Ultrahigh-resolution FMCW (frequency modulated continuous wave) system satellite-borne sliding bunching mode SAR (synthetic aperture radar) imaging method

Abstract

The invention discloses an ultrahigh-resolution FMCW (Frequency Modulated Continuous Wave) system satellite-borne sliding bunching mode SAR (Synthetic Aperture Radar) imaging method. The method comprises the following steps: receiving reflected signals of sub-blocks of an area to be imaged; performing Fourier transform on the fast time in the reflected signal, then performing phase compensation on the slow time, and converting the obtained signal into a virtual equivalent echo signal after direct sampling and matched filtering; pre-filtering in the range direction is carried out; flipping and translating in a frequency domain fi domain into a translation amount under a unified reference slant distance; carrying out fast time domain time domain symbol and proportion replacement; converting the distance time domain into a frequency domain, and performing down-sampling; imaging the downsampled signal to obtain a sub-block image; fusing the plurality of sub-block images to obtain an SAR image of the to-be-imaged region; through the spaceborne ultrahigh-resolution SAR fast imaging method based on range direction frequency domain downsampling, the operation amount of block imaging processing can be greatly reduced.

Description

Technical Field

[0001] The invention belongs to the technical field of microwave remote sensing, and in particular relates to an ultra-high resolution FMCW system satellite-borne sliding spotlight mode SAR imaging method. Background Art

[0002] The strip mode is the most basic working mode of spaceborne SAR. Its advantage is that it can continuously image a large area. However, due to limitations such as antenna gain, the azimuth resolution of the radar cannot be increased arbitrarily as the antenna length decreases. Its azimuth resolution cannot be higher than half of the azimuth antenna size.

[0003] In the spotlight mode, as the platform moves, the radar beam is always illuminated to a small area on the ground by controlling the direction of the beam. This beam pointing control increases the synthetic aperture time of the target, thereby obtaining a higher azimuth resolution. The disadvantage is that the imaging area in azimuth is small and cannot be continuously imaged.

[0004] The sliding beam mode is a more general mode between the strip mode and the spotlight mode. Compared with the traditional strip mode, it has a higher spatial resolution and a larger azimuth mapping width. The spotlight and strip modes can be regarded as special cases of the sliding beam mode. When the moving speed of the irradiated area on the ground is zero, it is the spotlight imaging mode; when the moving speed of the irradiated area on the ground is the speed of the load platform, it is the strip imaging mode.

[0005] Traditional sliding beam mode SAR often adopts pulse transmission system; full aperture whole image echo data serial processing; unified reference slant range and unified coordinate system algorithm processing; imaging geometry model based on straight track.

[0006] Traditional X-band and C-band spaceborne sliding-beam high-resolution SAR often uses the whole-block imaging processing method when the azimuth width is small. When entering the high-frequency band (such as Ka-band), due to the high frequency, when the system pulse repetition frequency (PRF) is fixed, the maximum Doppler bandwidth that can be tolerated by the imaging processing will be much lower than that of the X-band or C-band, and thus the instantaneous illumination range of the antenna is also much lower than that of the X-band or C-band.

[0007] Therefore, for high-frequency (such as Ka-band) satellite-borne sliding spotlight high-resolution SAR imaging that requires a large azimuth width, if the traditional whole-block imaging processing method is used, the number of azimuth points will be huge, which will cause a sharp increase in the amount of calculation. On the other hand, in satellite-borne ultra-high-resolution sliding spotlight mode SAR imaging, due to the long synthetic aperture length, the traditional straight track approximation model has a large error. In order to reduce the spatial variation error caused by the track curvature, block processing is also required in the range direction. Summary of the invention

[0008] The purpose of the present invention is to provide an ultra-high resolution FMCW system satellite-borne sliding spotlight mode SAR imaging method to reduce the computational complexity of imaging processing.

[0009] The present invention adopts the following technical solution: an ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method, comprising the following steps:

[0010] receiving a reflection signal of a sub-block of the area to be imaged;

[0011] After Fourier transforming the fast time in the reflected signal, phase compensation is performed on the slow time to obtain the signal S if_com1 (f i ,t m ), where f i represents frequency, t m Indicates slow time;

[0012] S if_com1 (f i ,t m ) is converted into a virtual equivalent direct sampling matched filtered echo signal S if_com2 (f i ,t m );

[0013] For S if_com2 (f i ,t m ) to perform distance pre-filtering to obtain S if_com3 (f i ,t m );

[0014] S if_com3 (f i ,t m ) where the target is in the frequency domain f i The domain flip translation is the translation amount under the uniform reference slant distance, and S if_com4 (f i ,t m );

[0015] S if_com4 (f i ,t m ) is used to replace the time domain symbols and ratios in the fast time domain to obtain in, Indicates fast time;

[0016] Will The distance in the time domain is converted to the frequency domain to obtain s(ω,t m ), and for s(ω,t m) downsamples in the ω domain; wherein ω represents the angular frequency domain;

[0017] Imaging the downsampled signal to obtain a sub-block image;

[0018] Several sub-block images are fused to obtain the SAR image of the area to be imaged.

