CN-116338685-B - Dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind detection SAMP

CN116338685BCN 116338685 BCN116338685 BCN 116338685BCN-116338685-B

Abstract

The invention discloses a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind detection SAMP, which comprises the steps of conducting distance-direction filtering on a baseband echo signal, calculating a forward view field Jing Sanshe coefficient observation value according to the filtering signal by utilizing an acceleration phase compensation function and an envelope declivity function, constructing an overcomplete dictionary matrix, conducting forward-looking super-resolution imaging according to the forward view field Jing Sanshe coefficient observation value and the overcomplete dictionary matrix by utilizing an improved SAMP algorithm, conducting sparsity estimation by the algorithm based on the divide-and-conquer method, reconstructing a foreground view signal based on the sparsity, constructing a single-channel forward-looking scanning imaging geometric model facing the dynamic platform by utilizing two functions and the overcomplete dictionary matrix, and introducing three-dimensional acceleration of the dynamic platform to accurately represent the motion track of the dynamic platform. The invention solves the problem of forward-looking super-resolution imaging blurring under the dynamic platform and effectively realizes forward-looking super-resolution imaging of the dynamic platform running at high speed.

Inventors

LIANG YI
GUO YIHENG
WANG TINGTING
LIANG YUJIE
XING MENGDAO

Assignees

西安电子科技大学

Dates

Publication Date: 20260512
Application Date: 20230222

Claims (6)

