A three-dimensional super-resolution imaging method and device based on forward-looking SAR

Abstract

The application discloses a kind of three-dimensional super-resolution imaging method and device based on forward-looking SAR, in the process of airplane flight, every time interval adopts one-dimensional linear array to emit linear frequency modulation signal and receive echo, obtains multiple groups of echo data;In single echo reception, each array element sequentially receives in order, which is equivalent to single pulse radar produces movement along array direction, obtains similar echo data with SAR imaging in forward-looking area;Azimuth desloping processing is carried out to each group of echo data, and is transformed to frequency domain using iterative adaptive method, i.e. azimuth direction super-resolution can be realized;Subsequently, all echo data of the same azimuth are selected for processing, the echo is phase compensated by grid search, and combined with data extrapolation algorithm, super-resolution is realized in track-height direction.Compared with the traditional forward-looking SAR imaging algorithm, the application overcomes the limitation of array antenna size on resolution, uses super-resolution method, so that super-resolution is realized in three-dimensional direction.

Description

Technical Field

[0001] This invention belongs to the field of radar imaging technology, specifically relating to a three-dimensional super-resolution imaging method and device based on forward-looking SAR. Background Technology

[0002] As an important branch of radar technology, radar imaging technology plays a crucial role in battlefield reconnaissance, terrain mapping, autonomous driving, and blind landing of helicopters. With the iterative upgrading of science and technology and industry, the requirements for radar imaging technology have gone beyond two-dimensional imaging, leading to a series of studies on radar three-dimensional imaging.

[0003] There are many radar 3D imaging technologies, and most range super-resolution can be achieved through range pulse compression. However, obtaining high azimuth-elevation resolution is crucial for the design of radar 3D imaging methods. Most existing algorithms for improving azimuth-elevation are based on SAR (Radar Analog SAR) systems. However, when SAR images the forward-looking region, the Doppler gradient change of the target is very small, making high-resolution imaging impossible. How to achieve 3D imaging of the forward-looking region has become a challenge. To solve the technical difficulties of high-resolution forward-looking 3D imaging, many high-resolution forward-looking 3D imaging algorithms have been proposed, which can be categorized according to imaging system, such as forward-looking SAR, array radar, and real-aperture radar.

[0004] Forward-looking SAR was initially applied to forward-looking two-dimensional imaging. This involves placing a linear array perpendicular to the aircraft's flight direction, with individual elements transmitting and the remaining elements receiving data in turn at regular time intervals; or each element transmitting and receiving individually, with only one element active at a time. Through these element-based reception methods, the echo data received by the array can be equivalent to the echo data received by the SAR, thus allowing classic high-resolution SAR algorithms to be used for azimuth high-resolution imaging. With the aid of other algorithms, such as InSAR and monopulse imaging algorithms, forward-looking three-dimensional imaging can be further achieved. However, due to antenna aperture limitations, the imaging resolution cannot break the Rayleigh limit, resulting in limited improvement.

[0005] Compared to traditional solid-aperture radar, array radar offers higher spatial degrees of freedom, enabling better differentiation of adjacent targets. It can image a target in a single scan. Furthermore, the introduction of array super-resolution methods can further improve imaging resolution, achieving super-resolution imaging. To address the issue that one-dimensional arrays cannot simultaneously achieve super-resolution in both azimuth and elevation directions, a two-dimensional planar array can be introduced onto a moving platform, using two-dimensional super-resolution methods to achieve both simultaneously. However, traditional two-dimensional super-resolution methods (such as 2D MUSIC and 2D ESPRIT) require multiple snapshots, which is difficult to implement on a moving platform. Therefore, algorithms with single or few snapshot requirements, such as compressed sensing algorithms and 2D iterative adaptive algorithms, have been applied to airborne forward-looking 3D imaging with good results. However, these algorithms suffer from high computational costs.

