Method, device and storage medium for rebuilding an aerial target recognition template library

Abstract

The application relates to an air target identification template library reconstruction method, device and equipment and a storage medium. The method combines an attribute scattering center model prior and derivation of a scattering coefficient amplitude and azimuth angle relationship under a physical optical method, constructs a scattering center evolution model reflecting a coupling relationship between a target scattering center and an azimuth angle, realizes model parameter estimation based on Newton orthogonal matching pursuit and simplex method, and finally reconstructs a target identification template library based on the SCEM. The method can realize accurate and stable identification template library reconstruction under the condition of incomplete ISAR image samples and does not depend on the number of samples.

Description

TECHNICAL FIELD

[0001] The application belongs to the technical field of target recognition, and relates to an air target recognition template library reconstruction method, device, equipment and storage medium. BACKGROUND

[0002] Inverse synthetic aperture radar (ISAR) can image air targets such as airplanes, unmanned aerial vehicles and the like. A distance-azimuth two-dimensional high-resolution image obtained by the ISAR can depict details such as the size, shape, structure and attitude of a target, thereby greatly improving the intelligent sensing capability of the radar and providing strong support for radar target feature extraction and classification recognition. At present, air target recognition based on the ISAR image mainly adopts a template-based method.

[0003] The main work of the template-based recognition method is concentrated on the construction of a recognition template library. However, according to the ISAR imaging mechanism and the scattering characteristics of a radar target, the ISAR images of the same target at different viewing angles are quite different, and the positions and intensities of strong scattering centers are very sensitive to the viewing angle change of the target, that is, the ISAR image has significant attitude angle sensitivity. Therefore, in order to accurately recognize a certain type of target, a template library containing various attitude angles of the target needs to be established, which not only occupies storage space but also has large matching overhead. Even if there is no problem with data storage, it takes more than several weeks to establish a complete ISAR image recognition template library according to typical indexes in a test field. Therefore, one of the main technical difficulties in air target ISAR image recognition is to accurately and stably recognize a target based on incomplete ISAR image samples.

[0004] To solve the above problems, the traditional solution is to obtain a more complete attitude recognition template library by the method of template averaging in equal interval azimuth angle range, that is, template averaging (TA). But the ISAR image changes with the attitude in a nonlinear evolution rule, and this template averaging method is difficult to accurately represent, thereby affecting the recognition performance of the ISAR image. The deep learning method has become another important means for ISAR image data expansion and more complete feature template library construction due to its unique advantages in the image domain, and the most representative network structure is the generative adversarial network (GAN) with image interpolation and extrapolation capabilities. However, GAN cannot generate SAR or ISAR image data with controllable attitude angles, and it is an unsupervised deep learning method. By introducing labels, SAR or ISAR image data generation at a controllable angle can be realized, that is, the conditional generative adversarial network (CGAN). The continuous representation method is another effective means for ISAR image data expansion of air targets, which maps the position coordinates to the corresponding values in the continuous space through a function, which is highly similar to the task of constructing a more complete ISAR image template library, and the most representative method is the implicit neural representation (INR), which is parameterized as a continuous function through a neural network, thereby realizing seamless interpolation and super-resolution. However, both the GAN and the INR method are general image domain template library reconstruction methods, and the robustness cannot be guaranteed without considering the electromagnetic scattering mechanism of the target. On the other hand, the deep learning method is heavily dependent on the number of training samples, but it is very challenging to obtain a large number of public ISAR image training samples due to the application background of ISAR image data. SUMMARY

[0005] To solve the problems in the above-mentioned traditional methods, the present application provides an air target recognition template library reconstruction method, device, equipment and storage medium, which can realize accurate and robust recognition template library reconstruction under the condition of incomplete ISAR image samples, and does not depend on the number of samples.

[0006] To achieve the above-mentioned purpose, the embodiments of the present application adopt the following technical solutions:

[0007] On the one hand, an air target recognition template library reconstruction method is provided, and the method comprises the following steps:

[0008] Analyzing the attribute scattering center model prior information, and deducing the coupling relationship between the target scattering coefficient amplitude and the azimuth angle under the physical optical method.

[0009] According to the attribute scattering center model prior information and the coupling relationship between the target scattering coefficient amplitude and the azimuth angle, a scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle is constructed.

[0010] According to ISAR image samples of a part of a posture of the air target to be reconstructed, the Newton orthogonal matching pursuit and the simplex method are used to estimate parameters of the scattering center evolution model, and the scattering center evolution model of the air target to be reconstructed is obtained.

[0011] The ISAR image of the air target to be reconstructed at an arbitrary azimuth angle is generated by using the scattering center evolution model of the air target to be reconstructed, and the reconstruction of the air target recognition template library to be reconstructed is realized.

[0012] In another aspect, an air target recognition template library reconstruction device is also provided, which comprises:

[0013] A prior information and physical optics derivation module is configured to analyze attribute scattering center model prior information and derive a coupling relationship between a target scattering coefficient amplitude and an azimuth angle under a physical optics method.

[0014] A scattering center evolution model construction module is configured to construct a scattering center evolution model reflecting a coupling relationship between a target scattering center and an azimuth angle according to the attribute scattering center model prior information and the coupling relationship between the target scattering coefficient amplitude and the azimuth angle.

[0015] A scattering center evolution model parameter estimation module is configured to estimate parameters of the scattering center evolution model according to ISAR image samples of a part of a posture of the air target to be reconstructed, and to obtain the scattering center evolution model of the air target to be reconstructed by using the Newton orthogonal matching pursuit and the simplex method.

[0016] An air target recognition template library reconstruction module is configured to generate the ISAR image of the air target to be reconstructed at an arbitrary azimuth angle by using the scattering center evolution model of the air target to be reconstructed, and to realize the reconstruction of the air target recognition template library to be reconstructed.

[0017] In still another aspect, a computer device is also provided, which comprises a memory and a processor, the memory stores a computer program, and the processor realizes steps of any of the above air target recognition template library reconstruction methods when executing the computer program.

