CN105204018A - Two-dimensional DOA tracking method by means of multi-frame information - Google Patents

Abstract

The invention discloses a two-dimensional DOA tracking method by means of multi-frame information, and belongs to the field of radar signal processing. A movement model of a target is considered for a weak target under the environment of low signal-to-noise ratio and few snapshot data, the estimation result of the target state is not declared in single frame, and combined processing is performed on multi-frame data by adopting a dynamic programming algorithm according to the state transition relationship of the target so that multi-frame space-time information of the target can be fully utilized and real-time tracking of the two-dimensional DOA of the target can be realized. The beneficial effects of the two-dimensional DOA tracking method by means of the multi-frame information are that real-time tracking of the two-dimensional DOA of the weak target can be performed under the environment of low signal-to-noise ratio and few snapshot data. Compared with an existing two-dimensional DOA method, higher tracking convergence probability and better DOA estimation precision can be obtained by the method.

Description

A kind of two-dimentional DOA tracking utilizing multiframe information

Technical field

The invention belongs to radar signal processing field.

Background technology

Direction of arrival (DOA) estimation technique is an important branch of Array Signal Processing, plays a part more and more important in fields such as radar, communication, sonar, astronomy.Existing DOA algorithm for estimating, major part depends on location technology, such as based on the undistorted response algorithm of minimum variance (MVDR), the multiple signal classification algorithm (Multiplesignalclassification of Wave beam forming, MUSIC), based on Signal parameter estimation (ESPRIT) algorithm of ESPRIT, the equal hypothetical target of these algorithms is static and only utilizes target to work as the spatial information of pre-test frame.In reality, when signal is kept in motion, the normally time dependent function of its DOA, the DOA of adjacent measurement frame is height correlation.In this case, traditional DOA algorithm for estimating is owing to only utilizing the spatial information when pre-test frame, cause the temporal information come from the estimation of former frame position to be dropped, in the situations such as low signal-to-noise ratio, few fast beat of data, performance all can be made to decline rapidly, even complete failure.Therefore, research is needed DOA to be carried out to the algorithm of real-time follow-up.

In radar tracking algorithm, before dynamic programming detects, track algorithm does not do threshold processing due to its single frames, but frame data digitizing is stored, to multiframe data aggregate process, thus the time of day of estimating target.This processing mode simultaneously utilizing multiframe information to carry out target locating, can realize the better tracking performance of weak target.X.Zhong proposes a kind of two-dimentional DOA tracking based on particle filter in document " Particlefilteringfor2-Ddirectionofarrivaltrackingusingan acousticvectorsensor ", the DOA estimation method that the method is more traditional can improve follows the tracks of convergent probability and tracking accuracy, but under low signal-to-noise ratio and few snap data cases, performance can decline to a great extent.

Summary of the invention

Technical matters to be solved by this invention is, provides a kind of two-dimentional DOA tracking utilizing multiframe information be applicable under low signal-to-noise ratio, few snap data cases, reaches the object that tracking progress is high, convergence is good.

The present invention for solving the problems of the technologies described above adopted technical scheme is, a kind of two-dimentional DOA tracking utilizing multiframe information, comprises the following steps:

Step 1, initializes system parameters: uniform rectangular array elements number is M, the position of each array element, export that data fast umber of beats is N, dynamic programming detect before track algorithm process frame number K, observation interval T, state transfer number q, error threshold ε;

The datum plane that step 2, acquisition K frame two dimension MUSIC spectral function are formed:

Step 2.1, kth frame echoed signal incide the position angle of uniform rectangular array, the angle of pitch is respectively θ _k, i-th signal observing array element receive is i=1 ..., M, wherein, λ is the wavelength of incoming signal, (x _i, y _i) be the coordinate of i-th array element, s (n) is far field narrow band signal, v _in () is additive noise, uniform rectangular array received to signal vector be x (n)=[x ₁(n), x ₂(n) ..., x _m(n)] ^t, T represents matrix transpose operation;

