CN103954942A - Method for partial combination clutter suppression in airborne MIMO radar three-dimensional beam space - Google Patents

Abstract

The invention belongs to the technical field of radar signal processing, relates to localized suppression processing of radar clutter, and discloses a partly combined clutter suppression method in three-dimensional beam space of an airborne MIMO radar. The method includes: Step 1. Obtain the space-time data vector x expression and construct the space-time two-dimensional steering vector of the target; Step 2. Use the three-dimensional discrete Fourier transform to transform the space-time data vector x from the array element-pulse domain to Three-dimensional beam domain; Step 3. Obtain the space-time data vector z after dimensionality reduction and the target space-time two-dimensional steering vector after dimensionality reduction; Step 4. Calculate the dimensionality reduction processor according to the cost function of dimensionality reduction space-time adaptive processing Weight vector; Step 5. Using the weight vector of the dimensionality reduction processor to obtain space-time data that has undergone clutter suppression. The invention can solve the problems of large calculation amount and large sample demand, and is applied to the scenario of radar clutter processing.

Description

Partial Joint Clutter Suppression Method in Three-Dimensional Beam Space of Airborne MIMO Radar

technical field

The invention belongs to the technical field of radar signal processing, and relates to localized suppression processing of radar clutter, in particular to a partial combined clutter suppression method in three-dimensional beam space of an airborne MIMO radar.

Background technique

Since MIMO radar has become the development direction of the next-generation radar, MIMO radar has also become a current research hotspot. MIMO radar uses digital array technology at the transmitter and receiver at the same time, transmits multiple orthogonal or incoherent signals, and uses matched filtering to separate the transmitted signal components at the receiver. MIMO radar can obtain a large The virtual array aperture and system degrees of freedom can greatly improve the radar parameter estimation accuracy, angular resolution and clutter suppression ability. Since MIMO radar adopts digital array technology at both the transmitting end and receiving end, MIMO radar has obvious advantages compared with traditional radar in terms of anti-clutter, anti-interference, low interception and other performance.

Space Time Adaptive Processing (STAP) is a key technology for airborne early warning radars to detect slow-moving targets, but it is affected by the large amount of calculations caused by the inversion of the high-dimensional data covariance matrix and the large number of independent distribution samples required. , making it difficult to realize in practical applications. Under the system of phased array radar, people have proposed many dimensionality reduction STAP methods aimed at reducing the number of samples and the amount of computation, such as the principal component method (Principle Component, PC), the factorization method (Factored Approach, FA) and the expansion factor Method (Extended Factored Approach, EFA), etc. The basic idea of the PC method is to use the eigenvectors of the clutter subspace to cancel the clutter of the main branch; the FA method first uses a set of Doppler filters with high out-of-band attenuation to output the output of each receiving element of the MIMO radar. Filter, and then use the spatial domain Capon beamforming to adaptively process the output of the same Doppler channel; the EFA method combines m (m usually takes an odd number of 3, 5, etc.) Doppler channels for self-adaptive processing near the main clutter area. Adapt to handling. Although these methods are also applicable to airborne MIMO radar, due to the transmit waveform diversity, MIMO-STAP extends the traditional space-time two-dimensional processing to the space-time-code (waveform) three-dimensional space, and the data dimension increases sharply. As a result, problems such as computation load and covariance matrix estimation have become more prominent. Simply using these methods is still difficult to meet the real-time requirements. The amount of computation for estimating and inverting the variance matrix will also be very large, that is, the prior art is restricted by a large amount of computation and a large demand for samples. Therefore, there is an urgent need to study more efficient dimensionality-reduced STAP methods for airborne MIMO radars.

Contents of the invention

When the existing method is applied to airborne MIMO radar clutter suppression, the present invention proposes a partly combined clutter suppression method in the three-dimensional beam space of airborne MIMO radar, which is used to solve the problems of large calculation amount and large sample demand.

In order to achieve the above object, the present invention adopts the following technical solutions to realize.

A partial joint clutter suppression method in three-dimensional beam space of airborne MIMO radar, is characterized in that, comprises the following steps:

Step 1. Use the airborne MIMO radar antenna to receive space-time data, obtain the space-time data vector x expression and construct the space-time two-dimensional steering vector b(f _s,0 ,f _d,0 );

Step 2. Utilize the three-dimensional discrete Fourier transform to transform the space-time data vector x from the array element-pulse domain to the three-dimensional beam domain; the space-time data vector x is called a three-dimensional beam in the three-dimensional beam domain;

Step 3. Construct the dimensionality reduction matrix P, and use the dimensionality reduction matrix P to perform partial joint adaptive processing on the 3D beams in the 3D beam domain to obtain the space-time data vector z after dimensionality reduction and the target space-time two-dimensional guidance after dimensionality reduction vector c(f _s,0 ,f _d,0 );

Step 4. According to the linear constraint minimum variance criterion, that is, to minimize the output clutter and noise power under the premise of ensuring a certain target signal gain, use the dimensionally reduced space-time data vector z and the dimensionally reduced target space-time two-dimensional Steering vector c(f _s,0 ,f _d,0 ) constructs the cost function of dimensionality reduction space-time adaptive processing; calculates the weight vector w of dimensionality reduction processor according to the cost function of dimensionality reduction space-time adaptive processing;

Step 5. Use the weight vector w of the dimensionality reduction processor to weight and sum the space-time data vector x received by the airborne MIMO radar antenna to obtain the space-time data after clutter suppression.

