Distributed strong scattering point scene SAR image self-focusing method based on improved PGA

Abstract

The invention discloses a distributed strong scattering point scene SAR image self-focusing method based on an improved PGA, and the method comprises the steps: firstly, selecting candidate scattering points for phase error estimation from a defocused distributed strong scattering point scene SAR image; then, screening out high-quality isolated scattering points according to the fluctuation degree of the signal intensity of the distance compressed domain corresponding to the candidate scattering points; thirdly, applying an improved PGA to the screened scattering points to estimate an azimuth phase error; and finally, compensating an azimuth phase error in the defocused image to obtain an SAR image after self-focusing processing. According to the method, the influence of mutually close strong scattering points in a distributed strong scattering point scene on phase error estimation is overcome, and the SAR imaging effect is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of synthetic aperture radar image processing technology, and more particularly to a distributed strong scattering point scene SAR image autofocusing method based on an improved PGA method, which is denoted as the MOPGA method. Background Art

[0002] Synthetic Aperture Radar (SAR) is a coherent imaging radar system with high resolution. The motion error of the SAR platform and any non-uniformity in the radar echo propagation path will cause random changes in the echo phase, resulting in defocus of the SAR image.

[0003] In the book Radar Imaging Technology, edited by Liu Yongtan and published by Harbin Institute of Technology Press in October 1999, it is pointed out that Synthetic Aperture Radar (SAR) is installed on a moving platform and transmits and receives pulses at a certain repetition frequency to form echo signals. The structural block diagram of the SAR system is shown in the figure below. Figure 1 As shown, the SAR system consists of an onboard radar system, a satellite platform and data downlink system, and a ground system. Synthetic aperture radar imaging processing is performed in the ground system. The ground system receives echo signals from the satellite platform and data downlink system via a ground receiving station. These echo signals are processed by a SAR signal processor to produce SAR images, which are then stored in a backup file operation system.

[0004] As SAR system resolution increases, the requirements for SAR platform motion compensation accuracy are becoming increasingly stringent. Autofocus algorithms that estimate phase errors using the SAR data themselves are becoming indispensable. SAR image autofocus algorithms can be categorized into two types: those based on parametric models and those based on non-parametric models. The former, exemplified by the Map Drift Algorithm (MDA) and the Contrast Optimization Autofocusing (COA) algorithms, are limited by model accuracy and cannot correct for high-order phase errors, typically failing to meet the requirements of high-resolution imaging. The latter, exemplified by the Phase Gradient Autofocusing (PGA) algorithm, achieves image autofocus by estimating phase gradients and compensating for phase errors. PGA algorithms are capable of correcting high-order phase errors. However, in scenarios with distributed strong scattering points, where multiple strong scattering points exist within a limited azimuth range, the PGA algorithm selects strong scattering points based on their amplitude and applies a windowing function, which causes adjacent scattering points to be included in the window function. This severely impacts the accuracy of phase error estimation and leads to defocused SAR images. Summary of the Invention

[0005] To minimize the impact of azimuth phase error and resulting defocusing of SAR images output by the SAR processor, this paper proposes a SAR image autofocusing method for distributed strong scattering point scenes based on an improved PGA. This method first selects candidate scattering points for phase error estimation, then screens out high-quality isolated scattering points based on the fluctuation of the signal intensity in the range compression domain corresponding to the candidate scattering points. Phase error is then estimated using the MOPGA on these selected scattering points, thus overcoming the influence of adjacent scattering points in azimuth on the phase error estimation.

[0006] The technical solution of the present invention is:

[0007] Step 1: intercept candidate scattering points in the defocused SAR image according to the window function;

[0008] Construct a window function in each range gate of the defocused distributed strong scattering point scene SAR image, and intercept multiple candidate scattering points according to the amplitude as samples for phase error estimation, that is, the candidate scattering point arrangement matrix S sample ;

[0009] Step 2: transform to the distance compression domain to eliminate candidate scattering points above the threshold;

[0010] The candidate scattering points are transformed into the range compression domain, the fluctuation of the range compression domain signal intensity corresponding to the candidate scattering points is calculated, and the candidate scattering points above the threshold are removed from the sample to obtain the high-quality scattering point SR sample ;

[0011] Step 3: Use the MOPGA method to estimate the azimuth phase error;

[0012] The MOPGA algorithm is used to estimate the phase error of high-quality scattering points. During the estimation process, the phase curvature is calculated and the phase error function is obtained by summing it twice. This process removes the constant term and linear component in the phase function to obtain the phase error estimation result SS.

