CN104502907A - Stable ground moving/static target tracking method for airborne radar - Google Patents

Abstract

The invention relates to a stable ground moving/static target tracking method for airborne radar. According to the method, nonlinear frequency-modulated and stepped-frequency signals are simultaneously emitted to respectively detect moving/static targets according to the characteristics of the moving/static targets; the nonlinear frequency-modulated signals are mainly used for detecting the moving targets, and the stepped-frequency signals are mainly used for detecting the static targets; one-time point trajectories of the moving/static targets on the ground can be simultaneously obtained in one scan frame of the radar. Compared with the traditional methods, the method has the advantages that the moving/static targets on the ground can be noticed simultaneously, the overall situation of the targets on the ground is reflected, continuous tracking can be carried out when the targets become static from moving or become moving from static, and the continuity and stability of tracking of the moving/static targets on the ground are guaranteed.

Description

A kind of airborne radar ground sound target tenacious tracking method

Technical field

Belong to the present invention Area Objects data processing method, be under the jurisdiction of Radar Signal Detection and technical field of data processing, be specifically related to a kind of airborne radar ground sound target tenacious tracking method.

Background technology

There is notable difference in airborne radar over the ground multiple target tracking, technical difficulty is large compared with traditional air search radar.First, there is relative motion between airborne radar platform and target, target following is not only relevant with the motion state of target, but also relevant with the motion state of carrier aircraft self; Secondly, because tracking target is terrain object, there is stronger Clutter Background, make detections of radar target difficult; Finally, because ground static target and moving target, maneuvering target and nonmaneuvering target exist simultaneously, target tenacious tracking difficulty is significantly increased.

(1) target travel environment: aerial target motion is at three dimensions, and terrain object can be equivalent to two dimensional motion usually among a small circle, and, for the target of moving on road, motion in one dimension can be equivalent in certain stage, until run into road junction, now its direction of motion may change.Additionally the moving region of Area Objects can be subject to more restriction usually, and such as severe topographic condition makes target to pass through.Although aerial target also can run into the stops such as massif, the external environment condition restriction suffered by it is far less than terrain object.

(2) target travel characteristic: aerial target is except helicopter and dirigible, all be subject to the restriction of minimum speed, target lower than this restriction cannot run well, and terrain object then can be accelerated, slow down, completely static or within a period of time, keep motion state constant.Therefore the changeability of terrain object motion will far above aerial target.

(3) heavy dense targets degree: unless some special circumstances, usual aerial target can keep certain safe distance.And between terrain object, can move with very near distance, time static can also one connect a ground and park, therefore the closeness of terrain object is far above aerial target.

(4) sensor detection probability: due to terrain shading, sensor possibly cannot observe terrain object, there is minimum detection radial velocity (MDV) in Ground moving target indication (GMTI) in addition, when target is static or cannot detect with sensor when comparatively low velocity is moved, therefore the detection of sensor may be discontinuous.

(5) clutter: terrain object is moved in the environment that clutter is comparatively serious, the clutter recognition process adopted at present also cannot reach close to desirable inhibition, and the reduction and false-alarm number object that cause detection probability increase by the existence of clutter.

Airborne radar, when detecting on a surface target, needs to pay close attention to ground moving target and quiet target simultaneously.Traditional data processing method is subject to processing level and technical conditions restriction, usually by moving-target and the separately process of quiet target, the method can not the overall situation of Area Objects fully corresponsively, especially when target by move quiet or by quiet to dynamic motion time, can track rejection be caused.

Along with microelectric technique development, chip processing capabilities improves constantly, the continuous appearance of double-core, multi-core CPU and application, the hardware bottleneck limiting multiple batches of terrain object process disappears, develop a kind of ground sound target data disposal route, ensure that the dynamic and static in stable condition tracking of terrain object becomes possibility.

Summary of the invention

In order to avoid the deficiencies in the prior art part, the present invention proposes a kind of airborne radar ground sound target tenacious tracking method

A kind of airborne radar ground sound target tenacious tracking method, is characterized in that step is as follows:

Step 1: radar emission NLFM signal, for detecting moving-target, is launched stepped frequency signal and detected quiet target, obtain the once some mark of ground moving target and quiet target at a scanning frame of radar simultaneously;

Step 2: the once some mark carrying out moving-target and quiet target respectively carries out secondary lobe and weighting agglomeration process;

Step 3: adopt the track initiation method based on two-stage Hough transform that congealing point that is dynamic and quiet target is set up respective alternative flight path respectively; Alternative for former frame flight path is carried out motion compensation, compensates to present frame carrier aircraft position;

