Non-cooperative target radar image generation method and device, non-cooperative target radar image training method and device and medium

Abstract

The invention provides a non-cooperative target radar image generation method, a training method, a device and a medium, an image generation model is constructed, and the image generation model comprises a diffusion network used for image generation and a control network used for inputting control conditions to the diffusion network. According to the method, the control network takes the edge images of different postures as the control conditions of the diffusion network, so that the image generation model obtained through training can generate the corresponding radar image of the specified posture according to the edge image of the specified posture, and the formed radar image database can be more accurately applied to the recognition of the non-cooperative target. Besides, the non-cooperative target radar image is generated by adopting the image generation model comprising the diffusion network and the control network, the processing speed is high, and the processed image is clearer and better in visual effect due to the fact that the model adopts more grids.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a non-cooperative target radar image generation method, training method, device and medium. Background Art

[0002] Radar has broad application prospects in the field of target recognition due to its advantages such as long range, low impact of climatic conditions, and high resolution. In the process of radar detection, the target has a strong posture sensitivity, so building a radar image database of targets with different postures is the basis of target recognition.

[0003] Non-cooperative targets refer to the real position information of the detected target. Except for direct measurement by sensors, there is no other technical means to obtain the accurate position of the target, such as incoming missiles, enemy aircraft, failed or malfunctioning spacecraft, etc.

[0004] Generative adversarial networks can be used to train existing radar images to generate images that are consistent with the style of real radar images of non-cooperative targets. However, generative adversarial networks have the problem of random generation results and are difficult to generate radar images of specified postures of non-cooperative targets. At the same time, in terms of visual effects, there is a large gap between the images generated by generative adversarial networks and real radar images, which in turn affects the effective recognition of non-cooperative targets based on radar image databases. Summary of the invention

[0005] The present application proposes a non-cooperative target radar image generation method, training method, device and medium, which can solve the problem of difficulty in generating non-cooperative target specified posture radar images.

[0006] In order to achieve the above objectives, this application adopts the following technical solutions:

[0007] In a first aspect, a training method for a non-cooperative target radar image generation model is provided, wherein the image generation model includes a diffusion network and a control network for inputting control conditions into an image generation process of the diffusion network, and the training method includes:

[0008] Performing three-dimensional modeling on the non-cooperative target to obtain a target model;

[0009] Simulating and obtaining two-dimensional electromagnetic images and edge images of the target model in different postures;

[0010] The electromagnetic image and the edge image are used as training sets of the image generation model to train the image generation model, wherein the edge image is used as an input of the control network.

[0011] Based on the above technical solution, an image generation model is constructed, which includes a diffusion network for image generation and a control network for inputting control conditions to the diffusion network. The control network uses edge images of different postures as control conditions of the diffusion network. In this way, the trained image generation model can generate corresponding radar images of specified postures based on edge images of specified postures, and then the radar image database formed can be more accurately applied to the recognition of non-cooperative targets. In addition, the use of an image generation model including a diffusion network and a control network to generate radar images of non-cooperative targets not only has a fast processing speed, but also, because the model uses more grids, the processed images are clearer and have better visual effects.

[0012] In a possible design of the first aspect,

[0013] The two-dimensional electromagnetic images of the target model in different postures are obtained by simulation, specifically:

[0014] The large-surface physical optics method is used to simulate the electromagnetic field data of non-cooperative targets at specified azimuth and elevation angles:

[0015]

[0016] Among them, E s is the electromagnetic field data, M is the number of strong scattering points on the non-cooperative target, is the beam vector in the direction of radar electromagnetic wave propagation, A i represents the amplitude of the scattered field of the i-th strong scattering point, e is a natural constant, j represents an imaginary unit, k x and k y They represent the beam vectors The projection values ​​on the coordinate axis X-axis and Y-axis, xi and yi represent the projection values ​​of the position vector of the i-th strong scattering point on the coordinate axis X-axis and Y-axis respectively;

[0017] Perform a two-dimensional inverse Fourier transform on the electromagnetic field data to obtain a two-dimensional electromagnetic image:

[0018]

[0019] Among them, ISAR(x,y) is a two-dimensional electromagnetic image, δ(xx i ,yy i ) means at position (x i ,y i ) is the impulse function at .