[0013] Step 4, iterative phase error estimation and phase error compensation;

[0014] Iterate SS using steps 1 to 3, and record the phase error estimation result of the σth iteration as SS σ , until SS σ When convergence or the maximum number of iterations is reached, the SAR image processed by MOPGA is output. That is, the phase error estimation accuracy is improved through iteration, and the phase error is compensated to obtain the SAR image processed by autofocus.

[0015] The application of the MOPGA method in SAR image autofocusing of distributed strong scattering point scenes has the following advantages:

[0016] (1) By selecting multiple candidate scattering points in the same range gate, the potential high-quality scattering points in each range gate are fully utilized, thereby improving the phase error estimation accuracy.

[0017] (2) By removing candidate scattering points whose fluctuation of range compression domain signal intensity is higher than the threshold in the sample, candidate scattering points with interference from adjacent scattering points are excluded, thereby improving the accuracy of phase error estimation.

[0018] (3) The calculation of phase curvature by formula (5) to formula (8) removes the constant phase and linear phase, avoiding the circular shift operation in the standard PGA. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 This is the structural block diagram of the SAR system.

[0020] Figure 2 This is a flow chart of the distributed strong scattering point scene SAR image self-focusing method based on the improved PGA of the present invention.

[0021] Figure 3 It is a comparison diagram of the range compression domain signal amplitude corresponding to adjacent strong scattering points and isolated scattering points.

[0022] Figure 4 Schematic diagram of scattering point selection during PGA processing.

[0023] Figure 5 This is a comparison diagram of the azimuthal profiles of isolated scattering points in the imaging results of the PGA and MOPGA algorithms.

[0024] Figure 6 This is the imaging result of PGA processing.

[0025] Figure 7 This is a diagram of the imaging result of the MOPGA processing of the present invention.

[0026] Figure 8 This is a diagram of the highlighted points selected for imaging quality evaluation.

[0027] Figure 9 It is a cross-sectional view of the highlighted points in the PGA processing results.

[0028] Figure 10 It is a cross-sectional view of a highlighted point in the MOPGA processing result of the present invention. DETAILED DESCRIPTION

[0029] The present invention will be further described in detail below with reference to the accompanying drawings and examples. The parameters listed are merely exemplary embodiments of the present invention. It should be noted that those skilled in the art may make various improvements and modifications without departing from the principles of the present invention, and such improvements and modifications are also within the scope of protection of the present invention.

[0030] In the present invention, the improved PGA (Phase Gradient Autofocus) method is referred to as the MOPGA method.

[0031] To address the degradation of phase error estimation accuracy in standard PGA due to the influence of adjacent strong scattering points in distributed strong scattering point scenarios, this paper proposes an autofocusing method for SAR images in distributed strong scattering point scenarios based on an improved PGA. Specifically, the MOPGA method is applied to autofocus SAR images in distributed strong scattering point scenarios. This MOPGA method first selects candidate scattering points for phase error estimation, then screens out high-quality isolated scattering points based on the fluctuation of the signal intensity in the range compression domain corresponding to the candidate scattering points. Finally, MOPGA is applied to these screened scattering points to estimate phase error, thereby overcoming the influence of adjacent scattering points in azimuth on phase error estimation.

[0032] See also Figure 1 As shown, the MOPGA method of the present invention is stored in the 1B level signal processor of the SAR system. The SAR image information with distributed strong scattering points in the 1B level signal processor of the SAR system is recorded as S, that is, the defocused pre-processed image.

[0033] See also Figure 2 As shown, the distributed strong scattering point scene SAR image autofocusing method based on the improved PGA of the present invention includes the following processing steps:

[0034] Step 1: intercept candidate scattering points in the defocused SAR image according to the window function;

[0035] In the present invention, N candidate scattering points are cut out according to amplitude in each range gate of the defocus pre-processed image S according to a window function as samples for phase error estimation.

[0036] In the present invention, the matrix form of the defocus pre-processed image S is expressed as:

[0037]

[0038] N a Indicates the number of azimuth points;

[0039] N r Indicates the distance to the point number;

[0040] n is a variable whose value range is from 1 to N a , indicating the direction to the nth point;

[0041] m is a variable whose value range is from 1 to N r , indicating the distance to the mth point;

[0042] a 1,1 Indicates the distance from the point in row 1 to column 1 in the image;

[0043] a 1,2 Indicates the distance from the point in row 1 to column 2 in the image;

[0044] a 1,m Indicates the distance from the 1st row to the mth column of the image;

[0045] Indicates the distance from the first row to the Nth row in the image. r Points of the column;

[0046] a 2,1 Indicates the distance from the point in the 2nd row to the 1st column in the image;

