Dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind SAMP

Abstract

The application discloses a kind of dynamic platform forward-looking super-resolution imaging methods based on divide-and-conquer blind SAMP, comprising: the baseband echo signal is filtered in distance direction;According to filter signal, using acceleration phase compensation function and envelope deslope function calculates forward-looking scene scattering coefficient observation value;Construct overcomplete dictionary matrix;According to forward-looking scene scattering coefficient observation value and overcomplete dictionary matrix, using improved SAMP algorithm carries out forward-looking super-resolution imaging;The algorithm is based on divide-and-conquer method and carries out sparsity estimation, and reconstructs forward-looking scene signal based on sparsity;Wherein, two functions and overcomplete dictionary matrix are based on single-channel forward-looking scanning imaging geometry model of dynamic platform orientation and are constructed;The model introduces the three-dimensional acceleration of dynamic platform to accurately characterize the motion trajectory of dynamic platform.The application solves the problem of forward-looking super-resolution imaging blur under dynamic platform, effectively realizes the forward-looking super-resolution imaging of high-speed running dynamic platform.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing, specifically relating to a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP (sparseness adaptive matching pursuit). Background Technology

[0002] Forward-looking (FLAG) mode is an important imaging mode in the field of remote sensing. It can detect and image areas directly in front of the flight path at all times and in all weather conditions, and at long distances, showing broad application prospects. Furthermore, FLAG imaging results can be used to detect and identify high-value targets within regions of interest directly in front. Compared to single-pulse target detection methods, FLAG imaging significantly improves detection and identification accuracy, thus making FLAG radar imaging a research focus in recent years.

[0003] In traditional Synthetic Aperture Radar (SAR) imaging systems, to acquire high-resolution two-dimensional microwave remote sensing images, a sufficiently long synthetic aperture or a sufficiently large coherent accumulation angle needs to be accumulated along the range lateral direction. This allows for the formation of an equivalent virtual array in space, which then interferometrically forms an extremely narrow beam, achieving high-resolution range lateral detection of the illuminated area. In the field of microwave remote sensing imaging, high-resolution range capability can be achieved through pulse compression technology, the performance of which is determined by the bandwidth of the transmitted signal.

[0004] However, in forward-looking imaging geometry, the radar antenna beam points directly forward of the platform's flight direction, so the synthetic aperture length in the lateral range cannot accumulate with the platform's movement. In this case, the lateral range resolution within the forward-looking illumination area is provided solely by the azimuth real aperture length of the radar antenna. Furthermore, when the radar beam points directly forward of the flight direction, the left and right sides of the forward-looking scene have the same spatial cone angle, causing left-right Doppler blurring during forward-looking imaging. Considering these two major challenges—range lateral resolution limited by the azimuth real aperture length and left-right Doppler blurring—traditional single-channel SAR imaging systems and signal processing methods cannot be directly applied to forward-looking imaging problems.

[0005] Super-resolution sparse signal reconstruction theory offers an improved approach to resolving lateral distances. Single-channel forward-looking imaging, combining beam scanning with super-resolution techniques, is a feasible solution for forward-looking imaging on high dynamic range platforms. For example, existing technologies include forward-looking super-resolution imaging methods based on OMP (Orthogonal Matched Pursuit) or SAMP. However, OMP-based forward-looking super-resolution imaging methods require sparsity as prior knowledge, thus making them unsuitable when scene sparsity is unknown.

[0006] The SAMP-based forward-looking super-resolution imaging method does not require prior precise knowledge of scene sparsity. In this method, the range-azimuth echo model is first established, and the range direction is processed by range compression and other methods. A Doppler convolution model is established, and the reconstruction problem of the forward-looking scene with known observations is represented as a linear regression model, that is, the overcomplete dictionary matrix A is constructed. Then, the SAMP method is used as the sparse reconstruction kernel to restore the forward-looking scene.

[0007] However, in high-speed dynamic platforms, existing SAMP-based forward super-resolution imaging methods suffer from the problem of relatively blurry reconstructed scenes. Summary of the Invention

[0008] To address the aforementioned problems in the existing technology, this invention provides a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP.

[0009] The technical problem to be solved by this invention is achieved through the following technical solution:

[0010] A dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP includes:

[0011] Acquire the baseband echo signal and perform range-direction matched filtering on the baseband echo signal to obtain the range-direction matched filtered signal;

[0012] Based on the range-direction matched filter signal, the observed values ​​of the forward-looking scene scattering coefficients are calculated using a preset acceleration phase compensation function and an envelope deskewing function; wherein, the acceleration phase compensation function and the envelope deskewing function are both pre-constructed based on a single-channel forward-looking scanning imaging geometric model for dynamic platforms;

[0013] An overcomplete dictionary matrix was constructed based on radar motion parameters and the single-channel forward-looking scanning imaging geometric model.

[0014] Based on the observed scattering coefficients of the forward-looking scene and the overcomplete dictionary matrix, a forward-looking super-resolution imaging is performed using an improved SAMP algorithm. The improved SAMP algorithm first estimates the sparsity based on a divide-and-conquer method, and then reconstructs the forward-looking scene signal based on the estimated sparsity.

[0015] The single-channel forward-looking scanning imaging geometric model is based on a Cartesian coordinate system with the horizontal equivalent velocity direction of the dynamic platform at time zero as the y-axis, the celestial direction as the z-axis, and the lateral direction as the x-axis. The projection point of the dynamic platform on the xOy plane at time zero is taken as the origin of the coordinate system. The three-dimensional acceleration of the dynamic platform is introduced to accurately characterize the motion trajectory of the dynamic platform.

