CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition

Abstract

The invention discloses a CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition, and the method comprises the steps: generating a plurality of sub-aperture data according to the to-be-imaged data, and carrying out the coarse imaging of all sub-aperture data, and obtaining a corresponding coarse image; performing feature extraction on all the coarse images to obtain corresponding projection vectors; performing self-focusing processing on all the projection vectors to obtain corresponding self-focusing images; and obtaining an imaging image corresponding to the to-be-imaged data through image matching and incoherent superposition processing according to the self-focusing imaging picture. According to the invention, the precision and efficiency of CSAR imaging can be improved.

Description

Technical Field

[0001] The present invention relates to the field of radar technology, and in particular, to a CSAR autofocus imaging method based on ABP sub-aperture processing and incoherent superposition. Background Art

[0002] Synthetic Aperture Radar (SAR) obtains high range resolution by transmitting a large bandwidth signal. At the same time, the radar platform observes the scene target at a large angle, and can obtain high azimuth resolution. As one of the sensors of the new productive forces in the fields of microwave remote sensing and advanced arrays, SAR technology has developed rapidly and received extensive attention. SAR imaging can obtain more electromagnetic scattering information of the observed target by reconstructing the target scattering function, which helps to analyze, classify, and identify the target characteristics. SAR technology can provide all-weather and all-time reconnaissance capabilities, and has significant advantages in remote sensing observations, with very wide applications.

[0003] Circular SAR (CSAR) means that the SAR platform moves in a circular trajectory around the observation scene, and its antenna beam continuously covers the target observation area. The omnidirectional observation of the CSAR mode can obtain an image with higher resolution, can realize multi-angle observation imaging of the target, so as to obtain more complete target information and improve the image resolution.

[0004] In existing airborne CSAR imaging processing, high-precision imaging of airborne CSAR experimental data mainly faces two key problems to be solved urgently. The first problem is the motion error between the actual flight trajectory and the ideal trajectory of the radar carrier platform. This error is caused by various factors such as airflows, which will lead to distortion phenomena such as offset, distortion, and blurring in the imaging results, thus making subsequent processing more difficult. In the acquisition of experimental data, various sensors need to be carried on the radar carrier platform to obtain various information such as position and speed, so as to obtain the actual motion trajectory. The frequency-domain algorithm corrects the imaging model by using the radar platform information provided by these sensors to reduce the impact of motion error on imaging. The time-domain algorithm has the characteristic of being applicable to irregular SAR imaging geometries and can further improve the imaging accuracy. Therefore, as long as the radar flight trajectory can be accurately measured, the motion error can be theoretically completely compensated. The second problem is the error in the flight trajectory obtained by the sensor. The airborne CSAR system often uses the global positioning system and inertial navigation system for positioning. In current commercial products, their positioning accuracy cannot meet the requirements of high-precision CSAR imaging. Therefore, autofocus algorithms need to be used for further processing. Currently, the means to compensate for the accuracy of CSAR experimental data mainly include the following. AFRL in the United States, ONEGA in France, and DLR in Germany placed calibrators in their experiments to correct measurement errors, but it is very difficult to place calibrators in advance in the area to be observed in practical applications. The phase gradient autofocus method can be directly applied to the image domain, but it requires a Fourier relationship between the image domain and the phase history, which is difficult to hold in the CSAR imaging processing process. Summary of the Invention

[0005] To overcome the defects of the above-mentioned prior art, the present invention provides a CSAR autofocus imaging method and device based on ABP sub-aperture processing and incoherent superposition, which can improve the accuracy and efficiency of CSAR imaging.

[0006] An embodiment of the present invention provides a CSAR autofocus imaging method based on ABP sub-aperture processing and incoherent superposition, including the following steps:

[0007] Generate a plurality of sub-aperture data according to the data to be imaged, and perform rough imaging on all the sub-aperture data respectively to obtain corresponding rough imaging maps;

[0008] Extract features from all the rough imaging maps respectively to obtain corresponding projection vectors;

[0009] Perform autofocus processing on all the projection vectors respectively to obtain corresponding autofocus imaging maps;

[0010] According to the autofocus imaging maps, through image matching and incoherent superposition processing, obtain the imaging image corresponding to the data to be imaged.

[0011] Further, generating a plurality of sub-aperture data from the data to be imaged specifically includes:

[0012] The data to be imaged is specifically full-aperture data, and the data to be imaged is divided into 2 M pieces of the sub-aperture data; where M is the number of sub-aperture divisions.

[0013] Further, respectively performing rough imaging on all the sub-aperture data to obtain corresponding rough imaging graphs specifically includes:

[0014] Performing rough imaging processing on all the sub-aperture data respectively through the BP algorithm to obtain the corresponding rough imaging graphs of all the sub-aperture data.