[0006] Real aperture radar employs a multi-scan approach, which can be equivalent to an array receiver model through discretization and vectorization. Therefore, array super-resolution methods are also applicable to real aperture scanning radar. Compared to array radar, real aperture radar does not need to consider the element spacing registration problem; super-resolution imaging can be achieved through multiple scans and two-dimensional super-resolution methods. Based on the real aperture scanning radar platform, algorithms such as two-dimensional iterative adaptive methods and two-dimensional OMP have been applied to improve resolution. However, similar to its application in array radar, the high computational cost of two-dimensional super-resolution methods remains an insurmountable obstacle.

[0007] Considering the above algorithms, the forward-looking SAR imaging algorithm does not employ super-resolution methods during imaging processing, and its imaging resolution is limited by the antenna aperture. Array radar imaging algorithms cannot achieve three-dimensional imaging using only a one-dimensional array, while using a two-dimensional array and a two-dimensional super-resolution algorithm introduces a sharp increase in computational load. Therefore, it is necessary to study an imaging algorithm that can achieve three-dimensional imaging using only a one-dimensional array while applying super-resolution methods to improve resolution. Summary of the Invention

[0008] Purpose of the invention: The technical problem to be solved by the present invention is to address the deficiencies involved in the background technology by providing a three-dimensional super-resolution imaging method and device based on forward-looking SAR. By adopting preprocessing such as azimuth deskewing, interpolation, and phase compensation, and two one-dimensional super-resolution methods, the target can achieve three-dimensional super-resolution while effectively reducing computational complexity.

[0009] Technical solution: This invention provides a three-dimensional super-resolution imaging method based on forward-looking SAR, specifically including the following steps:

[0010] (1) During the flight of the aircraft, the airborne radar collects a set of echoes using a one-dimensional linear array at fixed intervals. During the reception of each set of echoes, each array element receives the data in sequence, resulting in multiple sets of SAR-like echo data.

[0011] (2) For each group of SAR echo data, after pulse compression and range migration correction are completed, azimuth deskewing is performed by multiplying the echo signal with the azimuth reference signal to complete the azimuth deskewing of the echo signal.

[0012] (3) Achieve super-resolution in the azimuth direction by using an iterative adaptive method: the estimated value of the echo signal in the frequency domain is solved by minimizing the error function, and the estimated value is iterated multiple times;

[0013] (4) Interpolate the echo signal to obtain multiple forward-looking SAR images; then, perform phase compensation on the same azimuth data of multiple forward-looking SAR images by grid search so that the phase generated by the forward motion of the aircraft can be canceled.

[0014] (5) Combine the data extrapolation algorithm to achieve super-resolution in the orbital-height direction, thereby achieving forward-looking three-dimensional super-resolution imaging.

[0015] Furthermore, the specific steps of step (1) are as follows:

[0016] In each echo reception, the array adopts a multiple transmit and multiple receive mode, and the array elements switch at pulse repetition intervals. Only one array element works in each PRI, which is equivalent to a single pulse radar moving along the array direction, thus obtaining a set of SAR-like echo data; by acquiring multiple data during flight, multiple types of SAR echo data are obtained.

[0017] Furthermore, the implementation process of step (2) is as follows:

[0018] Perform Fourier transform on the range signal of each group of SAR-like echo data to complete range pulse compression and range migration correction; for the azimuth echo signal under a single range gate: S a (t a ) = A a exp[jπk a (t a -x p / v s ) 2 ]; where S a (t a ) represents the azimuth echo signal, A a t represents the magnitude term. a Represented as slow time, x p v represents the azimuth coordinate of a point target. s k is the antenna switching speed, which is equal to the pulse repetition frequency. a Represents azimuth frequency modulation; azimuth reference signal S ref (t a ) is represented as: S ref(t a )=exp[-jπk a t a 2 ]; Let S a With S ref Multiplying and ignoring the phase constant yields the azimuth-de-slanted signal S. dechirp (t a ) = A a exp[-j2πk a x p / v s At this time, the signal is represented in the azimuth frequency domain as a peak value located at f. a =k a x p / v s The sinc signal at that location.

