CN103941256A - Passive distribution SAR (synthetic aperture radar) imaging process method based on double-stage multi-resolution reconstruction - Google Patents

Abstract

The invention relates to a passive distribution SAR imaging process method based on double-stage multi-resolution reconstruction and belongs to the technical field of SAR imaging signal processing methods. The method comprises the following steps of, firstly, performing filtering and modulation on target echo signals received by an airborne receiver, separating out the target echo signals from different radiation sources and modulating the target echo signals to a base band; secondly, processing various channel signals separated out in the first step through the Fourier reconstruction algorithm to obtain the coarse-resolution images of a target; thirdly, for the coarse-resolution images obtained in the second step, obtaining the high-definition image of the target in every coarse-definition pixel unit among the images through a sparse reconstruction method; lastly, obtaining the high-definition image of a whole scene through data screening and image splicing. The passive distribution SAR imaging process method based on the double-stage multi-resolution reconstruction can solves the grating lobe problem due to sparse sampling and meanwhile has higher computation efficiency compared with traditional sparse reconstruction methods.

Description

Based on two-stage, differentiate the passive distributed SAR image processing method of reconstruct more

Technical field

The present invention relates to a kind of passive distributed SAR image processing method of differentiating reconstruct based on two-stage more, belong to synthetic-aperture radar (synthetic aperture radar is called for short SAR) imaging signal processing method technical field.

Background technology

Passive radar is a kind ofly itself not provide radiation source, but utilizes external opportunities radiation source typical case as the radar as irradiation source such as radio station, TV station and radio communication base station.Owing to providing irradiation source without transmitter, passive radar has good concealment, antijamming capability is strong and low cost and other advantages, is a study hotspot of current field of radar.For conventional external opportunities irradiation source, the signal that they provide has the features such as carrier frequency is low, bandwidth is little conventionally, this makes passive radar often have poor distance and bearing resolution, thereby study hotspot also mainly concentrates on the location of target and follows the trail of.In recent years, be subject on the one hand high-resolution imaging demand driving, on the other hand the available a plurality of illuminators of opportunities of passive radar for high-resolution imaging provide may, so passive radar imaging has received increasing concern.Passive synthetic-aperture radar is a kind of important passive imaging radar system, it utilizes a plurality of chance radiation sources as irradiation source, radar receiver is loaded on the platform of motion target echo is received, and the space diversity of utilizing the space providing of a plurality of radiation sources and frequency diversity and the motion of receiver platform to bring can provide the high-resolution imaging to target.At present, formation method for passive synthetic-aperture radar can be divided into three major types, first class is based on Fourier's reconfiguration technique, he has utilized the Fourier transform relation existing between radar image data and imaging area objective function, directly reception data are done to Fourier transform and obtain target image, as document 1: " Multistatic synthetic aperture imaging of aircraft using reflected television signals " (the many bases synthetic-aperture radar based on TV signal is to aerial target imaging) and document 2: disclosed technology in " the distributed algorithms for passive radar imaging based on external radiation source ", second largest Lei Shi back projection class algorithm, as document 3: " Multistatic synthetic aperture radar image formation " (many bases synthetic aperture radar image-forming) and document 4: disclosed technology in " a kind of passive synthetic aperture radar image-forming method based on echo correlation ", the large class of the 3rd class is sparse reconstruct class, as document 5: " the sparse passive radar imaging based on compressed sensing ", document 6: " the many bases algorithms for passive radar imaging based on compressed sensing " and document 7: disclosed technology in " Compressed sensing of mono-static and multi-static SAR " (the compressed sensing imaging of single base and Duo Ji synthetic-aperture radar).Fourier's reconstruct Lei He back projection class class all requires signal sampling to meet Nyquist theorem, and passive distributed radar data acquisition normally segmentation is sparse, in each subband, sampling is to meet Nyquist theorem, but sparse really between different sub-band, so these two class methods all exist and how effectively suppress graing lobe problem.Sparse reconstruct class algorithm can solve sparse sampling problem, but such algorithm does not also have effective fast algorithm at present, for the processing of mass data, also has the too low problem of efficiency.

Summary of the invention

The present invention proposes a kind of passive distributed SAR image processing method of differentiating reconstruct based on two-stage more, taken into full account the data sampling feature of passive distributed synthetic-aperture radar, by the many resolution imagings of two-stage, can solve the graing lobe problem that sparse sampling causes, compare traditional sparse reconstructing method simultaneously and there is better counting yield.

