Space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar

Abstract

The present invention discloses a space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar, constructs a space-time two-dimensional signal receiving model, constructs the space-time snapshot signal corresponding to each distance-azimuth unit according to the multi-channel data after pulse compression; calculates the space-time guidance vector corresponding to each azimuth angle to form a space-time guidance vector matrix; establishes a cost function based on the minimum mean square error criterion, continuously iterates and updates the weight vector to make it close to the optimal, and weights the received space-time snapshot to obtain the azimuth super-resolution spectrum curve corresponding to the space-time snapshot of the distance-azimuth unit; according to the carrier movement speed and antenna scanning parameters, the super-resolution spectrum curve of each distance-azimuth unit is spliced ​​in azimuth, that is, the spectrum curve is accumulated in the distance-azimuth domain, and finally a distance-azimuth imaging image is formed. The present invention effectively improves the azimuth resolution of the existing airborne forward-looking array radar by jointly processing super-resolution technology in the space-time domain.

Description

Technical Field

[0001] The present invention belongs to the field of radar imaging technology, relates to an airborne radar forward-looking imaging signal processing technology, and specifically relates to a space-time adaptive iterative super-resolution imaging method for an airborne forward-looking array radar. Background Art

[0002] Airborne radar forward imaging plays an important role in military applications such as enemy detection, aircraft guidance, complex terrain and natural disaster rescue. However, when the target is located in the forward position, the equidistance line is parallel to the equi-Doppler line, making existing imaging technologies such as synthetic aperture radar (SAR) or Doppler beam sharpening (DBS) ineffective.

[0003] Real beam imaging technology can be used to complete forward-looking imaging. However, its azimuth resolution is limited by the antenna beam width and the range of action, and it is impossible to achieve high-resolution imaging in azimuth. In order to improve the resolution of real aperture radar (RAR), scientific researchers have proposed many algorithms to improve the azimuth resolution. At present, the main forward-looking imaging technologies include: real beam imaging technology, deconvolution imaging technology, dual-base SAR forward-looking imaging technology, single pulse imaging technology, array radar super-resolution imaging technology, etc.

[0004] Deconvolution imaging technology uses the convolution of the radar echo's azimuth direction as the horizontal plane of the antenna pattern and the azimuth scattering point, and the echo's range direction as the vertical plane of the antenna pattern and the range scattering point. Therefore, ideally, the accurate position of the target can be reconstructed by deconvolution in the range and azimuth directions respectively. However, because the antenna pattern used in the deconvolution process is different from the actual one, it is difficult to obtain an accurate and stable solution, and the ideal azimuth resolution cannot be achieved.

[0005] In the process of forward-looking imaging in bistatic SAR, the transmitter and receiver are placed on two different platforms, and the equidistance lines and the equi-Doppler lines are approximately orthogonal, which effectively improves the resolution of forward-looking imaging. However, this technology is not yet mature, and there are many new theoretical and technical problems, which cannot be used in actual situations.

[0006] The single pulse technology uses two antennas to receive echoes at the same time, and a single pulse can be used to obtain the angle of the scattering point deviating from the center of the beam. The single pulse has high angle measurement accuracy and a fast data acquisition rate, but when the imaging area has complex terrain and the carrier platform is moving, the apparent center of the target will deviate from the actual center of the target, resulting in angular scintillation. Moreover, it cannot distinguish multiple targets in a beam, so when there are multiple targets in a beam, the performance deteriorates sharply.

[0007] Compared with traditional single-channel radar, array radar super-resolution imaging technology uses an array receiving system, which has higher spatial freedom and improved spatial resolution. At the same time, the super-resolution technology in array signal processing is introduced into forward-looking imaging, thereby obtaining an angular resolution far exceeding the real aperture beam width, which can better separate multiple targets in a beam. In recent years, as a new technical means for airborne radar forward-looking imaging, super-resolution technology has gradually been recognized and valued by relevant research institutions. It is one of the simple and efficient means to solve the problem of airborne radar forward-looking imaging in the future.