[0020] Based on the above technical solution, the stripe noise caused by severe defocusing of strong scattering points can be removed and the image quality can be optimized.

[0021] In a possible design of the first aspect,

[0022] Before training:

[0023] The electromagnetic image is subjected to frequency filtering to remove high-frequency components corresponding to strong defocusing in the electromagnetic image, specifically:

[0024] H(x,y)=FFT(ISAR(x,y))·F(x,y)

[0025] in,

[0026]

[0027] FFT stands for Fast Fourier Transform, h(x,y) is the impulse response when imaging any scattering point of a non-cooperative target, and the image size is M×N.

[0028] In a possible design of the first aspect,

[0029] The edge image obtained by simulation is as follows:

[0030] The edge image is extracted from the simulation images of the target model from different viewing angles by the modeling software through the Canny operator.

[0031] In a possible design of the first aspect,

[0032] The output of the image generation model is:

[0033]

[0034] in, represents the diffusion network, x is the input image of the diffusion network during training, Θ represents the parameters of the diffusion network, Z(·;·) represents a convolution with a size of 1×1 and initial weights and biases of 0, Θ z1 and θ z2 denotes the parameters of two zero convolutions, Θ c represents a trainable parameter, c represents the control condition input into the control network, and yc is the output image incorporating the control condition.

[0035] In a second aspect, a non-cooperative target radar image generation method is provided, the generation method being based on the image generation model trained by the training method of the first aspect, the generation method comprising:

[0036] Inputting a current edge image of a specified posture into the trained image generation model;

[0037] The image generation model processes to obtain a current electromagnetic image corresponding to the current edge image.

[0038] In a third aspect, a training device for a non-cooperative target radar image generation model is provided, wherein the image generation model includes a diffusion network and a control network for inputting control conditions into an image generation process of the diffusion network, and the training device includes:

[0039] A modeling unit, used for performing three-dimensional modeling on the non-cooperative target to obtain a target model;

[0040] A simulation calculation unit, used for simulating and obtaining two-dimensional electromagnetic images and edge images of the target model in different postures;

[0041] A training unit is used to train the image generation model by using the electromagnetic image and the edge image as training sets of the image generation model, wherein the edge image is used as an input of the control network.

[0042] In a fourth aspect, a non-cooperative target radar image generation device is provided, the image generation device is based on the image generation model trained by the training device in the third aspect, and the image generation device includes:

[0043] An input unit, used for inputting a current edge image of a specified posture into the trained image generation model;

[0044] A processing unit is used for processing the image generation model to obtain a current electromagnetic image corresponding to the current edge image.

[0045] In a fifth aspect, a computer device is provided, the computer device comprising: a processor, and a memory coupled to the processor,

[0046] The memory is used to store computer programs;

[0047] The processor is used to execute the computer program stored in the memory, so that the computer device executes the training method of the non-cooperative target radar image generation model as described in the first aspect, or executes the non-cooperative target radar image generation method as described in the second aspect.

[0048] In a sixth aspect, a computer-readable storage medium is provided, wherein the computer-readable storage medium includes a computer program or instructions. When the computer program or instructions are executed on a computer, the computer executes the training method for the non-cooperative target radar image generation model as described in the first aspect, or executes the non-cooperative target radar image generation method as described in the second aspect. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1It is a schematic diagram of the overall process of the training method of the non-cooperative target radar image generation model and the non-cooperative target radar image generation method provided in the embodiment of the present application;

[0050] Figure 2 is a schematic diagram of non-cooperative target detection in an embodiment of the present application;

[0051] Figure 3 is a schematic diagram of a non-cooperative target radar imaging plane according to an embodiment of the present application;

[0052] Figure 4 is a schematic diagram of image denoising in an embodiment of the present application;

[0053] Figure 5 is a schematic diagram of a model image and an edge image at different azimuths and elevation angles according to an embodiment of the present application;

[0054] Figure 6 It is a schematic diagram for comparing the effects of the embodiments of the present application. DETAILED DESCRIPTION

[0055] The technical solution in this application will be described below in conjunction with the accompanying drawings.

[0056] The technical solution of the present application is mainly used for the establishment of a non-cooperative target radar image database, and the subsequent identification of non-cooperative targets based on the database.

