CN102253377A - Target detection method for polarimetric interferometry synthetic aperture radar on basis of eigenvalue analysis - Google Patents

Abstract

The invention discloses a target detection method for a polarimetric interferometry synthetic aperture radar on the basis of eigenvalue analysis, relating to a target detection method for a polarimetric interferometry synthetic aperture radar so as to solve the problem that a ground object can not be detected and identified through the traditional complete polarization and the single polarization synthetic aperture radars under the background of the stronger natural ground object clutter. The target detection method comprises the following steps of: reading in the data of a polarimetric interferometry synthetic aperture radar image according to an image data format; preprocessing the polarimetric interferometry synthetic aperture radar image; obtaining a simplified polarimetric interferometry matrix by two groups of polarized scattering vectors, and solving the characteristic value of a matrix; simplifying the matrix by similar diagonalization and a Jordan standard form to obtain an optimized scattering vector-based correlation coefficient; analyzing an eigenvalue and the statics characteristic of the correlation coefficient by selecting different samples; and constructing a target detector by the obtained statics characteristic to detect an interested target to obtain a result. The target detection method is used for the target detection for the polarimetric interferometry synthetic aperture radar.

Description

Polarization interference synthetic aperture radar object detection method based on Eigenvalue Analysis

Technical field

The present invention relates to a kind of polarization interference synthetic aperture radar object detection method, belong to the remote sensing technology field.

Background technology

Synthetic-aperture radar as a kind of unique can round-the-clock, the remote sensing means of round-the-clock earth observation imaging, has irreplaceable effect in this field, and polarization interference synthetic aperture radar (PolInSAR) can combine the careful geometric configuration of atural object, structure, sensing and material composition and coherence, elevation information such as (sea level elevations), greatly promoted synthetic-aperture radar and extracted and the ability of analyzing characters of ground object, had broad application prospects in the remote sensing field.

The economic develop rapidly of China at present, urbanization process is constantly accelerated, and researching and analysing of mushroom town and combination area of city and country area better is familiar with development of urbanization trend with the help people, provides effective guidance for formulating further development plan.Therefore utilize synthetic-aperture radar to carry out cities and towns zone terrain analysis and have using value and wide application prospect by force with assessment.Tradition complete polarization and single polarization synthetic-aperture radar can realize classification and the identification to some basic atural objects, but can't solve the detection and the identification of man-made target under the strong natural feature on a map clutter background.

Summary of the invention

The purpose of this invention is to provide a kind of polarization interference synthetic aperture radar object detection method, can't realize the detection of man-made target under the strong natural feature on a map clutter background and the problem of identification to solve traditional complete polarization and single polarization synthetic-aperture radar based on Eigenvalue Analysis.

The present invention addresses the above problem the technical scheme of taking to be:

Polarization interference synthetic aperture radar object detection method based on Eigenvalue Analysis of the present invention, described method is realized by following steps:

Step 1: obtain pending view data by polarization interference synthetic aperture radar system acquisition image, read in the data of polarization interference synthetic aperture radar image according to image data format;

Step 2: image pre-service: the polarization interference synthetic aperture radar image that step 1 is read in carries out filtering, registration pre-service, obtains the Polarization scattering vector;

Step 3: the eigenwert and the Jordan standard form of the Polarization scattering vector computational short cut polarization interference coherence matrix that obtains according to step 2, calculating is based on the coefficient of coherence of Eigenvalue Analysis, three kinds of different atural object eigenwerts in zone, cities and towns and coefficient of coherence statistical property are carried out statistical study, and concrete steps are as follows:

Step 3 A: simplify the calculating of polarization interference coherence matrix:

${\stackrel{\→}{k}}_1=\frac{1}{\sqrt{2}}{[S_{\mathrm{hh}1}+S_{\mathrm{vv}1},S_{\mathrm{hh}1}-S_{\mathrm{vv}1},2S_{\mathrm{hv}1}]}^T$ Formula one

${\stackrel{\→}{k}}_2=\frac{1}{\sqrt{2}}{[S_{\mathrm{hh}2}+S_{\mathrm{vv}2},S_{\mathrm{hh}2}-S_{\mathrm{vv}2},2S_{\mathrm{hv}2}]}^T$

$\left[{\mathrm{\&Omega;}}_{12}\right]=<{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H>$

