Isar high-resolution robust imaging method based on hybrid constraint unfolding network

Abstract

The application discloses an ISAR high-resolution robust imaging method based on a mixed constraint unfolding network, and comprises the following steps: inputting a multi-variable characteristic parameter of an echo signal and an RD image of the echo signal into a trained parameter generation network to obtain a coefficient set, wherein the coefficient set comprises the numerical value of the coefficient used by each level of the mixed constraint unfolding sub-network; inputting the echo signal into the trained mixed constraint unfolding network to obtain a target image; wherein the mixed constraint unfolding network comprises a plurality of cascaded mixed constraint unfolding sub-networks, the current level of the mixed constraint unfolding sub-network is used for generating a current level of the iterative image according to the last level of the iterative image, the current level of the coefficient, and the last level of the iterative image is the target image; and the optimization target of the mixed constraint unfolding network comprises a data-driven item constraint. The method provided by the application has better robustness, higher imaging integrity and higher accuracy.

Description

Technical Field

[0001] This invention belongs to the field of radar signal processing technology, specifically relating to an ISAR high-resolution robust imaging method based on a hybrid constraint expansion network. Background Technology

[0002] Inverse Synthetic Aperture Radar (ISAR) possesses advantages such as all-weather, all-day operation, long range, and high resolution, playing an irreplaceable role in fields such as aerial target surveillance and space situational awareness. ISAR imaging methods based on deep unfolding networks utilize deep unfolding technology. By extracting a single iteration step from a traditional iterative optimization algorithm and unfolding it into a layer of a network, the sub-network corresponds to a complete iteration of the iterative algorithm. Adjustable parameters in each iteration are then mapped to trainable parameters for each network layer. This allows for data-driven training of the network parameters, thereby improving the algorithm's imaging efficiency and performance, and has become one of the mainstream technologies in ISAR imaging.

[0003] However, traditional methods based on unfolded networks only focus on sparsity constraints in their optimization objective function. The limitations in the design of the optimization objective function lead to overly sparse reconstructed images and distortion of target scattering point amplitudes, thereby reducing the integrity and accuracy of image reconstruction. Furthermore, the internal parameters of deep unfolded network methods are fixed after training. When the characteristic parameters of the echo signal change, the network needs to be retrained, resulting in excessively high spatial and temporal complexity. This makes it difficult to achieve dynamic adaptation to complex observation scenarios, which restricts its application in practical imaging scenarios. The above problems pose a great challenge to the current ISAR imaging methods for aerospace targets based on unfolded networks. Summary of the Invention

[0004] This invention provides a high-resolution robust imaging method for ISAR based on a hybrid constraint expansion network, which can solve the problems of low integrity and accuracy of the reconstructed target image and poor robustness caused by the single constraint in traditional methods.

[0005] In a first aspect, embodiments of the present invention provide an ISAR high-resolution robust imaging method based on a hybrid constraint unfolding network, the method comprising: The variable characteristic parameters of the echo signal and the RD image of the echo signal are input into the trained parameter generation network to obtain a set of coefficients, wherein the set of coefficients includes the values ​​of the coefficients used in each level of the hybrid constraint expansion sub-network. The echo signal is input into a trained hybrid constraint unpacking network to obtain the target image; The hybrid constraint expansion network includes multiple cascaded hybrid constraint expansion sub-networks. The current-level hybrid constraint expansion sub-network is used to generate the current-level iterative image based on the previous-level iterative image and the current-level coefficients. The last-level iterative image is the target image. The optimization objective of the hybrid constraint expansion network includes data-driven constraints.

[0006] Secondly, embodiments of the present invention provide an ISAR high-resolution robust imaging device based on a hybrid constraint expansion network, including a parameter estimation module and an iterative imaging module; The parameter estimation module is used to input the variable feature parameters of the echo signal and the RD image of the echo signal into the trained parameter generation network to obtain a coefficient set, wherein the coefficient set includes the values ​​of the coefficients used in each level of the hybrid constraint expansion network. The iterative imaging module is used to input the echo signal into a trained hybrid constraint unpacking network to obtain the target image; The hybrid constraint expansion network includes multiple cascaded hybrid constraint expansion sub-networks. The current hybrid constraint expansion sub-network is used to generate the current iteration image based on the previous iteration image, and the last iteration image is the target image. The optimization objective of the hybrid constraint expansion network includes data-driven constraints.