The present invention adopts following technical scheme for solving its technical matters:

A passive distributed SAR image processing method of differentiating reconstruct based on two-stage, comprises the steps: more

(1) target echo signal airboarne receiver being received, by one group of bandpass filter, is separated the target echo signal from different radiation sources, and each subband signal of separating is demodulated to baseband signal by down coversion;

(2) each passage restituted signal that separation obtains to step 1 carries out respectively polar format conversion, then adopts Fourier reconstruction algorithm processes to obtain the coarse resolution image of target;

(3) step 2 is processed to a plurality of coarse resolution images that obtain, for each coarse resolution pixel cell, between image, utilize sparse reconstructing method to obtain the high-definition picture of this pixel cell internal object, finally by data screening and Image Mosaics, obtain the high-definition picture of whole scene.

Beneficial effect of the present invention is as follows:

Comparing direct Fourier's reconstruct and direct sparse reconstruct, there is following advantage in the two-stage reconstructing method that the present invention adopts:

(1) in data, do not meet under Nyquist sampling thheorem condition, the present invention can well overcome the graing lobe effect that direct Fourier's reconstructing method faces.

(2) the inventive method first order adopts Fourier's reconstruct, and counting yield is very high, although the second level is sparse reconstruct, data dimension reduces greatly, and therefore whole counting yield is compared direct sparse reconstruct will very big raising.

Accompanying drawing explanation

Fig. 1 (a) is double-basis radar data acquisition geometric model; Fig. 1 (b) is double-basis radar spatial frequency domain sampling location.

Fig. 2 (a) is distributed radar data acquisition geometric model; Fig. 2 (b) is distributed radar spatial frequency domain sampling location.

Fig. 3 is algorithm process process flow diagram of the present invention.

Fig. 4 is algorithm process schematic flow sheet of the present invention.

Fig. 5 is polar format conversion schematic diagram.

Fig. 6 is the sparse reconstruct schematic diagram in the second level.

Fig. 7 (a) is single resolution element essence resolution analysis schematic diagram; Fig. 7 (b) is many resolution element essence resolution analysis schematic diagram.

Embodiment

Below in conjunction with accompanying drawing, the invention is further described in detail.For convenience of description, the signal model of the distributed passive synthetic-aperture radar of paper and data characteristics.

Signal model:

Passive distributed synthetic-aperture radar utilizes a plurality of chances source (TV, broadcast or communication signal source) of space sparse distribution as irradiation source, then utilizes the radar receiver being loaded on motion platform to receive the signal of area-of-interest target reflection.Each radiation source and receiver form a double-basis pair.First we consider the signal model that single double-basis is right.

Tentation data acquisition geometry model is as shown in Fig. 1 (a), and to simplify the analysis, we only consider how much of bidimensionals.In this geometric model, region of interest centers is defined as true origin, and radiation source and receiver location can be definite by utmost point footpath and polar angle, i.e. (r _tc, θ _t) and (r _rc, θ _r) represent respectively radiation source and receiver location, subscript T wherein, R represents respectively transmitter and receiver, subscript c represents aperture center.Suppose that in scene, having a point target, its position is (x _t, y _t), subscript t represents target, scattering coefficient is σ.This target is respectively r to the position of radiation source and receiver _tand r _r.

Suppose that radiation source transmits as s (τ)=exp (i2 π f τ), in formula, i is imaginary unit, and f is signal frequency, τ be distance to the time, exp represents exponential function.Receiver receives this signal of target reflection.After demodulation, receiving signal can be expressed as

$r\left(\mathrm{\τ}\right)=\mathrm{\σ}\·\exp [-i2\mathrm{\πf}\frac{r_T+r_R}{c}]---\left(1\right)$

In formula, c is propagation velocity of electromagnetic wave, and σ is target scattering coefficient.

