CN106125039A - Improvement space-time adaptive Monopulse estimation method based on local Combined Treatment - Google Patents

Abstract

The invention discloses a kind of improvement space-time adaptive Monopulse estimation method based on local Combined Treatment, when object space frequency is with Doppler frequency mismatch simultaneously, by the Doppler frequency of target estimated and azimuth of target, steering vector when iteration updates Space-time domain dimensionality reduction matrix and the target detection sky of JDL algorithm, solve multichannel onboard radar system passage less, DOPPLER RESOLUTION is relatively low, the problem that angle error is big.The present invention only needs two step iteration can accurately estimate object space angle, it is easy to engineering construction.

Description

Improvement space-time adaptive Monopulse estimation method based on local Combined Treatment

Technical field

The present invention relates to airborne radar Monopulse estimation field, be specifically related to a kind of improvement based on local Combined Treatment empty Time adaptive monopulse angle-measuring method.

Background technology

Regarding work under airborne radar, land clutter seriously causes dopplerbroadening so that moving target is easily flooded by clutter, Affect radar target acquisition performance.

Brennan in 1973 etc. propose to utilize Space time pattern (STAP) to carry out clutter reduction.STAP enters Two-dimensional filtering during line space, by unit to be detected (CUT) adjacency unit selection training sample, adaptive polo placement wave filter Weights, have become as the technology of a core, and are considered as the strong tools of airborne radar detection target clutter reduction simultaneously.

Although application STAP technology can improve target detection performance, but is unable to estimate out angle on target.Nickel carries The adaptive monopulse technology gone out is a kind of High-precision angle method of estimation, and the method can extend to space-time two-dimensional.STAP When sky, two-dimensional space realizes self-adapting clutter suppression and the coherent accumulation of moving-target signal, can realize in theory at optimum Reason, but the operand required for full dimension process is surprising, it is assumed that and spatially and temporally hits is respectively N and K, the self adaptation obtained Weights need NK × NK dimension clutter correlation matrix is estimated and inverted, and its operand is O (NK)³, process in soft or hard in real time Huge difficulty is all there is on part.

Dimensionality reduction STAP, by solving of problem is down to lower dimensional space by the linear transformation of full dimension data, feasible system The reduction of degree of freedom.

Summary of the invention

The technical problem to be solved is for defect involved in background technology, it is provided that a kind of based on office The improvement space-time adaptive Monopulse estimation method of territory Combined Treatment, solves multichannel onboard radar system passage less, many General Le resolution is relatively low, the problem that angle error is big.

The present invention solves above-mentioned technical problem by the following technical solutions:

Improvement space-time adaptive Monopulse estimation method based on local Combined Treatment, comprises the steps:

Step 1), according to below equation acquisition airborne radar detecting distance unit each array element reception signal z:

Z=bs+n

Wherein, b represents the complex envelope of target, and n represents clutter plus noise, steering vector when s is target empty,

Amass for Kronecker, s_t=[1 e^j2πv ... e^j2πv(K-1)]^T, s_s=[1 e^j2πu ... e^j2πu(N-1)]^T, s_t、 s_sThe most corresponding time domain steering vector and spatial domain steering vector, subscript T represents transposition computing；

The target normalization Doppler frequency that v represents, u=dsin θ/λ represents target normalization spatial frequency, and d is for showing battle array Unit's spacing, λ is wavelength, and θ is target bearing Space Angle, and N is radar antenna element number of array, and K is arteries and veins in the coherent accumulation cycle Strokes per minute；

Step 2), steering vector s during the detection sky of target place to be detected angle-doppler cells₀:

$s_0=s_{t0}\⊗s_{s0}$

Wherein,s_t0、s_s0Point The time domain steering vector of the most corresponding doppler cells center to be detected and the spatial domain of angle-unit center to be detected guide vows Amount；

v₀Represent the normalization Doppler frequency at doppler cells center to be detected, u₀=d sin θ₀/ λ represents angle measurement to be checked The normalization spatial frequency of degree unit center, θ₀Angle is pointed to for launching antenna bearingt；

Step 3), local Combined Treatment (JDL) the dimensionality reduction square of angle-doppler cells to be detected is obtained according to below equation Battle array:

$T=T_t\⊗T_s$

Wherein, T_tFor time domain dimensionality reduction matrix, T_sFor spatial domain dimensionality reduction matrix:

$T_t={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(v_0-1/K)}& ...& e^{j2\mathrm{\π}(v_0-1/K)(K-1)}\\ 1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\\ 1& e^{j2\mathrm{\π}(v_0+1/K)}& ...& e^{j2\mathrm{\π}(v_0+1/K)(K-1)}\end{array}\right)}^T$