[0019] Furthermore, before receiving the reflection signal of the sub-block of the area to be imaged, the method further includes:

[0020] Calculate the distance deviation between the circular orbit of the SAR antenna and the approximately equivalent straight orbit at different apertures and different imaging points;

[0021] When the distance deviation within the imaging area of ​​the sub-block is not greater than the threshold, the imaging area of ​​the sub-block is determined.

[0022] Further, after receiving the reflection signal of the sub-block of the area to be imaged, the method further includes:

[0023] The imaging area of ​​the sub-block is compensated for the overall track curvature.

[0024] Furthermore, after the overall track curvature compensation is performed on the imaging area of ​​the sub-block, the following steps are further included:

[0025] The signal after overall track curvature compensation is subjected to variable reference slant distance de-slant reception.

[0026] Furthermore, for S if_com2 (f i ,t m ) to perform distance pre-filtering to obtain S if_com3 (f i ,t m )include:

[0027] S if_com3 (f i ,t m )=S if_com2 (f i ,t m )H filter (f i ),

[0028] Among them, H filter (f i ) is a filtering function, which retains the data of the range gate where the sub-block is located and sets the rest of the range gates to zero.

[0029] Further, The distance in the time domain to the frequency domain includes:

[0030]

[0031] in, Express exist domain to perform Fourier transform.

[0032] Furthermore, performing variable reference slant distance de-slant reception on the signal after overall track curvature compensation includes:

[0033]

[0034] in, The signal after overall track bending compensation, Represents the reference signal corresponding to the i-th reference slant range.

[0035] Furthermore, S if_com1 (f i ,t m ) is converted into a virtual equivalent direct sampling matched filtered echo signal S if_com2 (f i ,t m )include:

[0036]

[0037] Where c is the speed of light, f c Indicates the carrier frequency of the signal transmitted by the antenna, R Δ Represents the distance difference between the antenna and the target at the azimuth sampling position and the reference distance corresponding to the reference signal, R Δi =R t -R ref_i , R t Indicates the distance from the target to the antenna, R ref_i is the ith reference slope distance.

[0038] Furthermore, S if_com4 (f i ,t m ) Specifically:

[0039]

[0040] Where A represents the amplitude, T p represents the radar transmit pulse width, and γ represents the antenna frequency modulation slope.

[0041] The beneficial effects of the present invention are as follows: through a spaceborne ultra-high-resolution SAR rapid imaging method based on range-direction frequency domain downsampling, under a preprocessing framework of a variable reference slant range and a variable coordinate system, the overall pulse pressure is first analyzed in the range direction, and then blocks are performed in the pulse pressure domain. Then, pre-filtering in the range time domain is performed, and then downsampling in the range frequency domain is performed, thereby greatly reducing the amount of computation required for block imaging processing. BRIEF DESCRIPTION OF THE DRAWINGS

[0042] Figure 1 This is a schematic diagram of an ultra-high resolution FMCW system satellite-borne sliding spotlight mode SAR imaging method according to an embodiment of the present invention;

[0043] Figure 2 It is a schematic diagram of the instantaneous illumination geometry of sliding beam imaging in an embodiment of the present invention;

[0044] Figure 3 Schematic diagram of the sliding bunching geometric relationship in an embodiment of the present invention;

[0045] Figure 4 Schematic diagram of azimuth sub-block sub-aperture decomposition of sliding beam imaging in an embodiment of the present invention;

[0046] Figure 5 Schematic diagram of azimuth sub-block sub-aperture decomposition of sliding beam imaging in an embodiment of the present invention;

[0047] Figure 6 This is a flow chart of a parallel fast imaging processing method for Ka-band high-resolution and wideband SAR in an embodiment of the present invention;

[0048] Figure 7 Schematic diagram of squint geometric imaging in an embodiment of the present invention;

[0049] Figure 8 Schematic diagram of a distance migration curve after distance compression in a squint situation in an embodiment of the present invention;

[0050] Fig. 9 Schematic diagram of imaging of a variable reference coordinate system in an embodiment of the present invention;

[0051] Fig.10 Schematic diagram of the distance migration curve after distance compression in the variable reference coordinate system (equivalent distance movement correction) under the strabismus situation in an embodiment of the present invention. DETAILED DESCRIPTION

[0052] The present invention is described in detail below with reference to the accompanying drawings and specific embodiments.

[0053] The present invention proposes a parallel and fast imaging processing method for high-frequency band (such as Ka-band) satellite-borne high-resolution and wide-bandwidth SAR. By establishing a wide-bandwidth and high-resolution imaging geometric relationship and designing the instantaneous illumination range of the antenna, independent imaging blocks in azimuth and range are constructed, and finally parallel and fast imaging is completed on a GPU.