This signal is carried out, after motion compensation process, can obtaining

$r\left(\mathrm{\τ}\right)=\mathrm{\σ}\·\exp \left[i2\mathrm{\πf}\frac{r_{\mathrm{Tc}}+r_{\mathrm{Rc}}-r_T-r_R}{c}\right]---\left(2\right)$

Under plane wave front assumed condition, formula (2) can be reduced to

r(τ)＝σ·exp{i2πf[x _t(cosθ _T+cosθ _R)+y _t(sinθ _T+sinθ _R)]}＝σ·exp{i2πf[x _tk _x+y _tk _y]} (3)

Wherein

$\begin{array}{c}{k}_x=\frac{2\mathrm{\πf}}{c}(\cos {\mathrm{\θ}}_T+\cos {\mathrm{\θ}}_R)\\ k_y=\frac{2\mathrm{\πf}}{c}(\sin {\mathrm{\θ}}_T+\sin {\mathrm{\θ}}_R)\end{array}---\left(4\right)$

Be respectively the spatial frequency of x and y direction.They also available following polar format represent

$\begin{array}{c}{k}_r=\sqrt{k_x^2+k_y^2}=\frac{2\mathrm{\πf}}{c}\cos \left(\frac{{\mathrm{\θ}}_T-{\mathrm{\θ}}_R}{2}\right)\\ {\mathrm{\θ}}_b=a\tan \left(\frac{k_y}{k_x}\right)=a\tan \left(\frac{\sin {\mathrm{\θ}}_T+\sin {\mathrm{\θ}}_R}{\cos {\mathrm{\θ}}_T+\cos {\mathrm{\θ}}_R}\right)=\frac{{\mathrm{\θ}}_T+{\mathrm{\θ}}_R}{2}\end{array}---\left(5\right)$

From formula (3), we see, the radar echo signal after demodulation is actually a sampling in target Fourier frequency space, and sampling location is as shown in Fig. 1 (b) orbicular spot.This sampling location is determined by position angle and the emission signal frequency of emittor/receiver, wherein transmitter-side parallactic angle and emission signal frequency determine position and the size of circle in figure, transmitter and receiver position angle determines double-basis angle (position angle corresponding to dotted line in figure), and the intersection point that circle and double-basis angle are pointed to is frequency domain sample position.Change this three parameters (f, θ _t, θ _r), just can change data at the sampling location of frequency domain, θ _tfor the position angle of radiation source, θ _rposition angle for receiver.

From analysis above, we learn, single double-basis can provide a sampling of object space frequency domain to (single-frequency, fixed transmission and receiving position are observed target), Airborne Passive distributed radar is by transmitting and receiving diversity, the sampling of spatial frequency domain diverse location can be provided, thus for high resolution target reconstruct provide may.

Radiation source diversity

A plurality of chance radiation sources that distributed passive radar utilization is distributed in space diverse location irradiate target, to realize the multi-angle observation to target, therefore can obtain emission space diversity.

The signal of chance radiate source radiation all has certain bandwidth conventionally, if typical TV signal bandwidth is in 8MHz left and right, utilizes these signals with certain bandwidth to observe target, can obtain frequency diversity.

In addition, different chance radiation sources are usually operated at different wave bands, and the signal of different radiate source radiations has different carrier frequency, and the difference of this multi radiation sources signal carrier frequency also can obtain frequency diversity.

Receiver diversity

Typical receive diversity is to utilize a plurality of receivers that are distributed in space diverse location to receive target scattering signal.This diversity mode is on the one hand due to a plurality of phy receivers of needs thereby there is high cost problem, secondly how to keep or the amplitude-phase consistency that compensates between each receiver is also a challenging problem.Adopt airboarne receiver can overcome well above-mentioned shortcoming, airboarne receiver can be realized the multi-angle observation of single receiver to target by means of the motion of platform, thereby obtains space diversity.

Below we consider passive distributed synthetic-aperture radar signal model, without loss of generality, our hypothesis space is distributed with J radiation source, the signal of each radiate source radiation different frequency irradiates target, the carrier frequency that might as well establish j radiate source radiation signal is f _cj, bandwidth is B _j, the receiver of motion receives the echo of target scattering with certain recurrence interval, and at coherent accumulation, in the time, (accumulation angle is designated as Δ θ) collects echoed signal altogether K time.According to formula (3), be not difficult to obtain radar receiver and can be expressed as (radiation source works in different-waveband, therefore at receiving end, by simple filtering, just the signal of different radiation sources can be separated) at k target echo signal that receives j radiation source the time of reception

$\begin{array}{c}r(j,k,l)=\mathrm{\σ}\·\exp \left\{i\frac{2\mathrm{\π}(f_{\mathrm{cj}}+f_l)}{c}\right[x_t(\cos {\mathrm{\θ}}_{\mathrm{Tj}}+\cos {\mathrm{\θ}}_{\mathrm{Rk}})+y_t(\sin {\mathrm{\θ}}_{\mathrm{Tj}}+\sin {\mathrm{\θ}}_{\mathrm{Rk}})\left]\right\}\\ =\mathrm{\σ}\·\exp \left\{i\right[x_t{k}_x^{\mathrm{jkl}}+y_t{k}_y^{\mathrm{jkl}}\left]\right\}\end{array}---\left(6\right)$

