CN103675808A - Indistinguishable multi-target detection method for monopulse radar seekers - Google Patents

Abstract

The invention discloses an indistinguishable multi-target detection method for monopulse radar seekers. The method includes the steps of modeling multiple indistinguishable targets, and building signal vector and parameter vector relation; according to parameters of a built model, describing trace information by a particle set; using the particle set as a particle prior set, resampling the particle set, copying obtained particles to a posterior particle set, and repeating the operations Np times to obtain a posterior particles set of k-1 moment; calculating and normalizing weights of prior particles, and calculating marginal posterior density under target observation to obtain k-moment pointing vector of a radar antenna. The trace status information and particle filtering are effectively combined, the capacity of distinguishing multiple targets within same distance and a speed unit in radar beams is achieved, angular information of the multiple indistinguishable targets is acquired, angular tracking accuracy and stability of the radar seekers is improved, and angular glint resistance is also improved.

Description

A kind of monopulse radar seeker can not differentiate multi-target detection method

Technical field

The present invention relates to target following technology, what particularly relate to a kind of monopulse radar seeker can not differentiate multi-target detection method.

Background technology

No matter target following technology is to be all widely used as air defense, air traffic control, Air Missile defence, satellite monitoring, aerial target attack, airborne early warning, ground early warning system etc. at military field or civil area.Improve the hardware device of existing weaponry and equipment, such as adopting array antenna, high-resolution radar etc. can significantly improve target traceability, but monopulse radar is extensively equipped in guided weapon at present, if whole updating is regenerated, huge expense will be brought, therefore do not change existing weaponry and equipment, the multi-target detection that realizes target seeker is comparatively economic means.

The method of tradition monopulse radar seeker target detection mainly contains than three kinds of width methods (amplitude Directional Method), phase comparing method, phase-magnitude relative method.

In width method (amplitude Directional Method) monopulse radar, in order to determine the angular coordinate in a plane, just need to form two antenna radiation patterns that overlap mutually, and their the center line equity be respectively ± θ ° of angle that sense departs from by force.When target is when departing from equisignal direction θ place, by two wave beams, receive that the difference of vibration of signal represents the side-play amount of the strong sense of target equity, the symbol of amplitude difference represents that equisignal direction is with respect to the offset direction of target.When equisignal direction overlaps with target, by two wave beams, received that the amplitude of echoed signal equates, its difference just equals zero.

Phase comparing method is relatively carried out to determine the direction of target in a coordinate plane by the phase place of two antenna received signals.In region far away, two antennas are all irradiating the same space scope.Therefore, the signal being reflected by point target, is actually amplitude identical, and phase place is different.According to the phase differential of two echoed signals that antenna is received of separating, determine the angle of arrival.

In phase-magnitude relative method, adopt two to follow the tracks of coordinate, follow the tracks of in coordinate for one and use than width method, another is followed the tracks of in coordinate and uses phase comparing method.The method is except there is crosstalk item and may occurring fuzzy tracking point because of target noise, and other are with more identical than width, phase comparing method.

Above-mentioned three kinds of methods are traditional conventional target detection tracking method of monopulse radar, but these methods are restricted detecting in destination number, can detect at most two targets, can not meet the demand of modern battlefield to weaponry far away.

The deviation angle of tradition monopulse radar is the real part that multiple angle represents, and can not resolution target or the existence of multipath make its real part distortion, even if angle represents to produce error or mistake, thereby make it have imaginary part, this imaginary part can obtain easily by processing the quadrature phase part of difference signal.Therefore, two can not resolution target the measurement that can represent by the multiple angle in two recurrent intervals theoretically of angle, Amplitude Ratio and relative phase obtain, and thereby this multiple angle method for expressing obtains angle error without changing antenna radiation pattern, but the method by multiple matched filter, can extract at most two can not resolution target.

For multiobject test problems, be the thorny problem of the required solution of monopulse radar, the monopulse radar that is applied to be on active service adopts traditional ratio width mostly, than phase always, or further adopts maximum likelihood method, but detects at most only 2 targets.