[0016] Optionally, based on the observed scattering coefficients of the forward-looking scene and the overcomplete dictionary matrix, forward-looking super-resolution imaging is performed using an improved SAMP algorithm, including:

[0017] A. Initialize the signal residual to the observed values ​​of the scattering coefficients of the forward-looking scene, and initialize the atom set to an empty set; wherein, the signal residual is used to characterize the difference between the actual forward-looking scene signal and the estimated forward-looking scene signal;

[0018] B. Update the atom set according to the current signal residual and the overcomplete dictionary matrix, estimate the sparsity using the sparsity over-judgment criterion, and find the support set for recovering the forward-looking scene signal from the updated atom set based on the estimated sparsity.

[0019] C. Based on the current support set and the observed values ​​of the forward-looking scene scattering coefficients, the estimated forward-looking scene signal is solved using the least squares method. The signal residual is calculated and updated based on the estimated forward-looking scene signal, and the residual correlation is calculated. The residual correlation is used to characterize the correlation between the current residual and the support set.

[0020] D. Determine whether the algorithm has converged based on the convergence of the signal residuals and residual correlation. If it has converged, use the currently estimated forward-looking scene signal as the reconstructed forward-looking scene signal. If it has not converged, return to step B as the current support set as the atom set.

[0021] Optionally, step B includes:

[0022] B1. Update the atom set using the following formula based on the current signal residual and the overcomplete dictionary matrix:

[0023] λ t ={i:(|Ar t-1 |) i ≥thr×max(|Ar t-1 |) j};

[0024]

[0025] Where thr is a preset first threshold; A represents the overcomplete dictionary matrix, r t-1 Represents the current signal residual; j is the column index of A; λ t This is the first set of subscript indices, where the subscript indices are selected from column index j of A, and the selected subscripts are numbered with i; Indicates the use of λ t The column selected from A, A t Let A represent the updated set of atoms. t-1 This represents the set of atoms before the update;

[0026] B2. Calculation and from Find the largest I atom indices in the set and form the second subscript index set λ. t ': Where s represents the observed value of the scattering coefficient of the forward-looking scene, For A t The transpose of the matrix;

[0027] B3. Judgment Is it true? If true, let I = I + 1 and return to step B2, until... If the current I is not found, use the estimated sparsity as the current value and proceed to step B4; if not, set I = I-1 and return to step B2, until... At that time, the current I is used as the estimated sparsity, and step B4 is executed; where, Indicated by λ t 'from The selected columns; ||·||2 represents the 2-norm operator; (2K,δ) 2K ) represents the parameter that the overcomplete dictionary matrix satisfies the finite isometric RIP property;

[0028] B4. Using the second index set λ from which the sparsity is estimated. t From the current set of atoms A t The support set selected is used to recover the forward-looking scene signal.

[0029] Optionally, the convergence of the algorithm is determined based on the convergence of the signal residuals and residual correlation, including:

[0030] Judgment | ε t-1 -ε t |<δ or||r t Does ||2 < opt × ||s||2 hold true?

[0031] If any one of these conditions is true, the algorithm is considered to have converged.

[0032] If none of the conditions are met, continue to evaluate ||r t ||2≥||r t-1 Is ||2 true?

[0033] If ||r t ||2≥||r t-1 If ||2 holds true, the algorithm is considered to have converged;

[0034] If ||r t ||2≥||r t-1 If 2 is not true, the algorithm is considered not to have converged.

[0035] Where, ε t ε represents the residual correlation calculated in this iteration.t-1 This represents the residual correlation calculated in the previous iteration; r t r represents the signal residual calculated in this iteration. t-1 represents the signal residual calculated in the previous iteration, s represents the observed value of the scattering coefficient of the forward-looking scene; δ is the preset second threshold; opt is the preset third threshold; ||·||2 represents the 2-norm operator.

[0036] Optionally, the acceleration phase compensation function is:

[0037]

[0038] in, f represents the range-direction frequency domain variable. s f represents the sampling frequency. c Let represent the radar carrier frequency, j represent the imaginary unit, π represent pi, and c represent the speed of light; exp{·} represents the complex exponential function used to characterize the phase information of the echo signal; t represents the azimuth time slow; k2 and k3 represent the second and third order Maclaurin expansion coefficients of the true slant range history R(t) between the target and the dynamic platform, which varies with azimuth time and incorporates three-dimensional acceleration, at t=0; f ACC (f r ,t) represents the acceleration phase compensation function.

[0039] Optionally, the envelope descrambling function is:

[0040]

[0041] in, The three-dimensional velocity vector of the dynamic platform at the imaging center time is represented; α0 = β - χ represents the beam pointing of the dynamic platform at time zero. The cone angle formed by the clamping, β represents the ground-scraping angle of the dynamic platform. The dive angle of the dynamic platform, v y v z These represent the velocity components of the dynamic platform along the y-axis and z-axis, respectively; f SQ (f r ,t) represents the envelope deslope function.

[0042] Optionally, based on the range-direction matched filter signal, the observed values ​​of the forward-looking scene scattering coefficients are calculated using a preset acceleration phase compensation function and an envelope deskewing function, including:

[0043] The range-direction matched filter signal is compared with the acceleration phase compensation function f. ACC (f r ,t) and envelope deslope function f SQ (f rMultiply the product (t) and perform an inverse Fourier transform in the distance direction on the product to obtain the observed values ​​of the scattering coefficients of the forward-looking scene.