Step 4: the position of prediction present frame flight path: $\hat{K}(k/k-1)=F\hat{X}(k-1/k-1),$ Wherein: $F=\left[\begin{array}{ccc}1& T& T^2/2\\ 0& 1& T\\ 0& 0& 1\end{array}\right],$ $X={\left[\begin{array}{ccc}x& \stackrel{\·}{x}& \stackrel{\·\·}{x}\end{array}\right]}^T,$ with represent the position of present frame flight path respectively, speed and component of acceleration;

Step 5: repeat step 1 ~ step 2, obtains the congealing point of next some mark;

Step 6: centered by the predicted position of flight path, is associated the metric data and flight path that fall into Bo Mennei, obtains correlation combiner;

If metric data does not drop on the Bo Mennei of any one flight path, set up a new flight path with these data;

If there is no observation station at the Bo Mennei of flight path, in two kinds of situation:

Situation 1: when moving-target flight path is all uncorrelated with all moving-target congealing points, then carries out two hypothesis tracking, obtain correlation combiner;

Suppose 1: target is undetected, then flight path is extrapolated according to the characteristics of motion before;

Suppose 2: target is static, then that carries out between moving-target flight path to quiet target congealing point is relevant, if there is reference point, then carries out flight path renewal with it, if do not have reference point, then the position keeping flight path original is motionless, and speed is zero;

Situation 2: when quiet targetpath is all uncorrelated with all quiet target congealing points, then carry out two hypothesis tracking;

Suppose 1: target is undetected, then the position keeping flight path original is motionless, and speed is zero, and target is still static;

Suppose 2: target transfers motion to, that then carries out between quiet targetpath to moving-target congealing point is relevant, if there is reference point, then carry out flight path renewal with it, and estimate the speed of flight path, if do not have reference point, then the position keeping flight path original is motionless, wait for the relevant of next frame, obtain correlation combiner;

The correlation combiner of all moving-target flight paths and quiet targetpath is participated in computing according to the following formula, finds the optimum allocation of an overall Least-cost, carry out the final updated of flight path with it;

a ^*(k)＝arg min _a(k)C(k|a(k))

Wherein:

$C\left(k\right|a\left(k\right))={\mathrm{\Σ}}_{m=0_1,0_2,1}^{M\left(k\right)}{\mathrm{\Σ}}_{n=1}^{N(k-1)}\mathrm{\α}(k,m,n)c(k,m,n)$

Wherein:

C (k, m, n) is the cost value of distributing a (k, m, n): $c(k,m,n)=-\ln \left(\frac{\mathrm{\Φ}(k,m,n)}{\mathrm{\Φ}(k,m,0)}\right)$

When situation 1: Φ (k, m, n) is calculated as follows:

$\mathrm{\&Phi;}(k,m,n)=\left\{\begin{array}{cc}{P}_{D}P\left\{\right|v_r\left(k\right)|\&GreaterEqual;v_0\}\mathrm{\&Lambda;}(k,m,n)& m>0,n>0\\ \mathrm{\&lambda;}& m>0,n=0\\ (1-P_D)P\left\{\right|v_r\left(k\right)|\&GreaterEqual;v_0\}\mathrm{\&lambda;}& m=0_1,n>0\\ P\left\{\right|v_r\left(k\right)|<v_0\}\mathrm{\&lambda;}& m=0_2,n>0\end{array}\right.$

Wherein: m > 0, n > 0 is relevant to flight path for measuring; M > 0, n=0 is uncorrelated with already present flight path for measuring, and is false-alarm; M=0 ₁, n > 0 is for supposing that measurement 1 is relevant to flight path; M=0 ₂, n > 0 is for supposing that measurement 2 is relevant to flight path;

Λ (k, m, n) is the combined probability that Interactive Multiple-Model is estimated, v ₀minimum detectable, vr ₍k) be that the radial velocity of a kth scanning frame is not (if a kth scanning frame has measured value, vr ₍k) be predicted by kth-1 scanning frame to get).Suppose speed Gaussian distributed, then can calculate P{|v _r(k) |>=v ₀and P{|v _r(k) | < v ₀;

The space density λ of false-alarm is

Wherein P _fAbe false-alarm probability, the measuring accuracy of distance, orientation, distance speed three-dimensional is respectively with

When situation 2, Φ (k, m, n) is calculated as follows:

$\mathrm{\&Phi;}(k,m,n)=\left\{\begin{array}{cc}{P}_{D}P\left\{\right|v_r\left(k\right)|=\&GreaterEqual;v_0\}\mathrm{\&Lambda;}(k,m,n)& m>0,n>0\\ \mathrm{\&lambda;}& m>0,n=0\\ (1-P_D)P\left\{\right|v_r\left(k\right)|=0\&GreaterEqual;\}\mathrm{\&lambda;}& m=0_1,n>0\\ P\left\{\right|v_r\left(k\right)|!=0\}\mathrm{\&lambda;}& m=0_2,n>0\end{array}\right.$

Situation 2 needs to calculate P{|v _r(k) |=0} and P{|v _r(k) | unequal to 0};

Described ripple door comprises range gate, orientation Bo Men, pitching ripple door and Doppler's ripple door;

Step 7, filtering: according to the filtering method of alpha-beta-γ, carry out filtering to flight path

$\hat{X}(k/k)=\hat{X}(k/k-1)+K[y\left(k\right)-H\hat{X}(k/k-1)]$

$\hat{X}(k/k-1)=F\hat{X}(k-1/k-1)$

Wherein gain matrix $K=\left[\begin{array}{c}\mathrm{\&alpha;}\\ \mathrm{\&beta;}/T\\ \mathrm{\&gamma;}/T^2\end{array}\right],$ Observing matrix H=[1 0 0];

Step 8, flight path manage: when a flight path does not repeatedly have observation station to cause repeatedly to extrapolate, deleted by flight path; Repeat step 1 ~ step 8, carry out sound target tenacious tracking.

Described range gate 20 ~ 50m.

1.5 ~ 2.5 degree, described orientation ripple door.

Described 2 ~ 4 degree, pitching ripple door.

Described 3 ~ 6, Doppler's ripple door.

A kind of airborne radar ground sound target tenacious tracking method that the present invention proposes, according to dynamic and static target characteristic, adopts transmitting nonlinear frequency modulation and stepped frequency signal simultaneously to detect dynamic and static target respectively.NLFM signal is mainly for detection of moving-target, and stepped frequency signal is mainly for detection of quiet target.The once some mark of ground moving target and quiet target can be obtained at a scanning frame of radar simultaneously.

Compared with traditional method, the present invention can pay close attention to ground moving target and quiet target simultaneously, corresponsively the overall situation of Area Objects, when target by move quiet or by quiet to continuing time dynamic to follow the tracks of, ensure that the continous-stable of ground dynamic and static target is followed the tracks of.

Accompanying drawing explanation

Fig. 1: be the rate curve of static-dynamic-dynamic motion of target

Fig. 2: the geometric locus being static-dynamic-dynamic motion of target, square frame is the static position of target

Fig. 3: be the rate curve of quiet-static-dynamic motion of target

Fig. 4: be the geometric locus of quiet-static-dynamic motion of target, square frame is the static position of target

Fig. 5: be the relation schematic diagram of carrier aircraft NED and carrier aircraft LOS coordinate system

Fig. 6: be the related procedure that two hypothesis are followed the tracks of

Embodiment

Now in conjunction with the embodiments, the invention will be further described for accompanying drawing:

Radar data process, mainly for be aerial target and naval target, and terrain object mobility strong, noise jamming is large, targeted species is many, same scope internal object blocks etc., and influence factor is more, to detection, the merger of target, condenses, build boat and be proposed higher requirement.Need on the basis of radar data disposal route in the past, for terrain object feature, propose a kind ofly to be applicable to strong maneuvering target, and can simultaneously to the method for sound target tenacious tracking.Namely to Fig. 1 ~ Fig. 4 tetra-width figure, target is quiet again to dynamic by moving, or target by quiet to moving again to quiet, can tenacious tracking.

According to dynamic and static target characteristic, transmitting nonlinear frequency modulation and stepped frequency signal is simultaneously adopted to detect dynamic and static target respectively.NLFM signal is mainly for detection of moving-target, and stepped frequency signal is mainly for detection of quiet target.The once some mark of ground moving target and quiet target can be obtained at a scanning frame of radar simultaneously.

1) two hypothesis trackings

The basic ideas of two hypothesis trackings, when a target is when this frame is without observation station, are divided into Liang Ge branch according to two kinds of hypothesis, finally in distribution of being all correlated with, finds one group of optimum allocation to carry out more fresh target.

Situation one:

When a moving target does not detect at a certain frame, there are two kinds of possibilities, a kind of is setting due to environment or detection threshold, cause target undetected, another kind is because target velocity is too small, be less than Minimum detectable, cannot be detected by moving target detect, target appears in quiet target detection.

Situation two:

When a static target does not detect at a certain frame, have two kinds of possibilities, a kind of is setting due to environment or detection threshold, cause target undetected, another kind is because target transfers motion to, and the position before have left, target appears in moving target detect.