$T_s={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(u_0-1/N)}& ...& e^{j2\mathrm{\π}(u_0-1/N)(N-1)}\\ 1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\\ 1& e^{j2\mathrm{\π}(u_0+1/N)}& ...& e^{j2\mathrm{\π}(u_0+1/N)(N-1)}\end{array}\right)}^T;$

Step 4), according to the JDL dimensionality reduction data of equation below acquisition angle-doppler cells to be detected:

z_T=T^Hz

Wherein, subscript H represents complex conjugate transposition computing；

Step 5), neighbor distance unit carry out Maximum-likelihood estimation as sample and obtain JDL dimensionality reduction clutter and add interference and make an uproar Sound covariance matrix R_T；

Step 6), calculate JDL and wave beam adaptive weight w according to below equation_T:

$w_T=R_T^{-1}{s}_{T_0};$

Wherein,Represent steering vector when the detection of JDL dimensionality reduction is empty,

Step 7), calculate JDL gun parallax wave beam adaptive weight and JDL time domain difference beam adaptive weight respectively, set h =0；

Step 8), the estimated value of target normalization spatial frequency u is calculated according to equation belowTarget normalization Doppler The estimated value of frequency v

$\left[\begin{array}{c}\hat{u}\\ \hat{v}\end{array}\right]=\left[\begin{array}{c}{u}_0\\ v_0\end{array}\right]+{\left[\begin{array}{cc}{c}_{uu},& c_{uv}\\ c_{vu},& c_{vv}\end{array}\right]}^{-1}\left[\begin{array}{c}{r}_u-{\mathrm{\μ}}_u\\ r_v-{\mathrm{\μ}}_v\end{array}\right]$

Wherein, r_uFor gun parallax wave beam with and the pulse ratio of wave beam, μ_uFor r_uOffset correction, r_vPoor for time domain Wave beam with and the pulse ratio of wave beam, μ_vFor r_vOffset correction；Oblique for space-time adaptive pulse ratio Rate matrix；

Step 9), orderH=h+1；

Step 10), it is judged that whether h is less than m, if h is ＜ m, m is the integer more than 1 pre-set, by step 1 Steering vector s when detection is empty₀It is modified toWherein:

$s_{t0}={\left[\begin{array}{cccc}1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\end{array}\right]}^T,s_{s0}={\left[\begin{array}{cccc}1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\end{array}\right]}^T;$

And by the T in the JDL dimensionality reduction matrix T in step 1_tAnd T_sIt is modified to:

$T_t={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(v_0-1/K)}& ...& e^{j2\mathrm{\π}(v_0-1/K)(K-1)}\\ 1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\\ 1& e^{j2\mathrm{\π}(v_0+1/K)}& ...& e^{j2\mathrm{\π}(v_0+1/K)(K-1)}\end{array}\right)}^T,$

$T_s={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(u_0-1/N)}& ...& e^{j2\mathrm{\π}(u_0-1/N)(N-1)}\\ 1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\\ 1& e^{j2\mathrm{\π}(u_0+1/N)}& ...& e^{j2\mathrm{\π}(u_0+1/N)(N-1)}\end{array}\right)}^T;$

Step 11), repeated execution of steps 4) to step 10), until h=m；

Step 12), the estimated value of target bearing Space Angle θ is calculated according to equation below

$\widehat{\mathrm{\θ}}=arcsin(\mathrm{\λ}\hat{u}/d)$

Wherein, arcsin () is arcsine computing；

Step 13), the estimated value of output target bearing Space Angle.

Further optimize as present invention improvement based on local Combined Treatment space-time adaptive Monopulse estimation method Scheme, described step 10) in m equal to 2.

Further optimize as present invention improvement based on local Combined Treatment space-time adaptive Monopulse estimation method Scheme, step 7) described in JDL gun parallax wave beam adaptive weight and JDL time domain difference beam adaptive weight computing formula respectively As follows:

$d_{aT,u}=R_T^{-1}{d}_{T,u},d_{aT,v}=R_T^{-1}{d}_{T,v}$

Wherein, d_T,u=T^Hd_u, d_T,v=T^Hd_v,

Diagonal matrix D_N=diag (λ π i [0 1 ... N-1]/d), D_K=diag (2 π i [0 1 ... K-1]).