[0015] Further, respectively performing feature extraction on all the rough imaging graphs to obtain corresponding projection vectors specifically includes:

[0016] Determining strong energy regions in the rough imaging graphs where the energy is greater than a preset energy threshold through the maximum externally stable region detection method;

[0017] Composing the projection vectors corresponding to the rough imaging graphs by extracting the pixel points in the strong energy regions.

[0018] Further, respectively performing autofocus processing on all the projection vectors to obtain corresponding autofocus imaging graphs specifically includes:

[0019] Calculating the phase error of the projection vectors through the ABP algorithm;

[0020] Compensating the error of the projection vectors according to the phase error to obtain the corresponding autofocus imaging graphs of the projection vectors.

[0021] Further, according to the autofocus imaging graphs, obtaining the imaging image corresponding to the data to be imaged through image matching and incoherent superposition processing specifically includes:

[0022] Sequentially performing image matching on the autofocus imaging graphs corresponding to adjacent two sub-aperture data to obtain matching images;

[0023] Performing incoherent superposition on the autofocus imaging graphs corresponding to adjacent two sub-aperture data according to the matching images to obtain superposition images;

[0024] Repeating the processes of image matching and incoherent superposition processing on the superposition images obtained after incoherent superposition of adjacent two in the previous round until only one superposition image is generated after the latest round of incoherent superposition, ending the repetition process, and determining the superposition image generated in the last round as the imaging image.

[0025] Further, the step of performing image matching on the autofocus imaging maps corresponding to adjacent two sub-aperture data to obtain the matching image specifically includes:

[0026] Determine one of the autofocus imaging maps corresponding to adjacent two sub-aperture data as a reference image, and determine the other as a sub-image;

[0027] Determine the coordinate transformation parameters between the reference image and the sub-image through the RIFT image matching method;

[0028] Perform an affine transformation on the sub-image according to the coordinate transformation parameters to obtain the matching image.

[0029] Another embodiment of the present invention provides a CSAR autofocus imaging device based on ABP sub-aperture processing and incoherent superposition, including: a rough imaging module, a feature extraction module, an autofocus module, and an imaging module;

[0030] The rough imaging module is configured to generate a plurality of sub-aperture data according to the data to be imaged, and respectively perform rough imaging on all the sub-aperture data to obtain corresponding rough imaging maps;

[0031] The feature extraction module is configured to respectively extract features from all the rough imaging maps to obtain corresponding projection vectors;

[0032] The autofocus module is configured to respectively perform autofocus processing on all the projection vectors to obtain corresponding autofocus imaging maps;

[0033] The imaging module is configured to obtain an imaging image corresponding to the data to be imaged through image matching and incoherent superposition processing according to the autofocus imaging maps.

[0034] Further, the rough imaging module is configured to generate a plurality of sub-aperture data according to the data to be imaged, specifically including:

[0035] The data to be imaged is specifically full-aperture data, and the data to be imaged is divided into 2 M sub-aperture data; where M is the number of sub-aperture division times.

[0036] Further, the feature extraction module is configured to respectively extract features from all the rough imaging maps to obtain corresponding projection vectors, specifically including:

[0037] Determine a strong energy region in the rough imaging map where the energy is greater than a preset energy threshold through the maximum externally stable region detection method;

[0038] Extract the pixel points in the strong energy region to form the projection vector corresponding to the rough imaging map.

[0039] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0040] By means of the improved ABP algorithm, only strong scattering points are extracted as feature points for estimation, significantly reducing the number of imaging grid points required, thereby reducing the demand for computing resources. In addition, by dividing the full aperture into multiple sub-apertures and adopting an imaging method of alternating image matching and incoherent accumulation, the geometric distortion of the image is effectively reduced, the imaging accuracy is improved, and the computational complexity is also reduced. Generally speaking, the present invention not only improves the imaging performance of the CSAR system, but also enhances its feasibility and reliability in practical applications. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 It is a schematic flow chart of a CSAR autofocus imaging method based on ABP sub-aperture processing and incoherent superposition provided by an embodiment of the present invention.

[0042] Figure 2 It is a schematic flow chart of a BP algorithm provided by an embodiment of the present invention.

[0043] Figure 3 It is a schematic flow chart of an improved ABP algorithm provided by an embodiment of the present invention.

[0044] Figure 4 It is a schematic diagram of the imaging geometry of an airborne CSAR provided by an embodiment of the present invention.

[0045] Figure 5 It is a geometric interpretation diagram of an autofocus model provided by an embodiment of the present invention.

[0046] Figure 6 It is a schematic flow chart of image matching and incoherent superposition processing for an autofocus image provided by an embodiment of the present invention.

[0047] Figure 7 It is a schematic structural diagram of a CSAR autofocus imaging device based on ABP sub-aperture processing and incoherent superposition provided by another embodiment of the present invention.

[0048] Figure 8 It is a schematic diagram of random motion errors introduced by simulation data provided by an embodiment of the present invention.