[0044] Optionally, the process of constructing the overcomplete dictionary matrix includes:

[0045] Under the single-channel forward-looking imaging geometric model, calculate the beam center pointing vector and velocity vector at each point in the motion trajectory of the dynamic platform. in, The vector represents the three-dimensional velocity of the dynamic platform at the imaging center moment, where t represents the azimuth time slow. The three-dimensional acceleration vector representing the large dynamic platform at the imaging center moment;

[0046] Based on the beam center pointing vector and velocity vector Calculate the instantaneous cone angle at the target point, and then use the instantaneous cone angle and velocity vector... Calculate the instantaneous Doppler frequency;

[0047] Calculate the triangular basis phase based on the instantaneous Doppler frequency;

[0048] A complete dictionary matrix is ​​constructed based on the triangular basis phase and the radar antenna pattern.

[0049] Optionally, the formula for calculating the beam center pointing vector is:

[0050]

[0051] The formula for calculating the instantaneous cone angle is:

[0052]

[0053] The formula for calculating the instantaneous Doppler frequency is:

[0054]

[0055] The formula for calculating the phase of a triangular basis is:

[0056] D=exp{j2πf dc t};

[0057] The formula for constructing an overcomplete dictionary matrix is:

[0058] A = D⊙G;

[0059] in, The beam center pointing vector at point D in the motion trajectory of the dynamic platform is represented by α, and the instantaneous cone angle is represented by f. dc Indicates the instantaneous Doppler frequency; D represents the trigonometric basis phase; θ represents the instantaneous beam azimuth angle; λ represents the radar operating wavelength; R sLet H represent the slant range of the scene center at the imaging center moment, H represent the flight altitude of the large dynamic platform at the imaging center moment; exp{·} represents the complex exponential function, t represents the azimuth slow time, j represents the imaginary unit, π represents pi; G represents the radar antenna pattern, D represents the triangular basis phase matrix composed of various triangular basis phases D, A represents the overcomplete dictionary matrix, and ⊙ represents the Hadamard product.

[0060] The present invention provides a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP. Considering the strong mobility of the dynamic platform, a three-dimensional acceleration of the dynamic platform is introduced into the single-channel forward-looking scanning imaging geometry model for the dynamic platform to accurately characterize the forward-looking imaging geometry model of a large dynamic platform. Based on this new model, an acceleration phase compensation function is used to compensate for the higher-order phase terms in the observed echo caused by the three-dimensional acceleration of the dynamic platform, thereby calculating the observed values ​​of the forward-looking scene scattering coefficients. Furthermore, the present invention also considers the influence of the three-dimensional acceleration of the dynamic platform during the construction of the overcomplete dictionary matrix. Therefore, when the present invention uses the improved SAMP algorithm to perform forward-looking super-resolution imaging based on the observed values ​​of the forward-looking scene scattering coefficients and the overcomplete dictionary matrix, it can effectively avoid the mismatch of the sparse reconstruction kernel. In addition, the present invention improves the SAMP algorithm. The improved SAMP algorithm first estimates the sparsity based on the divide-and-conquer method, and then reconstructs the forward-looking scene signal based on the estimated sparsity, obtaining a more accurate sparse reconstruction result, thereby achieving more accurate imaging.

[0061] The present invention will now be described in further detail with reference to the accompanying drawings. Attached Figure Description

[0062] Figure 1 This is a schematic diagram of a single-channel forward-looking scanning imaging geometric model for a dynamic platform constructed in an embodiment of the present invention;

[0063] Figure 2 This is a flowchart of a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP provided in an embodiment of the present invention;

[0064] Figure 3 This is a flowchart of forward-looking super-resolution imaging using the improved SAMP algorithm in an embodiment of the present invention;

[0065] Figure 4 It is a raw bitmap;

[0066] Figure 5 It uses existing technical methods to Figure 4 The result of forward-looking super-resolution imaging of the dot matrix shown.

[0067] Figure 6 The method of this invention is used to... Figure 4The results of forward-looking super-resolution imaging of the dot matrix shown. Detailed Implementation

[0068] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0069] In the process of realizing this invention, the inventors discovered that in dynamic platforms with high operating speeds (such as large dynamic platforms like airplanes), the scenes reconstructed by existing SAMP-based forward super-resolution imaging methods are relatively blurry, and the scene reconstruction success rate decreases significantly with the increase of the dynamic platform's movement speed.

[0070] Through research, the inventors discovered that the three-dimensional acceleration of the dynamic platform not only introduces higher-order phase information into the observation echo, but also changes the true slant range history between the flight platform and the illumination scene. This fundamentally changes the construction method of the overcomplete dictionary, causing the sparse reconstruction dictionary to completely mismatch, resulting in sparse reconstruction failure. Ultimately, this makes forward-looking imaging unable to reflect the true ground object information and unable to acquire forward-looking remote sensing images.

[0071] Therefore, to solve the above problems, this invention provides a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP; this method is based on a novel single-channel forward-looking scanning imaging geometric model for dynamic platforms. This single-channel forward-looking scanning imaging geometric model uses the horizontal equivalent velocity direction of the dynamic platform at time zero as the y-axis, the celestial direction as the z-axis, and the lateral direction as the x-axis in a Cartesian coordinate system, with the projection point of the dynamic platform on the xOy plane at time zero as the origin. Furthermore, it introduces the three-dimensional acceleration of the dynamic platform to accurately characterize its motion trajectory.