[0018] In yet another aspect, a computer readable storage medium is also provided, which stores a computer program, and the computer program realizes steps of any of the above air target recognition template library reconstruction methods when executed by a processor.

[0019] One of the above technical solutions has the following advantages and beneficial effects:

[0020] The air target recognition template library reconstruction method, device, equipment and storage medium, the method reconstructs the target recognition template library based on the SCEM by constructing a scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle through the joint attribute scattering center model prior and the derivation of the scattering coefficient amplitude and azimuth angle relationship under the physical optical method, and realizing model parameter estimation based on the Newton orthogonal matching pursuit and simplex method. BRIEF DESCRIPTION OF DRAWINGS

[0021] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, a brief introduction will be given to the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor.

[0022] Figure 1 A flowchart of the air target recognition template library reconstruction method in one embodiment;

[0023] Figure 2 A flowchart of the air target recognition template library reconstruction method in one embodiment;

[0024] Figure 3 A comparison chart of the ISAR image original chart and the reconstruction results of three template library reconstruction methods based on the darkroom measured data at an azimuth angle of 10° in one embodiment, wherein Figure 3 (a) is the ISAR image original chart based on the darkroom measured data at an azimuth angle of 10°, Figure 3 (b) is a comparison chart of the reconstruction results based on the darkroom measured data at an azimuth angle of 10° using the TA method, Figure 3 (c) is a comparison chart of the reconstruction results based on the darkroom measured data at an azimuth angle of 10° using the INR method, Figure 3 (d) is a comparison chart of the reconstruction results based on the darkroom measured data at an azimuth angle of 10° using the SCEM method (the present method);

[0025] Figure 4 A comparison chart of the ISAR image original chart and the reconstruction results of three template library reconstruction methods based on the darkroom measured data at an azimuth angle of 50° in one embodiment, wherein Figure 4 (a) is the ISAR image original chart based on the darkroom measured data at an azimuth angle of 50°, Figure 4 (b) is a comparison chart of the reconstruction results based on the darkroom measured data at an azimuth angle of 50° using the TA method, Figure 4(c) is a comparison chart of the reconstruction result based on the darkroom measured data at an azimuth angle of 50° using the INR method, Figure 4 (d) is a comparison chart of the reconstruction result based on the darkroom measured data at an azimuth angle of 50° using the SCEM method (the present method),

[0026] Figure 5 Fig. 6 is a comparison chart of the ISAR image original chart based on the darkroom measured data at an azimuth angle of 80° and the reconstruction result of three template library reconstruction methods in an embodiment, wherein Figure 5 (a) is an ISAR image original chart based on the darkroom measured data at an azimuth angle of 80°, Figure 5 (b) is a comparison chart of the reconstruction result based on the darkroom measured data at an azimuth angle of 80° using the TA method, Figure 5 (c) is a comparison chart of the reconstruction result based on the darkroom measured data at an azimuth angle of 80° using the INR method, Figure 5 (d) is a comparison chart of the reconstruction result based on the darkroom measured data at an azimuth angle of 80° using the SCEM method (the present method). DETAILED DESCRIPTION

[0027] In order to make the purpose, technical scheme and advantages of the present application more clear, the present application is further described in detail below in combination with the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and are not intended to limit the present application.

[0028] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application belongs. The terminology used in the description of the application herein is for the purpose of describing particular embodiments only and is not intended to be limiting of the application.

[0029] It should be noted that the reference herein to "embodiments" means that the specific features, structures or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase is exhibited at various places in the specification does not necessarily mean the same embodiment, nor is it an independent or alternative embodiment to the other embodiments. Those skilled in the art can understand that the embodiments described herein can be combined with other embodiments. The term "and / or" used herein refers to any combination of one or more of the associated listed items and all possible combinations, and includes these combinations.

[0030] The embodiments of the present application will be described in detail below in combination with the drawings of the embodiments of the present application.

[0031] In one embodiment, as shown in Figure 1 Fig. 1, an air target identification template library reconstruction method is provided, which can include the following processing steps 100 to 103:

[0032] Step 100: Analyzing the attribute scattering center model prior information, deducing the coupling relationship between the target scattering coefficient amplitude and the azimuth angle under the physical optics method.

[0033] Specifically, the attribute scattering center model (ASC) is a mathematical model composed of seven parameters, which is a parameterized representation of the radar backscatter echo in the high-frequency region.

[0034] The radar backscatter echo can be represented as:

[0035] (1)

[0036] wherein, represents a parameter set (backscatter coefficient, frequency-dependent factor, length, tilt angle, azimuth-dependent factor, distance coordinate and azimuth coordinate respectively), represents the azimuth angle, represents the frequency, represents the additive noise, represents the total number of ASCs, and represents the echo signal of the th individual ASC, . Wherein

[0037] (2)

[0038] When characterizes a local scattering center, and when characterizes a distributed scattering center.

[0039] Step 101: According to the attribute scattering center model prior information and the coupling relationship between the target scattering coefficient amplitude and the azimuth angle, a scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle is constructed.

[0040] Specifically, the target scattering center is formed at the curvature discontinuity. For a smooth target, the curvature discontinuity is the outer contour edge and the vertex of the target. At the same time, two priors are combined: the ASC model describes that the target scattering center can be divided into two categories of local and distributed; the PO method deduces the coupling relationship between the target scattering coefficient amplitude and the azimuth angle. A scattering center evolution model (SCEM) reflecting the functional relationship between the target backscatter coefficient and the azimuth angle is constructed.

[0041] Step 102: According to the ISAR image samples of the part of the attitude of the air target to be reconstructed, the Newton orthogonal matching pursuit and the simplex method are used to estimate the parameters of the scattering center evolution model, and the scattering center evolution model of the air target to be reconstructed is obtained.

[0042] Specifically, based on the scattering center evolution model (SCEM), the Newton orthogonal matching pursuit (NOMP) is adopted to extract the end-point scattering center of the target ISAR image.