The characteristics and further improvement of the above-mentioned technical scheme are:

(1) Step 1 includes the following sub-steps:

1a) It is set that within a coherent processing interval, the M transmitting elements of the airborne MIMO radar antenna simultaneously radiate a burst waveform composed of K pulses, and the M transmitting waveforms emitted by the M transmitting elements are orthogonal to each other, At each of the N receiving array elements, M reference transmitting signals are used to perform matched filtering on the echo data of K pulse periods, and the outputs of all matched filters are arranged into the following MNK×1-dimensional space-time data Vector, the expression of the space-time data vector x is obtained as the following formula:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; \&pi;</span>}}\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; d\&theta;</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}$

Among them, the dimension of space-time data vector x is MNK×1 dimension, is the clutter scattering coefficient that obeys the zero-mean complex Gaussian distribution; Indicates the normalized clutter Doppler frequency, T _r is the pulse repetition period, λ is the working wavelength of the radar antenna, v is the speed of the carrier aircraft, θ is the azimuth angle of the observed ground, is the pitch angle; Indicates the normalized clutter spatial frequency; $b(f_{\mathrm{the\; s}},f_d)=a_d\left(f_d\right)\&CircleTimes;a_t\left(f_{\mathrm{the\; s}}\right)\&CircleTimes;a_r\left(f_{\mathrm{the\; s}}\right)$ is the space-time two-dimensional steering vector of the clutter in the space-time data x, where a _t (f _s )=[1 exp(j2πf _s α)…exp(j2πf _s α(M-1))] ^T is M× 1-dimensional transmitting array steering vector, α=d _T /d _R , M is the number of transmitting array elements, N is the number of receiving array elements, K is the number of transmitting pulses, d _T is the distance between transmitting array elements, and d _R is the distance between receiving array elements ; a _r (f _s )=[1 exp(j2πf _s )…exp(j2πf _s (N-1))] ^T is N×1-dimensional receiving array steering vector, a _d (f _d )=[1 exp(j2πf _d )...exp(j2πf _d (K-1))] ^T is the K×1 dimensional Doppler steering vector; Indicates the Kronecker product; is subject to 0 mean, variance is The MNK×1-dimensional Gaussian white noise vector, I _MNK represents the MNK×MNK identity matrix, and Respectively represent the input clutter power and noise power, ( ) ^T represents the vector transpose;

1b) Construct the space-time two-dimensional guidance vector of the target $b(f_{\mathrm{the\; s},0},f_{d,0})=a_d\left(f_{d,0}\right)\&CircleTimes;a_t\left(f_{\mathrm{the\; s},0}\right)\&CircleTimes;a_r\left(f_{\mathrm{the\; s},0}\right);$

Among them, a _d (f _d,0 ) is the target Doppler steering vector, at (f _s,0 ) is the transmitting array steering vector, _{a r (} f _s,0 ) is the receiving array steering vector, f _d,0 is the Doppler frequency of the target, f _s,0 is the normalized spatial frequency of the target, Represents the Kronecker product.

(2) Step 3 includes the following sub-steps:

3a) Construct a dimensionality reduction matrix P, and the dimensionality reduction matrix P is expressed as

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}$

Among them, G _d =[a _d (f _d,-1 ), a _d (f _d,0 ), a _d (f _d,1 )] is the Doppler filter composed of three adjacent Doppler steering vectors G _t =[a _t (f _s,-1 ), _at (f _s,0 ), _at (f _s,1 )] is the transmit beamformer composed of three adjacent transmit array steering vectors group, G _r =[a _r (f _s,-1 ), a _r (f _s,0 ), a _r (f _s,1 )] is the receiving beamformer group composed of three adjacent receiving array steering vectors , Indicates the Kronecker product;

3b) Select 3 or 5 3D beams from the 3D beam field;

3c) Using the dimensionality reduction matrix P to perform partial joint adaptive processing on the selected three-dimensional beams according to the following formula, to obtain the space-time data vector z after dimensionality reduction of the space-time data vector x; obtain the following formula:

z = P ^H x

Among them, P is the dimensionality reduction matrix, z is the space-time data vector after dimensionality reduction of the space-time data vector x, ( ) ^H represents the complex conjugate transposition of the matrix or vector;

3d) Calculate the dimensionality-reduced target space-time two-dimensional steering vector c(f _{s ,0} ,f _{d , 0} ), get the following formula:

c(f _s,0 ,f _d,0 )＝P ^H b(f _s,0 ,f _d,0 )

Among them, c(f _s,0 ,f _d,0 ) is the target space-time two-dimensional steering vector after dimension reduction of the target space-time two-dimensional steering vector b(f _s,0 ,f _d,0 ); the dimensionality reduction matrix P The dimension is MNK×r _M r _N r _K , where r _M represents the number of selected transmit beams, r _N represents the number of selected receive beams, and r _K represents the number of selected Doppler channels; where r _M is greater than or equal to r _N is an integer greater than or equal to 3 and less than N, r _K is an integer greater than or equal to 3 and less than K, f _d,0 is the Doppler frequency of the target, f _s,0 is The normalized spatial frequency of the target, (·) ^H represents the complex conjugate transpose of a matrix or vector, M is the number of transmitting array elements, N is the number of receiving array elements, and K is the number of transmitting pulses.

(3) Step 4 includes the following sub-steps:

4a) Use the space-time data vector z after dimension reduction to obtain the covariance matrix R _{z of the space-time data vector z after dimension reduction, and the expression of R z is} :

R _z =E{zz ^H }=P ^H RP

Among them, z represents the space-time data vector after dimensionality reduction, P is the dimensionality reduction matrix, R is the covariance matrix of space-time data x, ( ) ^H represents the complex conjugate transpose of matrix or vector, E{ } represents perform the desired operation;

4b) According to the linearly constrained minimum variance criterion, the cost function of dimensionality reduction space-time adaptive processing is expressed as:

$\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; min</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}\mathrm{<span\; class="notranslate">\; w</span>}\\ \mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; .</span>{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right.$

Among them, R _z is the covariance matrix of the space-time data vector z after dimensionality reduction, and the dimension is r _M r _N r _K ; r _M represents the number of selected transmit beams, r _N represents the number of selected receive beams, r _K represents the number of Doppler channels selected; where r _M is an integer greater than or equal to 3 and less than M, r _N is an integer greater than or equal to 3 and less than N, and r _K is an integer greater than or equal to 3 and less than K, M is the number of transmitting array elements, N is the number of receiving array elements, K is the number of transmitting pulses, (·) ^H represents the complex conjugate transposition of matrix or vector;

4c) Solve the cost function of dimensionality reduction space-time adaptive processing, and obtain the weight vector w of dimensionality reduction processor:

$\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; c</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}$

Among them, c(f _s,0 ,f _d,0 ) is the target space-time two-dimensional steering vector after dimensionality reduction, (·) ^H represents the complex conjugate transpose of matrix or vector, [·] ^-1 represents the pair matrix Inverse, f _d,0 is the Doppler frequency of the target, f _s,0 is the normalized spatial frequency of the target, R _z is the covariance matrix of the space-time data vector z after dimensionality reduction.