In addition, angle scintillations problem is one of key factor affecting monopulse radar seeker target detection, angle tracking accuracy, when target seeker approaches target, the scattering strength of complex target different parts and the random variation of relative phase, must cause the distortion before target echo phase place random wave, the inclination of wavefront on receiving antenna bore face and random swing must cause angle error.Target angle flicker belongs to clarification of objective signal in itself, be the problem that tracking radar itself cannot overcome, the angle of target property research, all sizes and wavelength are comparable, any body target with two or more equivalent scattering centers, all can produce linear glint error.

Summary of the invention

The deficiency existing for overcoming above-mentioned prior art, what one of the present invention object was to provide a kind of monopulse radar seeker can not differentiate multi-target detection method, by by tracking mode information and the effective combination of particle filter, realized in the interior same distance of radar beam and speed unit multiobject resolution characteristic, obtain and a plurality ofly can not differentiate multiobject angle information, improve radar seeker angular tracking accuracy and degree of stability, improved the ability of anti-angle scintillations simultaneously.

What another object of the present invention was to provide a kind of monopulse radar seeker can not differentiate multi-target detection method, and it can reduce the calculated amount that signal is processed, and meets the requirement of monopulse radar seeker real-time performance of tracking.

For reaching above-mentioned and other object, the present invention proposes a kind of monopulse radar seeker can not differentiate multi-target detection method, comprises the steps:

Step 1, to differentiating multiple goal modeling, sets up the relation of signal vector and parameter vector;

Step 2, according to the parameter in institute's established model, adopts particle assembly ${\{x_{k-1}^{j,i},w_{k-1}^{j,i}\}}_{\begin{array}{cc}j=1& i=1\end{array}}^{\begin{array}{cc}{N}_p& N_{k-1}\end{array}}$ Trace information is described, wherein, N _k-1for following the tracks of number, by k-1 recurrence above, determined N _pfor the population of every secondary tracking, state vector x comprises motion state and the average radar cross section of target,

for weighting coefficient;

Step 3, the priori set using this particle assembly as particle, resamples to this particle assembly, and resulting particle copies in posteriority particle assembly, and repeats this operation N _pinferior, obtain k-1 posteriority particle assembly constantly ${\{x_{k-1}^{j,i},w_{k-1}^{j,i}\}}_{\begin{array}{cc}j=1& i=1\end{array}}^{\begin{array}{cc}{N}_p& N_{k-1}\end{array}};$

Step 4, weights the normalization of calculating priori particle, and by calculating the edge posterior density under target observation amount, obtain k radar antenna sensing constantly vector ${\hat{p}}_k={\left[\begin{array}{ccc}{\hat{r}}_k& {\hat{a}}_k& {\hat{e}}_k\end{array}\right]}^T,$ Wherein be respectively and point to distance, orientation, pitching.

Further, signal vector z=[s _id _aId _eIs _qd _aQd _eQ] ^t; Parameter vector φ ⁿ=[α ₀₁..., α _0N, η _a1..., η _aN, η _e1..., η _eN] ^t, wherein, s _i, s _qbe respectively I phase and Q phase with signal; d _aI, d _aQbe respectively I phase and the Q phase of gun parallax signal; d _eI, d _eQbe respectively I phase and the Q phase of trim signal; N is for can not differentiate number of targets; α _i, η _ai, η _eibe respectively the amplitude, orientation of signal to DOA, pitching to DOA.

Further, in step 1, hypothetical target is Si Weilin type, sets up the relation of signal vector and parameter vector.

Further, before step 4, adopt range gate to divide into groups to following the tracks of to detect, every group of corresponding validation matrix.

Further, in step 3, in each recurrence, adopt Bayesian model to select to determine detection event.

Compared with prior art, a kind of monopulse radar seeker of the present invention can not be differentiated multi-target detection method and can not differentiate on the basis of multi-objective Model in foundation, by trace information, realize and can not differentiate multiobject detection, realized the multi-target detection in same processing unit in radar beam, obtain a plurality of angle informations that can not resolution target, improve radar seeker angular tracking accuracy and degree of stability, improved the ability of anti-angle scintillations simultaneously.

Accompanying drawing explanation

Fig. 1 is the flow chart of steps that a kind of monopulse radar seeker of the present invention can not be differentiated multi-target detection method.

Embodiment

Below, by specific instantiation accompanying drawings embodiments of the present invention, those skilled in the art can understand other advantage of the present invention and effect easily by content disclosed in the present specification.The present invention also can be implemented or be applied by other different instantiation, and the every details in this instructions also can be based on different viewpoints and application, carries out various modifications and change not deviating under spirit of the present invention.

