GPU parallel acceleration space-based ISAR self-focusing rapid motion compensation method

Through GPU parallel processing and staged self-focusing methods, the problem of low two-dimensional motion compensation efficiency in space-based ISAR imaging is solved, efficient and accurate motion compensation is achieved, and imaging quality is improved.

CN119087367AActive Publication Date: 2024-12-06XIDIAN UNIV

Patent Information

Application Number
CN202411185144.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-27
Publication Date
2024-12-06
Estimated Expiration
2044-08-27

AI Technical Summary

Technical Problem

In space-based ISAR imaging scenarios, due to the high-speed motion of radar and target, the imaging geometry is complex, and the prior art is difficult to achieve efficient and accurate two-dimensional motion compensation, and the signal-to-noise ratio is low in a low signal-to-noise environment, which affects the imaging quality. .

Method used

Adopting the GPU parallel processing architecture, the distance and orientation dimensions are interpolated by polar coordinate resampling and staged self-focusing methods, reducing the parallel processing of loop statements, combining phase gradient self-focusing and image minimum entropy self-focusing to achieve two-dimensional motion compensate.

Benefits of technology

The efficiency of polar coordinate resampling and the accuracy of two-dimensional motion compensation are improved, and the time consumption is increased from seconds to milliseconds, which significantly improves imaging quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119087367A_ABST
    Figure CN119087367A_ABST
Patent Text Reader

Abstract

The invention provides a GPU (Graphics Processing Unit) parallel acceleration space-based ISAR (Inverse Synthetic Aperture Radar) self-focusing rapid motion compensation method, which comprises the following steps of: measuring echo data and angle measurement information of a space target by using a space-based radar, and synchronously calculating and fitting radar corner information by using a GPU parallel processing architecture; parallel design is carried out through polar coordinate resampling, loop statements are reduced, front-back coupling relevance is reduced, and efficient two-dimensional motion decoupling is achieved; then combining a polar coordinate format algorithm and a motion error inversion scheme, carrying out distance down-sampling on the signal after two-dimensional motion decoupling, and intercepting an effective data segment to reduce the data volume; phase gradient self-focusing and image minimum entropy self-focusing are adopted in stages to estimate distance non-space-variant errors, parallel processing of the phase gradient self-focusing and the image minimum entropy self-focusing is achieved through a GPU parallel processing architecture, interpolation estimation is conducted on the space-variant errors through the estimated errors, the estimation precision and the calculation efficiency are improved, and the method is suitable for large-scale popularization and application. And steady focusing imaging of the space target is completed.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of radar image processing technology, and in particular to a space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization. Background Art

[0002] Inverse synthetic aperture radar (ISAR) plays an important role in the observation of aviation and aerospace targets due to its characteristics of all-day, all-weather, high resolution and long distance. In the space-based ISAR imaging scene of space targets, because both the radar and the target are moving at high speed, the imaging geometry of space-based ISAR becomes more complicated, causing strong two-dimensional motion coupling and phase space-variation problems in the imaging model, which in turn causes the motion compensation accuracy to decrease and the imaging result to be space-variant and defocused. In addition, because the satellite payload is limited and the radar power is low, when the target is far away from the radar, the received radar echo will have a low signal-to-noise ratio problem, which will further reduce the accuracy of space-based ISAR motion compensation and migration correction of space targets, causing the imaging result to be defocused, greatly reducing the imaging quality. Therefore, how to efficiently and accurately realize the two-dimensional motion compensation of space targets has become a key technology that must be solved to improve the imaging quality.

[0003] In the prior art, methods for motion correction of turntable models generally include spherical transformation and polar coordinate resampling methods. Traditional spherical transformation realizes motion correction of the distance dimension by means of scale transformation, but there is strong two-dimensional motion coupling in the ISAR imaging scenario of space targets, and there is still defocusing of the azimuth dimension after spherical transformation, so spherical transformation is not suitable for this scenario. Polar coordinate resampling is usually used to decouple the two-dimensional motion. Polar coordinate resampling is generally implemented by performing one-dimensional interpolation on the distance dimension and the azimuth dimension respectively. Through multiple loop iterations, the distance dimension and the azimuth dimension are interpolated to realize two-dimensional motion decoupling, which is inefficient, and the parallelism of the algorithm is low and the time complexity is high.

[0004] In the prior art, the method for realizing residual translation compensation is to estimate the non-space-variant error of the distance by using the minimum entropy self-focusing method of the image after distance downsampling, and then to estimate the space-variant error by interpolation, and finally to realize the residual translation compensation. Directly estimating the non-space-variant error term by using the minimum entropy self-focusing method of the image after distance downsampling is easy to fall into the local optimal situation, resulting in inaccurate estimation accuracy, affecting the subsequent interpolation results and imaging quality, and using the minimum entropy self-focusing method of the image, the number of iterations is large and the efficiency is low. Summary of the invention

[0005] In order to solve the above problems existing in the prior art, the present invention provides a space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization. The technical problem to be solved by the present invention is achieved by the following technical solutions:

[0006] The present invention provides a space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization, comprising:

[0007] S100, receiving echo data of space targets in a low signal-to-noise ratio environment and space-based radar angle measurement information, converting all echo data into a wave number domain to obtain echo data in the wave number domain, and composing all echo data into an original echo matrix;

[0008] S200, using a GPU parallel processing architecture to parallel fit the angle information of the target at each pulse moment to form angle information;