Compared with the prior art, the present invention has outstanding substantive features and remarkable progress. Compared with existing methods, the present invention has the following advantages:

(1) Compared with the existing methods, the target detection performance of the localization method of the present invention is better. As shown in Figure 4, Figure 4 compares the improvement factor curves of the four methods. The performance of the FA method drops sharply when the target Doppler frequency f _d,0 is close to zero, because the clutter has penetrated from the main lobe of the Doppler filter, and the spatial domain Adaptive processing obviously cannot effectively filter out clutter. The difference is that, in addition to detecting the Doppler channel f _d,0 in the EFA method and the localization method of the present invention, two groups of auxiliary channels f _d,-1 and f _d,1 are used to participate in self-adaptation, so compared FA method, the performance of these two methods is obviously improved in the main clutter area, and there are certain improvements in other areas. In addition, the localization of the present invention reduces the requirement for training samples through simultaneous dimensionality reduction in the space domain and the time domain. Therefore, the localized target detection performance of the present invention is better than that of the EFA method under the condition that the number of samples is only 300, and Compared with the optimal STAP, there is only about 2.7dB performance loss.

(2) The target power of the present invention is larger than the value of the residual clutter peak power in the localization method compared with the existing method, so that the clutter suppression performance of the present invention is better. Moving target signals with signal-to-noise ratios of 0dB and -15dB are respectively injected into range cells No. 235 and No. 260. The targets are all located in the side-looking direction of the carrier aircraft, and the normalized Doppler frequency is f _d,0 =0.125. As shown in Figure 5, Figure 5 shows the normalized output power of the four methods at distance units 200 to 300. It can be clearly seen from Fig. 5(a) and 5(b) that the residual clutter power of the PC and FA methods is relatively large, which leads to the obliteration of the weak target located at the No. 260 range gate. Figure 5(c) shows the output results of the EFA method. The target powers of No. 235 and No. 260 range gates are 21.68dB and 6.72dB higher than the peak power of residual clutter, respectively, and the target powers of No. 235 and No. 260 range gates are higher than the average power of residual clutter. 31.28dB and 16.60dB. In Fig. 5(d), using the localization method, the target powers of range gates 235 and 260 are 23.26dB and 6.98dB higher than the residual clutter peak power respectively, and the target powers of range gates 235 and 260 are higher than the residual clutter peak power respectively. Wave average power 33.55dB and 17.27dB. It can be seen that the residual clutter power of the method of the present invention is lower than that of the PC method, FA method and EFA method. The above results show that the clutter suppression performance of the localization method of the present invention is better than other methods.

(3) Compared with the existing method, the localization method of the present invention has a faster convergence speed, which reduces the amount of computation and the demand for samples. Fig. 6 is the change curve of the improvement factors of the four methods in the target Doppler channel with the number of samples. It can be seen from the comparison that the localization method has a faster convergence speed, especially under the condition of small samples, and its performance is significantly better than other methods. It can be seen that in the real clutter environment where it is difficult to obtain a large number of independent and identically distributed reference units, the localization method has great advantages. The present invention reduces the requirement on training samples through simultaneous dimensionality reduction processing in space domain and time domain. In terms of calculation amount, the calculation amount of the localization of the present invention is O[(r _M ×r _N ×r _K ) ³ ], while the calculation amount of the EFA method is as high as O[(r _K ×M×N) ³ ], where , M is the number of transmitting array elements, N is the number of receiving array elements, r _M represents the number of selected transmitting beams, r _N represents the number of selected receiving beams, r _K represents the number of selected Doppler channels; when r _M = 3. When r _N =5, r _K =3, M=5, N=10, the amount of calculation of the EFA method is about 37 times that of the present invention. It can be seen that the amount of calculation of the localization method of the present invention is higher than that of the existing method The amount of computation is low.

The present invention is specifically a new airborne multiple-input multiple-output (Multiple-Input-Multiple-Output, MIMO) radar three-dimensional beam space partial joint clutter suppression method, which greatly reduces the amount of computation and the requirements for the number of reference units. Conducive to project realization.

Description of drawings

The present invention will be further described below in conjunction with the accompanying drawings and specific embodiments.

Figure 1 is a schematic diagram of an airborne MIMO radar system model;

Fig. 2 is the data processing flowchart of the localization method of the present invention;

Fig. 3 is a schematic diagram of the localization method of the present invention;

Fig. 4 is the improvement factor graph of PC method, FA method, EFA method and localization method of the present invention;

Fig. 5 is the normalized output power figure of PC method, FA method, EFA method and the localization method of the present invention at No. 200～300 distance units, wherein Fig. 5 (a) is the normalized output power figure of PC method; Fig. 5 (b) is the FA method; Fig. 5 (c) is the normalized output power diagram of the EFA method; Fig. 5 (d) is the normalized output power diagram of the localization method of the present invention;

Fig. 6 is a graph showing the variation of the improvement factor of the PC method, the FA method, the EFA method and the localization method of the present invention in the target Doppler channel with the number of samples.

