CN1299123C - Parameter estimation method for modelling noise Doppler of airborne radar - Google Patents

Abstract

The present invention discloses a parameter estimation method for the modeling clutter Doppler of an airborne radar. A clutter model provided by the parameter estimation method for the modeling clutter Doppler of an airborne radar is the Doppler distribution clutter model (DDC), and a clutter covariance matrix described by the model is parsed and expressed by the four undetermined parameters of a Doppler center f<c>, a Doppler spectrum width spreading coefficient p<f>, a clutter roughness parameter sigma<c>2 and a noise roughness parameter sigma<v>2. The parameter estimation method for the modeling clutter Doppler of an airborne radar comprises the following steps: (1) the airborne radar transmits a pulse signal and collects echo data; (2) clutter data in a collection signal forms an interference covariance matrix of a sample; (3) the joint estimation of an unknown parameter in the DDC clutter model is carried out according to the interference covariance matrix of the sample. The DDC model of the parameter estimation method for the modeling clutter Doppler of an airborne radar simultaneously delineates the changes of the clutter Doppler center of the airborne radar and spectrum width, and the parameter estimation method for the modeling clutter Doppler of an airborne radar can break through Rayleigh limit and overcome the error propagation effect of the estimation of order. Therefore, the performance of the parameter estimation method for the modeling clutter Doppler of an airborne radar is superior to the performance of the existing parameter estimation method of the clutter Doppler of the airborne radar.

Description

A kind of modelling clutter Doppler parameter method of estimation of airborne radar

Technical field

The present invention relates to radar, more particularly, the present invention relates to a kind of clutter Doppler parameter method of estimation of airborne radar.

Background technology

Airborne radar has good characteristics such as round-the-clock, round-the-clock, penetrability, has a wide range of applications in civilian and military fields such as navigation, mapping, scouting, warning, fire control.Yet, there are many pendent technical barriers in the airborne radar signal Processing, and many problems wherein can be summed up as under the motion platform condition, the estimation problem of clutter Doppler parameter (as Doppler center, Doppler's spectrum width and doppler frequency rate etc.).

For example, in the moving object detection of airborne radar, received signal generally includes echo signal, noise signal and noise signal.Owing to look pattern under airborne radar is in usually, land clutter distributes extensively, intensity is big, and especially in city and area, mountain area, noise intensity can reach 60～90dB; On the other hand, platform motion causes clutter Doppler spectrum width f _BExpansion causes target to be submerged in the clutter greatly, and the Detection of Radar Target ability is seriously influenced (referring to document [1]: D.C.Schleher, " MTI and pulsed Doppler radar ", Artech House Inc., London, 1991).Therefore airborne radar need (comprise estimating Doppler spectrum width f by the distribution of estimating clutter spectrum adaptively _B, Doppler center f _cEtc. parameter) to distinguish and clutter reduction, reach the purpose that improves target detection performance.

Again for example, in the high-resolution imaging of airborne synthetic aperture radar (SAR), need accurately measure the attitude of platform and kinematic parameter with the motion compensation that is embodied as picture (referring to document [2]: J.C.Curlander, R.N.McDonough.Synthetic Aperture Radar:System ﹠amp; Signal processing, John Wiley ﹠amp; Sons, NewYork, 1991).And present measuring equipment (as inertial navigation etc.) because the restriction of hardware can not be satisfied the requirement of measuring accuracy.Therefore, need to extract relevant Doppler parameter (as Doppler center f by analyzing the data (being the clutter data) of actual acquisition _cWith doppler frequency rate etc.), with the accurate estimation and the inverting of implementation platform attitude and kinematic parameter, finally obtain satisfied imaging effect.

In the prior art, fast Fourier transform (FFT) is normally adopted in the estimation of the Doppler parameter of clutter, clutter power spectrum is carried out modeling and Doppler parameter extracts (referring to document [1]) at frequency domain.Yet the spectral resolution of these class methods is subjected to the restriction of Rayleigh limit, that is: resolution is greater than the relevant inverse of (CPI) at interval of handling.Prescribe a time limit when impulse sampling has, this restriction will inevitably influence accuracy of parameter estimation.In addition, estimation of the order is generally adopted in its estimation to Doppler parameter, as it to f _BWith the estimation of doppler frequency rate be to be based upon f _cOn the basis of estimating, so f _cThe error of parameter estimation is delivered in the follow-up parameter estimation inevitably.