[0072] Figure 1 The above-described single-channel forward-looking scanning imaging geometric model for a dynamic platform is shown. Here, ABC represents the motion trajectory of the dynamic platform, which is represented by a curved trajectory in the geometric model. D represents any point on the curved trajectory, D' represents the projection of D onto the xOy plane, T represents the beam center pointing towards the target point in the scene when the dynamic platform flies to point D, and R... s The slant distance from the scene center at the imaging center moment is represented by H, the flight altitude of the dynamic platform at the imaging center moment is represented by ω, and the beam scanning angular velocity is represented by ω. The three-dimensional velocity vector v of the dynamic platform at the imaging center moment represents the velocity vector of the platform. x v y v z These represent the velocity components of the dynamic platform along the x, y, and z axes, respectively. a represents the three-dimensional acceleration vector of the dynamic platform at the imaging center time. x ay a z Let θ represent the acceleration components of the large dynamic platform along the x, y, and z axes, respectively; θ = ωt represents the instantaneous beam azimuth angle; β represents the ground grazing angle formed by the ground and the beam; and t represents the slow time.

[0073] based on Figure 1 The single-channel forward-looking scanning imaging geometric model for dynamic platforms shown in this embodiment of the invention provides a pre-constructed acceleration phase compensation function to compensate for the higher-order phase terms in the observation echo caused by the three-dimensional acceleration of the dynamic platform. This acceleration phase compensation function is as follows:

[0074]

[0075] in, f represents the range-direction frequency domain variable. s f represents the sampling frequency. c Let represent the radar carrier frequency, j represent the imaginary unit, π represent pi, and c represent the speed of light; exp{·} represents the complex exponential function used to characterize the phase information of the echo signal; t represents the azimuth time slow; k2 and k3 represent the second and third order Maclaurin expansion coefficients of the true slant range history R(t) between the target and the dynamic platform, which varies with azimuth time and incorporates three-dimensional acceleration, at t=0; f ACC (f r ,t) represents the acceleration phase compensation function.

[0076] In this invention, the acceleration phase compensation function can eliminate the influence of three-dimensional acceleration on the slant range error envelope.

[0077] based on Figure 1 The single-channel forward-looking scanning imaging geometric model for dynamic platforms shown in this embodiment of the invention, along with the forward-looking super-resolution imaging method for dynamic platforms based on divide-and-conquer blind probe SAMP, also pre-constructs an envelope deslant function, which is as follows:

[0078]

[0079] in, The three-dimensional velocity vector of the dynamic platform at the imaging center time is represented; α0 = β - χ represents the beam pointing of the dynamic platform at time zero. The cone angle formed by the clamping, β represents the ground-scraping angle of the dynamic platform. The dive angle of the dynamic platform, v y v z These represent the velocity components of the dynamic platform along the y-axis and z-axis, respectively; f SQ (f r,t) represents the envelope deslope function.

[0080] In this embodiment of the invention, the envelope deskewing function can eliminate the distortion and translation effects of the dive angle on the Doppler spectrum, thereby converting the large dive imaging model with three-dimensional acceleration into a level flight model without three-dimensional acceleration.

[0081] See Figure 2 As shown, based on the pre-constructed acceleration phase compensation function and envelope descrambling function, the dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP provided in this embodiment of the invention includes the following steps:

[0082] S10: Acquire the baseband echo signal and perform range-direction matched filtering on the baseband echo signal to obtain the range-direction matched filtered signal.

[0083] Specifically, the dynamic platform transmits a linear frequency modulated pulse signal, which can be represented as:

[0084] s t (τ)=w r (τ)cos(2πf c τ+πγτ 2 );

[0085] Among them, s t (τ) represents the linear frequency modulated pulse signal emitted by the dynamic platform, w r (τ) represents the range-directed time-domain window function, τ represents the range-directed fast-time variable, f c γ represents the radar carrier frequency, π represents pi, γ represents the range modulation frequency, and t represents the azimuth time slow.

[0086] Then, the dynamic platform receives the echo signal generated by backscattering from the target area directly in front. The radar receiver demodulates the received echo signal to generate a demodulated baseband echo signal, which is represented as:

[0087]

[0088] Among them, s r (τ,t) represents the demodulated baseband echo signal, w a R(t) represents the azimuth-oriented time-domain window function, exp{·} represents the complex exponential function used to characterize the phase information of the echo signal, j represents the imaginary unit, λ represents the radar operating wavelength, c represents the speed of light, and R(t) represents the true slant range history between the target and the large dynamic platform, which varies with azimuth time and incorporates three-dimensional acceleration. Here, R(t) is expressed by the formula:

[0089]

[0090] Where ||·||2 represents the 2-norm operator, k i The i-th order Maclaurin expansion coefficients of the precise true slant distance history R(t) at t=0 are expressed by the following formula:

[0091]

[0092] Where i! represents the factorial from 1 to i, This represents the i-th derivative of R(t) with respect to the variable t.

[0093] Understandably, the receiver receives the echo, and then performs baseband processing inside the receiver to obtain the baseband echo signal.

[0094] In the baseband echo signal s r Before performing range-direction matched filtering on (τ,t), the baseband echo signal is first processed using the stationary phase principle and series inversion method.

[0095] Specifically, the baseband echo signal s r (τ,t) and the rotation factor of the Fourier transform exp{-j2πf r Multiply by τ and extract the phase from the result. The phase extraction formula is expressed as:

[0096]

[0097] in, f represents the range-direction frequency domain variable. s φ represents the sampling frequency and φ represents the phase.