[0047] a 2,2 Indicates the distance from the point in the 2nd row to the 2nd column in the image;

[0048] a 2,m Indicates the distance from the 2nd row to the mth column in the image;

[0049] Indicates the distance from the 2nd row to the Nth row in the image. r Points of the column;

[0050] a n,1 Indicates the distance from the nth row to the first column of the image;

[0051] a n,2 Indicates the distance from the nth row to the second column in the image;

[0052] a n,m Indicates the point in the image whose orientation is in the nth row and whose distance is in the mth column;

[0053] Indicates the distance from the nth row to the Nth row in the image. r Points of the column;

[0054] Indicates the Nth direction in the image a The distance of the row to the point in column 1;

[0055] Indicates the Nth direction in the image a The distance of the row to the point in column 2;

[0056] Indicates the Nth direction in the image a The distance of the row to the point in the mth column;

[0057] Indicates the Nth direction in the image a The distance of the row to the Nth r Column points.

[0058] In the present invention, the window function is set to:

[0059]

[0060] n is a variable whose value range is from 1 to N a , indicating the direction to the nth point;

[0061] ω p,q Represents the window function of the qth candidate scattering point in the pth column of the intercepted image;

[0062] l p,q represents the azimuth coordinate of the qth candidate scattering point in the pth column of the range in the intercepted image;

[0063] L represents the length of the window function.

[0064] In the present invention, the product of the azimuth signal of the defocus pre-processed image S and the corresponding window function is denoted as S p,q :

[0065]

[0066] Indicates the orientation of the captured image. The distance of the row to the point in the pth column.

[0067] Indicates the orientation of the captured image. The distance of the row to the point in the pth column.

[0068] N represents the total number of candidate scattering points.

[0069] In the present invention, formula (1) is used to extract an image from the defocused pre-processed image S, which is recorded as the image to be processed SS. The SS is a part of the S. The scattering points appearing in the SS are called candidate scattering points.

[0070] In the present invention, formula (2) is used to extract all candidate scattering points from SS, which are recorded as MSS, and MSS = {ms1, ms2, ..., ms m ,…,msM}, the subscript m is the identification number of the candidate scattering point, and the subscript M is the total number of candidate scattering points.

[0071] ms1 represents the first candidate scattering point in the captured image.

[0072] ms2 represents the second candidate scattering point in the captured image.

[0073] ms m represents the mth candidate scattering point in the captured image.

[0074] ms M Represents the last candidate scattering point in the captured image.

[0075] In the present invention, MSS={ms1,ms2,…,ms m ,…,ms M Each candidate scattering point in} is arranged column by column to form a candidate scattering point arrangement matrix, denoted as S sample ,and The S sample The samples that will be used to estimate the phase error.

[0076] Step 2: transform to the distance compression domain to eliminate candidate scattering points above the threshold;

[0077] In the present invention, the candidate scattering points are transformed into the range compression domain, and the range compression domain signal after range migration correction is used to represent the candidate scattering points. Then, the range compression domain signal after range migration correction is:

[0078]

[0079] s rc is the range compression domain signal after range migration correction;

[0080] A0 is a complex constant;

[0081] p r is the range compression pulse envelope;

[0082] τ is the distance-to-fast time;

[0083] R0 is the minimum distance between the radar and the target;

[0084] c is the speed of light;

[0085] w a is the antenna pattern in azimuth;

[0086] η is the azimuthal slow time;

[0087] η c is the beam center crossing time;

[0088] j is the imaginary part;

[0089] π is the ratio of the circumference of a circle to the circumference of a circle, and its value is 3.

[0090] f0 is the radar center frequency;

[0091] R is the distance between the radar and the target.

[0092] In the present invention, the fluctuation of the signal strength in the range compression domain is v:

[0093]

[0094] T is the synthetic aperture time.

[0095] s rc is the range compression domain signal after range migration correction;

[0096] d is the differential symbol.

[0097] η is the azimuthal slow time.

[0098] Step 21, performing distance compression domain processing;

[0099] In the present invention, formula (3) is used to The signal corresponding to the scattering point transformed into the range compression domain is recorded as RSS, and

[0100] rs1 represents the signal in the range compression domain corresponding to the candidate scattering point ms1.

[0101] rs2 represents the signal in the range compression domain corresponding to the candidate scattering point ms2.

[0102] rs m Represents the candidate scattering point ms m The corresponding signal in the range compression domain.

[0103] rs M Represents the candidate scattering point ms M The corresponding signal in the range compression domain.

[0104] Step 22, calculating the fluctuation of the signal strength in the range compression domain;

[0105] In the present invention, formula (4) is used to calculate The fluctuation of the signal strength in the compression domain at each distance is recorded as v_RSS, and

[0106] Indicates the fluctuation of the range compression domain signal rs1.