$=\frac{1}{2}\left[\begin{array}{ccc}<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){S_{\mathrm{hv}2}}^{*}\\ <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){S_{\mathrm{hv}2}}^{*}\\ 2<S_{\mathrm{hv}1}{(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& 2<S_{\mathrm{hv}1}{(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 4<S_{\mathrm{hv}1}{S_{\mathrm{hv}2}}^{*}>\end{array}\right]$

Formula two

Wherein, [Ω ₁₂] for simplifying the polarization interference coherence matrix;

(i=1,2) are Polarization scattering vector under the Pauli base; S _Pq(p, { h v}) is scattering amplitude to q ∈, and expression receives with q polarized state emission, p polarized state after the target of electromagnetic wave gained to multiple scattering coefficient, and h is a horizontal polarization state, and v is the vertical polarization state; The conjugate transpose of superscript H representing matrix; The transposition of superscript T representing matrix; Superscript * represents conjugation;

Step 3 B: the simplification polarization interference coherence matrix that step 3 A is calculated carries out characteristic value decomposition, and similar diagonalization obtains eigenwert and based on the optimization coefficient of coherence of Eigenvalue Analysis;

Step 3 C: choose the three kinds of different ground object sample of typical case in cities and towns: forest, farmland, buildings, with described three kinds of different ground object sample as detecting target, according to eigenwert and optimization coefficient of coherence that step 3 B obtains the eigenwert of described three kinds of different ground object sample and the statistical property of coefficient of coherence are analyzed, and then obtained being used for the result of feature extraction and target detection;

Step 4: the result who is used for feature extraction and target detection that step 3 C is obtained carries out the object detector design, is used for the polarization interference synthetic aperture radar image, obtains the man-made target testing result.

The invention has the beneficial effects as follows: the defective that 1, the invention solves polarimetric synthetic aperture radar man-made target feature extraction difficulty under the strong clutter background, and proposed to simplify the notion of polarization interference coherence matrix, utilize the eigenwert of this matrix and corresponding coefficient of coherence to distinguish the differently statistics difference on the thing feature of zone, cities and towns, and then the design object detecting device is realized the detection and the identification of man-made target under the strong natural feature on a map clutter background.2, the present invention assesses the cost less with respect to other polarization interference synthetic aperture radar analysis of image data methods, and can make full use of the atural object geometric properties, the information that does not comprise in single polarizations such as coherence, the polarimetric synthetic aperture radar image realizes that man-made target interested detects under the strong clutter background.

Description of drawings

Fig. 1 is the amplitude picture of polarization interference synthetic aperture radar image hh (horizontal polarization emission level polarized state receives the SAR image that obtains) passage; Fig. 2 a is a buildings Sample selection synoptic diagram; Fig. 2 b is a farmland Sample selection synoptic diagram; Fig. 2 c is a forest Sample selection synoptic diagram; Fig. 3 is buildings, the farmland distribution plan (among the figure, △ represents buildings, and represents the farmland) on the eigenwert plane; Fig. 4 is three kinds of different ground object sample coefficient of coherence distribution probability density maps (among the figure solid line represent buildings, dotted line represent that forest, dot-and-dash line represent the farmland); Fig. 5 is testing result figure of the present invention.

Embodiment

Embodiment one: the polarization interference synthetic aperture radar object detection method based on Eigenvalue Analysis of present embodiment, described method is realized by following steps:

Step 1: obtain pending view data by polarization interference synthetic aperture radar system acquisition image, read in the data (the hh channel image as shown in Figure 1) of polarization interference synthetic aperture radar image according to image data format;

Step 2: image pre-service: the polarization interference synthetic aperture radar image that step 1 is read in carries out filtering, registration pre-service, obtains the Polarization scattering vector;

Step 3: the eigenwert of the Polarization scattering vector computational short cut polarization interference coherence matrix that obtains according to step 2 and Jordan standard form are (referring to " matrix analysis study course " 68 pages, the Dong Zengfu chief editor, publishing house of Harbin Institute of Technology in April, 2005 second edition), calculating is based on the coefficient of coherence of Eigenvalue Analysis, three kinds of different atural object eigenwerts in zone, cities and towns and coefficient of coherence statistical property are carried out statistical study, and concrete steps are as follows:

Step 3 A: simplify the calculating of polarization interference coherence matrix:

${\stackrel{\&RightArrow;}{k}}_1=\frac{1}{\sqrt{2}}{[S_{\mathrm{hh}1}+S_{\mathrm{vv}1},S_{\mathrm{hh}1}-S_{\mathrm{vv}1},2S_{\mathrm{hv}1}]}^T$ Formula one