In a word, to the core link that the accurate modeling of clutter and parameter estimation are used as the airborne radar various engineering, caused and paid attention to widely and study.Because therefore the deficiency of classic method, just needs a kind of new more high-precision doppler method for parameter estimation.

Summary of the invention

The objective of the invention is to overcome the deficiency of Doppler parameter method of estimation in the prior art, by a kind of statistical property of the Clutter Model of clutter priori understanding accurately being described clutter that comprises is provided, thereby the modeled clutter Doppler parameter of a kind of airborne radar method of estimation is provided, realizes accurate estimation the Doppler parameter of clutter.

In order to realize the foregoing invention purpose, the modelling clutter Doppler parameter method of estimation of a kind of airborne radar provided by the invention comprises the steps:

(1) set up Clutter Model in the signal processing system of airborne radar, described Clutter Model is meant clutter and interference of noise covariance matrix:

${\left[R\right]}_{m,n=1}^M={{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">c</figure-callout>}}}^2{\&Integral;}_{f_{\min }}^{f_{\max }}{B}^2{(f,{\mathrm{\&alpha;}}_l)\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">e</figure-callout>}}^j\mathrm{df}+{{\mathrm{\&sigma;}}_v}^2{\mathrm{\&delta;}}_{\mathrm{mn}};$

Wherein, [R] _{M, n}Be interference covariance matrix the (m, n) element, M are relevant pulse numbers in handling at interval, σ _c ²Be the clutter roughness parameter, σ _v ²Be the noise bounce parameter, δ _MnBe the Delta function, f is a Doppler frequency, α _lBe the angular altitude of clutter ring, Δ is the radar pulse recurrence interval, f _Max=2V/ λ, f _Min=-2V/ λ, V are the speed of carrier aircraft, and λ is the wavelength of radar carrier frequency; B ²(f, α _l) be the bidirectional power directional diagram of radar beam, comprise Doppler center f _cWith Doppler's spectrum width spreading coefficient ρ _fTwo parameters undetermined.

This Clutter Model can be described as the distributed clutter of Doppler (DDC) model.

Work as B ²(f, α _l) when being Gaussian bidirectional power directional diagram, described Clutter Model (DDC) can be reduced to:

${\left[R\right]}_{m,n=1}^M={{\mathrm{\&sigma;}}_c}^2{\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">e</figure-callout>}}^j{e}^{(-\frac{{\left(2\mathrm{\&pi;}(m-n){\mathrm{\&rho;}}_f\mathrm{\&Delta;}\right)}^2}{2})}+{{\mathrm{\&sigma;}}_v}^2{\mathrm{\&delta;}}_{\mathrm{mn}}.$

Described Clutter Model comprises Doppler center f _c, Doppler's spectrum width spreading coefficient ρ _f, clutter roughness parameter σ _c ²With noise bounce parameter σ _v ²Four undetermined parameters.Doppler's spectrum width spreading coefficient ρ _fWith clutter Doppler spectrum width f _BHave simple linear relationship, concrete form is relevant with the shape of wave beam.Work as B ²(f, α _l) when being Gaussian bidirectional power directional diagram, have ${\mathrm{\&rho;}}_f=\frac{{2f}_B}{2.355}$ Set up, therefore ρ by estimating _fCan directly obtain f _BEstimation.

(2) airborne radar by transmitter and antenna system to search coverage transponder pulse signal;

(3) airborne radar is by the backscatter signal of antenna system and receiver reception search coverage, and described backscatter signal comprises target echo, noise signal and system noise;

(4) airborne radar is sent into signal processing system with received signal after mixing and A/D conversion, and signal processing system constitutes digitized received signal the interference covariance matrix of sample;

(5) signal processing system is united estimation unknown parameter [f by the interference covariance matrix of sample and the analytical expression of the interference covariance matrix described in the step (1) _c, ρ _f, σ _c ², σ _v ²]

Wherein, estimate in the step (5) that the method that unknown parameter adopts is a maximum likelihood method, described maximum likelihood method is newton's search procedure.Estimate in the step (5) that the method that unknown parameter adopts is pseudo-Subspace Decomposition method, described pseudo-Subspace Decomposition method comprises distribution signal parameter estimation (DSPE), spread signal parametrization estimation (DISPARE), adds weight subspace coupling (WSF).The method of estimating the unknown parameter employing in the step (5) is the approximation method based on covariance matrix, and described approximation method based on covariance matrix comprises the invariable rotary signal parameter estimation technique (S-ESPRIT) of expansion, the general invariable rotary signal parameter estimation technique (G-ESPRIT).