[0007] Thirdly, embodiments of the present invention provide an electronic device, including a processor and a memory, wherein the memory is used to store a computer program; the processor can be used to execute a calculator program (instructions) stored in the memory to implement the method of the first aspect described above.

[0008] Fourthly, embodiments of the present invention provide a computer-readable storage medium storing a computer program that, when executed, can implement the method described in the first aspect above.

[0009] The beneficial effects of this invention compared to the prior art are as follows: This invention generates a set of coefficients that can be used by the hybrid unfolding network by using variable feature parameters and RD images of echoes, so that the coefficients can be adaptively adjusted according to echo features, improving the dynamic adaptability and robustness to complex observation scenarios; by adding data-driven constraints to the optimization objective of the hybrid constraint unfolding network, its optimization process is no longer limited to sparsity, and the network's attention to the integrity of the target image is improved; thus, when generating target images through the trained hybrid constraint unfolding network, the imaging integrity and accuracy of the target image can be improved. Attached Figure Description

[0010] Figure 1 This is a schematic diagram of a hybrid constraint expansion subnetwork provided in an embodiment of the present invention; Figure 2 This is a schematic diagram of the structure of a sparse ISAR imaging network provided in an embodiment of the present invention; Figure 3 This is a schematic diagram of the specific structure of a parameter generation network provided in an embodiment of the present invention; Figure 4 The present invention provides a flowchart of the implementation of an ISAR high-resolution robust imaging method based on a hybrid constraint expansion network. Figure 5 A pseudocode diagram illustrating the process of generating a target image using a hybrid constraint unfolding network, as provided in an embodiment of the present invention; Figure 6 A schematic diagram of the structure of an ISAR high-resolution robust imaging device based on a hybrid constraint unfolding network provided in an embodiment of the present invention; Figure 7 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention; Figures 8a-8c This is a schematic diagram of ISAR imaging results using conventional techniques at different signal-to-noise ratios, provided by an embodiment of the present invention. Figures 9a-9c This is a schematic diagram of the ISAR imaging results of the present invention under different signal-to-noise ratios, provided by an embodiment of the present invention; Figures 10a-10c This is a schematic diagram of ISAR imaging results using conventional techniques under different defect rates, provided by an embodiment of the present invention. Figures 11a-11c This is a schematic diagram of ISAR imaging results under different defect rates provided in an embodiment of the present invention. Detailed Implementation

[0011] In the following description, specific details such as particular system architectures and techniques are set forth for illustrative purposes and not for limitation, in order to provide a thorough understanding of the embodiments of the invention. However, those skilled in the art will understand that the invention can be implemented in other embodiments without these specific details. In other instances, detailed descriptions of well-known systems, apparatuses, circuits, and methods are omitted so as not to obscure the description of the invention with unnecessary detail.

[0012] It should be understood that, when used in this specification and the appended claims, the term "comprising" indicates the presence of the described features, integrals, steps, operations, elements and / or components, but does not exclude the presence or addition of one or more other features, integrals, steps, operations, elements, components and / or collections thereof.

[0013] It should also be understood that the term “and / or” as used in this specification and the appended claims refers to any combination of one or more of the associated listed items and all possible combinations, and includes such combinations.

[0014] As used in this specification and the appended claims, the term "if" may be interpreted, depending on the context, as "when," "once," "in response to determination," or "in response to detection." Similarly, the phrase "if determined" or "if [described condition or event] is detected" may be interpreted, depending on the context, as meaning "once determined," "in response to determination," "once [described condition or event] is detected," or "in response to detection of [described condition or event]."

[0015] Furthermore, in the description of this invention and the appended claims, the terms "first," "second," "third," etc., are used only to distinguish descriptions and should not be construed as indicating or implying relative importance.

[0016] References to "one embodiment" or "some embodiments" as described in this specification mean that one or more embodiments of the invention include a specific feature, structure, or characteristic described in connection with that embodiment. Therefore, the phrases "in one embodiment," "in some embodiments," "in other embodiments," "in still other embodiments," etc., appearing in different parts of this specification do not necessarily refer to the same embodiment, but rather mean "one or more, but not all, embodiments," unless otherwise specifically emphasized. The terms "comprising," "including," "having," and variations thereof mean "including but not limited to," unless otherwise specifically emphasized.