[0009] S300, using the GPU parallel processing architecture to parallelly resample the echo matrix of the current iteration in polar coordinates to obtain a two-dimensional interpolated spectrum;

[0010] S400, intercepting the range data in the spectrum after the two-dimensional interpolation and downsampling it, then selecting the valid data segment in the azimuth direction, and assigning values ​​thereto to obtain a first echo matrix;

[0011] S500, inputting the first echo matrix into the GPU parallel processing architecture, and estimating the first stage range non-space-varying error in the GPU parallel processing architecture;

[0012] S600, inputting the non-space-varying distance error of the first stage into the GPU parallel processing architecture, interpolating and solving the two-dimensional space-varying error phase in the GPU parallel processing architecture, and using the two-dimensional space-varying error phase to roughly compensate the echo matrix in S500 to obtain the echo matrix of the current cycle, which is used as the echo matrix input of the next cycle, and repeating the process from S500 to S600 until the number of cycle iterations is reached to obtain the echo matrix after the compensation of the first stage;

[0013] S700, inputting the echo matrix compensated in the first stage into the GPU parallel processing architecture, and estimating the range non-space-varying error of the second stage in the GPU parallel processing architecture;

[0014] S800, inputting the non-space-varying distance error of the second stage into the GPU parallel processing architecture, solving the two-dimensional space-varying error phase by parallel interpolation, and using the two-dimensional space-varying error phase to accurately compensate the echo matrix in S700 to obtain the echo matrix after the compensation of the second stage, and using it as the echo matrix of the next iteration;

[0015] S900, repeat the process from S300 to S800 until the number of loop iterations is reached, and obtain a two-dimensional motion-compensated echo matrix, which is subjected to a two-dimensional Fourier transform to obtain a focused image.

[0016] Beneficial effects:

[0017] 1. The present invention makes full use of the characteristics of GPU computing cores to parallelize and improve the polar coordinate resampling algorithm. For two-dimensional sinc interpolation, the distance dimension and the azimuth dimension are interpolated separately, the loop statements in each dimension interpolation are reduced, and the front-to-back coupling correlation is reduced. When each thread interpolates, the search for interpolation points adopts a binary search method, which can effectively improve the efficiency and reduce the time consumption of polar coordinate resampling from the original seconds to close to milliseconds.

[0018] 2. The present invention considers that if the two-dimensional space-variant motion compensation directly uses the image minimum entropy self-focusing method to estimate the non-space-variant error, and inverts the space-variant error to compensate, there will be too many iterations, low efficiency, and may fall into the local optimum. In response to this problem, it is proposed to use different self-focusing methods in stages to estimate the non-space-variant error, invert and compensate for the two-dimensional space-variant error, which can effectively improve the estimation accuracy, and use the characteristics of GPU computing cores to further improve the efficiency, and the total time consumption of two-dimensional space-variant motion compensation is increased from seconds to milliseconds.

[0019] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 A flowchart of a GPU parallel-accelerated space-based ISAR autofocus fast motion compensation method provided by the present invention;

[0021] Figure 2a A schematic diagram of the compensation process of the space-based ISAR self-focusing fast motion compensation method with GPU parallel acceleration provided by the present invention;

[0022] Figure 2b A flowchart for quickly implementing two-dimensional sinc interpolation with GPU parallel acceleration provided by the present invention;

[0023] Figure 3 A simulation point target scattering model diagram provided by the present invention;

[0024] Figure 4 The RD imaging result diagram before two-dimensional motion decoupling of the space-based ISAR echo provided by the present invention after de-slant parametrization and rough translation compensation;

[0025] Figure 5 The frequency domain result diagram after resampling in polar coordinates of MATLAB and GPU provided by the present invention;

[0026] Figure 6 This is the RD imaging result diagram after two-dimensional motion compensation on the MATLAB side and the GPU side provided by the present invention. DETAILED DESCRIPTION

[0027] The present invention is further described in detail below with reference to specific embodiments, but the embodiments of the present invention are not limited thereto.

[0028] The idea of ​​realizing the purpose of the present invention is: based on the radar principle, the ISAR turntable model expression is derived, and the correlation between echo data can be used to effectively accumulate echo energy, which is particularly suitable for estimating target motion parameters in a low signal-to-noise ratio echo environment. The echo is transformed into the wavenumber domain, and polar coordinate resampling is adopted. By interpolating the echo in polar coordinates and transforming it into the rectangular coordinate system, the decoupling of distance and azimuth is realized, and the effect of rotation compensation and two-dimensional motion decoupling is achieved. According to the principle of polar coordinate resampling, a highly parallelized algorithm is designed and implemented to realize efficient two-dimensional motion decoupling on the GPU side. After resampling, the signal after two-dimensional motion decoupling is distance downsampled, and then the valid data segment is intercepted to reduce the amount of data. The polar coordinate format algorithm is combined with the motion error inversion technology, and the phase gradient self-focusing and image minimum entropy self-focusing are used in stages to combine with the echo after polar coordinate resampling to estimate the non-space-variant error of the distance of the downsampled data, and the space-variant error is interpolated and estimated through the estimated error. The loop iteration realizes the precise compensation of the residual translation. According to the above-mentioned motion compensation principle, the phase gradient self-focusing, image minimum entropy self-focusing, etc. are parallelized to realize efficient motion compensation on the GPU side. Finally, the focused ISAR imaging result is obtained by performing two-dimensional Fourier transform on the motion compensated data.