Detailed ways

With reference to Fig. 1 and Fig. 2, the partial joint clutter suppression method of a kind of airborne MIMO radar three-dimensional beam space of the present invention is illustrated, and its specific steps are as follows:

Step 1. Use the airborne MIMO radar antenna to receive space-time data, and construct the space-time two-dimensional steering vector b(f _s,0 ,f _d,0 ) of the target;

1a) It is set that within a coherent processing interval, the M transmitting elements of the airborne MIMO radar antenna simultaneously radiate a burst waveform composed of K pulses, and the M transmitting waveforms emitted by the M transmitting elements are orthogonal to each other, At each of the N receiving array elements, M reference transmitting signals are used to perform matched filtering on the echo data of K pulse periods, and the outputs of all matched filters are arranged into the following MNK×1-dimensional space-time data Vector, the expression of the space-time data vector x is obtained as the following formula:

$\mathrm{<span\; class="notranslate">\; x</span>}<span\; class="notranslate">\; =</span>{<span\; class="notranslate">\; \&Integral;</span>}_{\mathrm{<span\; class="notranslate">\; 0</span>}}^{\mathrm{<span\; class="notranslate">\; \&pi;</span>}}\mathrm{<span\; class="notranslate">\; \&rho;</span>}<span\; class="notranslate">\; (</span>\mathrm{<span\; class="notranslate">\; \&theta;</span>}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; b</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; )</span>\mathrm{<span\; class="notranslate">\; d\&theta;</span>}<span\; class="notranslate">\; +</span>{\mathrm{<span\; class="notranslate">\; x</span>}}_{\mathrm{<span\; class="notranslate">\; w</span>}}$

Among them, the dimension of space-time data x is MNK×1 dimension, is the clutter scattering coefficient that obeys the zero-mean complex Gaussian distribution; Indicates the normalized clutter Doppler frequency, T _r is the pulse repetition period, λ is the working wavelength of the radar antenna, v is the speed of the carrier aircraft, θ is the azimuth angle of the observed ground, is the pitch angle; Indicates the normalized clutter spatial frequency; $b(f_{\mathrm{the\; s}},f_d)=a_d\left(f_d\right)\&CircleTimes;a_t\left(f_{\mathrm{the\; s}}\right)\&CircleTimes;a_r\left(f_{\mathrm{the\; s}}\right)$ is the space-time two-dimensional steering vector of the clutter in the space-time data x, where a _t (f _s )=[1 exp(j2πf _s α)…exp(j2πf _s α(M-1))] ^T is M ×1-dimensional transmitting array steering vector, α=d _T /d _R , M is the number of transmitting array elements, N is the number of receiving array elements, K is the number of transmitting pulses, d _T is the distance between transmitting array elements, and d _R is the receiving array element spacing; a _r (f _s )=[1 exp(j2πf _s )…exp(j2πf _s (N-1))] ^T is the N×1-dimensional receiving array steering vector, a _d (f _d )=[1 exp( j2πf _d )…exp(j2πf _d (K-1))] ^T is the K×1 dimensional Doppler steering vector; Indicates the Kronecker product; is subject to 0 mean, variance is The MNK×1-dimensional Gaussian white noise vector, I _MNK represents the MNK×MNK identity matrix, and Respectively represent the input clutter power and noise power, ( ) ^T represents the vector transpose;

1b) Construct the space-time two-dimensional guidance vector of the target $b(f_{\mathrm{the\; s},0},f_{d,0})=a_d\left(f_{d,0}\right)\&CircleTimes;a_t\left(f_{\mathrm{the\; s},0}\right)\&CircleTimes;a_r\left(f_{\mathrm{the\; s},0}\right);$

Among them, a _d (f _d,0 ) is the target Doppler steering vector, at (f _s,0 ) is the transmitting array steering vector, _{a r (} f _s,0 ) is the receiving array steering vector, f _d,0 is the Doppler frequency of the target, f _s,0 is the normalized spatial frequency of the target, Represents the Kronecker product.

In the prior art, the weight vector wopt of the space-time two-dimensional optimal processor of the airborne MIMO radar _is expressed as:

w _opt ＝μR ^-1 b(f _s,0 ,f _d,0 )

Among them, μ=1/[b ^H (f _s,0 ,f _d,0 )R ^-1 b(f _s,0 ,f _d,0 )] is a normalized complex constant, R=E{xx ^H } is the covariance matrix of the space-time data vector x, the dimension is MNK, M is the number of transmitting array elements, N is the number of receiving array elements, K is the number of transmitting pulses; ( ) ^H represents the complex conjugate transposition of the matrix or vector , [ ] ^-1 means to invert the matrix, and E( ) means mathematical expectation.

However, when solving the weight vector w _opt of the space-time two-dimensional optimal processor in the prior art, the M, N, and K involved in the actual MIMO radar are usually dozens or even hundreds, so the MNK dimensional data is directly There are two defects in adaptive processing: one is that the calculation amount O(M ³ N ³ K ³ ) for the inversion of the covariance matrix R is too large, and the processor hardware is difficult to realize; the other is that the independent and identical distribution reference The number of units is required to be no less than 2MNK, which is difficult to meet in practice, especially in the background of non-uniform clutter. Therefore, the application of STAP must adopt the dimension reduction processing scheme, so as to carry out the following steps 2:

Step 2. Utilize the three-dimensional discrete Fourier transform to transform the space-time data vector x from the array element-pulse domain to the three-dimensional beam domain; the space-time data vector x is called a three-dimensional beam in the three-dimensional beam domain;

As shown in Fig. 3, the transmit element coordinates of the space-time data vector x are transformed into transmit beam coordinates, the receive array element coordinates are transformed into receive beam coordinates, and the time-domain pulse coordinates are transformed into Doppler channel coordinates.