Step 2.2, N the fast beat of data utilizing array received to arrive, the spatial correlation matrix R of estimated signal:

$R=\frac{1}{N}\underset{n=1}{\overset{N}{\Σ}}x\left(n\right)x^H\left(n\right);$

Wherein, H represents its conjugate transpose;

Step 2.3, Eigenvalues Decomposition is carried out to R, and by the eigenwert that obtains by the arrangement of dull non-increasing order, i.e. λ ₁>=λ ₂>=...>=λ _l> λ _l+1=λ _l+2=...=λ _m=σ ², wherein, L represents signal source number, and normalization characteristic vector corresponding to these eigenwerts is u respectively ₁..., u _l, u _l+1..., u _m, wherein, u ₁..., u _land u _l+1..., u _mopen into signal subspace E respectively _swith noise subspace E _n; Definition matrix G=[u _l+1u _l+2u _m];

Step 2.4, two-dimentional MUSIC algorithm is utilized to build spatial spectrum function, for with θ and the steering vector of the array of change;

Step 3, value function accumulate

When k=1 frame, to each state x _kcorresponding value function initialize, the value function initial value of each state is the first frame radar measurement corresponding to this state, i.e. the two-dimentional MUSIC algorithm spatial spectrum function plane of the first frame;

When 2≤k≤K frame, upgrade the value function that each state is corresponding; arbitrary quantification state of kth frame, θ _krepresent that the position angle in this state quantizes state, represent that the angle of pitch in this state quantizes state, the renewal relational expression of value function is i (x _k) represent state x _kcorresponding value function, τ (x _k) represent that kth-1 frame likely transfers to x _kstate set, simultaneously recording status x _kwith it at k-1 frame state set τ (x _k) transfer relationship between state corresponding to median maximal value;

If step 4 k<K, makes k=k+1, return step 3; If k=K, perform step 5;

Step 5, make azimuth angle theta at (0 °, 180 °) and the angle of pitch in (0 °, 90 °) scope, change, carries out spectrum peak search to K frame data plane, and the angle coordinate found out corresponding to K frame value function maxima point is the estimated value of signal source two dimension DOA;

Step 6, flight path recover:

To the target that step 5 obtains, utilize the state x that step 3 records _kwith it at k-1 frame state set τ (x _k) transfer relationship between state corresponding to median maximal value, recover the flight path of target.

The present invention is directed to low signal-to-noise ratio, weak target under few snap data environment, by considering the motion model of target, the estimated result of dbjective state is not announced in single frames, but employing dynamic programming algorithm, according to the state transfer relationship of target, Combined Treatment is carried out to multiframe data, make full use of the space time information of target multiframe, the real-time follow-up to target two dimension DOA can be realized.

The invention has the beneficial effects as follows, under the environment of low signal-to-noise ratio, few fast beat of data, real-time follow-up can be carried out to the two-dimentional DOA of weak target.The more existing two-dimentional DOA method of the method, can obtain higher tracking convergent probability and better DOA estimated accuracy.

Accompanying drawing explanation

Fig. 1 is FB(flow block) of the present invention.

Fig. 2 applies the inventive method and target following convergent probability comparison diagram under traditional different signal to noise ratio (S/N ratio) of two-dimentional MUSIC algorithm in the embodiment of the present invention.

Fig. 3 is the root-mean-square error comparison diagram applying the inventive method and target location under traditional different signal to noise ratio (S/N ratio) of two-dimentional MUSIC algorithm in the embodiment of the present invention.

Fig. 4 applies target following convergent probability comparison diagram under the inventive method fast umber of beats different from traditional two-dimentional MUSIC algorithm in the embodiment of the present invention.

Fig. 5 is the root-mean-square error comparison diagram applying target location under the inventive method fast umber of beats different from traditional two-dimentional MUSIC algorithm in the embodiment of the present invention.

