A multi-pass SAR three-dimensional imaging self-focusing method

Abstract

The present application relates to a kind of multi-pass SAR three-dimensional imaging self-focusing method, in particular to a kind of multi-pass track error estimation and compensation method based on error dimension reduction and track-target position joint estimation, belong to radar three-dimensional imaging residual signal error estimation and compensation technical field, can effectively realize SAR three-dimensional imaging signal data processing, provide help for obtaining multi-pass tomographic SAR system high-resolution three-dimensional image.This method uses the information of multiple point targets, can effectively estimate and compensate the spatial variation error in large scene;And this method does not depend on the track measured by inertial navigation system, only depends on the echo information collected by radar system, which has advantages for radar system without inertial navigation, after the estimation and compensation of track error by this method, the image with good focusing effect can be obtained.

Description

Technical Field

[0001] This invention relates to a multi-pass SAR three-dimensional imaging autofocus method, and more particularly to a multi-pass track error estimation and compensation method based on error dimensionality reduction and track-target position joint estimation. It belongs to the field of radar three-dimensional imaging residual signal error estimation and compensation technology, and can effectively realize SAR three-dimensional imaging signal data processing, providing assistance for obtaining high-resolution three-dimensional images of multi-pass tomographic SAR systems. Background Technology

[0002] Synthetic Aperture Radar (SAR) is a high-resolution microwave imaging radar capable of detecting illuminated areas around the clock and in all weather conditions, acquiring high-resolution radar images of those areas. SAR systems are typically mounted on moving platforms, achieving high range resolution through transmitting and receiving broadband signals and pulse compression, and high azimuth resolution through platform motion and synthetic aperture imaging processing. Therefore, SAR can obtain two-dimensional high-resolution images of the observed scene.

[0003] While slow-moving unmanned platform SAR two-dimensional imaging technology is relatively mature and has some applications, it is still constrained by two main factors. First, current SAR image resolution and swath width are limited: the narrow relative bandwidth of current SAR systems limits range resolution, especially at lower frequencies. Narrow relative bandwidth translates to narrow signal bandwidth, resulting in lower range resolution. Furthermore, the narrow beamwidth of current SAR systems leads to lower azimuth resolution and a narrower imaging swath. Second, current SAR images suffer from limited imaging dimensionality: SAR two-dimensional images inherently suffer from overlay, geometric distortion, and missing dimensions, failing to accurately reflect the true three-dimensional scattering information of objects, leading to difficulties in image identification and application.

[0004] Multi-pass tomographic SAR acquires multi-baseline 3D imaging data by creating a third-dimensional aperture through multiple passes at different altitudes along the same flight path. Its resolution is not limited by the real array and it boasts the advantage of high tomographic resolution. However, the motion trajectories of each pass in multi-pass tomographic SAR on slow-moving unmanned platforms are complex, and there are no rigid body constraints between passes. Therefore, it is affected by both azimuth defocus due to complex trajectory errors within the passes and tomographic defocus due to time-varying baseline errors between passes. Existing autofocusing algorithms cannot detect low-order errors between passes, thus failing to effectively process multi-pass tomographic SAR data. Therefore, in-depth research is needed on autofocusing 3D imaging algorithms for multi-pass tomographic SAR on slow-moving unmanned platforms.

[0005] To date, scholars such as Feng Dong have proposed a three-dimensional autofocus method for multi-pass tomographic SAR, applying traditional autofocus phase error estimation methods to estimate phase errors between multi-pass data, thus achieving phase calibration between multiple passes in spaceborne multi-pass SAR. These methods all assume that only time-invariant phase errors exist between passes, neglecting the first, second, and higher-order errors that remain with azimuth and time variations from two-dimensional autofocus, and assuming these errors to be spatially invariant. However, in slow-moving unmanned platform multi-pass tomographic SAR, the two-dimensional images obtained through autofocusing of two-dimensional images not only contain constant-order errors but also time-varying errors, namely first, second, and higher-order errors, leading to defocusing of the three-dimensional image. Furthermore, the impact of track errors on the entire scene is spatially variable, especially under wide-beam conditions, where this spatial variation cannot be ignored. Therefore, the above methods are not applicable to autofocus processing in slow-moving unmanned platform multi-pass tomographic SAR. Tebaldini S. et al. proposed a tomographic autofocusing method based on phase center dual positioning, which can correct time-varying vertical heading track errors in multi-flight tomographic data. However, this method cannot adapt to defocusing and geometric distortion caused by heading track errors, and therefore cannot achieve good 3D focusing for slow-speed unmanned platforms in multi-flight tomographic SAR. Therefore, the problem of autofocusing 3D imaging for slow-speed unmanned platforms in multi-flight tomographic SAR remains to be solved. Summary of the Invention