The invention has the advantages that:

1) the DDC model can be delineated the variation of clutter doppler spectral center and spectrum width simultaneously, is particularly suitable for describing clutter dopplerbroadening and the spectrum off-centring that the airborne radar platform motion causes.

2) existing Clutter Model lays particular emphasis on the spectral property of describing in the frequency domain, and the DDC model is then described the covariance matrix character in the clutter time domain, therefore can directly obtain sample covariance matrix by sample, thereby avoid the transform domain processing.

3) estimate to adopt usually the time domain combined method of estimation of analyzing based on covariance matrix based on the Doppler parameter of DDC model, can break through Rayleigh limit, so performance is better than traditional estimation of the order method.

4), therefore can under the less situation of sample number, obtain the parameter estimation of degree of precision because the DDC model itself has comprised the priori to clutter statistical characteristics.

Description of drawings

Fig. 1 is the structural representation of airborne radar;

Fig. 2 is the work synoptic diagram of single channel onboard radar system;

Fig. 3 is the process flow diagram of the modelling clutter Doppler parameter method of estimation of airborne radar of the present invention;

Fig. 4 (a) is that the estimated performance of doppler angle frequency Center Parameter under different CNR conditions compares;

Fig. 4 (b) is that the estimated performance of doppler angle frequency expansion parameter under different CNR conditions compares;

Fig. 5 (a) is that the estimated performance of doppler angle frequency Center Parameter under the expansion condition of different angles compares;

Fig. 5 (b) is that the estimated performance of doppler angle frequency expansion parameter under the expansion condition of different angles compares;

Fig. 6 (a) is that the estimated performance of doppler angle frequency Center Parameter under the different pulse number condition compares;

Fig. 6 (b) is that the estimated performance of doppler angle frequency expansion parameter under the different pulse number condition compares;

Fig. 7 is the imaging results of airborne radar measured data.

Embodiment

Below in conjunction with the drawings and specific embodiments the present invention is described in further detail.

Fig. 1 shows the structural representation of conventional airborne radar, mainly partly is made up of receive-transmit system 1, signal processing system 2 and terminal display system etc.Receive-transmit system is made up of transmitter, receiver and antenna system again.Transmitter is transmitted into the pulse signal modulation of particular form in the space by antenna system to radio-frequency carrier.After the target and clutter reflections in search coverage of transmitting, to scattered signal by airborne radar antenna system received thereafter, obtain intermediate-freuqncy signal by receiver after with the radiofrequency signal mixing, this intermediate-freuqncy signal through transforming to the signal that is fit to collection after the multistage mixing, is delivered in the follow-up signal processing system more then.

The A/D converter of signal processing system is transformed to digital signal with the received signal of simulation, by being handled by polylith DSP disposable plates digital signal, realizes multiple functions such as target detection, parameter estimation, imaging identification again.

The display of terminal display system by secondary treating (data processing), various ways, man-machine interface etc. are dynamically, alternately, show result intuitively.

Fig. 2 shows the work synoptic diagram of airborne radar, and airborne radar adopts pulse-Doppler's system usually, promptly realizes detection or imaging to target by the pulse signal that constantly transmits and receives coherent.Wherein, the Doppler signal of airborne radar clutter is all multifactor relevant with position, platform motion speed and the radar beam etc. of noise source. is the position angle herein; The clutter unit of u () expression  correspondence, V is a carrier aircraft speed; f _pAnd Δ=1/f _pRepresent radar repetition frequency (PRF) and radar recurrence interval (PRT) respectively; R _lAnd α _lBe respectively the radar slant-range and the angular altitude of l rang ring correspondence; F (, α _l) then be by radar antenna bi-directional voltage directional diagram.