[0019] Furthermore, the process of minimizing the error function to solve for the estimated value of the echo signal in the frequency domain in step (3) is as follows:

[0020] According to the weighted least squares criterion, the weighted sum of squared errors is taken as the error function E:

[0021] E = [S - Aσ] H W[S-Aσ]

[0022] S = [S dechirp (t1),S dechirp (t2), ..., S dechirp (t N )]

[0023] A=[exp(j2πf1t),exp(j2πf2t),…,exp(j2πf L t)]

[0024] W={E[σ*σ H ]} -1

[0025] Where S is the set of matrices for sampling the target echo azimuth signal by different array elements, N is the number of array elements; A is an overcomplete dictionary containing different inverse Fourier transform bases, f1 to f L It is the frequency component obtained by dividing the pulse repetition frequency into L equal parts, t=[t1,t2,…,t N ] T Let σ be the sampling time of N array elements; σ = [σ1, σ2, ..., σ] L [] represents the estimated target signal values ​​at different frequency components; W is the weight matrix, and E[] represents the mean operation;

[0026] Minimize the error function to obtain the estimated value for each frequency component:

[0027]

[0028] Where i = 1, 2, ..., L.

[0029] Furthermore, the process of iterating the estimated value multiple times in step (3) is as follows:

[0030] First, update the signal power matrix P:

[0031] P = [p1, p2, ..., p L ]

[0032] Where, p i The estimated value from the previous estimation process The power P is estimated in the first iteration using the least squares method;

[0033] Subsequently, W and σ are updated sequentially, with W rewritten as:

[0034] W = [Rp] i exp(j2πf i t)*exp(j2πf i t) H ] -1

[0035] R = APA H

[0036] R is the covariance matrix of the signal; after updating P, R, and W sequentially, according to... Solving the equation updates σ; after multiple iterations, σ converges, thus completing the conversion of the azimuth signal from the time domain to the frequency domain, achieving azimuth super-resolution. The echo data after azimuth super-resolution is represented as:

[0037]

[0038] in, The distance between the array element and the target in the (yz) plane is obtained using the principle of stationary phase.

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

[0040] After interpolation is completed, the data at azimuth x = x0 of all forward-looking SAR image data are selected and represented as follows:

[0041]

[0042] Among them, R yz Let x = x0 be the distance between the array element in the yz plane and the target, denoted as: tir ,t y These represent the interpolated fast time and the aircraft's forward flight time, respectively; A is the amplitude term, c is the speed of light, and y... p Let z be the coordinates of the point target along the orbit. p Here, h represents the altitude coordinate of the point target, and f represents the altitude of the airborne radar. c Let be the carrier frequency; let the phase compensation function be:

[0043]

[0044] in y grid ,z grid f represents the along-axis coordinates and height coordinates of the current search grid, respectively. ir For the distance frequency domain; for S yz (t ir ,t y Compensating in the range frequency domain and transforming back to the range time domain, we get:

[0045]

[0046] When there is exactly R yz =R c At that time, the peak value of the sinc function is obtained at the center of the scene, and the peak value is obtained for different t values. y The data below are accumulated to obtain the target intensity estimate S under the current grid. res Let M be the aircraft at different t y The number of array samplings is then... A j Let be the target intensity during the j-th acquisition.

[0047] Furthermore, the implementation process of step (5) is as follows:

[0048] By using a data extrapolation algorithm, the number of array samplings M is increased to extend the synthetic aperture, thereby achieving in-orbit-height two-dimensional super-resolution.

[0049] Based on the same inventive concept, the present invention also proposes an apparatus comprising a memory and a processor, wherein:

[0050] Memory is used to store computer programs that can run on a processor;

[0051] A processor is configured to execute the steps of the forward-looking SAR-based three-dimensional super-resolution imaging method as described above when running the computer program.

[0052] Based on the same inventive concept, the present invention also proposes a storage medium storing a computer program, which, when executed by at least one processor, implements the steps of the three-dimensional super-resolution imaging method based on forward-looking SAR as described above.