For above two kinds of situations, two hypothesis trackings can be adopted.

A) target travel and measurement model

T _kthe state value of moment target is

$X\left(t_k\right)=\left[\begin{array}{c}x\left(t_k\right)\\ \stackrel{\&CenterDot;}{x}\left(t_k\right)\\ y\left(t_k\right)\\ \stackrel{\&CenterDot;}{y}\left(t_k\right)\end{array}\right]---\left(1\right)$

Wherein x (t _k) and y (t _k) be respectively cartesian coordinate system x, the position of y-axis, with be respectively x, the speed of y-axis.

The state model of target is

X(t _k)＝F(t _k)X(t _k-1)+Γ(t _k)v(t _k-1) (2)

Wherein F is state-transition matrix, and Γ is excitation matrix, v (t _k-1) be process noise.

T _kmoment measures vector and comprises distance, azimuth-range speed and doppler information.

$Z\left(t_k\right)=\left[\begin{array}{c}r\left(t_k\right)\\ \mathrm{\&theta;}\left(t_k\right)\\ \stackrel{\&CenterDot;}{r}\left(t_k\right)\end{array}\right]+W\left(t_k\right)---\left(3\right)$

Wherein W (t _k) be measurement noises vector.

B) two hypothesis track side ratio juris

The subject matter that two hypothesis trackings will solve is, the optimum attaching problem of measured value M (k) of kth frame and the flight path T (k-1) of kth-1 frame.It is minimum that optimum is exactly the overall associated costs realizing measuring between flight path.

Define a scale-of-two and distribute variable a (k, m, n),

Make the complete or collected works of a (k) for distributing, then

$a\left(k\right)=\left\{\begin{array}{c}\mathrm{\&alpha;}(k,m,n):m=0_1,0_2,\mathrm{1,2},...,M\left(k\right);\\ n=\mathrm{0,1},...,N(k-1)\end{array}\right\}---\left(5\right)$

Wherein M (k) is the sum of sound target detection measured value, and N (k-1) is the number of flight path.

Comprise distribution a (k, 0 that the first hypothesis is measured ₁, n) represent owing to not detecting, flight path T _nk () to measure M (k) all uncorrelated with any one.

Comprise distribution a (k, 0 that the second hypothesis is measured ₂, n), represent because target is static or radial velocity is less than Minimum detectable and causes flight path T in situation one _nk () measures all uncorrelated with any one moving-target in M (k); Represent in situation two and cause flight path T because target travel have left original position _nk () is all uncorrelated with any one quiet target measurement in M (k).

Distribute a (k, m, 0) and represent measurement m all uncorrelated with any one already present flight path (assuming that it is a false-alarm point).

Our object finds an optimum allocation a ^*k (), it can make overall associated costs minimum, namely

a ^*(k)＝arg min _a(k)C(k|a(k)) (6)

Wherein

$C\left(k\right|a\left(k\right))={\mathrm{\&Sigma;}}_{m=0_1,0_2,1}^{M\left(k\right)}{\mathrm{\&Sigma;}}_{n=1}^{N(k-1)}\mathrm{\&alpha;}(k,m,n)c(k,m,n)---\left(7\right)$

Wherein c (k, m, n) is the cost value of distributing a (k, m, n).

$c(k,m,n)=-\ln \left(\frac{\mathrm{\&Phi;}(k,m,n)}{\mathrm{\&Phi;}(k,m,0)}\right)---\left(8\right)$

Situation one:

$\mathrm{\&Phi;}(k,m,n)=\left\{\begin{array}{cc}{P}_{D}P\left\{\right|v_r\left(k\right)|\&GreaterEqual;v_0\}\mathrm{\&Lambda;}(k,m,n)& m>0,n>0\\ \mathrm{\&lambda;}& m>0,n=0\\ (1-P_D)P\left\{\right|v_r\left(k\right)|\&GreaterEqual;v_0\}\mathrm{\&lambda;}& m=0_1,n>0\\ P\left\{\right|v_r\left(k\right)|<v_0\}\mathrm{\&lambda;}& m=0_2,n>0\end{array}\right.---\left(9\right)$

Wherein m > 0, n > 0 is relevant to flight path for measuring; M > 0, n=0 is uncorrelated with already present flight path for measuring, i.e. false-alarm; M=0 ₁, n > 0 is for supposing that measurement 1 is relevant to flight path (due to P _d< 1); M=0 ₂, n > 0 is hypothesis measurement 2 relevant to flight path (due to Minimum detectable restriction).