In order to estimate the Doppler parameter of noise signal that airborne radar is gathered, can adopt the modelling clutter Doppler parameter method of estimation of airborne radar provided by the invention, the process flow diagram of this method is as shown in Figure 3.The basis of method of the present invention provides a kind of doppler distributed clutter (ddc) model (DDC).Therefore, at first provide the DDC model of airborne radar here by theoretical derivation.Before derivation DDC model, provide following 2 prerequisites hypothesis:

1) in the CPI (the relevant processing at interval), the relative geometrical relation of radar and noise source is constant, that is: carrier aircraft displacement is much smaller than the oblique distance between radar and the clutter;

2) carrier aircraft is made linear uniform motion;

These two hypothesis are set up in the ordinary course of things, also can make it satisfied better by shortening CPI in actual applications.Based on above-mentioned hypothesis, m impulse sampling of the undesired signal that radar receives (clutter plus noise) can be expressed with following integrated form

Wherein, τ _m=(m-1) Δ, s _(t) represent the scattered signal of clutter unit u (), and to can be considered be stochastic process independently; The noise samples of m pulse of expression, M is a relevant pulse number of handling in the interval, they are the white Gaussian noise on the time domain normally, its simple crosscorrelation E[v _m(t) v _n(t)]=σ _v ²δ _MnMake normalized doppler angle frequency

With its substitution (1) Shi Kede

$={\&Integral;}_{f_{\min }}^{f_{\min }}B(f,{\mathrm{\&alpha;}}_l)s_f(t+{\mathrm{\&tau;}}_m)\mathrm{df}+v_m\left(t\right)$

$={\&Integral;}_{f_{\min }}^{f_{\min }}B(f,{\mathrm{\&alpha;}}_l)s_f\left(t\right){\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">e</figure-callout>}}^j\mathrm{df}+v_m\left(t\right)---\left(3\right)$

Herein, s _f(t) the time dependent scattered signal of representing the inside fluctuation of the scattering point among the clutter unit u () to cause.When clutter on the orientation evenly, and its internal motion has s can ignore the time _f(t)=s _f, that is: the scattering function of clutter unit deteriorates to the stochastic variable of independent same distribution (i.i.d), and $E\left(s_f{s}_{f\&prime;}\right)={{\widetilde{\mathrm{\&sigma;}}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">c</figure-callout>}}}^2{\mathrm{\&delta;}}_{{\mathrm{ff}}^{\&prime;}}.$ So far, the sample vector of interference is shown below

$x={\&Integral;}_{f_{\min }}^{f_{\max }}B(f,{\mathrm{\&alpha;}}_l)s_{f}a\left(f\right)\mathrm{df}+v---\left(4\right)$

X=[x herein ₁(t), x ₂(t) ..., x _M(t)] ^T, v=[v ₁(t) ..., vM ₍T)] ^T, a (f)=[1, e ^{J2 π f Δ}..., e ^{J2 π f (M-1) Δ}] ^TInterference covariance matrix R=E (xx then ^H) (m, n) element can be expressed as

${\left[R\right]}_{m,n=1}^M={{\widetilde{\mathrm{\&sigma;}}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">c</figure-callout>}}}^2{\&Integral;}_{f_{\min }}^{f_{\max }}{B}^2(f,{\mathrm{\&alpha;}}_l){\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">e</figure-callout>}}^j\mathrm{df}+{{\mathrm{\&sigma;}}_v}^2{\mathrm{\&delta;}}_{\mathrm{mn}}---\left(5\right)$

Wherein, following formula is the DDC model of the broad sense of airborne radar clutter, and " broad sense " here is meant that the formula that embodies of the covariance element that following formula provides can be with B ²(f, α _l) difference concrete expression is arranged.Provide B below ²(f, α _l) special case when being Gaussian.Suppose that airborne radar has the bidirectional power directional diagram of Gaussian, promptly

(wherein  refers to the position angle,  _cBe the bearing sense of wave beam, ρ _Be the spreading coefficient of wave beam, α _lBe the angular altitude of rang ring correspondence), this normally satisfies in the single channel airborne radar of reality.Can prove and work as ρ _→ 0 and  _c→ pi/2 has following formula to set up

$B^2(f,{\mathrm{\&alpha;}}_l)=\frac{\mathrm{<figure-callout\; id="2"\; label="d"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">d</figure-callout>}}{\sqrt{2\mathrm{\&pi;}}{\mathrm{\&rho;}}_f}\exp (-\frac{{(f-f_c)}^2}{{{2\mathrm{\&rho;}}_f}^2})---\left(6\right)$

Herein D is a constant of being determined by known parameters.(6) formula explanation: under the situation of narrow beam, little stravismus, the clutter doppler spectral with airborne radar of Gaussian directional diagram also is a Gaussian.