${\stackrel{\&RightArrow;}{k}}_2=\frac{1}{\sqrt{2}}{[S_{\mathrm{hh}2}+S_{\mathrm{vv}2},S_{\mathrm{hh}2}-S_{\mathrm{vv}2},2S_{\mathrm{hv}2}]}^T$

$\left[{\mathrm{\&Omega;}}_{12}\right]=<{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H>$

$=\frac{1}{2}\left[\begin{array}{ccc}<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){S_{\mathrm{hv}2}}^{*}\\ <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){S_{\mathrm{hv}2}}^{*}\\ 2<S_{\mathrm{hv}1}{(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& 2<S_{\mathrm{hv}1}{(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 4<S_{\mathrm{hv}1}{S_{\mathrm{hv}2}}^{*}>\end{array}\right]$

Formula two

Wherein, [Ω ₁₂] for simplifying the polarization interference coherence matrix;

(i=1,2) be Pauli (referring to Cloude S.R.and Papathanassiou K.P., Polarimetric SAR interferometry, IEEE Trans.on GRS, 1998,36 (5), 1551-1565) the following Polarization scattering vector of base; S _Pq(P, { h v}) is scattering amplitude to q ∈, and expression receives with q polarized state emission, p polarized state after the target of electromagnetic wave gained to multiple scattering coefficient, and h is a horizontal polarization state, and v is the vertical polarization state; The conjugate transpose of superscript H representing matrix; The transposition of superscript T representing matrix; Superscript * represents conjugation;

Step 3 B: the simplification polarization interference coherence matrix that step 3 A is calculated carries out characteristic value decomposition, and similar diagonalization obtains eigenwert and based on the optimization coefficient of coherence of Eigenvalue Analysis;

Step 3 C: choose the three kinds of different ground object sample of typical case in cities and towns: forest, farmland, buildings, with described three kinds of different ground object sample as detecting target, according to eigenwert and optimization coefficient of coherence that step 3 B obtains the eigenwert of described three kinds of different ground object sample and the statistical property of coefficient of coherence are analyzed, and then obtained being used for the result of feature extraction and target detection;

Step 4: the result who is used for feature extraction and target detection that step 3 C is obtained carries out the object detector design, is used for the polarization interference synthetic aperture radar image, obtains the man-made target testing result.

As shown in Figures 3 and 4, utilize eigenwert and linear restriction condition to remove the farmland clutter, utilize a coefficient of coherence and a constant thresholding to remove the forest clutter, testing result as shown in Figure 5.

Embodiment two: among the step 3 B of present embodiment, describedly carry out characteristic value decomposition to simplifying the polarization interference coherence matrix, similar diagonalization, obtain eigenwert and based on the optimization coefficient of coherence of eigenwert, concrete eigenwert, optimize that coefficient of coherence calculates and the similar diagonalization technical process of matrix is as follows:

The eigenwert of a matrix can comprehensively, accurately characterize the characteristic of matrix, is the important means and the instrument of matrix analysis, therefore finds the solution simplification polarization interference coherence matrix eigenwert and is used for extracting and analyzing based on the characters of ground object of simplifying the polarization interference coherence matrix; Simplifying the polarization interference coherence matrix is 3 * 3 complex matrix, can calculate its three complex eigenvalues and three corresponding multiple eigenvectors;

Can notice that off-diagonal element is the tolerance of different polarization interchannel signal coherency in the formula two, the coherence of obviously different polarization interchannel signals plays the effect of decoherence to the signal of equipolarization state.Therefore, only needing utilization to make off-diagonal element someway is the zero relevant optimization that can realize the equipolarization status signal.Certain this imagination can realize with the method for Eigenvalue Analysis equally;

If simplifying the polarization interference coherence matrix is simple matrix, can carry out similar diagonalization, utilize following formula three～12 to obtain described optimization coefficient of coherence:

[Ω ₁₂]=[E] ^-1[Λ] [E] formula three

$\left[\mathrm{\&Lambda;}\right]=\left[E\right]\left[{\mathrm{\&Omega;}}_{12}\right]{\left[E\right]}^{-1}=\left[E\right]{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H{\left[E\right]}^{-1}=\left(\right[E\left]{\stackrel{\&RightArrow;}{k}}_1\right){\left({\left({\left[E\right]}^{-1}\right)}^H{\stackrel{\&RightArrow;}{k}}_2\right)}^H$ Formula four