[0017] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.

[0018] Example 1 As an example, a hybrid constraint expansion network can be built using deep expansion technology. Through multiple cascaded hybrid constraint expansion subnetworks, the echo signal can be iteratively reconstructed to finally obtain the target image.

[0019] In some embodiments, since the images of targets such as aircraft and satellites in ISAR images are "sparse," traditional techniques utilize this prior knowledge to reconstruct targets in ISAR images through L1 norm constraints. However, the L1 function overemphasizes sparsity and lacks consideration for the integrity of image reconstruction, focusing too much on the removal of noise and false points. This may lead to a decrease in the amplitude of some scattering points in the target, or even the loss of some weak scattering points, resulting in an overly sparse reconstructed image and distorted amplitude of target scattering points, causing a large reconstruction error. Therefore, this invention adds a data-driven constraint to the L1 norm sparsity constraint to compensate for the shortcomings of the original optimization objective function in not paying enough attention to image integrity.

[0020] In one possible implementation, a sparse observation model for the ISAR imaging problem with echo defects can be constructed first. Then, based on the optimization objective function with L1 norm sparsity constraints, data-driven constraints are added to obtain a hybrid constraint optimization objective function that places greater emphasis on image integrity.

[0021] For example, the optimization objective of the hybrid constraint expansion network provided by the present invention (i.e., the hybrid constraint optimization objective function mentioned above) can be expressed as: (1.1) in, This represents the optimization objective of the hybrid constraint expansion network. Indicates minimization. , The first and second coefficients are respectively. For data-driven constraints, For the target image, As a data-driven auxiliary variable, It is an L1 norm. Represents the L2 norm. for The regularization coefficient, , These are the observation matrices for the range and azimuth directions, respectively.

[0022] Specifically, the first coefficient is the regularization coefficient of the L1 norm, and the second coefficient is the regularization coefficient of the L2 norm.

[0023] In one possible implementation, the optimization problem of the hybrid constraint unfolding network can be decomposed into two problems: one for the target image and the other for the data-driven auxiliary variables. Then, the two problems are solved alternately, and a specific hybrid constraint unfolding network is constructed based on the update method used during the alternating solution of the two problems.

[0024] For example, the problem of solving the segmented target image can be represented as: (1.2).

[0025] For example, the problem of solving data-driven auxiliary variables can be represented as: (1.3).

[0026] In one example, the Iterative Soft-Thresholding Algorithm (ISTA) can be used to solve for the target image.

[0027] For example, the target image can be updated using the following formula: (1.4) in: (1.5) in, For the (t+1)th iteration image, The first coefficient of the t-th order. This is a soft threshold shrinkage function; , These are the auxiliary variables for the (t+1)th and tth levels of the image, respectively. , These represent the second coefficient at level t and the gradient descent step size, respectively. For the t-th iteration image, Indicates transpose. The echo signal, For the t-th level data-driven auxiliary variable.

[0028] Accordingly, an image-assisted update module and an image update module can be set in the hybrid constraint expansion sub-network. The image-assisted update module implements the calculation process of formula (1.5), and the image update module implements the calculation process of formula (1.4), thereby realizing the update of the target image.

[0029] In one example, since the data-driven constraint is a data regularization constraint, an analytical solution cannot be written. Therefore, a gradient update module can be used. The solution is performed on the data-driven auxiliary variables, and its structure is a residual block containing two convolutional layers.

[0030] For example, the update process of data-driven auxiliary variables can be represented as: (1.6) in, It is a data-driven auxiliary variable for the (t+1)th level.

[0031] Accordingly, a data-driven auxiliary update module can also be set in the hybrid constraint expansion network. For example, a gradient update module. The update of data-driven auxiliary variables is achieved through the data-driven auxiliary update module.

[0032] Therefore, see Figure 1The hybrid constraint expansion subnetwork may include an image-assisted update module, an image update module, and a data-driven auxiliary update module. The image-assisted update module is used to update the current-level image auxiliary variables based on the previous-level iterative image, the current-level coefficients, and the previous-level data-driven auxiliary variables; the image update module is used to generate the current-level iterative image based on the current-level image auxiliary variables and the current-level coefficients; and the data-driven auxiliary update module is used to update the current-level data-driven auxiliary variables based on the current-level iterative image.