[0053] Beneficial Effects: Compared with the prior art, the beneficial effects of this invention are as follows: This invention solves the problem of forward-looking SAR azimuth resolution being limited by antenna aperture by using an iterative adaptive method, breaking the Rayleigh limit and achieving super-resolution in the azimuth direction; This invention obtains the changes in target along the trajectory and altitude based on sampling of the aircraft at different flight times, and achieves super-resolution in both the trajectory and altitude directions through phase compensation and data extrapolation; In addition, the method proposed in this invention only uses a one-dimensional array for data acquisition, and can achieve forward-looking three-dimensional super-resolution imaging using a one-dimensional super-resolution algorithm, overcoming the shortcomings of traditional array radars that require a two-dimensional planar array for three-dimensional imaging, and effectively avoiding the huge computational load of two-dimensional super-resolution algorithms. Attached Figure Description

[0054] Figure 1 This is a schematic diagram of the working model of a forward-looking SAR array;

[0055] Figure 2 This is a schematic diagram of array element switching;

[0056] Figure 3 This is a diagram analyzing the positions of the array elements.

[0057] Figure 4 For scene simulation targets;

[0058] Figure 5 This is a three-dimensional image of a forward-looking SAR without using azimuth super-resolution methods;

[0059] Figure 6 A three-dimensional image of forward-looking SAR using the azimuth super-resolution method;

[0060] Figure 7 This is a forward-looking SAR 3D image without data extrapolation.

[0061] Figure 8 This is a forward-looking SAR 3D image obtained using data extrapolation. Detailed Implementation

[0062] The present invention will now be described in further detail with reference to the accompanying drawings.

[0063] This invention proposes a forward-looking 3D super-resolution method based on forward-looking SAR, comprising the following steps:

[0064] Step 1: During the flight of the aircraft, the airborne radar collects a set of echoes at fixed intervals using a one-dimensional linear array. During the reception of each set of echoes, each array element receives the data in sequence, resulting in multiple sets of SAR-like echo data.

[0065] like Figure 1 As shown, the aircraft travels at a constant speed v y The aircraft flies along the y-axis and a linear array is placed along the x-axis, with each element spaced d apart. Here, x represents the azimuth direction, y the orbital direction, z the altitude direction, and θ1 and θ2 the pitch angles relative to the target at different flight times. During forward flight, the aircraft samples the target area at fixed time intervals, obtaining multiple sets of echo data. Within each echo set, an element transmits and receives data individually; only one element operates within each PRI (Primary Airway Processing) unit. Element switching is as follows: Figure 2 As shown. Since the array elements work alternately, it can be equivalent to a single array element operating at a velocity v. s =d / PRI moves along the x-axis, forming a certain synthetic aperture, and the resulting echo data is similar to SAR echo data. For example... Figure 3 As shown, considering that the aircraft moves forward simultaneously during sampling, the array elements are not positioned on the same vertical axis. When v s Much greater than v y At this time, the change in the position of the array elements caused by the aircraft's motion is negligible, so we should choose the largest possible PRF. PRF is the pulse repetition frequency, which satisfies PRF = 1 / PRI.

[0066] Step 2: For each group of SAR-like echo data, after pulse compression and range migration correction are completed, azimuth deskewing is performed by multiplying the echo signal with the azimuth reference signal to complete the azimuth deskewing of the echo signal.

[0067] For a target echo at a certain point, it can be represented as:

[0068]

[0069] Among them, t r ,t a ,t y These represent the fast time, slow time, and the array sampling time during a specific forward flight of the aircraft, respectively. (x p ,y p ,z p Let k be the three-dimensional spatial coordinates of the point target. r Here, h is the range modulation frequency, h is the radar altitude, and λ is the wavelength. After pulse compression and range migration correction are performed, the azimuth signal of a single range gate in a single set of echo data can be expressed as:

[0070] S a (t a ) = Aa exp[jπk a (t a -x p / v s ) 2 ]

[0071] Among them, S a (t a ) represents the azimuth echo signal, A a k represents the magnitude term. a Represents azimuth frequency modulation; x p v represents the azimuth coordinate of a point target. s This represents the antenna switching speed, which is equal to the pulse repetition frequency.

[0072] Because the aperture of the actual array antenna is much smaller than the synthetic aperture of traditional SAR, it is impossible to directly distinguish targets in the time domain. Through deskewing processing, it is multiplied by the reference signal S. ref (t a After FFT, the target can be distinguished in the frequency domain. The reference signal can be represented as:

[0073] S ref (t a )=exp[-jπk a t a 2 ]

[0074] Let S a (t a ) and S ref (t a Multiplying these signals and ignoring the phase constant yields the azimuth-de-slanted signal.