[0049] Figure 9 It is a comparison diagram of imaging results obtained from airborne CSAR simulation data provided by an embodiment of the present invention.

[0050] Figure 10 It is an imaging cross-sectional view of the target at the center point of the scene provided by an embodiment of the present invention.

[0051] Figure 11 Schematic diagram of the flight trajectory of a radar in the actual measurement data of Gotcha provided by an embodiment of the present invention.

[0052] Figure 12 Schematic diagram of the random motion error introduced by the actual measurement data of Gotcha provided by an embodiment of the present invention.

[0053] Figure 13 Schematic diagram of the BP imaging result obtained based on the actual measurement data of Gotcha provided by an embodiment of the present invention.

[0054] Figure 14 Schematic diagram of the sub-aperture imaging result obtained based on the actual measurement data of Gotcha provided by an embodiment of the present invention.

[0055] Figure 15 Schematic diagram of the autofocus imaging result obtained based on the actual measurement data of Gotcha provided by an embodiment of the present invention. Detailed implementation manners

[0056] The accompanying drawings are only for illustrative purposes and should not be construed as limitations on this patent;

[0057] For those skilled in the art, it is understandable that some well-known structures and their descriptions in the accompanying drawings may be omitted.

[0058] The technical solutions of the present invention will be further described below with reference to the accompanying drawings and embodiments.

[0059] Referring to Figure 1 , which is a schematic flowchart of a CSAR autofocus imaging method based on ABP sub-aperture processing and incoherent superposition provided by an embodiment of the present invention, including the following steps:

[0060] S1: Generate a plurality of sub-aperture data according to the data to be imaged, and respectively perform rough imaging on all the sub-aperture data to obtain corresponding rough imaging maps;

[0061] S2: Respectively perform feature extraction on all the rough imaging maps to obtain corresponding projection vectors;

[0062] S3: Respectively perform autofocus processing on all the projection vectors to obtain corresponding autofocus imaging maps;

[0063] S4: According to the autofocus imaging maps, through image matching and incoherent superposition processing, obtain the imaging image corresponding to the data to be imaged.

[0064] For step S1, specifically, the generating a plurality of sub-aperture data according to the data to be imaged specifically includes:

[0065] The to-be-imaged data is specifically full-aperture data, and the to-be-imaged data is divided into 2 M sub-aperture data; where M is the number of sub-aperture divisions.

[0066] In a preferred embodiment, before performing rough imaging and subsequent operations, the to-be-imaged data needs to be divided from the full aperture into a sub-apertures. And in order to facilitate the pairwise matching and incoherent superposition operations of the subsequent sub-aperture data corresponding to the autofocus imaging map, a needs to be set as an even number, that is, a = 2 M , M ∈ R.

[0067] For step S1, further, the respectively performing rough imaging on all the sub-aperture data to obtain corresponding rough imaging maps specifically includes:

[0068] Performing rough imaging processing on all the sub-aperture data respectively through the BP algorithm to obtain the rough imaging maps corresponding to the respective sub-aperture data.

[0069] In a preferred embodiment, referring to Figure 2 , it is a schematic flowchart of a BP algorithm provided by an embodiment of the present invention. It can be seen from Figure 2 that the BP algorithm can be summarized into the following steps: First, generate an imaging area grid, then calculate the instantaneous slant range between the antenna phase center and the point target, and finally, after compensating the time-delay phase term of the demodulated echo signal, project it onto the imaging area grid according to the instantaneous slant range and coherently accumulate to obtain the BP imaging result, that is, the rough imaging map.

[0070] For step S2, specifically, the respectively performing feature extraction on all the rough imaging maps to obtain corresponding projection vectors specifically includes:

[0071] Determining a strong energy area in the rough imaging map where the energy is greater than a preset energy threshold through the maximum externally stable region detection method;

[0072] Composing the projection vector corresponding to the rough imaging map by extracting the pixel points in the strong energy area.

[0073] In a preferred embodiment, the ABP algorithm needs to store the imaging results of all radar apertures and use all the imaging results for phase error estimation, and its computational complexity and storage space are in a proportional relationship with the number of imaging grid points. And in image processing, the image maximum energy criterion will concentrate the stronger energy, that is, the sharpness of the image is determined by the bright spots in the image. Therefore, the estimation of the phase error is mainly determined by the strong scattering points, which means that in the autofocus process, it is not necessary to estimate all the imaging grid points, but only to estimate the strong scattering points, which can greatly reduce the imaging grid required by the ABP algorithm.

[0074] Therefore, before the subsequent autofocusing process is performed, it is necessary to first determine the region with stronger energy, i.e., the strong energy region, from the coarse image. In this preferred embodiment, the strong energy region is determined by the maximum external stable region detection method. Subsequently, the projection vector required by the ABP algorithm can be generated based on the strong energy region.