[0054] In the spaceborne sliding spotlight mode SAR, the traditional pulse transmission system requires high pulse power and high cost. By designing the FMCW system, the peak power and cost can be reduced. The traditional pulse system high-resolution (such as 0.5m) sliding spotlight mode SAR imaging usually adopts direct sampling reception and processing. When entering the ultra-high resolution (such as 0.05m) sliding spotlight mode SAR imaging, the FMCW de-slant (beat) reception system can reduce the large sampling data caused by direct sampling reception under large bandwidth.

[0055] The present invention proposes a virtual equivalent direct sampling method for the FMCW de-skew (beat) receiving system, by establishing a correlation transformation relationship between the echo signal after direct sampling and the signal after de-skew reception, and then the FMCW de-skew receiving data can be processed based on the imaging processing method of direct sampling.

[0056] In addition, in the case of satellite-borne ultra-high resolution de-slant receiving SAR imaging, due to the high resolution and long synthetic aperture length, if the traditional unified reference slant range processing method is used, the range sampling bandwidth will increase sharply, and then the sampling rate will increase sharply. Furthermore, under large squint, the data matrix brought by the ultra-high resolution range movement will increase sharply in the range direction.

[0057] The present invention proposes a large sampling data dimensionality reduction processing technology under space-borne large slant angle ultra-high resolution SAR imaging, and can greatly reduce the amount of sampling data through a sampling data dimensionality reduction preprocessing method that combines a variable coordinate system processing mode with a variable reference slant range.

[0058] In addition, for high-frequency ultra-high-resolution imaging, the shorter wavelength leads to a lower tolerance for distance blocking. Furthermore, the longer synthetic aperture length corresponding to high-frequency ultra-high resolution also has a lower tolerance for distance blocking. Therefore, it is necessary to increase the number of distance blocks in the echo data domain, but this also brings another problem of increased overall data computational complexity in the distance direction.

[0059] The present invention proposes an ultra-high resolution SAR fast imaging method based on range-direction frequency domain downsampling. Under the preprocessing framework of variable reference slant range and variable coordinate system, the overall pulse pressure is first analyzed in the range direction, and then blocks are performed in the pulse pressure domain. Then, pre-filtering is performed in the range direction in the time domain, and then downsampling is performed in the range direction in the frequency domain, which can greatly reduce the amount of computational complexity of block imaging processing.

[0060] The present invention discloses an ultra-high resolution FMCW system satellite-borne sliding spotlight mode SAR imaging method, such as Figure 1 As shown, the method comprises the following steps: receiving the reflection signal of the sub-block of the imaging area; performing Fourier transform on the fast time in the reflection signal and then performing phase compensation on the slow time to obtain a signal S if_com1 (f i ,t m), where f i represents frequency, t m Indicates slow time; S if_com1 (f i ,t m ) is converted into a virtual equivalent direct sampling matched filtered echo signal S if_com2 (f i ,t m ); if_com2 (f i ,t m ) to perform distance pre-filtering to obtain S if_com3 (f i ,t m );S if_com3 (f i ,t m ) where the target is in the frequency domain f i The domain flip translation is the translation amount under the uniform reference slant distance, and S if_com4 (f i ,t m );S if_com4 (f i ,t m ) is used to replace the time domain symbols and ratios in the fast time domain to obtain in, Expressing fast time; The distance in the time domain is converted to the frequency domain to obtain s(ω,t m ), and for s(ω,t m ) downsamples in the ω domain; wherein ω represents the angular frequency domain; imaging the downsampled signal to obtain a sub-block image; fusing several sub-block images to obtain a SAR image of the area to be imaged.

[0061] Through the spaceborne ultra-high-resolution SAR rapid imaging method based on range-direction frequency domain downsampling, in the preprocessing framework of variable reference slant range and variable coordinate system, the overall pulse pressure is first analyzed in the range direction, and then blocks are performed in the pulse pressure domain. Then, pre-filtering in the range time domain is performed, and then downsampling in the range frequency domain is performed, which can greatly reduce the amount of computational complexity of block imaging processing.

[0062] In SAR imaging, in order to improve the resolution in azimuth, the sliding beam mode can be used to synthesize a longer aperture by rotating the antenna angle to focus on a certain area. Figure 2 The figure shows the instantaneous illumination geometry of sliding beam imaging. In this mode, a longer synthetic aperture produces a larger Doppler bandwidth. If the PRF in strip mode is used, it will cause azimuth Doppler blur. To solve this problem, compression (Doppler blur removal) processing is required in azimuth before imaging. Although this preprocessing method can effectively reduce the PRF requirement, it also has a minimum PRF requirement, which is specifically constrained by the following formula:

[0063]

[0064] Among them, Δ uc is the spatial sampling step of the original data in the azimuth direction, R c is the slant distance of the center point of the imaging instantaneous illumination area, Y0 is the azimuth width of the imaging instantaneous illumination area, λ represents the wavelength of the electromagnetic wave, and v s It indicates the speed of satellite movement, and PRF is the pulse repetition frequency.