[0033] Since the optimization objective of the hybrid constraint unfolding network includes data-driven constraints, and these data-driven constraints can directly learn the statistical regularities of real data through end-to-end training, thereby guiding the imaging results to converge towards the real data distribution and making up for the limitations of traditional sparse constraints, adding data-driven constraints to the optimization objective function can preserve the amplitude and location information of key scattering points while maintaining image sparsity, avoiding the loss of details caused by "excessive sparsity", enhancing structural rationality and physical consistency, and ultimately achieving more accurate and robust ISAR imaging.

[0034] Example 2 Figure 2 The diagram shown illustrates the structure of a sparse ISAR imaging network provided in an embodiment of the present invention. As an example and not a limitation, the network may include a parameter generation network and a hybrid constraint unrolling network.

[0035] In some embodiments, traditional methods use fixed internal parameters after training. When the characteristic parameters of the echo signal change, the network needs to be retrained, resulting in excessively high spatial and temporal complexity. This makes it difficult to dynamically adapt to complex observation scenarios, limiting its application in practical imaging scenarios. Therefore, see... Figure 2 The parameter generation network designed in this invention can be used to predict the values ​​of coefficients used in each level of the hybrid constraint expansion sub-network based on the variable characteristic parameters of the echo signal and the RD image of the echo signal. The hybrid constraint expansion network can include multiple cascaded hybrid constraint expansion sub-networks for iterative prediction of the target image.

[0036] In one possible implementation, see Figure 3 The parameter generation network mainly consists of convolutional layers. It can extract relevant features from the RD of the echo signal and its variable feature parameters, and generate the optimal internal parameters of the hybrid constraint unfolding network, i.e., the coefficient set. .

[0037] In one example, variable characteristic parameters refer to features in the echo signal that are prone to change, such as the defect rate and signal-to-noise ratio. Here, the defect rate is mainly used.

[0038] For example, the echo signal can be preprocessed to obtain an initial ISAR image that contains noise and defects. After that Input the feature parameter extraction module to extract the defect rate information.

[0039] Specifically, the feature parameter extraction module may contain convolutional layers and fully connected layers of multiple residual blocks.

[0040] In one possible implementation, the hybrid constraint expansion network here can be the hybrid constraint expansion network provided in Embodiment 1 above.

[0041] In one example, when solving for auxiliary variables, U-Net can be used instead of the gradient update module. U-Net is used as the data-driven auxiliary update module. In this case, the data-driven auxiliary variables can be updated using the following formula: (1.7).

[0042] For example, the parameter generation network and the hybrid constraint expansion network can be jointly trained to obtain a trained parameter generation network and a hybrid constraint expansion network.

[0043] Specifically, the RD map and defect rate of the sample echo signal can be input into the parameter generation network to obtain the sample coefficient set. Then, the sample coefficient set and the sample echo signal are input into the hybrid constraint expansion network to obtain the sample target image. After imaging is completed, the hybrid loss of the sample target image and the label image is estimated. The internal parameters of each hybrid constraint expansion sub-network and the parameter generation network are updated through backpropagation to complete one round of iterative training.

[0044] Alternatively, the loss function used during training can be: ,in, For mixed loss, , These are the first training sessions. Each sample target image and label image This represents the set of samples used in this training.

[0045] Traditional methods for ISAR imaging using depth unfolding principles to construct networks do not take into account the variable characteristics of echoes received by ISAR systems in actual observation scenarios. The network parameters are fixed after training, and the network needs to be retrained to adapt to new observation scenarios under different loss rates and signal-to-noise ratios. This lack of robustness in imaging under complex observation conditions not only results in high spatial and temporal complexity but also restricts its application in real-world complex observation scenarios.