[0098] According to the stationary phase principle, by taking the partial derivative of the phase φ with respect to the variable τ and setting the result to zero, we obtain its extreme points, which are expressed by the following formula:

[0099]

[0100] Subsequently, using the series inversion method based on the previous formula, the inversely solved range time-domain variable expression τ (i.e., the extreme point) is substituted into the phase φ of the phase extraction formula to obtain the range-to-Fourier transform echo signal s. f (f r ,t), is represented as:

[0101]

[0102] Among them, W r (f r ) is the range-oriented frequency domain window function.

[0103] Construct a range-direction matched filter function in the range frequency domain and azimuth time domain. Using this range-direction matched filter function to perform range-direction matched filtering, the range-direction matched filtered signal is obtained, as follows:

[0104]

[0105] Among them, s RMF (f r ,t) represents the range-direction matched filter signal.

[0106] S20: Based on the range-direction matched filter signal, calculate the observed values ​​of the forward-looking scene scattering coefficient using the preset acceleration phase compensation function and envelope descrambling function.

[0107] Specifically, the distance-oriented matched filter signal s RMF (f r ,t) and acceleration phase compensation function f ACC (f r ,t) and envelope deslope function f SQ (f r Multiply the product (t, t) and perform an inverse Fourier transform to the distance direction on the product to obtain the observed values ​​of the forward-looking scene scattering coefficients. This process can be expressed by the formula:

[0108]

[0109] Among them, s pre (τ,t) represents the observed scattering coefficient of the forward-looking scene, which can be abbreviated as s thereafter; R e ' represents the slant range history expression with phase compensation, which can be calculated by referring to the expression for R(t).

[0110] S30: A complete dictionary matrix is ​​constructed based on radar motion parameters and a single-channel forward-looking scanning imaging geometric model.

[0111] The radar motion parameters include the speed, angle, and altitude of the radar during its movement, all of which are acquired in real time.

[0112] Specifically, considering the impact of the dynamic platform's three-dimensional velocity vector on the actual slant range history, based on radar motion parameters and Figure 1 The single-channel forward-looking scanning imaging geometric model for dynamic platforms shown here is constructed with an improved overcomplete dictionary matrix. The specific construction process includes:

[0113] (1) Calculate the beam center pointing vector and velocity vector at each point in the motion trajectory of the dynamic platform under the single-channel forward-looking scanning imaging geometric model. in, The vector represents the three-dimensional velocity of the dynamic platform at the imaging center moment, where t represents the azimuth time slow. The three-dimensional acceleration vector representing the large dynamic platform at the imaging center moment; The vector represents the three-dimensional velocity of the dynamic platform at the imaging center moment, and t represents the azimuth slow time.

[0114] Here, the formula for calculating the beam center pointing vector is:

[0115]

[0116] (2) Based on the beam center pointing vector and velocity vector Calculate the instantaneous cone angle at the target point, and then use the instantaneous cone angle and velocity vector... Calculate the instantaneous Doppler frequency.

[0117] Here, the formula for calculating the instantaneous cone angle is:

[0118]

[0119] The formula for calculating the instantaneous Doppler frequency is:

[0120]

[0121] (3) Calculate the triangular basis phase based on the instantaneous Doppler frequency.

[0122] Here, the formula for calculating the triangular basis phase is:

[0123]

[0124] (4) Construct an overcomplete dictionary matrix based on the triangular basis phase and the radar antenna pattern.

[0125] Here, the formula for constructing an overcomplete dictionary matrix is:

[0126] A = D⊙G.

[0127] In the above formula, The beam center pointing vector at point D in the motion trajectory of the dynamic platform is represented by α, and the instantaneous cone angle is represented by f. dc Indicates the instantaneous Doppler frequency; D represents the trigonometric basis phase; θ represents the instantaneous beam azimuth angle; λ represents the radar operating wavelength; R s Let H represent the slant range of the scene center at the imaging center moment, H represent the flight altitude of the large dynamic platform at the imaging center moment; exp{·} represents the complex exponential function, t represents the azimuth slow time, j represents the imaginary unit, and π represents pi; G represents the radar antenna pattern, and D represents the phase of each triangular basis. The triangular basis phase matrix is ​​formed, where A represents the overcomplete dictionary matrix and ⊙ represents the Hadamard product.

[0128] D can be specifically represented as:

[0129]

[0130] G can specifically be:

[0131]

[0132] Where K = N a -N b +1, N b N represents the number of scan points within a beamwidth in the scene. a The number of azimuth sampling points is represented by h, which represents the spatial sampling value of the radar's antenna pattern.

[0133] S40: Based on the observed scattering coefficients of the forward-looking scene and the overcomplete dictionary matrix, forward-looking super-resolution imaging is performed using the improved SAMP algorithm.

[0134] Here, the improved SAMP algorithm first estimates sparsity based on the divide-and-conquer method, and then reconstructs the forward-looking scene signal based on the estimated sparsity.

[0135] See Figure 3 As shown, step S40 specifically includes the following sub-steps:

[0136] A. Initialize the signal residual to the observed values ​​of the scattering coefficients of the forward-looking scene, and initialize the atom set to an empty set.

[0137] The signal residual is used to characterize the difference between the actual forward-looking scene signal and the estimated forward-looking scene signal.

[0138] At the same time, in step A, the iteration loop variable t = 1 can also be initialized. Note that in the algorithm iteration process, t represents the iteration loop variable, not the azimuth time.

[0139] B. Update the atom set based on the current signal residual and the overcomplete dictionary matrix, estimate the sparsity using the sparsity over-judgment criterion, and search for the support set for recovering the forward-looking scene signal from the updated atom set based on the estimated sparsity.

