A Phase Tracking-Based Correction Method for High-Order Range Migration of Moving SAR Targets
Through phase tracking and EKF technology, the computational complexity and stability of high-order distance migration correction in SAR dynamic target imaging are solved, and high-precision correction and high-resolution imaging of non-ideal motion models are achieved.
Patent Information
- Application Number
- CN202410509140.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-04-26
- Publication Date
- 2024-11-19
- Estimated Expiration
- 2044-04-26
AI Technical Summary
The existing SAR dynamic target imaging technology has problems such as high computational complexity, poor stability and high signal-to-noise ratio threshold in high-order distance migration correction (RCMC), especially in the face of non-ideal motion models and platform jitters. .
Using a phase tracking method, first-order distance movement correction is completed through KT, the second-order distance bending compensation factor is constructed using the platform speed, and extended Kalman filtering (EKF) is introduced after the distance frequency domain segmentation to track the phase history to build high The order phase compensation factor is used to achieve high-resolution high-order parameter estimation and distance migration correction in all frequency bands.
High-precision high-order distance migration correction under any motion model is realized, which reduces the computational complexity, ensures estimation accuracy and signal-to-noise ratio thresholds, and is suitable for high-resolution dynamic target imaging.
Smart Images

Figure CN118425962B_ABST
Abstract
Description
Technical Field
[0001] The invention belongs to the field of radar technology, and in particular relates to a SAR moving target high-order range migration correction method based on phase tracking. Background Art
[0002] Imaging of moving targets is a hot research topic in Synthetic Aperture Radar (SAR) in recent years, and its core lies in the estimation of motion parameters. Compared with stationary targets, moving targets will bring more complex range cell migration (RCM). However, there is an inherent contradiction between parameter estimation and range cell migration correction (RCMC), that is, RCMC depends on estimating motion parameters from SAR echoes, and the estimation of motion parameters is based on RCMC. Therefore, RCMC is a key step in achieving SAR moving target imaging. In recent years, with the development of technology, the resolution requirements of SAR have been continuously improved. On the other hand, the diversification of radar-carrying platforms and the complexity of motion forms have aggravated the complexity of RCM. Accurate RCMC has become an urgent problem to be solved in high-resolution moving target imaging.
[0003] In recent years, there have been a lot of studies on RCMC methods. According to whether the relative motion between the target and the radar is modeled as a model of exact order and the model parameters are estimated, the existing methods can be divided into parametric methods and non-parametric methods. According to the RCMC order considered, it can be divided into low-order (range walk and second-order range bending) and high-order RCMC methods. In the field of SAR moving target imaging, the commonly used low-order RCMC methods are mainly KT class and Radon Fourier Transform (RFT) class. The KT method was proposed by RCDiPietro in 1999. It is a typical non-parametric search transform method. It eliminates the range walk (Range Cell Walk, RCW) by stretching the transform along the slow time axis. Subsequently, in order to broaden the scope of application of Keystone and realize the correction of second-order range bending, the second-order Keystone transform (Second order KT, SOKT) was proposed. KT methods can achieve RCMC when the target prior information is unknown, but KT can only correct linear motion. Although SOKT can correct range curvature, it usually needs to use Hough Transform (HT) or Radon Transform (RT) with non-coherent accumulation to estimate the slope of linear motion and finally compensate RCW. However, the introduction of non-coherent accumulation methods improves the signal-to-noise ratio (SNR) threshold. In 2011, J.Xu proposed the RFT method based on target motion parameter search based on the relationship between target speed and RCW, but this method will produce a high blind speed sidelobe (BSSL). In order to solve the range curvature and Doppler spread caused by target acceleration, the RFRFT (Radon-fractional Fourier Transform, RFRFT) algorithm was proposed based on RFT. The maneuvering target parameter estimation is completed by searching for distance, speed and acceleration, and the computational complexity is high. In addition, in 2014, X. Rao proposed the Axis Rotation (AR) method to correct RCW by rotating two-dimensional echo data. In 2019, the RCMC of high-resolution SAR moving targets was achieved by combining the energy balance method, HT and Map-drift for parameter estimation, with small computational complexity and stable performance.
[0004] The above methods can only correct first-order linear walk and second-order range curvature. However, in SAR moving target imaging, in order to achieve long coherent accumulation and high-resolution imaging, high-order RCM cannot be ignored. The methods suitable for high-order RCMC mainly include generalized Radon Fourier Transform (GRFT), adjacent cross correlation function (ACCF), frequency axis rotation (FAR) and envelope alignment algorithms. For high-order RCM, it can be compensated with the envelope alignment method in ISAR. Commonly used envelope alignment methods include accumulation cross correlation method, minimum correlation distance method and global minimum entropy method. These non-parametric model methods do not need to build motion models and have wide applicability, but have high SNR requirements. In order to reduce SNR, some methods model the motion model as a polynomial, such as envelope alignment methods based on swarm intelligence optimization, gradient descent, and time-frequency distribution. However, these methods have poor robustness, and not all motion models can be modeled as polynomials.