[0029] Combination Figure 1 and Figure 2a The present invention provides a space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization, comprising:

[0030] S100, receiving echo data of space targets in a low signal-to-noise ratio environment and space-based radar angle measurement information, converting all echo data into a wave number domain to obtain echo data in the wave number domain, and composing all echo data into an original echo matrix;

[0031] Assuming that the target has completed de-oblique echo parametrization and horizontal coarse compensation, the moving radar echo is received in a low signal-to-noise ratio environment, and the echo data is converted into the wavenumber domain, which is expressed as

[0032] W(k r ,θ)=P(k r )exp(-jk r (xsinθ+ycosθ))exp(-jk r ΔR(θ))(1)

[0033] Among them, P(k r ) represents the signal envelope, k rrepresents the radial wave number, θ represents the rotation angle of the target, (x, y) represents the position of the target scattering point in space, and ΔR(θ) represents the residual translational momentum.

[0034] S200, using a GPU parallel processing architecture to parallel fit the angle information of the target at each pulse moment to form angle information;

[0035] Assuming that the number of echo points in range is M, the total number of points in azimuth is N, the number of points required for the angular accumulation pulse number of the i-th ISAR imaging is N i , four data storage areas are opened in the GPU memory, where the four data storage areas are data storage areas echo_all, M×N i ×8 bytes of data storage area echo_cur, N×8 bytes of data storage area Cta and N i ×8 bytes of data storage area Cta_temp, M is the number of points in the range direction of the echo signal, N is the total number of points in the azimuth direction of the echo data, and the number of points of the angular accumulation pulses required for imaging the i-th ISAR image is N i ; Use the GPU parallel processing architecture to set up (N×8+512-1) / 512 blocks, each block contains at least 512 threads;

[0036] Run multiple threads in parallel to fit the rotation angle θ at the nth pulse time from the radar angle measurement information using the rotation angle fitting formula n , and store the calculation results in the data storage area Cta;

[0037] Among them, a single block thread can set up to 1024 threads. Then through the cudaMemcpyAsync() function, address offset and multi-stream parallel processing operations, the device-side Cta is assigned to the device-side Cta_temp. Among them, the GPU performs operations with double type to ensure the operation accuracy. echo_all stores the echo data collected within a period of time, echo_cur stores the i-th echo data that currently meets the imaging conditions, Cta stores the angle information of the space target within a period of time, and Cta_temp stores the angle information required for the i-th ISAR imaging.

[0038] S300, using the GPU parallel processing architecture to perform resampling processing on the echo matrix of the current iteration in polar coordinates in parallel to obtain a two-dimensional interpolated spectrum;

[0039] The echo matrix is ​​resampled in polar coordinates, that is, k r sinθ i →k x ,k r cosθi →k y , the resampling formula is expressed as:

[0040] W(k y ,k x )=W(k r cosθ i ,k r sinθ i )=P(k y )exp(-jk x x-jk y y)exp(-jφ e )

[0041]

[0042] θ i =arctan(k x / k y )

[0043] where φ e represents the error phase caused by the residual translation, φ r0 (k y0 ,k x ) represents the non-space-varying error, φ rd (k y ,k x ) represents the space-variant error, k y0 represents the distance to the wave number center, φ r0 and φ rd Only the values ​​of the independent variables are different, so we can get r0 Interpolation gives φ rd ,θ i represents the change in rotation angle after resampling of the i-th ISAR image, k x represents the azimuthal wave number, k y represents the distance wave number, and arctan() represents the inverse tangent.

[0044] In a specific implementation of the present invention, S300 includes:

[0045] S310, constructing the matrix size of the range wave number and the azimuth wave number according to the size of the original echo matrix; constructing the matrix size of the radial wave number matrix k r and the angle θ at the i-th pulse moment i Construct the values ​​of the first matrix of distance wavenumbers and the second matrix of azimuth wavenumbers; allocate M×N matrices of size in the GPU global memory respectively i × 8 bytes of distance to the wave number data storage area kx_new and M × N i×8 bytes of azimuth wavenumber data storage area ky_new; using the GPU parallel processing architecture, set (M×N i +512-1) / 512 blocks, each block containing at least 512 threads; using multiple threads to update the values ​​of the first matrix and the second matrix in parallel, to obtain the first updated first matrix and the second matrix, that is, to construct a first matrix of range wavenumbers and a second matrix of azimuth wavenumbers;

[0046] Among them, the size of the original echo matrix is ​​M×N i ×8×2 bytes.

[0047] S320, using GPU parallel processing architecture, set ((ceil(log2M×N i )) 2 +512-1) / 512 blocks, where ceil means rounding up. Each block contains at least 512 threads. Using the parallel reduction idea, M×N i The data is divided into M×N levels i / 2 groups, each group contains two data sampling points, using the GPU's M×N i / 2 threads parallely calculate the maximum value of each set of data sampling points; for M×N i / 2 extreme values ​​are grouped in pairs, and the solution process of finding the extreme values ​​is iterated cyclically until one group of extreme values ​​remains; the maximum and minimum values ​​of the updated range wavenumber and azimuth wavenumber are solved by using the processing scheme for solving the remaining group of extreme values; the first matrix and the second matrix updated for the first time are updated according to the maximum and minimum values ​​to obtain the first matrix and the second matrix updated for the second time;

[0048] It is worth noting that: setting the shared memory kx_temp

[512] , the maximum shared memory can be set to 1024, assigning global memory to shared memory can further improve efficiency. i If it is not a multiple of 2, kx_temp is supplemented. If the maximum value is required, it is supplemented to INT_MIN. Otherwise, it is supplemented to INT_MAX, so as to solve the maximum and minimum values ​​of the distance wave number.

