CN111551934A - A motion-compensated self-focusing method and device for unmanned aerial vehicle SAR imaging - Google Patents

Abstract

Aiming at the problem that there is a serious motion error in the echo data of the UAV airborne synthetic aperture radar, and the inertial navigation system cannot be used for high-resolution imaging, the invention discloses a motion compensation automatic method for UAV airborne SAR imaging. Focusing method and apparatus. The method of the present invention proposes an improved sub-aperture correlation algorithm, which has higher Doppler modulation frequency estimation accuracy than the existing sub-aperture correlation algorithm under the condition of obtaining sufficient distance space variation information, and can compensate more effectively Motion error with distance variation. In order to avoid the occurrence of singular points in the sub-aperture Doppler frequency modulation estimation due to the uniformity of the local scene or the absence of strong scattering points, the present invention adopts a random consistency method to effectively suppress the interference of the singular points on the imaging results. The present invention proves the effectiveness of the proposed method through the simulation experiment of the measured data.

Description

A motion-compensated self-focusing method and device for unmanned aerial vehicle SAR imaging

technical field

The invention relates to a motion compensation self-focusing method and device for unmanned aerial vehicle SAR imaging, and belongs to the technical field of radar imaging.

Background technique

Synthetic Aperture Radar (SAR) can work under all-weather conditions, using broadband signals to obtain high range resolution, and forming a virtual large aperture through platform motion to improve azimuth resolution, and finally achieve high-resolution two-dimensional imaging. Synthetic aperture radar imaging requires that the motion vector of the platform is constant during the synthetic aperture time. The spaceborne synthetic aperture radar usually meets the requirements, but the airborne synthetic aperture radar, especially the unmanned airborne synthetic aperture radar, is subject to unstable atmospheric conditions. Due to the influence of airflow and other factors, the moving platform deviates from the predetermined track, the antenna phase center shifts, and the flight attitude changes, that is, there is a serious motion error. How to perform motion compensation for UAV synthetic aperture radar system is a key technology of UAV synthetic aperture radar imaging. At present, the mainstream synthetic aperture radar motion compensation imaging technology abroad mainly uses high-precision inertial navigation system and global satellite positioning system to compensate for the stability of the carrier and antenna beams, and then processes the self-focusing imaging algorithm to obtain a relatively ideal synthetic aperture. Radar image. At present, the domestic UAV synthetic aperture radar can only be equipped with a small and lightweight low-precision inertial navigation system, which cannot realize the high-precision motion compensation of the UAV-borne synthetic aperture radar. In addition, in order to break through the enemy's radar line of defense, UAVs often use low-altitude flight methods. At this time, in order to obtain wide-scene synthetic aperture radar imaging, the radar must work at a low pitch angle, which will inevitably lead to strong distance-to-air variability in motion errors. It further increases the difficulty of high-precision synthetic aperture radar imaging. At present, in order to compensate the motion error of the airborne synthetic aperture radar platform, many scholars have proposed some motion compensation methods. [Bezvesilniy, O.O., I.M.Gorovyi, and D.M.Vavriv."Estimation of Phase Errors in SAR Data by Local-Quadratic Map-Drift Autofocus." International RadarSymposium 2012:376-381] proposed a motion error compensation method, which can compare It is good for compensating the motion error in the azimuth direction, but it cannot effectively compensate the motion error in the range direction. While M.Xing et al [M.Xing, X.Jiang, R.Wu, F.Zhou, and Z.Bao, "Motion compensation for UAV SAR based on raw radar data," IEEETrans.Geosci.Remote Sens.,vol.47, no.8, pp.2870–2883, Aug.2009] proposed a new motion compensation method to divide the azimuth direction into multiple sub-apertures, and then divide the range direction into blocks in each sub-aperture, using the existing sub-aperture correlation ( The Map-Drift, MD) algorithm estimates the local Doppler rate (Doppler Rate, DR), which can simultaneously compensate for the motion errors in the azimuth and range directions. However, after the range direction is divided into blocks, the Doppler modulation frequency of each range block is obtained separately, and each Doppler modulation frequency is fitted, which often brings many problems. If the number of distance blocks is small and each block has enough distance units, the sub-aperture correlation algorithm can obtain a more accurate solution, but cannot provide sufficient distance space variation information; when the number of blocks is large, although it can provide sufficient distance However, when the sub-block contains only a small number of distinctive points, the estimation accuracy is difficult to guarantee.