[0008] The physical essence of super-resolution angle measurement based on array signals lies in the expansion of the antenna aperture in the direction of the tangent track. One method is to directly expand the array data through the linear prediction method, and the other method is to perform super-resolution estimation of the direction of arrival by the echo signals received by each array element. For example, the Capon spectrum estimation method and the multiple signal classification (MUSIC) algorithm have been applied to airborne radar forward imaging, and their feasibility has been verified through simulation and measured data processing. However, the Capon algorithm has poor robustness, and the MUSIC method relies on prior information on the number of targets and cannot estimate the target amplitude. For the angle of arrival estimation, in order to more accurately estimate the autocorrelation matrix of the array received signal, the above traditional methods need to accumulate a large number of independent and identically distributed snapshot samples. However, the echo data on the same range gate are often coherent, which will greatly affect the estimation results. Summary of the invention

[0009] Purpose of the invention: The present invention provides a space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar, which does not rely on a large amount of independent and identically distributed snapshot data. It mainly estimates the autocorrelation matrix of the array received signal in an iterative manner by establishing an objective function under the minimum mean square error criterion.

[0010] Technical solution: The space-time iterative super-resolution method for airborne radar forward imaging described in the present invention comprises the following steps:

[0011] (1) Constructing a space-time two-dimensional signal reception model: Using an airborne forward-looking array radar to obtain multi-channel echoes, the radar operates in a step-scan mode, transmitting several coherent pulses at the same azimuth in each coherent pulse interval, and sampling the echoes to obtain range-pulse-array three-dimensional echo data;

[0012] (2) constructing a space-time snapshot signal corresponding to each range-azimuth unit based on the multi-channel data after pulse compression; calculating the space-time steering vector corresponding to each azimuth angle to form a space-time steering vector matrix;

[0013] (3) Processing the space-time snapshots in each range-azimuth unit, using the space-time steering vector matrix as the initial weight vector to obtain the initial value of the iteration; establishing a cost function based on the minimum mean square error criterion, calculating and simplifying the weight vector expression; then substituting the initial value of the iteration to start the iteration process, updating the current weight vector through continuous iteration to make it close to the optimal value, and obtaining the final weight vector; using the final weight vector to perform weighted processing on the space-time snapshot, and calculating the super-resolution spectrum curve corresponding to the space-time snapshot of the range-azimuth unit;

[0014] (4) According to the carrier motion speed and antenna scanning parameters, the super-resolution spectrum curve of each range-azimuth unit is spliced ​​in azimuth, that is, the spectrum curve is accumulated in the range-azimuth domain, and finally a range-azimuth imaging image is formed.

[0015] Furthermore, the implementation process of step (1) is as follows: a horizontal array antenna along the azimuth direction is used to receive multi-channel echoes to obtain spatial sampling; the radar operates in a step scanning mode, and when the beam is directed to a fixed azimuth, a linear frequency modulated pulse (LFM) is emitted every pulse repetition frequency (PRF) or pulse repetition interval (PRI) within a coherent pulse interval (CPI), and a number of coherent pulses are received to obtain time sampling; the beam scans one circle to obtain range-pulse-array three-dimensional echo data.

[0016] Furthermore, the implementation process of step (2) is as follows:

[0017] According to the data obtained by the horizontal array antenna system of the airborne array radar, pulse compression processing is performed in the range direction, and the two-dimensional receiving data generated by each array element in each range-azimuth unit is selected to form the space-time snapshot signal of the range-azimuth unit; the space-time snapshot signal of each range-azimuth unit is stacked to convert the two-dimensional matrix into a one-dimensional vector, and then the corresponding space-time guidance vectors are calculated for all azimuth angles in the imaging area, and the space-time guidance vector matrix is ​​combined; the specific implementation process is as follows:

[0018] Assume that there are K targets in the front imaging area, and their azimuth angles relative to the radar are θ1, θ2, …, θ K ; In one cycle scan, the received data is a three-dimensional sampling of range-pulse-array. For a certain range-azimuth unit, the pulse-array sampling data constitutes a two-dimensional space-time snapshot:

[0019] y=Ax+N

[0020] Where x is the echo signal reflected from the target, N is additive white Gaussian noise, independent of other components; A is the NM×K-dimensional space-time guidance vector matrix, A = [a(θ1), a(θ2), …, a(θ K )],in as (θ k ) is an M×1-dimensional spatial guidance vector composed of the phase difference of the target reaching the horizontal linear array; a t (θ k ) is an N×1-dimensional time-steering vector composed of the Doppler frequencies generated by the relative motion between the radar and the target, represents the Kronecker product.