Step 3. Construct the dimensionality reduction matrix P, and use the dimensionality reduction matrix P to perform partial joint adaptive processing on the 3D beams in the 3D beam domain to obtain the space-time data vector z after dimensionality reduction and the target space-time two-dimensional guidance after dimensionality reduction vector c(f _s,0 ,f _d,0 );

3a) Construct a dimensionality reduction matrix P, and the dimensionality reduction matrix P is expressed as:

$\mathrm{<span\; class="notranslate">\; P</span>}<span\; class="notranslate">\; =</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; t</span>}}<span\; class="notranslate">\; \&CircleTimes;</span>{\mathrm{<span\; class="notranslate">\; G</span>}}_{\mathrm{<span\; class="notranslate">\; r</span>}}$

Among them, G _d =[a _d (f _d,-1 ), a _d (f _d,0 ), a _d (f _d,1 )] is the Doppler filter composed of three adjacent Doppler steering vectors G _t =[a _t (f _s,-1 ), _at (f _s,0 ), _at (f _s,1 )] is the transmit beamformer composed of three adjacent transmit array steering vectors group, G _r =[a _r (f _s,-1 ), a _r (f _s,0 ), a _r (f _s,1 )] is the receiving beamformer group composed of three adjacent receiving array steering vectors , Represents the Kronecker product.

3b) Select 3 or 5 3D beams from the 3D beam field.

3c) Using the dimensionality reduction matrix P to perform partial joint adaptive processing on the selected three-dimensional beams according to the following formula, to obtain the space-time data vector z after dimensionality reduction of the space-time data vector x; obtain the following formula:

z = P ^H x

Among them, P is the dimensionality reduction matrix, z is the space-time data vector after dimensionality reduction of space-time data vector x, (·) ^H represents the complex conjugate transpose of matrix or vector.

3d) Calculate the dimensionality-reduced target space-time two-dimensional steering vector c(f _{s ,0} ,f _{d , 0} ), get the following formula:

c(f _s,0 ,f _d,0 )＝P ^H b(f _s,0 ,f _d,0 )

Among them, c(f _s,0 ,f _d,0 ) is the target space-time two-dimensional steering vector after dimension reduction of the target space-time two-dimensional steering vector b(f _s,0 ,f _d,0 ); the dimensionality reduction matrix P The dimension is MNK×r _M r _N r _K , where r _M represents the number of selected transmit beams, r _N represents the number of selected receive beams, and r _K represents the number of selected Doppler channels; where r _M is greater than or equal to r _N is an integer greater than or equal to 3 and less than N, r _K is an integer greater than or equal to 3 and less than K, f _d,0 is the Doppler frequency of the target, f _s,0 is The normalized spatial frequency of the target, (·) ^H represents the complex conjugate transpose of a matrix or vector, M is the number of transmitting array elements, N is the number of receiving array elements, and K is the number of transmitting pulses.

Since the processor dimension of the localization method of the present invention is r _M r _N r _K , it is required that the number of independent and identically distributed reference units is not less than 2r _M r _N r _K , and the number of independent and identically distributed reference units in the present invention is the number of samples . R _z ＝E{zz ^H }＝P ^H RP is the covariance matrix of the space-time data vector z after dimensionality reduction, and the calculation amount of inversion of this matrix R _z is Usually r _M , r _N and r _K take 3 or 5; r _M r _N r _K << MNK, in actual radar, M, N, K are usually dozens or even hundreds, in the prior art, a The reason is that the amount of calculation O(M ³ N ³ K ³ ) for the inversion of the covariance matrix R is too large to be realized by the processor hardware; the second is that the number of independent and identically distributed reference units required for estimating the covariance matrix is not less than 2MNK, so The number of IID reference units required is far greater than the number of IID reference units required by the present invention, so the localization method of the present invention can greatly reduce the amount of computation and the requirement for the number of reference units. In fact, the larger the selected local area, the better the clutter suppression effect, but at the same time the amount of calculation involved will increase accordingly, so the performance and calculation amount should be considered comprehensively to choose a reasonable local area size . In the simulation of the present invention, a local area size of 3×5×3 is adopted, and near optimal clutter suppression performance is obtained without considering errors. However, in practice, due to the influence of array errors, the clutter spectrum will broaden. At this time, several more airspace channels should be selected to increase the degree of freedom in the airspace, and the time domain accuracy is higher, so a few Doppler units can be selected less.

Step 4. According to the linear constraint minimum variance criterion, that is, to minimize the output clutter and noise power under the premise of ensuring a certain target signal gain, use the dimensionally reduced space-time data vector z and the dimensionally reduced target space-time two-dimensional Steering vector c(f _s,0 ,f _d,0 ) constructs the cost function of dimensionality reduction space-time adaptive processing; calculates the weight vector w of dimensionality reduction processor according to the cost function of dimensionality reduction space-time adaptive processing;

4a) Use the space-time data vector z after dimension reduction to obtain the covariance matrix R _{z of the space-time data vector z after dimension reduction, and the expression of R z is} :

R _z =E{zz ^H }=P ^H RP

Among them, z represents the space-time data vector after dimensionality reduction, P is the dimensionality reduction matrix, R is the covariance matrix of space-time data x, ( ) ^H represents the complex conjugate transpose of matrix or vector, E{ } represents Perform the desired operation.

4b) According to the linearly constrained minimum variance criterion, the cost function of dimensionality reduction space-time adaptive processing is expressed as:

$\left\{\begin{array}{c}\mathrm{<span\; class="notranslate">\; min</span>}{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}\mathrm{<span\; class="notranslate">\; w</span>}\\ \mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; .</span>\mathrm{<span\; class="notranslate">\; t</span>}<span\; class="notranslate">\; .</span>{\mathrm{<span\; class="notranslate">\; w</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span><span\; class="notranslate">\; =</span>\mathrm{<span\; class="notranslate">\; 1</span>}\end{array}\right.$

Among them, R _z is the covariance matrix of the space-time data vector z after dimensionality reduction, and the dimension is r _M r _N r _K ; r _M represents the number of selected transmit beams, r _N represents the number of selected receive beams, r _K represents the number of Doppler channels selected; where r _M is an integer greater than or equal to 3 and less than M, r _N is an integer greater than or equal to 3 and less than N, and r _K is an integer greater than or equal to 3 and less than K, M is the number of transmitting array elements, N is the number of receiving array elements, K is the number of transmitting pulses, (·) ^H represents the complex conjugate transposition of matrix or vector;