[0005] In summary, the existing parametric methods are prone to motion model mismatch and high computational complexity during the RCMC process; the non-parametric methods have high SNR requirements and unstable performance. More importantly, the high-order RCM caused by platform jitter or yaw, as well as complex target motion, may not conform to the polynomial model. In this case, the above-mentioned parametric high-order RCMC method will fail. Summary of the invention
[0006] In order to meet the requirements of long coherent accumulation and high-resolution imaging for high-precision RCMC, this application proposes a high-order RCMC method applicable to any motion model for this type of non-ideal motion model. Considering that the main component of RCM comes from linear movement, the second-order range curvature mainly comes from the movement of the radar platform and the platform movement speed is usually much higher than the target azimuth speed, this application has moderate computational complexity, high RCMC accuracy, and is not limited to the RCM model. In theory, high-order RCMC can be achieved under any model; in addition, EKF can make suboptimal estimates of state quantities under linear Gaussian noise conditions with the minimum mean square error as the criterion, while ensuring the estimation performance, filtering denoising is achieved, so the estimation accuracy and SNR threshold are guaranteed. This method is used to process multiple sets of simulation data. The experimental results show that this application can effectively correct high-order RCM under non-ideal motion models and is suitable for high-resolution moving target imaging.
[0007] The present application discloses a SAR moving target high-order range migration correction method based on phase tracking, comprising the following steps:
[0008] Receive SAR echo signals and perform distance compression on the SAR echo signals;
[0009] The RCW correction is completed by KT, and the compensation factor is constructed by using the platform speed to achieve the rough correction of the second-order range bending;
[0010] By segmenting the distance frequency domain, the bandwidth is reduced, the range migration tolerance within the sub-band is widened, and the residual high-order RCM falls into the same distance unit;
[0011] The extended Kalman filter is introduced, and the state space equation of the phase history is constructed to track the signal phase history of the unit where the target is located;
[0012] By tracking the obtained phase inversion to obtain the RCM component and constructing the compensation factor, high-order parameter estimation and range migration correction can be achieved in the full frequency band with high resolution.
[0013] Preferably, the method of using KT to complete RCW correction and using the platform speed to construct a compensation factor to achieve a rough correction of the second-order range bending includes:
[0014] Perform FFT along the range direction on the SAR echo signal after range compression to obtain the range frequency domain azimuth time domain echo signal S(f r , t a ),as follows:
[0015]
[0016] Among them, f r t r The corresponding frequency variable, t r and t a They are respectively the fast time in distance and the slow time in azimuth, W r (f r ) is w r (t r ) corresponds to the frequency domain window function, w r (t r ) and w a (t a ) are the range and azimuth envelopes respectively; δ s is the signal amplitude after pulse compression, c is the speed of light, f c represents the carrier frequency, v r is the slant range velocity, ∈(t a ) is a high-order term caused by non-ideal factors, is the second-order range curvature, R0 is the closest slope distance, and the first exponential term in the formula is f r and t a The coupling term represents the range migration; the second exponential term is related to the fast time frequency f r Irrelevant, indicating Doppler migration term;
[0017] Use KT to correct RCW. The expression of KT is as follows:
[0018] (f r +f c )t a =f c η a
[0019] Among them, η a represents the new slow time variable;
[0020] The signal after getting KT is:
[0021]
[0022] In the above formula, f r With η a The coupling of has been eliminated, that is, the first-order RCW has been corrected, and f still exists in the third exponential term. r and The coupling of , produces second-order distance bending; there is f in the fourth exponential term r and non-ideal factor error∈(η a )’s high-order coupling produces high-order RCM;
[0023] Next, the second-order range curvature is corrected: first, the compensation factor is constructed using the carrier platform speed to make a rough correction for the second-order range curvature and eliminate most of the range curvature; the constructed compensation factor H1 (f r , η a )as follows:
[0024]
[0025] have:
[0026]
[0027] λ is the signal wavelength, V is the speed of the carrier platform;
[0028] Finally, perform IFFT along the range direction on the above equation to obtain the pulse compression signal after RCM coarse correction as follows:
[0029]
[0030] Among them, B r represents the transmission signal bandwidth, α=(Vv a ) 2 / (2R0)+a r / 2, α′=α-V 2 / 2R0,v a represents the target azimuth velocity, ar Represents the acceleration in the distance direction.
[0031] Preferably, the method of reducing the bandwidth and widening the range migration tolerance within the sub-band by segmenting the range frequency domain so that the residual high-order RCM falls into the same range unit includes:
[0032] The segmented echo signal is expressed as:
[0033]
[0034] Among them, N r Represents the number of frequency domain segments, the bandwidth of the subband signal is reduced by N r times, the corresponding distance resolution is reduced by N r times.