[0021] Furthermore, the implementation process of step (3) is as follows:

[0022] Accurately reconstruct the reflection cross-sectional area x of the target, assuming that the estimate of x is:

[0023]

[0024] Among them, w is the weight coefficient matrix to be determined, y is the two-dimensional space-time snapshot of a certain range-azimuth unit, and the minimum mean square error criterion (MMSE) is used to establish the cost function:

[0025] min J{||xw H y|| 2}

[0026] Derivative solution of the above equation yields the minimum point:

[0027] w=(E{yy H}) -1 E{yx H}

[0028] Substituting the above formula into the simplified form, we get:

[0029] w=(APA H +R N ) -1 AP

[0030] in, is the covariance matrix of the noise, P is the autocorrelation matrix of the echo signal x reflected from the target, P = diag{p1,p2…,p K}, where the diagonal elements are:

[0031]

[0032] Calculate the first estimation result, at this time the weight matrix w1 = A, Based on this, the signal's autocorrelation matrix P1 is calculated, and then the weight matrix is ​​updated:

[0033] w2=(AP1A H +R N ) -1 AP1

[0034] The second estimation result is obtained. And so on, using the estimate from the previous iteration Calculate the matrix P k , and then substitute into the formula to update the weight matrix w k+1 ; Finally, the new weight matrix is ​​applied to Get a new estimate, that is, the iterative expression is:

[0035] w k+1 =(AP k A H +R N ) -1 AP k

[0036]

[0037] The echo autocorrelation matrix R = AP k A H It does not rely on a large amount of snapshot data, but is obtained through iteration.

[0038] Furthermore, the scanning parameters in step (4) include a beam scanning range and a beam scanning speed.

[0039] Furthermore, the implementation process of step (4) is as follows:

[0040] The azimuth super-resolution spectrum curve corresponding to the space-time snapshot of each range-azimuth unit obtained in step (3) is spliced ​​at the corresponding azimuth angle for the azimuth super-resolution spectrum curves of different azimuth units in the same range gate, that is, the accumulation of the spectrum curve in the range-azimuth domain is achieved, and the range-azimuth imaging result can be obtained by displaying the accumulated stored data.

[0041] Beneficial effects: Compared with the prior art, the beneficial effects of the present invention are as follows: This method establishes a geometric model for forward-looking imaging of airborne array radars, obtains the space-time two-dimensional sampling information of echoes, and uses a single space-time snapshot to reconstruct the corresponding azimuth super-resolution spectrum curve through iteration without relying on a large number of independent and identically distributed snapshot data, which solves the problem of being unable to accumulate a large number of snapshot samples in actual situations, and the estimation results are more accurate and stable. While adopting multi-channel technology, the Doppler frequency information generated by the relative motion between the radar and the target in airborne forward-looking imaging is effectively utilized, and the imaging scene can be reconstructed more accurately. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 is a flow chart of the present invention;

[0043] Figure 2 This is a schematic diagram of the geometric model of airborne array radar forward imaging;

[0044] Figure 3 This is a schematic diagram of the azimuth space-time snapshot sampling model of the airborne array radar;

[0045] Figure 4 This is a schematic diagram of azimuthal super-resolution spectrum splicing;

[0046] Figure 5 This is the result diagram of real beam imaging of point target;

[0047] Figure 6 This is the result of space-time iterative super-resolution imaging of point targets;

[0048] Figure 7 It is the real beam imaging profile of point target;

[0049] Figure 8 It is a spatial and temporal iterative super-resolution imaging profile of a point target;

[0050] Fig. 9 is the target scene graph;

[0051] Fig.10 Comparison diagram of target scene imaging results; (a) is the real beam imaging result diagram, (b) is the adaptive iterative super-resolution imaging result diagram based on spatial one-dimensional echo, and (c) is the imaging result diagram after five iterations of spatial-temporal two-dimensional echo.