[0075] For step S3, specifically, performing self-focusing processing on all the projection vectors to obtain corresponding self-focusing imaging images specifically includes:

[0076] The phase error of the projection vector is obtained by calculating the ABP algorithm;

[0077] According to the phase error, the projection vector is error compensated to obtain the self-focusing imaging image corresponding to the projection vector.

[0078] In a preferred embodiment, the ABP algorithm is a BP-based self-focusing method, and after combining imaging processing with self-focusing, an improved ABP algorithm proposed in this preferred embodiment can be obtained. The improved ABP algorithm can be applied to any SAR imaging geometry.

[0079] Reference Figure 3 , is a flow chart of an improved ABP algorithm provided by an embodiment of the present invention. Figure 3 It can be seen that after obtaining the projection vector, the phase error corresponding to the projection vector can be estimated by the ABP algorithm, and finally the phase error is compensated to obtain the final self-focusing imaging image.

[0080] The following is the reasoning process of the improved ABP algorithm principle proposed in this preferred embodiment:

[0081] (1) Analysis of airborne CSAR motion errors

[0082] Reference Figure 4 , is a schematic diagram of an airborne CSAR imaging geometry provided by an embodiment of the present invention. Figure 4 It can be seen that the radar system is on a plane parallel to the xy plane, with the z axis as the center and a radius of r xy , a circular trajectory with a height of H and a tangential velocity of V. Assume that the flight platform transmits and receives N pulses with the same pulse repetition period T when flying in a circular trajectory. az When the flight platform transmits and receives the signal for the nth time, the position vector of the phase center of the radar antenna is d(n) = (x(n), y(n), z(n)). Let point p be any point target in the observation scene, and its position vector is d p =(x p ,y p ,zp ) Then, the instantaneous slant range between the antenna phase center d(n) and the point target p is:

[0083]

[0084] Suppose the airborne CSAR system transmits a linear frequency modulated signal that repeats N times with a period T, i.e.: az times, that is:

[0085]

[0086] The linear frequency modulated signal of a single pulse is:

[0087]

[0088] where rect(·) is the rectangular window function, f c is the carrier frequency of the signal, T p is the pulse width of the signal, and K r is the linear frequency modulation rate.

[0089] After the transmitted signal is reflected by any point target p in the observed scene, its echo signal is:

[0090]

[0091] where σ p (d(n), d p ) is the backscattering coefficient of target p, G t (d(n), d p ) is the directional pattern function of the transmitting antenna, G r (d(n), d p ) is the directional pattern function of the receiving antenna, and c is the speed of light. After performing range-direction pulse compression on the quadrature demodulated echo signal received at the nth radar aperture position, we get:

[0092]

[0093] where B is the signal bandwidth, and p rc is the pulse function after range compression. In the BP imaging result, the imaging result of any point d g =(x g , y g , z g ) on the imaging grid can be expressed as:

[0094]

[0095] Denote b n as the projection result of the echo signal of the nth radar aperture. Then, the imaging result of this radar aperture position for the point d g can be denoted as:

[0096]

[0097] Then it is located at point d g The imaging result can be expressed as:

[0098]

[0099] The imaging result can be expressed as:

[0100]

[0101] In actual CSAR imaging, the measurement result is easily affected by the motion error of the radar-carrying platform. The motion error can be modeled as an unknown phase shift of the phase history data of the radar echo signal, and the phase error will cause blurring in the CSAR image.

[0102] Let the position vector of the nth radar antenna phase center measured be d′(n) = (x′(n), y′(n), z′(n)), then the phase error caused by it is:

[0103]

[0104] where R(d′(n), d g ) is the distance between the radar antenna phase center and the point d on the imaging grid g . Generally speaking, it is assumed that the phase error is independent of the points on the imaging grid, that is, φ n (d g ) = φ n . When the above errors exist, affected by the phase error, the imaging result becomes:

[0105]

[0106] If direct imaging processing is carried out, that is, directly summing , it is difficult to obtain a good imaging result when φ n is large.

[0107] If you want to reduce or eliminate the influence of the phase error on the imaging result, it is necessary to estimate the phase error at each sampling position, and its estimated value is Then the imaging result obtained by compensating the phase with this estimated value is:

[0108]

[0109] From the above analysis, it can be seen that the model assumes that the phase error is independent of the points on the imaging grid. Specifically, this assumption holds that the distance errors from the radar antenna phase center to all imaging grid points caused by motion errors are the same. However, if the differences in distance errors are large, it may be difficult to obtain good focusing results. Therefore, when using the autofocus algorithm based on Equation (12), it is necessary to evaluate whether the differences in distance errors are within a certain range so that the phase error is also within a certain range.

[0110] (2) Autofocus (ABP) algorithm combined with BP

[0111] Based on the model shown in Equation (12), the autofocus algorithm uses the sharpest image criterion to estimate the phase error, that is, to solve the following optimization equation:

[0112]

[0113] where,

[0114]

[0115] is used to evaluate the sharpness of the image. Next, how to solve the above optimization equation will be introduced.