[0107] Indicates the fluctuation of the range compression domain signal rs2.

[0108] Represents the range compression domain signal rs m volatility.

[0109] Represents the range compression domain signal rs M volatility.

[0110] Step 23, comparison of volatility;

[0111] Set the volatility threshold to v in .

[0112] In the working environment, the motion error of the SAR system's platform and any non-uniformity in the radar echo propagation path will cause random changes in the echo phase, resulting in defocusing of the SAR image. This scenario is called a distributed strong scattering point scenario. In a distributed strong scattering point scenario, the signal intensity of the scattering point in the range compression domain varies greatly due to interference from adjacent scattering points and clutter, while the signal intensity of the high-quality scattering point in the range compression domain varies little, such as Figure 3 The fluctuation is actually the variance normalized by the average power to reflect the degree of signal strength change. Therefore, samples with fluctuations higher than the threshold are removed to screen out high-quality scattering points. In the present invention, the fluctuation of the high-quality scattering points from the compressed domain signal strength is generally less than 0.1 (v in <0.1).

[0113] Compare With v in ,like Removal The corresponding candidate scattering point; if reserve The corresponding candidate scattering points are recorded as Then record the Corresponding candidate scattering point ms m .

[0114] Similarly, comparison With v in ,like Removal The corresponding candidate scattering point; if reserve The corresponding candidate scattering points are recorded as Then record the The corresponding candidate scattering point ms1.

[0115] Similarly, comparison With v in ,like Removal The corresponding candidate scattering point; if reserve The corresponding candidate scattering points are recorded as Then record the The corresponding candidate scattering point ms2.

[0116] Similarly, comparison With v in ,like Removal The corresponding candidate scattering point; if reserve The corresponding candidate scattering points are recorded as Then record the Corresponding candidate scattering point ms M .

[0117] Longitudinal fluctuation threshold v in After comparison, the scattering point corresponding to the retained range compression domain signal is obtained and recorded as the high-quality scattering point SR sample ,and

[0118] Then based on The retained candidate scattering points can be obtained.

[0119] In the present invention, the value higher than the fluctuation threshold v in The candidate scattering points corresponding to the range compression domain signal intensity are removed from the sample, which is beneficial to improving the phase error accuracy.

[0120] In the present invention, the process of PGA in processing the defocused image of the scene is analyzed. According to the implementation steps of PGA, the scattering points for phase error estimation are first selected according to the amplitude, and the azimuth signals of the range gates where the three scattering points are located are plotted, as shown in FIG. Figure 4 As shown in Figure 2, two peaks can be seen in the azimuth signal, corresponding to the yellow box area and the red box area in the defocused image. The yellow box represents adjacent strong scattering points, while the red box represents isolated scattering points with weaker intensity.

[0121] Since PGA directly selects the scattering points for phase error estimation based on the amplitude, the signal of the yellow dotted part will be windowed and the center will be shifted to the center of the aperture for phase error estimation during the execution of PGA. The selected scattering points are adjacent strong scattering points, which will affect the accuracy of phase error estimation. The present invention uses the MOPGA algorithm to select the scattering points for phase error estimation. The fluctuation degree defined by formula (4) is used to screen out isolated high-quality scatterers for phase error estimation to improve the accuracy of phase error estimation when processing distributed strong scattering point scenes. The improved algorithm based on PGA of the present invention first selects the scattering points for phase error estimation during the execution of processing defocused images. Figure 4 The yellow and red boxes in the middle are first selected as candidate scattering points for phase error estimation, and then transformed into the range compression domain after windowing. The range compression and signal amplitude corresponding to adjacent strong scattering points and isolated scattering points are as follows Figure 3 As shown. Figure 3 The results show that the range compression domain signal corresponding to the adjacent strong scattering points has a higher instantaneous amplitude, but a large fluctuation, while the isolated scattering point has a smaller instantaneous amplitude, but a small fluctuation. According to formula (4), the fluctuation of the range compression domain signal corresponding to them is calculated, and the fluctuation threshold condition is used. Finally, the remaining ones are Figure 4 The scattering points in the red box are isolated high-quality scattering points and are used for phase error estimation.

[0122] Step 3: Use the MOPGA method to estimate the azimuth phase error;

[0123] In the present invention, the improved PGA (Phase Gradient Autofocus) method is referred to as the MOPGA method.

[0124] The difference between the MOPGA method of the present invention and the traditional PGA method is that the phase error is calculated by curvature and two cumulative summations without circular shift.