Further optimize as present invention improvement based on local Combined Treatment space-time adaptive Monopulse estimation method Scheme, described step 8) in r_v、μ_v、r_u、μ_uWithIn the computing formula of each element be respectively as follows:

$r_v=\mathrm{Re}\left\{\frac{{d_{aT,v}}^H{z}_T}{{w_T}^H{z}_T}\right\},{\mathrm{\μ}}_v=\mathrm{Re}\left\{\frac{{d_{aT,v}}^H{s}_{T_0}}{{w_T}^H{s}_{T_0}}\right\},r_u=\mathrm{Re}\left\{\frac{{d_{aT,u}}^H{z}_T}{{w_T}^H{z}_T}\right\},{\mathrm{\μ}}_u=\mathrm{Re}\left\{\frac{{d_{aT,u}}^H{s}_{T_0}}{{w_T}^H{s}_{T_0}}\right\}$

$c_{uu}=\frac{\mathrm{Re}\left\{{d_{aT,u}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,u}}^H{s}_{T_0}{d_{T,u}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{u}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,u}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{vu}=\frac{\mathrm{Re}\left\{{d_{aT,v}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,v}}^H{s}_{T_0}{d_{T,u}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{v}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,u}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{uv}=\frac{\mathrm{Re}\left\{{d_{aT,u}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,u}}^H{s}_{T_0}{d_{T,v}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{u}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,v}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{vv}=\frac{\mathrm{Re}\left\{{d_{aT,v}}^H{d}_{T,v}{s}_{T_0}{w}_T+{d_{aT,v}}^H{s}_{T_0}{d_{T,v}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{v}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,v}}{{w_T}^H{s}_{T_0}}\right\}.$

Further optimize as present invention improvement based on local Combined Treatment space-time adaptive Monopulse estimation method Scheme, step 5) described in R_TEstimated valueFor:

${\hat{R}}_T=\frac{1}{L}\underset{i=1}{\overset{L}{\Σ}}{x}_{Ti}{x}_{Ti}^H$

Wherein, x_Ti=T^Hx_iRepresent i-th training sample x_iOutput after dimensionality reduction matrix T carries out dimension-reduction treatment, L is sample This number.

Further optimize as present invention improvement based on local Combined Treatment space-time adaptive Monopulse estimation method Scheme, L=27.

The present invention uses above technical scheme compared with prior art, has following technical effect that

1., when object space frequency is with Doppler frequency mismatch simultaneously, the method is by estimating Doppler's frequency of target Steering vector when rate updates Space-time domain dimensionality reduction matrix and the target empty of JDL algorithm with azimuth iteration, it is possible to reduce Doppler across More loss, and then improve output letter miscellaneous noise ratio, angle measurement accuracy more higher than conventional adaptive monopulse can be obtained.

2. fast convergence rate, it is easy to engineering construction.

Accompanying drawing explanation

Fig. 1 is improvement space-time adaptive Monopulse estimation method flow diagram based on JDL；

Fig. 2 is that JDL-STAM algorithm angle on target is estimated with iterations change curve；

Fig. 3 is that JDL-STAM Yu JDL-MSTAM algorithm angle on target is estimated with iterations change curve；

Fig. 4 is that the normalization target Doppler frequency that JDL-STAM Yu JDL-MSTAM algorithm is estimated changes with iterations Curve；

Fig. 5 is that JDL-STAM Yu JDL-MSTAM algorithm angle on target estimates that RMSE is with SCNR change curve；

Fig. 6 is that JDL-MSTAM algorithm angle on target under different pulse number estimates that RMSE is with SCNR change curve.

Detailed description of the invention

Below in conjunction with the accompanying drawings technical scheme is described in further detail:

The invention discloses a kind of improvement space-time adaptive Monopulse estimation method based on local Combined Treatment, in target During spatial frequency and Doppler frequency mismatch simultaneously, by the Doppler frequency of target estimated and azimuth of target, iteration Steering vector when updating Space-time domain dimensionality reduction matrix and the target detection sky of JDL algorithm, it is possible to significantly reduce Doppler and cross over loss, And then improving output letter miscellaneous noise ratio, the method only needs two step iteration can accurately estimate object space angle.

Assume that radar antenna has array element N number of, each array-element antenna isotropic, in the coherent accumulation cycle, umber of pulse is K, Then the radar signal doppler cells signal model of its detector unit can be expressed as form:

Z=bs+n

Wherein, z represents that array of detection units receives signal phasor, and b represents the complex envelope of target, and n represents clutter plus noise； Assuming that clutter plus noise n obeys average is 0, and covariance is the Gauss distribution of R, and clutter, noise and target are orthogonal；S is mesh Steering vector when mark is empty:

$s=s_t\⊗s_s$

Wherein, s_t=[1 e^j2πv ... e^j2πv(K-1)]^T, s_s=[1 e^j2πu ... e^j2πu(N-1)]^T, s_t、s_sRespectively to correspondence time Territory steering vector and spatial domain steering vector, subscript T represents transposition computing, and v represents target normalization Doppler frequency, u=dsin θ/λ represents target normalization spatial frequency, and d is for showing array element distance, and λ is wavelength, and θ is target bearing Space Angle.