θ in formula _tjbe the position angle of j radiation source, fl is l sampling frequency (L the frequency of supposing to have sampled altogether) in radiation signal bandwidth range, θ _rkfor the position angle of airboarne receiver k the time of reception, $k_x^{\mathrm{jkl}}=\frac{2\mathrm{\π}(f_{\mathrm{cj}}+f_l)}{c}(\cos {\mathrm{\θ}}_{\mathrm{Tj}}+\cos {\mathrm{\θ}}_{\mathrm{Rk}})$ With $k_y^{\mathrm{jkl}}=\frac{2\mathrm{\π}(f_{\mathrm{cj}}+f_l)}{c}(\sin {\mathrm{\θ}}_{\mathrm{Tj}}+\sin {\mathrm{\θ}}_{\mathrm{Rk}})$ Be expressed as the spatial frequency of x and y direction.Due to space and frequency diversity, the data that airborne distributed radar collects are a plurality of samplings of purpose-function space frequency domain, sampling location determines by the mode of space and frequency diversity completely, and subscript Tj represents j radiation source, and subscript Rk represents k position of receiver.In formula (6), common f _lthe intensive sampling in radiation signal bandwidth, θ _rkthe intensive sampling in receiver coherent accumulation angle, and the locus θ of radiation source _tjwith radiation signal carrier frequency f _cjsparse distribution normally, therefore in the sampling of frequency domain, to support normally segmentation sparse for airborne distributed radar.As Fig. 2 (b) has provided the airborne distributed radar that comprises three radiation sources (supposing configuration as shown in Fig. 2 (a)) echo data at object space frequency domain sample position view, the effective Support that can see spatial frequency domain consists of three sub-Support, every sub-Support is corresponding with a radiation source, the size of sub-Support depends on radiation signal bandwidth and receiver coherent accumulation time, simultaneously relevant with carrier frequency and double-basis angle again, the interval of different sub-Support is determined jointly by position and the carrier frequency difference of each radiation source.

Treatment scheme:

From analysis above, we know, the data of radar admission are actually the discrete sampling in objective function Fourier frequency domain space, for passive distributed synthetic-aperture radar, each radiation source provides signal that a sub-band sample of spatial frequency domain is provided, this sampling is that segmentation is sparse, in subband, sampling meets Nyquist theorem, but subband and intersubband are normally sparse.Consider this feature of passive distributed synthetic-aperture radar echo data, the invention provides a kind of two-stage imaging processing algorithm.The first order is processed each subband signal, because sampling in subband meets Nyquist theorem, therefore the present invention utilizes Fourier's reconfiguration technique each subband signal processing to be obtained to the coarse resolution image of target, and the second level utilizes sparse reconfiguration technique to obtain the smart image in different resolution of target to each coarse resolution pixel between coarse resolution image.The side lobe effect of considering first order Fourier reconstruct can exist and disturb second level essence resolution reconstruct, therefore after sparse reconstruct, also need data to screen, reject secondary lobe disturbing effect, finally smart resolution imaging result is spliced to the full resolution pricture that obtains whole scene.As shown in Figure 3, its schematic diagram is as Fig. 4 for whole treatment scheme.Treatment scheme is done to more detailed description below.

The reconstruct of first order coarse resolution

By formula (6), be not difficult to learn, each subband data is the bidimensional discrete sampling of scene objects function Fourier, and generally, sampling is at (f _l, θ _rk) territory is uniformly, but is mapped to spatial frequency domain (k _x, k _y) after, but by polar format, sample.In order to utilize Fast Fourier Transform (FFT) to realize Fourier's reconstruct, polar format sampled data need be converted to matrix format sampled data.This process as shown in Figure 5.