[0052] Fig.11 for Fig.10 The final reconstructed result image projected into the geodetic coordinate system, where (a) is the spatial domain adaptive iterative super-resolution imaging image, and (b) is the spatial-temporal adaptive iterative super-resolution imaging image. DETAILED DESCRIPTION

[0053] The present invention is further described in detail below with reference to the accompanying drawings.

[0054] The present invention provides a space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar, and its processing flow is as follows: Figure 1 As shown, the following steps are included:

[0055] Step 1: Construct a space-time two-dimensional signal reception model. Use an airborne forward-looking array radar to obtain multi-channel echoes. The radar works in a step-scan mode, emitting several coherent pulses at the same azimuth in each coherent pulse interval, and sampling the echoes to obtain range-pulse-array three-dimensional echo data.

[0056] A horizontal array antenna receiving system along the azimuth direction is used to receive multi-channel echoes to obtain spatial sampling data; the array radar works in a step-scan mode. When the beam is directed to a fixed azimuth, a linear frequency-modulated pulse is emitted every pulse repetition frequency (PRF) or pulse repetition interval (PRI) within a coherent pulse interval (CPI), and several coherent pulse echoes are received to obtain time sampling data.

[0057] The geometric model of airborne array radar forward imaging is as follows: Figure 2 As shown, the multi-channel forward-looking radar is at a speed v a Flying at a constant speed, the array antenna system is fixed on the mobile platform along the track direction. The array antennas are evenly spaced to form M receiving channels with a channel spacing of d. The radar scans the area ahead with a narrow beam step. When the beam center points to a fixed azimuth angle θ (θ is the angle between the beam center and the carrier speed), within a coherent pulse interval (CPI), the radar transmitter transmits N coherent linear frequency modulated pulses (LFM), and then the beam points to the next azimuth angle and repeats the operation. Finally, the different pulse echoes received by the same array element are sampled and arranged in columns to obtain the original range-pulse two-dimensional echo matrix. The received data of multiple array elements are combined to obtain the range-pulse-array three-dimensional echo data.

[0058] When the beam center points to the azimuth angle θ, the M×1-dimensional spatial steering vector composed of the phase difference of the target reaching the horizontal linear array is:

[0059]

[0060] The N×1-dimensional time steering vector consists of the Doppler frequency generated by the relative motion between the radar and the target:

[0061]

[0062] Therefore, the NM×1-dimensional space-time steering vector formed by the two steering vectors is:

[0063]

[0064] in, represents the Kronecker product, d is the channel spacing, λ is the signal wavelength, θ is the beam center pointing angle, and v a The carrier speed.

[0065] Step 2: Construct the space-time snapshot signal corresponding to each range-azimuth unit based on the multi-channel data after pulse compression; calculate the space-time steering vector corresponding to each azimuth angle to form a space-time steering vector matrix.

[0066] According to the data obtained by the horizontal array antenna system of the airborne array radar, pulse compression processing is performed in the range direction, and the two-dimensional receiving data generated by each array element in each range-azimuth unit is selected to form the space-time snapshot signal of the range-azimuth unit, such as Figure 3 As shown. The obtained space-time snapshot signals of each range-azimuth unit are stacked to convert the two-dimensional matrix into a one-dimensional vector. Then, for all azimuth angles in the imaging area, the corresponding space-time steering vector a(θ) is calculated and combined to form the space-time steering vector matrix A.

[0067] Assume that there are K targets in the front imaging area, and their azimuth angles relative to the radar are θ1, θ2, …, θ K In one cycle scan, the received data is a three-dimensional sampling of range-pulse-array. For a certain range-azimuth unit, the pulse-array sampling data constitutes a two-dimensional space-time snapshot:

[0068] y=Ax+N

[0069] Where A is the NM×K-dimensional space-time steering vector matrix, A=[a(θ1),a(θ2),…,a(θ K )],in x is the echo signal reflected from the target, and N is additive white Gaussian noise, which is independent of other components.