[0140] Step B specifically includes:

[0141] B1. Update the atom set using the following formula based on the current signal residual and the overcomplete dictionary matrix:

[0142] λ t ={i:(|Ar t-1 |) i ≥thr×max(|Ar t-1 |) j};

[0143]

[0144] Where thr is the preset first threshold; A represents the overcomplete dictionary matrix, r t-1 Represents the current signal residual; j is the column index of A; λ t This is the first set of subscript indices, where the subscript indices are selected from column index j of A, and the selected subscripts are numbered with i; Indicates the use of λ t The column selected from A, A t Let A represent the updated set of atoms. t-1 This represents the set of atoms before the update.

[0145] B2. Calculation and from Find the largest I atom indices in the set and form the second subscript index set λ. t ': Where s represents the observed value of the scattering coefficient of the forward-looking scene, For A t The transpose of .

[0146] Here, the size of I can be set to I = M / 4, where M is equal to the dimension of the observed scattering coefficient of the forward-looking scene.

[0147] B3. Judgment Is it true? If true, let I = I + 1 and return to step B2, until... If the current I is not found, use the estimated sparsity as the current value and proceed to step B4; if not, set I = I-1 and return to step B2, until... At that time, the current I is used as the estimated sparsity, and step B4 is executed; where, Indicated by λ t 'from The selected columns; ||·||2 represents the 2-norm operator; (2K,δ) 2K ) represents the parameter of an overcomplete dictionary matrix that satisfies the finite isometric RIP property.

[0148] B4. Using the second index set λ when estimating sparsity t ', from the current set of atoms A t The support set selected is used to recover the forward-looking scene signal.

[0149] After step B4 is completed, proceed to step C.

[0150] C. Based on the current support set and the observed values ​​of the scattering coefficients of the forward-looking scene, the estimated forward-looking scene signal is solved using the least squares method. The signal residuals are then calculated and updated based on the estimated forward-looking scene signal, and the residual correlation is calculated.

[0151] Among them, residual correlation is used to characterize the correlation between the current residual and the support set.

[0152] Specifically, based on the current support set and the observed scattering coefficients of the forward-looking scene, the estimated forward-looking scene signal can be obtained using the least squares method, and can be expressed as:

[0153]

[0154] Where s is substituted into the observed value of the scattering coefficient of the forward-looking scene. Substitute the support set, The foreground scene signal to be solved. This represents the solution obtained in the t-th iteration.

[0155] The formula for calculating signal residuals is:

[0156]

[0157] Where, r t This represents the signal residual calculated in the current iteration, i.e., the t-th iteration.

[0158] The formula for calculating residual correlation is:

[0159] ε t =max| <r t ,a j >|;

[0160] Where <·> denotes the operation of taking the dot product of vectors, a j express The j-th column can be understood as the area to be traversed. Each column, thus finding the relationship with r t The column with the largest inner product.

[0161] D. Determine whether the algorithm has converged based on the convergence of the signal residuals and residual correlation.

[0162] If convergence occurs, the currently estimated forward scene signal is used as the reconstructed forward scene signal; if convergence fails, the current support set is used as the atom set and the process returns to step B.

[0163] The method for determining whether the algorithm has converged based on the convergence of the signal residuals and residual correlation includes:

[0164] Judgment | ε t-1 -ε t |<δ or||r t Does ||2 < opt × ||s||2 hold true?

[0165] If any one of these conditions is true, the algorithm is considered to have converged.

[0166] If none of the conditions are met, continue to evaluate ||r t ||2≥||r t-1 Is ||2 true?

[0167] If ||r t ||2≥||r t-1 If ||2 holds true, the algorithm is considered to have converged;

[0168] If ||r t ||2≥||r t-1 If ||2 is not true, the algorithm is considered not to have converged.

[0169] Where, ε t ε represents the residual correlation calculated in this iteration. t-1 This represents the residual correlation calculated in the previous iteration; r t r represents the signal residual calculated in this iteration. t-1 represents the signal residual calculated in the previous iteration, s represents the observed value of the scattering coefficient of the forward-looking scene; δ is the preset second threshold; opt is the preset third threshold; ||·||2 represents the 2-norm operator.

[0170] It should be noted that while existing SAMP algorithms do not require precisely known scene sparsity beforehand, they still require setting a step size for sparsity estimation when the scene sparsity is unknown. This step size determines the accuracy of sparsity estimation and the computation time of the algorithm, making it difficult to simultaneously satisfy reconstruction accuracy and reconstruction speed. In contrast, this invention uses a divide-and-conquer method to estimate sparsity, which can improve the accuracy of sparsity estimation and reduce the computation time.

[0171] Furthermore, existing SAMP algorithms suffer from drawbacks such as excessive pre-selection of atoms during the atom set pre-selection stage, leading to long reconstruction times and fixed step sizes. This results in a significant decrease in sparse reconstruction performance, or even failure to complete sparse reconstruction. In contrast, in this embodiment of the invention, the initial atom set is empty, and the atom set is updated based on the signal residual and the overcomplete dictionary matrix during subsequent iterations. After each iteration, if the algorithm fails to converge, the support set is used as the atom set for continued iteration, improving reconstruction efficiency and success rate.