SUMMARY OF THE INVENTION

Purpose of the invention: In view of the above-mentioned problem of the number of range-wise blocks in the prior art, the purpose of the present invention is to propose a new motion compensation self-focusing method and device for synthetic aperture radar imaging based on the existing sub-aperture correlation algorithm. The invention can solve the problem that it is difficult to simultaneously obtain good Doppler modulation frequency estimation accuracy and sufficient distance space variation information in the prior art, and can effectively suppress the interference of singular points on the imaging results, and can be suitable for the small size of the unmanned aerial vehicle. , Synthetic aperture radar imaging that cannot carry a high-precision inertial navigation system and has serious motion errors in the synthetic aperture radar echo data.

Technical solution: In order to achieve the above purpose of the invention, a motion compensation self-focusing method for unmanned aerial vehicle SAR imaging according to the present invention includes the following steps:

(1) After performing coarse compensation, RCM correction, range-matched filtering, and De-ramping operations on the received echo signal using inertial navigation data, the obtained signal with residual phase error is subjected to azimuth block, and each sub-element is obtained. Aperture data;

(2) Dividing each sub-aperture in the distance direction, and selecting characteristic point samples in all distance units;

(3) Divide the distance unit where the characteristic point sample is located into two parts, and perform correlation processing to obtain the second-order phase error of the distance unit where it is located. Based on the linear change characteristics of the second-order error along the distance direction, use The data of the distance unit is calculated to obtain the second-order phase error coefficient, so as to estimate the second-order error in the distance unit without distinctive points;

(4) After the second derivative of the residual phase error of each sub-aperture of each distance unit is obtained, a random consistency algorithm is used to eliminate outliers;

(5) The residual phase error is obtained by the least square method and phase compensation is performed to improve the focusing effect of synthetic aperture radar imaging.

Preferably, in the step (3), the second-order phase error coefficient is calculated by the following formula:

为快时间，f表示多普勒频率，r_k表示挑选出的第k个距离单元的斜距，

表示对第k个距离单元前后半段回波序列多普勒谱值做相关运算，K表示挑选出有特显点的距离单元总数，|·|表示取绝对值运算。in,

is the fast time, f represents the Doppler frequency, r _k represents the slant range of the k-th range unit selected,

Indicates that the correlation operation is performed on the Doppler spectrum values of the echo sequence before and after the k-th distance unit, K indicates the total number of distance units with distinctive points selected, and |·| indicates the operation of the absolute value.

Preferably, in the step (4), adopting a random consistency algorithm to eliminate outliers includes:

(4.1) Approximate the phase error function as a Q (Q<N) order polynomial f(a ₀ , a ₁ ,...a _Q ), where N represents the number of sub-apertures, then its second-order derivative can be expressed as:

Among them, t represents the slow time, and a _q is the q-th degree coefficient of the polynomial f(a ₀ , a ₁ ,...a _Q );

n＝0,2...N-1，其中t_n＝(2n+1)T_s/2表示子孔径中心时间，在这N个点中，随机选择Q个作为内点，构成初始内点集，得到其对应的多次项系数a_q；设定阈值G，分别计算剩余的N-Q个点到该多项式曲线的距离，若某点对应的距离小于阈值，则该点被认为内点，将其列入内点集；(4.2) Corresponding to N sub-apertures, there are N second-order derivatives in total

n=0,2...N-1, where t _n =(2n+1)T _s /2 represents the center time of the sub-aperture. Among these N points, Q are randomly selected as interior points to form the initial interior points set the corresponding multi-term coefficients a _q ; set the threshold G, and calculate the distances from the remaining NQ points to the polynomial curve respectively. its inclusion in the inset point set;

(4.3) Count the number of interior points in the interior point set;

(4.4) Repeat steps (4.2) and (4.3) S times, compare and select the inlier set with the most inliers;

(4.5) All inliers in the inlier set with the most inliers are valid values.