[0057] For example, Figure 1 As shown, the embodiment of the present application provides a training method covering a non-cooperative target radar image generation model and an overall process of a non-cooperative target radar image generation method, specifically including: non-cooperative target geometric modeling, data acquisition and preprocessing, model training and model reasoning processes, wherein data acquisition and preprocessing specifically include: echo calculation, echo imaging, image denoising, edge image acquisition and other processes.

[0058] The above overall process is described in detail below.

[0059] 1. Non-cooperative target geometry modeling

[0060] This example selects the non-cooperative target E-3 early warning aircraft as the object of simulation modeling. The aircraft is a Boeing 707 passenger aircraft with a rotating radar module installed on the fuselage, and has the characteristics of both civil aircraft and military aircraft. According to the public parameters, the CAD model of the E-3 early warning aircraft is established. Figure 2 As shown, the center of mass of the aircraft is the origin O, the direction pointing to the nose is the X axis, the Y axis points to the direction perpendicular to the OX axis in the horizontal plane, and the Z axis points to the right-hand rule. The pitch angle of the radar relative to the target is θ, and the azimuth angle is

[0061] 2. Data acquisition and preprocessing

[0062] 2.1 Echo calculation

[0063] In order to simulate the small-bandwidth, small-angle turntable imaging process in ISAR radar imaging, the millimeter-wave frequency band stepped frequency signal used in the simulation process has a center frequency of 80GHz, a signal bandwidth of 640MHz, 128 range image sampling points, and a range resolution of 0.23m; the azimuth rotation angle is 0.45 degrees, the azimuth sampling points are 128, and the resolution is 0.23m. In order to shorten the simulation time and save memory consumption, the large-surface physical optics method is selected as the simulation calculation method, and the electromagnetic field data under different postures, that is, different azimuth and elevation angles, are simulated.

[0064] 2.2 Echo imaging

[0065] The target scattered electromagnetic field can be regarded as the superposition of the scattered fields of all the strong scattering centers on the target. The ISAR image is the electromagnetic scattering intensity distribution of all the strong scattering centers on the target in the azimuth-range plane. The two-dimensional imaging plane is as follows: Figure 3 shown.

[0066] The total scattered field of the non-cooperative target obtained in Section 2.1 can be expressed as:

[0067]

[0068] The scattered field E in formula (1) s It is regarded as the linear superposition of the scattering fields of M strong scattering points on the non-cooperative target, and its magnitude is proportional to the beam vector in the direction of radar electromagnetic wave propagation. About. i represents the amplitude of the scattered field of the i-th strong scattering point, e is a natural constant, j represents an imaginary unit, is the position vector of the i-th strong scattering point on the target relative to the coordinate origin O. and It can be expressed as formula (2), where and Represents the unit vectors in the X-axis and Y-axis directions, k x and k y and x i and i They represent the beam vectors and the position vector of the i-th strong scattering point The projection value on the coordinate axis.

[0069]

[0070] According to formula (1) and formula (2), the final scattered field can be expressed as follows:

[0071]

[0072] In the small-bandwidth, small-angle turntable ISAR imaging scenario, the scattering field data of the non-cooperative target is subjected to a two-dimensional inverse Fourier transform to obtain the spatial distribution of the target's strong scattering points, that is, the ISAR image of the target:

[0073]

[0074] In formula (4), ISAR(x,y) represents the imaging result after two-dimensional inverse Fourier transform, π represents pi, and δ(x i ,yy i ) means at position (x i ,y i ) is the impulse function at .

[0075] 2.3 Image Denoising

[0076] It can be seen from formula (4) that, in an ideal situation (angle and bandwidth are infinite), the ISAR image of the target will be composed of M strong scattering points on non-cooperative targets. However, in reality, due to the limitations of illumination angle and transmission bandwidth, the upper and lower limits of the integral are not infinite. According to the actual situation, formula (4) can be rewritten as:

[0077]

[0078] h(x,y) in formula (5) is the point spread function, which is defined as the impulse response when imaging any scattering point of a non-cooperative target. It can be seen from formula (5) that when the bandwidth and angle are limited, the ISAR image of the scattering point is no longer an ideal point target, but a two-dimensional sinc function, resulting in defocus. At some strong scattering points (such as the nose, tail, engine, etc.), the defocus phenomenon is very serious, which may form strip noise that runs through the image, seriously deteriorating the image quality, such as Figure 4 (a) shown.

