CN105807267A - MIMO radar extended target detection method - Google Patents

Abstract

The invention relates to a MIMO radar extended target detection method, and the purpose is that multichannel process advantage of a MIMO radar is utilized, so that extended targets in a compound Gaussian clutter can be preferably detected. The core idea of the method comprises: according to target echo received by the MIMO radar, estimation of target motion parameters and clutter parameters are performed, a GLRT basic principle, test statistics is acquired, and the detection is complete at last. Target Doppler mismatching, compound Gaussian clutter shape parameters and the like can influence detection performance if GLRT, so that an MLE method is adopted, estimated values of the target Doppler and clutter shape parameters are obtained, and the target detection is performed. At last, advantages of MIMO radar multichannel processing are used for realizing accumulation of target echo energy, so that performance of a GLRT detector is greatly improved, and the performance improvement relates to the number of MIMO radar transmitting and receiving antenna.

Description

A kind of detection method of MIMO radar Extended target

[technical field]

The invention belongs to traditional Radar Targets'Detection field, be specifically related to current anti-clutter under multiple-input and multiple-output (Multi-inputMulti-output, MIMO) radar system and target detection technique.It is further in Compound-Gaussian Clutter, for the MIMO radar adopting the broadband signals such as linear frequency modulation, it is proposed that the detection method of a kind of Extended target.

[background technology]

Conventional radar adopts individual antenna to launch and receives signal, carries out target acquisition；And MIMO radar adopts multiple transmitting antennas and multiple reception antenna to carry out target acquisition, thus possess good space diversity characteristic and waveform diversity advantage.It is said that in general, MIMO radar can be divided into distributed and centralized two kinds.To different configuration modes, currently mainly adopt narrow band signal, the point target detecting method under Gaussian Clutter has been studied.But, when MIMO radar launches the bandwidth increase of signal, resolution constantly promotes, and the distribution of target scattering point presents the characteristic of extension, and point target model has been difficult to accurately simulate real target characteristic, and it will be more accurate for adopting Extended target model.Additionally, under high-resolution background, when low target is detected, the statistical distribution of clutter echo occurs in that serious hangover, deviate from conventional Gaussian distribution model.In low grazing angle situation, adopt the clutter distribution character that complex Gaussian distribution can describe under high-resolution background more truly.Speckle component (being generally zero-mean gaussian process) and the texture component (being generally non-negative reality stochastic process) of change of slow time that complex Gaussian distribution was changed by the fast time are described.Owing to wideband MIMO radar have employed transmitted wide band signal, having possessed high resolution, Compound-Gaussian Clutter model considers the correlation properties between sample, and analytic properties is good, is more suitable for describing the clutter environment that wideband MIMO radar faces.Currently, broadband signal is widely used, studies wideband MIMO radar object detection method in Compound-Gaussian Clutter environment, has highly important using value.

In field of radar, for target detection problems, the method based on Generalized Likelihood Ratio (generalizedlikelihoodratiotest, GLRT) is obtained in that good performance.In Compound-Gaussian Clutter, adopt GLRT ultimate principle, Extended target parameter is estimated and to carry out the method for target detection relatively common.But, under MIMO radar system, along with the increase of receiver treatment channel number, GLRT method becomes more complicated, and the characteristic of target and clutter is different, and the processing mode of detector also presents different features.For utilizing the advantage of MIMO radar band-wise processing, for the Extended target in Compound-Gaussian Clutter, it is proposed that based on the MIMO radar object detection method of GLRT.

[summary of the invention]

The purpose of this method is in that to utilize the band-wise processing advantage of MIMO radar so that the Extended target in Compound-Gaussian Clutter can be detected better.The core concept of this method is: the target echo received according to MIMO radar, carries out the parameters of target motion and the estimation of clutter parameter, then utilizes GLRT ultimate principle, it is thus achieved that statistic of test, and be finally completed detection.