[0049] S330, using the GPU parallel processing architecture to set the same number of blocks and threads as in S320, and combining the first matrix and the second matrix updated for the second time to calculate the radial wave number matrix k r Perform parallel updates; allocate M×N size in GPU global memory i × 8 bytes of data storage ind, and solve in parallel on the new k x -k yThe index value ind in the coordinate system is stored in the data storage area ind; the size of M×N is allocated in the GPU global memory i ×8×2 bytes of data storage area s1; using the GPU parallel processing architecture and sinc interpolation principle, setting the same number of blocks and threads as in S320, combining the index value ind to perform one-dimensional range interpolation on the original echo matrix to obtain an interpolated spectrum, and storing it in the data storage area s1;

[0050] It is worth noting that the interpolated spectrum is of complex double type. The formula used for interpolation is:

[0051]

[0052] Among them, s represents the echo spectrum of the original echo matrix, x interp represents the spectrum after one-dimensional interpolation, i represents the number of interpolation rounds for each point, round() represents the rounding function, sinc represents the Sinker function,

[0053] S340, using the GPU parallel architecture to set the same number of blocks and threads as in S320, and combine the first matrix and the second matrix updated for the second time to calculate the rotation angle θ i Perform parallel updates to obtain the updated turning angle theta_new; construct a __device__ function on the device side for implementing binary search, and call the __device__ function in the kernel function, so that each thread in the kernel function can perform binary search in parallel to obtain the updated turning angle theta_new. i The neighboring points on the distance are used as index values, and the index values ​​are linearly interpolated in parallel to obtain a new index value ind1; the new index value ind1 and the spectrum s after distance interpolation are combined to obtain a two-dimensional interpolated spectrum ss, with a size of M×N i ×8×2; returns the two-dimensional interpolated spectrum ss.

[0054] It is worth noting that, after the distance interpolation, the interpolation points change nonlinearly, but the change between two adjacent points is small, so the change between two adjacent points can be approximately regarded as linear, and binary search can be implemented by constructing the __device__ function on the device side. After returning the spectrum ss after the two-dimensional interpolation, the unnecessary memory space opened up on the GPU side is released.

[0055] S400, intercepting the range data in the spectrum after the two-dimensional interpolation and downsampling it, then selecting the valid data segment in the azimuth direction, and assigning values ​​thereto to obtain a first echo matrix;

[0056] In a specific implementation of the present invention, S400 includes:

[0057] S410, intercepting part of the data close to the center of the distance wavenumber in the spectrum after the two-dimensional interpolation, and downsampling it to obtain downsampled data;

[0058] S420, selecting valid data segments in azimuth from the downsampled data, and assigning values ​​to them in parallel by means of kernel function indexes to obtain a first echo matrix.

[0059] It is worth noting that by changing φ of formula (3) e In k y =k y0 After Taylor expansion, it can be found that the closer the distance is to the wave number center, the two-dimensional space-variant error φ rd The smaller it is, the smaller the k y0 Part of the data is downsampled, and then the valid data segment in the azimuth is selected to further reduce the amount of matrix data, and parallel assignment is realized through the kernel function index to obtain a new downsampled echo matrix yPol_new, whose memory space size is Nr×Na×8×2 bytes, where Nr represents the number of range points after distance downsampling, and Na represents the number of azimuth points after the valid data segment in the azimuth is intercepted, which is recorded as the first echo matrix.

[0060] S500, inputting the first echo matrix into the GPU parallel processing architecture, and estimating the first stage range non-space-varying error through the GPU parallel processing architecture;

[0061] In a specific embodiment of the present invention, reference Figure 2b , S500 includes:

[0062] S510, using the parallel architecture of the GPU to set a kernel function, then input the first echo matrix into the kernel function, and use the parallel reduction idea to select special points in the first echo matrix through the kernel function, and form the special points into a second echo matrix; at this time, the size of the second echo matrix becomes Nr1×Na, where Nr1 represents the number of special points.

[0063] S520, allocate a data storage area X of Nr1×Na×8×2 bytes and a data storage area Hpha of Na×8×2 bytes in the GPU global memory, wherein the data storage area Hpha is used to store the distance non-space-varying error; interpolate the second echo matrix using the set interpolation multiple k to obtain a third echo matrix, and store it in the data storage area Y; the third echo matrix is ​​expressed as [0(X.*Hpha)0], wherein 0 is an all-zero matrix with Nr1 rows and (k-1)×Na / 2 columns.

[0064] S530, calling the FFT function of the CUFFT library to perform FFT transformation on the third echo matrix along the azimuth dimension to obtain a fourth echo matrix Tmp, and then performing Fourier transformation on the fourth echo matrix Tmp in the azimuth dimension;

[0065] This step calls cufftHandle in the CUFFT library to formulate an FFT plan, sets the parameter nx to the number of azimuth points Na, sets the parameter batch to the number of range points Nr1, and performs Fourier transform on the fourth echo matrix in the azimuth dimension, expressed as Tmp=CUDA_FFT(Y,2).