4c) Solve the cost function of dimensionality reduction space-time adaptive processing, and obtain the weight vector w of dimensionality reduction processor:

$\mathrm{<span\; class="notranslate">\; w</span>}<span\; class="notranslate">\; =</span>\frac{{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}{{\mathrm{<span\; class="notranslate">\; c</span>}}^{\mathrm{<span\; class="notranslate">\; h</span>}}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>{\mathrm{<span\; class="notranslate">\; R</span>}}_{\mathrm{<span\; class="notranslate">\; z</span>}}^{<span\; class="notranslate">\; -</span>\mathrm{<span\; class="notranslate">\; 1</span>}}\mathrm{<span\; class="notranslate">\; c</span>}<span\; class="notranslate">\; (</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; the\; s</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; ,</span>{\mathrm{<span\; class="notranslate">\; f</span>}}_{\mathrm{<span\; class="notranslate">\; d</span>}<span\; class="notranslate">\; ,</span>\mathrm{<span\; class="notranslate">\; 0</span>}}<span\; class="notranslate">\; )</span>}$

Among them, c(f _s,0 ,f _d,0 ) is the target space-time two-dimensional steering vector after dimensionality reduction, (·) ^H represents the complex conjugate transpose of matrix or vector, [·] ^-1 represents the pair matrix Inverse, f _d,0 is the Doppler frequency of the target, f _s,0 is the normalized spatial frequency of the target, R _z is the covariance matrix of the space-time data vector z after dimensionality reduction.

Step 5. Use the weight vector w of the dimensionality reduction processor to weight and sum the space-time data vector x received by the airborne MIMO radar antenna to obtain the space-time data after clutter suppression.

Because in the present invention, the weight vector w of the dimensionality reduction processor is obtained according to the linear constraint minimum variance criterion, the clutter output power in the space-time data after clutter suppression is the smallest, and the energy of the target signal is kept constant, thereby completing the clutter inhibition.

Since the processor dimension of the partial joint clutter suppression method of the present invention is r _M r _N r _K , it is required that the number of independent and identically distributed reference units is not less than 2r _M r _N r _K ; matrix R _z only needs to match the The covariance matrix of domain clutter is estimated and inverted, and the calculation amount is Usually r _M , r _N and r _K take 3 or 5, and r _M r _N r _K <<MNK. In actual radar, M, N, and K are usually dozens or even hundreds. In the prior art, one It is that the amount of computation O(M ³ N ³ K ³ ) for the inversion of the covariance matrix R is far greater than the amount of computation for the inversion of the dimension-reduced covariance matrix in the present invention, which is difficult for the processor hardware to realize; The number of i.d. reference units is required to be no less than 2MNK. The number of i.d. reference units required in the prior art is far greater than the number of i.d. The quantity and the requirement for the number of reference units are beneficial to the realization of the project.

In order to further illustrate the partial joint clutter suppression method of airborne MIMO radar three-dimensional beam space of the present invention, also can be referred to as localization method, with respect to the superiority of existing method (such as PC method, FA method and EFA method) , through the following simulation.

The effects of the present invention will be further described below in combination with simulation experiments.

(1) Experimental conditions:

The local area size selected in the present invention is r _M ×r _N ×r _K =3×5×3, and the time domain filtering of the FA method adopts -60dB Chebyshev weighting. The experimental parameters are set as follows: aircraft speed v=150m/s, height h=9km, radar antenna operating wavelength λ=0.3m, pulse number K=16, pulse period T _r =5×10 ^-4 s, radar range resolution is 150m, the number of transmitting array elements M=5, the number of receiving array elements is N=10, the spacing between transmitting array elements d _t =1.5m, and the spacing between receiving array elements d _r =0.15m, that is, the spacing between receiving array elements is half a wavelength and The emission array is sparse. The ground clutter fluctuation is not considered in the computer simulation, and it is considered that the clutter of each distance unit is independent of each other, and the noise power The clutter-to-noise ratio is 40dB; the clutter covariance matrix is obtained from the data of 300 range gates from 100 to 399; it is assumed that the radar detection direction is always the side-looking direction, that is, the normalized spatial frequency f _s,0 =0.

In the present invention, the space-time data vector x received by one range unit is taken as a sample, and the space-time data samples received by the radar at different range units are independent of each other, and the number of independent samples is the number of range units.

(2) Analysis of experimental results

Experiment 1: As shown in Figure 4, the abscissa is the normalized Doppler frequency, and the ordinate is the improvement factor. Figure 4 compares the regression of the PC method, the FA method, the EFA method and the localization method of the present invention. Improvement factor curve for normalized Doppler frequency change. The performance of the FA method drops sharply when the target Doppler frequency f _d,0 is close to zero, because the clutter has penetrated from the main lobe of the Doppler filter, and the spatial domain is only performed in the detection Doppler channel. Adaptive processing obviously cannot effectively filter out clutter. The difference is that, in addition to detecting the Doppler channel f _d,0 in the EFA method and the localization method of the present invention, two groups of auxiliary channels f _d,-1 and f _d,1 are used to participate in self-adaptation, so compared FA method The performance of these two methods is obviously improved in the main clutter area, and there are certain improvements in other areas. In addition, the localization method of the present invention reduces the requirements for training samples through simultaneous dimensionality reduction in the space and time domains. Therefore, the performance of the localization method is better than that of the EFA method when the number of samples is only 300, and compared with the optimal STAP , only about 2.7dB performance loss. In terms of calculation volume, the calculation volume of the localization method of the present invention is O[(3×5×3) ³ ], while the calculation volume of the EFA method is as high as O[(3×5×10) ³ ], which is about 37 times that of localized methods. Although the PC method takes advantage of the low-rank characteristics of clutter to reduce the requirement on the number of samples, its performance is poor.