[0116] Since there is no closed-form solution to the optimization equation in Equation (13), the present invention uses the coordinate descent method for iterative optimization. In each optimization step, each parameter in the phase error is optimized to obtain a local extreme value of the objective function while keeping other parameters fixed. Since all parameters interact with each other, multiple iterations of the entire parameter set are required. Let be the nth phase error compensation value after the lth iteration, then the nth result of the (l + 1)th iteration is:

[0117]

[0118] Substituting the phase error compensation value obtained by iterating Equation (14) into Equation (12), we get:

[0119]

[0120] where x includes the sum of the BP projections of all pulses (including those after and before iteration) except the nth pulse, y is the BP projection result of the nth pulse without phase error compensation, then The sharpness of the ith pixel of

[0121]

[0122] where the fixed part is (v0) i = |xi | 2 +|y i | 2 , and the part of the parameter to be estimated is Then:

[0123] v = v0 + v φ (18)

[0124] For After arrangement, we can get:

[0125]

[0126] Then the problem is transformed into the following problem:

[0127]

[0128] The above problem can also be explained from a geometric perspective. Refer to Figure 5 , which is a geometric interpretation diagram of a self-focusing model provided by an embodiment of the present invention. Let the number of imaging grid points be H, then v = v0 + v φ , a, b are vectors in. v can determine an ellipse s on the plane ∑, where the intersection point of v0 and ∑ is the center of the circle, a and b are the major and minor axes, and the points on the ellipse correspond to one by one. Therefore, the problem can be transformed into finding an optimal corresponding point on the ellipse such that ‖v‖ is the largest.

[0129] It can be found that the ABP algorithm needs to store the imaging results of all radar apertures and use all the imaging results for the calculation of phase error estimation. Its computational complexity and storage space are proportional to the number of imaging grid points. Therefore, if the number of imaging grids required by the ABP algorithm can be reduced, the resources required for calculation can be greatly reduced.

[0130] By observing Equation (15) and Equation (16), it can be found that the maximum image energy criterion will concentrate the stronger energy, that is, the sharpness of the image is determined by the bright spots in the image. Therefore, in the process of phase error estimation, the strong scattering points play a dominant role, which means that in the self-focusing process, it is not necessary to estimate all the imaging grid points, and only the strong scattering points need to be estimated, which can greatly reduce the imaging grids required by the ABP algorithm. For the area with stronger energy, it can be determined by the maximum externally stable region detection method, so as to form a new projection vector Then estimate the phase error.

[0131] For step S4, specifically, obtaining the imaging image corresponding to the data to be imaged through image matching and incoherent superposition processing according to the autofocus imaging map specifically includes:

[0132] Sequentially perform image matching on the autofocus imaging maps corresponding to adjacent two sub-aperture data to obtain a matching image;

[0133] Perform incoherent superposition on the autofocus imaging maps corresponding to adjacent two sub-aperture data according to the matching image to obtain a superposition image;

[0134] Repeat the processes of image matching and incoherent superposition processing on adjacent two superposition images obtained after the previous round of incoherent superposition until only one superposition image is generated after the latest round of incoherent superposition, then end the repetition process, and determine the superposition image generated in the last round as the imaging image.

[0135] Further, performing image matching on the autofocus imaging maps corresponding to adjacent two sub-aperture data to obtain the matching image specifically includes:

[0136] Determine one of the autofocus imaging maps corresponding to adjacent two sub-aperture data as a reference image, and determine the other as a sub-image;

[0137] Determine the coordinate transformation parameters between the reference image and the sub-image through the RIFT image matching method;

[0138] Perform affine transformation on the sub-image according to the coordinate transformation parameters to obtain the matching image.

[0139] In a preferred embodiment, in CSAR sub-aperture imaging, the imaging image will be defocused due to motion errors. Even after autofocus processing, there will still be phenomena such as offset, scale stretching, and geometric deformation. Moreover, due to different motion errors of different sub-apertures, the deformation types and degrees of sub-aperture images are also different. At this time, if the sub-aperture images are directly superimposed, the same points in different images cannot be superimposed at the same coordinate position, resulting in phenomena such as blurring and ghosting in the final imaging result. To solve the above problems, a coordinate transformation method is introduced to transform the same points in different images to the same coordinate position for superposition, and then the imaging quality will be greatly improved. Therefore, the present invention will adopt an image matching method to solve the problem of superposition of projection results of different sub-apertures.

[0140] Image matching is to find the same or similar point pairs in two images through a certain algorithm, so as to establish a one-to-one correspondence between the coordinate relationships of each pixel point in different images. According to different methods, image matching methods are mainly divided into gray-scale matching methods and feature point matching methods. In this preferred embodiment, the image matching method adopted is specifically the Radiation-Invariant Feature Transform (RIFT) image matching method.