[0043] After the extraction of the scattering center, the parameters to be estimated mainly include the edge scattering intensity coefficient C, the end-point relative edge intensity proportion coefficient P, the edge evolution function order β, and the correction factor d. According to the previous steps, the positions and intensities of the corresponding scattering centers can be extracted, and then the evolution curves of the intensities of the two types of scattering centers with respect to the azimuth angle can be obtained, that is, and .

[0044] Obviously, the estimation of these parameters only needs to fit and according to the minimum mean square error criterion, and it is worth noting that according to the expression of the evolution function, the estimation of the correction factor d needs to solve a transcendental equation, so the estimation is relatively difficult, and therefore the bisection method is needed to estimate the approximate range of d, and then the optimization estimation is performed. The optimization algorithm applied in this step is the simplex method (Nelder-Mead), which is a non-gradient heuristic search algorithm for solving non-linear unconstrained problems of the objective function that is not derivable.

[0045] Step 103: generating the ISAR image of the air target to be reconstructed at any azimuth angle by using the scattering center evolution model of the air target to be reconstructed, to realize the reconstruction of the target recognition template library.

[0046] Specifically, based on the ISAR image sample of the partial attitude, the scattering center evolution model (SCEM) of the target is obtained by using the model parameter estimation method mentioned in step 102, and then the ISAR image at any azimuth angle can be generated by using this model, to realize the reconstruction of the target recognition template library.

[0047] Starting from the scattering characteristics of the target, the scattering center evolution model (SCEM) reflecting the functional relationship between the backscattering coefficient of the target and the azimuth angle is constructed, based on which the accurate reconstruction of the target recognition template library can be realized under the condition of incomplete ISAR image sample. By comparing the template library reconstruction results of the TA method, the INR method, and the method, it is shown that the performance of the method is optimal, and the reconstruction similarity index is more than 80%.

[0048] The above method for reconstructing the air target recognition template library combines the attribute scattering center model priori and the derivation of the scattering coefficient amplitude and azimuth angle relationship under the physical optics method, constructs a scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle, and realizes model parameter estimation based on Newton orthogonal matching pursuit and simplex method. Finally, the SCEM is used to reconstruct the target recognition template library. The method can accurately and stably reconstruct the recognition template library under the condition of incomplete ISAR image samples, and does not depend on the number of samples.

[0049] In one embodiment, the coupling relationship between the target scattering coefficient amplitude and the azimuth angle under the physical optics method includes: when the electric dimension is large, the scattering field is determined by the physical optics method as:

[0050] (3)

[0051] wherein, represents the scattering field, represents the unit normal vector outward of the surface; represents the incident magnetic field intensity at a point on the surface; represents the radial vector from the source point to the integration point on the surface; represents the integration of the illumination surface; represents the wave number; represents the radial unit vector from the source point to the field point, , represents the free space wave impedance, represents the distance from the source point to the field point, represents the illumination surface;

[0052] When the irradiated target is an ideal spherical surface or a cylindrical surface, the target scattering coefficient under the physical optics method is isotropic in the azimuth angle, and the relationship between the scattering field amplitude and the azimuth angle is:

[0053] (4)

[0054] wherein, represents the remaining term excluding the azimuth angle coupling term when the irradiated target is an ideal spherical surface or a cylindrical surface, which is independent of the azimuth angle, is the azimuth angle.

[0055] When the irradiated target is an ideal plane, the relationship between the scattering field amplitude and the azimuth angle is:

[0056] (5)

[0057] wherein, represents the remaining term excluding the azimuth angle coupling term when the irradiated target is an ideal plane, which is independent of the azimuth angle.

[0058] When the illuminated target is an ideal plane, the relationship between the scattering field amplitude and the azimuth angle is shown in equation (5).

[0059] (6)

[0060] wherein, represents the remaining term excluding the azimuth angle coupling term when the illuminated target is an ideal plane, which is irrelevant to the azimuth angle.

[0061] Specifically, the specific process of deriving the coupling relationship between the scattering field and the azimuth angle under the physical optics method (PO) includes:

[0062] When the electrical size (the electrical size refers to the ratio of the physical length of the signal transmission line to the wavelength) is large, the physical optics method (PO) can accurately calculate the backscattering field (ignoring diffraction), ignoring the harmonic factor, and the PO method gives the scattering field expression as shown in equation (3).

[0063] When the illuminated target is an ideal sphere or cylinder, the target scattering coefficient under the PO method is isotropic in the azimuth angle, i.e., constant, and the relationship between the scattering field amplitude and the azimuth angle is shown in equation (4).

[0064] When the illuminated target is an ideal plane, its scattering field can be calculated according to equation (3) as:

[0065] (7)

[0066] (8)

[0067] wherein, is the unit vector of the azimuth direction of incidence, is the pitch angle, is the side length of the ideal plane, is the wave number.

[0068] Then the scattering field amplitude and the azimuth angle are shown in equation (5).

[0069] When the illuminated target is an ideal plane, its scattering field can be calculated according to equation (3) as:

[0070] (9)

[0071] wherein, is the pitch angle, is the radius of the base circle of the ideal curved surface, is the wave number, is the height of the ideal curved surface.

[0072] The relationship between the amplitude of the scattered field and the azimuth angle is shown in formula (6).