With (6) formula substitution (5) formula, utilize the character of Fourier transform, have

${\left[R\right]}_{m,n=1}^M={{\mathrm{\&sigma;}}_c}^2{\mathrm{<figure-callout\; id="2"\; label="e\; j"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">e</figure-callout>}}^j{e}^{(-\frac{{\left(2\mathrm{\&pi;}(m-n){\mathrm{\&rho;}}_f\mathrm{\&Delta;}\right)}^2}{2})}+{{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="v"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">v</figure-callout>}}}^2{\mathrm{\&delta;}}_{\mathrm{mn}}---\left(7\right)$

Wherein ${{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">c</figure-callout>}}}^2=d{{\widetilde{\mathrm{\&sigma;}}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">c</figure-callout>}}}^2.$ So far, obtain airborne radar clutter Gaussian DDC model shown in (7) formula.

Refer again to Fig. 3, can carry out the estimation of clutter Doppler parameter, comprise the steps: according to the DDC model that obtains

(1) airborne radar by transmitter and antenna system to search coverage transponder pulse signal;

(2) airborne radar receives backscatter signal by search coverage by antenna system and receiver, and described backscatter signal comprises target echo, noise signal and system noise; Airborne radar obtains an independent identically distributed N interference sample from reflected signal.This N interference sample constitutes vector X=[x (t ₁) ..., x (t _N)].In the practical application, X can constitute with the sample vector near N the unit that comprises the pending range unit.

(3) received signal of airborne radar is sent into signal processing system after mixing and A/D conversion, the received signal after this signal processing system will be changed constitutes the interference covariance matrix of sample, promptly $\hat{R}={\mathrm{XX}}^H/N.$

(4) unite the unknown parameter x=[f that estimates aforementioned resulting Clutter Model according to the interference covariance matrix of sample by the signal processing system of airborne radar _c, ρ _f, σ _c ², σ _v ²].The invention provides based on covariance matrix approach, based on the subspace analysis of covariance matrix, three major types clutter Doppler parameters such as approximate model based on covariance matrix are estimated new method, these three class methods all can realize the estimation of the unknown parameter vector x of DDC model, and they all can obtain the Doppler parameter estimation performance that obviously is better than the traditional frequency domain method.

First kind clutter Doppler parameter method of estimation of the present invention is based on the method that covariance matrix approaches, as maximal possibility estimation (ML), weighted least-squares coupling methods such as (WLS) etc.This instructions with the ML method (referring to document [3]: T.Trump and B.Otterson.Estimation of nominal direction of arrival andangular spread using an array of sensors, Signal Processing, Vol.50, pp57～69, Apr.1996), estimate that the ML method of clutter Doppler parameter needs the following negative log-likelihood function of minimization for example provides its implementation procedure

$l({\mathrm{\&sigma;}}_s^2,{\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="n"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">n</figure-callout>}}^2,f_c,{\mathrm{\&rho;}}_f)=\log \left|R\right|+\mathrm{tr}\left(R^{-1}\hat{R}\right)---\left(8\right)$

Wherein | R| is the determinant of matrix R.The non-linear minimization problem of above-mentioned 4 dimensions can adopt searching method to find the solution, and for example adopts newton-type searching algorithm (referring to document [3]).The present invention provides the covariance matrix of the parameter estimating error vector that the ML method obtains

D= vecR in the formula ₀/  z ^T,

Be the parameter vector of estimating,  is that Kronecker is long-pending.The parameter estimating error that while is obtained by (9) formula, i.e. C _MLThe diagonal angle vector, also provided Cramer-Rao (CRB) boundary of each estimated parameter error.

The present invention's second class clutter Doppler parameter method of estimation is based on the method for the subspace analysis of covariance matrix, as distribution signal parameter estimation (DSPE) (referring to document [4]: S.Valaee, B.Champagne, P.Kabal, " Parametric localization of distributed sources; " IEEE Tans.Signal Processing, vol.43, no.9, pp2144-2153,1995), the spread signal parametrization estimates that (DISPARE) is (referring to document [5]: Y.Meng, P.Stoica, and K.M.Wong, " Estimation of the directions of arrival of spatially dispersedsignals in array processing; " Proc.Inst.Elect.Eng., Radar, Sonar, Navigat., vol.143, pp.1-9, Feb.1996), add weight subspace coupling (WSF) (referring to document [6]: Bengtsson M, Ottersten B, " AGeneralization of Weighted Subspace Fitting to Full-Rank Models; " IEEE Trans.SignalProcessing, vol.49, no.5, the algorithm of pseudo-subspace class such as pp1002-1012,2001).This instructions is that example provides its implementation procedure with the WSF method.Be following signal subspace and noise subspace at first with covariance matrix R svd