Λ (k, m, n) is the combined probability that Interactive Multiple-Model is estimated, v ₀minimum detectable, v _rk () is that the radial velocity of a kth scanning frame is not (if a kth scanning frame has measured value, v _rk () is predicted by kth-1 scanning frame to get).Suppose speed Gaussian distributed, then can calculate P{|v _r(k) |>=v ₀and P{|v _r(k) | < v ₀.

The space density λ of false-alarm is

$\mathrm{\&lambda;}=\frac{P_{\mathrm{FA}}}{{12}^{3/2}{\mathrm{\&sigma;}}_r{\mathrm{r\&sigma;}}_{\mathrm{\&theta;}}{\mathrm{\&sigma;}}_{\stackrel{\&CenterDot;}{r}}}---\left(10\right)$

Wherein P _fAbe false-alarm probability, the measuring accuracy of distance, orientation, distance speed three-dimensional is respectively with

Situation two:

$\mathrm{\&Phi;}(k,m,n)=\left\{\begin{array}{cc}{P}_{D}P\left\{\right|v_r\left(k\right)|=\&GreaterEqual;v_0\}\mathrm{\&Lambda;}(k,m,n)& m>0,n>0\\ \mathrm{\&lambda;}& m>0,n=0\\ (1-P_D)P\left\{\right|v_r\left(k\right)|=0\&GreaterEqual;\}\mathrm{\&lambda;}& m=0_1,n>0\\ P\left\{\right|v_r\left(k\right)|!=0\}\mathrm{\&lambda;}& m=0_2,n>0\end{array}\right.---\left(11\right)$

Situation two needs to calculate P{|v _r(k) |=0} and P{|v _r(k) | unequal to 0}.

C) restrictive condition of two hypothesis tracking

A () confirms.Measuring for one only may be relevant to a flight path, and must meet

${[Z_m\left(t_k\right)-{\hat{Z}}_m\left(t_k\right)]}^{\&prime;}{S}_m{\left(t_k\right)}^{-1}[Z_m\left(t_k\right)-{\hat{Z}}_m\left(t_k\right)]\&le;\mathrm{\&gamma;}---\left(12\right)$

γ confirms thresholding, prediction measured value, S _m(t _k) be covariance.

(b) man-to-man restriction.A flight path is measured relevant to one at most, and supposes that flight path can be relevant to any number of measurement.A flight path is distributed at most in a measurement, and hypothesis measures (m=0 ₁or m=0 ₂) multiple flight path can be distributed to.

C () non-NULL is correlated with.Suppose that measurement can not distribute to hypothesis flight path.

D) state updating

Situation one:

Suppose not measure at kth frame to upgrade flight path, then from k-1 frame to k frame, prediction obtains the state of flight path with covariance P (k|k-1).Radial velocity obeys average variance is P _r(k|k-1) Gaussian distribution.If flight path is relevant to hypothesis measurement 1 (not detecting), so the state estimation of target current time is the predicted value of last scanning frame.If flight path is relevant to hypothesis measurement 2 (Minimum detectable limits), so the position of target keeps motionless, and speed is zero.

Situation two:

Suppose not measure at kth frame to upgrade flight path, then no matter flight path is relevant to hypothesis measurement 1 (not detecting), or relevant to hypothesis measurement 2 (target transfers motion to), and all keep the position of target motionless, speed is zero.

Embodiment:

1) launch nonlinear frequency modulation and stepped frequency signal simultaneously, obtain the once some mark of ground moving target and quiet target;

2) target agglomeration process

Carry out the agglomeration process that moving-target and quiet target are once put respectively, all comprise secondary lobe and two steps are condensed in weighting.

● remove secondary lobe

First delimit the scope of a Distance geometry speed, carry out Amplitude Ratio comparatively for the point within the scope of this, the excessive point of amplitude difference thinks secondary lobe, is deleted.

$\frac{F\left(i\right)}{F\left(j\right)}>D---\left(13\right)$

● amplitude weighting

Because radar beam is when continuous sweep, wave beam lobe has one fixed width, several pulses are had at least to sweep to target continuously, the corresponding orientation values of each pulse, same target is caught in repeatedly, orientation values repeatedly during target acquisition is all different, and it is larger that this just causes azimuthal splitting degree.Therefore multiple Plot coherence of target same in single pass are needed to become a some mark.

By delimiting some M altogether at the four-dimensional Bo Mennei of distance, orientation, pitching and Doppler simultaneously, carry out amplitude weighting.