[0073] In one embodiment, the scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle in step 101 is as follows:

[0074] (10)

[0075] in, This represents the evolution model of the scattering center. This represents the number of clusters in the edge scattering set; Indicates the first j The number of scattering centers at the edge of the strip. , These represent the upward widening of the j-th edge in terms of orientation and the upward widening of the j-th edge in terms of distance, respectively. Indicates the first j The scattering intensity coefficient at the edge of the strip; This represents the occlusion function, which includes two types: gradual occlusion and complete occlusion. Indicates the first j The initial angle of the strip edge; Indicates the first j Evolutionary symbols for the edge of a strip; Indicates the first j Edge occlusion angle; express j The order of the edge evolution function; Represents the rotation function; Indicates azimuth. This represents the main lobe portion of the sinc function; , The coordinates represent the location of the scattering center. , Indicates the first j The coordinates of the edge scattering center; These represent the distance and azimuth resolution, respectively. Indicates the number of endpoint scattering centers; This represents the function for complete occlusion. This represents the initial angle corresponding to the m-th endpoint. Indicates the first m endpoints Evolutionary symbols Indicates the first m endpoints Angle of obstruction This represents the scattering intensity coefficient of the nth edge connected to the mth endpoint. This represents the initial angle of the nth edge connected to the mth endpoint. The evolution symbol representing the nth edge connected to the mth endpoint. This represents the order of the edge evolution function of the nth edge connected to the mth endpoint. Indicates the first m The number of edges connected to each endpoint; Indicates the first m The coordinates of the scattering center at each endpoint; Indicates the first m The endpoint corresponding to the first n The strength ratio factor of the strip edge; The mapping function between endpoints and edges, Let represent the evolution function, and the expression for the evolution function is:

[0076] (11)

[0077] in d This represents a correction factor to prevent the denominator of the evolution function from taking a value of 0.

[0078] Specifically, the first term of formula (10) is the edge scattering center and the second term is the endpoint scattering center.

[0079] In one embodiment, step 102 includes: extracting the scattering center of the target ISAR image using Newton's orthogonal matching pursuit based on the ISAR image samples of the partial attitude of the airborne target to be reconstructed; after the scattering center is extracted, estimating the parameters to be estimated using the simplex method based on the ISAR image samples of the partial attitude of the airborne target to be reconstructed; the parameters to be estimated include: edge scattering intensity coefficient C, endpoint relative edge intensity ratio coefficient P, edge evolution function order β, and correction factor d; and obtaining the scattering center evolution model of the airborne target to be reconstructed based on the endpoint scattering center, the estimated values ​​of the parameters to be estimated, and the scattering center evolution model.

[0080] In one embodiment, based on ISAR image samples of the partial attitude of the aerial target to be reconstructed, Newton's orthogonal matching pursuit is used to extract the scattering center of the target's ISAR image, including: simplifying the scattering center evolution model to obtain the SCEM simplified function; and establishing a dictionary based on initial values ​​given by the SCEM simplified function and the range of the scattering center; the dictionary is as follows:

[0081] (12)

[0082] (13)

[0083] in, For a dictionary, for The normalization result, For a given azimuth angle, the simplified SCEM function is the first... k One portion, The coordinates of the scattering center are... is the parameter of the echo signal of the k th individual ASC, is the parameter set, , , are respectively the edge scattering center parameter set, the end point scattering center parameter set, , are respectively the edge scattering center intensity coefficient parameter set, the edge scattering center transverse coordinate parameter set, the edge scattering center longitudinal coordinate parameter set, , are respectively the end point scattering center intensity coefficient parameter set, the end point scattering center transverse coordinate parameter set, the end point scattering center longitudinal coordinate parameter set.

[0084] Initialize the residual image RSE as the input image, select the atom most matched with RSE from the dictionary; take the atom most matched with RSE as the initial point, solve the atom parameter fine estimation by using Newton method, and add the atom parameter fine estimation result to the estimated parameter set; wherein the atom parameter fine estimation is:

[0085] (14)

[0086] wherein, is the atom parameter fine estimation result, represents the normalized result of , and the conjugate is taken, represents the residual image.

[0087] Determine the optimal coefficient of the input image I approximated by the atom in the estimated parameter set through least square estimation; update the residual image RSE according to the input image I , the optimal coefficient, and the atom parameter fine estimation result; continue the next round of solving until the current residual image RSE cannot extract effective end point scattering centers or the number of extracted end points meets the requirement, and obtain the target ISAR image scattering center extraction result; wherein the residual image RSE update formula is:

[0088] (15)

[0089] wherein, represents the optimal coefficient of the optimal coefficient group corresponding to the i th atom, k is the input image, I is the th component of the SCEM simplified function optimization result under the given azimuth angle, k is the coordinate of the scattering center position, is the coordinate of the scattering center position, This is the result of a fine estimation of atomic parameters.

[0090] In one embodiment, the optimal coefficients for the input image I are:

[0091] (16)

[0092] in, This represents the coefficient group corresponding to each atom. Indicates the first k The coefficients corresponding to each atom k =1,2,……, i ; Indicates use Using atoms to approximate the input image I The optimal coefficient, Indicates use The Middle k Atoms are used to approximate the input image. I The optimal coefficient.

[0093] Specifically, Newton's Orthogonal Matching Pursuit (NOMP) is used to extract the endpoint scattering center of the target ISAR image. The specific steps are as follows:

[0094] (1) Simplified SCEM function. The simplified SCEM function for a given azimuth angle... k The components are:

[0095] (17)

[0096] (2) Establish a dictionary. Initial values ​​need to be given, which can be given according to the approximate range of the scattering center, as shown in formulas (12) and (13).

[0097] (3) Atom selection. Initialize the residual image RSE as the input image, and select the atoms that best match RSE from the dictionary:

[0098] (18)

[0099] in Representing the i The atoms selected in the next iteration.

[0100] (4) Fine estimation of atomic parameters. The atomic parameters obtained in step 3 are... As the initial point, Newton's method is used to solve formula (14).

[0101] Add the refined estimation results to The set of estimated parameters is as follows:

[0102] (19)

[0103] (5) Find the least square solution. Approximate the input image I with atoms in the dictionary D by least square estimation as shown in equation (16).

[0104] (6) Calculate the residual. Update the residual image RSE as shown in equation (15). Repeat the above steps until the current residual image RSE cannot extract valid end-point scatter centers or the number of extracted end-points has met the requirement, and exit the loop.