[0141] In different sub-aperture imaging, the motion errors of the radar carrying platform are not the same, and there will be a situation where one image is relatively blurred while the other image is relatively clear. In addition, due to different target scattering angles, there will be a situation where the same target appears in one image and disappears in the other image. Therefore, if the sub-aperture images are directly matched without processing, there may be a situation of feature point mismatching, which will affect the subsequent image superposition and ultimately lead to a significant decrease in imaging quality. This is also one of the reasons why it is necessary to first determine the strong energy region and extract the projection vector in the previous step before performing subsequent autofocus processing.

[0142] Image matching can correspond the same feature points between two images, so as to obtain the corresponding coordinate transformation relationship. After coordinate transformation, non-coherent superposition of different images can make the same points in different images be superposed at the same coordinate position, thus obtaining a better imaging result. In the actual imaging scenario, the backscattering coefficients of the target in different directions are not the same, so the projection results of the same target in different sub-aperture images are also not the same. The change of the backscattering coefficient is relatively small within a small angle, and the imaging results under adjacent sub-apertures are also relatively close. Based on this characteristic, for the convenience of matching operation, this preferred embodiment selects the images generated by two adjacent sub-apertures as a pair of reference images and matching images for matching and non-coherent superposition.

[0143] Refer to Figure 6 , which is a schematic flow chart of image matching and non-coherent superposition processing for an autofocus image provided by an embodiment of the present invention. As can be seen from Figure 6 , the entire process of image matching and non-coherent superposition processing can be summarized as: dividing the entire full aperture into 2 M sub-apertures, and pairwise matching and non-coherently superposing the adjacent sub-aperture images to generate 2 M-1 sub-aperture images. Repeat the above operation on these 2 M-1 sub-aperture images to generate the final full aperture data. The following are the specific processing flow steps:

[0144] The first step is to use the maximum external stable region detection method to determine the region for feature point selection in the image;

[0145] In the second step, obtain the matching feature points of the sub-images generated by two adjacent sub-apertures;

[0146] In the third step, calculate the transformation parameters between the sub-image and the global reference image by using the coordinate parameters of the feature points;

[0147] In the fourth step, perform an affine transformation on the sub-image by using the coordinate transformation parameters;

[0148] In the fifth step, perform non-coherent addition on the results of the affine transformation to obtain the result of image matching;

[0149] In the sixth step, repeat the process from the second step to the fifth step for the sub-aperture images obtained in the fifth step until the final image matching result is obtained.

[0150] Generally speaking, the advantages of the method described in the embodiments of the present invention are as follows:

[0151] (1) By adopting the image matching method, the geometric deformation of the image is effectively reduced, and the imaging accuracy is improved;

[0152] (2) By adopting the non-coherent accumulation method, the imaging result is smoother, the influence of coherent speckles is reduced, and the quality of the image is improved;

[0153] (3) By the improved ABP algorithm, only the strong scattering points are estimated, significantly reducing the required number of imaging grid points, thereby reducing the demand for computing resources;

[0154] (4) By dividing the full aperture into multiple sub-apertures and adopting the non-coherent accumulation method, the computational complexity is further reduced.

[0155] The experimental results show that the proposed method not only maintains good imaging quality, but also significantly reduces the computing time and storage space. Therefore, the method proposed in the present invention not only improves the imaging performance of the CSAR system, but also enhances its feasibility and reliability in practical applications.

[0156] Refer to Figure 7 , which is a schematic structural diagram of a CSAR autofocus imaging device based on ABP sub-aperture processing and non-coherent superposition provided by another embodiment of the present invention, including: a rough imaging module 101, a feature extraction module 102, an autofocus module 103, and an imaging module 104;

[0157] The rough imaging module 101 is used to generate a plurality of sub-aperture data according to the data to be imaged, and perform rough imaging on all the sub-aperture data respectively to obtain corresponding rough imaging maps;

[0158] The feature extraction module 102 is used to perform feature extraction on all the rough imaging maps respectively to obtain corresponding projection vectors;

[0159] The autofocus module 103 is configured to perform autofocus processing on all the projection vectors respectively to obtain corresponding autofocus imaging graphs.

[0160] The imaging module 104 is configured to obtain an imaging image corresponding to the data to be imaged through image matching and incoherent superposition processing according to the autofocus imaging graphs.

[0161] Further, the rough imaging module 101 is configured to generate a plurality of sub-aperture data according to the data to be imaged, specifically including:

[0162] The data to be imaged is specifically full-aperture data, and the data to be imaged is divided into 2 M sub-aperture data; where M is the number of sub-aperture division times.

[0163] Further, the feature extraction module 102 is configured to perform feature extraction on all the rough imaging graphs respectively to obtain corresponding projection vectors, specifically including:

[0164] Determine a strong energy region in the rough imaging graph with energy greater than a preset energy threshold through a maximum externally stable region detection method;

[0165] Form the projection vector corresponding to the rough imaging graph by extracting pixel points in the strong energy region.