[0035] Preferably, the introduction of extended Kalman filtering and the construction of a state space equation of the phase history to track the signal phase history of the unit where the target is located include:
[0036] The RCM compensation factor is constructed by tracking the signal phase history of the distance unit where the sub-band signal target is located;
[0037] The signal of the distance unit where the target is located in the sub-band is:
[0038]
[0039] Among them A s represents the signal amplitude, then the Doppler phase history generated by the relative motion between the radar and the target is
[0040]
[0041] Phase φ(η a ) satisfies the smoothness assumption, that is, the phase difference between adjacent pulse sampling points does not fluctuate violently, and its mathematical expression is:
[0042] φ(k+1)-φ(k)≈φ(k)-φ(k-1)
[0043] Wherein, k is the azimuth sampling index, and k = 1, 2, ..., N a , N a represents the number of sampling points in azimuth, φ(k+1) and φ(k) represent the phases at the k+1th and kth sampling moments respectively;
[0044] In order to obtain the phase history, EKF is used for phase tracking, and the EKF state space equation representing the phase change relationship is established:
[0045] First, the three adjacent phase sampling points constitute the state vector, then the state vector corresponding to the (k-1) and k sampling moments is and They are:
[0046]
[0047] in(·) T represents transpose;
[0048] According to the recursive relationship between state vectors, the state equation is as follows:
[0049]
[0050] Wherein, B is the state transfer matrix, and B = [0, 1, 0; 0, 0, 1; 0, -1, 2];
[0051] The observation vector is constructed as follows:
[0052]
[0053] Among them, s k Represents the sampling of observation values between three adjacent pulses, n r (k) and n i (k) represent the real and imaginary sampling of the observation noise, respectively;
[0054] Considering the influence of noise terms in practice, the EKF state space equation for phase tracking is constructed as follows:
[0055]
[0056] Among them, w k-1 Represents the process noise vector, which describes the approximate error of the smoothing assumption. The corresponding process noise covariance matrix is recorded as W k-1 , n k represents the observation noise vector, and n k =[n r (k), n i (k)] T , the corresponding observation noise covariance matrix is recorded as R k , The expression is
[0057]
[0058] In tracking phase At the same time, the amplitude A s (k) Tracking, that is, constructing a binary state space equation
[0059] The amplitude state vector at sampling time k is denoted as p k, construct the state equation corresponding to the amplitude, and set the constant amplitude A s Use p k Substituting, we obtain the binary state space equation as follows:
[0060]
[0061] Among them, the amplitude state vector p k It is composed of three adjacent amplitude sampling values, p k =[A s (k-1), A s (k), A s (k+1)] T , v k Represents the noise vector, and its corresponding noise covariance matrix is recorded as V k-1 ; In addition, in the observation equation, ⊙ represents the Hadamard product, and h1(p k )=[p k , p k ] T , h1(p k )and The corresponding Jacobian matrices are H 1k and H 2k ,pass For p k and Find the partial derivative to obtain;
[0062] Processing binary state equations through ADMM algorithm;
[0063] Solve the binary state space equation through EKF and get the phase state vector After traversing all sampling points, the Doppler phase history φ(η a )
[0064] Preferably, after obtaining the Doppler phase history φ(η a ) Then, the estimated distance between the radar platform and the target is calculated. for:
[0065]
[0066] is the estimated value of the Doppler phase history. It is worth noting that t a and η both represent slow time and have the same physical meaning;
[0067] Construct a high-order phase compensation function H2(f r , ta ),as follows:
[0068]
[0069] Perform IFFT along the distance direction to obtain the results of high-order parameter estimation of the full-band data and range migration correction, as follows:
[0070]
[0071] At this point, high-order parameter estimation and range migration correction are completed. BRIEF DESCRIPTION OF THE DRAWINGS
[0072] Figure 1 Schematic diagram of the geometric relationship of airborne SAR;
[0073] Figure 2 Data preprocessing results;
[0074] Figure 3 Doppler phase history tracking results;
[0075] Figure 4 RCMC results of full-band data;
[0076] Figure 5 Scattering point model;
[0077] Figure 6 The two-dimensional focusing results of this application at 5db;
[0078] Figure 7 The two-dimensional focusing results of this application at -5db. DETAILED DESCRIPTION
[0079] The present invention is further described below in conjunction with the accompanying drawings, but the present invention is not limited in any way. Any changes or substitutions made based on the teachings of the present invention belong to the protection scope of the present invention.
[0080] The geometric relationship of airborne SAR observing ground moving targets is as follows: Figure 1 As shown. O is the origin of the coordinate system, and the carrier platform flies along the azimuth Y at a speed of V. In the slant range plane, suppose there is a moving target T at the closest slant range R0, and its slant range velocity and acceleration are v respectively. r and a r , the azimuthal velocity and acceleration are v a and a a . In the azimuth slow time t a When the aircraft moves to point P on the real trajectory, which corresponds to point P′ on the ideal trajectory, the actual instantaneous slant distance between the aircraft platform and the target is recorded as R(t a ), the instantaneous slant distance under the ideal trajectory is recorded as R′(t a ).
[0081] Ideally, the aircraft platform is considered to fly along an ideal straight line trajectory. Figure 1 In the geometric model, the instantaneous slope distance R′(t a ) can be expressed as:
[0082]
[0083] In practice, due to the influence of non-ideal factors such as platform shaking, yaw, and terrain undulation, the actual instantaneous slant range R(t a ) will usually be in R′(t a ) based on the introduction of a slow time t a Complex distance error ε(t a ),Right now:
[0084] R(t a )=R′(t a )+ε(t a ) (2)
[0085] Combining formula (1), Taylor expansion of formula (2) yields
[0086]
[0087] Where α=(Vv a ) 2 / (2R0)+a r / 2.∈(t a ) represents the high-order residual term O(t a ) and the slant range error ε(t a ), that is, ∈(t a )=ε(t a )+O(t a ), in this application, ∈(t a ) are called higher-order terms.
[0088] Assuming that the radar transmits a linear frequency modulated (LFM) signal, the baseband echo signal of the SAR observing the moving target can be expressed as:
[0089]
[0090] Among them, δ0 is the echo reflection coefficient, t r and t a They are respectively the fast time in distance and the slow time in azimuth, w r (t r ) and w a (t a) are the envelopes in the range and azimuth directions, usually rectangular window functions. r is the modulation frequency of LFM, c is the speed of light, λ is the signal wavelength, and λ=c / f c , f c Indicates the carrier frequency.