[0105] In one embodiment, the parameters to be estimated include: edge scattering intensity coefficient C, end-point relative edge intensity proportion coefficient P, edge evolution function order β, and correction factor d; after the scatter centers are extracted, the parameters to be estimated are estimated by using the simplex method according to the ISAR image samples of the part of the aerial target to be reconstructed, including: determining the evolution curve of the edge scattering and end-point scattering center intensity with respect to the azimuth angle according to the position and intensity of the scatter centers extracted; estimating the parameters to be estimated by using the simplex method according to the evolution curve of the edge scattering and end-point scattering center intensity with respect to the azimuth angle and the input objective function; the input objective function is:

[0106] (20)

[0107] wherein, represents the function curve between the intensity and the azimuth angle of two different scatter centers, i.e. and , represents the parameter set to be estimated, represents the end-point relative edge intensity proportion coefficient P, represents the edge scattering intensity coefficient C, represents the edge evolution function order β, represents the correction factor d, is the evolution function, represents the azimuth angle independent variable, and the input tolerance TolFun is 0.05.

[0108] In some embodiments, experimental verification is also provided to illustrate the examples. First, it is illustrated that: a certain passenger aircraft darkroom scale model is wrapped with tin foil paper to enhance its scattering intensity. The radar center frequency thereof is 220 GHz, the bandwidth is 3 GHz, the number of distance sampling points is 2048 points, the sampling rate is 80 MHz, the pulse repetition period is 1000 Hz, the turntable speed is 10° / s, and the accumulation rotation angle is 4°.

[0109] Figures 3 to 5 ​The result is based on the template library reconstruction experiment of darkroom measured data. In the experiment, the echoes are obtained by darkroom test of a scaled model of an aircraft, and the ISAR imaging results of some postures are obtained by imaging the echoes as the template library. The results of TA method, INR method and SCEM method are compared.

[0110] Figure 3 (a) is the ISAR image original graph based on darkroom measured data at 10° in azimuth, Figure 3 (b) is the reconstruction result comparison graph based on darkroom measured data at 10° in azimuth using TA method, Figure 3 (c) is the reconstruction result comparison graph based on darkroom measured data at 10° in azimuth using INR method, Figure 3 (d) is the reconstruction result comparison graph based on darkroom measured data at 10° in azimuth using SCEM method (the method), Figure 4 (a) is the ISAR image original graph based on darkroom measured data at 50° in azimuth, Figure 4 (b) is the reconstruction result comparison graph based on darkroom measured data at 50° in azimuth using TA method, Figure 4 (c) is the reconstruction result comparison graph based on darkroom measured data at 50° in azimuth using INR method, Figure 4 (d) is the reconstruction result comparison graph based on darkroom measured data at 50° in azimuth using SCEM method (the method), Figure 5 (a) is the ISAR image original graph based on darkroom measured data at 80° in azimuth, Figure 5 (b) is the reconstruction result comparison graph based on darkroom measured data at 80° in azimuth using TA method, Figure 5 (c) is the reconstruction result comparison graph based on darkroom measured data at 80° in azimuth using INR method, Figure 5 (d) is the reconstruction result comparison graph based on darkroom measured data at 80° in azimuth using SCEM method (the method).

[0111] The specific steps are: for the scaled model, the fixed pitch angle is 0°, the azimuth angle is from 0 to 90°, and the interval is 15°. Seven darkroom measured data ISAR imaging results are taken as training samples for feature template library reconstruction experiment. The training samples are reconstructed by TA method, INR method and SCEM method, so as to reconstruct a more complete ISAR image recognition template library, such as Figures 3 to 5The original drawing and the reconstruction results of three kinds of template library reconstruction methods of a certain passenger model at azimuth angles of 10°, 50° and 80° are shown, and the parameter setting of the INR method is consistent with the simple model. The INR method is adopted in the present application, the parameters of which are batch_size of 7, training batch of 8000 times, Adam optimizer is adopted, and it is run in the tensorflow environment. In terms of qualitative analysis, the reconstruction accuracy of the SCEM method for the scattering center main lobe of the darkroom measured data is optimal.

[0112] In order to quantitatively analyze and identify the reconstruction result performance of the template library, the present application proposes an evaluation index suitable for radar images, that is, image main lobe similarity (IMLS).

[0113] (1) Local energy calculation. The local energy of images A and B at pixel position ( x , y ) is defined as follows:

[0114] (21)

[0115] Wherein W x and W y respectively represent the main lobe width parameters in the range direction and the azimuth direction.

[0116] (2) Definition of weight matrix. The weight is defined based on the geometric mean of local energy:

[0117] (22)

[0118] (3) IMLS calculation. The numerator is the weighted dot product, and the denominator is the weighted normalization term:

[0119] (23)

[0120] The IMLS index of the three reconstruction methods is shown in Table 1, and it can be seen that the reconstruction similarity of the SCEM method is higher than 80%. Not only is the index better than the other two methods, but also the results of Figure 3 can be seen that the SCEM method reconstructs the image with higher quality and clarity. Figure 3

[0121] Table 1 Comparison of image main lobe similarity (IMLS) results of a certain passenger model

[0122]

[0123] It should be understood that although the above process Figure 1 ​Each of the steps in the flowchart is shown in sequence according to the arrows, but the steps are not necessarily performed in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order requirement for the performance of the steps, and the steps can be performed in other orders. Moreover, at least some of the steps of the above-described flowchart Figure 1 may include multiple sub-steps or stages, which are not necessarily performed at the same time, but can be performed at different times, and the order of the performance of the sub-steps or stages is not necessarily sequential, but can be performed in rotation or alternation with other steps or sub-steps or stages of other steps.

[0124] In one embodiment, there is also provided an apparatus for reconstructing a library of identification templates of an airborne target, the apparatus comprising:

[0125] a priori information and physical optics derivation module configured to analyze a priori information of an attribute scattering center model and derive a coupling relationship between a target scattering coefficient amplitude and an azimuth angle under a physical optics method.