[0166] In a preferred embodiment, to reflect the advantages of the method described in the embodiments of the present invention over the prior art, this preferred embodiment verifies the correctness and effectiveness of the method described in the embodiments of the present invention using simulation experiments and Gotcha measured data.

[0167] First, a simulation experiment of point target imaging is carried out. The parameters of the radar simulation system are shown in Table 1, and the parameters of the point targets in the observation scene are shown in Table 2.

[0168] Table 1 Radar simulation system parameters

[0169]

[0170]

[0171] Table 2 Observation scene point target parameters

[0172]

[0173] When simulating the echo signal, random motion errors need to be introduced in the x, y, and z directions respectively, as Figure 8 shown.

[0174] Refer to Figure 9, which is a comparison diagram of imaging results obtained from airborne CSAR simulation data provided by an embodiment of the present invention. Among them, the BP imaging results under the conditions of no motion error and with motion error are as shown in Figure 9 (a) and Figure 9 (b), Figure 9 (c) is the result after autofocusing the imaging image with introduced random motion error using the ABP algorithm.

[0175] By comparing the imaging results in Figure 9 , it can be found that when the ABP algorithm is not used, the target at the center point of the scene is defocused, and the surrounding point targets are submerged by the defocusing result of the center point target; after using the ABP algorithm, the target at the center point of the scene obtains good focusing, and the surrounding point targets also become clear. Therefore, the ABP algorithm can obtain good focusing effect.

[0176] Referring to Figure 10 , which is an imaging profile diagram of the target at the center point of the scene provided by an embodiment of the present invention. Table 3 gives the resolution and peak side-lobe ratio (PSLR) of the center point target.

[0177] Table 3 Focusing quality evaluation of the target at the center point of the scene

[0178]

[0179] Through Figure 10 and Table 3, it can be found that after using the ABP algorithm, the side lobes of the center point target are significantly suppressed, indicating that the ABP algorithm has excellent performance in improving the image focusing effect. At the same time, it can be found that after using the ABP algorithm, the center point target has a displacement in both the x direction and the y direction. Therefore, in the implementation of the subsequent CSAR autofocus algorithm, it is necessary to first perform image matching and affine transformation on the sub-aperture images, and then perform non-coherent superposition, otherwise it will affect the quality of the final imaging.

[0180] Then, in order to verify the effectiveness of the CSAR autofocus algorithm proposed by the present invention for processing measured data, the CSAR data (Gotcha measured data) publicly available from the US AFRL is used for verification. The observation scene of this experiment is a parking lot where multiple civilian vehicles are parked, and the specific parameters adopted are shown in Table 4.

[0181] Table 4 Parameters of the Gotcha dataset

[0182] Parameter Parameter value Polarization direction HH <![CDATA[Carrier frequency f0]]> 9.6 GHz Signal bandwidth B 0.62 GHz <![CDATA[Number N of radar apertures az > 42208 Imaging area size 120 m × 120 m

[0183] The entire radar aperture is divided into 64 sub-apertures, that is, the accumulation angle of each aperture is 5.625°. The flight altitude is about 7250 m, the flight radius is about 7250 m, and the flight trajectory is as shown inFigure 11 As shown. Introduce random motion errors into the original flight trajectory, such as Figure 12 As shown. Figure 13 Then, the imaging results obtained with and without introducing random motion errors are given. It can be found that when there are random motion errors, the imaging quality significantly deteriorates.

[0184] First, verify the autofocus ability of the ABP algorithm. In the experiment, a segment of sub-aperture data is extracted for ABP autofocus, and the focusing results are as Figure 14 (b) shown. It can be found that the result obtained by direct BP imaging has obvious blurring in the range direction, while the blurring degree of the ABP autofocus imaging result in the range direction is significantly reduced. Therefore, the image quality after ABP autofocus is significantly better than that of the direct BP imaging image. Then, perform imaging processing on the entire Gotcha data, set the imaging grid to 120m×120m (501 pixels×501 pixels), and the sampling interval of the scene is 0.24m. Using the CSAR autofocus algorithm proposed in the present invention, the obtained imaging result is as Figure 15 shown. It can be found that Figure 15 Compared with Figure 13 (b), it has the advantages of being clearer and smoother. Therefore, it can be proved that the method proposed in the present invention has good autofocus performance.

[0185] Obviously, the above embodiments of the present invention are merely examples for clearly illustrating the present invention, rather than limitations on the implementation manners of the present invention. For those of ordinary skill in the art, other different forms of changes or modifications can be made based on the above description. It is not necessary and impossible to enumerate all the implementation manners here. Any modifications, equivalent replacements, and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the claims of the present invention. This patent application relies on the Shenzhen Science and Technology Plan Funding (Project No.: JCYJ20240813151238049), the Shenzhen Science and Technology Plan Funding (Project Nos.: 202206193000001, 20220815171723002), the Guangdong Basic and Applied Basic Research Foundation (Project No.: 2023A1515011588), and the project "Research on Airborne Staring SAR Moving Target Detection and Tracking Algorithm" of Beijing Institute of Radio Measurement (Contract No.: 20242467).