[0075] S dechirp (t a ) = A a exp[-j2πk a x p / v s ]

[0076] At this point, the signal changes to a frequency with a peak value located at f. a =k a x p / v s The sinc signal at that location.

[0077] Step 3: Use an iterative adaptive method to replace the azimuth Fourier transform. Solve the estimated value of the echo signal in the frequency domain by minimizing the error function. Repeated iterations make the estimated value more accurate, thus achieving azimuth super-resolution.

[0078] Due to S dechirp (t aIn the time domain, the function is a rectangular window function. Directly transforming it to the frequency domain using FFT results in a sinc function, which exhibits significant sidelobes. Furthermore, the main lobe width of the sinc function is related to the antenna aperture, preventing the azimuth resolution from exceeding the Rayleigh limit. To suppress sidelobes while achieving super-resolution, this invention employs an iterative adaptive method instead of FFT, enabling azimuth super-resolution.

[0079] For the iterative adaptive method, an error function is constructed using the weighted least squares criterion. Minimizing the error function yields a preliminary estimate of the target, and repeated iterations refine the estimate. The specific steps are as follows:

[0080] Step 3.1: Assume that the target's time-domain signal S can be obtained from the target's frequency-domain signal using IFFT. dechirp (t a In matrix form, we have:

[0081] S=Aσ+n

[0082] Where S is the set of matrices for sampling the target echo azimuth signal by different array elements, which can be written as:

[0083] S = [S dechirp (t1),S dechirp (t2), ..., S dechirp (t N )]

[0084] N is the number of array elements; A is an overcomplete dictionary containing different inverse Fourier transform bases, which can be written as:

[0085] A=[exp(j2πf1t),exp(j2πf2t),…,exp(j2πf L t)]

[0086] Where f1 to f L It is the frequency component obtained by dividing the PRF into L equal parts, t=[t1,t2,…,t N ] T Let σ be the sampling time of N array elements; σ = [σ1, σ2, ..., σ] L [Equation] represents the estimated target signal values ​​at different frequency components. According to the weighted least squares criterion, the error function E can be written as:

[0087] E = [S - Aσ] H W[S-Aσ]

[0088] Where W is the weight matrix, which can be written as:

[0089] W={E[σ*σ H ]} -1

[0090] E[] represents the mean operation. Minimizing the error function yields the estimated values ​​for the frequency components.

[0091]

[0092] Where i = 1, 2, BD, ..., L.

[0093] Step 3.2: After obtaining the estimated value, multiple iterations are required. During the iteration process, the power P of the estimated value, the covariance matrix R of the signal, the weight matrix W, and the estimated signal value σ are updated sequentially. Where P = [p1, p2, ..., p...] L ], where p i The estimated value from the previous estimation process The power, where i = 1, 2, ..., L. Furthermore, for the initial estimate, P can be obtained using the least squares method. R can be obtained using R = APA. H To update, W can be accessed via W = [Rp] i exp(j2πf i t)*exp(j2πf i t) H ]- 1 The σ value is then updated using the derivation in step 3.1. After multiple iterations, the estimate of σ becomes more accurate, eventually converging and achieving azimuth super-resolution. The echo data after azimuth super-resolution can be represented as:

[0094]

[0095] in, Given the distance between an array element and the target in a certain yz plane, the principle of stationary phase can also be used to obtain...

[0096] Step 4: After interpolating the echo signal, multiple forward-looking SAR images are obtained. Then, phase compensation is performed on the same azimuth data of the multiple forward-looking SAR images through grid search, so that the phase generated by the forward motion of the aircraft can be canceled.