[0070] Therefore, for the position at the azimuth angle θ k The goal of the department:

[0071] y=a(θ k )x k +N

[0072] Where N is additive Gaussian white noise, independent of other components. k ) is the NM×1 dimensional space-time steering vector.

[0073] Step 3: Process the space-time snapshots one by one using the space-time steering vector matrix as the initial weight vector to obtain the initial value of the iteration; establish the cost function based on the minimum mean square error criterion, and calculate and simplify the weight vector expression. Then substitute the initial value of the iteration to start the iteration process, and update the current weight vector through continuous iteration to make it close to the optimal value to obtain the final weight vector. Use the final weight vector to perform weighted processing on the space-time snapshot, and calculate the super-resolution spectrum curve corresponding to the space-time snapshot of the range-azimuth unit.

[0074] In order to minimize the mean square error between the reconstructed target amplitude and the actual signal, the present invention establishes a cost function based on the minimum mean square error (MMSE) criterion. The optimal weight vector expression is obtained by calculation and simplification, and it is found that the optimal weight vector is related to the autocorrelation matrix of the space-time snapshot of the unit and the cross-correlation matrix between the space-time snapshot and the actual target reflection signal.

[0075] Now we need to accurately reconstruct the reflection cross-sectional area x of the target, assuming that the estimation of x is:

[0076]

[0077] Where w is the weight coefficient matrix to be determined, and y is the two-dimensional space-time snapshot of a certain range-azimuth unit. The minimum mean square error criterion (MMSE) can be used to establish the cost function:

[0078] min J{||xw H y|| 2}

[0079] The derivative of the above formula can be solved to get the minimum point:

[0080] w=(E{yy H}) -1 E{yx H}

[0081] Substituting the above formula into the simplified form, we get:

[0082] w=(E{yy H}) -1 E{yx H}

[0083] =(E{(Ax+N)(Ax+N) H}) -1 E{(Ax+N)x H}

[0084] =(APA H +R N ) -1 AP

[0085] in, is the covariance matrix of the noise. P is the autocorrelation matrix of the echo signal x reflected from the target, P = diag{p1,p2…,p K}, where the diagonal elements are:

[0086]

[0087] When α=2, p k is the power of each target. However, in actual processing, the exponent α takes a number between 1 and 2 to maintain the convergence and stability of the matrix inversion.

[0088] When an accurate estimate of x is required, an iterative method can be used to continuously reduce the error between the estimated value and the true value. Next, an iterative process is established. It is necessary to continuously update the autocorrelation matrix of the estimated signal, and then iteratively update the weight vector according to the weight vector calculation formula to obtain a more accurate signal wave direction estimation result. The iterative process is established as follows:

[0089] First, calculate the first estimation result. At this time, the weight matrix w1=A. Based on this, the signal autocorrelation matrix P1 is calculated, and the subscript represents the number of iterations. Then the weight matrix is ​​updated:

[0090] w2=(AP1A H +R N ) -1 AP1

[0091] The second estimation result is obtained. And so on, using the estimate from the previous iteration Calculate the matrix P k , and then substitute into the formula to update the weight matrix w k+1 Finally, the new weight matrix is ​​applied to Get a new estimated value. That is, the iterative expression is:

[0092] w k+1 =(AP k A H +R N ) -1 AP k

[0093]

[0094] Usually, a good reconstruction effect can be obtained by iterating each snapshot about 5 times. The autocorrelation matrix of the echo R = AP k A H It does not rely on a large amount of snapshot data, but is obtained through iteration.

[0095] Table 1 summarizes the algorithm flow of the space-time iterative adaptive algorithm.

[0096] Table 1. Space-time iterative adaptive super-resolution algorithm process

[0097]

[0098] Step 4: According to the carrier movement speed and antenna scanning parameters (beam scanning range, beam scanning speed), the spectrum curve of each range-azimuth unit is spliced ​​in azimuth, that is, the spectrum curve is accumulated in the range-azimuth domain, and finally a range-azimuth imaging image is formed.