Embodiment

The present invention mainly adopts the method for Computer Simulation to verify, institute in steps, conclusion all on MATLAB-R2010b checking correct.Concrete implementation step is as follows:

Step 1, initializes system parameters: uniform rectangular array elements number is M=12, the position of each array element, export that data fast umber of beats is N, dynamic programming detect before track algorithm process frame number K, observation interval T=1, state transfer number q=4, error threshold ε=3 °;

The datum plane that step 2, acquisition K frame two dimension MUSIC spectral function are formed:

Step 2.1, kth frame echoed signal incide the position angle of uniform rectangular array, the angle of pitch is respectively θ _k, i-th signal observing array element receive is i=1 ..., M, wherein, λ is the wavelength of incoming signal, (x _i, y _i) be the coordinate of i-th array element, s (n) is far field narrow band signal, v _in () is additive noise, uniform rectangular array received to signal vector be x (n)=[x ₁(n), x ₂(n) ..., x _m(n)] ^t, T represents matrix transpose operation;

Step 2.2, N the fast beat of data utilizing array received to arrive, the spatial correlation matrix R of estimated signal:

$R=\frac{1}{N}\underset{n=1}{\overset{N}{\&Sigma;}}x\left(n\right)x^H\left(n\right);$

Wherein, H represents its conjugate transpose;

Step 2.3, Eigenvalues Decomposition is carried out to R, and by eigenwert by the arrangement of dull non-increasing order, i.e. λ ₁>=λ ₂>=...>=λ _l> λ _l+1=λ _l+2=...=λ _m=σ ², wherein, L represents signal source number, and normalization characteristic vector corresponding to these eigenwerts is u respectively ₁..., u _l, u _l+1..., u _m, wherein, u ₁..., u _land u _l+1..., u _mopen into signal subspace E respectively _swith noise subspace E _n.Definition matrix G=[u _l+1u _l+2u _m];

Step 2.4, two-dimentional MUSIC algorithm is utilized to build spatial spectrum function, for with θ and the steering vector of the array of change;

Step 3, dynamic programming value function accumulate

When k=1 frame, to each state x _kcorresponding value function initialize, the value function initial value of each state is the first frame radar measurement corresponding to this state, i.e. the two-dimentional MUSIC algorithm spatial spectrum function plane of the first frame;

When 2≤k≤K frame, upgrade the value function that each state is corresponding; arbitrary quantification state of kth frame, θ _krepresent that the position angle in this state quantizes state, represent that the angle of pitch in this state quantizes state, the renewal relational expression of value function is i (x _k) represent state x _kcorresponding value function, τ (x _k) represent that kth-1 frame likely transfers to x _kstate set, simultaneously recording status x _kwith it at k-1 frame state set τ (x _k) transfer relationship between state corresponding to median maximal value;

If step 4 k<K, makes k=k+1, return step 3; If k=K, perform step 5;

Step 5, make azimuth angle theta at (0 °, 180 °) and the angle of pitch in (0 °, 90 °) scope, change, carries out spectrum peak search to K frame data plane, and the angle coordinate found out corresponding to K frame value function maxima point is the estimated value of signal source two dimension DOA;

Step 6, flight path recover:

To the target that step 5 obtains, utilize the state x that step 3 records _kwith it at k-1 frame state set τ (x _k) transfer relationship between state corresponding to median maximal value, recover the flight path of target.

Follow the tracks of convergent probability (PROC) to be defined as estimated value and to be converged in probability within the scope of target actual value certain error, converging factor is defined as

Follow the tracks of convergent probability $PROC=\underset{k=1}{\overset{K}{\&Sigma;}}{\mathrm{\&chi;}}_k/K\&times;100\%.$

Figure 2 shows in the present embodiment the tracking convergent probability comparison diagram utilizing the different frame numbers of the inventive method process and existing two-dimentional MUSIC algorithm under different signal to noise ratio (S/N ratio) condition, fig. 3 gives in the present embodiment the comparison diagram of the root-mean-square error utilizing the different frame numbers of the inventive method process and existing two-dimentional MUSIC algorithm under different signal to noise ratio (S/N ratio) condition, it is the statistics of 500 Monte Carlo experiments, result shows that the method is in low signal-to-noise ratio situation, has higher tracking convergent probability and better estimated accuracy.

