division results to whole Polarimetric SAR Image data, ask the cluster centre V of each class according to following formula _i:

$V_i=\frac{\underset{j=1}{\overset{N_i}{\mathrm{\&Sigma;}}}{C}_j}{N_i},i=\mathrm{1,2}...9,j=\mathrm{1,2}...N_j$

Wherein, C _jrepresent to belong to the covariance matrix of j class pixel, N _irepresent the number of the pixel that belongs to i class;

Vi in this step is initial cluster center, and in Wishart iteration, iteration cluster centre all can change each time, until reach the iterations of regulation or other stopping criterion for iteration, V _ijust as final cluster centre;

3b) calculate the distance of each pixel to i class cluster centre according to following formula:

$d(\&lang;C\&rang;,V_i)=\ln \left[V_i\right]+\mathrm{Tr}(V_i^{-1}\&lang;C\&rang;),i=\mathrm{1,2}...9$

Wherein C is the covariance matrix of pixel, and < > represents by looking number average, the determinant of [] representing matrix, and the mark of Tr representing matrix,

represent matrix V _iinvert;

3c) Polarimetric SAR Image data are repartitioned to the distance of i class cluster centre according to each pixel:

If d is (< C >, V _i)≤d (< C >, V _j) i, j=1,2,9, j ≠ i, is divided into i class by this pixel, if d is (< C >, V _i) > d (< C >, V _j) i, j=1,2,9, j ≠ i, is divided into j class by this pixel, wherein d (< C >, V _j) represent the distance of this pixel to j class cluster centre;

3e) repeating step 3a)-3c) until iterations equals given iterations n, wherein n=5, here be the stopping criterion for iteration that the present invention stipulates, for regulation iterations, other end condition can be also that twice of front and back cluster centre floats, reach artificial specialized range, or the variation of classification results is less than a number percent, pixel ownership is stable.

Multiple Wishart iteration is a kind of alternative manner being proposed by foreign scholar for 1994, see document Lee J S, Grunes M R.Classification of multi-look polarimetric SAR imagery based on complex Wishart distribution[J] .Int.J.Remote Sensing.1994,15 (11): 2299-2311.

Step 4, with red R, green G, tri-color components of blue B as three primary colours, paint to Polarimetric SAR Image Data classification result in step 3 according to the principle of three primary colours, by the mark classification of sorted view data, different mark classifications is composed with different colors, obtained final color classification result figure.

Effect of the present invention further illustrates by following experiment simulation.

1. experiment condition and method

Experiment simulation environment is: MATLAB7.0.4, Intel (R) Pentium (R) 1CPU2.4GHz, Window XP Professional.

Experimental technique: be respectively existing H/ α method and H/ α-Wishart method and the present invention, wherein existing these two kinds of methods are all to quote more classical way in polarization SAR Data classification.

2. experiment content and result

Experiment content: the present invention uses respectively two groups of polarization SAR data shown in Fig. 2 to do test experiments, wherein to be first group be San Francisco Bay data to Fig. 2 (a), be four depending on number, Fig. 2 (b) is second group of polarization SAR data, it is the atural object distribution situation in Dutch Flevoland area,, deriving from AIRSAR, image size is 300 × 270.

Experiment one, with existing H/ α and H/ α-Wishart sorting technique and the present invention to the emulation of classifying of San Francisco Bay data, classification results is shown in Fig. 3, wherein, Fig. 3 (a) is H/ α classification result, Fig. 3 (b) is the result of H/ α-Wishart classification, and Fig. 3 (c) is classification results of the present invention.

From Fig. 3 (a), although visible H/ α method is classical, and classification results is very undesirable, a lot of regions all do not distinguish;

From Fig. 3 (b), be obviously better than original H/ α method in conjunction with H/ α-Wishart sorting technique classification results of H/ α method and Wishart sorter, it is more careful that region is divided, and divides unclear but also have compared with multizone;

From 3 (c), classification results of the present invention is from visually seeing better effects if, wherein these the consistance in territorial classification region such as golf course, racecourse, parking lot is significantly better than first two method, and between zones of different, sorted edge is also more level and smooth.

Experiment two, the sorting technique of decomposing with existing H/ α-Wishart and traditional F reeman and the present invention are to the emulation of classifying of Flevoland data, classification results is shown in Fig. 4, wherein, Fig. 4 (a) is H/ α-Wishart classification result, Fig. 4 (b) is the classification results that traditional F reeman decomposes, and Fig. 4 (c) is the result that the present invention classifies.

As can be seen from Figure 4, result of the present invention is significantly better than the result of existing two kinds of classical ways classification, and it is more careful and more accurate that region is divided.