Owing to the detection performance of GLRT all can be produced impact by target Doppler mismatch, Compound-Gaussian Clutter form parameter etc., method initially with maximal possibility estimation (MLE), obtain the Doppler of target and the estimated value of clutter form parameter, then carry out target detection.Finally, utilizing the advantage of MIMO radar band-wise processing, it is achieved the accumulation of target echo energy, thus being greatly improved the performance of GLRT detector, the raising of performance is relevant with the number of MIMO radar dual-mode antenna.For realizing above-mentioned target detection process, this method adopts following steps to realize:

Step one: signal MIMO radar received carries out pretreatment:

First, signal receiver received is filtered and the pretreatment such as low noise amplification.Considering centralized MIMO radar, have M transmitting antenna and N number of reception antenna, launching signal and adopt M mutually orthogonal broadband signal, pulse number is K.For each reception antenna, comprise M matched filter, the target echo received is carried out matched filtering, then each reception antenna can obtain M echo, and N number of reception antenna can obtain M × N road echo-signal altogether.Assume that the transmitting signal of m-th transmitting antenna carries out, after matched filtering, obtaining echo-signal x through the n-th reception antenna_m,n.If each pulse once being sampled, then x_m,nVector for K × 1.For Extended target, ignoring the impact of noise and Range cell migration, when the scattering unit number of target is L, the Hypothesis Testing Problem of the t target scattering unit can be expressed as follows:

H₀:x_m,n,t=c_m,n,tT=1 ..., L+R

$H_1:\left\{\begin{array}{cc}{x}_{m,n,t}={\mathrm{\α}}_t{s}_{m,n,t}\left(f_d\right)+c_{m,n,t}& t=1,...,L\\ x_{m,n,t}=c_{m,n,t}& t=L+1,...,L+R\end{array}\right.---\left(1\right)$

H₀Represent that distance unit to be detected does not have target, H₁Represent there is target.Wherein, x_m,n,tIt is the echo-signal of t scattering unit, c_m,n,tIt it is noise signal；α_tIt it is target scattering coefficient；f_dDoppler frequency for target；Here, d_tWith d_rExpression is the spacing between transmitting antenna and reception antenna respectively, θ_tmWithIt is transmitting antenna and reception antenna and horizontal angle, T_pRepresent the pulse repetition period；m∈[1,M]；n∈[1,N].Owing to MIMO radar is centralized, it is believed that the target scattering coefficient α that each transmitting antenna is corresponding with reception antenna_tIdentical, and the scattering coefficient between each scattering unit is different, thus α_tDetermined by L；R is auxiliary unit number, only comprises noise signal, for trying to achieve accurate clutter covariance matrix, usually requires that R >=K.

Under MIMO radar system, the clutter vector c of t scattering unit_m,n,tAdopt complex Gaussian model, it is possible to be expressed as

$c_{m,n,t}=\sqrt{{\mathrm{\τ}}_t}{\mathrm{\η}}_t---\left(2\right)$

Wherein, τ_tBeing the texture component of t scattering unit, this component is slow change, it is possible to the clutter power describing different distance unit rises and falls；η_tBeing the speckle component of t scattering unit, be independent identically distributed multiple Gaussian random variable, texture component is separate with speckle component.

For the t scattering unit, clutter covariance matrix C_tCan be expressed as

$C_t=E\left\{c_{m,n,t}{c}_{m,n,t}^H\right\}={\mathrm{\τ}}_{t}E\left\{{\mathrm{\η}}_t{\mathrm{\η}}_t^H\right\}={\mathrm{\τ}}_t\mathrm{\Σ}---\left(3\right)$

Wherein, ()^HThe conjugate transpose computing of representing matrix；

Not limiting scattering unit, if texture component τ meets Gamma distribution, its probability density function can be expressed as

$p\left(\mathrm{\τ}\right)=\frac{2b^{2v}{\mathrm{\τ}}^{2v-1}}{\mathrm{\Γ}\left(v\right)}\exp (-b^2{\mathrm{\τ}}^2),\mathrm{\τ}\≥0---\left(4\right)$

Wherein, b is scale parameter, represents the size of clutter mean power；V is form parameter；Γ (v) is Gamma function；Exp () is exponent arithmetic.Form parameter v characterizes the non-gaussian degree of clutter amplitude, and v is more little, and clutter amplitude distribution is more sharp-pointed, and clutter fluctuation characteristic is more violent, and non-gaussian degree is more big；Otherwise clutter amplitude is distributed closer to Gauss distribution.