[0079] For example Figure 4 The image shown in (a) is subjected to a two-dimensional Fourier transform to obtain Figure 4 (b) shows that when there is horizontal strip noise in the original image, there will be a vertical high-frequency component at the center of the image spectrum. This high-frequency component can be filtered out by setting a suitable filter.

[0080] Assuming that the size of the image to be processed is M×N, the corresponding two-dimensional spectrum graph has the same size as the original image, and the filter F(x,y) can be expressed as:

[0081]

[0082] The filtering process can be expressed as:

[0083] H(x,y)=FFT(ISAR(x,y))·F(x,y) (7)

[0084] H(x,y) represents the result after frequency domain denoising, and FFT represents two-dimensional fast Fourier transform.

[0085] Perform a two-dimensional inverse Fourier transform on the filtered spectrum H(x,y) and convert it to the spatial domain to obtain the denoised result. Figure 4 (c) Figure 4 (d) shows the two-dimensional spectrum of the denoised image. It can be seen that after denoising, the original high-frequency component no longer exists in the center of the spectrum.

[0086] 2.4 Edge Image Acquisition

[0087] Under different azimuth angles, the target points in different directions; under different elevation angles, the size of the target will change due to projection. By importing the target model into the modeling software and modifying the angle of view according to the simulation conditions, you can get the model images under different angles (postures). Using the Canny operator to extract edge information from the model image, you can get the edge map of the target under different angles, such as Figure 5 shown.

[0088] 3. Model training

[0089] The embodiment of the present application adopts the ControlNet model structure for training, using the image obtained by electromagnetic calculation as the input image and the corresponding edge image as the control image to obtain the final generation model weight. ControlNet is developed from the diffusion model. In order to retain the original image generation capability of the pre-trained diffusion model, ControlNet freezes the backbone structure of the pre-trained diffusion model, only trains the copied part, and injects the features extracted during the training process into the decoder of the backbone structure through zero convolution, thereby achieving the introduction of additional control conditions to fine-tune the network parameters while retaining the generation capability of the pre-trained model. The output of the improved network is:

[0090]

[0091] represents the original backbone network, x is the input image of the backbone network during training, Θ represents the parameters of the backbone network; Z(·;·) represents a convolution with a size of 1×1 and initial weights and biases of 0, Θ z1 and θ z2 denotes the parameters of two zero convolutions, Θ crepresents a trainable parameter, c represents a control condition input into the control network, and in the embodiment of the present application, it is an edge image at a specified angle of the target; y c is the output image incorporating the control condition.

[0092] The contents of points 1-3 above mainly correspond to the training method of the non-cooperative target radar image generation model in the embodiment of the present application.

[0093] 4. Model Reasoning

[0094] Using the trained model above, the edge image of the non-cooperative target at the specified angle obtained above is used as the control image input to output the corresponding non-cooperative target radar image.

[0095] To further demonstrate the effectiveness of the present invention, the following simulation results are used to provide additional explanations. Using the same electromagnetic simulation radar image as a training sample, the method proposed by the present invention and the traditional method are used to generate radar images of non-cooperative targets at a specified angle. The comparison results are shown in Figure 2. Figure 6 shown.

[0096] Depend on Figure 6 It can be seen that the method used in the embodiment of the present application can accurately generate a non-cooperative target radar image at a specified azimuth and elevation angle. This method can not only achieve the generation of a specified angle, but also the visual effect of the generated image is significantly better than the generation result of the traditional method, and is closer to the real image.

[0097] The content of the above point 4 mainly corresponds to the non-cooperative target radar image generation method of the embodiment of the present application.

[0098] Accordingly, an embodiment of the present application further provides a training device for a non-cooperative target radar image generation model, wherein the image generation model includes a diffusion network and a control network for inputting control conditions into the image generation process of the diffusion network, and the training device includes:

[0099] A modeling unit, used for performing three-dimensional modeling on the non-cooperative target to obtain a target model;

[0100] A simulation calculation unit, used for simulating and obtaining two-dimensional electromagnetic images and edge images of the target model in different postures;

[0101] A training unit is used to train the image generation model by using the electromagnetic image and the edge image as training sets of the image generation model, wherein the edge image is used as an input of the control network.