[0172] The present invention provides a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP. Considering the strong mobility of the dynamic platform, a three-dimensional acceleration of the dynamic platform is introduced into the single-channel forward-looking scanning imaging geometry model for the dynamic platform to accurately characterize the forward-looking imaging geometry model of a large dynamic platform. Based on this new model, an acceleration phase compensation function is used to compensate for the higher-order phase terms in the observed echo caused by the three-dimensional acceleration of the dynamic platform, thereby calculating the observed values ​​of the forward-looking scene scattering coefficients. Furthermore, the present invention also considers the influence of the three-dimensional acceleration of the dynamic platform during the construction of the overcomplete dictionary matrix. Therefore, when the present invention uses the improved SAMP algorithm to perform forward-looking super-resolution imaging based on the observed values ​​of the forward-looking scene scattering coefficients and the overcomplete dictionary matrix, it can effectively avoid the mismatch of the sparse reconstruction kernel. In addition, the present invention improves the SAMP algorithm. The improved SAMP algorithm first estimates the sparsity based on the divide-and-conquer method, and then reconstructs the forward-looking scene signal based on the estimated sparsity, obtaining a more accurate sparse reconstruction result, thereby achieving more accurate imaging.

[0173] Figure 4 The image shows an original bitmap; Figure 5 The image shows imaging results obtained using the existing SAMP algorithm without considering the three-dimensional acceleration of the dynamic platform in the prior art; Figure 6 The image shows the imaging results obtained using the method provided in the embodiments of the present invention; in comparison Figures 4 to 6 As can be seen, the imaging results of the embodiments of the present invention are clearer and significantly superior to those of the prior art. Therefore, the embodiments of the present invention solve the problem of blurred forward super-resolution imaging under dynamic platforms, effectively realizing forward super-resolution imaging of high-speed dynamic platforms.

[0174] The method provided in this invention can be applied to electronic devices. Specifically, the electronic devices referred to herein may include radar, aircraft, satellites, and computers, etc.

[0175] The present invention also provides a computer-readable storage medium. A computer program is stored in the computer-readable storage medium, and when executed by a processor, the computer program implements the steps described in any of the above-described dynamic platform forward-looking super-resolution imaging methods based on divide-and-conquer blind probe SAMP.

[0176] Optionally, the computer-readable storage medium may be non-volatile memory (NVM), such as at least one disk storage device.

[0177] In another embodiment of the present invention, a computer program product containing instructions is also provided, which, when run on a computer, causes the computer to perform the steps of the method described in any of the above-described dynamic platform forward super-resolution imaging methods based on divide-and-conquer blind probe SAMP.

[0178] It should be noted that the description of the storage medium / computer program product embodiment is relatively simple because it is basically similar to the method embodiment. For relevant details, please refer to the description of the method embodiment.

[0179] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this disclosure described herein can be implemented in orders other than those illustrated or described herein. The implementations described in the following exemplary embodiments do not represent all implementations consistent with this disclosure. Rather, they are merely examples of apparatuses and methods consistent with some aspects of this disclosure.

[0180] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., indicate that a specific feature or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. In addition, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.

[0181] Although this application has been described herein in conjunction with various embodiments, those skilled in the art, by reviewing the accompanying drawings and the disclosure, will understand and implement other variations of the disclosed embodiments in carrying out the claimed application. In the description of this invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.

[0182] Those skilled in the art will understand that embodiments of this application can be provided as methods, apparatus (devices), or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects, all of which are collectively referred to herein as "modules" or "systems." Furthermore, this application can take the form of a computer program product implemented on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code. The computer program may be stored / distributed in a suitable medium, provided with or as part of other hardware, or may take other distribution forms, such as via the Internet or other wired or wireless telecommunications systems.

[0183] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (devices), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0184] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0185] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0186] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.

Claims

1. A dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP, characterized in that, include: Acquire the baseband echo signal and perform range-direction matched filtering on the baseband echo signal to obtain the range-direction matched filtered signal; Based on the range-direction matched filter signal, the observed values ​​of the forward-looking scene scattering coefficients are calculated using a preset acceleration phase compensation function and an envelope deskewing function; wherein, the acceleration phase compensation function and the envelope deskewing function are both pre-constructed based on a single-channel forward-looking scanning imaging geometric model for dynamic platforms; An overcomplete dictionary matrix was constructed based on radar motion parameters and the single-channel forward-looking scanning imaging geometric model. Based on the observed scattering coefficients of the forward-looking scene and the overcomplete dictionary matrix, a forward-looking super-resolution imaging is performed using an improved SAMP algorithm. The improved SAMP algorithm first estimates the sparsity based on a divide-and-conquer method, and then reconstructs the forward-looking scene signal based on the estimated sparsity. The single-channel forward-looking scanning imaging geometric model is based on the Cartesian coordinate system with the horizontal equivalent velocity direction of the dynamic platform at time zero as the y-axis, the celestial direction as the z-axis, and the lateral direction as the x-axis. The projection point of the dynamic platform on the xOy plane at time zero is taken as the origin of the coordinate system. The three-dimensional acceleration of the dynamic platform is introduced to accurately characterize the motion trajectory of the dynamic platform. The process of performing forward-looking super-resolution imaging using an improved SAMP algorithm, based on the observed scattering coefficients of the forward-looking scene and the overcomplete dictionary matrix, includes: A. Initialize the signal residual to the observed values ​​of the scattering coefficients of the forward-looking scene, and initialize the atom set to an empty set; wherein, the signal residual is used to characterize the difference between the actual forward-looking scene signal and the estimated forward-looking scene signal; B. Update the atom set according to the current signal residual and the overcomplete dictionary matrix, estimate the sparsity using the sparsity over-judgment criterion, and find the support set for recovering the forward-looking scene signal from the updated atom set based on the estimated sparsity. C. Based on the current support set and the observed values ​​of the forward-looking scene scattering coefficients, the estimated forward-looking scene signal is solved using the least squares method. The signal residual is calculated and updated based on the estimated forward-looking scene signal, and the residual correlation is calculated. The residual correlation is used to characterize the correlation between the current residual and the support set. D. Determine whether the algorithm has converged based on the convergence of the signal residuals and residual correlation. If it has converged, use the currently estimated forward-looking scene signal as the reconstructed forward-looking scene signal. If it has not converged, return to step B as the current support set as the atom set. Step B includes: B1. Update the atom set using the following formula based on the current signal residual and the overcomplete dictionary matrix: ； ； in, The first threshold is preset; This represents the overcomplete dictionary matrix. This represents the current signal residual; j for Column index; For the first subscript index set, the subscript indexes are from column index j Select from the options, and use the selected subscripts. i serial number; Indicates the use of from The selected column This represents the updated set of atoms. This represents the set of atoms before the update; B2. Calculation and from Find the largest one I Atom indices constitute the second subscript index set. :in, This represents the observed value of the scattering coefficient of the forward-looking scene. for The transpose of the matrix; B3. Judgment Is it true? If it is true, let I = I After adding 1, return to step B2 until... At that time, the current I As the estimated sparsity, proceed to step B4; if not, then let I = I After -1, return to step B2 until... At that time, the current I As the estimated sparsity, step B4 is performed; where, Indicates using from The selected columns; Represents the 2-norm operator; The parameter represents the overcomplete dictionary matrix satisfying the finite isometric RIP property; B4. Utilize the second index set when estimating the sparsity. From the current set of atoms The support set used to recover the forward-looking scene signal is selected from the data. The method for determining whether the algorithm has converged based on the convergence of the signal residuals and residual correlation includes: judge or Is it valid? If any one of these conditions is met, the algorithm is considered to have converged. If none of the conditions are met, continue the evaluation. Is it valid? like If the condition is met, the algorithm is considered to have converged. like This is not valid; the algorithm is considered not to have converged. in, This represents the residual correlation calculated in this iteration. This represents the residual correlation calculated in the previous iteration; This represents the signal residual calculated in this iteration. This represents the signal residual calculated in the previous iteration. This represents the observed value of the scattering coefficient of the forward-looking scene; This is the preset second threshold; This is the preset third threshold; This represents the 2-norm operator.