[0006] The technical problem solved by this invention is to overcome the shortcomings of existing technologies and propose a multi-pass SAR three-dimensional imaging autofocusing method. Specifically, it relates to a multi-pass track error estimation and compensation method based on error dimensionality reduction and joint track-target position estimation, belonging to the field of radar three-dimensional imaging residual signal error estimation and compensation technology. It can effectively realize SAR three-dimensional imaging signal data processing and provides assistance in acquiring high-resolution three-dimensional images of multi-pass tomographic SAR systems.

[0007] The method of this invention is achieved through the following technical solution:

[0008] A multi-pass SAR three-dimensional imaging autofocus method, the steps of which include:

[0009] Step one: The two-dimensional self-focusing track inversion method is used to estimate and compensate for the track errors of each orbit, specifically as follows:

[0010] Let the coordinates of any ground point target within the radar beam coverage area be (x0, y0), then the azimuth time is t. a The slope distance error at that time is ΔR(t) a (x0, y0), azimuth time is t a The trajectory error at time ΔT(t) a )=[Δx(t a ),Δy(t a ),Δz(ta )] T , where Δx(t) a ) represents the azimuth error, Δy(t) a ) represents the track error in the ground distance direction, Δz(t) a The slant range error is the trajectory error in the celestial direction. The relationship between the slant range error and the trajectory error is as follows:

[0011] ΔR(t a ;x0,y0)=-cosβ(t a ;x0,y0)·sinθ(t a ;x0,y0)Δx(t a )-cosθ(t a ;x0,y0)Δy(t a )+sinβ(t a ;x0,y0)·sinθ(t a ;x0,y0)Δz(t a (1)

[0012] Wherein, β(t) a x0, y0) is the instantaneous ground rub angle, θ(t) a x0, y0) is the instantaneous forward oblique angle, which together form the line of sight direction for radar observation of ground point targets;

[0013] Suppose there are M ground point targets within the radar beam coverage area, and the position vector of the m-th ground point target is P. m =[x m ,y m ,z m ] T The relationship between the slant range error and the track error of the M ground point targets is expressed as:

[0014] R E (t a )=M(t a )·ΔT c (t a (2)

[0015] Among them, R Ε (t a Let M(t) be a vector consisting of the slant range errors of M ground point targets. a () represents the line-of-sight vectors for M ground point targets;

[0016] Substituting formula (1) into formula (2) yields:

[0017] R E (t a )=[ΔR(t a ;x1,y1) ΔR(ta ;x2,y2) … ΔR(t a ;x M ,y M )] T (3)

[0018]

[0019] Using weighted least squares method to determine the position and time t a The three-dimensional track error ΔT at that time c (t a The solution is performed to complete the estimation and compensation of the track error for each orbit.

[0020] Step two involves jointly estimating and compensating for the low-order errors of the residual track and the target position. The specific method is as follows:

[0021] Assuming the unmanned platform flew a total of N tracks, n = 1, 2, 3, ..., N, then the nth track is:

[0022]

[0023] Among them, T i,n (t a ) represents the ideal trajectory of the radar flight on the nth orbit; ΔT c,n (t a ) represents the track error ΔT(t) of the nth orbit. a The compensation result, ΔΔT cl,n (t a Let be the current value of the low-order track error of the nth track, expressed as: A n B is a constant term. n C represents the coefficient of the linear term. n Let be the coefficient of the quadratic term. Then, the echo of the m-th ground point target after deslant removal is expressed as:

[0024]

[0025] Where c is the speed of light, f c For bandwidth, f a f is the azimuth Doppler frequency. r For the range frequency, B r For bandwidth, It is a descrambling function. This is the current value of the slant range history;

[0026] A three-dimensional image is obtained from the echo after deskewing. When constructing the slant range history using the correct target point location and the correct low-order track error, the three-dimensional image I... mThis allows for the achievement of the maximum coherent superposition effect, meaning the energy accumulation of the point target reaches its maximum value. for:

[0027]

[0028] Therefore, the joint estimation problem is expressed as:

[0029]

[0030] To achieve a faster solution to the optimization problem, the Particle Swarm Optimization (PSO) algorithm is adopted. Furthermore, PSO escapes local minima and obtains the global optimum. The fast solution method for this optimization problem is as follows:

[0031] Step 1: Select the local image where the feature points are located and obtain the pulse compression echo of all feature points;

[0032] Step 2, according to A all B all C all ,P all Constructing the deslope function H from the current value deramp,m After deskewing, the deskewing echo S is obtained. deramp,m ;

[0033] Step 3: According to equation (7), calculate the cumulative amplitude value I at each target point. p,m Thus, the energy accumulation index is obtained.