[0097] After processing each set of echo data in step 3, multiple sets of super-resolution data in the range-time domain and azimuth frequency domain are obtained. For two coordinate systems with different slant ranges but the same azimuth coordinate, different Doppler frequencies will be generated. Therefore, data with the same azimuth coordinate cannot be processed uniformly; the data must be transformed back into the range-azimuth two-dimensional time domain. This can be achieved through two-dimensional interpolation. The interpolated echo signal is represented as follows:

[0098]

[0099] Among them, t irThis represents the time in the range direction within the yz plane. Data with the azimuth coordinate x = x0 is selected. The term is a constant; the echo signal is represented as:

[0100]

[0101] Analysis S yz (t ir ,t y From the expression, we can see that for the same target at different aircraft flight times t... y When observing below, the phase term and latency This will change with the relative distance between the array element and the target. To determine the target's position along the orbit and its altitude, compensation should be applied to the latter two parameters to ensure that different values ​​at different t values ​​are within acceptable limits. y The data below is uniformly corrected. These two terms can be canceled out by transforming to the range frequency domain and multiplying by a suitable phase compensation function. S yz (t ir ,t y Transforming to the range-frequency domain yields:

[0102]

[0103] Among them, f ir This represents the distance frequency domain. The phase to be compensated can be written as:

[0104]

[0105] However, because the exact location of the observed target is unknown, R yz It is also impossible to know. Furthermore, different observed targets will have different R values. yz Therefore, phase compensation cannot be directly performed. In this invention, the yz plane of the observed area is divided into multiple grids, each with different along-orbit coordinates and height coordinates, allowing the generation of different phase compensation functions. grid ,z grid For a grid of ), its phase compensation function can be written as:

[0106]

[0107] in, Connect the phase compensation function with S yz (f ir ,t y Multiplying these together and converting back to the time domain, we get:

[0108]

[0109] When (ygrid ,z grid When R is exactly the coordinate of the observed target, yz =R c The phase term and the time delay term are canceled out. Superimposed with different t... y The signal at time t and let t ir =0 has:

[0110]

[0111] Among them, S res For target amplitude estimation in the azimuth coordinates x = x0, A j Let M be the target intensity collected in the j-th sampling, and M be the number of sampling points when the aircraft is flying forward.

[0112] Step 5: Combine the data extrapolation algorithm to achieve super-resolution in the orbital-height direction, thereby completing forward-looking 3D super-resolution imaging.

[0113] By S res As can be seen from the expression, under fixed flight speed and sampling interval, the more sampling points the aircraft flies, the larger the synthesized aperture, and the more accurate the estimation of target amplitude. However, in some special situations (such as battlefield), the aircraft cannot fly a sufficiently large distance, resulting in an excessively small synthesized aperture, making it impossible to effectively distinguish targets in the altitude and along-track directions. Considering this situation, this invention uses data extrapolation to increase the number of sampling points, effectively improving the resolution in the along-track and altitude directions. By sequentially performing step 4 and data extrapolation processing on the data in each azimuth direction, forward-looking three-dimensional super-resolution of the observed target can be achieved.

[0114] Based on the same inventive concept, the present invention also proposes an apparatus comprising a memory and a processor, wherein the memory is used to store a computer program capable of running on the processor; and the processor is used to execute the steps of the three-dimensional super-resolution imaging method based on forward-looking SAR as described above when running the computer program.

[0115] Based on the same inventive concept, the present invention also proposes a storage medium storing a computer program, which, when executed by at least one processor, implements the steps of the three-dimensional super-resolution imaging method based on forward-looking SAR as described above.

[0116] The present invention is compared with a forward-looking 3D imaging algorithm that does not apply super-resolution methods to verify the improvement of resolution by super-resolution methods in the azimuth and height directions, thereby verifying the effectiveness of the method.

[0117] First, we verify the improvement in azimuth resolution achieved by the super-resolution method. We employ methods such as... Figure 4The simulated scene shown has a radar distance of 5000m from the center of the scene, a radar altitude of 1000m, an array antenna length of 3m, 60 array elements, and a pulse repetition frequency of 9000Hz. Other parameters are similar to those of traditional radar forward-looking imaging. In the imaging processing, the forward-looking 3D imaging algorithm without super-resolution is directly FFT-transformed to the azimuth frequency domain after azimuth de-skewing processing. The imaging results of the comparison method and the method proposed in this invention are shown below. Figure 5 , Figure 6 As shown. Analysis Figure 5 It can be seen that the imaging results reconstruct the target scene relatively accurately, but there are obvious side lobes and aliasing of point targets in different orientations, resulting in a deterioration in the imaging results. Figure 5 Compare and analyze the imaging results. Figure 6 It can be seen that the imaging results further improve the resolution in the azimuth direction, distinguishing adjacent point targets in different azimuth directions, and there is no sidelobe phenomenon, thus achieving super-resolution of the target in the azimuth direction.