[0099] The azimuth spectrum curve corresponding to each range-azimuth unit space-time snapshot obtained in step 3 is summed up for the corresponding azimuth angles of the azimuth spectrum curves of different azimuth units in the same range gate. The schematic diagram is shown in Figure 4 As shown, the distance and azimuth imaging results can be obtained by displaying the accumulated and stored data.

[0100] Next, we simulate point targets. The five point targets are located at [750m, -1.5°], [500m, 0°], [750m, 0°], [1000m, 0°] and [750m, 1.5°], with the same amplitude. The 3dB width of the real beam is about 4 degrees. The real beam scanning imaging results are shown in Figure 2. Figure 5 As shown in the figure, it can be seen that after real beam imaging, three close targets at the same range gate cannot be distinguished. However, after imaging processing by the space-time adaptive iterative super-resolution algorithm, the five point targets are clearly visible. The imaging results are shown in Figure 6 shown. Figure 7 , Figure 8 This is the azimuth imaging profile at the same range gate (750m). Figure 7 This is the real beam imaging profile result. The three targets cannot be distinguished. Figure 8 This is the imaging profile result of the space-time adaptive iterative super-resolution algorithm. The three targets in the same range gate are located at -1.5°, 0° and 1.5° respectively after imaging. This proves that the algorithm can effectively distinguish multiple targets in the main lobe, providing feasibility for airborne radar forward imaging.

[0101] A high-resolution SAR image is selected as the ground simulation scene. The simulation scene is the airport runway area. The target scene is shown in the figure below. Fig. 9 As shown, radar echo is generated in this way. Fig.10 For the comparison of target scene imaging results, Figure 10(a) shows the real beam imaging result. It can be seen that the angular resolution is low, point targets cannot be distinguished, and the contour features of surface targets are blurred. Fig.10 (b) and (c) both use adaptive iterative super-resolution algorithms for imaging, but Fig.10 (b) Adaptive iterative super-resolution imaging based on spatial one-dimensional echo, which is the imaging result after 5 iterations. Fig.10 (c) is the imaging result after 5 iterations of the space-time two-dimensional echo. From the imaging results, it can be seen that the azimuth resolution of Figure (b) has been greatly improved, but the runway resolution of Figure (c) is more obvious than that of Figure (b), the strong scattering targets are clearly visible, and the outlines of surface targets such as roofs are clearer, which proves the effectiveness of the algorithm. It is proved that while adopting multi-channel technology, the Doppler frequency information generated by the relative motion between the radar and the target in airborne forward imaging can be effectively used to more accurately reconstruct the imaging scene. Fig.11 for Fig.10The final reconstruction result projected into the geodetic coordinate system, Fig.11 (a) is spatial domain adaptive iterative super-resolution imaging, and Figure 11(b) is space-time adaptive iterative super-resolution imaging.