The distance of a jth congealing point of the i-th frame is:

$r_i\left(j\right)=\frac{{\mathrm{\&Sigma;}}_{k=0}^{M}r\left(k\right)*F\left(k\right)}{{\mathrm{\&Sigma;}}_{k=0}^{M}F\left(k\right)}---\left(14\right)$

Wherein, M is once counting of four-dimensional Bo Mennei, the distance that r (k) once puts for kth, the amplitude that F (k) once puts for kth, r _ij () is the distance of a jth congealing point of the i-th frame.

3) boat is built

The congealing point of dynamic and static target is set up alternative flight path; Adopt the track initiation method based on two-stage Hough transform.

● choose displacement---the time, as parameter, does as down conversion to twice adjacent measurement:

$\begin{array}{c}{\mathrm{\&Delta;}}_{\mathrm{\&rho;i}}\left[{\mathrm{\&theta;}}_n\right]={\mathrm{\&rho;}}_{i+1}\left[{\mathrm{\&theta;}}_n\right]-{\mathrm{\&rho;}}_i\left[{\mathrm{\&theta;}}_n\right]=(x_{i+1}-x_i){\cos \mathrm{\&theta;}}_n\\ +k(t_{i+1}-t_i)\sin {\mathrm{\&theta;}}_n\end{array}---\left(15\right)$

Wherein

${\mathrm{\&theta;}}_n=(n-\frac{1}{2}-\frac{N_{\mathrm{\&theta;}}}{2})\mathrm{\&pi;}/N_{\mathrm{\&theta;}},(n=\mathrm{1,2},...,N_{\mathrm{\&theta;}})---\left(16\right)$

N in formula _θfor the segmentation hop count of parameter θ.

K=(v _max+ α _maxt)/12 is scale-up factor, v _maxand α _maxbe respectively maximal rate and the peak acceleration of target, T is the sampling period.

Find θ _n=θ ₀, (n=1,2 ..., N _θ), make Δ ρ _i[θ _n]=0.

● first order Hough transform

First to x _iand y _ido the Hough transform based on motion state.

If flight path meets with then think that this flight path may be reliable flight path, if do not met, then this flight path is false track, directly deletes from alternative flight path.

● second level Hough transform

Owing to being greater than 1km place, oblique distance r _iwith time t _isubstantially linear, therefore can to oblique distance r _ido the Hough transform based on motion state, if flight path meets then this flight path can be initial.

4), when next frame once puts mark arrival, the cohesion of dynamic and static target is carried out equally respectively;

5) alternative for former frame flight path is carried out motion compensation, compensate to present frame carrier aircraft position;

The history point mark of flight path comprises two steps (with reference to Fig. 5) to the carrier aircraft NED down conversion of this frame: (1) oblique distance compensates: oblique distance should according to displacement but not distance compensate.In fact, need the displacement compensated should be the air line distance of carrier aircraft between two frames between corresponding moment NED coordinate system (treating compensation point mark) initial point, calculate the oblique distance of target under current carrier aircraft NED again according to sine, this computing also obtains the azimuth deviation introduced by carrier aircraft displacement.In fact, this displacement calculating means is, all previous move distance of all inertial guidance data between two frame initial times is projected to this motion vector direction (being determined by the crab angle of two frame initial times) and sued for peace.(2) orientation compensates: need to carry out in two steps, first the orientation values of some mark needs driftage declinate filtered between two frames (to treat compensation point mark, the crab angle difference of the corresponding moment NED of consecutive frame) deduct, its secondary azimuth deviation carrier aircraft displacement introduced compensates.

6) predict

According to the characteristics of motion of flight path, the position of prediction present frame flight path;

$\hat{X}(k*k-1)=F\hat{X}(k-1/k-1)---\left(17\right)$

Wherein $F=\left[\begin{array}{ccc}1& T& T^2/2\\ 0& 1& T\\ 0& 0& 1\end{array}\right],$ $X={\left[\begin{array}{ccc}x& \stackrel{\&CenterDot;}{x}& \stackrel{\&CenterDot;\&CenterDot;}{x}\end{array}\right]}^T,$ Here, with represent position respectively, speed and component of acceleration.

7) relevant

Centered by the predicted position of flight path, the relevant ripple door of setpoint distance, orientation, pitching and Doppler, the metric data and flight path that fall into Bo Mennei are associated, this metric data may be the new observation of the corresponding target of this flight path, also may be false-alarm, be likely also the initial of a fresh target, carry out again after finally the determining of these problems obtains more observation data by the time.If metric data does not drop on the Bo Mennei of any one flight path, then these data directly produce a new flight path.If there is no observation station at the Bo Mennei of flight path, then carry out course extrapolation.