[0091] Substituting equation (3) into equation (4), we can obtain the SAR echo signal after range compression:
[0092]
[0093] Among them, δ s is the signal amplitude after pulse compression, B r Indicates the transmit signal bandwidth.
[0094] From Equation (5), we can see that the high-order term ∈(t a ) will introduce high-order RCM and Doppler migration. Under long-term coherent accumulation and high-resolution imaging, the high-order RCM generated by this high-order term cannot be ignored. Therefore, the focus of this application is to eliminate ∈(t a ) provides a good basic condition for high-resolution imaging.
[0095] In order to correct the high-order RCM generated by non-ideal factors so that it can meet the purpose of long coherent accumulation and high-resolution imaging, this application proposes a method for realizing high-order RCMC based on phase tracking. This application mainly includes RCM coarse correction, range frequency domain sub-band phase tracking and RCM compensation factor construction.
[0096] Considering that the main component of RCM comes from the first-order linear motion (v r t a ), and secondly from the second-order distance bending This application first performs RCM rough correction, that is, first uses KT to achieve first-order linear movement correction, and then constructs a second-order range bending compensation factor according to the speed of the carrier platform to eliminate most of the second-order range bending, as follows:
[0097] Performing FFT along the range direction on equation (5) can obtain the range frequency domain azimuth time domain echo signal S(f r , t a ),as follows:
[0098]
[0099] Among them, f r t r The corresponding frequency variable, W r (f r ) is w r (t r) corresponds to the frequency domain window function. The first exponential term in the above formula is f r and t a The coupling term represents the range migration; the second exponential term is related to the fast time frequency f r Irrelevant, indicating Doppler migration term.
[0100] In this application, considering that KT can complete the first-order linear motion correction without the need for prior information of the target motion, KT is first used to correct RCW. The expression of KT is as follows:
[0101] (f r +f c )t a =f c η a (7)
[0102] Among them, η a Represents the new slow time variable.
[0103] Substitute equation (7) into equation (6), and according to f c / (f r +f c )≈1-f r / f c , ignoring the envelope window function, the signal after KT is:
[0104]
[0105] It can be seen that f r With η a The coupling has been eliminated, that is, the first-order RCW has been corrected. However, in the above formula, the third exponential term still has f r and The coupling of , produces second-order distance bending; there is f in the fourth exponential term r and non-ideal factor error∈(η a )’s high-order coupling produces high-order RCM.
[0106] Next, the second-order range curvature is corrected. From the third exponential term in equation (8), it can be seen that the second-order range curvature is related to α, and α is determined by the carrier platform speed V and the target azimuth speed v a and the acceleration in the distance a r Since the motion parameters of the non-cooperative target are unknown, namely v a and a r Unknown, but considering that the second-order range curvature is mainly caused by the carrier platform speed V, that is, V is usually much larger than v a and a rTherefore, this application first uses the carrier platform speed to construct a compensation factor to make a rough correction for the second-order range curvature and eliminate most of the range curvature. The constructed compensation factor is as follows:
[0107]
[0108] Multiplying equation (9) and (8), we have
[0109]
[0110] Finally, by performing IFFT along the range direction on equation (10), the pulse compression signal after RCM coarse correction can be obtained as follows:
[0111]
[0112] From the above formula, we can find that the first-order RCW and most of the second-order range curvature have been corrected. When the resolution requirement is not high and the platform stability is good, it is believed that the RCMC can meet the imaging requirements, that is, The RCM caused by this method can be ignored. However, under the requirements of long-term coherent accumulation and high-resolution imaging, the residual RCM cannot be ignored and needs further correction.
[0113] After the RCM rough correction, the phenomenon of the target signal after pulse compression crossing the distance unit is significantly weakened, and the distance unit occupied by the high-order residual RCM is limited. Therefore, next, through the range frequency domain sub-band, the bandwidth in the sub-band is reduced, and then the range migration tolerance in the sub-band is widened, so that the residual high-order migration amount is distributed in the same sub-band distance unit. The segmented echo signal is expressed as:
[0114]
[0115] Among them, N r Represents the number of frequency domain segments, the bandwidth of the subband signal is reduced by N r times, the corresponding distance resolution will also be reduced by N r However, it can be seen from formula (12) that the high-order residual is still reflected in the azimuth Doppler phase. In order to achieve high-order RCMC of the full-band signal, the present application constructs the RCM compensation factor by tracking the signal phase history of the distance unit where the sub-band signal target is located.
[0116] From formula (12), we can get that the signal of the distance unit where the target is located in the sub-band is:
[0117]
[0118] Among them A s represents the signal amplitude. Then the Doppler phase history generated by the relative motion between the radar and the target is:
[0119]
[0120] Normally, the phase φ(η a ) is modeled as a polynomial phase signal (PPS) model. However, due to the non-ideal factors ∈(η a ) is produced. In some cases, PPS is used to describe the phase φ(η a ) is not optimal, or the required order of PPS is high. But without loss of generality, the phase φ(η a ) satisfies the smoothness assumption, that is, the phase difference between adjacent pulse sampling points will not fluctuate violently, and its mathematical expression is:
[0121] φ(k+1)-φ(k)≈φ(k)-φ(k-1) (15)
[0122] Wherein, k is the azimuth sampling index, and k = 1, 2, ..., N a , N a Indicates the number of sampling points in the azimuth direction.