1. A dynamic platform forward-looking super-resolution imaging method based on blind detection SAMP based on divide and conquer is characterized by comprising the following steps: acquiring a baseband echo signal, and performing distance matching filtering processing on the baseband echo signal to obtain a distance matching filtering signal; Calculating a front view field Jing Sanshe coefficient observation value by using a preset acceleration phase compensation function and an envelope declivity function according to the distance matched filtering signal, wherein the acceleration phase compensation function and the envelope declivity function are both pre-constructed based on a single-channel forward-looking scanning imaging geometric model facing a dynamic platform; constructing an overcomplete dictionary matrix based on radar motion parameters and the single-channel forward-looking scanning imaging geometric model; Performing forward-looking super-resolution imaging by utilizing an improved SAMP algorithm according to the forward-looking Jing Sanshe coefficient observation value and the overcomplete dictionary matrix, wherein the improved SAMP algorithm firstly performs sparsity estimation based on a divide-and-conquer method, and then reconstructs foreground signals of the forward-looking field based on the estimated sparsity; The single-channel forward-looking scanning imaging geometric model takes the horizontal equivalent speed direction of the dynamic platform at zero time as a y axis, the vertical direction as a z axis and the lateral direction as an x axis under a Cartesian coordinate system, takes a projection point of the dynamic platform at zero time on an x-O-y plane as an origin of coordinates, and introduces three-dimensional acceleration of the dynamic platform to accurately represent the motion track of the dynamic platform; Wherein, based on the front field Jing Sanshe coefficient observations and the overcomplete dictionary matrix, performing front-view super-resolution imaging by using a modified SAMP algorithm, comprising: A. initializing a signal residual as the front view Jing Sanshe coefficient observed value and initializing an original subset as an empty set, wherein the signal residual is used for representing the difference between a real foreground view signal and an estimated foreground view signal; B. Updating the original subset according to the current signal residual error and the overcomplete dictionary matrix, utilizing a sparsity excess judgment criterion to divide and treat the estimated sparsity, and searching a support set for recovering foreground signals from the updated atomic set based on the estimated sparsity; C. According to the current support set and the front view Jing Sanshe coefficient observation value, utilizing a least square method to solve an estimated front view scene signal, calculating and updating a signal residual according to the estimated front view scene signal, and calculating a residual correlation degree, wherein the residual correlation degree is used for representing the correlation between the current residual and the support set; D. B, judging whether an algorithm is converged according to convergence conditions of signal residual errors and residual error relativity, if so, taking the current estimated foreground scene signal as a reconstructed foreground scene signal, and if not, taking the current support set as a primary subset and returning to the step B; Wherein, step B includes: b1, updating the original subset according to the current signal residual and the overcomplete dictionary matrix by using the following formula: ; ; Wherein, the Is a preset first threshold; Representing the overcomplete dictionary matrix, Representing the current signal residual, j is Is a column label of (2); for a first set of index indices, where index indices are derived from The index j of the list is selected, and the selected index is numbered by i; representation utilization From the slave The column selected from the group consisting of the rows, Representing the set of atoms after the update, Representing the original subset before updating; b2, calculating And from Finding out the maximum I atom indexes to form a second index set Wherein, the method comprises the following steps of, Representing the front field Jing Sanshe coefficient observations, Is that Is a transposed matrix of (a); b3, judging If so, returning to the step B2 until the I=I+1 is satisfied When the current I is used as the estimated sparsity, executing the step B4; if not, returning to the step B2 after I=I-1 until And when the current I is used as the estimated sparsity, executing a step B4, wherein, For representation From the slave The selected column; representing a 2-norm operator; parameters representing that the overcomplete dictionary matrix satisfies a finite equidistant RIP property; B4, utilizing a second index set of subscripts when the sparsity is estimated From the current set of atoms Selecting a support set for recovering foreground view signals; judging whether the algorithm converges according to the convergence condition of the signal residual error and the residual error correlation degree, wherein the method comprises the following steps: Judging Or (b) Whether or not to establish; If any one of the two is true, the algorithm is considered to be converged; if none of the above is true, continuing to judge Whether or not to establish; If it is If true, the algorithm is considered to be converged; If it is If not, the algorithm is considered not to be converged; Wherein, the Representing the residual correlation calculated in this iteration, Representing the residual correlation calculated in the last iteration; representing the signal residuals computed in this iteration, Representing the signal residual calculated in the last iteration, Representing the front field Jing Sanshe coefficient observations; is a preset second threshold; is a preset third threshold; Representing a 2-norm operator.
2. The dynamic platform forward looking super resolution imaging method based on divide-and-conquer blind detection SAMP according to claim 1, wherein the acceleration phase compensation function is: ; Wherein, the Representing the variation of the distance to the frequency domain, Representing the sampling frequency of the sample, Representing the radar carrier frequency, Representing the units of an imaginary number, The circumference ratio, c, the speed of light; representing a complex exponential function to characterize phase information of the echo signal; Indicating that the azimuth is slow for a time, And Representing the true pitch history between an azimuth time-varying object incorporating three-dimensional acceleration and a dynamic platform, respectively At the position of 2 Nd and 3 rd order majulin expansion coefficients at, Representing the acceleration phase compensation function.
3. The dynamic platform forward looking super resolution imaging method based on divide-and-conquer blind detection SAMP according to claim 2, wherein the envelope declivity function is: ; Wherein, the A three-dimensional velocity vector representing the dynamic platform at the moment of the imaging center; Beam pointing representing a time-zero dynamic platform The taper angle formed by the clamping is equal to the taper angle, Representing the ground-wiping angle of the dynamic platform, Representing the angle of depression of the dynamic platform, Representing velocity components of the dynamic platform along the y-axis and the z-axis respectively; representing an envelope deskewing function.
4. A dynamic platform forward looking super resolution imaging method based on divide-and-conquer blind detection SAMP according to claim 3, wherein calculating a forward field Jing Sanshe coefficient observation value by using a preset acceleration phase compensation function and an envelope declivity function according to the distance direction matched filtering signal comprises: matching the distance direction with the filtered signal and the acceleration phase compensation function And envelope deskewing function Multiplying and performing distance inverse Fourier transform on the product result to obtain a front view Jing Sanshe coefficient observation value.
5. The dynamic platform forward looking super resolution imaging method based on blind detection SAMP of claim 1, wherein the construction process of the overcomplete dictionary matrix comprises: calculating beam center pointing vectors and speed vectors of each point in motion tracks of a dynamic platform under the single-channel forward-looking scanning imaging geometric model Wherein, the method comprises the steps of, A three-dimensional velocity vector representing the dynamic platform at the moment of the imaging center, Indicating that the azimuth is slow for a time, Representing a three-dimensional acceleration vector of a large dynamic platform at the moment of an imaging center; from beam centre pointing vector and velocity vector Calculating the instantaneous cone angle of the target point and according to the instantaneous cone angle and the speed vector Calculating an instantaneous Doppler frequency; calculating a triangle base phase according to the instantaneous Doppler frequency; an overcomplete dictionary matrix is constructed from the triangle base phases and the antenna pattern of the radar.
6. The dynamic platform forward-looking super-resolution imaging method based on blind detection SAMP of claim 5, wherein, The calculation formula of the beam center pointing vector is as follows: ; The calculation formula of the instantaneous cone angle is as follows: ; The calculation formula of the instantaneous Doppler frequency is as follows: ; The calculation formula of the triangle base phase is: ; the building formula of the overcomplete dictionary matrix is as follows: A=D⊙G; Wherein, the The beam center pointing vector representing the D point in the motion trajectory of the dynamic platform, Represents the instantaneous cone angle; Representing instantaneous Doppler frequency; representing the triangle base phase; Representing the instantaneous beam azimuth angle, Indicating the operating wavelength of the radar, The scene center slant distance at the moment of the imaging center is represented, and H represents the flying height of a large dynamic platform at the moment of the imaging center; which represents a complex exponential function that is a function of the complex index, Indicating that the azimuth is slow for a time, Representing the units of an imaginary number, Representing the circumference ratio; representing the antenna pattern of the radar, Representing the phase of each triangle base A triangular base phase matrix is formed by the two components, Representing an overcomplete dictionary matrix, +..