After polar format conversion, subband data can be expressed as

$r(j,k,l)=\mathrm{\σ}\·\exp \left\{i\right[x_t{k}_x^{\mathrm{jkl}}+y_t{k}_y^{\mathrm{jl}}\left]\right\}---\left(7\right)$

In formula $k_x^{\mathrm{jk}}=k_{\mathrm{xc}}^j+(k-\frac{K}{2}){\mathrm{\Δ}}_{k_x},$ K=1 ..., K and $k_y^{\mathrm{jl}}=k_{\mathrm{yc}}^j+(l-\frac{L}{2}){\mathrm{\Δ}}_{k_y},$ L=1 ..., L represents respectively k _xand k _ythe resampling position in territory, wherein K be orientation to sampling number, with be respectively orientation to sampling central value and sampling interval, L be distance to sampling number, L with be respectively distance to sampling central value and sampling interval.Convenient for subsequent treatment, different sub-band resampling position is defined in a unified rectangular coordinate system.Different sub-band resampling position has different supporting domain centers.To j subband, its wavenumber domain center is determined by following formula

$k_{\mathrm{xc}}^j=\frac{2\mathrm{\π}{f}_{\mathrm{cj}}}{c}(\cos {\mathrm{\θ}}_{\mathrm{Tj}}+\cos {\mathrm{\θ}}_{\mathrm{Rc}}),k_{\mathrm{yc}}^j=\frac{2\mathrm{\π}{f}_{\mathrm{cj}}}{c}(\sin {\mathrm{\θ}}_{\mathrm{Tj}}+\sin {\mathrm{\θ}}_{\mathrm{Rc}})---\left(8\right)$

After subband data is changed by polar format, then do the coarse resolution image that a bidimensional FFT can obtain target

$r(j,n,m)=\mathrm{\σ}\·A^j(\frac{n}{K{\mathrm{\Δ}}_{k_x}}-x_t,\frac{m}{L{\mathrm{\Δ}}_{k_y}}-y_t)\·\exp \left\{i\right[x_t{k}_{\mathrm{xc}}^j+y_t{k}_{\mathrm{yc}}^j\left]\right\}---\left(9\right)$

A wherein ^j(x, y) is the response function of point target after j subband data processing, with be respectively the pixel size of x and y direction, n represents n pixel of x direction, and m represents m pixel of y direction.

By first order Fourier reconstruct, can obtain the coarse resolution location estimation of target

$\begin{array}{c}{\hat{x}}_t=\frac{n_0}{K{\mathrm{\Δ}}_{k_x}}\\ {\hat{y}}_t=\frac{m_0}{L{\mathrm{\Δ}}_{k_y}}\end{array}---\left(10\right)$

N wherein ₀and m ₀be respectively the pixel cell at target place.X and y direction estimated accuracy are respectively with

Second level essence is differentiated reconstruct

Sparse reconstruct

Due to emitter Signals bandwidth and radar receiver coherent accumulation angle conventionally all smaller, so the common resolution of subband signal reconstructed image is lower.Fortunately, conventionally have a plurality of emitter Signals, if a plurality of subband signals that therefore we can provide a plurality of radiation sources fully utilize, high-resolution imaging is possible.When each subband data is when spatial frequency domain continuous seamless distributes, between coarse resolution image, for each pixel cell, the Fourier analysis that tries again, just can obtain the high-definition picture of target.And for passive distributed radar, as noted earlier, subband data is sparse distribution at spatial frequency domain conventionally, therefore cannot directly carry out Fourier analysis.Sparse reconstruct theory is told us, for irregular sparse sampling signal, utilize non-linear reconstructing method can realize the Accurate Reconstruction to sparse target, so let us consider to adopt sparse reconstructing method to solve the high-resolution reconstruction problem of coarse resolution unit internal object.

For each coarse resolution unit, we are refined as N ²equal portions (distance and bearing is divided into N equal portions), are refined as distance and bearing pixel cell with like this, the echo signal in each coarse resolution unit can be modeled as

t＝Φw (11)

T=(t in formula ₁..., t _j) ^tfor the pixel value of same position pixel cell in J coarse resolution image, for the N after refinement ²the scattering coefficient of target on individual smart pixel cell, T represents vectorial transposition, Φ is J * N ²the perception matrix of size, the element that the capable n of its j lists is

(x in formula _c, y _c) be the coordinate of coarse resolution unit center, (x _n, y _n) be the coordinate of n essence resolution pixel cell in this coarse resolution unit, A ^jthe amplitude response that represents j coarse resolution image.