[0123] In order to obtain the phase history, this application proposes to use EKF for phase tracking. The key is to establish an EKF state space equation that can represent the phase change relationship. First, assuming that the three adjacent phase sampling points constitute the state vector, then the state vector corresponding to the (k-1) and k sampling moments is and They are:
[0124]
[0125] in(·) T Indicates transpose.
[0126] Then, combining equations (15) and (16), we can get the recursive relationship between the state vectors, and the state equation can be obtained as follows:
[0127]
[0128] Wherein, B is the state transfer matrix, and B=[0, 1, 0; 0, 0, 1; 0, -1, 2].
[0129] In addition, EKF needs to construct the observation equation based on the observation data. From formula (17), it can be seen that the state vector is composed of 3 adjacent phase sampling points, so the observation vector should also be composed of 3 adjacent observation values. In this application, the constructed observation vector is as follows:
[0130]
[0131] Among them, s k represents the sampling of the observation value between three adjacent pulses, that is, s in formula (13) T (η a ) is a discrete sampling value. Specifically, s k =[s T (k-1), s T (k), s T (k+1)] T .n r (k) and n i (k) represent the real and imaginary samples of the observation noise, respectively.
[0132] Combining equations (17) and (18), and considering the influence of noise terms in practice, the EKF state space equation for phase tracking can be constructed as:
[0133]
[0134] Among them, w k-1 represents the process noise vector, which describes the approximate error of the smoothing assumption in equation (15). The corresponding process noise covariance matrix is recorded as W k-1 .n k represents the observation noise vector, and n k =[n r (k), n i (k)] T , the corresponding observation noise covariance matrix is recorded as R k In addition, according to formula (18), The expression is
[0135] Theoretically, the estimated value of the Doppler phase history can be obtained by iteratively solving Equation (19) through EKF. However, considering a more practical situation, the amplitude A in Equation (13) s A may be affected by factors such as antenna pattern weighting and noise interference. s It is not constant and known, but usually changes and is unknown, i.e. A s Should be recorded as A s (k). At this time, in the tracking phase At the same time, the amplitude A s (k) To track, it is necessary to construct a binary state space equation.
[0136] Assume that the amplitude state vector at sampling time k is denoted by p k , similar to equation (17), we can construct the state equation corresponding to the amplitude and replace the constant amplitude A in equation (18) with s Use p kInstead, we can obtain a binary state space equation that is more in line with the actual situation, as follows:
[0137]
[0138] Among them, the amplitude state vector p k It is composed of three adjacent amplitude sampling values, p k =[A s (k-1), A s (k), A s (k+1)] T .v k Represents the noise vector, and its corresponding noise covariance matrix is recorded as V k-1 In addition, in the observation equation, ⊙ represents the Hadamard product, and h1(p k )=[p k , p k ] T , h1(p k )and The corresponding Jacobian matrices are denoted as H 1k and H 2k , which can be achieved through For p k and Obtain the partial derivative.
[0139] From equation (20), we can see that the binary state space equation consists of two linear state equations and one nonlinear observation equation. Considering that the ADMM algorithm is widely used to solve multivariable problems, this application uses the ADMM idea to process the binary state equation. Finally, the EKF is used to solve the binary state space equation in equation (20) to obtain the phase state vector After traversing all sampling points, the Doppler phase history φ(η a ) The specific process of iteratively solving equation (20) using EKF is shown in Table 1.
[0140] In obtaining the Doppler phase history φ(η a ) Then, combining equations (3) and (14), we can get the estimated distance history between the radar platform and the target: for:
[0141]
[0142] It is worth noting that in this application, what is tracked is the phase It includes all order motion components of the radar and the target and is not limited to a specific motion model. In other words, is an accurate representation of the motion model. In addition, it can be seen from equation (21) that in the inverted distance history When the phase tracking result is Multiply by the factor λ / (4π), and usually, λ is much smaller than 1, then λ / (4π)<<1, so when inverting the distance history, the present application can have a better tolerance for the error of the phase tracking result.
[0143] Next, we use equation (21) to construct a high-order phase compensation function H2(f r , t a ),as follows:
[0144]
[0145] Finally, by multiplying equation (22) with (6) and performing IFFT along the distance direction, the result after high-order RCMC of the full-band data can be obtained, as follows:
[0146]
[0147] At this point, the high-order RCMC is completed.
[0148] Table 1 Iterative solution process of binary state space equations
[0149]
[0150]
[0151] In this section, the computational complexity of the proposed RCMC method is analyzed. First, for ease of explanation, the parameters involved are explained. Assume that N r and N a They represent the number of sampling points in the range direction and the number of pulses in the azimuth direction of the SAR echo data respectively. In addition, in the present application, M=3.
[0152] For this application, the main steps include KT, second-order range curvature coarse correction and phase tracking. The computational complexity required to implement KT through interpolation operation is The second-order range bending rough correction can be achieved only by complex multiplication, and its computational complexity is O(N r N a ); Finally, when using EKF for phase tracking, the computational complexity of each step is given in Table 1. It can be seen that in each iteration, the computational complexity of steps 3 to 12 is approximately O(M 3 ), and need to be (N a -3) iterations, the computational complexity of phase tracking is about O(M 3N a ). In summary, the computational complexity of the improved order RCMC method is
[0153] In this section, the airborne SAR simulation parameters and target motion parameters are shown in Tables 3 and 4, respectively. In the simulation, it is assumed that the slant range error ε(t a ) makes cosine waves with slow time, and ε(t a )=0.6cos(2πt a ).
