CN103675768B - A kind of suppression method for satellite-borne SAR radio-frequency interference - Google Patents

Abstract

The invention provides a kind of suppression method for satellite-borne SAR radio-frequency interference.Technical scheme comprises: the first step: build data matrix, and the range line echo according to gained builds data matrix; Second step: estimate covariance correlation matrix, the data matrix constructed by utilization estimates its covariance correlation matrix; 3rd step: feature decomposition, obtain eigenwert and characteristic of correspondence vector and sort; 4th step: wavelet decomposition estimates dominant eigenvalue number, the proper vector structure interference space corresponding to the dominant eigenvalue estimated by utilization; 5th step: AF panel and data reconstruction, completes wide band radio-frequency AF panel.The present invention all can obtain good inhibition to narrow radio frequency AF panel and wide band radio-frequency AF panel, and Robust Performance.

Description

A kind of suppression method for satellite-borne SAR radio-frequency interference

Technical field

The invention belongs to the interleaving techniques field of space flight and microwave remote sensing, particularly Radio frequency interference (RFI) suppressing method in a kind of satellite-borne SAR (Synthetic Aperture Radar, synthetic-aperture radar) data.

Background technology

Spaceborne SAR System can complete wide swath, high resolving power earth observation, is the effective means of carrying out military mapping and battlefield surveillance.In order to meet growing military and civilian demand, the particularly active demand of more high resolving power and concealed target detection, the working frequency range of satellite-borne SAR is expanded gradually toward higher and lower both direction, the electromagnetic environment faced will become increasingly complex, particularly low band Spaceborne SAR System, a large amount of radio and television and communication system is there is within the scope of frequency spectrums of operation, they are in continuous duty, and emissive power is higher, even the bandwidth of operation of some system is wider, such as DVB-T (Digital VideoBroadcasting--Terrestrial, DTTB) system, bandwidth has 8MHz, these undesired signals enter into Spaceborne SAR System receiver, Space-borne SAR Imaging quality will be had a strong impact on, cause image interpretation and interpretation difficulty.

The SAR Radio frequency interference (RFI) suppressing method of the feature based Subspace Decomposition that Feng Zhou proposes is relatively good to narrow radio frequency interference suppressioning effect.But wide band radio-frequency is disturbed, due to be with wide, power is high, the large eigenwert that number is more can be produced, the error that the dominant eigenvalue number decision method that existing eigendecomposition method adopts can cause dominant eigenvalue number to judge when carrying out Eigenvalue Decomposition.Based on the advantage of wavelet decomposition in detection of Singular Point, the present invention proposes the suppression method for satellite-borne SAR radio-frequency interference of a kind of wavelet decomposition and eigendecomposition associating.This method all has good inhibition to narrow radio frequency interference and wide band radio-frequency interference, and Robust Performance.

Summary of the invention

The object of the invention is: the satellite-borne SAR wide band radio-frequency disturbance restraining method proposing a kind of wavelet decomposition and eigendecomposition associating, the method takes full advantage of the advantage of wavelet decomposition in detection of Singular Point, good inhibition is all had to narrow radio frequency interference and wide band radio-frequency interference, and Robust Performance.

The thinking of technical solution of the present invention is: the covariance correlation matrix first constructing raw data, feature decomposition is carried out to it, the method of wavelet decomposition is utilized to carry out dominant eigenvalue judgement, what dominant eigenvalue was corresponding is interference components, utilize dominant eigenvalue characteristic of correspondence vector structure interference space, then the mapping component of original signal on feature interference subspace is deducted, undesired signal can be curbed.

Technical solution of the present invention is:

Suppose that x represents the data that a certain orientation moment satellite-borne SAR receives, x=[x ₁..., x _m..., x _m] ^t, wherein M represents that distance is to sampling number, and subscript T represents matrix transpose operation.

The first step, structure data matrix

Data x is divided into the subvector of K L dimension, L determines according to actual needs, usual 32≤L≤128, wherein K=M-L+1.A definition kth sub-vector x _kfor:

x _k＝[x _k,…x _k+L-1] ^T,k＝1,2,…,K

Then, the data matrix X of L × K is built:

X＝[x ₁,…,x _k]

Second step, estimate covariance correlation matrix

Data matrix X is utilized to estimate its covariance correlation matrix

$\hat{R}=\frac{1}{M}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{x}_k{x}_k^T$

3rd step, feature decomposition

Right carry out Eigenvalues Decomposition, eigenwert is arranged according to descending order, obtain characteristic value sequence λ ₁, λ ₂..., λ _l, its corresponding proper vector is u ₁, u ₂..., u _l.