[0065] From formula 1, we can get:

[0066]

[0067] Therefore, in sliding beamforming mode, the minimum PRF is:

[0068]

[0069] Among them, λ min is the minimum wavelength of the electromagnetic wave bandwidth. Conversely, when the PRF is fixed, the maximum azimuth width of the instantaneous illumination area of ​​the imaging is:

[0070]

[0071] When the instantaneous irradiation width of the antenna is used as the total imaging width in the azimuth direction, the whole block imaging processing method can be used. When entering the high frequency band such as the Ka band, due to the high frequency, when the system PRF is fixed, the maximum Doppler bandwidth that the imaging processing can tolerate will be much lower than the X band or C band, and thus the instantaneous irradiation range of the antenna is also much lower than the X band or C band.

[0072] For the Ka band, when the orbit altitude is 800 km, the viewing angle is 30°, the scene center slant distance is 923.76 km, the carrier frequency is 35 GHz (wavelength 0.0086 m), the PRF is 3000 Hz, and the satellite speed is 7617 m / s, the maximum azimuth width of the allowed instantaneous imaging illumination area is 1.5453 km (i.e. ±0.7726 km) according to Formula 4. If the azimuth width requirement is 15 km to 20 km, the PRF needs to be increased to increase the instantaneous illumination range. After calculation, for the azimuth width of 15 km, the PRF needs to reach 30,000 Hz to meet the requirement, but a higher PRF will bring system design difficulties.

[0073] If a lower PRF is used, according to the sliding beam geometry, the ratio of the total azimuth width to the instantaneous illumination range is large, resulting in a longer total synthetic aperture length brought by the total azimuth width. Therefore, the total number of sampling points in azimuth increases dramatically. If the traditional whole-block imaging processing method is used, the amount of calculation increases dramatically.

[0074] In order to solve this problem, the present invention proposes a parallel fast imaging processing method for high-frequency spaceborne high-resolution wide-band SAR, which establishes a wide-band high-resolution imaging geometric relationship, designs the instantaneous illumination range of the antenna, and then constructs independent imaging blocks in azimuth, and finally completes parallel fast imaging on the GPU. The solution is described in detail below.

[0075] For equation 4, when the antenna azimuth size is determined, the maximum azimuth width of the instantaneous imaging illumination area is:

[0076]

[0077] Where D represents the aperture of the antenna.

[0078] Considering that the sliding beam mode gradually slides the above footprints and illuminates different footprints with different apertures, a larger azimuth width than one footprint can be obtained in the end. When the azimuth resolution is determined, the effective synthetic aperture length of each imaging point (referring to the imaging pixel position) is:

[0079]

[0080] Among them, r A is the azimuth resolution. According to Formula 6, to ensure that the azimuth resolution of each imaging point meets the requirements, Figure 3 As shown, the total synthetic aperture length of the entire imaging area is:

[0081]

[0082] Where L is the total synthetic aperture length of the entire imaging area, L0 is the effective synthetic aperture length of each imaging point, Y is the total width of the entire imaging area, and v s is the speed of the satellite, v f The speed at which the footprint moves.

[0083] When entering high-frequency bands such as the Ka band, after the above calculations, due to the large ratio of the total azimuth width to the instantaneous illumination range, the total synthetic aperture length brought by the total azimuth width is longer, and therefore the total number of sampling points in azimuth increases dramatically. If the traditional whole-block imaging processing method is used, the amount of calculation will increase dramatically.

[0084] like Figure 4As shown in the figure, considering the relative independence of each imaging footprint during the wide-band sliding process of the Ka band, and the relative independence of each imaging sub-aperture corresponding to each imaging footprint, the entire imaging area can be decomposed into independent sub-blocks and sub-apertures for separate imaging. The number of sampling points of each independent sub-block and sub-aperture will be greatly reduced, so the memory and computational complexity of each imaging sub-block are greatly reduced. At the same time, due to the relative independence of each imaging sub-block, the parallel architecture of the GPU is used to perform parallel processing on each imaging sub-block, and finally the stitching is completed in the image domain.

[0085] Due to the relative independence of each imaging sub-block, the final stitching in the image domain after the imaging of each imaging sub-block is completed will have a higher error tolerance than the stitching based on coherent superposition in the signal domain, so it is easier to complete the stitching in the image domain. Furthermore, block imaging can also reduce the impact of spatial variability in the imaging area and improve the accuracy of imaging.

[0086] In spaceborne ultra-high resolution sliding spotlight SAR imaging, due to the long synthetic aperture length, the influence of orbit curvature cannot be ignored. If a straight track model is directly used for approximation, it will cause a large error. Therefore, it can be solved by dividing the image into blocks in the range direction and compensating for the curved track error of each imaging sub-block. Figure 5 The diagram shown is a schematic diagram of a circular curved track.