In sum, the sorting technique to polarization SAR data that the present invention proposes, by first data being carried out to Freeman decomposition, three kinds of scattering powers that characterize polarization characteristic are extracted, according to power matrix, polarization SAR data are done to initial division, then the dual threshold Otsu method of each class being applied based on particle group optimizing is being subdivided into three classes, finally 9 class division results of polarization SAR data is carried out to multiple Wishart iteration, has further improved the division result of each class.And inventive concept is fairly simple, computation complexity is relatively little, easily understands and application.

Claims (3)

1. the Classification of Polarimetric SAR Image method based on Freeman decomposition and particle group optimizing, comprises the steps:

(1) with Lee filtering, the Polarimetric SAR Image of input is done to pre-service;

(2) filtered data are carried out to Freeman decomposition by the method for introducing above, obtain three kinds of scattering power matrix: P _spresentation surface scattering power matrix wherein, P _drepresent dihedral angle scattering power matrix, P _vrepresent volume scattering power matrix;

(3) according to power matrix P _s, P _d, P _vpolarimetric SAR Image data are carried out to initial division:

A) according to max (P _s, P _v, P _d) value, be three classes by Polarimetric SAR Image data initial division, meet max (P _s, P _v, P _d)=P _spixel be divided into in-plane scatter class, max (P _s, P _d, P _v)=P _dcorresponding pixel is divided into dihedral angle scattering class, max (P _s, P _d, P _v)=P _vcorresponding pixel is divided into volume scattering class;

B) each class is applied to the dual threshold Otsu method selected threshold based on particle group optimizing, threshold value choose formula as shown in the formula, in the time that following formula is obtained maximal value, vector (u ₁, v ₁), (u ₂, v ₂) be the optimum thresholding vector of two-dimentional dual threshold Otsu

we have just obtained two threshold values of each class so.Adopting said method does Further Division, and view picture Polarimetric SAR Image is divided into 9 classes;

$\mathrm{tr}\left({\mathrm{\&sigma;}}_B\right)=\frac{{\mathrm{\&omega;}}_1[{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{1,}{v}_1\right)]}^2+{\mathrm{\&omega;}}_0[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2+[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_1\left(u_{1,}{v}_1\right)-{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2}{{\mathrm{\&omega;}}_0\&times;{\mathrm{\&omega;}}_1\&times;(1-{\mathrm{\&omega;}}_0-{\mathrm{\&omega;}}_1)}$

$\frac{{\mathrm{\&omega;}}_1[{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{1,}{v}_1\right)]}^2+{\mathrm{\&omega;}}_0[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_{\mathrm{Ti}}-{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2+[{\mathrm{\&omega;}}_1{\mathrm{\&mu;}}_1\left(u_{1,}{v}_1\right)-{\mathrm{\&omega;}}_0{\mathrm{\&mu;}}_i{\left(u_{2,}{v}_2\right)]}^2}{{\mathrm{\&omega;}}_0\&times;{\mathrm{\&omega;}}_1\&times;(1-{\mathrm{\&omega;}}_0-{\mathrm{\&omega;}}_1)}$

(4) 9 class division results of whole Polarimetric SAR Image data are carried out to multiple Wishart iteration, obtain classification results more accurately;

(5) use red R, green G, tri-color components of blue B as three primary colours, obtain classification results colouring more accurately according to the principle of three primary colours to step (5), obtain final color classification result figure.

2. Classification of Polarimetric SAR Image method according to claim 1, what wherein step (2) was described carries out Freeman decomposition to input data, carries out as follows:

2a) each pixel of reading data is 3 × 3 polarization covariance matrix C that contain 9 elements;

$C=\left[\begin{array}{ccc}\&lang;|S_{\mathrm{HH}}{|}^2\&rang;& \sqrt{2}\&lang;S_{\mathrm{HH}}{S}_{\mathrm{HV}}^{*}\&rang;& \&lang;S_{\mathrm{HH}}{S}_{\mathrm{VV}}^{*}\&rang;\\ \sqrt{2}\&lang;S_{\mathrm{HV}}{S}_{\mathrm{HH}}^{*}\&rang;& 2\&lang;|S_{\mathrm{HV}}{|}^2\&rang;& \sqrt{2}\&lang;S_{\mathrm{HV}}{S}_{\mathrm{VV}}^{*}\&rang;\\ \&lang;S_{\mathrm{VV}}{S}_{\mathrm{HH}}^{*}\&rang;& \sqrt{2}\&lang;S_{\mathrm{VV}}{S}_{\mathrm{HV}}^{*}\&rang;& \&lang;|S_{\mathrm{VV}}{|}^2\&rang;\end{array}\right]---\left(1\right)$

Wherein, H represents horizontal polarization, and V represents vertical polarization, S _hHexpression level to transmitting and level to reception echo data, S _vVrepresent vertically to transmitting and the vertical echo data to receiving, S _hVexpression level to transmitting vertically to the echo data receiving, ^*represent the conjugation of these data, represent by looking number average;