[0118] Next, we verify the improvement of the super-resolution method on the height resolution. In this invention, the principle of achieving super-resolution in the height and along-orbit directions is the same, so we only need to verify the improvement of the super-resolution method on one dimension. We select three point targets with the same along-orbit and azimuth coordinates, and whose height coordinates differ by 3m. The radar position information and parameter information are the same as those in the above scenario simulation. Figure 7 and Figure 8 These are the imaging results without data extrapolation and after data extrapolation, respectively. Figure 7 It can be seen that without data extrapolation, adjacent targets overlap, making it impossible to distinguish targets 3 meters apart. Furthermore, due to the low resolution, the algorithm cannot accurately estimate the target's orbital information; as can be seen in the figure, the estimated orbital coordinates of the target show a shift. Figure 8 It can be seen that after data extrapolation, the synthesized virtual aperture is further increased, and the resolution is further improved, enabling the differentiation of three adjacent 3m targets. Meanwhile, in comparison... Figure 7 It can be seen Figure 8 The trajectory information of the three point targets was also accurately reconstructed. The above simulations fully demonstrate the effectiveness of the method proposed in this invention.

Claims

1. A three-dimensional super-resolution imaging method based on forward-looking SAR, characterized in that, Includes the following steps: (1) During the flight of the aircraft, the airborne radar collects a set of echoes using a one-dimensional linear array at fixed intervals. During the reception of each set of echoes, each array element receives the data in sequence, resulting in multiple sets of SAR-like echo data. (2) For each group of SAR echo data, after pulse compression and range migration correction are completed, azimuth deskewing is performed by multiplying the echo signal with the azimuth reference signal to complete the azimuth deskewing of the echo signal. (3) Achieve super-resolution in the azimuth direction by using an iterative adaptive method: the estimated value of the echo signal in the frequency domain is solved by minimizing the error function, and the estimated value is iterated multiple times; (4) Interpolate the echo signal to obtain multiple forward-looking SAR images; then, perform phase compensation on the same azimuth data of multiple forward-looking SAR images by grid search so that the phase generated by the forward motion of the aircraft can be canceled. (5) Combine the data extrapolation algorithm to achieve super-resolution in the orbital-height direction, thereby achieving forward-looking three-dimensional super-resolution imaging.

2. The three-dimensional super-resolution imaging method based on forward-looking SAR according to claim 1, characterized in that, The specific steps of step (1) are as follows: In each echo reception, the array adopts a multiple transmit and multiple receive mode, and the array elements switch at pulse repetition intervals. Only one array element works in each PRI, which is equivalent to a single pulse radar moving along the array direction, thus obtaining a set of SAR-like echo data; by acquiring multiple data during flight, multiple types of SAR echo data are obtained.

3. The three-dimensional super-resolution imaging method based on forward-looking SAR according to claim 1, characterized in that, The implementation process of step (2) is as follows: Perform Fourier transform on the range signal of each group of SAR-like echo data to complete range pulse compression and range migration correction; for the azimuth echo signal under a single range gate: S a (t a ) = A a exp[jπk a (t a -x p / v s ) 2 ]; where S a (t a ) represents the azimuth echo signal, A a t represents the magnitude term. a Represented as slow time, x p v represents the azimuth coordinate of a point target. s k is the antenna switching speed, which is equal to the pulse repetition frequency. a Represents azimuth frequency modulation; azimuth reference signal S ref (t a ) is represented as: S ref (t a )=exp[-jπk a t a 2 ]; Let S a With S ref Multiplying and ignoring the phase constant yields the azimuth-de-slanted signal S. dechirp (t a ) = A a exp[-j2πk a x p / v s At this time, the signal is represented in the azimuth frequency domain as a peak value located at f. a =k a x p / v s The sinc signal at that location.