Preferably, the number of repetitions S is selected according to the following formula:

Among them, P is the confidence probability, and ε is the data error rate.

Based on the same inventive concept, the present invention discloses a motion-compensated self-focusing device for unmanned aerial vehicle SAR imaging, comprising a memory, a processor and a computer program stored in the memory and running on the processor. When the computer program is loaded into the processor, the described motion-compensated self-focusing method for unmanned aerial vehicle SAR imaging is implemented.

Beneficial effects: Based on the existing sub-aperture correlation algorithm, the present invention proposes a new motion compensation self-focusing method for unmanned aerial vehicle SAR imaging. After dividing the distance into blocks, all distances in a certain azimuth block Find the same number of distinctive points in each block, and according to the principle that the Doppler modulation frequency has a linear relationship with the slant range of the point target in the Doppler domain, the distance unit where the distinctive point is located is matched to the maximum linear change with the distance. The excellent Doppler modulation frequency is used as the Doppler modulation frequency corresponding to each range gate of the azimuth block. This operation solves the problem in the prior art that it is difficult to obtain good Doppler modulation frequency estimation accuracy and sufficient range space variation information at the same time. At the same time, since the azimuth block may cause no strong scattering points in the sub-aperture or the scene is uniform, the estimation will generate singular values, which will affect the imaging results. Therefore, the present invention adopts the random consistency method to eliminate the singular values and uses the remaining effective values. The Doppler modulation frequency of the range gate is estimated, which further improves the estimation accuracy of the Doppler modulation frequency.

Description of drawings

FIG. 1 is a model diagram of a synthetic aperture radar signal with distance space variation in an embodiment of the present invention.

FIG. 2 is a flowchart of a method according to an embodiment of the present invention.

Figure 3 is a simulation motion error diagram.

Fig. 4 is the image to be processed, 4(a) is the original image, and 4(b) is the defocused image after adding motion error.

Figure 5 is the result of the existing method of dividing the distance into 4 blocks, 5(a) does not remove the singular value, 5(b) removes the singular value.

Fig. 6 is the result diagram of the existing method of dividing the distance into 128 blocks.

Figure 7 Self-focusing image of the improved sub-aperture correlation algorithm, 7(a) does not remove singular values, 7(b) removes singular values.

Figure 8 is a partial enlarged view of the original image.

FIG. 9 is a partial enlarged view of the self-focusing result. 9(a) Partial enlargement of the existing method, 9(b) Partial enlargement of the improved sub-aperture correlation algorithm.

FIG. 10 is a result diagram of the change of the entropy value of the block along the distance.

Detailed ways

The present invention will be further described below with reference to specific embodiments and accompanying drawings.

The airborne synthetic aperture radar signal model with distance space variation error involved in the embodiment of the present invention is shown in FIG. 1 , and the positive X axis in the figure represents the flight direction of the airborne aircraft. The height of the carrier from the ground is H. Under ideal conditions, the antenna phase center changes at a uniform speed along the X-axis, but in practice, there is a positional deviation in the route. r represents the shortest distance from the target to the ideal route, and R(t) is the slow time t. The real distance from the target to the antenna phase center at time, R ₀ (t) is the ideal distance from the target to the antenna phase center at time t, the point P in the figure is the real position of the antenna phase center at time t, and the instant between it and the ideal position P ₀ The motion error is Δ=[Δx(t), Δy(t), Δz(t)], where Δx(t), Δy(t) and Δz(t) represent the antenna phase center on the X-axis, Y-axis and Z-axis, respectively Instantaneous position deviation in the axis direction. (x, y, z) is the coordinates of a target on the ground.

Assume that the synthetic aperture radar transmits a chirp signal as

为快时间，j表示虚数单位，

f_c为信号载频，T_p为发射信号脉宽，γ为调频率。经过相干检波，接收回波信号表示为：in,

is fast time, j represents imaginary unit,

f _c is the signal carrier frequency, T _p is the pulse width of the transmitted signal, and γ is the modulation frequency. After coherent detection, the received echo signal is expressed as:

Among them, c is the speed of light, R(t; r,x) is R in Figure 1, and the specific expression is:

First, use inertial navigation data to perform coarse compensation processing on the received echo signal, secondly perform Range Cell Migration (RCM) correction and range matched filtering, and then perform azimuth de-ramping (De-ramping) operation on the signal. Eliminating the quadratic FM component in the ideal slant range, the signal with residual phase error is obtained as:

为剩余相位误差，s_ref(r_l,t)为无剩余相位误差下理想的回波信号。Among them, r _l is the slant distance from the center of the lth (l=1, 2...L) distance unit to the ideal route, L is the total number of distance units,

is the residual phase error, and s _ref (r _l ,t) is the ideal echo signal without residual phase error.