In formula (11), conventionally, J<<N ², so this equation is underdetermined equation, cannot obtain the unique solution of w.But suitably relax when condition, while for example needing the target of reconstruct sparse, by suitable non-linear reconfiguration technique, still can obtain the high-definition picture of target.Here, we just suppose to meet this sparse condition, this is because really there is the sparse situation of target on the one hand, on the other hand, even target is not sparse on stricti jurise, but the strong scattering point that conventionally only has minority in target, when we only put when interested strong scattering, all the other non-strong scattering points can be thought noise.Consider in radar return signal simultaneously and inevitably have thermonoise, therefore actual signal model can be expressed as

t＝Φw+n (13)

Wherein n represents noise.

Want sparse reconstruct target, except meeting the sparse condition of target, perception matrix F is also had to certain requirement.We consider the situation that radiation source is compact, and with this understanding, in formula (12), different coarse resolution image mid point target response functions are approximate identical, so perception matrix can be approximated to be the irregular sparse sampling of Fourier's matrix.Already proved, this perception matrix meets the RIP condition of sparse reconstruct with very large probability.

Under above condition satisfies condition, by solving following optimization procedure, just can obtain the high-definition picture of target

$\underset{w}{\min }\left|\right|t-\mathrm{\&Phi;w}{\left|\right|}_2^2+\mathrm{\&lambda;}{\left|\right|w\left|\right|}_0---\left(14\right)$

In formula || || ₀with || || ₂represent respectively 0 norm and 2 norms, λ is regularization parameter.

To solving of this problem, existing a lot of algorithms, repeat no more here.

To each coarse resolution unit, all utilize above-mentioned sparse reconstructing method to obtain target high-definition picture, last splicing more just can obtain the high-definition picture of whole scene.Whole process as shown in Figure 6

Sidelobe Suppression

What consider the imaging of first order coarse resolution we adopt is Fourier's reconstructing method, and we know, in Fourier's reconstructed image, inevitably can there is side lobe effect, even therefore for a single point target, in coarse resolution image, except unit, main lobe place, close on unit and also can have the side-lobe signal of this target.If side lobe effect is not suppressed, when second level essence resolution imaging, except can going out target by high-resolution reconstruction, also can produce undesirable diplopia closing on unit in main lobe unit.Below we illustrate this phenomenon with the example that is reconstructed into of one-dimensional signal.Suppose to have a point target, its coarse resolution image is as shown in the first row in Fig. 7 (a), wherein dotted line representative point target response, empty circles represents coarse resolution sampling location, can see that this target bit is i coarse resolution unit, but due to side lobe effect, on its adjacent cells, also there is the signal of this target.When i coarse resolution unit carried out to smart resolved analysis, no doubt can correctly obtain the high-definition picture of this target, as shown in the third line in figure.But when adjacent cells is carried out to smart resolved analysis, when to i-1 coarse resolution element analysis, this target also can be aliased into present analysis unit, thereby causes diplopia, as shown in last column in figure.The diplopia phenomenon causing in order to overcome coarse resolution image side lobe effect, we can solve by increasing smart resolved analysis cell width.For example, we expand to three coarse resolution unit by the width of smart resolved analysis, and the side lobe effect of adjacent cells target just can be effectively suppressed.As shown in Fig. 7 (b), when analyzing i-1 and i+1 coarse resolution unit, in i unit, the side-lobe signal of target also can correctly navigate to its actual position place, and aliasing can not occur.Now, only need screen data, retain unit object to be analyzed, and reject all the other, close on unit object, can realize effective secondary lobe and disturb inhibition.After final splicing, in smart image in different resolution, there will not be diplopia, as shown in last column in Fig. 7 (b).

Image Mosaics

Obtain each coarse resolution unit after resolution imaging result, directly by the splicing of position, coarse resolution unit, can obtain the high-resolution picture of whole scene.

Claims (1)

1. based on two-stage, differentiate a passive distributed SAR image processing method for reconstruct more, it is characterized in that, comprise the steps:

(1) target echo signal airboarne receiver being received, by one group of bandpass filter, is separated the target echo signal from different radiation sources, and each subband signal of separating is demodulated to baseband signal by down coversion;

(2) each passage restituted signal that separation obtains to step 1 carries out respectively polar format conversion, then adopts Fourier reconstruction algorithm processes to obtain the coarse resolution image of target;

(3) step 2 is processed to a plurality of coarse resolution images that obtain, for each coarse resolution pixel cell, between image, utilize sparse reconstructing method to obtain the high-definition picture of this pixel cell internal object, finally by data screening and Image Mosaics, obtain the high-definition picture of whole scene.