Steering vector s during the detection sky of target place to be detected angle-doppler cells₀For:

$s_0=s_{t0}\⊗s_{s0}$

Wherein,s_t0、s_s0Point The time domain steering vector of the most corresponding doppler cells center to be detected and the spatial domain of angle-unit center to be detected guide vows Amount, v₀Represent the normalization Doppler frequency at doppler cells center to be detected, u₀=d sin θ₀/ λ represents angle list to be detected The normalization spatial frequency at unit center, θ₀Angle is pointed to for launching antenna bearingt.

When JDL algorithm is the most empty, signal data transforms to angle-Doppler by bidimensional discrete Fourier transform (DFT) Territory.At angle Doppler domain, owing to radar emission energy is concentrated mainly on observed direction, therefore in observed direction, angle is many General Le unit is grouped, and often consists of a local processing region (LPR).Assume that radar array antenna has N to arrange, and once phase In dry-cure interval, time domain impulse number is K.There are 3 angle-unit and 3 doppler cells in setting LPR, are then mapped to formulation The conversion matrix T of LPR realizes.The dimensionality reduction that wherein T is a series of spatial domain and time domain steering vector Kronecker amasss composition turns Put matrix, be defined as:

$T=T_t\⊗T_s$

In formula,Amass for Kronecker；T_tFor time domain dimensionality reduction matrix, it may be assumed that

$T_t={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(v_0-1/K)}& ...& e^{j2\mathrm{\π}(v_0-1/K)(K-1)}\\ 1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\\ 1& e^{j2\mathrm{\π}(v_0+1/K)}& ...& e^{j2\mathrm{\π}(v_0+1/K)(K-1)}\end{array}\right)}^T$

T_sFor spatial domain dimensionality reduction matrix, it may be assumed that

$T_s={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(u_0-1/N)}& ...& e^{j2\mathrm{\π}(u_0-1/N)(N-1)}\\ 1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\\ 1& e^{j2\mathrm{\π}(u_0+1/N)}& ...& e^{j2\mathrm{\π}(u_0+1/N)(N-1)}\end{array}\right)}^T$

It is assumed herein that clutter is defined as R with the Gauss distribution that noise obedience average is 0 and clutter plus noise covariance matrix. After matrix T operates, the conversion of data realizes with dimensionality reduction simultaneously, and wherein angle domain is divided into 1/N interval, and Doppler domain divides It is spaced for 1/K, is i.e. equivalent to the bidimensional DFT transform of JDL algorithm.

After conversion, JDL dimensionality reduction data z of angle-doppler cells to be detected_T=T^HZ, clutter adds interference noise covariance Matrix R_T=E{z_Tz_T ^H, E{ } represent mathematic expectaion computing, subscript^HRepresent complex conjugate transposition computing, JDL and wave beam self adaptation WeightsWhereinRepresent steering vector when the detection of JDL dimensionality reduction is empty；Owing to miscellaneous characteristic of making an uproar is unknown, actual Above formula covariance matrix R in application_TCalculating by its Maximum-likelihood estimation formReplace:

${\hat{R}}_T=\frac{1}{L}\underset{i=1}{\overset{L}{\Σ}}{x}_{Ti}{x}_{Ti}^H$

Wherein x_Ti=T^Hx_iRepresent i-th training sample x_iOutput after dimensionality reduction matrix T carries out dimension-reduction treatment.L is sample Number, for ensureing estimated accuracy, sample needs statistically to meet independent same distribution condition with the miscellaneous component of making an uproar of unit to be detected, can Take L=27 independent same distribution training sample.

If the normalization Doppler frequency of target is strictly limited to detector unit center Doppler frequency, the returning of target simultaneously One changes spatial frequency also at corresponding detector unit center, then during empty in angle-Doppler domain, steering vector is

s_T=[0 ... 010 ... 0]^T

Wherein, " 1 " represents detection angles-doppler cells, and remaining element is " 0 ".But, when there is deviation, the most just When being target Doppler with angle equal offset detection unit center, during target empty after JDL transition matrix T and conversion, guide time domain Inevitable mismatch.Further resulting in target Doppler and cross over the loss of loss and output signal-to-noise ratio, follow-up target detection error increases Greatly.It is to say, want to promote target angle measurement accuracy further, need to be modified compensating to target Doppler and angle parameter.

Data, through JDL process, should calculate space-time adaptive pulse at angle-doppler cells to be detected corresponding respectively Empty, time domain difference beam adaptive weight.Then JDL gun parallax wave beam adaptive weight and JDL time domain difference beam adaptive weight divide It is not:

$d_{aT,u}=R_T^{-1}{d}_{T,u},d_{aT,v}=R_T^{-1}{d}_{T,v}$

Wherein, d_T,u=T^Hd_u, d_T,v=T^Hd_v,Diagonal matrix D_N= Diag (λ π i [0 1 ... N-1]/d), D_K=diag (2 π i [0 1 ... K-1]).