2b) covariance matrix C is resolved into following expression:

C＝＜C＞ _s+＜C＞ _d+＜C＞ _v

$=f_s\left[\begin{array}{ccc}|\mathrm{\&beta;}{|}^2& 0& \mathrm{\&beta;}\\ 0& 0& 0\\ {\mathrm{\&beta;}}^{*}& 0& 1\end{array}\right]+f_d\left[\begin{array}{ccc}|a{|}^2& 0& \mathrm{\&alpha;}\\ 0& 0& 0\\ {\mathrm{\&alpha;}}^{*}& 0& 1\end{array}\right]+f_v\left[\begin{array}{ccc}1& 0& 1/3\\ 0& 2/3& 0\\ 1/3& 0& 1\end{array}\right]---\left(2\right)$

Wherein, < C > _sfor the covariance matrix of surface scattering component, < C > _dfor the covariance matrix of dihedral angle scattering component, < C > _vfor the covariance matrix of volume scattering component, f _sfor the coefficient of dissociation of in-plane scatter component, f _dfor the coefficient of dissociation of dihedral angle scattering component, f _vfor the coefficient of dissociation of volume scattering component, β is that horizontal emission level receives the back scattering reflection coefficient ratio that receives back scattering emission ratio vertical with Vertical Launch, and α is defined as α=R _ghr _vhr _gvr _vv, R _ghand R _gvrepresent respectively level and the vertical reflection coefficient on earth's surface, R _vhand R _vvrepresent level and the vertical reflection coefficient of vertical body of wall;

2c) by entry of a matrix element and formula 1 in formula (2)) element of middle covariance matrix C is corresponding, and obtain one and there are five unknown number f _s, f _v, f _d, α, the system of equations of β and four equations is as follows:

$\left\{\begin{array}{c}\&lang;|S_{\mathrm{HH}}{|}^2\&rang;=f_s|\mathrm{\&beta;}{|}^2+f_d|\mathrm{\&alpha;}{|}^2+f_v\\ \&lang;|S_{\mathrm{VV}}{|}^2\&rang;=f_s+f_d+f_v\\ \&lang;S_{\mathrm{HH}}{S}_{\mathrm{VV}}^{*}\&rang;=f_s\mathrm{\&beta;}+f_d\mathrm{\&alpha;}+f_v/3\\ \&lang;|S_{\mathrm{HV}}{|}^2\&rang;=f_v/3\end{array}\right.---\left(3\right)$

2d) the Re in calculating pixel point covariance matrix C value and judge positive and negative, if

make α=-1, if

make β=1, after the value of given α or β, remaining 4 unknown numbers solve according to formula (3), and wherein Re () represents to get real part;

2e) according to the f solving _s, f _v, f _d, α, β, solves scattering power matrix P according to formula (4) _s, P _d, P _v:

$P_v=\frac{{8f}_v}{3},P_d=f_d(1+|\mathrm{\&alpha;}{|}^2),P_s(1+|\mathrm{\&beta;}{|}^2)---\left(4\right)$

Wherein, P _vrepresent volume scattering power matrix, P _drepresent dihedral angle scattering power matrix, P _spresentation surface scattering power matrix.

3. Classification of Polarimetric SAR Image method according to claim 1, wherein the described 9 class division results to whole Polarimetric SAR Image data of step (4) are carried out multiple Wishart iteration, carry out as follows:

3a) the 9 class division results to whole Polarimetric SAR Image data, ask the cluster centre V of each class according to following formula _i:

$V_i=\frac{\underset{j=1}{\overset{N_i}{\mathrm{\&Sigma;}}}{C}_j}{N_i},i=\mathrm{1,2}...9,j=\mathrm{1,2}...N_j$

Wherein, C _jrepresent to belong to the covariance matrix of j class pixel, N _irepresent the number of the pixel that belongs to i class;

3b) calculate the distance of each pixel to i class cluster centre according to following formula:

$d(\&lang;C\&rang;,V_i)=\ln \left[V_i\right]+\mathrm{Tr}(V_i^{-1}\&lang;C\&rang;),i=\mathrm{1,2}...9$

Wherein C is the covariance matrix of pixel, and < > represents by looking number average, the determinant of [] representing matrix, and the mark of Tr representing matrix,

represent matrix V _iinvert;

3c) Polarimetric SAR Image data are repartitioned to the distance of i class cluster centre according to each pixel:

If d is (< C >, V _i)≤d (< C >, V _j) i, j=1,2,9, j ≠ i, is divided into i class by this pixel, if d is (< C >, V _i) > d (< C >, V _j) i, j=1,2,9, j ≠ i, is divided into j class by this pixel, wherein d (< C >, V _j) represent the distance of this pixel to j class cluster centre;

3d) repeating step 3a)-3c) until iterations equals given iterations n, wherein n=5, obtains classification results more accurately.

Images