${\stackrel{\&RightArrow;}{k}}_1^{\&prime;}=\left[E\right]{\stackrel{\&RightArrow;}{k}}_1,$ ${\stackrel{\&RightArrow;}{k}}_2^{\&prime;}={\left({\left[E\right]}^{-1}\right)}^H{\stackrel{\&RightArrow;}{k}}_2$ Formula five

Wherein: [Ω ₁₂] for simplifying the polarization interference coherence matrix; The conjugate transpose of superscript H representing matrix;

(i=1,2) are Polarization scattering vector under the Pauli base;

(i=1,2) are the Scattering of Vector under the new scattering mechanism; By the proper vector of simplifying the polarization interference coherence matrix

Constitute; [Λ]=diag (λ ₁, λ ₂, λ ₃), by the eigenvalue of simplifying the polarization interference coherence matrix ₁, λ ₂, λ ₃Constitute; [E] ^H[E] ^-1Column vector can be counted as three groups of Polarization scattering mechanism, that is,

${\left[E\right]}^H=[{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{11},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{13}]$ Formula seven

${\left[E\right]}^{-1}=[{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{21},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{22},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{23}]$ Formula eight

So

${\stackrel{\&RightArrow;}{k}}_1^{\&prime;}=\left[E\right]$ ${\stackrel{\&RightArrow;}{k}}_1=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{11}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H\end{array}\right]$ ${\stackrel{\&RightArrow;}{k}}_1=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{11}^H{\stackrel{\&RightArrow;}{k}}_1\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H{\stackrel{\&RightArrow;}{k}}_1\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H{\stackrel{\&RightArrow;}{k}}_1\end{array}\right]=\left[\begin{array}{c}{\mathrm{\&mu;}}_{11}\\ {\mathrm{\&mu;}}_{12}\\ {\mathrm{\&mu;}}_{13}\end{array}\right]$ Formula nine

${\stackrel{\&RightArrow;}{k}}_2^{\&prime;}={\left[{\left[E\right]}^{-1}\right]}^H$ ${\stackrel{\&RightArrow;}{k}}_2=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{21}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{22}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{23}^H\end{array}\right]$ ${\stackrel{\&RightArrow;}{k}}_2=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{21}^H{\stackrel{\&RightArrow;}{k}}_2\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{22}^H{\stackrel{\&RightArrow;}{k}}_2\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{23}^H{\stackrel{\&RightArrow;}{k}}_2\end{array}\right]=\left[\begin{array}{c}{\mathrm{\&mu;}}_{21}\\ {\mathrm{\&mu;}}_{22}\\ {\mathrm{\&mu;}}_{23}\end{array}\right]$ Formula ten

Wherein:

[E] ^H[E] ^-1Column vector

I=1,2; J=1,2,3 are counted as three groups of Polarization scattering mechanism; The conjugate transpose of superscript H representing matrix;

Be Polarization scattering vector under the Pauli base; μ _Ij(i=1,2; J=1,2,3) be the projection value of Scattering of Vector i on the j scattering mechanism, scattering mechanism is by [E] ^H[E] ^-1Column vector determine;

(i=1,2) are the Scattering of Vector under the new scattering mechanism;

Under these three groups of Polarization scattering mechanism, coherences between different polarization channel signals are inhibited and decoherence between the equipolarization channel signal is minimized, the coherence is optimized, utilize following formula 11 and 12 to calculate three and optimize coefficient of coherence, and be used from polarization interference synthetic aperture radar analysis of image data and feature extraction with simplification polarization interference coherence matrix one;

${\stackrel{\&RightArrow;}{k}}_1^{\&prime;}={({\mathrm{\&mu;}}_{11},{\mathrm{\&mu;}}_{12},{\mathrm{\&mu;}}_{13})}^T$ Formula 11

${\stackrel{\&RightArrow;}{k}}_2^{\&prime;}={({\mathrm{\&mu;}}_{21},{\mathrm{\&mu;}}_{22},{\mathrm{\&mu;}}_{23})}^T$

${\widetilde{\mathrm{\&gamma;}}}_i=\frac{<{\mathrm{\&mu;}}_{1i},{\mathrm{\&mu;}}_{2i}^{*}>}{\sqrt{<{\mathrm{\&mu;}}_{1i}{\mathrm{\&mu;}}_{\mathrm{li}}^{*}><{\mathrm{\&mu;}}_{2i}{\mathrm{\&mu;}}_{2i}^{*}>}},$ $0\&le;\left|{\widetilde{\mathrm{\&gamma;}}}_i\right|\&le;1$ Formula 12