The target bearing Space Angle estimated by space-time adaptive pulse and the formula of Doppler frequency are as follows:

$\left[\begin{array}{c}\hat{u}\\ \hat{v}\end{array}\right]=\left[\begin{array}{c}{u}_0\\ v_0\end{array}\right]+{\left[\begin{array}{cc}{c}_{uu},& c_{uv}\\ c_{vu},& c_{vv}\end{array}\right]}^{-1}\left[\begin{array}{c}{r}_u-{\mathrm{\μ}}_u\\ r_v-{\mathrm{\μ}}_v\end{array}\right]$

WhereinWithIt is respectively the estimated value of the v of target normalization spatial frequency u and target normalization Doppler frequency, r_u For spatial domain difference beam with and the pulse ratio of wave beam, μ_uFor r_uOffset correction, its computing formula is respectively as follows:

$r_u=\mathrm{Re}\left\{\frac{{d_{aT,u}}^H{z}_T}{{w_T}^H{z}_T}\right\},{\mathrm{\μ}}_u=\mathrm{Re}\left\{\frac{{d_{aT,u}}^H{s}_{T_0}}{{w_T}^H{s}_{T_0}}\right\},$

r_vFor time domain difference beam with and the pulse ratio of wave beam, μ_vFor r_vOffset correction, its computing formula is respectively For:

$r_v=\mathrm{Re}\left\{\frac{{d_{aT,v}}^H{z}_T}{{w_T}^H{z}_T}\right\},{\mathrm{\μ}}_v=\mathrm{Re}\left\{\frac{{d_{aT,v}}^H{s}_{T_0}}{{w_T}^H{s}_{T_0}}\right\},$

For the slope matrix of space-time adaptive pulse ratio, its each element computing formula is respectively as follows:

$c_{uu}=\frac{\mathrm{Re}\left\{{d_{aT,u}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,u}}^H{s}_{T_0}{d_{T,u}}^T{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{u}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,u}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{vu}=\frac{\mathrm{Re}\left\{{d_{aT,v}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,v}}^H{s}_{T_0}{d_{T,u}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{v}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,u}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{uv}=\frac{\mathrm{Re}\left\{{d_{aT,u}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,u}}^H{s}_{T_0}{d_{T,v}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{u}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,v}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{vv}=\frac{\mathrm{Re}\left\{{d_{aT,v}}^H{d}_{T,v}{s}_{T_0}{w}_T+{d_{aT,v}}^H{s}_{T_0}{d_{T,v}}^H{w}_T\right\}}{{\left|{w_T}^H{s}_{T_0}\right|}^2}-{\mathrm{\μ}}_{v}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,v}}{{w_T}^H{s}_{T_0}}\right\}.$

In the case of variable determines, space-time adaptive pulse (abbreviated here as JDL-STAM) based on JDL can use Estimation in target Doppler frequency Yu space angle.But, when target Doppler frequency departure detects doppler cells center During frequency and in the case of spatial frequency mismatch simultaneously, the angle error of JDL-STAM algorithm increases accordingly.JDL-STAM realizes JDL process after detection steering vector coupling, but the Doppler not compensating target crosses over loss.Multistep pulse is permissible Reduce the deviation of estimated value and actual value further.In order to improve angle measurement accuracy further, should be able to be according to the spatial frequency estimated Steering vector s when simultaneously updating detection sky with Doppler frequency₀Propose with the Space-time domain dimensionality reduction matrix T of JDL algorithm, the i.e. present invention Based on JDL improve space-time adaptive single burst algorithm (JDL-MSTAM) given observed direction spatial frequency and normalizing After change Doppler-frequency estimation goes out u Yu v, as new setting initial value u₀And v₀, update detection time domain steering vector s_t0 With spatial domain steering vector s_s0, and dimensionality reduction matrix when redesigning JDL algorithm emptyThat is:

$T_s={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(u_0-1/N)}& ...& e^{j2\mathrm{\π}(u_0-1/N)(N-1)}\\ 1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\\ 1& e^{j2\mathrm{\π}(u_0+1/N)}& ...& e^{j2\mathrm{\π}(u_0+1/N)(N-1)}\end{array}\right)}^T,$

$T_t={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(v_0-1/K)}& ...& e^{j2\mathrm{\π}(v_0-1/K)(K-1)}\\ 1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\\ 1& e^{j2\mathrm{\π}(v_0+1/K)}& ...& e^{j2\mathrm{\π}(v_0+1/K)(K-1)}\end{array}\right)}^T;$

Steering vector s during the detection sky revised₀ForWherein:

$s_{t0}={\left[\begin{array}{cccc}1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\end{array}\right]}^T,s_{s0}={\left[\begin{array}{cccc}1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\end{array}\right]}^T;$

According to update detection empty time steering vector and JDL dimensionality reduction matrix reappraise sample covariance matrix and corresponding from Adapt to and wave beam, time-domain adaptive difference beam and the weights of space domain self-adapted difference beam, and iterative estimate space-time adaptive simple venation Object space angle and Doppler frequency are estimated in punching, have both improve target Doppler coherent accumulation gain, and have reduced again target many The cross-domain loss of general Le and the mismatch loss of steering vector time empty, therefore can obtain more preferable angle measurement accuracy.The simulation experiment result table Bright: by two step interative computations, can accurately estimate target normalization spatial frequency u and normalization Doppler frequency v.Estimate Target bearing Space Angle is

θ=arcsin (λ u/d)

Wherein arcsin () is arcsine computing.