[0046] The parameter generation network provided by this invention can extract features strongly correlated with imaging quality from variable feature parameters and echo signals that can affect imaging performance and change frequently. Based on these features, it dynamically estimates a set of coefficients suitable for the current conditions. Through this coefficient set, the nonlinear mapping and constraint strength of each stage of the subsequent hybrid constraint expansion network can be adjusted, ensuring that the iterative optimization process within the network matches the actual characteristics of the current echo. This ensures that the hybrid constraint expansion network can effectively converge under varying conditions, stably reconstructing a clear target image and improving imaging robustness. Furthermore, this approach avoids manual parameter tuning.

[0047] Example 3 The ISAR high-resolution robust imaging method based on hybrid constraint expansion network provided in this embodiment of the invention can be applied to electronic devices such as mobile terminals, personal laptops, supercomputers, and radar equipment, or electronic devices with radar signal interfaces. This embodiment of the invention does not impose any restrictions on the specific type of electronic device.

[0048] Figure 4 The diagram illustrates a flowchart of an ISAR high-resolution robust imaging method based on a hybrid constraint unfolding network, provided by an embodiment of the present invention. As an example and not a limitation, this method can be applied to the aforementioned electronic device. The method may include steps S401-S402, which are described below.

[0049] S401, input the variable characteristic parameters of the echo signal and the RD image of the echo signal into the trained parameter generation network to obtain the coefficient set.

[0050] S402, the echo signal is input into the trained hybrid constraint unfolding network to obtain the target image.

[0051] For example, the parameter generation network and the hybrid constraint expansion network can be the parameter generation network and the hybrid constraint expansion network included in the sparse ISAR imaging network in the above embodiment 2, respectively.

[0052] Specifically, see Figure 5 The pseudocode shown allows inputting echo signals, coefficient sets, and the maximum number of iterations into a hybrid constraint expansion network. When the hybrid constraint expands the subnetwork to a higher level t When the number of iterations is less than or equal to the maximum number of iterations, the image auxiliary variables, iterative image, and data-driven auxiliary variables are updated sequentially based on the coefficients of the previous level and the previous iterative image to obtain the current iterative image, and then the next iteration is performed. If the number of levels t is greater than the maximum number of iterations, the current last iterative image is directly output as the target image.

[0053] This invention generates a set of coefficients usable by a hybrid unfolding network using variable feature parameters and RD images of echoes. This allows the coefficients to adaptively adjust with echo characteristics, improving the dynamic adaptability and robustness to complex observation scenarios. By adding data-driven constraints to the optimization objective of the hybrid constraint unfolding network, its training and optimization process is no longer limited to sparsity, increasing the network's focus on the integrity of the target image. Thus, when generating target images using the trained hybrid constraint unfolding network, the imaging integrity and accuracy of the target image can be improved.

[0054] Example 4 Figure 6 The diagram shown is a structural schematic of an ISAR high-resolution robust imaging device based on a hybrid constraint unfolding network provided by an embodiment of the present invention. It is an example and not a limitation. The device may include a parameter estimation module and an iterative imaging module.

[0055] For example, the parameter estimation module is used to input the variable feature parameters of the echo signal and the RD image of the echo signal into the trained parameter generation network to obtain a coefficient set, wherein the coefficient set includes the values ​​of the coefficients used by each level of the hybrid constraint expansion sub-network; the iterative imaging module is used to input the echo signal into the trained hybrid constraint expansion network to obtain the target image; wherein the hybrid constraint expansion network includes multiple cascaded hybrid constraint expansion sub-networks, the current level hybrid constraint expansion network is used to generate the current level iterative image based on the previous level iterative image and the current level coefficients, and the last level iterative image is the target image; the optimization objective of the hybrid constraint expansion network includes data-driven constraints.

[0056] This invention generates a set of coefficients usable by a hybrid unfolding network using variable feature parameters and RD images of echoes. This allows the coefficients to adaptively adjust with echo characteristics, improving the dynamic adaptability and robustness to complex observation scenarios. By adding data-driven constraints to the optimization objective of the hybrid constraint unfolding network, its training and optimization process is no longer limited to sparsity, increasing the network's focus on the integrity of the target image. Thus, when generating target images using the trained hybrid constraint unfolding network, the imaging integrity and accuracy of the target image can be improved.

[0057] Example 5 Figure 7 The diagram shown is a structural schematic of an electronic device provided in an embodiment of the present invention. Figure 7 The illustrated electronic device 700 may include: at least one processor 710 ( Figure 7 The diagram shows only one processor, a memory 720, and a computer program 730 stored in the memory 720 and executable on the at least one processor 710, which, when executing the computer program 730, implements the steps in any of the above method embodiments.