Wherein, μ _Ij(i=1,2; J=1,2,3) be the projection value of Scattering of Vector i on the j scattering mechanism, scattering mechanism is by [E] ^H[E] ^-1Column vector determine;

I=1,2,3 is three pairs of optimization coefficient of coherence under the scattering mechanism;

(i=1,2) are the Scattering of Vector under the new scattering mechanism; The transposition of superscript T representing matrix;

If simplify the condition (under a few cases) that the polarization interference coherence matrix does not satisfy simple matrix, utilize the Jordan standard form to come the approximate similarity diagonalization, utilize formula six, four, five and seven～12 to obtain described optimization coefficient of coherence again:

[Ω ₁₂]=[P] ^-1[J] [P] formula six

Wherein: [Ω ₁₂] for simplifying the polarization interference coherence matrix; [J] is [Ω ₁₂] the Jordan standard form, [J] diagonal entry is [Ω ₁₂] eigenwert, [J] off diagonal element except that the minor diagonal adjacent with diagonal line is zero, is similar diagonalizable a kind of approximate; [P] is the similarity transformation matrix.Other method step is identical with embodiment one.

Embodiment three: among the step 3 C of present embodiment, the eigenwert and the optimization coefficient of coherence that obtain according to step 3 B are analyzed as follows the eigenwert of three kinds of selected different ground object sample and the statistical property of coefficient of coherence:

Eigenwert with above-mentioned three kinds of different ground object sample is coordinate constitutive characteristic space or plane, add up the distribution situation of above-mentioned three kinds of different atural object sample points on feature space or plane, analyze the separability of above-mentioned three kinds of samples in feature space or plane, statistics is drawn three coefficient of coherence distribution probability density maps of three kinds of different ground object sample respectively, analyze the difference of three kinds of samples on this coefficient of coherence probability density figure by described three coefficient of coherence distribution probability density maps, utilize above-mentioned analysis means to analyze the difference of described three kinds of different atural objects, and then obtain being used for feature extraction and (promptly add up the distribution situation of eigenwert on characteristic plane and the distribution situation of coefficient of coherence of above-mentioned three kinds of different atural objects with the result of target detection, therefrom analyze the difference of described three kinds of different atural objects, and then be used for feature extraction and target detection).Other method step is identical with embodiment one.

Embodiment four: in the step 4 of present embodiment, the method for designing of described object detector is as follows:

The result who is used for feature extraction and target detection who obtains according to step 3 C is to determine the judgement plane on the feature space that constitutes of coordinate axis in eigenwert; Utilize three coefficient of coherence distribution probability density map setting thresholds of three kinds of different ground object sample, comprehensive above two kinds of methods realize target detection.Other method step is identical with embodiment one, two or three.

Present embodiment is an example with Fig. 3, Fig. 4, the linear judgment condition that is provided with on the characteristic plane that first and second eigenwert constitutes is as shown in Figure 3, Figure 4 removed the farmland, first coefficient of coherence is provided with the constant thresholding filtering forest about 1, realizes the detection of building target.Judgment condition in the practical application and thresholding should be set flexibly according to the statistical property that step 3 C obtains.

Embodiment five: among the step 3 A of present embodiment, the proposition of described simplification polarization interference coherence matrix (formula two) notion; Specifically describe as follows:

Conventional polar interference synthetic aperture radar image coherence's tolerance adopts the polarization interference coherence matrix, promptly

$\left[T_6\right]=<\left[\begin{array}{c}{\stackrel{\&RightArrow;}{k}}_1\\ {\stackrel{\&RightArrow;}{k}}_2\end{array}\right]\left[\begin{array}{cc}{\stackrel{\&RightArrow;}{k}}_1^H& {\stackrel{\&RightArrow;}{k}}_2^H\end{array}\right]>=\left[\begin{array}{cc}{[T}_{11}]& \left[{\mathrm{\&Omega;}}_{12}\right]\\ {\left[{\mathrm{\&Omega;}}_{12}\right]}^H& \left[T_{22}\right]\end{array}\right]$ Formula 13

[T ₆] be one 2 * 2 partitioned matrix (polarization interference coherence matrix), by three 3 * 3 complex matrix [T ₁₁], [T ₂₂] and [Ω ₁₂] constitute, comprised the polarization information and the interference information of two width of cloth interference complex pattern simultaneously; Wherein, [T ₁₁], [T ₂₂] and [Ω ₁₂] be defined as respectively