[0066] S540, using the parallel architecture and parallel reduction concept of GPU to select strong and distinctive points in the fourth echo matrix; performing a circular displacement on the original echo matrix, and moving the peak of the strong and distinctive point to the image center corresponding to the original echo matrix to obtain a fifth echo matrix YImg, the matrix size is Nr1×Na;

[0067] This step selects strong distinctive points based on the same parallel reduction idea as in S320.

[0068] S550, input the fifth echo matrix into the GPU parallel architecture, and calculate the sixth echo matrix YImgnew, which is expressed as

[0069]

[0070] Where, CUDA_IFFT([],2) indicates inverse Fourier transform along the azimuth direction, Win indicates window function, and conj() indicates conjugation.

[0071] S560, using the GPU parallel architecture and in accordance with the parallel reduction concept, setting a block size of Nr1, and a thread size of Na under each block; solving the mean and correlation coefficient for the same distance unit data in each block, and solving the weight in parallel according to the results; using the sixth echo matrix to solve the distance non-space-varying error for the threads in each block in parallel;

[0072] This step uses the GPU parallel architecture. According to the parallel reduction idea, the block size is set to Nr1 and the thread size is set to Na, where Na<=1024. The mean, correlation coefficient, etc. are solved for the same distance unit data in each block, and the weight is solved in parallel based on the results. If Na>1024, it is necessary to iterate and solve the mean, etc. The non-empty error function of the thread parallel distance in each block is expressed as:

[0073]

[0074] In the formula, α represents the weight;

[0075] S570, loop iterate S520-S560 until the set number of loops is reached, and output the distance non-space-varying error of the first stage.

[0076] S600, inputting the non-space-varying distance error of the first stage into the GPU parallel processing architecture, interpolating and solving the two-dimensional space-varying error phase in the GPU parallel processing architecture, and using the two-dimensional space-varying error phase to roughly compensate the echo matrix in S500 to obtain the echo matrix of the current cycle, which is used as the echo matrix input of the next cycle, and repeating the process from S500 to S600 until the number of cycle iterations is reached to obtain the echo matrix after the compensation of the first stage;

[0077] In a specific embodiment of the present invention, reference Figure 2b , S600 includes:

[0078] S610, using the characteristics of GPU multi-stream parallel processing, transfers the first-stage distance non-empty error Hpha estimated by the device side back to the host side through the cudaMemcpyAsync() function. The use of multi-stream data asynchronous copy can improve efficiency.

[0079] S620, calling the cubic spline interpolation function of the gsl library, combining the distance non-space-variant error Hpha of the first stage and the rotation angle information before and after resampling, interpolating the distance non-space-variant residual translation error and the two-dimensional space-variant residual translation error before interception in S400 to obtain the interpolation result, and then transmitting the interpolation result back to the GPU end with the multi-stream parallel function of step S610;

[0080] S630, using the GPU parallel processing architecture to write kernel functions, thread blocks, thread sizes, and construct a residual translation error compensation function H CEr ;

[0081] This step constructs the residual translation error compensation function H according to the following formula CEr :

[0082]

[0083] Where ΔR(arctan(k x / k y )) can be expressed by ΔR(arctan(k x / k y0 )) Interpolation is obtained, φ r0_interp (k y0 ,k x ) is the distance non-space-varying error φ r0 (k y0 ,k x ) is interpolated to obtain the two-dimensional space-varying error phase before downsampling.

[0084] S640, using the GPU parallel processing architecture to write a kernel function, set the same thread block and thread size as S630, and compensate the first echo matrix in echo_cur in parallel for each block of threads to obtain the echo matrix of the current cycle, and use it as the echo matrix input for the next cycle, repeating the process from S500 to S600 until the number of loop iterations is reached to obtain the echo matrix after the first stage compensation.

[0085] S700, inputting the echo matrix compensated in the first stage into the GPU parallel processing architecture, and estimating the range non-space-varying error of the second stage in the GPU parallel processing architecture;

[0086] In a specific implementation of the present invention, S700 includes:

[0087] S710, in the current cycle, using the echo matrix, the number of azimuth points and the number of range points Nr of the current cycle to solve various parameters required for the Newton method optimization solution in parallel;

[0088] In the loop process, the present invention uses the GPU parallel processing architecture, writes the kernel function, sets the thread size to 512, sets the thread block size to (Nr×Na+512-1) / 512, sets the matrix transposition operation to be performed in parallel, and then calls the cufftHandle in the CUFFT library to formulate the FFT plan, sets the parameter nx to the number of azimuth points Na, sets the parameter batch to the number of distance points Nr, and performs Fourier transform on the azimuth dimension. The parameters required by the Newton method are solved in parallel through the kernel function, and the summation function is implemented using the parallel reduction idea described above.

[0089] S720 uses the GPU parallel processing architecture and Newton method to optimize the required parameters, and uses the threads of each block to solve the Newton method gradient descent direction in parallel to obtain the non-space-varying error of the distance in the current cycle;

[0090] S730, determining whether the cycle is finished, if not finished, using the distance non-space-variant error of the current cycle to compensate the first echo matrix to obtain the first echo matrix of the next cycle;

[0091] S740, repeat S710-S730 until the loop ends to obtain the first echo matrix of the end loop;

[0092] S750, inputting the first echo matrix that ends the loop into the GPU parallel processing architecture, and estimating the distance non-space-varying error of the second stage in the GPU parallel processing architecture.