Experiment 2: Inject moving target signals with signal-to-noise ratios of 0dB and -15dB into range cells No. 235 and No. 260 respectively. The targets are located in the side-looking direction of the carrier aircraft, and the normalized Doppler frequency is f _d,0 = 0.125 . As shown in Figure 5, the abscissa is the distance unit, and the ordinate is the output power, which specifically refers to the normalized output power in the simulation of the present invention. Normalized output power for ~300 range cells. It should be noted that the FA method in the present invention is also the 1DT method, and the EFA method is also the 3DT method. It can be clearly seen from Fig. 5(a) and 5(b) that the residual clutter power of the PC and FA methods is relatively large, which leads to the obliteration of the weak target located at the No. 260 range gate. Figure 5(c) shows the output results of the EFA method. The target powers of No. 235 and No. 260 range gates are 21.68dB and 6.72dB higher than the peak power of residual clutter, respectively, and the target powers of No. 235 and No. 260 range gates are higher than the average power of residual clutter. 31.28dB and 16.60dB. Among Fig. 5 (d), utilize the localization method of the present invention, gained No. 235 and No. 260 range gate target powers are respectively higher than residual clutter peak power 23.26dB and 6.98dB, gained No. 235 and No. 260 range gate target powers are respectively higher than The average power of residual clutter is 33.55dB and 17.27dB. It can be seen that the target power of the localization method of the present invention is higher than the value of the residual clutter power than the target power of other methods is higher than the value of the residual clutter power, that is, the residual clutter power of the method of the present invention is higher than that of the PC method, Both the FA method and the EFA method have low residual clutter power, and the above results show that the clutter suppression performance of the localization method of the present invention is better than other methods. It should be noted that the partial joint clutter suppression method studied in the present invention is the localization method shown in FIG. 4 , FIG. 5 and FIG. 6 .

Experiment three: as shown in Figure 6, the abscissa is the number of independent samples, and the ordinate is the improvement factor. Figure 6 shows the results of the PC method, the FA method, the EFA method and the localization method of the present invention in the target Doppler channel. The change curve of the improvement factor with the number of independent samples. It can be seen from the comparison that the localization method of the present invention has a faster convergence speed, especially under the condition of small samples, and the target detection performance is significantly better than other methods. In the present invention, the space-time data vector x received by one range unit is taken as a sample, and the space-time data samples received by the radar at different range units are independent of each other, and the number of independent samples is the number of range units. It can be seen that in the real clutter environment where it is difficult to obtain a large number of IID reference units, the localization method of the present invention has great advantages.

Claims (4)

1. the part of an airborne MIMO radar three-dimensional beam space associating clutter suppression method, is characterized in that, comprises the following steps:

Data when step 1. utilizes airborne MIMO radar antenna to receive sky, the space-time two-dimensional steering vector b (f of data vector x expression formula and establishing target while obtaining sky _{s, 0}, f _{d, 0});

When step 2. is utilized 3 d-dem Fourier transform by sky, data vector x transforms to three-dimensional Beam Domain by array element-pulse domain; When empty, data vector x is called three-dimensional wave beam in three-dimensional Beam Domain;

Step 3. structure dimensionality reduction matrix P, utilizes dimensionality reduction matrix P to carry out part associating self-adaptive processing to the three-dimensional wave beam in three-dimensional Beam Domain, two-dimensional guide vector C (f during target empty while obtaining empty after dimensionality reduction after data vector z and dimensionality reduction _{s, 0}, f _{d, 0});

Step 4. is according to linearly constrained minimum variance, in the clutter and the noise power that guarantee to minimize under the prerequisite that targeted signal gain is certain output, and two-dimensional guide vector C (f during target empty while utilizing empty after dimensionality reduction after data vector z and dimensionality reduction _{s, 0}, f _{d, 0}) build the cost function that dimensionality reduction space-time adaptive is processed; The weight vector w of the cost function calculation dimension-reduction treatment device of processing according to dimensionality reduction space-time adaptive;

The weight vector w that step 5. is utilized dimension-reduction treatment device receives airborne MIMO radar antenna when empty, and data vector x is weighted summation, obtain suppressing through clutter empty time data.

2. the part of a kind of airborne MIMO radar three-dimensional beam space according to claim 1 associating clutter suppression method, is characterized in that, step 1 comprises following sub-step:

1a) be set in relevant a processing in interval, the train of impulses waveform that the radiation simultaneously of the M of airborne MIMO radar antenna transmitting array element is comprised of K pulse, and M transmitted waveform of M transmitting array element transmitting is mutually orthogonal, at N each reception array element place that receives array element, by M reference transmitted signal, the echo data of K recurrence interval is carried out to matched filtering, the data vector when output of all matched filters is arranged in to following MNK * 1 dimension sky, while obtaining sky, data vector x expression formula is following formula:

$x={\&Integral;}_0^{\mathrm{\&pi;}}\mathrm{\&rho;}\left(\mathrm{\&theta;}\right)b(f_s,f_d)\mathrm{d\&theta;}+x_w$

Wherein, when empty, the dimension of data vector x is MNK * 1 dimension, it is the clutter scattering coefficient of obeying the multiple Gaussian distribution of zero-mean; represent normalization clutter Doppler frequency, T _rfor the pulse repetition time, λ is radar antenna operation wavelength, and v is carrier aircraft speed, and θ represents to observe the position angle on ground, for the angle of pitch; represent normalization clutter spatial frequency; $b(f_s,f_d)=a_d\left(f_d\right)\&CircleTimes;a_t\left(f_s\right)\&CircleTimes;a_r\left(f_s\right)$ The space-time two-dimensional steering vector of the clutter during for sky in data x, wherein a _t(f _s)=[1 exp (j2 π f _sα) ... exp (j2 π f _sα (M-1))] ^tfor M * 1 dimension emission array steering vector, α=d _t/ d _r, M is transmitting array number, and N is for receiving array number, and K is transponder pulse number, d _tfor transmitting array element distance, d _rfor receiving array element distance; a _r(f _s)=[1 exp (j2 π f _s) ... exp (j2 π f _s(N-1))] ^tfor N * 1 dimension receiving array steering vector, a _d(f _d)=[1 exp (j2 π f _d) ... exp (j2 π f _d(K-1))] ^tfor K * 1 dimension Doppler steering vector; represent that Kronecker is long-pending; be to obey 0 average, variance is mNK * 1 dimension white Gaussian noise vector, I _mNKrepresent MNK * MNK unit matrix, with represent respectively input clutter power and noise power, () ^trepresent vectorial transposition;