${[T}_{11}]=<{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_1^H>$

${[T}_{22}]=<{\stackrel{\&RightArrow;}{k}}_2{\stackrel{\&RightArrow;}{k}}_2^H>$ Formula 14

${[\mathrm{\&Omega;}}_{12}]=<{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H>$

Obviously [T ₁₁] and [T ₂₂] only comprise the polarization information of two radars, and interfere irrelevantly, in the polarization interference data analysis a kind of data redundancy, therefore, definition [Ω ₁₂] (Simplified Polarimetric Interferometric Coherency Matrix SPICM) is in order to simplify the polarization interference coherence matrix

${[\mathrm{\&Omega;}}_{12}]=<{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H>$

$=\frac{1}{2}\left[\begin{array}{ccc}<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){S_{\mathrm{hv}2}}^{*}\\ <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){S_{\mathrm{hv}2}}^{*}\\ 2<S_{\mathrm{hv}1}{(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& 2<S_{\mathrm{hv}1}{(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 4<S_{\mathrm{hv}1}{S_{\mathrm{hv}2}}^{*}>\end{array}\right]$

Formula two

Each element of being simplified the polarization interference coherence matrix by formula two as can be seen is the tolerance of coherence between the different polarization of polarization interference synthetic aperture radar two antennas channel signal, comprises the required enough polarization interference information of data analysis.

Claims (4)

1. polarization interference synthetic aperture radar object detection method based on Eigenvalue Analysis, it is characterized in that: described method is realized by following steps:

Step 1: obtain pending view data by polarization interference synthetic aperture radar system acquisition image, read in the data of polarization interference synthetic aperture radar image according to image data format;

Step 2: image pre-service: the polarization interference synthetic aperture radar image that step 1 is read in carries out filtering, registration pre-service, obtains the Polarization scattering vector;

Step 3: the eigenwert and the Jordan standard form of the Polarization scattering vector computational short cut polarization interference coherence matrix that obtains according to step 2, calculating is based on the coefficient of coherence of Eigenvalue Analysis, three kinds of different atural object eigenwerts in zone, cities and towns and coefficient of coherence statistical property are carried out statistical study, and concrete steps are as follows:

Step 3 A: simplify the calculating of polarization interference coherence matrix:

${\stackrel{\&RightArrow;}{k}}_1=\frac{1}{\sqrt{2}}{[S_{\mathrm{hh}1}+S_{\mathrm{vv}1},S_{\mathrm{hh}1}-S_{\mathrm{vv}1},2S_{\mathrm{hv}1}]}^T$ Formula one

${\stackrel{\&RightArrow;}{k}}_2=\frac{1}{\sqrt{2}}{[S_{\mathrm{hh}2}+S_{\mathrm{vv}2},S_{\mathrm{hh}2}-S_{\mathrm{vv}2},2S_{\mathrm{hv}2}]}^T$

$\left[{\mathrm{\&Omega;}}_{12}\right]=<{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H>$

$=\frac{1}{2}\left[\begin{array}{ccc}<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{hh}1}+S_{\mathrm{vv}1}){S_{\mathrm{hv}2}}^{*}\\ <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& <(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 2<(S_{\mathrm{vv}1}-S_{\mathrm{hh}1}){S_{\mathrm{hv}2}}^{*}\\ 2<S_{\mathrm{hv}1}{(S_{\mathrm{hh}2}+S_{\mathrm{vv}2})}^{*}>& 2<S_{\mathrm{hv}1}{(S_{\mathrm{vv}2}-S_{\mathrm{hh}2})}^{*}>& 4<S_{\mathrm{hv}1}{S_{\mathrm{hv}2}}^{*}>\end{array}\right]$

Formula two

Wherein, [Ω ₁₂] for simplifying the polarization interference coherence matrix; (i=1,2) are Polarization scattering vector under the Pauli base; S _Pq(p, { h v}) is scattering amplitude to q ∈, and expression receives with q polarized state emission, p polarized state after the target of electromagnetic wave gained to multiple scattering coefficient, and h is a horizontal polarization state, and v is the vertical polarization state; The conjugate transpose of superscript H representing matrix; The transposition of superscript T representing matrix; Superscript * represents conjugation;

Step 3 B: the simplification polarization interference coherence matrix that step 3 A is calculated carries out characteristic value decomposition, and similar diagonalization obtains eigenwert and based on the optimization coefficient of coherence of Eigenvalue Analysis;

Step 3 C: choose the three kinds of different ground object sample of typical case in cities and towns: forest, farmland, buildings, with described three kinds of different ground object sample as detecting target, according to eigenwert and optimization coefficient of coherence that step 3 B obtains the eigenwert of described three kinds of different ground object sample and the statistical property of coefficient of coherence are analyzed, and then obtained being used for the result of feature extraction and target detection;

Step 4: the result who is used for feature extraction and target detection that step 3 C is obtained carries out the object detector design, is used for the polarization interference synthetic aperture radar image, obtains the man-made target testing result.