To sum up, the improvement space-time adaptive Monopulse estimation method concrete signal flow chart based on JDL that the present invention proposes See Fig. 1.Computer Simulation assessment algorithm performance is carried out below based on radar clutter simulation data.Under radar system parameters reference Table:

Injecting a target to be detected at distance unit to be detected, its normalization Doppler frequency v=0, how general target place is Strangle normalization center Doppler frequency v of unit₀=1/2K, and LPR is 3 × 3.

Fig. 2 gives the JDL-STAM algorithm angle in target Doppler frequency offset detection unit center Doppler frequency Estimate figure.As it is shown in figure 1, through two step iteration, JDL-STAM algorithm, in the case of skew minimum, i.e. during v=0.004, is estimated The angle of meter is the most accurate.But, when target normalization Doppler frequency v off center normalization Doppler frequency, target angle Degree estimated accuracy reduces immediately, and along with Doppler shift amount increases, its angle estimation error strains greatly mutually.

As v=0.002, two kinds of algorithm each iteration angle estimation results of JDL-STAM Yu JDL-MSTAM such as Fig. 3 institute Show.Unlike JDL-STAM algorithm, the angle estimation error of JDL-MSTAM algorithm is less.JDL-MSTAM algorithm high-precision Degree has benefited from its Combined estimator target Doppler frequency and director space angle, updates JDL transition matrix by iteration, revises sky Time steering vector reduce target Doppler cross over loss and steering vector matching error.Fig. 4 gives JDL-STAM and JDL- The target normalization Doppler frequency of two kinds of each iterative estimates of algorithm of MSTAM.As shown in Figure 4, JDL-MSTAM algorithm is estimated Target Doppler frequency is the most accurate.

Root-mean-square error (RMSE) is used to carry out the angle estimation of the adaptive monopulse processor that quantitative analysis is studied herein Precision.Root-mean-square error is defined as

${\mathrm{\θ}}_{RMSE}=\sqrt{\frac{1}{M}\underset{m=1}{\overset{M}{\Σ}}{({\widehat{\mathrm{\θ}}}_m-\mathrm{\θ})}^2}$

In formula, M is Monte Carlo experiment number of times,Representing the azimuth of target estimated the m time, θ represents actual mesh Mark azimuth.Following result is 3 step space-time adaptive pulse iteration, the meansigma methods of 200 independent Monte Carlo experiments.Two kinds The RMSE of the target bearing Space Angle estimated by method is with signal clutter noise ratio (SCNR) change curve as shown in Figure 5.Target Normalization Doppler frequency v=0.002, SCNR changes to 30dB from-10dB.As can be seen from the figure two kinds of algorithm angle estimations RMSE all reduces along with the increase of SCNR, but the error of JDL-MSTAM algorithm is less.

Assuming that target Doppler frequency departure detection doppler cells mid frequency 40%, Fig. 6 gives JDL-MSTAM and calculates At different pulse number condition, i.e. K=128,64 and 32, angle on target, method estimates that RMSE is with SCNR change curve.From figure permissible Finding out, when umber of pulse is many, DOPPLER RESOLUTION is improved, and JDL-MSTAM algorithm can obtain higher angle measurement accuracy.

It is understood that unless otherwise defined, all terms used herein (include skill to those skilled in the art of the present technique Art term and scientific terminology) have with the those of ordinary skill in art of the present invention be commonly understood by identical meaning.Also It should be understood that those terms defined in such as general dictionary should be understood that have with in the context of prior art The consistent meaning of meaning, and unless defined as here, will not explain by idealization or the most formal implication.

Above-described detailed description of the invention, has been carried out the purpose of the present invention, technical scheme and beneficial effect further Describe in detail, be it should be understood that the detailed description of the invention that the foregoing is only the present invention, be not limited to this Bright, all within the spirit and principles in the present invention, any modification, equivalent substitution and improvement etc. done, should be included in the present invention Protection domain within.