4. The three-dimensional super-resolution imaging method based on forward-looking SAR according to claim 1, characterized in that, The process of minimizing the error function to solve for the estimated value of the echo signal in the frequency domain in step (3) is as follows: According to the weighted least squares criterion, the weighted sum of squared errors is taken as the error function E: E=[S-Aσ] H W[S-Aσ] S=[S dechirp (t1),S dechirp (t2),…,S dechirp (t N )] A=[exp(j2πf1t),exp(j2πf2t),…,exp(j2πf L t)] W={E[σ*σ H ]} -1 Where S is the set of matrices for sampling the target echo azimuth signal by different array elements, N is the number of array elements; A is an overcomplete dictionary containing different inverse Fourier transform bases, f1 to f L It is the frequency component obtained by dividing the pulse repetition frequency into L equal parts, t=[t1,t2,…,t N ] T Let σ be the sampling time of N array elements; σ = [σ1, σ2, ..., σ] L [] represents the estimated target signal values ​​at different frequency components; W is the weight matrix, and E[] represents the mean operation; Minimize the error function to obtain the estimated value for each frequency component: Where i = 1, 2, ..., L.

5. A three-dimensional super-resolution imaging method based on forward-looking SAR according to claim 1, characterized in that, The process of iterating the estimated value multiple times in step (3) is as follows: First, update the signal power matrix P: P=[p1,p2,…,p L ] Where, p i The estimated value from the previous estimation process The power P is estimated in the first iteration using the least squares method; Subsequently, W and σ are updated sequentially, with W rewritten as: W=[R-p i exp(j2πf i t)*exp(j2πf i t) H ] -1 R=WHAT H R is the covariance matrix of the signal; after updating P, R, and W sequentially, according to... Solving the equation updates σ; after multiple iterations, σ converges, thus completing the conversion of the azimuth signal from the time domain to the frequency domain, achieving azimuth super-resolution; the echo data after azimuth super-resolution is represented as: in, The distance between the array element and the target in the (yz) plane is obtained using the principle of stationary phase.

6. The three-dimensional super-resolution imaging method based on forward-looking SAR according to claim 1, characterized in that, The implementation process of step (4) is as follows: After interpolation is completed, the data at azimuth x = x0 of all forward-looking SAR image data are selected and represented as follows: Among them, R yz Let x = x0 be the distance between the array element in the yz plane and the target, denoted as: t ir ,t y These represent the interpolated fast time and the aircraft's forward flight time, respectively; A is the amplitude term, c is the speed of light, and y... p Let z be the coordinates of the point target along the orbit. p Here, h represents the altitude coordinate of the point target, and f represents the altitude of the airborne radar. c Let be the carrier frequency; let the phase compensation function be: in y grid ,z grid f represents the along-axis coordinates and height coordinates of the current search grid, respectively. ir For the distance frequency domain; for S yz (t ir ,t y Compensating in the range frequency domain and transforming back to the range time domain, we get: When there is exactly R yz =R c At that time, the peak value of the sinc function is obtained at the center of the scene, and the peak value is obtained for different t values. y The data below are accumulated to obtain the target intensity estimate S under the current grid. res Let M be the aircraft at different t y The number of array samplings is then... A j Let be the target intensity during the j-th acquisition.

7. A three-dimensional super-resolution imaging method based on forward-looking SAR according to claim 1, characterized in that, The implementation process of step (5) is as follows: By using a data extrapolation algorithm, the number of array samplings M is increased to extend the synthetic aperture, thereby achieving in-orbit-height two-dimensional super-resolution.

8. A device, characterized in that, Includes memory and processor, wherein: Memory is used to store computer programs that can run on a processor; A processor, configured to, while running the computer program, perform the steps of the three-dimensional super-resolution imaging method based on forward-looking SAR as described in any one of claims 1-7.

9. A storage medium, characterized in that, The storage medium stores a computer program that, when executed by at least one processor, implements the steps of the three-dimensional super-resolution imaging method based on forward-looking SAR as described in any one of claims 1-7.

Images