To utilize the sub-aperture correlation algorithm in stripe mode, we block the azimuth direction, dividing the data of duration T into N durations of T _s with a center time of t _n = (2n+1)(T _s /2), n=0,1,...,N-1 sub-apertures.

_Ts should be small enough to ensure that the residual phase error in each subaperture can be approximated by a second-order polynomial as follows, typically about one tenth of the coherent accumulation time:

Among them, τ∈-T _s /2<τ<T _s /2.

Based on the above-mentioned airborne synthetic aperture radar signal model with airborne distance variation error, a motion compensation self-focusing method for UAV airborne SAR imaging disclosed in the embodiment of the present invention uses inertial navigation for the received echo signal. After the data is subjected to coarse compensation, RCM correction, range matching filtering, and De-ramping operations, the obtained signal with residual phase error is subjected to azimuth block to obtain the data of each sub-aperture, and then each sub-aperture is subjected to range block and Select characteristic point samples from all distance units; then divide the distance unit where the samples are located into two parts, before and after, perform correlation processing, and estimate the second-order phase error coefficient to obtain the second-order phase error of each distance unit; After obtaining the second derivative of the residual phase error of each sub-aperture of the unit, in order to avoid the existence of a few outliers, a random consistency algorithm is used to eliminate outliers; the least squares method is used to obtain the residual phase error and perform phase compensation.

When applying the sub-aperture correlation algorithm, if the azimuth signal in a certain range cell has distinctive points, its estimation is much more accurate than that of the range cell with only stray points, so the samples should be selected. It is usually sufficient to select the one-tenth of the distance cell with the largest energy. Since the present invention considers the second-order phase error of the distance space variation, the selected samples should provide sufficient distance space variation information. Our approach is to select samples for each sub-aperture data separately, divide each sub-aperture data into some distance blocks, and select the required sample data from the distance units of each distance block by the energy criterion.

In the following, taking the nth sub-aperture as an example, the sub-aperture correlation algorithm is respectively applied to the N sub-apertures to perform the same processing. Divide the nth sub-aperture into two halves, which can be expressed as:

r_k表示挑选出的第k个距离单元的斜距，k＝1,2...K,K＜L，K为挑选出有特显点的距离单元总数。常数项

对估计没有任何影响可以忽略，同时

只是会让左右两幅图像在方位向上产生相同的平移，由(7)、(8)、(9)式知，可以等价于考虑如下两式：in,

_rk represents the slant distance of the k-th distance unit selected, k=1, 2...K, K<L, and K is the total number of distance units with distinctive points selected. Constant term

has no effect on the estimate and can be ignored, while

It just makes the left and right images produce the same translation in the azimuth direction. Known from equations (7), (8) and (9), it can be equivalent to considering the following two equations:

in,

The Fourier transform of (9) and (10) in the azimuth direction can be obtained:

为

做傅里叶变换，

为

做傅里叶变换。where f is the Doppler frequency,

for

Do the Fourier transform,

for

Do the Fourier transform.

Here, the Doppler modulation frequency error is considered to vary linearly along the range, namely:

Among them, k ₀ , k ₁ are the second-order phase error coefficients.