$R\left(x\right)=U_s{\left(x\right)\mathrm{\&Sigma;}}_s\left(x\right)U_s^H\left(x\right)+U_n\left(x\right){\mathrm{\&Sigma;}}_n\left(x\right)U_n^H\left(x\right)---\left(10\right)$

And but the identification condition of DDC model parameter estimation is set up, that is: x=x ₀Sufficient and necessary conditions is

span[U _s(x)]＝span[U _s(x ₀)] (11)

Utilize the distributed source parameter estimation of estimation of false signal subspace and pseudo-subspace matching process to be respectively

$\hat{x}=\arg \underset{z}{\min }\left|\right|{\hat{U}}_s-U_s\left(x\right)T\left|\right|---\left(12\right)$

$\hat{x}=\arg \underset{z}{\min }\left|\right|U_n^H{\widehat{\left(x\right)U}}_s\left|\right|---\left(13\right)$

Wherein ‖ ‖ is a matrix norm undetermined.As seen unknown parameter vector x can search for acquisition by the criterion that (12) or (13) formula is determined.But and know that by the identification condition of DDC model parameter estimation when N was enough big, what obtained by (12) formula and (13) formula was consistent Estimation.Consider weighting secondary norm, these two kinds of algorithms are asymptotic equivalences, and there is another weighted norm in each weighted norm in promptly corresponding (12) formula in (13) formula, and the parameter estimation that the two is obtained has identical large sample and distributes, and vice versa.Here the covariance matrix that provides the parameter estimating error vector that the WSF method obtains is

$C_{\mathrm{WSF}}=\frac{1}{2}{\left(\mathrm{Re}\right[D^H[\left(U_s{\mathrm{\&Sigma;}}_s^{-1}{U}_s^H\right)\&CircleTimes;\left(U_n{\mathrm{\&Sigma;}}_n^{-1}{U}_n^H\right)]D\left]\right)}^{-1}---\left(14\right)$

The present invention's the 3rd class clutter Doppler parameter method of estimation is at little angle expansion ρ _fUse the approximation method that proposes based on covariance matrix, comprise that S-ESPRIT (the invariable rotary signal parameter estimation technique of expansion) is (referring to document [7]: Shahbazpanahi S, Valaee S, Bastani M, " Distributed source localization usingESPRIT algorithm; " IEEE Trans.Signal Processing, vol.49, no.10, pp 2169-2178,2001), G-ESPRIT (the general invariable rotary signal parameter estimation technique) is (referring to document [8]: J S Jeong, KSakaguchi, " Generalization of MUSIC Using Extended Array Mode Vector for JointEstimation of Instantaneous DOA and Angular Spread; " IEICE Trans.Commun., E84-B, no.7, low complex degree method such as pp1781-1789,2001).Here, be that example provides its implementation procedure with S-ESPRIT.The expression of the covariance matrix that provides according to (5) formula is as can be known at little angle expansion ρ _fCondition under, only have a spot of big eigenwert in the eigenvalue distribution of covariance matrix.Therefore handle by 2 rank Taylor series expansions and first approximation, can get R near the perturbation component the doppler angle frequency center _c≈ AA ^H/ 2, matrix A=[a (f wherein ₁) a (f ₂)]=[a (f _c-σ _f) a (f _c+ σ _f)].Effective order of signal covariance matrix is 2, and noise signal can be approximated to be 2 Doppler frequencies and is respectively f ₁And f ₂The stack of some signal.Therefore, can by

$f_c=\frac{{\mathrm{<figure-callout\; id="1"\; label="f"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">f</figure-callout>}}_1+{\mathrm{<figure-callout\; id="2"\; label="f"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">f</figure-callout>}}_2}{2}---\left(15\right)$

${\mathrm{\&rho;}}_f=\frac{f_1-{\mathrm{<figure-callout\; id="2"\; label="f"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">f</figure-callout>}}_2}{2}---\left(16\right)$

Obtain f _cAnd ρ _fEstimation.The S-ESPRIT method is utilized R _cOrder be approximately 2 characteristics with R _cCarry out characteristic value decomposition, promptly

$R_c\&ap;U_s{\mathrm{\&Sigma;}}_s{U}_s^H---\left(17\right)$

The matrix U on M * 2 rank wherein _s2 column vectors be R _cThe eigenwert characteristic of correspondence vector of 2 maximums, the diagonal element of diagonal matrix sigma s is R _cThe eigenwert of 2 maximums, be f ₁And f ₂Estimation.