At H₀And H₁Under assuming, and joint probability density function p (x | τ_t,H₀) and p (x | τ_t,α_t,f_d,H₁) can be expressed as

$\begin{array}{c}p\left(x\right|{\mathrm{\τ}}_t,H_0)=\underset{t=1}{\overset{L}{\Π}}\underset{m=1}{\overset{M}{\Π}}\underset{n=1}{\overset{N}{\Π}}\frac{1}{{\left({\mathrm{\π\τ}}_t\right)}^K\det (\Σ)}\exp (-\frac{x_{m,n,t}^H{\Σ}^{-1}{x}_{m,n,t}}{{\mathrm{\τ}}_t})---\left(5\right)\\ p\left(x\right|{\mathrm{\τ}}_t,{\mathrm{\α}}_t,f_d,H_1)\\ =\underset{t=1}{\overset{L}{\Π}}\underset{m=1}{\overset{M}{\Π}}\underset{n=1}{\overset{N}{\Π}}(\frac{1}{{\left({\mathrm{\π\τ}}_t\right)}^K\det (\Σ)}\exp \left(-\frac{{\mathrm{\Φ}}_{m,n,t}{({\mathrm{\α}}_t,f_d)}^H{\Σ}^{-1}{\mathrm{\Φ}}_{m,n,t}({\mathrm{\α}}_t,f_d)}{{\mathrm{\τ}}_t}\right)---(6)\end{array}$

Wherein, Φ_m,n,t(α_t,f_d)=x_m,n,t-α_ts_m,n,t(f_d)；Det () is determinant computing；(·)^HFor conjugate transpose computing；All the other each parameters are with reference to definition above.

Step 2: carry out parameter estimation to the received signal, prepares for further object detection.The method of parameter estimation is as follows:

According to target echo and Compound-Gaussian Clutter feature, at H₀Under (target is absent from) condition, (5) formula can obtain clutter texture componentFor

${\widehat{\mathrm{\τ}}}_{t,H_0}=\frac{1}{MNK}\underset{m=1}{\overset{M}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{x}_{m,n,t}^H{\mathrm{\Σ}}^{-1}{x}_{m,n,t}---\left(7\right)$

At H₁Under (target existence) condition, (6) formula the MLE that can obtain target Doppler, target scattering coefficient and clutter texture component can be expressed as successivelyWith

${\hat{f}}_d=\arg \underset{f_d}{max}\left\{\underset{m=1}{\overset{M}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}\frac{{\left|{\left(s_{m,n,t}\right(f_d\left)\right)}^H{\mathrm{\Σ}}^{-1}{x}_{m,n,t}\right|}^2}{{\left(s_{m,n,t}\right(f_d\left)\right)}^H{\mathrm{\Σ}}^{-1}\left(s_{m,n,t}\right(f_d\left)\right)}\right\}---\left(8\right)$

${\widehat{\mathrm{\α}}}_t=\frac{{\left(s_{m,n,t}\right({\hat{f}}_d\left)\right)}^H{\mathrm{\Σ}}^{-1}{x}_{m,n,t}}{{\left(s_{m,n,t}\right({\hat{f}}_d\left)\right)}^H{\mathrm{\Σ}}^{-1}\left(s_{m,n,t}\right({\hat{f}}_d\left)\right)}---\left(9\right)$

${\widehat{\mathrm{\τ}}}_{t,H_1}=\frac{1}{MNK}\underset{m=1}{\overset{M}{\Σ}}\underset{n=1}{\overset{N}{\Σ}}{\mathrm{\Φ}}_{m,n,t}{({\widehat{\mathrm{\α}}}_t,{\hat{f}}_d)}^H{\mathrm{\Σ}}^{-1}{\mathrm{\Φ}}_{m,n,t}({\widehat{\mathrm{\α}}}_t,{\hat{f}}_d)---\left(10\right)$

Wherein, x_m,n,tBe the transmitting signal of the m-th transmitting antenna that the t scattering unit of target is corresponding after the n-th reception antenna matched filtering, obtain echo-signal；s_m,n,t(f_d) represent contain target Doppler information and the echo-signal of signal guide vector；The target Doppler information corresponding due to the every pair of dual-mode antenna is identical, and same target also immobilizes at the doppler information of different scattering units, therefore can use f here_dRepresent the Doppler of target.