Fig. 1 is the flow chart of steps that a kind of monopulse radar seeker of the present invention can not be differentiated multi-target detection method.As shown in Figure 1, a kind of monopulse radar seeker of the present invention can not be differentiated multi-target detection method, comprises the steps:

Step 101, to differentiating multiple goal modeling.

Typical monopulse radar will adopt two independently receiving branchs to each directed coordinate plane, be two branch roads in aximuthpiston, two branch roads in pitch plane, therefore the present invention is measurement for angle in 2 D, adopt four stravismus subpulses, receiving end can obtain three signals, with signal s, gun parallax signal d _a, trim signal d _e, be expressed as:

$s=\frac{1}{2}(A+B+C+D)---\left(1\right)$

$d_a=\frac{1}{2}[(C+D)-(A+B)]---\left(2\right)$

$d_e=\frac{1}{2}[(A+C)-(B+D)]---\left(3\right)$

Above the signal of three passages through separate being in harmonious proportion after matched filtering, be output as:

$s_I=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\α}}_i\cos {\mathrm{\φ}}_i+n_{\mathrm{sI}}---\left(4\right)$

$s_Q=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\α}}_i\sin {\mathrm{\φ}}_i+n_{\mathrm{sQ}}---\left(5\right)$

$d_{\mathrm{aI}}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\η}}_{\mathrm{ai}}{\mathrm{\α}}_i\cos {\mathrm{\φ}}_i+n_{\mathrm{daI}}---\left(6\right)$

$d_{\mathrm{aQ}}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\η}}_{\mathrm{ai}}{\mathrm{\α}}_i\sin {\mathrm{\φ}}_i+n_{\mathrm{daQ}}---\left(7\right)$

$d_{\mathrm{eI}}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\η}}_{\mathrm{ei}}{\mathrm{\α}}_i\cos {\mathrm{\φ}}_i+n_{\mathrm{deI}}---\left(8\right)$

$d_{\mathrm{eQ}}=\underset{i=1}{\overset{N}{\mathrm{\Σ}}}{\mathrm{\η}}_{\mathrm{ei}}{\mathrm{\α}}_i\sin {\mathrm{\φ}}_i+n_{\mathrm{deQ}}---\left(9\right)$

Wherein, s _i, s _qbe respectively I phase and Q phase with signal; d _aI, d _aQbe respectively I phase and the Q phase of gun parallax signal; d _eI, d _eQbe respectively I phase and the Q phase of trim signal; N is for can not differentiate number of targets; α _i, η _ai, η _eibe respectively the amplitude, orientation of signal to DOA(Direction of Arrival, Bo Dajiao), pitching is to DOA; φ _iit is the phasing degree of i target; Channel noise n _sI, n _sQfor zero-mean, variance is gaussian distribution; Channel noise n _daI, n _daQ, n _deI, n _deQfor zero-mean, variance is

gaussian distribution.

According to formula (4)-(9), can obtain signal vector z=[s _id _aId _eIs _qd _aQd _eQ] ^twith parameter vector φ ⁿ=[α ₀₁..., α _0N, η _a1..., η _aN, η _e1..., η _eN] ^t.Then hypothetical target is Si Weilin (Swerling) type, thereby sets up the relation of signal vector and parameter vector.

The signal model obtaining by above-mentioned approach, has actual operability, is convenient to analyze and subsequent treatment.

Step 102, adopts particle assembly ${\{x_{k-1}^{j,i},w_{k-1}^{j,i}\}}_{\begin{array}{cc}j=1& i=1\end{array}}^{\begin{array}{cc}{N}_p& N_{k-1}\end{array}}$ (wherein, N _k-1for following the tracks of number, by k-1 recurrence above, determined; N _ppopulation for every secondary tracking; State vector x comprises motion state and the average radar cross section of target;

for weighting coefficient) trace information is described, this descriptor has determined that next residence time internal antenna points to.