Then from equations (12) and (13) we can get:

与

在距离r_k上的频率移动，其可以由前后半段回波序列多普勒谱值做相关处理估计得到。此相关处理可以表示为：where Δf _r is the Doppler spectrum

and

The frequency shift at the distance _rk can be estimated by performing correlation processing on the Doppler spectrum values of the echo sequence in the front and back half segments. This correlation processing can be expressed as:

的相关峰值的位置。where φ represents the correlation of the two functions, and Δf _r is equivalent to

the position of the correlation peak.

l＝1,2,...L。上述的过程可以用下面的公式表述为：In the traditional sub-aperture correlation algorithm, the Doppler modulation frequency is assumed to be a constant. Accuracy of tuning frequency estimates. In practical applications, this is very important when there are strong clutter or no distinctive features in the scene. In the prior art, after the range direction is divided into blocks, the distance cells with strong scattering points in each range block are directly added to obtain the Doppler modulation frequency of the range block, and then the Doppler modulation frequency obtained by each range block is obtained. The operation of fitting frequency often brings many problems. When the number of range blocks is large, there are fewer range cells with strong scattering points in each block, and the Doppler modulation frequency cannot be estimated accurately; when the number of range blocks is small, although the Doppler modulation frequency of each block is small, The frequency can be accurately estimated, but due to the range-space variability of the Doppler modulation frequency, sufficient range-space variability information cannot be provided at this time, which we will clearly see in the simulation experiments in Part 3. Therefore, in order to use sufficient samples to make the algorithm robust and overcome the influence of noise, so that high precision and fast convergence can be achieved, we use the correlation peak position Δf _r of the cross-correlation function, combined with its linear movement along the distance unit. to estimate the parameters k ₀ , k ₁ , and then obtain by formula (13)

l=1,2,...L. The above process can be expressed by the following formula:

表示相关谱的积累，由(12)和(13)式可以看出，其值受参数k₀、k₁影响。同时，argmax表示取积累谱的最大峰值所在的位置Δf_r，而由(14)可以估计得到k₀、k₁。这里为了保证精度的同时减小计算量，可以应用文献[Zhang L,Duan J,Qiao Z J,et al.Phase adjustment and isar imaging of maneuvering targets withsparse apertures[J].IEEE Transactions on Aerospace&Electronic Systems,2014,50(3):1955-1973.]中的方法或其他加速算法加以解决。in,

Represents the accumulation of the correlation spectrum. It can be seen from equations (12) and (13) that its value is affected by parameters k ₀ and k ₁ . Meanwhile, argmax represents the position Δf _r where the maximum peak value of the accumulated spectrum is taken, and k ₀ and k ₁ can be estimated from (14). In order to ensure the accuracy and reduce the amount of calculation, the literature [Zhang L, Duan J, Qiao ZJ, et al. Phase adjustment and isar imaging of maneuvering targets with sparse apertures [J]. IEEE Transactions on Aerospace&Electronic Systems, 2014, 50 (3):1955-1973.] or other accelerated algorithms.

Compensating the distance samples in the time domain with the second-order phase error coefficient obtained by the current estimation can make the peak value of the cross-correlation function narrower, obtain a more accurate second-order phase error coefficient, and obtain a better focused image. Through repeated iterations, convergence can finally be achieved. In practice, a sufficiently accurate estimate of the second-order phase error can be obtained with 2-5 iterations.

n＝0,2...N-1后，为避免其中存在少数异常值的情况，我们采用随机一致性算法(RANdom SAmpleConsensus,RANSAC)剔除异常值。对各距离单元进行相同的操作，以第l个距离单元为例，具体步骤如下：Due to the sub-block processing of the scene azimuth, when the sub-block itself does not contain strong scattering points or the scene is uniform, due to the inherent limitations of the sub-aperture correlation algorithm, even if the above estimation method is used, the sub-aperture cannot be accurately obtained. If no processing is performed, this singular value will have an impact on the imaging results. Therefore, after obtaining L distance elements, the second derivative of the residual phase error of N sub-apertures

After n=0, 2...N-1, in order to avoid a few outliers, we use a random consensus algorithm (RANdom SAmple Consensus, RANSAC) to eliminate outliers. Perform the same operation on each distance unit, taking the lth distance unit as an example, the specific steps are as follows:

1) Approximate the phase error function as a Q (Q<N) order polynomial f(a ₀ , a ₁ ,...a _Q ), then its second-order derivative can be expressed as:

Among them, t represents the slow time, and a _q is the q-th degree coefficient of the polynomial f(a ₀ , a ₁ ,...a _Q );

n＝0,2...N-1中，随机选择Q个作为内点，构成初始内点集，按照式(17)计算，得到其对应的多次项系数a_q。设定阈值G，分别计算剩余的N-Q个点到该多项式曲线的距离，若某点对应的距离小于阈值，则该点被认为内点，将其列入内点集；2) in

In n=0, 2...N-1, Q are randomly selected as interior points to form an initial interior point set, which is calculated according to formula (17) to obtain the corresponding multi-term coefficients a _q . Set the threshold G, and calculate the distances from the remaining NQ points to the polynomial curve respectively. If the distance corresponding to a point is less than the threshold, the point is considered as an interior point and is included in the interior point set;

3) Count the number of interior points in the set of interior points;

4) Repeat step 2) and step 3) S times, compare and select the inlier set with the most inliers;

5) All inliers in the inlier set are what we consider valid values.

Obviously, if the Q points used to calculate the polynomial coefficients contain abnormal points, the number of inner points corresponding to the curve will not be the most. At the same time, the selection of the threshold G is very important. If it is too small, some inliers that should be selected will be lost, and if it is too large, some abnormal points will be misjudged as inliers. The choice of the threshold is related to the size of the error.

The number of repetitions S is selected according to the following formula:

Among them, P is the confidence probability, and ε is the data error rate.

的有效值，假设第l₁个距离单元上有M个有效值

m＝1,2...M，并根据(17)式可求得a_q,q＝2...Q的最小二乘解

这样我们便得到了多项式f(a₀,a₁,…a_Q)不含零次项和一次项的解析表达式，利用该表达式代回(7)式便可对该距离单元的剩余相位误差函数进行补偿，这样就可以有效地避免因某些子孔径内没有强散射点或场景均匀而造成多普勒调频率估计值奇异，使得聚焦结果得到改进。After the above processing, it can be obtained that each distance unit r _l (l=1, 2...L)

, assuming that there are M valid values on the _l1th distance unit

m=1, 2...M, and the least squares solution of a _q , q=2...Q can be obtained according to equation (17)

In this way, we have obtained the analytical expression of the polynomial f(a ₀ , a ₁ ,...a _Q ) without zero-order and first-order terms, and the residual phase of the distance unit can be obtained by substituting this expression back to Equation (7) The error function is used to compensate, so that it can effectively avoid the singularity of the Doppler modulation frequency estimate due to the absence of strong scattering points in some sub-apertures or the uniformity of the scene, so that the focusing results are improved.

In order to further verify the effect of the method of the present invention, the result of processing the measured data from the UAV synthetic aperture radar system of the Sandia laboratory by the method of the present invention is given below, and its specific system parameters are shown in Table 1.

Table 1 System simulation parameters

In the experiment, the synthetic aperture radar raw data is first transformed into the range pulse pressure domain, and the motion errors in the three coordinate directions are assumed to be tenth-order polynomials in slow time, as shown in Figure 3.

In order to illustrate the applicability of the improved sub-aperture correlation algorithm, the observation scene selected here is an open field, and only a few distance cells contain strong scattering points, which are used as the estimated samples of the phase gradient. In order to verify the self-focusing performance of the algorithm, image entropy is used as the quantitative evaluation index of imaging focusing. The entropy of a two-dimensional synthetic aperture radar image is:

Among them, _Na is the total number of image pulses, N _t is the total number of distance cells, D(q, k) is the scattering intensity density of the image, and its expression is:

Among them, s(I) is the total energy of the image, and I(q,k) is the complex reflection intensity of the synthetic aperture radar image. For an image with phase error, due to the blurred image and large uncertainty, the entropy value of the image is larger, so the entropy value of the image can well reflect the self-focusing performance of the algorithm.

Figure 4(a) shows the original image data, and its entropy value is 13.9467. After adding the distance space-variant motion error, the defocused image is shown in Figure 4(b), and the image entropy value is 14.2484. Compared with the original image, the entropy value is 14.2484. The difference between the values is about 0.3, at which time the road and the trees are no longer clearly distinguishable. The first one-eighth or so of the directional units of the original image data in Figure 4(a) contain very few strong scattering points, and the scene is relatively uniform. Imaging results are not good, so it is necessary to remove the singular values by the random consistency method mentioned earlier.