[0093] S800, inputting the distance non-space-variant error of the second stage into the GPU parallel processing architecture, solving the two-dimensional space-variant error phase by parallel interpolation, and using the two-dimensional space-variant error phase to accurately compensate the echo matrix in S700 to obtain the echo matrix after the compensation of the second stage;

[0094] This step is similar to S600 in solving the two-dimensional space-varying error phase φ rd (k y ,k x ) process is the same, and the processing process will not be repeated here.

[0095] S900, repeat the process from S300 to S800 until the number of loop iterations is reached, and obtain a two-dimensional motion-compensated echo matrix, which is subjected to a two-dimensional Fourier transform to obtain a focused image.

[0096] The present invention makes full use of the feature of GPU with multiple computing cores, improves the parallel design of the algorithm, and improves the parallel two-dimensional sinc interpolation method algorithm of polar coordinate resampling in ISAR two-dimensional space-variable translation compensation. The method of performing one-dimensional interpolation on the distance dimension and the azimuth dimension respectively is adopted to reduce loop statements and reduce the correlation between front and back coupling. At the same time, the interpolation point is queried by binary search to improve efficiency, so that the time consumption of polar coordinate resampling is reduced from the original second level to close to the millisecond level.

[0097] Considering the problem that two-dimensional space-varying motion compensation directly uses image minimum entropy self-focusing to estimate non-space-varying errors, it is easy to fall into local optimality, and the compensation is inaccurate and inefficient. The present invention can improve the estimation efficiency and estimation accuracy by downsampling the distance direction, and at the same time adopts a staged method to estimate the non-space-varying error. After the first resampling, the non-space-varying error is roughly estimated and roughly compensated by the phase gradient self-focusing method, and the image minimum entropy self-focusing method is subsequently used for estimation, which can improve the estimation accuracy and efficiency. At the same time, the effective data segment in the azimuth is intercepted to reduce the amount of data and improve the estimation efficiency. The above algorithm is parallelized and improved, and the characteristics of multiple GPU computing cores are used to further improve the efficiency. The total time consumption of two-dimensional space-varying motion compensation is increased from seconds to milliseconds. The present invention adopts a multi-stream parallel processing method to perform multi-stream parallel design on parameter generation and data transmission on the host side and device side, so as to improve the operation processing efficiency.

[0098] The effectiveness of the present invention can be further illustrated by the following simulation.

[0099] Simulation content and result analysis:

[0100] Emulated device: CPU model is Intel i9-10885H, GPU model is NVIDIA RTX 5000with Max-QDesign.

[0101] Simulation experiment: The simulated radar parameters are shown in Table 1, and the simulation results are shown in Tables 2, 3, Figures 4 to 6 .

[0102] Table 1 Simulation radar parameters

[0103]

[0104] Table 2 Comparison of entropy values ​​of imaging results after two-dimensional motion compensation

[0105] Polar coordinate resampling interpolation method 2D sinc interpolation MATLAB image entropy value 7.5396 GPU image entropy value 7.5398

[0106] Table 3 Comparison of two-dimensional self-focus motion compensation time between MATLAB and GPU

[0107]

[0108]

[0109] Figure 3 This is a diagram of the simulated point target scattering model; Figure 4 This is the RD imaging result before two-dimensional motion decoupling after the space-based ISAR echo has been processed by de-slant parametrization and rough translation compensation. It can be found that the general outline of the observed target has been presented, but the solar panels on both sides of the target have obvious widening and rotation, and each point still has obvious defocusing phenomenon; Figure 5 Figure (a) shows the frequency domain result after polar coordinate resampling on the MATLAB side, and Figure (b) shows the frequency domain result after polar coordinate resampling on the GPU side. It can be found that the two are basically consistent. After subtracting the two, the maximum difference is 0.0027, and the average relative error is about 1.6729e-05%, which is basically not much different. Figure 6 Figure (a) shows the RD imaging result after two-dimensional motion compensation on the MATLAB side, and Figure (b) shows the RD imaging result after two-dimensional motion compensation on the GPU side. It can be found that the imaging results are basically consistent, the rotation of the solar panels on both sides of the target is greatly alleviated, and each scattering point is clearer.

[0110] Table 2 is a comparison of the image entropy results after two-dimensional motion compensation on the MATLAB side and the GPU side. The entropy results of the two sides are basically the same. Table 3 is a comparison of the efficiency of two-dimensional self-focusing motion compensation, where the input echo matrix is ​​all complex double type. In the simulation experiment, the input matrix size is 512*1024. The two-dimensional self-focusing motion compensation process consists of 2 polar coordinate resampling, 4 phase gradient self-focusing and 2 image minimum entropy self-focusing. The efficiency of GPU polar coordinate resampling is increased by about 70 times, the efficiency of phase gradient self-focusing is increased by about 7 times, and the efficiency of image minimum entropy self-focusing is increased by about 10 times. The overall process efficiency is increased by about 16 times, from the original second level to the millisecond level.

[0111] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions or substitutions can be made without departing from the concept of the present invention, which should be regarded as falling within the protection scope of the present invention.