2. The dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP according to claim 1, characterized in that, The acceleration phase compensation function is: ； in, Represents the range-direction frequency domain variable. Indicates the sampling frequency. Indicates the radar carrier frequency. Represents the imaginary unit. Represents pi (π). c Represents the speed of light; This represents a complex exponential function, used to characterize the phase information of the echo signal; Indicates direction in slow time. and These represent the true slant range history between the target and the dynamic platform, which varies with azimuth and time and incorporates three-dimensional acceleration. exist The coefficients of the second and third order Maclaurin expansions at that point, This represents the acceleration phase compensation function.

3. The dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP according to claim 2, characterized in that, The envelope descrambling function is: ； in, The three-dimensional velocity vector representing the dynamic platform at the imaging center moment; Indicates the beam pointing of the dynamic platform at time zero and The angle of the cone formed by the clamping, This indicates the corner of the dynamic platform. Indicates the dive angle of the dynamic platform. These represent the velocity components of the dynamic platform along the y-axis and z-axis, respectively. This represents the envelope deslope function.

4. The dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP according to claim 3, characterized in that, Based on the range-guided matched filter signal, the observed values ​​of the forward-looking scene scattering coefficients are calculated using a preset acceleration phase compensation function and envelope deskewing function, including: The range-direction matched filter signal and the acceleration phase compensation function are used. and envelope deslant function Multiply the results and perform a distance-to-inverse Fourier transform on the product to obtain the observed values ​​of the scattering coefficients of the forward-looking scene.

5. The dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP according to claim 1, characterized in that, The process of constructing the overcomplete dictionary matrix includes: Under the single-channel forward-looking imaging geometric model, calculate the beam center pointing vector and velocity vector at each point in the motion trajectory of the dynamic platform. ;in, The three-dimensional velocity vector representing the dynamic platform at the imaging center moment. Indicates direction in slow time. The three-dimensional acceleration vector representing the large dynamic platform at the imaging center moment; Based on the beam center pointing vector and velocity vector Calculate the instantaneous cone angle at the target point, and then use the instantaneous cone angle and velocity vector... Calculate the instantaneous Doppler frequency; Calculate the triangular basis phase based on the instantaneous Doppler frequency; A complete dictionary matrix is ​​constructed based on the triangular basis phase and the radar antenna pattern.

6. The dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind probe SAMP according to claim 5, characterized in that, The formula for calculating the beam center pointing vector is: ； The formula for calculating the instantaneous cone angle is: ； The formula for calculating the instantaneous Doppler frequency is: ； The formula for calculating the phase of a triangular basis is: ； The formula for constructing an overcomplete dictionary matrix is: A = D⊙G; in, In the motion trajectory of the dynamic platform D The beam center of the point points to the vector. Indicates the instantaneous cone angle; Indicates the instantaneous Doppler frequency; Indicates the trigonometric basis phase; Indicates the instantaneous beam azimuth angle. Indicates the radar operating wavelength. The slant distance from the scene center at the moment of image centering. H This indicates the flight altitude of the large dynamic platform at the moment of imaging center. Represents a complex exponential function. Indicates direction in slow time. Represents the imaginary unit. Represents pi; This represents the antenna pattern of the radar. Represents the phase of each triangular basis The triangular basis phase matrix is ​​formed. Let represent an overcomplete dictionary matrix, and ⊙ denotes the Hadamard product.

Images