Claims (6)

1. improvement space-time adaptive Monopulse estimation method based on local Combined Treatment, it is characterised in that comprise the steps:

Step 1), according to below equation acquisition airborne radar detecting distance unit each array element reception signal z:

Z=bs+n

Wherein, b represents the complex envelope of target, and n represents clutter plus noise, steering vector when s is target empty,

Amass for Kronecker, s_t=[1 e^j2πv ... e^j2πv(K-1)]^T, s_s=[1 e^j2πu ... e^j2πu(N-1)]^T, s_t、s_sPoint Not corresponding time domain steering vector and spatial domain steering vector, subscript T represents transposition computing；

The target normalization Doppler frequency that v represents, u=d sin θ/λ represents target normalization spatial frequency, and d is for showing between array element Away from, λ is wavelength, and θ is target bearing Space Angle, and N is radar antenna element number of array, and K is umber of pulse in the coherent accumulation cycle；

Step 2), steering vector s during the detection sky of target place to be detected angle-doppler cells₀:

$s_0=s_{t0}\⊗s_{s0}$

Wherein,s_t0、s_s0The most right Answer time domain steering vector and the spatial domain steering vector of angle-unit center to be detected of doppler cells center to be detected；

v₀Represent the normalization Doppler frequency at doppler cells center to be detected, u₀=d sin θ₀/ λ represents angle list to be detected The normalization spatial frequency at unit center, θ₀Angle is pointed to for launching antenna bearingt；

Step 3), according to local Combined Treatment (JDL) the dimensionality reduction matrix of below equation acquisition angle-doppler cells to be detected:

$T=T_t\⊗T_s$

Wherein, T_tFor time domain dimensionality reduction matrix, T_sFor spatial domain dimensionality reduction matrix:

$T_t={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(v_0-1/K)}& ...& e^{j2\mathrm{\π}(v_0-1/K)(K-1)}\\ 1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\\ 1& e^{j2\mathrm{\π}(v_0+1/K)}& ...& e^{j2\mathrm{\π}(v_0+1/K)(K-1)}\end{array}\right)}^T$

$T_s={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(u_0-1/N)}& ...& e^{j2\mathrm{\π}(u_0-1/N)(N-1)}\\ 1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\\ 1& e^{j2\mathrm{\π}(u_0+1/N)}& ...& e^{j2\mathrm{\π}(u_0+1/N)(N-1)}\end{array}\right)}^T;$

Step 4), according to the JDL dimensionality reduction data of equation below acquisition angle-doppler cells to be detected:

z_T=T^Hz

Wherein, subscript H represents complex conjugate transposition computing；

Step 5), neighbor distance unit carry out Maximum-likelihood estimation acquisition JDL dimensionality reduction clutter as sample and add interference noise association Variance matrix R_T；

Step 6), calculate JDL and wave beam adaptive weight w according to below equation_T:

$w_T=R_T^{-1}{s}_{T_0};$

Wherein,Represent steering vector when the detection of JDL dimensionality reduction is empty,

Step 7), calculate JDL gun parallax wave beam adaptive weight and JDL time domain difference beam adaptive weight respectively, set h=0；

Step 8), the estimated value of target normalization spatial frequency u is calculated according to equation belowTarget normalization Doppler frequency v Estimated value

$\left[\begin{array}{c}\hat{u}\\ \hat{v}\end{array}\right]=\left[\begin{array}{c}{u}_0\\ v_0\end{array}\right]+{\left[\begin{array}{cc}{c}_{uu},& c_{uv}\\ c_{vu},& c_{vv}\end{array}\right]}^{-1}\left[\begin{array}{c}{r}_u-{\mathrm{\μ}}_u\\ r_v-{\mathrm{\μ}}_v\end{array}\right]$

Wherein, r_uFor gun parallax wave beam with and the pulse ratio of wave beam, μ_uFor r_uOffset correction, r_vFor time domain difference beam with With the pulse ratio of wave beam, μ_vFor r_vOffset correction；Slope matrix for space-time adaptive pulse ratio；

Step 9), orderH=h+1；

Step 10), it is judged that whether h is less than m, if h is ＜ m, m is the integer more than 1 pre-set, by the detection in step 1 Steering vector s time empty₀It is modified toWherein:

$s_{t0}={\left[\begin{array}{cccc}1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\end{array}\right]}^T,s_{s0}={\left[\begin{array}{cccc}1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\end{array}\right]}^T;$

And by the T in the JDL dimensionality reduction matrix T in step 1_tAnd T_sIt is modified to:

$T_t={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(v_0-1/K)}& ...& e^{j2\mathrm{\π}(v_0-1/K)(K-1)}\\ 1& e^{j2{\mathrm{\πv}}_0}& ...& e^{j2{\mathrm{\πv}}_0(K-1)}\\ 1& e^{j2\mathrm{\π}(v_0+1/K)}& ...& e^{j2\mathrm{\π}(v_0+1/K)(K-1)}\end{array}\right)}^T,$