Description

Dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind detection SAMP Technical Field The invention belongs to the field of remote sensing detection, and particularly relates to a dynamic platform forward-looking super-resolution imaging method based on divide-and-conquer blind detection SAMP (sparsity self-adaptive matching pursuit). Background The forward-looking mode is an important imaging mode in the field of remote sensing detection, can be used for detecting and imaging the area right in front of the flight direction in all weather and long distance all the day, and has wide application prospect. Meanwhile, the forward-looking imaging result can be used for detecting and identifying high-value targets in the region of interest in front of the target, compared with a single-pulse target detection method, the forward-looking imaging greatly improves the detection and identification precision, and therefore forward-looking radar imaging becomes a research focus in recent years. In a conventional synthetic aperture radar (SYNTHETIC APERTURE RADAR, abbreviated as SAR) imaging system, in order to acquire a two-dimensional high-resolution microwave remote sensing image, a sufficiently long synthetic aperture or a sufficiently large coherent accumulation angle needs to be accumulated along a distance transverse direction, so that an equivalent virtual array can be formed in space, and an extremely narrow wave beam is formed by interference, so that the distance transverse high-resolution detection of an irradiation region is realized. For the field of microwave remote sensing imaging, the high resolution capability of the distance direction can be realized by a pulse compression technology, and the performance of the high resolution capability is determined by the bandwidth of a transmitted signal. However, in forward looking imaging geometry, the radar antenna beam is directed directly in front of the platform's flight direction, so the synthetic aperture length from the lateral cannot accumulate with the platform's motion. In addition, when the radar beam is directed to the right front of the flight direction, the left and right sides of the forward-looking scene have the same space cone angle, and the problem of left and right Doppler blurring during forward-looking imaging can be caused. The distance and transverse resolution of the two problems are limited by the length of the azimuth real aperture and the blurring of left and right Doppler, and the traditional single-channel SAR imaging system and the signal processing method cannot be directly applied to the forward-looking imaging problem. The super-resolution sparse signal reconstruction theory provides an improved thought for the difficult problem of resolution in the transverse direction, and the single-channel forward-looking imaging method combining beam scanning and super-resolution technology is a feasible scheme in the field of forward-looking imaging of a large dynamic platform. For example, there are prior art forward looking super resolution imaging methods based on OMP (orthogonal matching pursuit) or SAMP. In the OMP-based forward-looking super-resolution imaging method, sparsity is required as a priori knowledge, so that the method cannot be applied under the condition that the scene sparsity is unknown. The front-view super-resolution imaging method based on the SAMP does not need to accurately know scene sparsity in advance, in the method, firstly, a distance-azimuth echo model is established, distance direction processing is completed in a distance compression mode and the like, a Doppler convolution model is established, the reconstruction problem of the front-view scene under the condition of a known observation value is represented in a linear regression model mode, namely, the construction of an overcomplete dictionary matrix A is completed, and then the SAMP method is used as a sparse reconstruction kernel to recover the foreground view. However, in a dynamic platform with higher running speed, the existing forward-looking super-resolution imaging method based on SAMP has the problem that the reconstructed scene is blurred. Disclosure of Invention In order to solve the problems in the prior art, the invention provides a dynamic platform forward-looking super-resolution imaging method based on blind detection SAMP. The technical problems to be solved by the invention are realized by the following technical scheme: A dynamic platform forward-looking super-resolution imaging method based on blind detection SAMP of divide and conquer comprises the following steps: acquiring a baseband echo signal, and performing distance matching filtering processing on the baseband echo signal to obtain a distance matching filtering signal; Calculating a front view field Jing Sanshe coefficient observation value by using a preset acceleration phase compensation function and an envelope declivity function