Step 3: according at H₀And H₁Echo data under condition, by the joint probability density function (5) being previously obtained and (6) and GLRT criterion, it is possible to obtaining statistic of test is

$GLR=\frac{\underset{{\mathrm{\&tau;}}_t,{\mathrm{\&alpha;}}_t,f_d}{max}\left(p\right(x|{\mathrm{\&tau;}}_t,{\mathrm{\&alpha;}}_t,f_d,H_1\left)\right)}{\underset{{\mathrm{\&tau;}}_t}{max}\left(p\right(x|{\mathrm{\&tau;}}_t,H_0\left)\right)}\underset{H_0}{\overset{H_1}{\begin{array}{c}>\\ <\end{array}}}\mathrm{\&gamma;}---\left(11\right)$

Wherein, γ is the detection threshold under certain false-alarm probability.

Step 4: relevant parameter estimated value step 2 obtained substitutes into (11) formula abbreviation, it is possible to obtain

Formula (12) is taken the logarithm computing, it is possible to obtaining statistic of test is

$\begin{array}{c}T=MNK\underset{t=1}{\overset{L}{\&Sigma;}}(\ln \left(\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}{x}_{m,n,t}^H{\&Sigma;}^{-1}{x}_{m,n,t}\right)\\ -\ln \left(\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}{x}_{m,n,t}^H{\&Sigma;}^{-1}{x}_{m,n,t}-\frac{{\left|s_{m,n,t}{\left({\hat{f}}_d\right)}^H{\&Sigma;}^{-1}{x}_{m,n,t}\right|}^2}{s_{m,n,t}{\left({\hat{f}}_d\right)}^H{\&Sigma;}^{-1}{s}_{m,n,t}\left({\hat{f}}_d\right)}\right))\underset{H_0}{\overset{H_1}{\begin{array}{c}>\\ <\end{array}}}{\mathrm{\&gamma;}}^{\&prime;}\end{array}---\left(13\right)$

Wherein, γ '=ln (γ), by threshold judgement, when statistic of test T is more than thresholding γ ', it is believed that target detected, on the contrary the target of being not detected by.

According to (13) formula it can be seen that statistic of test and target Doppler frequencyRadar transmit-receive antenna number M and N, signal pulse number K and target scattering unit number L are relevant.Target Doppler information is generally unknown, if the Doppler frequency obtained by other means exists deviation, it will cause GLRT hydraulic performance decline, it is therefore necessary to Doppler is estimated.In addition, under Compound-Gaussian Clutter, the clutter amplitude acuity that different form parameter v obtains is different, when v is less, clutter amplitude is comparatively sharp-pointed, for reaching certain false-alarm probability, detection threshold will improve, and by effective estimation of clutter covariance matrix Σ in (13) formula, utilizes in detector, it is possible to reduce echo x_m,n,tThe impact on detection performance of the middle amplitude sharp-pointed clutter.For MIMO radar, the increase of dual-mode antenna number can make GLRT obtain enough energy accumulations, thus improving detection performance, the raising of radar resolution also can increase target scattering unit number on the other hand, equally possible improvement detection performance.

The beneficial effects of the present invention is:

First, utilize MLE, effectively achieve target and the estimation of clutter unknown parameter information, provide good basis for detection, corresponding method of estimation can also be applied in the Radar Targets'Detection problem of other system.

Second, detector is design under the background of Compound-Gaussian Clutter, therefore, it is possible to the Compound-Gaussian Clutter type adapted under difformity parameter, has well adapting to property and range of application.For different form parameters, detection performance also has certain difference.

3rd, due to MIMO radar employing is wideband orthogonal signal, and the transmitting signal between different antennae is mutually orthogonal, therefore, when each reception antenna carries out matched filtering, it is possible to eliminate the impact that adjacent antenna signal brings, it is ensured that the independence of signal launched by each antenna.Adopting broadband signal, it is possible to increase the resolution capability of MIMO radar, target is become Extended target from point target, and the scattering properties of different scattering units is different so that detector can adapt to different target properties.

4th, utilize the advantage of MIMO radar MIMO, be greatly improved the energy accumulation effect of detector, thus reaching to detect preferably performance.Therefore, under different dual-mode antenna configuration modes, the detection performance of MIMO radar is also different.Expression formula according to detector, in dual-mode antenna number sum one timing, detection performance reaches optimum when dual-mode antenna number is equal.