[0087] Let the antenna sampling coordinates under the circular orbit be (X1, Y1, Z1), and the corresponding antenna sampling coordinates under the approximate straight line orbit be (X0, Y0, Z0). The X direction is the direction of the projection of the antenna on the ground toward the imaged area, the Y direction is the direction of the antenna's travel (i.e., the direction of the satellite's movement), and the Z direction is the direction of the projection of the antenna on the ground toward the antenna. The height reference of the imaging plane is used as the zero-point coordinate in the Z direction, and the middle position of the antenna in the synthetic aperture is used as the zero-point coordinate in the X and Y directions, so:

[0088]

[0089] Where H is the height of the satellite, R is the radius of the earth, and O is the center of the earth. look_angle is the viewing angle of the antenna, L is half the effective synthetic aperture length, and theta_curve is the geometric angle between the circular orbit and the center of the earth when the circular orbit is at the aperture at L. theta is the geometric angle between the approximate straight line orbit and the center of the earth. Let the coordinates of the imaging point be (x, y1, 0), then according to the calculation formula of the three-dimensional space coordinate system, the distance deviation (i.e. phase difference) between the circular orbit and the approximate equivalent straight line orbit at different apertures and different imaging point positions can be calculated.

[0090] That is to say, before receiving the reflected signal of the sub-block of the area to be imaged, the distance deviation between the circular orbit where the SAR antenna is located and the approximately equivalent straight-line orbit at different apertures and different imaging points is calculated; when the distance deviation in the imaging area of ​​the sub-block is not greater than the threshold (such as not exceeding 45°), the spatial variability can be ignored, thereby determining the imaging area of ​​the sub-block. Then, the imaging area of ​​the sub-block can be compensated for the overall orbit curvature, and the distance deviation value of the center point of the imaging area of ​​the sub-block is used as the reference to compensate all positions of the imaging area of ​​the sub-block.

[0091] According to the above principles, the block design in the range direction can be completed. In the imaging processing of each imaging sub-block, Doppler deambiguation must be performed first, followed by imaging processing. Figure 6 As shown, it is a flowchart for implementing the parallel fast imaging processing method for Ka-band high-resolution and wide-bandwidth SAR.

[0092] The following will describe in detail the virtual equivalent direct sampling method for the FMCW de-slanted (beat) receiving system proposed in the ultra-high resolution FMCW system satellite-borne sliding beam mode SAR imaging algorithm, the large sampling data dimensionality reduction processing technology under satellite-borne large squint angle ultra-high resolution SAR imaging, and the ultra-high resolution SAR fast imaging method based on range-direction frequency domain downsampling.

[0093] Assume that the received signal after azimuth segmentation and track bending compensation is:

[0094]

[0095] in, It indicates the received signal after azimuth segmentation and track bending compensation. Indicates the relative time in fast time, with the emission pulse as the reference time, t m Indicates slow time, R t represents the distance from the target to the antenna, c represents the speed of light, T p represents the radar transmit pulse width, f c represents the carrier frequency of the radar transmitting signal, t represents the absolute time in fast time, and γ represents the radar frequency modulation slope.

[0096] Since the bandwidth after de-slant reception depends on the distance difference between the radar and the target at the azimuth sampling position and the reference slant range corresponding to the reference signal, if the distance difference is larger, the corresponding bandwidth after de-slant reception will be larger. Therefore, reducing the distance difference can effectively reduce the bandwidth after de-slant reception, thereby reducing the subsequent sampling rate.

[0097] For each azimuth sampling point, the distance from the sampling point to the center of the imaging scene is used as the reference distance, that is, the reference slant distance of each azimuth sampling point is different. This variable reference slant distance processing method can reduce the sampling bandwidth of the range direction (the direction in which the antenna points to the target is called the range direction, and the direction in which the antenna moves forward is the azimuth direction) corresponding to each azimuth sampling point, thereby reducing the number of sampling points in the range direction and reducing memory consumption. Different reference slant distances are used for de-slanting, that is, the signal after the overall track curvature compensation is de-slanted by variable reference slant distance reception:

[0098]

[0099] Among them, R Δi =R t -R ref_i , R ref_i is the i-th reference slope distance, Represents the reference signal corresponding to the i-th reference slant range.

[0100] Fast time in formula 10 (Taking the time of the reference point as the benchmark, the time along the distance direction is the fast time, and the time along the azimuth direction is the slow time), perform Fourier transform and get:

[0101]

[0102] Where A represents the amplitude, f i represents frequency, S if (f i ,t m ) refers to the Fourier transform of the previous equation in the fast time domain.

[0103] In the slow time domain, that is, at each t m , perform phase compensation on equation 11, that is, multiply by get:

[0104]

[0105] The echo signal expression after direct sampling matched filtering is:

[0106]

[0107] Among them, A IQ is the amplitude value, B is the coefficient, R Δ Refers to the distance difference between the radar's distance to the target at the azimuth sampling position and the reference distance corresponding to the reference signal.