So far, the present invention has obtained the DDC model, reaches three class clutter Doppler parameter New Estimation Method based on the DDC model.In order to verify the validity of new model and new method, the X-band airborne synthetic aperture radar (SAR) with science and technology Huadong Electronic Engineering Inst. of group of China Electronics (ECRIEE) development is a prototype here, and it is as shown in table 1 to design the airborne radar parameter of emulation.In order to improve Doppler's dynamic range that can detect target, this paper brings up to 2000Hz with the PRF of prototype from 700Hz.Setting this radar is positive side-looking mode of operation, i.e. f _cNear fluctuation 0; Umber of pulse M=32 in the CPI.The clutter bandwidth f of the main beam correspondence of clutter _B=2v β/λ ≈ 167.5Hz is by (6) formula parameter ρ of the Gaussian DDC model of this radar clutter correspondence as can be known _f≈ 142.5Hz.The performance of each class methods relatively from two aspects of emulation and measured data below.Below adopt f _cAnd ρ _fAngular frequency θ _cAnd ρ _θIndex as a comparison.

Table 1 onboard radar system parameter list

Systematic parameter	Parameter value
		The flying height orientation is to beam angle carrier wavelength signal bandwidth pulse repetition rate carrier aircraft speed	h＝3000m β＝1.5° λ＝0.03125m B＝70MHz PRF＝700/2000Hz v＝100m/s

1) simulation performance analysis

According to the radar parameter of setting, adopt three kinds of methods of preamble introduction to carry out the Monte Carlo experiment of 500 DDC parameter estimation.Fig. 4 (a) has provided under different miscellaneous noise ratios (CNR) situation with Fig. 4 (b), respectively by ML, WSF, based on three kinds of θ that method obtains of FFT _cAnd ρ _θThe root-mean-square error of estimating (RMSE).Herein $\mathrm{CNR}=10\log ({\mathrm{\&sigma;}}_{\mathrm{<figure-callout\; id="2"\; label="c"\; filenames="C0316005200141.png"\; state="\{\{state\}\}">c</figure-callout>}}^2/{\mathrm{\&sigma;}}_n^2).$ As seen, 1) when CNR was big, the performance of WSF was very near ML; 2) based on the FFT method owing to be subjected to the restriction of Rayleigh limit, to θ _cEstimation have a fixing deviation, promptly it is not the consistent Estimation with CNR.Fig. 5 (a) and Fig. 5 (b) have provided signal to noise ratio (S/N ratio) CNR=35dB, under the spreading number situation of different angles, and the RMSE of the DDC parameter estimation that obtains by above-mentioned three kinds of methods respectively.As seen, the error of three kinds of methods all is extended to direct ratio with the angle.Fig. 6 (a) and Fig. 6 (b) have then provided under the different pulse number situation, three kinds of results that method obtains.As seen, the error of three kinds of methods all is inversely proportional to umber of pulse.Generally speaking, as seen all significantly surpassed traditional method from Fig. 4～Fig. 6 based on FFT based on the ML of DDC model and the performance of WSF method.

2) measured data checking

Fig. 7 is the imaging results of the partial data that the ECRIEE-SAR system gathers Yellow River Channel area in first half of the year calendar year 2001.This image is looked processing through four, and resolution is about 3m * 3m.Its corresponding region is about 3 kilometers (vertically) * 10 kilometer (laterally), and wherein image vertically is the platform heading, laterally is the radar slant-range direction.The type of ground objects that comprises in this imaging scene is abundanter, and the left part of image is the Yellow River, and right part is mountain area and plateau.Zone 1 (area1) and zone 2 (area2) are the corresponding water surface and plateau flat site respectively, will be from the real data in these two zones respectively by WSF method and the existing estimation of real data having been carried out Doppler parameter based on the method for FFT based on the DDC model provided by the invention, table 2 has provided the Doppler parameter estimated result of two kinds of methods, and the numerical value in the table is 100 adjacent statisticses that range unit is obtained by two kinds of methods in zone 1 and the zone 2.As seen from Table 2, unite estimation because the WSF algorithm is a kind of two dimension of super-resolution, the standard deviation of its estimation (STD) is significantly less than frequency domain method.Table 2 illustrates that also model of the present invention and method are correct simultaneously.