4th step, estimation dominant eigenvalue number, structure interference space

To characteristic value sequence λ ₁, λ ₂..., λ _lcarry out multiple dimensioned one-dimensional discrete wavelet decomposition, the wavelet coefficient c of i-th layer that obtains _ifor

c _i＝[cA _i，cD _i，cD _i-1，…，cD ₁],i＝1,2,…,N，

Wherein cA _irepresent the low frequency coefficient that i-th layer of decomposition obtains; CD _irepresent the high frequency coefficient that i-th layer of decomposition obtains.N gets at most , represent and round downwards.

Utilize high frequency coefficient to be reconstructed signal and obtain detail signal d, obtain the sequence number that in detail signal d, maximum amplitude is corresponding, this sequence number is exactly the number r of dominant eigenvalue, then construct interference space F to be:

F＝span{u ₁,u ₂,…,u _r},1≤r＜L

5th step, AF panel and data reconstruction

Following formula is utilized to obtain interfering data I _k:

$I_k=\underset{i=1}{\overset{r}{\mathrm{\Σ}}}{u}_i^T{x}_k{u}_i$

Following formula is utilized to calculate the signal after suppressing Radio frequency interference (RFI)

${\hat{x}}_k=x_k-I_k$

Thus, the data matrix after the Radio frequency interference (RFI) that is inhibited

$\hat{X}={[{\hat{x}}_1,...,{\hat{x}}_k]}^T$

Rearrange data matrix namely the inverse process that the first step builds data matrix is implemented, the signal vector x ' after the Radio frequency interference (RFI) that is inhibited.

Employing the present invention is desirable obtains following technique effect:

The suppression method for satellite-borne SAR radio-frequency interference that the small echo that the present invention proposes is combined with eigendecomposition, can suppress, in interfering process, to utilize wavelet decomposition accurately to estimate dominant eigenvalue number in employing eigendecomposition method.Overcome existing Subspace Decomposition method and inaccurate defect is judged to broadband interference dominant eigenvalue number, all can obtain good inhibition to narrow radio frequency AF panel and wide band radio-frequency AF panel, and Robust Performance.

Accompanying drawing explanation

Fig. 1 is the principle flow chart of suppression method for satellite-borne SAR radio-frequency interference provided by the invention;

When Fig. 2 is for utilizing emulated data, the dominant eigenvalue number that the inventive method and existing method generate is estimated to contrast schematic diagram;

Fig. 3 is the imaging results of the raw data used in emulation experiment;

Fig. 4 is the imaging results of Fig. 3 being added to the rear data of wide band radio-frequency interference;

Fig. 5 utilizes existing method to carry out the imaging results after AF panel to Fig. 4;

Fig. 6 utilizes method provided by the invention to carry out the imaging results after AF panel to Fig. 4.

Embodiment

Fig. 1 is the principle flow chart of suppression method for satellite-borne SAR radio-frequency interference provided by the invention.Whole flow process is divided into five steps: with SAR data range line echoed signal for input, the first step: build data matrix, and the range line echo according to gained builds data matrix; Second step: estimate covariance correlation matrix, the data matrix constructed by utilization estimates its covariance correlation matrix; To second step, 3rd step: feature decomposition, estimates that the covariance correlation matrix obtained carries out feature decomposition, obtain eigenwert and characteristic of correspondence vector, carry out descending order arrangement to eigenwert, corresponding proper vector order also corresponds to the order of eigenwert.4th step: wavelet decomposition estimates dominant eigenvalue number, build interference space, multiple dimensioned one-dimensional discrete wavelet decomposition is carried out to the characteristic value sequence after sequence, obtain corresponding low frequency coefficient and high frequency coefficient, utilize high frequency coefficient to be reconstructed characteristic value sequence, obtain the detailed information of characteristic value sequence, obtain the sequence number that maximum amplitude in detail signal is corresponding, just can determine the number of dominant eigenvalue, the proper vector structure interference space corresponding to the dominant eigenvalue estimated by utilization; 5th step: AF panel and data reconstruction, data matrix is projected to the interfering data estimated by interference space acquisition, the interfering data deducting estimation from data matrix can obtain the data matrix after AF panel, data matrix after AF panel is rearranged and can obtain the vector of the echo after AF panel, complete the wide band radio-frequency AF panel of this range line.

When Fig. 2 is for utilizing emulated data, the dominant eigenvalue number that the inventive method and existing method generate is estimated to contrast schematic diagram.Emulated data utilizes one-dimensional linear FM signal to increase the DVB-T interfere generation of 8MHz, is with rhombus solid line in figure represent the characteristic value sequence (according to arranging from big to small) of covariance correlation matrix gained after feature decomposition; Band square frame dotted line represent and adopt the dominant eigenvalue estimated sequence that obtains of existing ratio calculation method, namely what wherein maximal value was corresponding 5 is judged to dominant eigenvalue number; The line of band Diamond spot represent and adopt the dominant eigenvalue estimated sequence that obtains of method provided by the invention, namely what wherein maximal value was corresponding 7 is judged to dominant eigenvalue number; Dotted line " ... " represents that dominant eigenvalue number that existing ratio calculation method estimates corresponds to the position of primitive character value sequence; Dot-and-dash line represent that dominant eigenvalue number that method provided by the invention obtains corresponds to the position of primitive character value sequence.As seen from the figure, compared with the conventional method, the inventive method can accurately estimate dominant eigenvalue number under wide band radio-frequency interference exists situation, and estimated result is better than existing method.