[0108] Therefore, it is necessary to study the conversion relationship from equation 12 to equation 13, and then transform the echo signal after de-slanting reception into a virtual equivalent echo signal after direct sampling matched filtering, and then the imaging processing algorithm after direct sampling matched filtering can be fully used for imaging processing. Comparing equations 12 and 13, it is necessary to first perform secondary phase compensation on equation 12 to make it equivalent to the phase at the unified reference slant range in equation 13. The compensation factor is set to So we have:

[0109]

[0110] Where c is the speed of light, f c Indicates the carrier frequency of the signal transmitted by the antenna, R Δ Represents the distance difference between the antenna and the target at the azimuth sampling position and the reference distance corresponding to the reference signal, R Δi =R t -R ref_i , R t Indicates the distance from the target to the antenna, R ref_i is the ith reference slope distance.

[0111] Then, for equation 14 at f i domain, filter the corresponding sub-block to be imaged in the entire range unit (this operation is the same as the following equation 25 after s(ω,t m ) is matched after the downsampling is completed in the ω domain to achieve the purpose of reducing the number of operation points and thus reducing the amount of operation), that is, the data of the range gate where the range imaging block is located is retained, and the remaining range gates are set to zero, that is:

[0112] S if_com3 (f i ,t m )=S if_com2 (f i ,t m )H filter (f i ) (15)

[0113] Among them, H filter (f i ) is the filtering function, and then the target in Equation 15 needs to be transformed into the frequency domain f i The domain flip translation is the translation under the uniform reference slant distance, which becomes:

[0114]

[0115] Among them, R Δ It refers to the distance difference between the radar's distance to the target at the azimuth sampling position and the reference distance corresponding to the reference signal.

[0116] When performing the above translation operation in the digital domain, the translation unit in the digital domain needs to be calculated, because:

[0117]

[0118] Where dt is the time domain sampling interval during de-slant reception, ω move is the digital angular frequency translation, k move is the translation of the digital unit, and N is the number of de-skew sampling points. Therefore:

[0119]

[0120] After the translation is completed, it is necessary to study the conversion relationship of Equation 16 to Equation 13 in the digital domain. For Equation 16, the translation amount corresponding to the unified reference slant distance is:

[0121]

[0122] For Equation 13, the translation corresponding to the unified reference slant distance is:

[0123]

[0124] dt' is the time domain sampling interval under virtual equivalent direct sampling, k' move_ref is the digital shift under virtual equivalent direct sampling. To make equation 13 and equation 16 equivalent, it is necessary to satisfy:

[0125] k move_ref =k' move_ref (twenty one)

[0126] So we have:

[0127]

[0128] At this point, the subsequent imaging processing can be performed according to the imaging algorithm under the time domain sampling interval dt' under the virtual equivalent direct sampling.

[0129] In the case of satellite-borne high-frequency band (such as Ka-band) ultra-high-resolution SAR imaging, Figure 7 As shown in Fig. 1, in order to obtain a SAR image with a wider width, it is necessary to perform block imaging through the forward squint, front side view and rear squint sliding beamforming modes. For any imaging point, the corresponding effective synthetic aperture length is relatively long. In the case of squint, when the imaging algorithm under the time domain sampling interval dt' under the above virtual equivalent direct sampling is used for imaging processing, the number of distance moving units is relatively large (such as Figure 8 As shown in Figure 2, the number of data processing matrices increases dramatically as the distance increases.

[0130] In order to reduce the number of processing units in the distance direction and thus reduce the amount of computation, such as Figure 8As shown in FIG. 1 , the reference coordinate system is adjusted accordingly with the change of the oblique angle to be the equivalent front view under the oblique angle. By projecting the azimuth sampling point onto a straight line perpendicular to the oblique angle, it becomes a virtual synthetic aperture, that is, the azimuth sampling point S is projected onto S'. Under the virtual synthetic aperture, the imaging area can be equivalent to the front view under the new coordinate system, which is equivalent to correcting the distance movement part (such as Fig.10 As shown in the figure, it is the distance migration curve after the correction and the distance compression. After the correction, the span of the entire migration unit is significantly reduced). Then, the imaging processing is performed using the positive side view imaging algorithm. After obtaining the imaging result, the imaging result is rotated according to the oblique angle, and finally the imaging result in the original coordinate system is obtained. This method can significantly reduce the memory consumption.

[0131] In the equivalent front side view processing, R in Equation 13 and Equation 16 Δ It will become R' in the equivalent positive side view, that is, the equivalent reference slant distance corresponding to the virtual projection sampling point S'. At the same time, the virtual sampling step becomes:

[0132] du'=du·cos(theta_c) (23)

[0133] After the coordinate system is transformed, the virtual sampling step and other transformed time-frequency variables are determined according to the coordinates of the sampling points after the geometric transformation. Then, the imaging algorithm under the original front-side view can be used for imaging.