Step 103, the priori set using this particle assembly as particle, resamples to this set, and resulting particle copies in posteriority particle assembly, repeats this operation N _pinferior, obtain k-1 posteriority particle assembly constantly ${\{x_{k-1}^{j,i},w_{k-1}^{j,i}\}}_{\begin{array}{cc}j=1& i=1\end{array}}^{\begin{array}{cc}{N}_p& N_{k-1}\end{array}}.$

Step 104, weights the normalization of calculating priori particle.Now, by calculating the edge posterior density under target observation amount

can obtain k radar antenna sensing constantly vector ${\hat{p}}_k={\left[\begin{array}{ccc}{\hat{r}}_k& {\hat{a}}_k& {\hat{e}}_k\end{array}\right]}^T$ (wherein

be respectively and point to distance, orientation, pitching).

Although particle filter technology has good effect to processing non-Gauss's nonlinear problem, the shortcomings such as it exists, and particle is degenerated, calculated amount is large, calculation of complex.In order to realize validity and the real-time of particle filter multi-target detection, the present invention adopts method for resampling in step 103, and the particle of low weighting is rejected, and the particle that those high multiplied by weight obtain has identical weight, thereby weakens particle decline.According to radar return, can obtain all types of target and detect event, this has just increased the burden that signal is processed.Meanwhile, in order to reduce calculated amount, before step 104, service range door of the present invention divides into groups to follow the tracks of detecting, every group of corresponding validation matrix, and by validation matrix can build likely the difference of target check event suppose.In each recurrence of particle filter, adopt Bayesian model to select to determine detection event, be conducive to like this calculate multiple integral by Monte Carlo method.

For Efficient Evaluation the present invention's the quality that can not differentiate multi-target detection method, the present invention utilizes Cramer-Rao lower limit (CRLB) to evaluate unbiased estimator, and the variance that is unbiased estimator is determined a lower limit.According to CRLB theorem, derive by the required satisfied minimum value of the variance of estimated parameter, for analyzing tracking error performance.Meanwhile, the present invention utilizes Matlab to carry out emulation to the present invention, and in outfield, the present invention is carried out to principle checking.

Utilize Matlab to carry out emulation to the present invention's the multi-target detection method of can not differentiating, adopt and test with the following method and checking:

The first, suppose a target, two can not resolution target, three can not three kinds of situations of resolution target under the performance of analytical algorithm, and utilize the CRLB deriving to assess detection algorithm, draw tracking error curve;

The second, in theory maximum likelihood estimator module can only estimate two can not resolution target, and helpless for a plurality of estimations that can not resolution target.For the performance of the algorithm that project proposes is described better, we analyze this algorithm and maximum likelihood estimator module.This analysis can not be differentiated under target conditions and carry out two, three of existence, draws the graph of errors of two kinds of algorithms, compares the real-time of two kinds of algorithms simultaneously;

The 3rd, it is distinguishable supposing to start most three targets, and lasting close to each other with linear movement, and then wherein two targets continue coordinate turns.Within the duration, three targets start to become can not resolution target, until three targets become completely can not resolution target after, target starts to continue coordinate turn, then does to continue linear movement.Under this scene, the tracking performance of simulation analysis algorithm;

The 4th, suppose the current target that only exists, then there is another one target, two target trajectories are parallel, and can not differentiate, and analyze algorithm under this scene and detect performance.

On the basis of Matlab theoretical simulation, make full use of lab resources, algorithm is carried out to field trial checking.Adopting four circuit receiver is four single-conversion double sideband receivers that magnitude-phase characteristics is identical, and four circuit receiver shares the phase-locked local oscillator of same 3GHz and carries out mixing, and receiver output frequency is 20MHz, bandwidth is the radiofrequency signal of ± 3MHz.First with E4422B type signal generator, produce the high-frequency signal of 2.98GHz, then with loudspeaker emitting antenna, this signal amplitude is shot out.Arrange three and can not differentiate multiple goal, its RCS is different, i.e. a corner reflector, a spherical metal, a rectangular metal body, and according to default rule, target is placed in different distance.Can not differentiate multiobject radiation signal, through antenna, enter receiver, utilize data collecting card to obtain raw data, finally utilize the present invention to carry out the estimation that object wave reaches angle DOA.