[0058] The electronic device 700 can be a robot or other processing device capable of implementing the above methods. This embodiment of the invention does not impose any restrictions on the specific type of electronic device.

[0059] Those skilled in the art will understand that Figure 7 This is merely an example of electronic device 700 and does not constitute a limitation on the electronic device. It may include more or fewer components than shown, or combine certain components, or use different components. For example, the electronic device 700 may also include input / output interfaces.

[0060] The processor 710 may be a Central Processing Unit (CPU), or it may be other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASTCs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0061] In some embodiments, the memory 720 may be an internal storage unit, such as a hard disk or RAM. In other embodiments, the memory 720 may be an external storage device, such as a plug-in hard disk, a smart memory card (SMC), a secure digital card (SD), or a flash card. Furthermore, the memory 720 may include both internal and external storage units. The memory 720 is used to store the operating system, applications, a boot loader, data, and other programs, such as the program code of the computer program. The memory 720 can also be used to temporarily store data that has been output or will be output.

[0062] It should be understood that the sequence number of each step in the above embodiments does not imply 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 invention.

[0063] Those skilled in the art will understand that, for the sake of convenience and brevity, the above-described division of functional units and modules is merely an example. In practical applications, the functions described above can be assigned to different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiments can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit. Furthermore, the specific names of the functional units and modules are only for easy differentiation and are not intended to limit the scope of protection of this invention. The specific working process of the units and modules in the above system can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0064] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the steps described in the various method embodiments above.

[0065] This invention provides a computer program product that, when run on an electronic device, enables the electronic device to perform the steps described in the various method embodiments above.

[0066] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include at least: any entity or device capable of carrying computer program code to a photographing device / terminal device, a recording medium, a computer memory, a read-only memory (ROM), a random access memory (RAM), an electrical carrier signal, a telecommunication signal, and a software distribution medium. Examples include USB flash drives, portable hard drives, magnetic disks, or optical disks. In some jurisdictions, according to legislation and patent practice, computer-readable media cannot be electrical carrier signals or telecommunication signals.

[0067] In the above embodiments, the descriptions of each embodiment have different focuses. For parts that are not described in detail or recorded in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0068] Those skilled in the art will recognize that the units and algorithm steps of the various examples 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 implemented in hardware or software depends on the specific application and design constraints of the technical solution. Those skilled in the art can use different methods to implement the described functions for each specific application, but such implementations should not be considered beyond the scope of this invention.

[0069] To better illustrate the beneficial effects of the present invention, the following simulation experiments were conducted: Simulation Experiment 1 For example, Experiment 1 compared the imaging results of conventional techniques and the present invention under the same defect rate and different signal-to-noise ratios. Specifically, see Figures 8 and 9. Figure 8a ~c represents the ISAR imaging results of traditional techniques under signal-to-noise ratios of 0dB, 5dB, and 10dB, with a defect rate of 50%. Figure 9a ~c represents the ISAR imaging results of the method provided by this invention under signal-to-noise ratios of 0dB, 5dB, and 10dB, with a defect rate of 50%. Furthermore, referring to Table 1 below, it can be seen that the imaging effect of this invention is better when the signal-to-noise ratio changes.

[0070] Table 1 Evaluation Indicators of Imaging Results of Measured Data at 50% Missing Rate and Different Signal-to-Noise Ratios

[0071] Simulation Experiment 2 For example, Experiment 2 compared the imaging results of the conventional technique and the present invention under the same signal-to-noise ratio but different defect rates. Specifically, see Figures 10 and 11. Figure 10a ~c represents the ISAR imaging results of traditional techniques under the conditions of 25%, 50%, and 75% defect rates, and a signal-to-noise ratio of 5dB. Figure 11a ~c represents the ISAR imaging results of the method provided by this invention under the conditions of defect rates of 25%, 50%, and 75%, and a signal-to-noise ratio of 5dB. Furthermore, referring to Table 2 below, it can be seen that the imaging effect of this invention is still better when the defect rate changes.