[0034] Step 4, estimate the low-order track error parameters, the specific method is as follows:

[0035] For the nth track among N tracks, the optimal low-order track error A is estimated using PSO. n B n C n And update H deramp,m S deramp,m and

[0036] Step 5, estimate the target location, the specific method is as follows:

[0037] For the m-th target point among M ground point targets, PSO is used to estimate the target point's location. And update H deramp,m S deramp,m and

[0038] Step 6: Repeat steps 2-5 until the change in the energy accumulation index is less than the threshold, and obtain the estimation result ΔΔT of the low-order track error. cl,n (t aThe estimation results of the target location.

[0039] Step 7, Estimation of the low-order track error ΔΔT cl,n (t a The estimation results of the target location. A joint solution is performed to estimate and compensate for the low-order errors of each residual track and the target position, resulting in the compensated multi-flight tomographic SAR data and the estimated target position.

[0040] Step 3: Estimate and compensate for the higher-order errors of the residual track. The specific method is as follows:

[0041] After compensation in step two, the estimated results for the M target points are as follows: The orientation of N flight paths is given by time t. a The higher-order error of the residual track at that time is ΔΔT ch,n (t a N azimuth directions with time t a The slant distance error at that time is The relationship between slant range error and track error is as follows:

[0042] R ch,E (t a ) = M ch (t a )·ΔΔT ch,n (t a (9)

[0043] Among them, R ch,E (t a Let M be the vector consisting of the slant range errors of M point targets. ch (t a () represents the line-of-sight vectors for M point targets;

[0044] Substituting formula (1) into formula (9) yields

[0045]

[0046]

[0047] in, To wipe the corner of the ground instantly, The instantaneous forward angle, together with the other two, forms the line-of-sight direction for radar observation of ground point targets;

[0048] Using weighted least squares method to determine the position and time t a The higher-order error of the residual track at that time is ΔΔT ch,n (t a The solution is performed to estimate and compensate for the higher-order errors of the residual track.

[0049] To ensure that the residual error is negligible, the method in steps two and three is iterated, with 1 to 3 iterations to obtain the high-order estimation results of the residual track error and the compensated multi-flight tomography SAR data;

[0050] Step 4: Image the multi-flight tomography SAR data after compensation in Step 3;

[0051] The obtained tracks were used for 3D imaging, resulting in good 3D imaging effects. Finally, the complete track estimation results were used. Three-dimensional fine imaging was performed, and the three-dimensional BP algorithm was used to observe the autofocus imaging effect.

[0052] Beneficial effects

[0053] Compared with existing UAV SAR 3D imaging methods, this method uses information from multiple point targets, which can effectively estimate and compensate for spatial variation errors in large scenes. Furthermore, this method does not rely on the track measured by the inertial navigation system, but only on the echo information collected by the radar system. This has advantages for radar systems without inertial navigation. After estimating and compensating for track errors using this method, images with excellent focusing effects can be obtained. Attached Figure Description

[0054] Figure 1 The overall flowchart of the multi-flight tomographic SAR autofocus imaging method is shown below;

[0055] Figure 2 The relative track errors of each track in the simulation experiment (excluding constant and linear term errors; for ease of observation, only tracks 1, 11, 21, and 30 are shown).

[0056] Figure 3 The three-axis trajectory error for each flight path in the simulation experiment;

[0057] Figure 4 Point cloud map, representing the result of a raster 3D imaging.

[0058] Figure 5 This is a diagram of the experimental scene.

[0059] Figure 6 The three-dimensional imaging results are based on measured data, and the relative elevation within the scene is used for coloring. Detailed Implementation

[0060] The implementation of the method of the present invention will be described in detail below with reference to the accompanying drawings and embodiments.

[0061] The multi-flight tomographic SAR autofocusing imaging method includes the following steps:

[0062] 1. Two-dimensional self-focusing track inversion for error estimation and compensation of each track;

[0063] 2. Joint estimation and compensation of low-order errors in residual tracks and target positions;

[0064] 3. Estimation results and compensation for higher-order errors in residual tracks;

[0065] 4. Imaging using multi-flight tomographic SAR data.