[0125] The PGA method is referenced in "Principles of Synthetic Aperture Radar Imaging," by Pi Yiming et al., Chengdu: University of Electronic Science and Technology Press, March 2007, pages 115-120. In the PGA method flowchart on page 119, the present invention's method does not perform "circular shifting" because the estimated phase curvature of the present invention does not require circular shifting to remove linear phase errors.

[0126] For high-quality scattering points SR sample The MOPGA method is used to estimate the azimuth phase error and improve the estimation accuracy.

[0127] In the present invention, the phase of the range compression domain signal after expansion is for:

[0128]

[0129] x is the discrete position time.

[0130] unwrap is the phase unwrapping operator.

[0131] Arg is the argument principal value operator.

[0132] s rc is the distance compression domain signal.

[0133] In the present invention, the phase curvature after the range compression domain signal is expanded is for:

[0134]

[0135] In the present invention, the coherent average of the phase curvature after the range compression domain signal is expanded is for:

[0136]

[0137] K is the total number of samples in SRSS.

[0138] k is any sample in SRSS.

[0139] is the phase curvature of the expanded range compression domain signal corresponding to the kth sample.

[0140] Since x is the discretized azimuthal slow time, it is necessary to calculate the curvature of each phase error sample and perform coherent averaging, and then perform two cumulative summations on the average curvature to obtain the phase error estimation result.

[0141] In the present invention, the estimation result of the phase error for:

[0142]

[0143] x is the discrete position time.

[0144] I is the serial number of the outer summation symbol.

[0145] i is the serial number of the inner summation symbol.

[0146] is the coherent average of the phase curvature of the i-th range compression domain signal after expansion in the inner layer.

[0147] In this invention, equations (5) through (8) are used to compensate for the circular shift that occurs in conventional PGA methods. However, the present invention differs in that the phase error is calculated using curvature and two cumulative sums without requiring circular shift. The unwarped phase of the range-compression domain signal corresponding to the high-quality scattering point is calculated, and the curvature of the phase error is calculated using the central difference formula.

[0148] Step 31, phase unwrapping processing;

[0149] In the MOPGA estimation process, first use formula (5) to calculate The unwrapped phase of each range compression domain signal in is denoted as SW. This step is used to solve the periodic jump problem of the phase of the range compression domain signal during processing, ensuring the continuity of the phase and the accuracy of subsequent analysis.

[0150] Step 32, calculating the phase curvature;

[0151] The curvature of the phase error is calculated using formula (6) for SW, which is denoted as SU 。

[0152] In this invention, constant phase error and linear phase error do not affect the focusing quality of SAR images. Compared with calculating the phase gradient in PGA, calculating the phase curvature removes both the constant phase and the linear phase. Furthermore, formula (6) estimates the phase curvature at the center point using the central difference method, utilizing the values at multiple discrete azimuth time points on both sides of the center point, improving estimation accuracy and noise immunity.

[0153] Step 33, calculating the average phase curvature;

[0154] The average curvature of the phase error samples of SU is calculated using formula (7), which is recorded as ST 。

[0155] In this invention, the phase curvature of a single scattering point may be affected by noise or local interference (such as multipath), resulting in outliers. Directly using this value can lead to inaccurate phase error estimates. Calculating the average phase curvature essentially eliminates the impact of some outliers on the phase error estimate through statistical synergy across multiple scattering points. This step ensures the accuracy of the phase error estimate and the algorithm's noise immunity.

[0156] Step 34, calculating the phase error;

[0157] Using formula (8), the average curvature of the phase error samples of ST is accumulated twice to obtain the phase error estimation result, which is recorded as SS 。

[0158] Step 4, iterative phase error estimation and phase error compensation;

[0159] In the present invention, the phase error threshold is denoted as ζ in The number of iterations is denoted as σ. The maximum number of iterations is set to 10.

[0160] In the present invention, in order to conveniently record the phase error estimation results and the number of iterations, the phase error estimation result of the σth iteration is recorded as SS σ The SS σ Also called the current iteration result SS σ , will be located in SS σThe previous one is called the previous iteration result SS σ-1 During the iteration process, the previous iteration result SS σ-1 As the result of the current iteration SS σ input.

[0161] Iterate SS using steps 1 to 3, and record the phase error estimation result of the σth iteration as SS σ , until SS σ When convergence or the maximum number of iterations is reached, the SAR image processed by MOPGA is output.

[0162] During the iteration process, if the SS of the current iteration σ is less than or equal to ζ in (||SS σ ||≤ζ in ), the iteration ends and the SAR image processed by MOPGA is output.