Using the processing method using the sub-aperture correlation algorithm proposed in the prior art to perform self-focusing on the defocused image with the distance space-variant motion error added, the number of distance blocks will have a serious impact on the imaging result. If the number of distance blocks is small and each block has enough distance units, the sub-aperture correlation algorithm can obtain a more accurate solution, but cannot provide enough distance space variation information; when the number of blocks is large, although it can provide sufficient distance Space variation information, but this will reduce the number of distinctive points contained in the sub-block, resulting in a decrease in estimation accuracy. When there are few distance blocks, the result of dividing into 4 blocks is shown in Figure 5. Figure 5(a) is the result of not removing singular values, and its entropy value is 14.0830; Figure 5(b) is the method proposed by the present invention. Combined with the existing method, the entropy value of the result after removing the singular value is 14.0205. It can be seen that the method proposed in the present invention can further focus the image and effectively improve the image quality by removing the singular value. Figure 5(b) The difference between the entropy value and the original image is 0.0738, which is significantly lower than that of the defocused image. The Doppler modulation frequency can be estimated more accurately, but the image entropy value still has a certain gap compared with the original image. This is because the algorithm can effectively correct the non-space-variant part, but the influence of distance space-variation is not well eliminated, and the remaining distance-space-variant phase error causes the image to have obvious defocus; and when there are many distance blocks , divided into 128 blocks without using random consistency to remove the singular value as shown in Figure 6, its entropy value is 14.4093, because the entropy value of Figure 6 is 0.1636 higher than the defocused image before compensation, which means that due to the distance There are too many blocks, and each distance block contains few distance units containing strong scattering points. Using the existing sub-aperture correlation algorithm can no longer obtain accurate estimates, the existing processing methods have failed, and the obtained estimates are all singular values. , so it doesn't make sense to use random consistency any more.

In order to solve the above problems, in the case of the same distance space-variable error, the improved sub-aperture correlation algorithm proposed by the present invention is used to perform self-focusing, and the result of dividing the distance into 128 blocks is shown in Fig. The result when the singular value is removed, its entropy value is 14.0035; Figure 7(b) is the result after removing the singular value, and its entropy value is 13.9897. It can be seen from Figure 7(b) that the image has been well focused after being processed by the method of the present invention, and the difference between the image entropy value and the original image entropy value is only 0.043.

In order to further illustrate the effectiveness of the improved sub-aperture correlation algorithm for space-variant phase error correction, Fig. 8 shows a partial enlarged view of Fig. 4(a) of the original image, and Fig. 9(a) and Fig. 9(b) respectively give Figure 5(b) of the self-focusing image of the existing method and the self-focusing image of the improved sub-aperture correlation algorithm in Figure 7(b) are shown as enlarged images. It can be seen from the figure that the method proposed in this paper can accurately focus the strong scattering points in the figure, while the existing method neutron aperture correlation method cannot accurately focus, which fully shows that the method proposed in the present invention can better compensate for the Motion error for distance space variation.

Fig. 10 shows the difference between the entropy value and the original image entropy value obtained according to the existing sub-aperture correlation algorithm and the improved sub-aperture correlation algorithm proposed by the present invention, with respect to the number of distance blocks changes. It can be seen from the figure that no matter the number of blocks, the effect of the method proposed in the present invention is very stable and better than the existing sub-aperture correlation algorithms. When the number of blocks is small, the performance of the two methods is similar, but with the increase of The increase of distance blocks gradually shows the advantages of the method proposed in this paper in effectively utilizing the distance space-variable information, but the existing methods gradually fail and cannot effectively self-focus the image. Figure 10 fully demonstrates the effectiveness of the proposed method for motion errors with distance variation.

To sum up, when using the sub-aperture correlation algorithm to estimate the Doppler modulation frequency, it is necessary to have strong scattering points in the selected range unit. Too many distance blocks will reduce the number of strong scattering points, resulting in a decrease in the estimation accuracy. Sufficient distance space variation information. The improved sub-aperture correlation algorithm proposed by the present invention can effectively solve this problem by utilizing the characteristic that the Doppler modulation frequency changes linearly along the range gate. When the scene contains only a few strong scattering points or the scene is uniform, the random consistency method is adopted to eliminate the singular values, which can improve the focusing effect of the image.