Fig. 3, Fig. 4, Fig. 5 and Fig. 6 are the results of carrying out emulation experiment.Emulation experiment have chosen a glitch-free RadarSat-1 echo measured data, and in these data, with the addition of the DVB-T undesired signal that bandwidth is 8MHz, jamming-to-signal ratio is 20dB.

Fig. 3 is the imaging results of the raw data used in emulation experiment.As can be seen from the figure, clear picture, contrast is high.Fig. 4 is the imaging results of Fig. 3 being added to the rear data of wide band radio-frequency interference.As can be seen from the figure, due to the existence of interference, image object information is capped, and picture contrast declines.Fig. 5 utilizes existing method to carry out the imaging results after AF panel to Fig. 4, and what wherein dominant eigenvalue number estimation method adopted is existing radiometer algorithm, and after AF panel, picture quality recovers less, and contrast is not high.Fig. 6 utilizes method provided by the invention to carry out the imaging results after AF panel to Fig. 4, and after AF panel, picture quality obviously promotes, and detailed information recovers comparatively obvious, and picture contrast is significantly improved.

Above emulation experiment describes the inventive method still estimating dominant eigenvalue number exactly by under wide band radio-frequency disturbing effect, ensure that the accuracy that interference space constructs, and wide band radio-frequency is disturbed can be effectively suppressed.Meanwhile, for narrow radio frequency interference, the present invention also can be suppressed result preferably.

Claims (2)

1. a suppression method for satellite-borne SAR radio-frequency interference, is characterized in that, comprises the steps,

Suppose that x represents the data that a certain orientation moment satellite-borne synthetic aperture radar receives, x=[x ₁..., x _m..., x _m] ^t, wherein M represents that distance is to sampling number, and subscript T represents matrix transpose operation;

The first step, structure data matrix:

Data x is divided into the subvector of K L dimension, wherein K=M-L+1; A definition kth sub-vector x ' _kfor:

x′ _k＝[x _k,…x _k+L-1] ^T,k＝1,2,…,K

Then, the data matrix X of L × K is built:

X＝[x′ ₁,…,x′ _K]

Second step, estimate covariance correlation matrix:

Data matrix X is utilized to estimate its covariance correlation matrix

$\hat{R}\frac{1}{M}\underset{k=1}{\overset{K}{\mathrm{\Σ}}}{x}_k^{\′}{x}_k^{\′T}$

3rd step, feature decomposition:

To covariance correlation matrix carry out Eigenvalues Decomposition, eigenwert is arranged according to descending order, obtain characteristic value sequence λ ₁, λ ₂..., λ _l, its corresponding proper vector is u ₁, u ₂..., u _l;

4th step, estimation dominant eigenvalue number, structure interference space:

To characteristic value sequence λ ₁, λ ₂..., λ _lcarry out multiple dimensioned one-dimensional discrete wavelet decomposition, the wavelet coefficient c of i-th layer that obtains _ifor

c _i＝[cA _i，cD _i，cD _i-1，…，cD ₁],i＝1,2,…,N，

Wherein cA _irepresent the low frequency coefficient that i-th layer of decomposition obtains; CD _irepresent the high frequency coefficient that i-th layer of decomposition obtains; N maximal value is represent and round downwards;

Utilize high frequency coefficient to be reconstructed signal and obtain detail signal d, obtain the sequence number that in detail signal d, maximum amplitude is corresponding, this sequence number is exactly the number r of dominant eigenvalue, then construct interference space F to be:

F＝span{u ₁,u ₂,…,u _r},1≤r＜L

5th step, AF panel and data reconstruction:

Following formula is utilized to obtain interfering data I _k:

$I_k=\underset{i=1}{\overset{r}{\mathrm{\Σ}}}{u}_i^T{x}_k^{\′}{u}_i$

Following formula is utilized to calculate the signal after suppressing Radio frequency interference (RFI)

${\hat{x}}_k=x_k^{\′}-I_k$

Thus, the data matrix after the Radio frequency interference (RFI) that is inhibited

$\hat{X}={[{\hat{x}}_1,\·\·\·,{\hat{x}}_k]}^T$

Rearrange data matrix namely the inverse process that the first step builds data matrix is implemented, the signal vector x ' after the Radio frequency interference (RFI) that is inhibited.

2. suppression method for satellite-borne SAR radio-frequency interference according to claim 1, is characterized in that, 32≤L≤128.