[0134] Replace the symbols in equation 16 (replace f i Replace with This step does not perform calculations), and we get:

[0135]

[0136] Where A represents the amplitude, T p represents the radar transmit pulse width, and γ represents the antenna frequency modulation slope.

[0137] Next, each imaging sub-block needs to be imaged. When imaging each imaging sub-block, Doppler deblurring needs to be performed first, followed by imaging processing.

[0138] make Where k = ω / c, X c , Y c and R c Corresponding to the coordinates of the central reference point of the imaging sub-block, firstly, Equation 24 is transformed in the fast time domain (i.e. domain), perform Fourier transform, and get s(ω,t m ),Right now:

[0139]

[0140] in, Express exist domain to perform Fourier transform.

[0141] For ultra-high resolution imaging, in order to reduce the error of curved tracks, the number of range blocks will increase accordingly. If the same number of sampling points is used in the ω domain, the amount of data will increase dramatically. Next, we need to study the processing method to reduce the amount of data. As the number of range blocks increases, the range width of the corresponding imaging sub-block becomes smaller. Therefore, using this prior information, the sampling interval in the ω domain (i.e., the angular frequency domain) can be increased, that is, s(ω,t m ) is downsampled in the ω domain, and the downsampling multiple is equal to the multiple of the distance block, which can reduce s(ω,t m ) in the ω domain, thereby reducing the amount of computation. m domain) to obtain:

[0142]

[0143] For Equation 26 in the slow time domain (i.e., t m Domain) is transformed by Fourier transform, and after the transformation, we have:

[0144]

[0145] In k u The domain is subjected to zero-adding operation to obtain the up-sampling result, that is, S c (ω,k u ) becomes S cd (ω,k u ). cd (ω,k u ) in k u Do an inverse Fourier transform on the domain and get:

[0146]

[0147] Decompressing equation 28 yields:

[0148] s d (ω,t m )=s cd (ω,t m )s ref (ω,t m ) (29)

[0149] For Equation 29, at t m Doppler frequency shift correction is performed in the domain, namely:

[0150] s d1 (ω,t m )=s d(ω,t m )exp(-j*2*k c *sin(theta_c)*vt m ) (30)

[0151] theta_c is the oblique angle of the sub-block. When a variable coordinate system is used for processing, the oblique angle can be equivalent to 0.

[0152] Next, we will carry out imaging processing. m Domain) to do Fourier transform and get:

[0153]

[0154] For S in formula 31 11 (ω,k u ) in k u The domain is sampled, that is, down-sampled, and S(ω,k u ). In order to make the coordinate system centered at the reference point, let Then we have:

[0155]

[0156] make k y (ω,k u ) = k u , for S o (ω,k u ) to interpolate (after interpolation, F s Points can be combined with S o (ω,k u ) are basically consistent) to obtain:

[0157] F s =F[k x (ω,k u ),k y (ω,k u )] (33)

[0158] F s Perform a two-dimensional inverse Fourier transform to obtain the imaging result:

[0159]

[0160] The imaged result is rotated in the coordinate system, namely:

[0161] f(x,y)=rot(f(x,y) theta_c ) (35)

[0162] After imaging the imaging sub-blocks, the sub-blocks are then stitched together in the image domain to eventually complete the full-width image.

[0163] In summary, the present invention proposes a parallel and rapid imaging processing method for high-frequency band (such as Ka-band) spaceborne high-resolution and wide-bandwidth SAR. By establishing the geometric relationship of spaceborne wide-bandwidth and high-resolution imaging, the instantaneous illumination range of the antenna is designed, and then independent imaging sub-blocks in azimuth and range are constructed, and finally parallel and rapid imaging is completed on the GPU.

[0164] At the same time, the present invention proposes a virtual equivalent direct sampling method for the FMCW de-skewed (beat) receiving system. By establishing a correlation transformation relationship between the echo signal after direct sampling and the signal after de-skewed reception, the FMCW de-skewed reception data can be processed based on the imaging processing method of direct sampling.

[0165] Moreover, the present invention proposes a large sampling data dimensionality reduction processing technology under satellite-borne large slant angle ultra-high resolution SAR imaging. The sampling data dimensionality reduction preprocessing method combining a variable coordinate system processing method with a variable reference slant range can greatly reduce the amount of sampling data.

[0166] In addition, the present invention proposes a space-borne ultra-high resolution SAR rapid imaging method based on range-direction frequency domain downsampling. Under the preprocessing framework of variable reference slant range and variable coordinate system, the overall pulse pressure is first analyzed in the range direction, and then blocks are performed in the pulse pressure domain. Then, pre-filtering in the range direction is performed in the time domain, and then downsampling in the range direction frequency domain is performed, which can greatly reduce the amount of computational complexity of block imaging processing.