[0126] a scattering center evolution model construction module configured to construct a scattering center evolution model reflecting a coupling relationship between a target scattering center and an azimuth angle according to the a priori information of the attribute scattering center model and the coupling relationship between the target scattering coefficient amplitude and the azimuth angle.

[0127] a scattering center evolution model parameter estimation module configured to estimate parameters of the scattering center evolution model according to ISAR image samples of a part of a posture of the airborne target to be reconstructed, using a Newton orthogonal matching pursuit and a simplex method, to obtain the scattering center evolution model of the airborne target to be reconstructed.

[0128] a library of identification templates of the airborne target to be reconstructed.

[0129] In one embodiment, the a priori information and physical optics derivation module is further configured to, when the target to be illuminated is of a large size, determine a scattering field using a physical optics method, as shown in formula (3). When the target to be illuminated is an ideal spherical surface or a cylindrical surface, the target scattering coefficient under the physical optics method is isotropic in the azimuth angle, and the relationship between the scattering field amplitude and the azimuth angle is as shown in formula (4). When the target to be illuminated is an ideal plane, the relationship between the scattering field amplitude and the azimuth angle is as shown in formula (5). When the target to be illuminated is an ideal curved surface, the relationship between the scattering field amplitude and the azimuth angle is as shown in formula (6).

[0130] In one embodiment, the scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle in the scattering center evolution model construction module is shown in Equation (10). The expression of the evolution function is shown in Equation (11).

[0131] In one embodiment, the scattering center evolution model parameter estimation module is further used to extract the scattering center of the target ISAR image using Newton's orthogonal matching pursuit based on the ISAR image samples of the partial attitude of the airborne target to be reconstructed; after the scattering center is extracted, the parameters to be estimated are estimated using the simplex method based on the ISAR image samples of the partial attitude of the airborne target to be reconstructed; the parameters to be estimated include: edge scattering intensity coefficient C, endpoint relative edge intensity ratio coefficient P, edge evolution function order β, and correction factor d; based on the endpoint scattering center, the estimated values ​​of the parameters to be estimated, and the scattering center evolution model, the scattering center evolution model of the airborne target to be reconstructed is obtained.

[0132] In one embodiment, the scattering center evolution model parameter estimation module is also used to simplify the scattering center evolution model to obtain the SCEM simplified function; based on the SCEM simplified function and the range of the scattering center, an initial value is given, and a dictionary is established; the dictionary is shown in formula (12) and formula (13). The residual image RSE is initialized as the input image, and the atom that best matches RSE is selected from the dictionary; the atom that best matches RSE is used as the initial point, and the Newton method is used to solve the fine estimation of the atomic parameters, and the fine estimation result of the atomic parameters is added to the estimated parameter set; wherein, the fine estimation of atomic parameters is shown in formula (14).

[0133] By using least squares estimation, the optimal coefficients for approximating the input image I using the atoms in the estimated parameter set are determined; based on the input image... I The optimal coefficients and atomic parameters are finely estimated, and the residual image RSE is updated. The next round of solving is carried out until the current residual image RSE can no longer extract effective endpoint scattering centers or the number of extracted endpoints has met the requirements, and the target ISAR image scattering center extraction result is obtained. The residual image RSE update formula is shown in formula (15).

[0134] In one embodiment, the optimal coefficients of the input image I in the scattering center evolution model parameter estimation module are shown in Equation (16).

[0135] In one embodiment, the parameters to be estimated include: an edge scattering intensity coefficient C, an end point relative edge intensity proportion coefficient P, an edge evolution function order β, and a correction factor d; the scattering center evolution model parameter estimation module is further configured to: determine an evolution curve of edge scattering and end point scattering center intensity with respect to an azimuth angle according to the position and intensity of the scattering center extracted by the scattering center extraction module; estimate the parameters to be estimated by using a simplex method according to the evolution curve of edge scattering and end point scattering center intensity with respect to the azimuth angle and an input objective function; and the input objective function is shown in formula (20).

[0136] It can be understood that the specific explanation of the air target recognition template library reconstruction device can refer to the corresponding explanation of the embodiments of the air target recognition template library reconstruction method in the above, and will not be repeated here. Each module in the above air target recognition template library reconstruction device can be realized by software, hardware and combinations thereof in whole or in part. The above modules can be embedded in or independent of a device with data processing function in hardware form, or can be stored in the memory of the above-mentioned device in software form, so as to call and execute the operations corresponding to each module by the processor. The above-mentioned device can be, but is not limited to, various types of data processing computer devices in the prior art.

[0137] In one embodiment, a computer device is also provided, which includes a memory and a processor, the memory stores a computer program, and the processor implements the steps in the above embodiments of the air target recognition template library reconstruction method when executing the computer program.

[0138] It can be understood that the above computer device, in addition to the above-mentioned memory and processor, also includes other software and hardware components not listed in the specification, which can be determined according to the specific model of the image processing computer in different application scenarios, and the specification will not be listed in detail.

[0139] In one embodiment, the processor executing the computer program can also implement the steps or sub-steps added in the above embodiments of the air target recognition template library reconstruction method.

[0140] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when the computer program is executed, the processes of the above-mentioned embodiments of the methods can be included. Any reference to memory, storage, database or other medium used in the embodiments provided in the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus dynamic random access memory (Rambus DRAM, RDRAM for short) and interface dynamic random access memory (DRDRAM).

[0141] The technical features of the above embodiments can be combined in any way. In order to make the description simple, all possible combinations of the technical features in the above embodiments are not described, but as long as the combination of the technical features does not exist, it should be considered as the scope of the present application.

[0142] The above embodiments only express several implementation manners of the present application, and the description is specific and detailed, but it should not be understood as a limitation on the protection scope of the present application. It should be pointed out that for ordinary skilled in the art, without departing from the concept of the present application, some modifications and improvements can be made, which all belong to the protection scope of the present application.