5th, the ultimate principle of this method can extend to other application scenarios of MIMO radar target detection, only need to according to different noise performances, corresponding detector of deriving.

[accompanying drawing explanation]

The detection that Fig. 1 is this method realizes overall procedure.

Fig. 2 is M matched filtering processing procedure of the n-th antenna.

Fig. 3 is Doppler's estimated bias that under different antennae configuration mode, MLE algorithm obtains.

Fig. 4 a is clutter amplitude distribution situation corresponding for difformity parameter v.

Fig. 4 b is the 2 × 2MIMO radar detedtion probability curve under difformity parameter in v situation.

2 × 2MIMO radar detedtion probability curve when Fig. 5 is Doppler mismatch.

Fig. 6 is the MIMO radar detection probability curve under different antennae configuration mode.

[detailed description of the invention]

This method is applicable to the MIMO radar of various employing broadband signals.Shown in accompanying drawing 1-6, concrete steps and effect to this method are further explained below.Mainly comprise the following steps:

Step one: first set simulation parameter: MIMO radar transmitted signal bandwidth 200MHz；PRF is 500Hz；Pulse number K is 16；False-alarm probability isHypothetical target only has orientation to extension, and scattering unit number is 4, and corresponding scattering coefficient is set to α_t=[1.22.33.12.5]；The speed of target is 25m/s.The signal obtained by N number of reception antenna carries out down coversion, obtains intermediate-freuqncy signal, and the matched filtering for next step is prepared.

Step 2: the signal received is carried out matched filtering process.MIMO radar adopts MIMO system, keeps orthogonal property between the transmitting signal of each transmitting antenna.Therefore, at receiving terminal, each receiver can receive M launch signal and signal, utilize M matched filter, it is possible to respectively obtain M filter export echo-signal.For N number of reception antenna, it is possible to obtain M × N number of filtering output.

Step 3: utilize MIMO radar to close on the covariance matrix Σ of the echo-signal calculating clutter that distance unit obtains in target, then according to above-mentioned MLE algorithm, calculate target and clutter relevant parameter.

Fig. 3 is under the MIMO radar background that dual-mode antenna is 1 × 3,2 × 2 and 2 × 3, according to MLE algorithm (formula (8)), the target Doppler estimated bias obtained.It appeared that the algorithm for estimating in this method is after signal to noise ratio (SCR) reaches 10dB, estimated bias can reach less level.

According to MLE algorithm, after obtaining Doppler's estimated value, utilizing formula (9) can calculate the estimated value of target scattering coefficient, result of calculation here isContrast actual value, it is believed that the estimated value of scattering coefficient is more accurately.Utilize above parameter estimation result, it is possible to carry out target detection.

Step 4: according to the false-alarm probability set, obtain detection threshold, then utilizes parameter estimation result and echo-signal to obtain detector output, carries out threshold judgement, finally obtain testing result.

Effectiveness and method for this method is described realizes the importance of the parameter estimation mentioned in process, according to simulation parameter above, under the MIMO radar background that dual-mode antenna is 2 × 2, first give when clutter form parameter difference, the target detection curve that flow chart according to Fig. 1 obtains, result is as shown in Figure 4 b.It appeared that under different clutter form parameters, detector remains to detect preferably target, and less in clutter form parameter, when clutter amplitude distribution comparatively sharp-pointed (as shown in fig. 4 a), detection performance still increases.

Fig. 5 is the target detection curve under target Doppler mismatch condition, it is possible to find when Doppler mismatch is approximately 2Hz, and along with the increase of SCR, detection performance remains to reach 1.When Doppler mismatch reaches 6Hz, detection performance finally can only be maintained at about 0.2.Visible, the accuracy that target Doppler is estimated will have a strong impact on the performance of detector.Therefore, it is necessary to target Doppler is estimated accurately by MLE algorithm, to ensure the performance of detector.