Claims

1. A GPU-parallel accelerated method for space-based ISAR self-focusing fast motion compensation. It is characterized in that include: S100, receiving echo data of space targets in a low signal-to-noise ratio environment and space-based radar angle measurement information, converting all echo data into a wave number domain to obtain echo data in the wave number domain, and composing all echo data into an original echo matrix; S200, using a GPU parallel processing architecture to fit the angle information of the target at each pulse moment in parallel; S300, using the GPU parallel processing architecture to parallelly resample the echo matrix of the current iteration in polar coordinates to obtain a two-dimensional interpolated spectrum; S400, intercepting the range data in the spectrum after the two-dimensional interpolation and downsampling it, then selecting the valid data segment in the azimuth direction, and assigning values ​​thereto to obtain a first echo matrix; S500, inputting the first echo matrix into the GPU parallel processing architecture, and estimating the first stage range non-space-varying error in the GPU parallel processing architecture; S600, inputting the non-space-varying distance error of the first stage into the GPU parallel processing architecture, interpolating and solving the two-dimensional space-varying error phase in the GPU parallel processing architecture, and using the two-dimensional space-varying error phase to roughly compensate the echo matrix in S500 to obtain the echo matrix of the current cycle, which is used as the echo matrix input of the next cycle, and repeating the process from S500 to S600 until the number of cycle iterations is reached to obtain the echo matrix after the compensation of the first stage; S700, inputting the echo matrix after the compensation in the first stage into the GPU parallel processing architecture, and estimating the range non-space-varying error in the second stage in the GPU parallel processing architecture; S800, inputting the non-space-varying distance error of the second stage into the GPU parallel processing architecture, solving the two-dimensional space-varying error phase by parallel interpolation, and using the two-dimensional space-varying error phase to accurately compensate the echo matrix in S700 to obtain the echo matrix after the compensation of the second stage, and using it as the echo matrix of the next iteration; S900, repeat the process from S300 to S800 until the number of loop iterations is reached, and obtain a two-dimensional motion-compensated echo matrix, which is subjected to a two-dimensional Fourier transform to obtain a focused image.

2. The space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization according to claim 1, It is characterized in that S200 includes: Four data storage areas are opened in the GPU memory, where the four data storage areas are data storage areas echo_all, M×N×8 bytes in size, i ×8 bytes of data storage area echo_cur, N×8 bytes of data storage area Cta and N i ×8 bytes of data storage area Cta_temp, M is the number of points in the range direction of the echo signal, N is the total number of points in the azimuth direction of the echo data, and the number of points of the angular accumulation pulses required for imaging the i-th ISAR image is N i ; Use the GPU parallel processing architecture to set up (N×8+512-1) / 512 blocks, each block contains at least 512 threads; Run multiple threads in parallel to fit the rotation angle θ at the nth pulse time from the radar angle measurement information using the rotation angle fitting formula n , and store the calculation results in the data storage area Cta; Among them, echo_all stores the echo signals collected within a period of time, echo_cur stores the i-th echo data that currently meets the imaging conditions, Cta stores the angle information of the target within a period of time, and Cta_temp stores the angle information required for the i-th ISAR imaging.

3. The GPU parallel accelerated space-based ISAR self-focusing fast motion compensation method according to claim 1, It is characterized in that S300 includes: S310, constructing the matrix size of the range wave number and the azimuth wave number according to the size of the original echo matrix; constructing the matrix size of the radial wave number matrix k r and the angle θ at the i-th pulse moment i Construct the values ​​of the first matrix of distance wavenumbers and the second matrix of azimuth wavenumbers; allocate M×N matrices of size in the GPU global memory respectively i × 8 bytes of distance to the wave number data storage area kx_new and M × N i ×8 bytes of azimuth wavenumber data storage area ky_new; using the GPU parallel processing architecture, set (M×N i +512-1) / 512 blocks, each block containing at least 512 threads; using multiple threads to update the values ​​of the first matrix and the second matrix in parallel, to obtain the first updated first matrix and the second matrix; S320, using GPU parallel processing architecture, set ((ceil(log 2 M×Ni)) 2 +512-1) / 512 blocks, where ceil means rounding up. Each block contains at least 512 threads. Using the parallel reduction idea, M×N i The data is divided into M×N levels i / 2 groups, each group contains two data sampling points, using the GPU's M×N i / 2 threads parallely calculate the maximum value of each set of data sampling points; for M×N i / 2 extreme values ​​are grouped in pairs, and the solution process of finding the extreme values ​​is iterated cyclically until one group of extreme values ​​remains; the maximum and minimum values ​​of the updated range wavenumber and azimuth wavenumber are solved by using the processing scheme for solving the remaining group of extreme values; the first matrix and the second matrix updated for the first time are updated according to the maximum and minimum values ​​to obtain the first matrix and the second matrix updated for the second time; S330, using the GPU parallel processing architecture to set the same number of blocks and threads as in S320, and combining the first matrix and the second matrix updated for the second time to calculate the radial wave number matrix k r Perform parallel updates; allocate M×N size in GPU global memory i × 8 bytes of data storage ind, and solve in parallel on the new k x -k y The index value ind in the coordinate system is stored in the data storage area ind; the size of M×N is allocated in the GPU global memory i ×8×2 bytes of data storage area 1 ; Using the GPU parallel processing architecture and sinc interpolation principle, set the same number of blocks and threads as in S320, and perform one-dimensional distance interpolation on the original echo matrix in conjunction with the index value ind to obtain the interpolated spectrum s, and store it in the data storage area s 1 ; S340, using the GPU parallel architecture to set the same number of blocks and threads as in S320, and combine the first matrix and the second matrix updated for the second time to calculate the rotation angle θ i Perform parallel updates to obtain the updated turning angle theta_new; construct a __device__ function on the device side for implementing binary search, and call the __device__ function in the kernel function, so that each thread in the kernel function can perform binary search in parallel to obtain the updated turning angle theta_new. i The neighboring points on the distance are used as index values, and the index values ​​are linearly interpolated in parallel to obtain a new index value ind1; the new index value ind1 and the spectrum s after distance interpolation are combined to obtain a two-dimensional interpolated spectrum ss, with a size of M×N i ×8×2; returns the two-dimensional interpolated spectrum ss.