Claims

1. A method for reconstructing an aerial target recognition template library, characterized in that, Including the following steps: Analyze the prior information of the attribute scattering center model to derive the coupling relationship between the target scattering coefficient amplitude and azimuth angle under the physical optics method; Based on the prior information of the attribute scattering center model and the coupling relationship between the target scattering coefficient amplitude and the azimuth angle, a scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle is constructed. Based on the ISAR image samples of the partial attitude of the aerial target to be reconstructed, the parameters of the scattering center evolution model are estimated by Newton's orthogonal matching pursuit and simplex method, and the scattering center evolution model of the aerial target to be reconstructed is obtained. The ISAR image of the aerial target to be reconstructed at any azimuth angle is generated by using the scattering center evolution model of the aerial target to be reconstructed, thereby realizing the reconstruction of the aerial target identification template library. The evolution model of the scattering center, which reflects the coupling relationship between the target scattering center and the azimuth angle, is as follows: in, This represents the evolution model of the scattering center. This represents the number of clusters in the edge scattering set; Indicates the first j The number of scattering centers at the edge of the strip. , These represent the upward widening of the j-th edge in terms of orientation and the upward widening of the j-th edge in terms of distance, respectively. Indicates the first j The scattering intensity coefficient at the edge of the strip; This represents the occlusion function, which includes two types: gradual occlusion and complete occlusion. Indicates the first j The initial angle of the strip edge; Indicates the first j Evolutionary symbols for the edge of a strip; Indicates the first j Edge occlusion angle; express j The order of the edge evolution function; Represents the rotation function; Indicates azimuth. This represents the main lobe portion of the sinc function; , The coordinates represent the location of the scattering center. , Indicates the first j The coordinates of the edge scattering center; These represent the distance and azimuth resolution, respectively. Indicates the number of endpoint scattering centers; This represents the function for complete occlusion. This represents the initial angle corresponding to the m-th endpoint. Indicates the first m endpoints Evolutionary symbols Indicates the first m endpoints Angle of obstruction This represents the scattering intensity coefficient of the nth edge connected to the mth endpoint. This represents the initial angle of the nth edge connected to the mth endpoint. The evolution symbol representing the nth edge connected to the mth endpoint. This represents the order of the edge evolution function of the nth edge connected to the mth endpoint. Indicates the first m The number of edges connected to each endpoint; Indicates the first m The coordinates of the scattering center at each endpoint; Indicates the first m The endpoint corresponding to the first n The strength ratio factor of the strip edge; The mapping function between endpoints and edges, Let represent the evolution function, and the expression for the evolution function is: in d This represents a correction factor to prevent the denominator of the evolution function from taking a value of 0. Represents the azimuth angle as the independent variable. Represents a positive integer.

2. The method for reconstructing an aerial target recognition template library according to claim 1, characterized in that, The coupling relationship between the target scattering coefficient amplitude and azimuth angle under physical optics is derived, including: When the electrical size is large, the scattering field is determined using physical optics: in, Represents the scattered field. Represents free-space wave impedance. Indicates wave number, This represents the radial unit vector from the source point to the field point. Represents the unit normal vector pointing outwards from the surface; This represents the intensity of the incident magnetic field at a point on the surface. This represents the radial vector on the surface from the source point to the integration point; Represents the integral over the illuminated surface. This represents the distance from the source point to the field point. Indicates the illuminated surface; When the illuminated target is an ideal sphere or cylinder, its target scattering coefficient under physical optics is isotropic in the azimuth angle, and the relationship between the scattered field amplitude and the azimuth angle is as follows: in, This represents the remaining terms after excluding the azimuth coupling term when the illuminated target is an ideal sphere or cylinder, and is independent of the azimuth angle. It is the azimuth angle; When the illuminated target is an ideal plane, the relationship between the amplitude of the scattered field and the azimuth angle is as follows: in, This represents the remaining terms after excluding the azimuth coupling term when the illuminated target is an ideal plane, and is independent of the azimuth angle; When the illuminated target is an ideal curved surface, the relationship between the amplitude of the scattered field and the azimuth angle is as follows: in, This represents the remaining terms excluding the azimuth coupling term when the illuminated target is an ideal curved surface, and is independent of the azimuth angle.

3. The method for reconstructing an aerial target recognition template library according to claim 1, characterized in that, Based on ISAR image samples of the partial attitude of the aerial target to be reconstructed, the parameters of the scattering center evolution model are estimated using Newton's orthogonal matching pursuit and the simplex method, resulting in the scattering center evolution model of the aerial target to be reconstructed, including: Based on the ISAR image samples of the attitude of the aerial target to be reconstructed, the scattering center of the target ISAR image is extracted by Newton's orthogonal matching pursuit. After the scattering center is extracted, the simplex method is used to estimate the parameters to be estimated based on the ISAR image samples of the attitude of the aerial target to be reconstructed. The parameters to be estimated include: edge scattering intensity coefficient C, endpoint relative edge intensity ratio coefficient P, edge evolution function order β, and correction factor d. Based on the endpoint scattering center, the estimated values ​​of the parameters to be estimated, and the scattering center evolution model, the scattering center evolution model of the aerial target to be reconstructed is obtained.