The estimated result of table 2 Doppler parameter

Estimated result (Hz) in the zone 1				Estimated result (Hz) in the zone 2
								f dcAverage (FFT)	358.7	f dcAverage (WSF)	358.1	f dcAverage (FFT)	361.2	f dcAverage (WSF)	359.6
f dcStandard deviation (FFT)	5.7	f dcStandard deviation (WSF)	2.7	f dcStandard deviation (FFT)	6.3	f dcStandard deviation (WSF)	3.2
								f BAverage (FFT)	171.0	f BAverage (WSF)	168.6	f BAverage (FFT)	173.6	f BAverage (WSF)	169.5
f BStandard deviation (FFT)	6.6	f BStandard deviation (WSF)	3.3	f BStandard deviation (FFT)	8.7	f BStandard deviation (WSF)	4.1

Claims (8)

1, a kind of modelling clutter Doppler parameter method of estimation of airborne radar comprises the steps:

(1) set up Clutter Model in the signal processing system of airborne radar, described Clutter Model is meant clutter and interference of noise covariance matrix:

${\left[R\right]}_{m,n=1}^M={{\mathrm{\&sigma;}}_c}^2{\&Integral;}_{f_{\min }}^{f_{\max }}{B}^2(f,{\mathrm{\&alpha;}}_l)e^{j2\mathrm{\&pi;f}(m-n)\mathrm{\&Delta;}}\mathrm{df}+{{\mathrm{\&sigma;}}_v}^2{\mathrm{\&delta;}}_{\mathrm{mn}};$

Wherein, [R] _{M, n}Be interference covariance matrix the (m, n) element, M are relevant pulse numbers in handling at interval, σ _c ²Be the clutter roughness parameter, σ _v ²Be the noise bounce parameter, δ _MnBe the Delta function, f is a Doppler frequency, α _lBe the angular altitude of clutter ring, Δ is the radar pulse recurrence interval, f _Max=2V/ λ, f _Min=-2V/ λ, V are the speed of carrier aircraft, and λ is the wavelength of radar carrier frequency; B ²(f, α _l) be the bidirectional power directional diagram of radar beam, comprise Doppler center f _cWith Doppler's spectrum width spreading coefficient ρ _fTwo parameters undetermined;

Described Clutter Model comprises Doppler center f _c, Doppler's spectrum width spreading coefficient ρ _f, clutter roughness parameter σ _c ²With noise bounce parameter σ _v ²Four undetermined parameters;

(2) airborne radar by transmitter and antenna system to search coverage transponder pulse signal;

(3) airborne radar receives backscatter signal by search coverage by antenna system and receiver, and described backscatter signal comprises target echo, noise signal and system noise;

(4) airborne radar is sent into signal processing system with received signal after mixing and A/D conversion, and signal processing system constitutes digitized received signal the interference covariance matrix of sample;

(5) signal processing system is united estimation unknown parameter [f by the interference covariance matrix of sample and the analytical expression of the interference covariance matrix described in the step (1) _c, ρ _f, σ _c ², σ _v ²].

2, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 1 is characterized in that, works as B ²(f, α _l) when being Gaussian bidirectional power directional diagram, described interference covariance matrix is reduced to:

${\left[R\right]}_{m,n=1}^M={\mathrm{\&sigma;}}_c^2{e}^{j2\mathrm{\&pi;}{f}_c\mathrm{\&Delta;}}{e}^{(-\frac{{\left(2\mathrm{\&pi;}(m-n){\mathrm{\&rho;}}_f\mathrm{\&Delta;}\right)}^2}{2})}+{{\mathrm{\&sigma;}}_v}^2{\mathrm{\&delta;}}_{\mathrm{mn}}.$

3, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 1 is characterized in that, estimates in the step (5) that the method that unknown parameter adopts is a maximum likelihood method.

4, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 3 is characterized in that described maximum likelihood method is newton's search procedure.

5, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 1 is characterized in that, estimates in the step (5) that the method that unknown parameter adopts is pseudo-Subspace Decomposition method.

6, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 5, it is characterized in that described pseudo-Subspace Decomposition method comprises that distribution signal parameter estimation, spread signal parametrization estimate, add weight subspace and mate three class methods.

7, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 1 is characterized in that, the method for estimating the unknown parameter employing in the step (5) is the approximation method based on covariance matrix.

8, the modelling clutter Doppler parameter method of estimation of airborne radar according to claim 7, it is characterized in that described approximation method based on covariance matrix comprises the invariable rotary signal parameter estimation technique of expansion, the general invariable rotary signal parameter estimation technique two class methods.

Images