[0163] In the present invention, the SAR image S is obtained by processing the defocused pre-processed image S with MOPGA. MOPGA , S MOPGA The focus quality is better compared to the S.

[0164] The imaging results of the MOPGA algorithm of the present invention show that two strong scattering points in the adjacent strong scattering point area can be distinguished from the image, and the weaker isolated scattering points are also correctly focused. The azimuth profile of the weaker isolated scattering point in the simulation scene is as follows: Figure 5 As shown in the figure, the azimuth cross-section shows that the focusing quality of the MOPGA is significantly better than that of the PGA. PSLR, ISLR, and Res were further calculated to evaluate the imaging quality, and the results are shown in Table 1.

[0165] Table 1 Simulation data imaging quality evaluation

[0166] Azimuth Algorithm used PSLR(dB) ISLR(dB) Res(m) PGA -3.57 0.54 0.15 MOPGA -12.30 -8.46 0.15

[0167] In azimuth, the PSLR (Peak Sidelobe Ratio), ISLR (Integrated Sidelobe Ratio), and Res (Resolution) metrics of the MOPGA algorithm compared to those of the PGA algorithm. In range, the focus quality of the MOPGA and PGA algorithms is nearly identical. Therefore, the MOPGA algorithm can effectively improve the accuracy of phase error estimation and enhance image focus quality when processing scenes with distributed strong scattering points.

[0168] Example 1

[0169] In order to verify the effectiveness of the SAR image autofocusing method for distributed strong scattering point scenes based on the improved PGA of the present invention, experimental verification is carried out using airborne SAR measured data. The parameters of the measured airborne SAR data are shown in Table 2.

[0170] Table 2 Measured data airborne SAR parameters

[0171] parameter Numerical Radar band Ka Radar carrier frequency (GHz) 35 Pulse repetition frequency (Hz) 4000 Transmit pulse duration (μs) 25 Signal bandwidth (MHz) 420 Flight altitude (km) 3.11 Aircraft flight speed (m / s) 90.5 Slant distance from scene center (km) 26.5 Squint angle (°) 0

[0172] Using PGA and MOPGA algorithms for processing, the imaging results are as follows Figure 6 and Figure 7 As shown in the figure, due to the presence of a large number of strong scattering points close to each other in the scene, PGA is affected by the adjacent strong scattering points when performing phase error estimation, and the phase error estimation accuracy is significantly reduced. However, the method of the present invention reduces the influence of adjacent scattering points in the azimuth direction on the phase error estimation, achieving a better focusing effect than PGA.

[0173] Then, point targets are selected to calculate their RES, PSLR and ISLR to evaluate the imaging quality. The selected point targets are as follows: Figure 8 As shown, the circled points are T1, T2 and T3 from left to right. The azimuth and range profiles of the PGA processing results are shown in Figure 9 As shown in the figure, the azimuth and range profiles of the MOPGA algorithm processing results are shown in the figure. Figure 10 shown. Figure 9 and Figure 10 (a) T1 azimuth profile, (b) T2 azimuth profile, (c) T3 azimuth profile, (d) T1 range profile, (e) T2 range profile, and (f) T3 range profile. The RES, PSLR, and ISLR of point targets are shown in Table 3.

[0174] Table 3. Imaging quality evaluation index of point targets in processing results

[0175] Azimuth Distance Target algorithm RES(m) PSLR(dB) ISLR(dB) RES(m) PSLR(dB) ISLR(dB) T1 PGA 0.21 -4.15 4.19 0.45 -9.30 -6.13 T1 MOPGA 0.17 -6.47 -6.94 0.37 -9.41 -9.96 T2 PGA 0.21 -4.34 3.84 0.48 -15.37 -5.39 T2 MOPGA 0.17 -11.54 -9.94 0.35 -15.63 -9.48 T3 PGA 0.18 -5.35 3.65 0.47 -13.07 -7.48 T3 MOPGA 0.15 -8.35 -10.27 0.35 -13.16 -10.08

[0176] from Figure 8 、 Figure 9 、 Figure 10 From the results in Table 3, it can be seen that the RES of the MOPGA algorithm in azimuth and range directions are improved by about 0.4m and 0.12m respectively. The PSLR and ISLR of the MOPGA algorithm in azimuth direction are lowered by 3dB and 10dB respectively compared with PGA. The PSLR of the MOPGA algorithm in range direction is close to that of PGA, and the ISLR is lowered by 3dB. Therefore, the MOPGA algorithm is effective, and its focusing quality is significantly better than that of PGA.