[0102] Correspondingly, the embodiment of the present application further provides a non-cooperative target radar image generation device, the image generation device is based on the image generation model trained by the above-mentioned training device, and the image generation device includes:

[0103] An input unit, used for inputting a current edge image of a specified posture into the trained image generation model;

[0104] A processing unit is used for processing the image generation model to obtain a current electromagnetic image corresponding to the current edge image.

[0105] Accordingly, an embodiment of the present application further provides a computer device, the computer device comprising: a processor, and a memory coupled to the processor,

[0106] The memory is used to store computer programs;

[0107] The processor is used to execute the computer program stored in the memory, so that the computer device executes part or all of the above-mentioned overall process.

[0108] Accordingly, an embodiment of the present application also provides a computer-readable storage medium, which includes a computer program or instructions. When the computer program or instructions are executed on a computer, the computer executes part or all of the above-mentioned overall process.

[0109] It should be noted that:

[0110] In the embodiments of the present application, the words "exemplarily", "for example", etc. are used to indicate examples, illustrations or explanations. Any embodiment or design described as "exemplary" in the present application should not be understood as being more preferred or more advantageous than other embodiments or designs. Specifically, the use of the word "exemplarily" is intended to present concepts in a specific way.

[0111] In the embodiments of the present application, (multiple words) can sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings to be expressed are the same. "of", "corresponding, relevant" and "corresponding" can sometimes be used interchangeably. It should be noted that when the distinction between them is not emphasized, the meanings to be expressed are the same.

[0112] It should be understood that the processor in the embodiments of the present application may be a central processing unit (CPU), and the processor may also be other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may also be any conventional processor, etc.

[0113] It should also be understood that the memory in the embodiments of the present application may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may be a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may be a random access memory (RAM), which is used as an external cache. By way of example and not limitation, many forms of random access memory (RAM) are available, such as static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link DRAM (SLDRAM), and direct rambus RAM (DR RAM).

[0114] The embodiments of the present application do not limit the execution order of the steps in the above method. For example, A may first execute step 1 and then execute step 2, or A may first execute step 2 and then execute step 1, or A may execute step 1 and step 2 at the same time.

[0115] The above embodiments can be implemented in whole or in part by software, hardware (such as circuits), firmware or any other combination. When implemented using software, the above embodiments can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions or computer programs. When the computer instructions or computer programs are loaded or executed on a computer, the process or function described in the embodiment of the present application is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium. For example, the computer instructions can be transmitted from one website site, computer, server or data center to another website site, computer, server or data center by wired (such as infrared, wireless, microwave, etc.). The computer-readable storage medium can be any available medium that a computer can access or a data storage device such as a server or data center that contains one or more available media sets. The available medium can be a magnetic medium (for example, a floppy disk, a hard disk, a tape), an optical medium (for example, a DVD), or a semiconductor medium. The semiconductor medium can be a solid-state hard disk.

[0116] It should be understood that in the various embodiments of the present application, the size of the serial numbers of the above-mentioned processes does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present application.

[0117] Those of ordinary skill in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0118] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0119] In the several embodiments provided in the present application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0120] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0121] In addition, each functional unit in each embodiment of the present application may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit.

[0122] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application, or the part that contributes to the prior art, or the part of the technical solution, can be embodied in the form of a software product, which is stored in a storage medium and includes several instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the various embodiments of the present application.

[0123] The aforementioned storage media include: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), magnetic disk or optical disk and other media that can store program codes.

[0124] The above is only a specific implementation of the present application, but the protection scope of the present application is not limited thereto. Any person skilled in the art who is familiar with the present technical field can easily think of changes or substitutions within the technical scope disclosed in the present application, which should be included in the protection scope of the present application. Therefore, the protection scope of the present application should be based on the protection scope of the claims.