4. The method for reconstructing an aerial target recognition template library according to claim 3, characterized in that, Based on the ISAR image samples of the partial attitude of the aerial target to be reconstructed, Newton's orthogonal matching pursuit is used to extract the scattering center of the target's ISAR image, including: The scattering center evolution model is simplified to obtain the SCEM simplified function; Based on initial values ​​given by the SCEM reduction function and the range of the scattering centers, a dictionary is established; the dictionary is as follows: in, For a dictionary, for The normalization result, For a given azimuth angle, the simplified SCEM function is the first... k One portion, The coordinates of the scattering center are... For the first k Parameters of the echo signal of a single ASC. For parameter set, , , These are the sets of parameters for edge scattering centers and the sets of parameters for endpoint scattering centers, respectively. , These are the parameter sets for the intensity coefficients of the edge scattering centers, the lateral coordinates of the edge scattering centers, and the longitudinal coordinates of the edge scattering centers, respectively. , These are the parameter sets for the intensity coefficients of the endpoint scattering center, the horizontal coordinates of the endpoint scattering center, and the vertical coordinates of the endpoint scattering center, respectively. The residual image RSE is initialized as the input image, and the atoms that best match RSE are selected from the dictionary; Using the atom that best matches RSE as the initial point, Newton's method is used to solve for a fine estimate of the atomic parameters. The result of the fine estimate of the atomic parameters is then added to the estimated parameter set. The fine estimate of the atomic parameters is as follows: in, For the fine estimation results of atomic parameters, Represented as The normalized result is taken as conjugate. Represents the residual image; The optimal coefficients for approximating the input image I using the atoms in the estimated parameter set are determined by least squares estimation. Based on the input image I The optimal coefficients and atomic parameters are used to refine the estimation results, and the residual image RSE is updated. The next round of solving continues until the current residual image RSE can no longer extract effective endpoint scattering centers or the number of extracted endpoints meets the requirements, thus obtaining the target ISAR image scattering center extraction result. The residual image RSE update formula is: in, Indicates the first i The optimal coefficient group corresponding to the _ atom is the _ _th k One optimal coefficient; I For the input image, The first result of the optimization of the simplified SCEM function under a given azimuth angle. k One portion, The coordinates of the scattering center are... This is the result of a fine estimation of atomic parameters.

5. The method for reconstructing an aerial target recognition template library according to claim 4, characterized in that, Input image I The optimal coefficient is: in, This represents the coefficient group corresponding to each atom. Indicates the first k The coefficients corresponding to each atom k =1,2,……, i ; Indicates use Using atoms to approximate the input image I The optimal coefficient, Indicates use The Middle k Atoms are used to approximate the input image. I The optimal coefficient.

6. The method for reconstructing an aerial target recognition template library according to claim 3, characterized in that, After the scattering center is extracted, the parameters to be estimated are estimated using the simplex method based on the ISAR image samples of the partial attitude of the aerial target to be reconstructed, including: Based on the location and intensity of the scattering center extracted from the scattering center, determine the evolution curves of the intensity of the edge scattering and endpoint scattering centers with respect to the azimuth angle; Based on the evolution curves of the intensity of edge scattering and endpoint scattering centers with respect to azimuth angle and the input objective function, the simplex method is used to estimate the parameters to be estimated; the input objective function is: in, The curves representing the function curves between the intensity and azimuth of two different scattering centers, i.e. and , This represents the set of parameters to be estimated. P represents the relative strength ratio of the endpoint to the edge. The edge scattering intensity coefficient C is represented. β represents the order of the marginal evolution function. This represents the correction factor d.

7. An aerial target identification template library reconstruction device, characterized in that, include: The prior information and physical optics derivation module is used to analyze the prior information of the attribute scattering center model and derive the coupling relationship between the target scattering coefficient amplitude and azimuth angle under the physical optics method. The scattering center evolution model construction module is used to construct a scattering center evolution model that reflects the coupling relationship between the target scattering center and the azimuth angle based on the prior information of the attribute scattering center model and the coupling relationship between the target scattering coefficient amplitude and the azimuth angle. The scattering center evolution model parameter estimation module is used to estimate the parameters of the scattering center evolution model based on the ISAR image samples of the partial attitude of the air target to be reconstructed, using Newton's orthogonal matching pursuit and simplex method, so as to obtain the scattering center evolution model of the air target to be reconstructed. The aerial target recognition template library reconstruction module is used to generate ISAR images of the aerial target to be reconstructed at any azimuth angle using the scattering center evolution model of the aerial target to be reconstructed, thereby realizing the reconstruction of the aerial target recognition template library to be reconstructed. Among them, the scattering center evolution model reflecting the coupling relationship between the target scattering center and the azimuth angle in the scattering center evolution model construction module is as follows: in, This represents the evolution model of the scattering center. This represents the number of clusters in the edge scattering set; Indicates the first j The number of scattering centers at the edge of the strip. , These represent the upward widening of the j-th edge in terms of orientation and the upward widening of the j-th edge in terms of distance, respectively. Indicates the first j The scattering intensity coefficient at the edge of the strip; This represents the occlusion function, which includes two types: gradual occlusion and complete occlusion. Indicates the first j The initial angle of the strip edge; Indicates the first j Evolutionary symbols for the edge of a strip; Indicates the first j Edge occlusion angle; express j The order of the edge evolution function; Represents the rotation function; Indicates azimuth. This represents the main lobe portion of the sinc function; , The coordinates represent the location of the scattering center. , Indicates the first j The coordinates of the edge scattering center; These represent the distance and azimuth resolution, respectively. Indicates the number of endpoint scattering centers; This represents the function for complete occlusion. This represents the initial angle corresponding to the m-th endpoint. Indicates the first m endpoints Evolutionary symbols Indicates the first m endpoints Angle of obstruction This represents the scattering intensity coefficient of the nth edge connected to the mth endpoint. This represents the initial angle of the nth edge connected to the mth endpoint. The evolution symbol representing the nth edge connected to the mth endpoint. This represents the order of the edge evolution function of the nth edge connected to the mth endpoint. Indicates the first m The number of edges connected to each endpoint; Indicates the first m The coordinates of the scattering center at each endpoint; Indicates the first m The endpoint corresponding to the first n The strength ratio factor of the strip edge; The mapping function between endpoints and edges, Let represent the evolution function, and the expression for the evolution function is: in d This represents a correction factor to prevent the denominator of the evolution function from taking a value of 0. Represents the azimuth angle as the independent variable. Represents a positive integer.

8. A computer device, comprising a memory and a processor, characterized in that, The memory stores a computer program, and when the processor executes the computer program, it implements the steps of the aerial target recognition template library reconstruction method according to any one of claims 1 to 6.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the aerial target recognition template library reconstruction method according to any one of claims 1 to 6.

Images