Fig. 4 shows in the present embodiment the tracking convergent probability comparison diagram utilizing the different frame numbers of the inventive method process and existing two-dimentional MUSIC algorithm under different snap said conditions, figure 5 provides in the present embodiment the comparison diagram of the root-mean-square error utilizing the different frame numbers of the inventive method process and existing two-dimentional MUSIC algorithm under different snap said conditions, it is the statistics of 500 Monte Carlo experiments, result shows that the method is in fast umber of beats situation less, tracking convergent probability can be improved, obtain better estimated accuracy.

Claims (1)

1. utilize a two-dimentional DOA tracking for multiframe information, comprise the following steps:

Step 1, initializes system parameters: uniform rectangular array elements number is M, the position of each array element, export that data fast umber of beats is N, dynamic programming detect before track algorithm process frame number K, observation interval T, state transfer number q, error threshold ε;

The datum plane that step 2, acquisition K frame two dimension MUSIC spectral function are formed:

Step 2.1, kth frame echoed signal incide the position angle of uniform rectangular array, the angle of pitch is respectively θ _k, i-th signal observing array element receive is i=1 ..., M, wherein, λ is the wavelength of incoming signal, (x _i, y _i) be the coordinate of i-th array element, s (n) is far field narrow band signal, v _in () is additive noise, uniform rectangular array received to signal vector be x (n)=[x ₁(n), x ₂(n) ..., x _m(n)] ^t, T represents matrix transpose operation;

Step 2.2, N the fast beat of data utilizing array received to arrive, the spatial correlation matrix R of estimated signal:

$R=\frac{1}{N}\underset{n=1}{\overset{N}{\&Sigma;}}x\left(n\right)x^H\left(n\right);$

Wherein, H represents its conjugate transpose;

Step 2.3, Eigenvalues Decomposition is carried out to R, and by the eigenwert that obtains by the arrangement of dull non-increasing order, i.e. λ ₁>=λ ₂>=...>=λ _l> λ _l+1=λ _l+2=...=λ _m=σ ², wherein, L represents signal source number, and normalization characteristic vector corresponding to these eigenwerts is u respectively ₁..., u _l, u _l+1..., u _m, wherein, u ₁..., u _land u _l+1..., u _mopen into signal subspace E respectively _swith noise subspace E _n; Definition matrix G=[u _l+1u _l+2u _m];

Step 2.4, two-dimentional MUSIC algorithm is utilized to build spatial spectrum function, for with θ and the steering vector of the array of change;

Step 3, value function accumulate

When k=1 frame, to each state x _kcorresponding value function initialize, the value function initial value of each state is the first frame radar measurement corresponding to this state, i.e. the two-dimentional MUSIC algorithm spatial spectrum function plane of the first frame;

When 2≤k≤K frame, upgrade the value function that each state is corresponding; arbitrary quantification state of kth frame, θ _krepresent that the position angle in this state quantizes state, represent that the angle of pitch in this state quantizes state, the renewal relational expression of value function is i (x _k) represent state x _kcorresponding value function, τ (x _k) represent that kth-1 frame likely transfers to x _kstate set, simultaneously recording status x _kwith it at k-1 frame state set τ (x _k) transfer relationship between state corresponding to median maximal value;

If step 4 k<K, makes k=k+1, return step 3; If k=K, perform step 5;

Step 5, make azimuth angle theta at (0 °, 180 °) and the angle of pitch in (0 °, 90 °) scope, change, carries out spectrum peak search to K frame data plane, and the angle coordinate found out corresponding to K frame value function maxima point is the estimated value of signal source two dimension DOA;

Step 6, flight path recover:

To the target that step 5 obtains, utilize the state x that step 3 records _kwith it at k-1 frame state set τ (x _k) transfer relationship between state corresponding to median maximal value, recover the flight path of target.