When moving-target flight path is all uncorrelated with all moving-target congealing points, then carry out two hypothesis tracking.Suppose that a target is undetected, then flight path is extrapolated according to the characteristics of motion before; Suppose that two targets are static, then that carries out between moving-target flight path to quiet target congealing point is relevant, if there is reference point, then carries out flight path renewal with it, if do not have reference point, then the position keeping flight path original is motionless, and speed is zero.

When quiet targetpath is all uncorrelated with all quiet target congealing points, then carry out two hypothesis tracking.Suppose that a target is undetected, then the position keeping flight path original is motionless, and speed is zero, and target is still static; Suppose that two targets transfer motion to, then that carries out between quiet targetpath to moving-target congealing point is relevant, if there is reference point, then carry out flight path renewal with it, and estimate the speed of flight path, if there is no reference point, the position then keeping flight path original is motionless, waits for the relevant of next frame.

The correlation combiner of all moving-target flight paths and quiet targetpath is participated in computing, finds the optimum allocation of an overall Least-cost according to formula 6, carry out the final updated of flight path with it.

8) filtering

According to alpha-beta-γ filtering method, filtering is carried out to flight path.

$\hat{X}(k/k)=\hat{X}(k/k-1)+K[y\left(k\right)-H\hat{X}(k/k-1)]---\left(18\right)$

$\hat{X}(k/k-1)=F\hat{X}(k-1/k-1)---\left(19\right)$

Wherein gain matrix $K=\left[\begin{array}{c}\mathrm{\&alpha;}\\ \mathrm{\&beta;}/T\\ \mathrm{\&gamma;}/T^2\end{array}\right],$ Observing matrix H=[1 0 0].

9) flight path management

When a flight path does not repeatedly have observation station repeatedly to extrapolate, flight path is deleted.

10) flight path is newly-built

The point mark do not shut mutually, sets up new alternative flight path.

11) 4 are repeated) ~ 10).

Claims (5)

1. an airborne radar ground sound target tenacious tracking method, is characterized in that step is as follows:

Step 1: radar emission NLFM signal, for detecting moving-target, is launched stepped frequency signal and detected quiet target, obtain the once some mark of ground moving target and quiet target at a scanning frame of radar simultaneously;

Step 2: the once some mark carrying out moving-target and quiet target respectively carries out secondary lobe and weighting agglomeration process;

Step 3: adopt the track initiation method based on two-stage Hough transform that congealing point that is dynamic and quiet target is set up respective alternative flight path respectively; Alternative for former frame flight path is carried out motion compensation, compensates to present frame carrier aircraft position;

Step 4: the position of prediction present frame flight path: $\hat{X}(k/k-1)=F\hat{X}(k-1/k-1),$ Wherein: $F=\left[\begin{array}{ccc}1& T& T^2/2\\ 0& 1& T\\ 0& 0& 1\end{array}\right],$ $X={\left[\begin{array}{ccc}x& \stackrel{\&CenterDot;}{x}& \stackrel{\&CenterDot;\&CenterDot;}{x}\end{array}\right]}^T,$ X, with represent the position of present frame flight path respectively, speed and component of acceleration;

Step 5: repeat step 1 ~ step 2, obtains the congealing point of next some mark;

Step 6: centered by the predicted position of flight path, is associated the metric data and flight path that fall into Bo Mennei, obtains correlation combiner;

If metric data does not drop on the Bo Mennei of any one flight path, set up a new flight path with these data;

If there is no observation station at the Bo Mennei of flight path, in two kinds of situation:

Situation 1: when moving-target flight path is all uncorrelated with all moving-target congealing points, then carries out two hypothesis tracking, obtain correlation combiner;

Suppose 1: target is undetected, then flight path is extrapolated according to the characteristics of motion before;

Suppose 2: target is static, then that carries out between moving-target flight path to quiet target congealing point is relevant, if there is reference point, then carries out flight path renewal with it, if do not have reference point, then the position keeping flight path original is motionless, and speed is zero;

Situation 2: when quiet targetpath is all uncorrelated with all quiet target congealing points, then carry out two hypothesis tracking;

Suppose 1: target is undetected, then the position keeping flight path original is motionless, and speed is zero, and target is still static;

Suppose 2: target transfers motion to, that then carries out between quiet targetpath to moving-target congealing point is relevant, if there is reference point, then carry out flight path renewal with it, and estimate the speed of flight path, if do not have reference point, then the position keeping flight path original is motionless, wait for the relevant of next frame, obtain correlation combiner;

The correlation combiner of all moving-target flight paths and quiet targetpath is participated in computing according to the following formula, finds the optimum allocation of an overall Least-cost, carry out the final updated of flight path with it;

a ^*(k)＝arg min _a(k)C(k|a(k))

Wherein:

$C\left(k\right|a\left(k\right))={\mathrm{\&Sigma;}}_{m=0_1,0_2,1}^{M\left(k\right)}{\mathrm{\&Sigma;}}_{n=1}^{N(k-1)}a(k,m,n)c(k,m,n)$

Wherein:

C (k, m, n) is the cost value of distributing a (k, m, n): $c(k,m,n)=-\ln \left(\frac{\mathrm{\&Phi;}(k,m,n)}{\mathrm{\&Phi;}(k,m,0)}\right)$

When situation 1: Ф (k, m, n) is calculated as follows:

$\mathrm{\&Phi;}(k,m,n)=\left\{\begin{array}{cc}{P}_{D}P\left\{\right|v_r\left(k\right)|\&GreaterEqual;v_0\}\mathrm{\&Lambda;}(k,m,n)& m>0,n>0\\ \mathrm{\&lambda;}& m>0,n=0\\ (1-P_D)P\left\{\right|v_r\left(k\right)|\&GreaterEqual;v_0\}\mathrm{\&lambda;}& m=0_1,n>0\\ P\left\{\right|v_r\left(k\right)|<v_0\}\mathrm{\&lambda;}& m=0_2,n>0\end{array}\right.$

Wherein: m>0, n>0 are relevant to flight path for measuring; M>0, n=0 are uncorrelated with already present flight path for measuring, and are false-alarm; M=0 ₁, n>0 is for supposing that measurement 1 is relevant to flight path; M=0 ₂, n>0 is for supposing that measurement 2 is relevant to flight path;

Λ (k, m, n) is the combined probability that Interactive Multiple-Model is estimated, v ₀minimum detectable, v _rk () is that the radial velocity of a kth scanning frame is not (if a kth scanning frame has measured value, v _rk () is predicted by kth-1 scanning frame to get).Suppose speed Gaussian distributed, then can calculate P{|v _r(k) |>=v ₀and P{|v _r(k) | <v ₀;

The space density λ of false-alarm is: $\mathrm{\&lambda;}=\frac{P_{\mathrm{FA}}}{{12}^{3/2}{\mathrm{\&sigma;}}_r{\mathrm{r\&sigma;}}_{\mathrm{\&theta;}}{\mathrm{\&sigma;}}_{\stackrel{\&CenterDot;}{r}}}$

Wherein P _fAbe false-alarm probability, the measuring accuracy of distance, orientation, distance speed three-dimensional is respectively with

When situation 2, Ф (k, m, n) is calculated as follows:

$\mathrm{\&Phi;}(k,m,n)=\left\{\begin{array}{cc}{P}_{D}P\left\{\right|v_r\left(k\right)|=0\}\mathrm{\&Lambda;}(k,m,n)& m>0,n>0\\ \mathrm{\&lambda;}& m>0,n=0\\ (1-P_D)P\left\{\right|v_r\left(k\right)|=0\}\mathrm{\&lambda;}& m=0_1,n>0\\ P\left\{\right|v_r\left(k\right)|!=0\}\mathrm{\&lambda;}& m=0_2,n>0\end{array}\right.$

Situation 2 needs to calculate P{|v _r(k) |=0} and P{|v _r(k) | unequal to 0};

Described ripple door comprises range gate, orientation Bo Men, pitching ripple door and Doppler's ripple door;

Step 7, filtering: according to the filtering method of alpha-beta-γ, carry out filtering to flight path

$\hat{X}(k/k)=\hat{X}(k/k-1)+K[y\left(k\right)-H\hat{X}(k/k-1)]$

$\hat{X}(k/k-1)=F\hat{X}(k-1/k-1),$

Wherein gain matrix $K=\left[\begin{array}{c}\mathrm{\&alpha;}\\ \mathrm{\&beta;}/T\\ \mathrm{\&gamma;}/T^2\end{array}\right],$ Observing matrix H=[1 0 0];

Step 8, flight path manage: when a flight path does not repeatedly have observation station to cause repeatedly to extrapolate, deleted by flight path; Repeat step 1 ~ step 8, carry out sound target tenacious tracking.

2. airborne radar ground sound target tenacious tracking method according to claim 1, is characterized in that: described range gate 20 ~ 50m.

3. airborne radar ground sound target tenacious tracking method according to claim 1, is characterized in that: 1.5 ~ 2.5 degree, described orientation ripple door.

4. airborne radar ground sound target tenacious tracking method according to claim 1, is characterized in that: described 2 ~ 4 degree, pitching ripple door.

5. airborne radar ground sound target tenacious tracking method according to claim 1, is characterized in that: described 3 ~ 6, Doppler's ripple door.