[0154] Table 2 Radar simulation parameters
[0155] Parameter name Parameter Value Center frequency (GHz) 10 Signal bandwidth (MHz) 300 Range sampling rate (MHz) 360 Signal pulse width (10μs) 10 Pulse repetition frequency (Hz) 2000 Center slope distance (Km) 10 Distance sampling points 4096 Azimuth sampling points 8192
[0156] Table 3 Platform and target motion parameters
[0157] Parameter name V <![CDATA[υ r ]]> <![CDATA[a r ]]> <![CDATA[υ a ]]> <![CDATA[a a ]]> Parameter Value 100m / s -10m / s <![CDATA[-2m / s 2 ]]> 10m / s <![CDATA[2m / s 2 ]]>
[0158] The RCMC method proposed in this application mainly includes data preprocessing and phase tracking. Data preprocessing includes pulse compression, KT correction RCW, second-order range bending rough correction and range frequency domain segmentation. In order to verify the feasibility of this application, first Figure 2 The results of data preprocessing are given in . In this section, SNR=30dB, and the SNRs mentioned in this application are all after pulse compression.
[0159] from Figure 2 As can be seen in (a), due to the influence of target motion and non-ideal factors, the envelope curve after pulse compression has not only first-order RCW and second-order range curvature, but also high-order RCM. After KT, the dominant component of RCM, namely the first-order RCW, is corrected, and the result is as follows: Figure 2 Next, the platform velocity is used to construct the second-order range bending rough correction compensation factor. The compensated result is shown in Figure 2 (c) It can be seen that most of the RCM has been corrected, and the remaining high-order RCM spans a limited range unit. Combined with the range resolution, it can be seen that if the range frequency domain is divided into three sub-bands, the range migration tolerance can be widened, so that the remaining high-order RCM is in the same range unit of the sub-band signal. The sub-band pulse pressure signal is as follows: Figure 2 As shown in (d), it can be seen that the target signal is in the same distance unit at this time.
[0160] Then, the echo signal of the target range unit is extracted in the frequency domain subband for phase tracking. The Doppler phase history obtained by phase tracking is as follows: Figure 3 (a) shows that the error between it and the theoretical value is Figure 3(b) As shown. It can be seen that the present application can effectively obtain the estimated value of the Doppler phase. In addition, it is worth noting that when performing phase tracking, the initial values of the parameters required for the EKF iteration in Table 1 are set as follows: Phase state vector initial value It does not affect the phase change trend of the estimated results, so setting The initial value of the amplitude state vector p2 = mean|s T (k)|*1 3×1 , which is the mean amplitude of the observed data; in addition, the initial values of the phase state vector, the state equation noise covariance matrix Φ2 and W2, the amplitude state vector, the state equation noise covariance matrix P2 and V2 are all set to I 3×3 , which will be adaptively adjusted according to the observed data during iteration, and the present application introduces a phase smoothness constraint to reduce the sensitivity of the phase tracking process to the initial parameters. Finally, the initial value R2 of the noise covariance matrix of the observation equation is obtained by statistically analyzing the noise units of the measured data.
[0161] Finally, the phase tracked by EKF is used to construct the RCM compensation function in combination with equations (21) and (22). The compensation function is multiplied with the range frequency domain data after SAR echo pulse compression to obtain the RCMC result of the full-band data, as shown in Figure 4 As shown, it can be seen that the residual high-order RCM is completely corrected and the targets are distributed in the same distance unit, which also illustrates the effectiveness of the present application in processing high-order RCM.
[0162] The high-order RCMC results of this application at different SNRs are given below. It should be noted that when the SNR is low, it is impossible to directly determine whether the target signal has been corrected to the same distance unit from the pulse pressure results after RCMC. Therefore, in this application, the azimuth focusing function is constructed using theoretical values for the results after RCMC to achieve two-dimensional focusing. In theory, the better the RCMC effect, the better the image quality after two-dimensional focusing.
[0163] Under different SNRs, the RCMC of the present application is followed by azimuth focusing. Even when SNR = -5dB, the present application can still obtain good RCMC results. The reason is that, on the one hand, the KT and second-order range bending coarse correction involved in the present application have no requirements for SNR; on the other hand, when performing EKF phase tracking on the sub-band signal, it essentially performs filtering and denoising, combined with phase smoothing phase estimation values, thereby achieving higher-precision RCMC. In addition, it should be noted that for the present application, SNR refers to the SNR of the full-band data. When performing sub-band division processing, the SNR will be reduced. In the present application, the number of sub-bands is 3, that is, the SNR of the sub-band signal will lose about 5dB. In other words, the proposed EKF phase tracking can still obtain phase tracking results that meet the accuracy required by RCMC when SNR = -10dB.
[0164] To further illustrate the SNR advantage of the present application, the Monte Carlo simulation results of the 2D focused image contrast after RCMC of the proposed method are given in Table 4, where the number of Monte Carlo simulations at each SNR is 100. It can be seen that the present application is not limited to a specific model assumption, and can correct RCM of any order, and the filtering effect of EKF makes its performance stable at low SNR.