[0167] In high-frequency band (such as Ka-band) spaceborne sliding spotlight high-resolution SAR imaging with large azimuth width, if the traditional whole-block imaging processing method is used, the huge number of azimuth points will cause a sharp increase in the amount of calculation. The present invention proposes a parallel fast imaging processing method for high-frequency band (such as Ka-band) spaceborne high-resolution and large-width SAR, which can greatly improve the calculation efficiency.

[0168] Traditional pulse SAR usually adopts direct sampling reception processing, and when entering ultra-high resolution (such as 0.05m) sliding beam mode SAR imaging, the amount of sampling data increases dramatically. The virtual equivalent direct sampling method for FMCW de-slanted (beat) receiving system proposed in the present invention can significantly reduce the amount of sampling data. In the case of satellite-borne ultra-high resolution de-slanted receiving SAR imaging, due to the high resolution and long synthetic aperture length, if the traditional unified reference slant range processing method is adopted, the range sampling bandwidth will increase sharply, and then the sampling rate will increase sharply. Furthermore, under large squint, the data matrix brought by ultra-high resolution range movement will increase sharply in the range data amount. The large sampling data dimensionality reduction processing technology under satellite-borne large squint ultra-high resolution SAR imaging proposed in the present invention can greatly reduce the amount of sampling data. In addition, the ultra-high resolution SAR fast imaging method based on range frequency domain downsampling proposed in the present invention can greatly reduce the amount of calculation during block imaging processing.

Claims

1. An ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method, characterized in that: The following steps are involved: receiving a reflection signal of a sub-block of the area to be imaged; After Fourier transforming the fast time in the reflected signal, phase compensation is performed on the slow time to obtain a signal S if_com1 (f i ,t m ), where f i represents frequency, t m Indicates slow time; S if_com1 (f i ,t m ) is converted into a virtual equivalent direct sampling matched filtered echo signal S if_com2 (f i ,t m ); For S if_com2 (f i ,t m ) to perform distance pre-filtering to obtain S if_com3 (f i ,t m ); S if_com3 (f i ,t m ) where the target is in the frequency domain f i The domain flip translation is the translation amount under the uniform reference slant distance, and S if_com4 (f i ,t m ); S if_com4 (f i ,t m ) is used to replace the time domain symbols and ratios in the fast time domain to obtain in, Indicates fast time; Will The distance in the time domain is converted to the frequency domain to obtain s(ω,t m ), and for s(ω,t m ) downsamples in the ω domain; wherein ω represents the angular frequency domain; Imaging the downsampled signal to obtain a sub-block image; Several sub-block images are fused to obtain the SAR image of the area to be imaged.

2. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 1, characterized in that: Before receiving the reflection signal of the sub-block of the area to be imaged, the method further includes: Calculate the distance deviation between the circular orbit of the SAR antenna and the approximately equivalent straight orbit at different apertures and different imaging points; When the distance deviation within the imaging area of ​​the sub-block is not greater than a threshold, the imaging area of ​​the sub-block is determined.

3. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 2, characterized in that: After receiving the reflection signal of the sub-block of the area to be imaged, the method further includes: The imaging area of ​​the sub-block is compensated for the overall track curvature.

4. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 3, characterized in that: After the imaging area of ​​the sub-block is subjected to overall track curvature compensation, the method further includes: The signal after overall track curvature compensation is subjected to variable reference slant distance de-slant reception.

5. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to any one of claims 2 to 4, characterized in that: For S if_com2 (f i ,t m ) to perform distance pre-filtering to obtain S if_com3 (f i ,t m )include: S if_com3 (f i ,t m )=S if_com2 (f i ,t m )H filter (f i ), Among them, H filter (f i ) is a filtering function, which retains data for the range gate where the sub-block is located and performs zeroing operations on the remaining range gates.

6. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 5, characterized in that: Will The distance in the time domain to the frequency domain includes: in, Express exist domain to perform Fourier transform.

7. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 4, characterized in that: The variable reference slant distance de-slant reception of the signal after the overall track curvature compensation includes: in, The signal after overall track bending compensation, Represents the reference signal corresponding to the i-th reference slant range.

8. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 6, characterized in that: S if_com1 (f i ,t m ) is converted into a virtual equivalent direct sampling matched filtered echo signal S if_com2 (f i ,t m )include: Where c is the speed of light, f c Indicates the carrier frequency of the signal transmitted by the antenna, R Δ Represents the distance difference between the antenna and the target at the azimuth sampling position and the reference distance corresponding to the reference signal, R Δi =R t -R ref_i , R t Indicates the distance from the target to the antenna, R ref_i is the ith reference slope distance.

9. The ultra-high resolution FMCW system spaceborne sliding spotlight mode SAR imaging method according to claim 8, characterized in that: S if_com4 (f i ,t m ) Specifically: Where A represents the amplitude, T p represents the radar transmit pulse width, and γ represents the antenna frequency modulation slope.

Images