Claims

1. A CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition, characterized in that: The steps include: Generate a plurality of sub-aperture data according to the data to be imaged, and perform coarse imaging on all the sub-aperture data respectively to obtain a corresponding coarse imaging map; Performing feature extraction on all the coarse images respectively to obtain corresponding projection vectors; Performing self-focusing processing on all the projection vectors respectively to obtain corresponding self-focusing imaging images; According to the self-focusing imaging image, an imaging image corresponding to the data to be imaged is obtained through image matching and incoherent superposition processing.

2. The CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition according to claim 1, characterized in that: The generating of a plurality of sub-aperture data according to the data to be imaged specifically includes: The data to be imaged is specifically full aperture data, and the data to be imaged is divided into 2 M The sub-aperture data; wherein M is the number of sub-aperture divisions.

3. The CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition according to claim 1, characterized in that: The performing coarse imaging on all the sub-aperture data respectively to obtain corresponding coarse imaging images specifically includes: The coarse imaging process is performed on all the sub-aperture data respectively by using the BP algorithm to obtain the coarse imaging map corresponding to each sub-aperture data.

4. The CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition according to claim 1, characterized in that: The feature extraction of all the coarse images is performed respectively to obtain corresponding projection vectors, specifically including: Determine a strong energy region in the coarse image whose energy is greater than a preset energy threshold by using a maximum external stable region detection method; The projection vector corresponding to the coarse image is formed by extracting pixel points in the strong energy region.

5. The CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition according to claim 1, characterized in that: The self-focusing processing is performed on all the projection vectors respectively to obtain corresponding self-focusing imaging images, specifically including: The phase error of the projection vector is obtained by calculating the ABP algorithm; According to the phase error, the projection vector is error compensated to obtain the self-focusing imaging image corresponding to the projection vector.

6. The CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition according to claim 1, characterized in that: The step of obtaining an imaging image corresponding to the data to be imaged by image matching and incoherent superposition processing according to the self-focus imaging image specifically includes: Sequentially performing image matching on the self-focusing images corresponding to two adjacent sub-aperture data to obtain a matching image; Incoherently superimposing the self-focusing imaging images corresponding to two adjacent sub-aperture data according to the matching image to obtain a superimposed image; Repeat the image matching and incoherent superposition processing for two adjacent superposition images obtained after the previous round of incoherent superposition until only one superposition image is generated after the latest round of incoherent superposition, then terminate the repetitive process and determine that the superposition image generated in the last round is the imaging image.

7. The CSAR self-focusing imaging method based on ABP sub-aperture processing and incoherent superposition according to claim 6, characterized in that: The step of performing image matching on the self-focusing imaging images corresponding to two adjacent sub-aperture data to obtain the matching image specifically includes: Determine one of the self-focusing imaging images corresponding to two adjacent sub-aperture data as a reference image, and determine the other as a sub-image; Determine the coordinate transformation parameters between the reference image and the sub-image by using the RIFT image matching method; Affine transformation is performed on the sub-image according to the coordinate transformation parameters to obtain the matching image.

8. A CSAR self-focusing imaging device based on ABP sub-aperture processing and incoherent superposition, characterized in that: include: Coarse imaging module, feature extraction module, autofocus module and imaging module; The coarse imaging module is used to generate a plurality of sub-aperture data according to the data to be imaged, and to perform coarse imaging on all the sub-aperture data respectively to obtain a corresponding coarse imaging map; The feature extraction module is used to extract features from all the coarse images respectively to obtain corresponding projection vectors; The self-focusing module is used to perform self-focusing processing on all the projection vectors respectively to obtain corresponding self-focusing imaging images; The imaging module is used to obtain an imaging image corresponding to the data to be imaged through image matching and incoherent superposition processing according to the self-focusing imaging image.

9. The CSAR self-focusing imaging device based on ABP sub-aperture processing and incoherent superposition according to claim 8, characterized in that: The coarse imaging module is used to generate a plurality of sub-aperture data according to the data to be imaged, specifically including: The data to be imaged is specifically full aperture data, and the data to be imaged is divided into 2 M The sub-aperture data; wherein M is the number of sub-aperture divisions.

10. The CSAR self-focusing imaging device based on ABP sub-aperture processing and incoherent superposition according to claim 8, characterized in that: The feature extraction module is used to extract features from all the coarse images to obtain corresponding projection vectors, specifically including: Determine a strong energy region in the coarse image whose energy is greater than a preset energy threshold by using a maximum external stable region detection method; The projection vector corresponding to the coarse image is formed by extracting pixel points in the strong energy region.

Images