[0177] Aiming at the self-focusing of imaging in scenes with distributed strong scattering points, the present invention proposes a phase error correction method based on PGA. The algorithm first selects candidate targets based on pixel amplitude in the pre-processed defocused image, then calculates the fluctuation of the range compression domain signal corresponding to the candidate targets, and then screens out high-quality isolated scattering points based on the threshold condition. The improved PGA is then used to estimate the phase error of the screened high-quality isolated scattering points, and finally the phase error is compensated to obtain a well-focused image. This method can overcome the problem that when PGA processes scenes with distributed strong scattering points, the phase error estimation sample selected based on amplitude will be affected by adjacent scatterers, thereby reducing the accuracy of phase error estimation, and the problem that the scattering points selected in the iterative process are inconsistent, resulting in difficulty in convergence. Simulation experiments and the processing of measured data have verified the effectiveness of the method proposed in the present invention.

Claims

1. A SAR image autofocusing method for scenes with distributed strong scattering points based on an improved PGA, wherein a defocused preprocessed image S with distributed strong scattering points exists in a level 1B signal processor of a SAR system; The following steps are included: Step 1: intercept candidate scattering points in the defocused SAR image according to the window function; Construct a window function in each range gate of the defocused distributed strong scattering point scene SAR image, and intercept multiple candidate scattering points according to the amplitude as samples for phase error estimation, that is, the candidate scattering point arrangement matrix S sample ; Step 2: transform to the distance compression domain to eliminate candidate scattering points above the threshold; The candidate scattering points are transformed into the range compression domain, the fluctuation of the range compression domain signal intensity corresponding to the candidate scattering points is calculated, and the candidate scattering points above the threshold are removed from the sample to obtain the high-quality scattering point SR sample ; Step 3: Use the MOPGA method to estimate the azimuth phase error; The MOPGA algorithm is used to estimate the phase error of high-quality scattering points. During the estimation process, the phase curvature is calculated and the phase error function is obtained by summing it twice. This process removes the constant term and linear component in the phase function to obtain the phase error estimation result SS. Step 4, iterative phase error estimation and phase error compensation; Iterate SS using steps 1 to 3, and record the phase error estimation result of the σth iteration as SS σ , until SS σ When convergence or the maximum number of iterations is reached, the SAR image processed by MOPGA is output.

2. The distributed strong scattering point scene SAR image autofocusing method based on improved PGA according to claim 1 is characterized in that In step 1: N candidate scattering points are intercepted according to the amplitude in each range gate of the defocused pre-processed image S according to the window function as samples for phase error estimation; In the present invention, the matrix form of the defocus pre-processed image S is expressed as: N a Indicates the number of azimuth points; N r Indicates the distance to the point number; n is a variable whose value range is from 1 to N a , indicating the direction to the nth point; m is a variable whose value range is from 1 to N r , indicating the distance to the mth point; a 1,1 Indicates the distance from the point in row 1 to column 1 in the image; a 1,2 Indicates the distance from the point in row 1 to column 2 in the image; a 1,m Indicates the distance from the 1st row to the mth column of the image; Indicates the distance from the first row to the Nth row in the image. r Points of the column; a 2,1 Indicates the distance from the point in the 2nd row to the 1st column in the image; a 2,2 Indicates the distance from the point in the 2nd row to the 2nd column in the image; a 2,m Indicates the distance from the 2nd row to the mth column in the image; Indicates the distance from the 2nd row to the Nth row in the image. r Column points; a n,1 Indicates the distance from the nth row to the first column of the image; a n,2 Indicates the distance from the nth row to the second column in the image; a n,m Indicates the point in the image whose orientation is in the nth row and whose distance is in the mth column; Indicates the distance from the nth row to the Nth row in the image. r Points of the column; Indicates the Nth direction in the image a The distance of the row to the point in column 1; Indicates the Nth direction in the image a The distance of the row to the point in column 2; Indicates the Nth direction in the image a The distance of the row to the point in the mth column; Indicates the Nth direction in the image a The distance of the row to the Nth r Points of the column; Set the window function to: n is a variable whose value range is from 1 to N a , indicating the direction to the nth point; ω p,q represents the window function of the qth candidate scattering point in the pth column of the intercepted image; l p,q represents the azimuth coordinate of the qth candidate scattering point in the pth column of the intercepted image; L represents the length of the window function; The product of the azimuth signal of S and the corresponding window function is: Indicates the orientation of the captured image. The distance of the row to the point in the pth column; Indicates the orientation of the captured image. The distance of the row to the point in the pth column; N represents the total number of candidate scattering points; Formula (2) is used to extract all candidate scattering points from SS, which are recorded as MSS, and MSS = {ms1,ms2,…,ms m ,…,ms M }, subscript m is the identification number of the candidate scattering point, and subscript M is the total number of candidate scattering points; ms1 represents the first candidate scattering point in the intercepted image; ms2 represents the second candidate scattering point in the intercepted image; ms m represents the mth candidate scattering point in the intercepted image; ms M Indicates the last candidate scattering point in the intercepted image; Set MSS = {ms1,ms2,…,ms m ,…,ms M Each candidate scattering point in} is arranged column by column to form a candidate scattering point arrangement matrix, denoted as S sample ,and The S sample The samples that will be used to estimate the phase error.