Fig. 6 is under the different antennae configuration mode that this method obtains, the target detection curve of MIMO radar.From simulation result it appeared that, along with number of antennas increases, the detection performance of MIMO radar effectively improves, this is because when number of antennas increases, target echo can be effectively taking place accumulation, make statistic of test increase and thresholding is constant, therefore the performance of the MIMO radar of 2 × 3 is better than the situation of 1 × 3.In addition, under identical dual-mode antenna number (M+N fixes), as M=N, detection performance is more excellent than under other configurations, this is because when dual-mode antenna number is equal, M × N reaches maximum, and therefore in simulation result, the detection performance of the MIMO radar of 2 × 2 is better than the detection performance under 1 × 3 configuration mode.

By the detection performance contrasting that Doppler is known and under unknown situation, it is believed that when Doppler's the unknown, MLE algorithm the Doppler's estimated value obtained is more accurate, it is possible to ensure the detection performance of GLRT.The simulation result of Fig. 6, can be MIMO radar under Compound-Gaussian Clutter in the problem of extension target detection, the arrangement of dual-mode antenna provides foundation, when number of antennas is certain, make dual-mode antenna number equal, it is possible to make the detection best performance of MIMO radar.

Claims (4)

1. the detection method of a MIMO radar Extended target, it is characterised in that comprise the steps:

Step one: signal MIMO radar received carries out pretreatment:

First, the pretreatment that signal receiver received is filtered and low noise is amplified；

For Extended target, ignoring the impact of noise and Range cell migration, when the scattering unit number of target is L, the Hypothesis Testing Problem of the t target scattering unit is expressed as follows:

H₀:x_m,n,t=c_m,n,tT=1 ..., L+R

$H_1:\left\{\begin{array}{cc}{x}_{m,n,t}={\mathrm{\&alpha;}}_t{s}_{m,n,t}\left(f_d\right)+c_{m,n,t}& t=1,...,L\\ x_{m,n,t}=c_{m,n,t}& t=L+1,...,L+R\end{array}\right.---\left(1\right)$

H₀Represent that distance unit to be detected does not have target, H₁Represent there is target；Wherein, x_m,n,tIt is the echo-signal of t scattering unit, c_m,n,tIt it is noise signal；α_tIt it is target scattering coefficient；f_dDoppler frequency for target；Here, d_tWith d_rExpression is the spacing between transmitting antenna and reception antenna respectively, θ_tmWithIt is transmitting antenna and reception antenna and horizontal angle, T_pRepresent the pulse repetition period；m∈[1,M]；n∈[1,N]；

Under MIMO radar system, the clutter vector c of t scattering unit_m,n,tAdopt complex Gaussian model, be expressed as

$c_{m,n,t}=\sqrt{{\mathrm{\&tau;}}_t}{\mathrm{\&eta;}}_t---\left(2\right)$

Wherein, τ_tBeing the texture component of t scattering unit, this component is slow change, and the clutter power describing different distance unit rises and falls；η_tBeing the speckle component of t scattering unit, be independent identically distributed multiple Gaussian random variable, texture component is separate with speckle component；

For the t scattering unit, clutter covariance matrix C_tIt is expressed as

$C_t=E\left\{C_{m,n,t}{C}_{m,n,t}^H\right\}={\mathrm{\&tau;}}_{t}E\left\{{\mathrm{\&eta;}}_t{\mathrm{\&eta;}}_t^H\right\}={\mathrm{\&tau;}}_t\mathrm{\&Sigma;}---\left(3\right)$

Wherein, ()^HThe conjugate transpose computing of representing matrix；

Not limiting scattering unit, if texture component τ meets Gamma distribution, its probability density function is expressed as

$p\left(\mathrm{\&tau;}\right)=\frac{2b^{2v}{\mathrm{\&tau;}}^{2v-1}}{\mathrm{\&Gamma;}\left(v\right)}\exp (-b^2{\mathrm{\&tau;}}^2),\mathrm{\&tau;}\&GreaterEqual;0---\left(4\right)$

Wherein, b is scale parameter, represents the size of clutter mean power；V is form parameter；Γ (v) is Gamma function；Exp () is exponent arithmetic；Form parameter v characterizes the non-gaussian degree of clutter amplitude, and v is more little, and clutter amplitude distribution is more sharp-pointed, and clutter fluctuation characteristic is more violent, and non-gaussian degree is more big；Otherwise clutter amplitude is distributed closer to Gauss distribution；