[0066] Example

[0067] This section uses scene simulation to first compare the channel error estimation results with and without motion error compensation, and then compares the three-dimensional imaging results before autofocus, two-dimensional autofocus track inversion, and multi-way tomographic SAR track inversion.

[0068] Table 2 Parameters of the Azimuth Narrow Beam Distributed SAR Simulation System

[0069]

[0070] In the simulation, the interval between each pass of the ideal trajectory is 3m, and the ideal trajectory is a uniform straight line trajectory with a velocity of 8m / s in the positive Y-axis direction. Based on this, constant, first-order, and higher-order trajectory errors are added to each track. Unknown random constant-order trajectory errors and three-axis velocity errors (first-order trajectory errors) are also added to each track. The actual collected UAV trajectory is used as the true relative trajectory error for each track (excluding constant and first-order errors), such as... Figure 2 As shown, the relative track errors of each flight path exhibit randomness. The three-axis position error (constant order error) is a uniformly distributed random number between -0.8 and 0.8 m, while the three-axis velocity error is a uniformly randomly distributed random number between -0.05 and 0.05 m / s. The added three-axis position and velocity errors are as follows: Figure 3 As shown by the blue square, in the simulation, only the ideal trajectory is assumed to be known. The three-axis constants, first-order errors, and higher-order errors of each trajectory are all unknown. Therefore, the multi-passage data under this trajectory contains unknown, complex, time-varying trajectory errors.

[0071] Since the trajectory is fully sampled and ideally has a uniform baseline, the three-dimensional backpropagation algorithm can be used for three-dimensional imaging processing of the entire scene. Figure 3 The red asterisks in the diagram represent the error estimation results of the proposed algorithm, demonstrating that the algorithm can correctly infer the track errors of each order. The 3D imaging results are shown below. Figure 4 (a) to (c) are point cloud images of the three-dimensional imaging results of the non-autofocused, two-dimensional autofocused track inversion, and the proposed algorithm, respectively. It can be seen that the processing results of the non-autofocused and two-dimensional autofocused track inversion are significantly defocused, while the proposed algorithm achieves good three-dimensional focusing after autofocusing.

[0072] To further verify the effectiveness of the multi-flight tomographic SAR autofocus imaging algorithm, an equivalent verification experiment was conducted in Chongqing using a self-developed UAV-borne ultra-wideband / ultra-widebeam SAR system to image a building scene. Figure 5 This is an optical image of the experimental scene. Figure 6 The three-dimensional imaging results of each method are shown in point cloud maps, which are colored with the height of the scattering points. Figure 6 (a) shows the imaging result before autofocusing. Figure 6 (b) is the imaging result after using traditional two-dimensional autofocus processing. Figure 6 (c) shows the imaging result after processing by the proposed algorithm. As can be seen from the figure, the autofocusing results in both the front and height dimensions are significantly defocused. Even after the two-dimensional autofocusing track inversion, the height dimension is still significantly defocused. However, the proposed algorithm achieves good three-dimensional focusing effect for the entire scene after autofocusing.

[0073] Furthermore, the three-dimensional image entropy of the full-scene three-dimensional imaging results was evaluated. Before autofocus, the full-scene three-dimensional image entropy was 20.6988. After processing with traditional two-dimensional autofocus, the full-scene three-dimensional image entropy was improved to 20.3046. After processing with the proposed algorithm, the full-scene image entropy was further improved to 18.4591. It can be seen that the proposed algorithm has the best image entropy index, which proves the effectiveness of the proposed method.

[0074] In summary, the above are merely preferred embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A multi-pass SAR three-dimensional imaging autofocusing method, characterized in that... The steps of this method include: Step 1: The two-dimensional self-focusing track inversion method is used to estimate and compensate for the track errors of each track. Step 2: Jointly estimate and compensate for the low-order errors of the residual track and the target position; Step 3: Estimate and compensate for the higher-order errors of the residual tracks to obtain compensated multi-flight tomographic SAR data; Step 4: Perform three-dimensional imaging on the multi-flight tomography SAR data after compensation in Step 3; In step one, the specific method for estimating and compensating for the track errors of each orbit using the two-dimensional self-focusing track inversion method is as follows: Let the coordinates of any ground point target within the radar beam coverage area be... Then the location and time are The slant distance error at that time is Location and time are The trajectory error at that time was ,in, This represents the trajectory error in the azimuth direction. This represents the track error in the ground distance direction. The relationship between the slant range error and the trajectory error is as follows: (This refers to the trajectory error in the sky direction.) （1） in, To wipe the corner of the ground instantly, The instantaneous forward angle, together with the other two, forms the line-of-sight direction for radar observation of ground point targets; Suppose there are M ground point targets within the radar beam coverage area, and the position vector of the m-th ground point target is... The relationship between the slant range error and the track error of the M ground point targets is expressed as: （2） in, Let be a vector consisting of the slant range errors of M ground point targets. Let M be the line-of-sight vectors for M ground point targets; Substituting formula (1) into formula (2) yields: （3） （4） Using weighted least squares method to determine the position and time Three-dimensional track error at time The solution is performed to complete the estimation and compensation of the track error for each orbit.