[0165] Table 4 Image contrast at different SNRs
[0166] This application 491.58 243.74 157.03 112.7 110.76 100.62 100.22 SNR 5dB 2dB 0dB -2dB -4dB -5dB -6dB
[0167] In actual situations, the SAR moving target detected from the SAR echo is composed of multiple scattering points. In order to be more realistic and further illustrate the effectiveness of this application, a simulation experiment is conducted on the multi-scattering point model in this section. In addition, the distribution of the multi-scattering points is as follows: Figure 5 Under the multi-scattering point model, the results of RCMC and two-dimensional focusing using this application are shown as follows Figure 6 and Figure 7 As shown. It can be seen that when SNR = 5dB, the focusing quality of the present application is relatively high. In addition, even when SNR = -5dB, the contour information of the target can be obtained from the two-dimensional focusing result after RCMC of the present application. This shows that the present application is also applicable to moving targets composed of multiple scattering points in practice.
[0168] The beneficial effects of this application are as follows:
[0169] This application studies a high-order range migration correction method for high-resolution SAR moving targets. In order to solve the problems of large computational complexity and unsatisfactory results faced by existing methods in processing high-order RCM, a method for phase tracking based on EKF and constructing an RCM compensation function is proposed. This application makes full use of the advantages of EKF's low computational complexity and strong anti-noise performance, and what is obtained during the phase tracking process is the phase history, which is not limited to the relative motion model between the platform and the target. Compared with existing methods, this application has moderate computational complexity and wide applicability. The processing results of multiple sets of simulation data verify the correctness and effectiveness of this text application.
[0170] As used herein, the word "preferred" is intended to be used as an example, instance, or illustration. Any aspect or design described herein as "preferred" is not necessarily to be construed as being more advantageous than other aspects or designs. On the contrary, the use of the word "preferred" is intended to present concepts in a specific way. The term "or" as used in this application is intended to mean an inclusive "or" rather than an exclusive "or". That is, unless otherwise specified or clear from the context, "X uses A or B" means any one of the naturally included permutations. That is, if X uses A; X uses B; or X uses both A and B, then "X uses A or B" is satisfied in any of the foregoing examples.
[0171] Moreover, although the present disclosure has been shown and described with respect to one or implementations, those skilled in the art will think of equivalent variations and modifications based on the reading and understanding of this specification and the accompanying drawings. The present disclosure includes all such modifications and variations, and is limited only by the scope of the appended claims. In particular, with respect to the various functions performed by the above-mentioned components (such as elements, etc.), the terms used to describe such components are intended to correspond to any component (unless otherwise indicated) that performs the specified function of the component (such as it is functionally equivalent), even if the structure is not equivalent to the disclosed structure of the function in the exemplary implementation of the present disclosure shown herein. In addition, although the specific features of the present disclosure have been disclosed with respect to only one of several implementations, such features can be combined with one or other features of other implementations that may be desired and advantageous for a given or specific application. Moreover, insofar as the terms "including", "having", "containing" or their variations are used in specific embodiments or claims, such terms are intended to be included in a manner similar to the term "comprising".
[0172] The functional units in the embodiments of the present invention may be integrated into a processing module, or each unit may exist physically separately, or multiple or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The above-mentioned storage medium may be a read-only memory, a disk or an optical disk, etc. The above-mentioned devices or systems may execute the storage method in the corresponding method embodiment.
[0173] To sum up, the above embodiment is an implementation mode of the present invention, but the implementation mode of the present invention is not limited by the embodiment. Any other changes, modifications, substitutions, combinations, and simplifications that deviate from the spirit and principles of the present invention should be equivalent replacement methods and are included in the protection scope of the present invention.
Claims
1. A phase tracking-based high-order range migration correction method for SAR moving targets, characterized by: The following steps are involved: Receive SAR echo signals and perform distance compression on the SAR echo signals; RCW correction is performed using KT, and the platform velocity is used to construct a compensation factor to achieve a rough correction of the second-order range bending; By segmenting the range frequency domain, the bandwidth is reduced, the range migration tolerance within the sub-band is widened, and the residual high-order RCM falls into the same range unit; The extended Kalman filter is introduced, and the state space equation of the phase history is constructed to track the signal phase history of the unit where the target is located; By tracking the obtained phase inversion to obtain the RCM component and constructing the compensation factor, high-order parameter estimation and range migration correction can be achieved in the full frequency band with high resolution.