$T_s={\left(\begin{array}{cccc}1& e^{j2\mathrm{\π}(u_0-1/N)}& ...& e^{j2\mathrm{\π}(u_0-1/N)(N-1)}\\ 1& e^{j2{\mathrm{\πu}}_0}& ...& e^{j2{\mathrm{\πu}}_0(N-1)}\\ 1& e^{j2\mathrm{\π}(u_0+1/N)}& ...& e^{j2\mathrm{\π}(u_0+1/N)(N-1)}\end{array}\right)}^T;$

Step 11), repeated execution of steps 4) to step 10), until h=m；

Step 12), the estimated value of target bearing Space Angle θ is calculated according to equation below

$\widehat{\mathrm{\θ}}=\arcsin (\mathrm{\λ}\hat{u}/d)$

Wherein, arcsin () is arcsine computing；

Step 13), the estimated value of output target bearing Space Angle.

Improvement space-time adaptive Monopulse estimation method based on local Combined Treatment the most according to claim 1, it is special Levy and be, described step 10) in m equal to 2.

3., based on the improvement space-time adaptive Monopulse estimation method based on local Combined Treatment described in claim 1, it is special Levy and be, step 7) described in JDL gun parallax wave beam adaptive weight and JDL time domain difference beam adaptive weight computing formula divide As follows:

$d_{aT,u}=R_T^{-1}{d}_{T,u},d_{aT,v}=R_T^{-1}{d}_{T,v}$

Wherein, d_T,u=T^Hd_u, d_T,v=T^Hd_v,

Diagonal matrix D_N=diag (λ π i [0 1 ... N-1]/d), D_K=diag (2 π i [0 1 ... K-1]).

4., based on the improvement space-time adaptive Monopulse estimation method based on local Combined Treatment described in claim 1, it is special Levy and be, described step 8) in r_v、μ_v、r_u、μ_uWithIn the computing formula of each element be respectively as follows:

$r_v=\mathrm{Re}\left\{\frac{{d_{aT,v}}^H{z}_T}{{w_T}^H{z}_T}\right\},{\mathrm{\μ}}_v=\mathrm{Re}\left\{\frac{{d_{aT,v}}^H{s}_{T_0}}{{w_T}^H{s}_{T_0}}\right\},r_u=\mathrm{Re}\left\{\frac{{d_{aT,u}}^H{z}_T}{{w_T}^H{z}_T}\right\},{\mathrm{\μ}}_u=\mathrm{Re}\left\{\frac{{d_{aT,u}}^H{s}_{T_0}}{{w_T}^H{s}_{T_0}}\right\}$

$c_{uu}=\frac{\mathrm{Re}\left\{{d_{aT,u}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,u}}^H{s}_{T_0}{d_{T,u}}^H{w}_T\right\}}{|{w_T}^H{s}_{T_0}{|}^2}-{\mathrm{\μ}}_{u}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,u}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{vu}=\frac{\mathrm{Re}\left\{{d_{aT,v}}^H{d}_{T,u}{s}_{T_0}{w}_T+{d_{aT,v}}^H{s}_{T_0}{d_{T,u}}^H{w}_T\right\}}{|{w_T}^H{s}_{T_0}{|}^2}-{\mathrm{\μ}}_{v}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,u}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{uv}=\frac{\mathrm{Re}\left\{{d_{aT,u}}^H{d}_{T,v}{s}_{T_0}{w}_T+{d_{aT,u}}^H{s}_{T_0}{d_{T,v}}^H{w}_T\right\}}{|{w_T}^H{s}_{T_0}{|}^2}-{\mathrm{\μ}}_{u}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,v}}{{w_T}^H{s}_{T_0}}\right\},$

$c_{vv}=\frac{\mathrm{Re}\left\{{d_{aT,v}}^H{d}_{T,v}{s}_{T_0}{w}_T+{d_{aT,v}}^H{s}_{T_0}{d_{T,v}}^H{w}_T\right\}}{|{w_T}^H{s}_{T_0}{|}^2}-{\mathrm{\μ}}_{v}2\mathrm{Re}\left\{\frac{{w_T}^H{d}_{T,v}}{{w_T}^H{s}_{T_0}}\right\}.$

5., based on the improvement space-time adaptive Monopulse estimation method based on local Combined Treatment described in claim 1, it is special Levy and be, step 5) described in R_TEstimated valueFor:

${\hat{R}}_T=\frac{1}{L}\underset{i=1}{\overset{L}{\Σ}}{x}_{Ti}{x}_{Ti}^H$

Wherein, x_Ti=T^Hx_iRepresent i-th training sample x_iOutput after dimensionality reduction matrix T carries out dimension-reduction treatment, L is sample Number.

6., based on the improvement space-time adaptive Monopulse estimation method based on local Combined Treatment described in claim 5, it is special Levy and be, L=27.