At H₀And H₁Under assuming, and joint probability density function p (x | τ_t,H₀) and p (x | τ_t,α_t,f_d,H₁) be expressed as

$p\left(x\right|{\mathrm{\&tau;}}_t,H_0)=\underset{t=1}{\overset{L}{\mathrm{\&Pi;}}}\underset{m=1}{\overset{M}{\mathrm{\&Pi;}}}\underset{n=1}{\overset{N}{\mathrm{\&Pi;}}}\frac{1}{{\left({\mathrm{\&pi;\&tau;}}_t\right)}^K\det \left(\mathrm{\&Sigma;}\right)}\exp (-\frac{x_{m,n,t}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t}}{{\mathrm{\&tau;}}_t})---\left(5\right)$

$\begin{array}{c}p\left(x\right|{\mathrm{\&tau;}}_t,{\mathrm{\&alpha;}}_t,f_d,H_1)\\ =\underset{t=1}{\overset{L}{\mathrm{\&Pi;}}}\underset{m=1}{\overset{M}{\mathrm{\&Pi;}}}\underset{n=1}{\overset{N}{\mathrm{\&Pi;}}}\left(\frac{1}{{\left({\mathrm{\&pi;\&tau;}}_t\right)}^K\det \left(\mathrm{\&Sigma;}\right)}\exp \right(-\frac{{\mathrm{\&Phi;}}_{m,n,t}{({\mathrm{\&alpha;}}_t,f_d)}^H{\mathrm{\&Sigma;}}^{-1}{\mathrm{\&Phi;}}_{m,n,t}({\mathrm{\&alpha;}}_t,f_d)}{{\mathrm{\&tau;}}_t}))\end{array}---\left(6\right)$

Wherein, Φ_m,n,t(α_t,f_d)=x_m,n,t-α_ts_m,n,t(f_d)；Det () is determinant computing；(·)^HFor conjugate transpose computing；

Step 2: carry out parameter estimation to the received signal, prepares for further object detection；The method of parameter estimation is as follows:

According to target echo and Compound-Gaussian Clutter feature, under target is absent from condition, it is H₀Time, (5) formula obtain clutter texture componentFor

${\widehat{\mathrm{\&tau;}}}_{t,H_0}=\frac{1}{MNK}\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}{x}_{m,n,t}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t}---\left(7\right)$

Under target existence condition, it is H₁Time, (6) formula the MLE obtaining target Doppler, target scattering coefficient and clutter texture component is represented sequentially asWith

${\hat{f}}_d=\arg \underset{f_d}{max}\left\{\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}\frac{|{\left(s_{m,n,t}\right(f_d\left)\right)}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t}{|}^2}{{\left(s_{m,n,t}\right(f_d\left)\right)}^H{\mathrm{\&Sigma;}}^{-1}\left(s_{m,n,t}\right(f_d\left)\right)}\right\}---\left(8\right)$

${\widehat{\mathrm{\&alpha;}}}_t=\frac{{\left(s_{m,n,t}\right({\hat{f}}_d\left)\right)}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t}}{{\left(s_{m,n,t}\right({\hat{f}}_d\left)\right)}^H{\mathrm{\&Sigma;}}^{-1}\left(s_{m,n,t}\right({\hat{f}}_d\left)\right)}---\left(9\right)$

${\widehat{\mathrm{\&tau;}}}_{t,H_1}=\frac{1}{HK}\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}{\mathrm{\&Phi;}}_{m,n,t}{({\widehat{\mathrm{\&alpha;}}}_t,{\hat{f}}_d)}^H{\mathrm{\&Sigma;}}^{-1}{\mathrm{\&Phi;}}_{m,n,t}({\widehat{\mathrm{\&alpha;}}}_t,{\hat{f}}_d)---\left(10\right)$

Wherein, x_m,n,tBe the transmitting signal of the m-th transmitting antenna that the t scattering unit of target is corresponding after the n-th reception antenna matched filtering, obtain echo-signal；s_m,n,t(f_d) represent contain target Doppler information and the echo-signal of signal guide vector；The target Doppler information corresponding due to the every pair of dual-mode antenna is identical, and same target also immobilizes at the doppler information of different scattering units, therefore uses f_dRepresent the Doppler of target；