Claims

1. A training method for a non-cooperative target radar image generation model, characterized in that: The image generation model includes a diffusion network and a control network for inputting control conditions into the image generation process of the diffusion network, and the training method includes: Performing three-dimensional modeling on the non-cooperative target to obtain a target model; Simulating and obtaining two-dimensional electromagnetic images and edge images of the target model in different postures; The electromagnetic image and the edge image are used as training sets of the image generation model to train the image generation model, wherein the edge image is used as an input of the control network.

2. The training method for the non-cooperative target radar image generation model according to claim 1, characterized in that: The two-dimensional electromagnetic images of the target model in different postures are obtained by simulation, specifically: The large-surface physical optics method is used to simulate the electromagnetic field data of non-cooperative targets at specified azimuth and elevation angles: Among them, E s is the electromagnetic field data, M is the number of strong scattering points on the non-cooperative target, is the beam vector in the direction of radar electromagnetic wave propagation, A i represents the amplitude of the scattered field of the i-th strong scattering point, e is a natural constant, j represents an imaginary unit, k x and k y They represent the beam vectors The projection values ​​on the coordinate axis X-axis and Y-axis, xi and yi represent the projection values ​​of the position vector of the i-th strong scattering point on the coordinate axis X-axis and Y-axis respectively; Perform a two-dimensional inverse Fourier transform on the electromagnetic field data to obtain a two-dimensional electromagnetic image: Among them, ISAR(x,y) is a two-dimensional electromagnetic image, δ(xx i ,yy i ) means at position (x i ,y i ) is the impulse function at .

3. The training method for the non-cooperative target radar image generation model as claimed in claim 2, characterized in that: Before training: The electromagnetic image is subjected to frequency filtering to remove high-frequency components corresponding to strong defocusing in the electromagnetic image, specifically: H(x,y)=FFT(ISAR(x,y))·F(x,y) in, FFT stands for Fast Fourier Transform, h(x,y) is the impulse response when imaging any scattering point of a non-cooperative target, and the image size is M×N.

4. The training method for the non-cooperative target radar image generation model according to claim 1, characterized in that: The edge image obtained by simulation is as follows: The edge image is extracted from the simulation images of the target model from different viewing angles by the modeling software through the Canny operator.

5. The training method for the non-cooperative target radar image generation model according to claim 1, characterized in that: The output of the image generation model is: in, represents the diffusion network, x is the input image of the diffusion network during training, Θ represents the parameters of the diffusion network, represents a convolution of size 1×1 with initial weights and biases of 0, Θ z1 and θ z2 denotes the parameters of two zero convolutions, Θ c represents a trainable parameter, c represents the control condition input into the control network, and yc is the output image incorporating the control condition.

6. A method for generating a non-cooperative target radar image, characterized in that: The generation method is based on the image generation model trained by the training method as shown in any one of claims 1 to 5, and the generation method includes: Inputting a current edge image of a specified posture into the trained image generation model; The image generation model processes to obtain a current electromagnetic image corresponding to the current edge image.

7. A training device for a non-cooperative target radar image generation model, characterized in that: The image generation model includes a diffusion network and a control network for inputting control conditions into the image generation process of the diffusion network, and the training device includes: A modeling unit, used for performing three-dimensional modeling on the non-cooperative target to obtain a target model; A simulation calculation unit, used for simulating and obtaining two-dimensional electromagnetic images and edge images of the target model in different postures; A training unit is used to train the image generation model by using the electromagnetic image and the edge image as training sets of the image generation model, wherein the edge image is used as an input of the control network.

8. A non-cooperative target radar image generation device, characterized in that: The image generation device is based on the image generation model trained by the training device according to claim 7, and the image generation device comprises: An input unit, used for inputting a current edge image of a specified posture into the trained image generation model; A processing unit is used for processing the image generation model to obtain a current electromagnetic image corresponding to the current edge image.

9. A computer device, characterized in that: The computer device comprises: a processor, and a memory coupled to the processor, The memory is used to store computer programs; The processor is used to execute the computer program stored in the memory, so that the computer device executes the training method of the non-cooperative target radar image generation model as described in any one of claims 1 to 5, or executes the non-cooperative target radar image generation method as described in claim 6.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a computer program or instructions. When the computer program or instructions are executed on a computer, the computer executes the training method for a non-cooperative target radar image generation model as described in any one of claims 1 to 5, or executes the non-cooperative target radar image generation method as described in claim 6.

Images