2. the polarization interference synthetic aperture radar object detection method based on Eigenvalue Analysis according to claim 1, it is characterized in that: among the step 3 B, describedly carry out characteristic value decomposition to simplifying the polarization interference coherence matrix, similar diagonalization, obtain eigenwert and based on the optimization coefficient of coherence of eigenwert, concrete eigenwert, optimize that coefficient of coherence calculates and the similar diagonalization technical process of matrix is as follows:

If simplifying the polarization interference coherence matrix is simple matrix, can carry out similar diagonalization, utilize following formula three～12 to obtain described optimization coefficient of coherence:

[Ω ₁₂]=[E] ^-1[Λ] [E] formula three

$\left[\mathrm{\&Lambda;}\right]=\left[E\right]\left[{\mathrm{\&Omega;}}_{12}\right]{\left[E\right]}^{-1}=\left[E\right]{\stackrel{\&RightArrow;}{k}}_1{\stackrel{\&RightArrow;}{k}}_2^H{\left[E\right]}^{-1}=\left(\right[E\left]{\stackrel{\&RightArrow;}{k}}_1\right){\left({\left({\left[E\right]}^{-1}\right)}^H{\stackrel{\&RightArrow;}{k}}_2\right)}^H$ Formula four

${\stackrel{\&RightArrow;}{k}}_1^{\&prime;}=\left[E\right]{\stackrel{\&RightArrow;}{k}}_1,$ ${\stackrel{\&RightArrow;}{k}}_2^{\&prime;}={\left({\left[E\right]}^{-1}\right)}^H{\stackrel{\&RightArrow;}{k}}_2$ Formula five

Wherein: [Ω ₁₂] for simplifying the polarization interference coherence matrix; The conjugate transpose of superscript H representing matrix;

(i=1,2) are Polarization scattering vector under the Pauli base;

(i=1,2) are the Scattering of Vector under the new scattering mechanism; By the proper vector of simplifying the polarization interference coherence matrix

Constitute; [Λ]=diag (λ ₁, λ ₂, λ ₃), by the eigenvalue of simplifying the polarization interference coherence matrix ₁, λ ₂, λ ₃Constitute; [E] ^H[E] ^-1Column vector can be counted as three groups of Polarization scattering mechanism, that is,

${\left[E\right]}^H=[{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{11},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{13}]$ Formula seven

${\left[E\right]}^{-1}=[{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{21},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{22},{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{23}]$ Formula eight

So

${\stackrel{\&RightArrow;}{k}}_1^{\&prime;}=\left[E\right]$ ${\stackrel{\&RightArrow;}{k}}_1=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{11}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H\end{array}\right]$ ${\stackrel{\&RightArrow;}{k}}_1=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{11}^H{\stackrel{\&RightArrow;}{k}}_1\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H{\stackrel{\&RightArrow;}{k}}_1\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{12}^H{\stackrel{\&RightArrow;}{k}}_1\end{array}\right]=\left[\begin{array}{c}{\mathrm{\&mu;}}_{11}\\ {\mathrm{\&mu;}}_{12}\\ {\mathrm{\&mu;}}_{13}\end{array}\right]$ Formula nine

${\stackrel{\&RightArrow;}{k}}_2^{\&prime;}={\left[{\left[E\right]}^{-1}\right]}^H$ ${\stackrel{\&RightArrow;}{k}}_2=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{21}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{22}^H\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{23}^H\end{array}\right]$ ${\stackrel{\&RightArrow;}{k}}_2=\left[\begin{array}{c}{\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{21}^H{\stackrel{\&RightArrow;}{k}}_2\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{22}^H{\stackrel{\&RightArrow;}{k}}_2\\ {\stackrel{\&RightArrow;}{\mathrm{\&omega;}}}_{23}^H{\stackrel{\&RightArrow;}{k}}_2\end{array}\right]=\left[\begin{array}{c}{\mathrm{\&mu;}}_{21}\\ {\mathrm{\&mu;}}_{22}\\ {\mathrm{\&mu;}}_{23}\end{array}\right]$ Formula ten

Wherein:

[E] ^H[E] ^-1Column vector

I=1,2; J=1,2,3 are counted as three groups of Polarization scattering mechanism; The conjugate transpose of superscript H representing matrix; (i=1,2) are Polarization scattering vector under the Pauli base; μ _Ij(i=1,2; J=1,2,3) be the projection value of Scattering of Vector i on the j scattering mechanism, scattering mechanism is by [E] ^H[E] ^-1Column vector determine;

(i=1,2) are the Scattering of Vector under the new scattering mechanism;

Under these three groups of Polarization scattering mechanism, coherences between different polarization channel signals are inhibited and decoherence between the equipolarization channel signal is minimized, the coherence is optimized, utilize following formula 11 and 12 to calculate three and optimize coefficient of coherence, and be used from polarization interference synthetic aperture radar analysis of image data and feature extraction with simplification polarization interference coherence matrix one;

${\stackrel{\&RightArrow;}{k}}_1^{\&prime;}={({\mathrm{\&mu;}}_{11},{\mathrm{\&mu;}}_{12},{\mathrm{\&mu;}}_{13})}^T$ Formula 11

${\stackrel{\&RightArrow;}{k}}_2^{\&prime;}={({\mathrm{\&mu;}}_{21},{\mathrm{\&mu;}}_{22},{\mathrm{\&mu;}}_{23})}^T$

${\widetilde{\mathrm{\&gamma;}}}_i=\frac{<{\mathrm{\&mu;}}_{1i},{\mathrm{\&mu;}}_{2i}^{*}>}{\sqrt{<{\mathrm{\&mu;}}_{1i}{\mathrm{\&mu;}}_{\mathrm{li}}^{*}><{\mathrm{\&mu;}}_{2i}{\mathrm{\&mu;}}_{2i}^{*}>}},$ $0\&le;\left|{\widetilde{\mathrm{\&gamma;}}}_i\right|\&le;1$ Formula 12

Wherein, μ _Ij(i=1,2; J=1,2,3) be the projection value of Scattering of Vector i on the j scattering mechanism, scattering mechanism is by [E] ^H[E] ^-1Column vector determine;

I=1,2,3 is three pairs of optimization coefficient of coherence under the scattering mechanism;

(i=1,2) are the Scattering of Vector under the new scattering mechanism; The transposition of superscript T representing matrix;

If simplify the condition that the polarization interference coherence matrix does not satisfy simple matrix, utilize the Jordan standard form to come the approximate similarity diagonalization, utilize formula six, four, five and seven～12 to obtain described optimization coefficient of coherence again:

[Ω ₁₂]=[P] ^-1[J] [P] formula six

Wherein: [Ω ₁₂] for simplifying the polarization interference coherence matrix; [J] is [Ω ₁₂] the Jordan standard form, [J] diagonal entry is [Ω ₁₂] eigenwert, [J] off diagonal element except that the minor diagonal adjacent with diagonal line is zero, is similar diagonalizable a kind of approximate; [P] is the similarity transformation matrix.

3. the polarization interference synthetic aperture radar object detection method based on Eigenvalue Analysis according to claim 1, it is characterized in that: among the step 3 C, the eigenwert and the optimization coefficient of coherence that obtain according to step 3 B are analyzed as follows the eigenwert of three kinds of selected different ground object sample and the statistical property of coefficient of coherence:

Eigenwert with above-mentioned three kinds of different ground object sample is coordinate constitutive characteristic space or plane, add up the distribution situation of above-mentioned three kinds of different atural object sample points on feature space or plane, analyze the separability of above-mentioned three kinds of samples in feature space or plane, statistics is drawn three coefficient of coherence distribution probability density maps of three kinds of different ground object sample respectively, analyze the difference of three kinds of samples on this coefficient of coherence probability density figure by described three coefficient of coherence distribution probability density maps, utilize above-mentioned analysis means to analyze the difference of described three kinds of different atural objects, and then obtain being used for the result of feature extraction and target detection.Other method step is identical with embodiment one.

4. according to claim 1,2 or 3 described polarization interference synthetic aperture radar object detection methods based on Eigenvalue Analysis, it is characterized in that: in the step 4, the method for designing of described object detector is as follows:

The result who is used for feature extraction and target detection who obtains according to step 3 C is to determine the judgement plane on the feature space that constitutes of coordinate axis in eigenwert; Utilize three coefficient of coherence distribution probability density map setting thresholds of three kinds of different ground object sample, comprehensive above two kinds of methods realize target detection.

Images