2. The self-focusing method for multi-pass SAR three-dimensional imaging according to claim 1, the method comprising the following steps: In step two, the specific method for jointly estimating and compensating for the low-order error of the residual track and the target position is as follows: Assuming the unmanned platform flew N tracks in total, n=1, 2, 3, ..., N, then the nth track is: （5） in, The ideal trajectory for radar flight on the nth orbit; The track error of the nth orbit The compensation results Let be the current value of the low-order track error of the nth track, denoted as A n B is a constant term. n C represents the coefficient of the linear term. n Let be the coefficient of the quadratic term. Then, the echo of the m-th ground point target after deslant removal is expressed as: （6） in, At the speed of light, For bandwidth, The azimuth Doppler frequency, For range frequency, For bandwidth, It is a descrambling function. This is the current value of the slant distance history; A three-dimensional image is obtained from the echo after deskewing. When constructing the slant range history using the correct target point location and the correct low-order track error, the 3D image... This allows for the achievement of the maximum coherent superposition effect, meaning the energy accumulation of the point target reaches its maximum value. for: （7） Therefore, the joint estimation problem is expressed as: （8）。 3. The self-focusing method for multi-pass SAR three-dimensional imaging according to claim 2, the method comprising the following steps: To achieve a faster solution, a particle swarm optimization algorithm is used to solve the optimization problem, escaping local minima and obtaining the global optimum. The solution method for this optimization problem is as follows: Step 1: Select the local image where the feature points are located and obtain the pulse compression echo of all feature points; Step 2, according to Constructing a descrambling function based on the current value After deskewing, the deskewing echo is obtained. ; Step 3: According to equation (7), calculate the cumulative amplitude value at each target point. Thus, the energy accumulation index is obtained. ; Step 4, estimate the low-order track error parameters, the specific method is as follows: For the nth track among N tracks, PSO is used to estimate the optimal low-order track error. And update , and ; Step 5, estimate the target location, the specific method is as follows: For the m-th target point among M ground point targets, PSO is used to estimate the target point's location. and update , and ; Step 6: Repeat steps 2-5 until the change in the energy accumulation index is less than the threshold, and obtain the estimation result of the low-order track error. and the estimation results of the target location ; Step 7: Estimation results of low-order track errors and the estimation results of the target location A joint solution is performed to estimate and compensate for the low-order errors of each residual track and the target position, resulting in the compensated multi-flight tomographic SAR data and the estimated target position. .

4. The multi-pass SAR three-dimensional imaging autofocusing method according to claim 3, the method comprising the following steps: In step three, the specific method for estimating and compensating for the higher-order error of the residual track is as follows: After compensation in step two, the estimated results for the M target points are as follows: The time of N flight paths in azimuth direction is The higher-order error of the residual track at that time is The time of N flight paths in azimuth direction is The slant distance error at that time is The relationship between slant range error and track error is as follows: （9） in, Let be the vector composed of the slant range errors of M target points. Let M be the line-of-sight vectors for M point targets; Substituting formula (1) into formula (9) yields （10） （11） in, To wipe the corner of the ground instantly, The instantaneous forward angle, together with the other two, forms the line-of-sight direction for radar observation of ground point targets; Using weighted least squares method to determine the position and time The higher-order error of the residual track at that time is The solution is performed to estimate and compensate for the higher-order errors of the residual track.

5. A multi-pass SAR three-dimensional imaging autofocusing method according to claim 4, the method comprising the following steps: The high-order estimation results of residual track error and the compensated multi-flight tomographic SAR data were obtained by iterating 1 to 3 times.

6. A multi-pass SAR three-dimensional imaging autofocusing method according to claim 5, the method comprising the following steps: In step four, the method for performing three-dimensional imaging on the multi-flight tomography SAR data compensated in step three is as follows: The obtained tracks are used for 3D imaging to obtain 3D imaging results. Finally, the complete track estimation results are used. Three-dimensional fine imaging was performed, and the three-dimensional BP algorithm was used to observe the autofocus imaging effect.

Images