The present invention passes through Monte Carlo simulation 25000 times, has following result:

(1) when only there is a target, and signal to noise ratio snr=20dB, detection probability is 90%, false-alarm probability is 1%;

(2) when there is two targets, and the signal to noise ratio snr=13dB of two targets, detection probability is %92, and false-alarm probability is 1%, and the size of detection probability depends on the angle of two target separation;

(3) when there is three targets, and the signal to noise ratio snr=20dB of target 1, the signal to noise ratio snr=13dB of target 2 and target 3, detection probability is 94%, false-alarm probability is 5%.

Result based on Monte Carlo simulation, in field trial, three targets are respectively corner reflector, spherical metal, rectangular metal body.Basic and the Monte Carlo simulation of test findings comes to the same thing, but need to particularly point out, and when signal to noise ratio (S/N ratio) is less than 13dB, and the ripple of target reaches the half beam width that angle DOA departs from antenna, and false-alarm probability can increase to some extent.

In sum, the multi-target detection method of can not differentiating a kind of monopulse radar seeker of the present invention has realized a kind ofly can effectively carry out a plurality of methods that can not differentiate target detection, the method can make do not changing under existing weaponry and equipment configuration prerequisite, realize monopulse radar seeker to 3 and above multiobject detection, and can effectively improve anti-angle scintillations and the detectability to integrated target.

Above-described embodiment is illustrative principle of the present invention and effect thereof only, but not for limiting the present invention.Any those skilled in the art all can, under spirit of the present invention and category, modify and change above-described embodiment.Therefore, the scope of the present invention, should be as listed in claims.

Claims (5)

1. monopulse radar seeker can not be differentiated a multi-target detection method, comprises the steps:

Step 1, to differentiating multiple goal modeling, sets up the relation of signal vector and parameter vector;

Step 2, according to the parameter in institute's established model, adopts particle assembly ${\{x_{k-1}^{j,i},w_{k-1}^{j,i}\}}_{\begin{array}{cc}j=1& i=1\end{array}}^{\begin{array}{cc}{N}_p& N_{k-1}\end{array}}$ Trace information is described, wherein, N _k-1for following the tracks of number, by k-1 recurrence above, determined N _pfor the population of every secondary tracking, state vector x comprises motion state and the average radar cross section of target,

for weighting coefficient;

Step 3, the priori set using this particle assembly as particle, resamples to this particle assembly, and resulting particle copies in posteriority particle assembly, and repeats this operation N _pinferior, obtain k-1 posteriority particle assembly constantly ${\{x_{k-1}^{j,i},w_{k-1}^{j,i}\}}_{\begin{array}{cc}j=1& i=1\end{array}}^{\begin{array}{cc}{N}_p& N_{k-1}\end{array}};$

Step 4, weights the normalization of calculating priori particle, and by calculating the edge posterior density under target observation amount, obtain k radar antenna sensing constantly vector ${\hat{p}}_k={\left[\begin{array}{ccc}{\hat{r}}_k& {\hat{a}}_k& {\hat{e}}_k\end{array}\right]}^T,$ Wherein

be respectively and point to distance, orientation, pitching.

2. a kind of monopulse radar seeker as claimed in claim 1 can not be differentiated multi-target detection method, it is characterized in that: signal vector z=[s _id _aId _eIs _qd _aQd _eQ] ^t; Parameter vector φ ⁿ=[α ₀₁..., α _0N, η _a1..., η _aN, η _e1..., η _eN] ^t, wherein, s _i, s _qbe respectively I phase and Q phase with signal; d _aI, d _aQbe respectively I phase and the Q phase of gun parallax signal; d _eI, d _eQbe respectively I phase and the Q phase of trim signal; N is for can not differentiate number of targets; α _i, η _ai, η _eibe respectively the amplitude, orientation of signal to DOA, pitching to DOA.

3. a kind of monopulse radar seeker as claimed in claim 2 can not be differentiated multi-target detection method, it is characterized in that: in step 1, hypothetical target is Si Weilin type, sets up the relation of signal vector and parameter vector.

4. a kind of monopulse radar seeker as claimed in claim 3 can not be differentiated multi-target detection method, it is characterized in that: before step 4, adopt range gate to divide into groups to following the tracks of to detect, every group of corresponding validation matrix.

5. a kind of monopulse radar seeker as claimed in claim 4 can not be differentiated multi-target detection method, it is characterized in that: in step 3, adopt Bayesian model to select to determine detection event in each recurrence.

Images