3. The distributed strong scattering point scene SAR image autofocusing method based on improved PGA according to any one of claims 1 to 2, characterized in that In step 2: Step 21, performing distance compression domain processing; Using formula (3) The signal corresponding to the scattering point transformed into the range compression domain is recorded as RSS, and rs1 represents the signal in the range compression domain corresponding to the candidate scattering point ms1; rs2 represents the signal in the range compression domain corresponding to the candidate scattering point ms2; rs m Represents the candidate scattering point ms m The corresponding signal in the range compression domain; rs M Represents the candidate scattering point ms M The corresponding signal in the range compression domain; The candidate scattering points are transformed into the range compression domain and represented by the range compression domain signal after range migration correction. Then the range compression domain signal after range migration correction is: s rc is the range compression domain signal after range migration correction; A0 is a complex constant; p r is the range compression pulse envelope; τ is the distance-to-fast time; R0 is the minimum distance between the radar and the target; c is the speed of light; w a is the antenna pattern in azimuth; η is the azimuthal slow time; η c is the beam center crossing time; j is the imaginary part; π is the ratio of the circumference of a circle to the circumference of a circle, and its value is 3. f0 is the radar center frequency; R is the distance between the radar and the target; Step 22, calculating the fluctuation of the signal strength in the range compression domain; Calculate using formula (4) The fluctuation of the signal strength in the compression domain at each distance is recorded as v_RSS, and Indicates the fluctuation of the range compression domain signal rs1; Indicates the fluctuation of the range compression domain signal rs2; Represents the range compression domain signal rs m Volatility; Represents the range compression domain signal rs M Volatility; The fluctuation of the signal strength in the range compression domain is v: T is the synthetic aperture time; s rc is the range compression domain signal after range migration correction; d is the differential symbol; η is the azimuthal slow time; Step 23, comparison of volatility; Set the volatility threshold to v in ; and v in <0.1; Compare With v in ,like Removal The corresponding candidate scattering point; if reserve The corresponding candidate scattering point is recorded as Then record the Corresponding candidate scattering point ms m ; Longitudinal fluctuation threshold v in After comparison, the scattering point corresponding to the retained range compression domain signal is obtained and recorded as the high-quality scattering point SR sample ,and 4. The distributed strong scattering point scene SAR image autofocusing method based on improved PGA according to any one of claims 1 to 3, characterized in that In step three: Step 31, phase unwrapping processing; In the MOPGA estimation process, first use formula (5) to calculate The unwrapped phase of each range compression domain signal in is denoted as SW; Phase after expansion of range compression domain signal for: x is the discrete azimuth time; unwrap is the phase unwrapping operator; Arg is the operator for taking the principal value of the argument; s rc is the range compression domain signal; Step 32, calculating the phase curvature; The curvature of the phase error is calculated for SW using formula (6), which is denoted as SU. Phase curvature after expansion of range compression domain signal for: Step 33, calculating the average phase curvature; The average curvature of the phase error samples of SU is calculated using formula (7), which is recorded as ST; Coherent averaging of phase curvature after expansion of range-compressed signal for: K is the total number of samples in SRSS; k is any sample in SRSS; is the phase curvature of the expanded range compression domain signal corresponding to the kth sample; Step 34, calculating the phase error; The phase error estimation result is obtained by performing two cumulative sums of the average curvature of the phase error samples of ST using formula (8), which is recorded as SS. Phase error estimation results for: x is the discrete azimuth time; I is the serial number of the outer summation symbol; i is the serial number of the inner summation symbol; is the coherent average of the phase curvature of the i-th range compression domain signal after expansion in the inner layer.

5. The distributed strong scattering point scene SAR image autofocusing method based on improved PGA according to any one of claims 1 to 4, characterized in that In step 4: During the iteration process, if the SS of the current iteration σ is less than or equal to ζ in , the iteration ends and the SAR image processed by MOPGA is output.

6. The distributed strong scattering point scene SAR image autofocusing method based on improved PGA according to any one of claims 1 to 4, characterized in that In step 4: During the iteration process, if the number of current iterations reaches the maximum number of iterations, the SAR image processed by MOPGA is output.

Images