[0072] Table 2 Evaluation Indicators of Imaging Results Based on Measured Data at Different Defect Rates with a 5dB Signal-to-Noise Ratio

[0073] Therefore, this invention generates a set of coefficients that can be used by the hybrid unfolding network by using variable feature parameters and RD images of echoes, so that the coefficients can be adaptively adjusted according to the echo features, improving the dynamic adaptability and robustness to complex observation scenarios; by adding data-driven constraints to the optimization objective of the hybrid constraint unfolding network, its training and optimization process is no longer limited to sparsity, and the network's attention to the integrity of the target image is improved; thus, when generating target images through the trained hybrid constraint unfolding network, the imaging integrity and accuracy of the target image can be improved.

Claims

1. A robust high-resolution ISAR imaging method based on a hybrid constraint unfolding network, characterized in that, include: The variable characteristic parameters of the echo signal and the RD image of the echo signal are input into the trained parameter generation network to obtain a set of coefficients, wherein the set of coefficients includes the values ​​of the coefficients used in each level of the hybrid constraint expansion sub-network. The echo signal is input into a trained hybrid constraint unpacking network to obtain the target image; The hybrid constraint expansion network includes multiple cascaded hybrid constraint expansion sub-networks. The current-level hybrid constraint expansion sub-network is used to generate the current-level iterative image based on the previous-level iterative image and the current-level coefficients. The last-level iterative image is the target image. The optimization objective of the hybrid constraint expansion network includes data-driven constraints.

2. The method according to claim 1, characterized in that, The hybrid constraint expansion subnetwork includes an image-assisted update module, an image update module, and a data-driven assisted update module; The image-assisted update module is used to update the auxiliary variables of the current level image based on the previous level iterated image, the current level coefficients, and the previous level data-driven auxiliary variables; The image update module is used to generate the current-level iterative image based on the current-level image auxiliary variables and the current-level coefficients; The data-driven auxiliary update module is used to update the data-driven auxiliary variables of the current level based on the current level iterative image.

3. The method according to claim 1, characterized in that, The variable feature parameters include the defect rate, and the coefficients include a first coefficient, a second coefficient, and a gradient descent step size.

4. The method according to claim 3, characterized in that, Image auxiliary variables are updated using the following formula: in, , The first t +1 level, No. t Level image auxiliary variables, , The first t Second coefficient of the gradient descent, step size of gradient descent For the first t Iterative images, This indicates the conjugate transpose. The echo signal, For the t-th level data-driven auxiliary variable, , These are dictionary matrices for the range and azimuth directions, respectively.

5. The method according to claim 4, characterized in that, The iterative image is updated using the following formula: in, For the first t +1 level iterative image, For the first t First coefficient of level, This is the soft threshold shrinkage function.

6. The method according to claim 5, characterized in that, The optimization objective of the hybrid constraint expansion network satisfies the following formula: in, This represents the optimization objective of the hybrid constraint expansion network. Indicates minimization. , These are the first and second coefficients, respectively. For data-driven constraints, This refers to the target image. As a data-driven auxiliary variable, It is an L1 norm. Describing the L2 norm, for The regularization coefficient.

7. The method according to claim 2, characterized in that, The data-driven auxiliary update module is U-Net, and the data-driven auxiliary variables are updated using the following formula: in, For the first t +1 level data-driven auxiliary variable, For the first t +1 level iterative image.

8. A high-resolution robust ISAR imaging device based on a hybrid constraint unfolding network, characterized in that, Includes a parameter estimation module and an iterative imaging module; The parameter estimation module is used to input the variable feature parameters of the echo signal and the RD image of the echo signal into the trained parameter generation network to obtain a set of coefficients, wherein the set of coefficients includes the values ​​of the coefficients used in each level of the hybrid constraint expansion sub-network. The iterative imaging module is used to input the echo signal into a trained hybrid constraint unpacking network to obtain the target image; The hybrid constraint expansion network includes multiple cascaded hybrid constraint expansion sub-networks. The current hybrid constraint expansion sub-network is used to generate the current iteration image based on the previous iteration image, and the last iteration image is the target image. The optimization objective of the hybrid constraint expansion network includes data-driven constraints.

9. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, When the processor executes the computer program, it implements the method as described in any one of claims 1-7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by an electronic device, it implements the method as described in any one of claims 1-7.

Images