Claims

1. A space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar, characterized in that: The following steps are involved: (1) Constructing a space-time two-dimensional signal reception model: Using an airborne forward-looking array radar to obtain multi-channel echoes, the radar operates in a step-scan mode, transmitting several coherent pulses at the same azimuth in each coherent pulse interval, and sampling the echoes to obtain range-pulse-array three-dimensional echo data; (2) constructing the space-time snapshot signal corresponding to each range-azimuth unit based on the multi-channel data after pulse compression; Calculate the space-time steering vector corresponding to each azimuth angle to form a space-time steering vector matrix; (3) Processing the space-time snapshots in range-azimuth units, using the space-time steering vector matrix as the initial weight vector to obtain the initial value of the iteration; establishing a cost function based on the minimum mean square error criterion, calculating and simplifying the weight vector expression; then substituting the initial value of the iteration to start the iteration process, and continuously iterating and updating the current weight vector to make it close to the optimal value, to obtain the final weight vector; The space-time snapshot is weighted by using the final weight vector to calculate the super-resolution spectrum curve corresponding to the space-time snapshot of the range-azimuth unit; (4) According to the aircraft motion speed and antenna scanning parameters, the super-resolution spectrum curve of each range-azimuth unit is spliced ​​in azimuth, that is, the spectrum curve is accumulated in the range-azimuth domain, and finally a range-azimuth imaging image is formed; The implementation process of step (2) is as follows: According to the data obtained by the horizontal array antenna system of the airborne array radar, pulse compression processing is performed in the range direction, and the two-dimensional receiving data generated by each array element in each range-azimuth unit is selected to form the space-time snapshot signal of the range-azimuth unit; The space-time snapshot signals of each range-azimuth unit are stacked to convert the two-dimensional matrix into a one-dimensional vector. Then, for all azimuth angles in the imaging area, the corresponding space-time steering vectors are calculated and combined to form a space-time steering vector matrix. The specific implementation process is as follows: Assume that there are K targets in the front imaging area, and their azimuth angles relative to the radar are θ1, θ2, …, θ K ; In one cycle scan, the received data is a three-dimensional sampling of range-pulse-array. For a certain range-azimuth unit, the pulse-array sampling data constitutes a two-dimensional space-time snapshot: y=Ax+N Where x is the echo signal reflected from the target, N is additive white Gaussian noise, independent of other components; A is the NM×K-dimensional space-time guidance vector matrix, A = [a(θ1), a(θ2), …, a(θ K )],in a s (θ k ) is an M×1-dimensional spatial guidance vector composed of the phase difference of the target reaching the horizontal linear array; a t (θ k ) is the N×1-dimensional time steering vector composed of the Doppler frequency generated by the relative motion between the radar and the target, represents the Kronecker product; The implementation process of step (3) is as follows: Accurately reconstruct the reflection cross-sectional area x of the target, assuming that the estimate of x is: Among them, w is the weight coefficient matrix to be determined, y is the two-dimensional space-time snapshot of a certain range-azimuth unit, and the minimum mean square error criterion (MMSE) is used to establish the cost function: minJ{||xw H y|| 2 } Derivative solution of the above equation yields the minimum point: w=(E{yy H }) -1 E{yx H } Substituting the above formula into the simplified form, we get: w=(WHAT H +R N ) -1 AP in, is the covariance matrix of the noise, P is the autocorrelation matrix of the echo signal x reflected from the target, P = diag{p1,p2…,p K }, where the diagonal elements are: Calculate the first estimation result, at this time the weight matrix w1 = A, Based on this, the signal's autocorrelation matrix P1 is calculated, and then the weight matrix is ​​updated: w2=(AP1A H +R N ) -1 AP1 The second estimation result is obtained. And so on, using the estimate from the previous iteration Calculate the matrix P k , and then substitute into the formula to update the weight matrix w k+1 ; Finally, the new weight matrix is ​​applied to Get a new estimate, that is, the iterative expression is: w k+1 =(AP k A H +R N , -1 AP k The echo autocorrelation matrix R = AP k A H It does not rely on a large amount of snapshot data, but is obtained through iteration.

2. The space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar according to claim 1, characterized in that: The implementation process of step (1) is as follows: a horizontal array antenna along the azimuth direction is used to receive multi-channel echoes to obtain spatial sampling; the radar operates in a step scanning mode, and when the beam is directed to a fixed azimuth, a linear frequency modulation pulse (LFM) is transmitted every pulse repetition frequency (PRF) or pulse repetition interval (PRI) within a coherent pulse interval (CPI), and a number of coherent pulses are received to obtain time sampling; The beam scans one circle to obtain range-pulse-array three-dimensional echo data.

3. The space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar according to claim 1, characterized in that: The scanning parameters in step (4) include beam scanning range and beam scanning speed.

4. The space-time adaptive iterative super-resolution imaging method for airborne forward-looking array radar according to claim 1, characterized in that: The implementation process of step (4) is as follows: The azimuth super-resolution spectrum curve corresponding to the space-time snapshot of each range-azimuth unit obtained in step (3) is spliced ​​at the corresponding azimuth angle for the azimuth super-resolution spectrum curves of different azimuth units in the same range gate, that is, the accumulation of the spectrum curve in the range-azimuth domain is achieved, and the range-azimuth imaging result can be obtained by displaying the accumulated stored data.

Images