Based on the same inventive concept, an embodiment of the present invention discloses a motion-compensated self-focusing device for unmanned aerial vehicle SAR imaging, comprising a memory, a processor, and a computer program stored in the memory and running on the processor, When the computer program is loaded into the processor, the above-mentioned motion-compensated self-focusing method for unmanned aerial vehicle SAR imaging is implemented.

Claims (6)

1. A motion compensated autofocus method for unmanned on-board SAR imaging, the method comprising the steps of:

(1) after coarse compensation, RCM correction, distance matching filtering and De-tracking operation are carried out on the received echo signals by using inertial navigation data, azimuth blocking is carried out on the obtained signals with residual phase errors, and sub-aperture data are obtained;

(2) performing distance direction blocking on each sub-aperture, and selecting a special display point sample from all distance units;

(3) dividing a distance unit where the special display sample book is located into a front part and a rear part, performing correlation processing, estimating to obtain a second-order phase error of the distance unit where the special display sample book is located, and calculating to obtain a second-order phase error coefficient by using data of the distance unit where the special display sample book is located based on the linear change characteristic of the second-order error along the distance direction, so as to estimate to obtain the second-order error in the distance unit without the special display sample book;

(4) after second-order derivatives of residual phase errors of all sub-apertures of all distance units are obtained, eliminating abnormal values by adopting a random consistency algorithm;

(5) and obtaining residual phase errors by using a least square method and performing phase compensation to improve the imaging focusing effect of the synthetic aperture radar.

2. The motion-compensated self-focusing method for unmanned airborne SAR imaging according to claim 1, wherein the distance-wise blocking of each sub-aperture and the selection of the samples in step (2) are performed by: and (4) considering the second-order phase error of the distance space-variant, respectively selecting samples of each sub-aperture data, dividing each sub-aperture data into a plurality of distance blocks, and selecting required sample data from the distance units of each distance block by an energy criterion.

3. The motion compensated self-focusing method for unmanned airborne SAR imaging according to claim 1, wherein in said step (3) the second order phase error coefficient is calculated by the following formula:

wherein,

for fast time, f denotes the Doppler frequency, r_kRepresents the pitch of the selected kth distance unit,

indicating that the correlation operation is carried out on the Doppler spectrum values of the echo sequence of the front half section and the back half section of the kth distance unit, wherein K indicates the total number of the distance units with special display points, and | represents the absolute value operation.

4. The motion-compensated self-focusing method for unmanned airborne SAR imaging according to claim 1, wherein the step (4) of rejecting outliers by using a random consistency algorithm comprises:

(4.1) approximating the phase error function to a Q (Q < N) order polynomial f (a)₀,a₁,…a_Q) And N denotes the number of subapertures, the second derivative can be expressed as:

wherein t denotes a slow time, a_qIs a polynomial f (a)₀,a₁,…a_Q) Q-order item coefficients;

(4.2) corresponding to N sub-apertures, there are N second derivatives

Wherein t is_n＝(2n+1)T_sAnd/2 represents the sub-aperture center time, and Q points in the N points are randomly selected as interior points to form an initial interior point set, and the corresponding polynomial coefficient a of the initial interior point set is obtained_q(ii) a Setting a threshold value G, respectively calculating the distance from the remaining N-Q points to the polynomial curve, and if the distance corresponding to a certain point is less than the threshold value, considering the point as an interior point and listing the interior point into an interior point set;

(4.3) counting the number of inner points in the inner point set;

(4.4) repeating the step (4.2) and the step (4.3) for S times, comparing and selecting the interior point set with the most interior points;

and (4.5) all the interior points in the interior point set with the most interior points are effective values.

5. The motion-compensated autofocus method for unmanned-airborne SAR imaging according to claim 4, wherein the number of repetitions S is selected according to the following equation:

wherein, P is confidence probability and data error rate.

6. A motion compensated autofocus apparatus for unmanned on-board SAR imaging, comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the computer program, when loaded into the processor, implements the motion compensated autofocus method for unmanned on-board SAR imaging according to any of claims 1 to 5.

Images