Step 3: according at H₀And H₁Echo data under condition, by the joint probability density function (5) being previously obtained and (6) and GLRT criterion, obtaining statistic of test is

$GLR={\frac{\underset{{\mathrm{\&tau;}}_t,{\mathrm{\&alpha;}}_t,f_d}{max}\left(p\right(x|{\mathrm{\&tau;}}_t,{\mathrm{\&alpha;}}_t,f_d,H_1\left)\right)}{\underset{{\mathrm{\&tau;}}_t}{max}\left(p\right(x|{\mathrm{\&tau;}}_t,H_0\left)\right)}}_{\begin{array}{c}<\\ H_0\end{array}}^{\begin{array}{c}{H}_1\\ >\end{array}}\mathrm{\&gamma;}---\left(11\right)$

Wherein, γ is the detection threshold under certain false-alarm probability；

Step 4: relevant parameter estimated value step 2 obtained substitutes into (11) formula abbreviation, obtains

Formula (12) being taken the logarithm computing, obtaining statistic of test is

$\begin{array}{c}T=MNK\underset{t=1}{\overset{L}{\mathrm{\&Sigma;}}}\left(ln\right(\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}{x}_{m,n,t}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t})\\ -\ln (\underset{m=1}{\overset{M}{\&Sigma;}}\underset{n=1}{\overset{N}{\&Sigma;}}{x}_{m,n,t}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t}-\frac{|s_{m,n,t}{\left({\hat{f}}_d\right)}^H{\mathrm{\&Sigma;}}^{-1}{x}_{m,n,t}{|}^2}{s_{m,n,t}{\left({\hat{f}}_d\right)}^H{\mathrm{\&Sigma;}}^{-1}{s}_{m,n,t}\left({\hat{f}}_d\right)}){)}_{\begin{array}{c}<\\ H_0\end{array}}^{\begin{array}{c}{H}_1\\ >\end{array}}{\mathrm{\&gamma;}}^{\&prime;}\end{array}---\left(13\right)$

Wherein, γ '=ln (γ), by threshold judgement, when statistic of test T is more than thresholding γ ', it is believed that target detected, on the contrary the target of being not detected by；

Find out according to (13) formula, statistic of test and target Doppler frequencyRadar transmit-receive antenna number M and N, signal pulse number K and target scattering unit number L are relevant；Target Doppler information is unknown, if the Doppler frequency obtained by other means exists deviation, it will cause GLRT hydraulic performance decline, it is therefore necessary to Doppler is estimated；In addition, under Compound-Gaussian Clutter, the clutter amplitude acuity that different form parameter v obtains is different, when v is less, clutter amplitude is comparatively sharp-pointed, for reaching certain false-alarm probability, detection threshold will improve, and by effective estimation of clutter covariance matrix Σ in (13) formula, utilizes in detector, reduce echo x_m,n,tThe impact on detection performance of the middle amplitude sharp-pointed clutter；For MIMO radar, the increase of dual-mode antenna number can make GLRT obtain enough energy accumulations, thus improving detection performance, the raising of radar resolution also can increase target scattering unit number on the other hand, can improve detection performance equally.

2. the detection method of a kind of MIMO radar Extended target according to claim 1, it is characterised in that: described MIMO radar is centralized, has M transmitting antenna and N number of reception antenna, launches signal and adopts M mutually orthogonal broadband signal, and pulse number is K；For each reception antenna, comprise M matched filter, the target echo received is carried out matched filtering, then each reception antenna obtains M echo, and N number of reception antenna obtains M × N road echo-signal altogether.

3. the detection method of a kind of MIMO radar Extended target according to claim 1, it is characterised in that: the transmitting signal setting m-th transmitting antenna carries out, after matched filtering, obtaining echo-signal x through the n-th reception antenna_m,n；If each pulse once being sampled, then x_m,nVector for K × 1.

4. the detection method of a kind of MIMO radar Extended target according to claim 1, it is characterised in that: because MIMO radar is centralized, then it is assumed that the target scattering coefficient α that each transmitting antenna is corresponding with reception antenna_tIdentical, and the scattering coefficient between each scattering unit is different, thus α_tDetermined by L；R is auxiliary unit number, only comprises noise signal, for trying to achieve clutter covariance matrix accurately, it is desirable to R >=K.