4. The space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization according to claim 1, It is characterized in that S400 includes: S410, intercepting part of the data close to the center of the distance wavenumber in the spectrum after the two-dimensional interpolation, and downsampling it to obtain downsampled data; S420, selecting valid data segments in azimuth from the downsampled data, and assigning values ​​to them in parallel by means of kernel function indexes, to obtain a first echo matrix after new downsampling.

5. The GPU parallel accelerated space-based ISAR self-focusing fast motion compensation method according to claim 4, It is characterized in that S500 includes: S510, using the parallel architecture of the GPU to set a kernel function, then input the first echo matrix into the kernel function, and use the parallel reduction idea to select special points in the first echo matrix through the kernel function, and form a second echo matrix with the special points; S520, the size of the GPU global memory is Nr 1 ×Na×8×2 bytes of data storage area X, and Na×8×2 bytes of data storage area Hpha, wherein the data storage area Hpha is used to store the distance non-space-varying error; interpolating the second echo matrix using the set interpolation multiple k to obtain a third echo matrix, and storing it in the data storage area Y; S530, calling the FFT function of the CUFFT library to perform FFT transformation on the third echo matrix along the azimuth dimension to obtain a fourth echo matrix Tmp, and then performing Fourier transformation on the fourth echo matrix Tmp in the azimuth dimension; S540, using the parallel architecture and parallel reduction idea of ​​GPU to select strong and distinctive points in the fourth echo matrix; making a circular displacement on the original echo matrix, and moving the peak of the strong and distinctive point to the image center corresponding to the original echo matrix to obtain the fifth echo matrix YImg, the matrix size is Nr 1 ×Na; S550, inputting the fifth echo matrix into the GPU parallel architecture to calculate and obtain a sixth echo matrix YImgnew; S560, using GPU parallel architecture and following the parallel reduction idea, sets the size to Nr 1 The thread size under each block is Na; the mean and correlation coefficient are solved for the same distance unit data in each block, and the weight is solved in parallel according to the results; the distance non-space-varying error is solved in parallel for the threads in each block using the sixth echo matrix; S570, loop iterate S520-S560 until the set number of loops is reached, and output the distance non-space-varying error Hpha of the first stage.

6. The space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization according to claim 1, It is characterized in that S600 includes: S610, using the multi-stream parallel processing feature of GPU, transmits the first-stage distance non-empty-variant error Hpha estimated by the device side back to the host side through the cudaMemcpyAsync() function; S620, calling the cubic spline interpolation function of the gsl library, combining the distance non-space-variant error Hpha of the first stage and the rotation angle information before and after resampling, interpolating the distance non-space-variant residual translation error and the two-dimensional space-variant residual translation error before interception in S400 to obtain an interpolation result, and then transmitting the interpolation result back to the GPU end; S630, using the GPU parallel processing architecture to write kernel functions, thread blocks, thread sizes, and construct a residual translation error compensation function H CEr : S640, using the GPU parallel processing architecture to write a kernel function, set the same thread block and thread size as S630, and compensate the first echo matrix in echo_cur in parallel for each block of threads to obtain the echo matrix of the current cycle, and use it as the echo matrix input for the next cycle, repeating the process from S500 to S600 until the number of loop iterations is reached to obtain the echo matrix after the first stage compensation.

7. The space-based ISAR self-focusing fast motion compensation method accelerated by GPU parallelization according to claim 1, It is characterized in that S700 includes: S710, in the current cycle, using the echo matrix, the number of azimuth points and the number of range points Nr of the current cycle to solve various parameters required for the Newton method optimization solution in parallel; S720 uses the GPU parallel processing architecture and Newton method to optimize the required parameters, and uses the threads of each block to solve the Newton method gradient descent direction in parallel to obtain the non-space-varying error of the distance in the current cycle; S730, determining whether the cycle is finished, if not finished, using the distance non-space-variant error of the current cycle to compensate the first echo matrix to obtain the first echo matrix of the next cycle; S740, repeat S710-S730 until the loop ends to obtain the first echo matrix of the end loop; S750, inputting the first echo matrix that ends the loop into the GPU parallel processing architecture, and estimating the distance non-space-varying error of the second stage in the GPU parallel processing architecture.

Citation Information

Patent Citations

  • Missile-borne SAR forward-squint imaging method based on GPU

    CN104459693A

  • GPU-based parallel structure ISAR range alignment method

    CN106291551A

  • Related motion error compensation method based on BiSAR echoing, system and application

    CN111443349A

  • Envelope alignment rapid implementation method for GPU (Graphics Processing Unit) parallel acceleration

    CN112991140A

  • Missile-borne SAR front side view time domain imaging method based on embedded GPU

    CN113156435A

Cited By

  • Space target space-based microwave imaging high-order space-variant motion error compensation method and device

    CN122218706A

  • Space target space-based microwave imaging high-order space-variant motion error compensation method and device

    CN122218706B