1b) the space-time two-dimensional steering vector of establishing target $b(f_{s,0},f_{d,0})=a_d\left(f_{d,0}\right)\&CircleTimes;a_t\left(f_{s,0}\right)\&CircleTimes;a_r\left(f_{s,0}\right);$

Wherein, a _d(f _{d, 0}) be target Doppler steering vector, a _t(f _{s, 0}) be emission array steering vector, a _r(f _{s, 0}) be receiving array steering vector, f _{d, 0}for the Doppler frequency of target, f _{s, 0}for the normalization spatial frequency of target, represent that Kronecker is long-pending.

3. the part of a kind of airborne MIMO radar three-dimensional beam space according to claim 1 associating clutter suppression method, is characterized in that, step 3 comprises following sub-step:

3a) structure dimensionality reduction matrix P, dimensionality reduction matrix P is expressed as

$P=G_d\&CircleTimes;G_t\&CircleTimes;G_r$

Wherein, G _d=[a _d(f _{d ,-1}), a _d(f _{d, 0}), a _d(f _{d, 1})] be the Doppler filter group of Doppler's steering vector composition of three vicinities, G _t=[a _t(f _{s ,-1}), a _t(f _{s, 0}), a _t(f _{s, 1})] be the launching beam shaper group of the emission array steering vector composition of three vicinities, G _r=[a _r(f _{s ,-1}), a _r(f _{s, 0}), a _r(f _{s, 1})] be the receive beamformer group of the receiving array steering vector composition of three vicinities, represent that Kronecker is long-pending;

3b) from three-dimensional Beam Domain, select 3 or 5 three-dimensional wave beams;

3c) utilize dimensionality reduction matrix P according to following formula, to carry out part associating self-adaptive processing to the three-dimensional wave beam of selecting, data vector z during empty while obtaining sky after data vector x dimensionality reduction; Obtain following formula:

z＝P ^Hx

Wherein, P is dimensionality reduction matrix, data vector during empty when z is empty after data vector x dimensionality reduction, () ^hthe complex-conjugate transpose of representing matrix or vector;

Two-dimensional guide vector b (f during 3d) according to dimensionality reduction matrix P and target empty _{s, 0}, f _{d, 0}) two-dimensional guide vector C (f while asking for the target empty after dimensionality reduction _{s, 0}, f _{d, 0}), obtain following formula:

c(f _s,0,f _d,0)＝P ^Hb(f _s,0,f _d,0)

Wherein, c (f _{s, 0}, f _{d, 0}) two-dimensional guide vector b (f while being target empty _{s, 0}, f _{d, 0}) two-dimensional guide vector during target empty after dimensionality reduction; Dimensionality reduction matrix P dimension is MNK * r _mr _nr _k, r _mthe number of the launching beam that expression is chosen, r _nthe number of the received beam that expression is chosen, r _krepresent Doppler's number of active lanes of choosing; R wherein _mfor being more than or equal to 3 and be less than the integer of M, r _nfor being more than or equal to 3 and be less than the integer of N, r _kfor being more than or equal to 3 and be less than the integer of K, f _{d, 0}for the Doppler frequency of target, f _{s, 0}for the normalization spatial frequency of target, () ^hthe complex-conjugate transpose of representing matrix or vector, M is transmitting array number, and N is for receiving array number, and K is transponder pulse number.

4. the part of a kind of airborne MIMO radar three-dimensional beam space according to claim 1 associating clutter suppression method, is characterized in that, step 4 comprises following sub-step:

The covariance matrix R of data vector z when data vector z asks for empty after dimensionality reduction while 4a) utilizing empty after dimensionality reduction _z, R _zexpression formula is:

R _z＝E{zz ^H}＝P ^HRP

Wherein, data vector when z represents empty after dimensionality reduction, P is dimensionality reduction matrix, the covariance matrix of data x when R is empty, () ^hthe complex-conjugate transpose of representing matrix or vector, E{} represents to carry out desired operation;

4b) according to linearly constrained minimum variance, the cost function that dimensionality reduction space-time adaptive is processed is expressed as:

$\left\{\begin{array}{c}\min w^H{R}_{z}w\\ s.t.w^{H}c(f_{s,0},f_{d,0})=1\end{array}\right.$

Wherein, R _zthe covariance matrix of data vector z during for empty after dimensionality reduction, dimension is r _mr _nr _k; r _mthe number of the launching beam that expression is chosen, r _nthe number of the received beam that expression is chosen, r _krepresent Doppler's number of active lanes of choosing; R wherein _mfor being more than or equal to 3 and be less than the integer of M, r _nfor being more than or equal to 3 and be less than the integer of N, r _kfor being more than or equal to 3 and be less than the integer of K, M is transmitting array number, and N is for receiving array number, and K is transponder pulse number, () ^hthe complex-conjugate transpose of representing matrix or vector;

4c) solve the cost function that dimensionality reduction space-time adaptive is processed, obtain the weight vector w of dimension-reduction treatment device:

$w=\frac{R_z^{-1}c(f_{s,0},f_{d,0})}{c^H(f_{s,0},f_{d,0})R_z^{-1}c(f_{s,0},f_{d,0})}$

Wherein, c (f _{s, 0}, f _{d, 0}) two-dimensional guide vector while being the target empty after dimensionality reduction, () ^hthe complex-conjugate transpose of representing matrix or vector, [] ^-1expression is to matrix inversion, f _{d, 0}for the Doppler frequency of target, f _{s, 0}for the normalization spatial frequency of target, R _zthe covariance matrix of data vector z during for empty after dimensionality reduction.