2. The phase tracking-based high-order range migration correction method for SAR moving targets according to claim 1, characterized in that: The RCW correction is performed using KT, and the compensation factor is constructed using the platform velocity to achieve the second-order range curvature rough correction, including: Perform FFT along the range direction on the SAR echo signal after range compression to obtain the range frequency domain azimuth time domain echo signal S(f r , t a ),as follows: Among them, f r t r The corresponding frequency variable, t r and t a They are respectively the fast time in distance and the slow time in azimuth, W r (f r ) is w r (t r ) corresponds to the frequency domain window function, w r (t r ) and w a (t a ) are the range and azimuth envelopes respectively; δ s is the signal amplitude after pulse compression, c is the speed of light, f c Indicates the carrier frequency, υ r is the slant range velocity, ∈(t a ) is a high-order term caused by non-ideal factors, is the second-order range curvature, R0 is the closest slope distance, and the first exponential term is f r and t a The coupling term represents the range migration; the second exponential term is related to the fast time frequency f r Irrelevant, indicating Doppler migration term; Use KT to correct RCW. The expression of KT is as follows: (f r +f c )t a =f c η a Among them, η a represents the new slow time variable; The signal after getting KT is: In the above formula, f r and η a The coupling has been eliminated, that is, the first-order RCW is corrected, and f still exists in the third exponential term. r and The coupling of , produces the second-order distance bending; there is f in the fourth exponential term r and non-ideal factor error∈(η a )’s high-order coupling, resulting in high-order RCM; Next, the second-order range curvature is corrected: the compensation factor is constructed using the carrier platform speed to make a rough correction for the second-order range curvature and eliminate most of the range curvature; the constructed compensation factor H1(f r , η a )as follows: have: λ is the signal wavelength, V is the speed of the carrier platform; Finally, perform IFFT on the above equation along the range direction to obtain the pulse compression signal after RCM coarse correction as follows: Among them, B r Indicates the bandwidth of the transmitted signal, α=(V-υ a ) 2 / (2R0)+a r / 2, α′=α-V 2 / 2R0,υ a represents the target azimuth velocity, a r represents the acceleration in the distance direction.
3. The phase tracking-based high-order range migration correction method for SAR moving targets according to claim 2, characterized in that: The method of reducing bandwidth and widening the range migration tolerance within the sub-band by segmenting the range frequency domain so that the residual high-order RCM falls into the same range unit includes: The segmented echo signal is expressed as: Among them, N r Indicates the number of frequency domain segments, the bandwidth of the sub-band signal is reduced by N r times, the corresponding distance resolution is reduced by N r times.
4. The phase tracking-based high-order range migration correction method for SAR moving targets according to claim 3, wherein: The extended Kalman filter is introduced, and the state space equation of the phase history is constructed to track the signal phase history of the unit where the target is located, including: The RCM compensation factor is constructed by tracking the signal phase history of the range unit where the sub-band signal target is located; The signal of the distance unit where the target is located in the sub-band is: Among them A s represents the signal amplitude, then the Doppler phase history generated by the relative motion between the radar and the target is: Phase φ(η a ) satisfies the smoothness assumption, that is, the phase difference between adjacent pulse sampling points does not fluctuate violently. Its mathematical expression is: φ(k+1)-φ(k)≈φ(k)-φ(k-1) Where k is the azimuth sampling index, and k = 1, 2, ..., N a , N a represents the number of sampling points in azimuth, φ(k+1) and φ(k) represent the phases at the k+1th and kth sampling moments, respectively; In order to obtain the phase history, EKF is used for phase tracking, and the EKF state space equation representing the phase change relationship is established: First, the three adjacent phase sampling points constitute the state vector, then the state vector corresponding to the (k-1) and k sampling moments is and They are: in(·) T represents transpose; According to the recursive relationship between state vectors, the state equation is as follows: Wherein, B is the state transition matrix, and B=[0, 1, 0; 0, 0, 1; 0, 1, 2]; The observation vector is constructed as follows: Among them, s k Represents the sampling of observation values between three adjacent pulses, n r (k) and n i (k) represents the real and imaginary sampling of the observation noise, respectively; Considering the influence of noise terms in practice, the EKF state space equation for phase tracking is constructed as follows: Among them, ω k-1 Represents the process noise vector, which describes the approximate error of the smoothing assumption. The corresponding process noise covariance matrix is recorded as W k-1 , n k represents the observation noise vector, and n k =[n r (k), n i (k)] T , the corresponding observation noise covariance matrix is recorded as R k , The expression is In tracking phase At the same time, the amplitude A s (k) Perform tracking, that is, construct a binary state space equation: The amplitude state vector at sampling time k is denoted as p k , construct the state equation corresponding to the amplitude, and set the constant amplitude A s Use p k Substituting, we obtain the binary state space equation as follows: Among them, the amplitude state vector p k It is composed of three adjacent amplitude sampling values, p k =[A s (k-1), A s (k), A s (k+1)] T , v k Represents the noise vector, and its corresponding noise covariance matrix is recorded as V k-1 ; In addition, in the observation equation, ⊙ represents Hadamard product, and h1(p k )=[p k , p k ] T , h1(p k )and The corresponding Jacobian matrices are H 1k and H 2k ,pass For p k and Obtain the partial derivative; Processing binary state equations through ADMM algorithm; Solve the binary state space equation by EKF and get the phase state vector After traversing all sampling points, the Doppler phase history φ(η a ) 5. The phase tracking-based high-order range migration correction method for SAR moving targets according to claim 4, characterized in that: In obtaining the Doppler phase history φ(η a ) Then, calculate the estimated distance between the radar platform and the target for: is an estimate of the Doppler phase history; Construct a high-order phase compensation function H2(f r , t a ),as follows: Perform IFFT along the range direction to obtain the results of high-order parameter estimation and range migration correction for the full-band data, as follows: At this point, high-order parameter estimation and range migration correction are completed.
Citation Information
Patent Citations
High-resolution wide-swath SAR frequency domain NLCS imaging method based on frequency domain correction model
CN111337922A
SAR weak maneuvering target imaging method based on adaptive phase tracking
CN116299458A
Method and device for synthethic aperture radar imaging based on non-linear frequency modulation signal
EP3144702A1
