Target image construction method and device and electronic equipment

Abstract

The invention relates to a target image construction method and apparatus, and an electronic device. The method comprises the steps of obtaining an echo signal obtained by detecting a target by a radar; inputting the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain a structural characteristic parameter of the target; wherein the target electromagnetic scattering parameter estimation model is obtained by unfolding and solving a specified iterative algorithm and an attribute scattering center model; the iterative algorithm is an alternating direction multiplier method TSPN-ADMM algorithm based on a truncated Schattert-p norm; the attribute scattering center model is used for representing the incidence relation between the echo signal and the structural characteristic parameter of the target; and constructing a target image according to the structural feature parameters. According to the embodiment of the invention, the target electromagnetic scattering parameter estimation model is determined based on the attribute scattering center model and the expanded TSPN-ADMM algorithm, and high-quality structural characteristic parameters can be obtained through inversion from the echo signal with the missing frequency spectrum, so that the quality of the constructed target image is improved.

Description

Technical Field

[0001] This application belongs to the field of data processing, and specifically relates to a method, apparatus and electronic device for constructing a target image. Background Technology

[0002] In the field of target recognition, extracting the structural feature parameters of a target is a crucial step in achieving accurate target identification and tracking. However, due to hardware limitations or environmental interference, the accurate extraction of the target's structural feature parameters often faces challenges such as difficult data processing, low estimation accuracy, and structural distortion affecting the estimation results.

[0003] In some related solutions, target images acquired via microwave methods (such as radar) are utilized to overcome these challenges. These images contain rich phase information, providing an additional source of information for extracting structural feature parameters such as the target's geometric and physical properties. However, in terms of estimating the target's structural feature parameters, the echo signals obtained by radar detection of the target suffer from two problems: missing spectral interference imaging and target feature parameter inversion. This results in low-quality target images constructed based on the target feature parameters. Summary of the Invention

[0004] The purpose of this application is to provide a target image construction method, apparatus, and electronic device to overcome or at least partially solve the above-mentioned problems.

[0005] To solve the above-mentioned technical problems, this application is implemented as follows: A method for constructing a target image, the method comprising: Acquire echo signals obtained from radar detection of targets; The echo signal is input into a preset target electromagnetic scattering parameter estimation model to obtain the structural characteristic parameters of the target; wherein, the target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model; the iterative algorithm is the alternating direction multiplier method TSPN-ADMM algorithm based on the truncated Schatter-p norm; the property scattering center model is used to characterize the correlation between the echo signal and the structural characteristic parameters of the target; The target image is constructed based on the structural feature parameters.

[0006] In the above embodiments, the target electromagnetic scattering parameter estimation model is determined based on the attribute scattering center model and the expanded TSPN-ADMM algorithm. The target electromagnetic scattering parameter estimation model can accurately analyze the sparse low-rank characteristics of echo signals in non-standard environments, and uniformly solve the two coupled problems of missing spectrum interference imaging and target structural feature parameter inversion. High-quality structural feature parameters can be obtained from the echo signal with missing spectrum through the target electromagnetic scattering parameter estimation model. The target image can be constructed based on the structural feature parameters, which can improve the quality of the constructed target image.

[0007] In one embodiment of this application, before inputting the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural characteristic parameters of the target, the method further includes: A target electromagnetic scattering parameter estimation model is constructed based on the attribute scattering center model and the TSPN-ADMM algorithm. The target electromagnetic scattering parameter estimation model is developed using an unrolling network; The expanded target electromagnetic scattering parameter estimation model is trained using a preset loss function and training samples to obtain the predicted complete echo signal output by the expanded target electromagnetic scattering parameter estimation model for the missing real echo signal; the training samples include at least the missing real echo signal and the complete real echo signal corresponding to the missing real echo signal. The predicted complete echo signal and the complete real echo signal are input into the loss function to obtain the loss value output by the loss function; When the loss value output by the loss function satisfies the preset convergence condition, the trained target electromagnetic scattering parameter estimation model is obtained.

[0008] In one embodiment of this application, the loss function is:

[0009] in, Indicates the loss value. Indicates training samples, This represents the predicted complete echo signal corresponding to the i-th missing true echo signal in the training samples. This represents the i-th complete true echo signal in the training samples. This represents the square of the Frobenius norm.

[0010] In one embodiment of this application, the target electromagnetic scattering parameter estimation model is as follows:

[0011] in, This indicates the prediction of the complete echo signal. This indicates the truncated p-norm. Describing the L1 norm, Represents the L2 norm. This indicates a missing echo signal. This represents a simulated echo signal, which is generated based on candidate structure feature parameters. This represents the transformation from the image domain to the pulse compression domain. This indicates echo sampling processing. Represents the imaging operator, Represents sparse constraint auxiliary variables. , is the regularization constraint coefficient.

[0012] In one embodiment of this application, the method further includes: The particle swarm optimization algorithm is used to determine the optimized structural feature parameters for generating the simulated echo signal based on the predicted complete echo signal and the candidate structural feature parameters. The difference between the simulated echo signal generated based on the optimized structural feature parameters and the predicted complete echo signal satisfies the preset optimization objective.

[0013] In the above embodiments, the particle swarm optimization algorithm is used to iteratively optimize the candidate structural feature parameters so that the difference between the simulated echo signal generated based on the optimized structural parameters and the complete echo signal predicted by the target electromagnetic scattering parameter estimation model can reach the preset optimization target, such as minimization. This provides the inversion accuracy and reliability of the target's structural feature parameters, thereby enabling the construction of a high-quality target image based on the optimized structural feature parameters.

[0014] In one embodiment of this application, the structural feature parameters include at least the range and azimuth coordinates of the scattering center in the relative coordinate system, the type parameter of geometric diffraction classification, the length of the distributed target, and the angle at which the distributed target deviates from the normal of the azimuth direction; after acquiring the echo signal obtained by radar detection of the target, the method further includes: The echo signal is preprocessed; wherein the preprocessing includes at least noise reduction, deskewing, and phase compensation.

[0015] In the above embodiments, before inputting the echo signal into the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET), the input echo signal needs to be preprocessed. This reduces interference factors in the echo signal input into the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) for structural feature parameter estimation, thereby improving the accuracy of the estimated structural feature parameters and thus improving the quality of the constructed target image.

[0016] A target image construction apparatus, the apparatus comprising: The echo signal acquisition module is used to acquire the echo signal obtained by radar detection of the target; The structural feature parameter determination module is used to input the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural feature parameters of the target; wherein, the target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model; the iterative algorithm is the TSPN-ADMM algorithm based on the truncated Schatter-p norm; the property scattering center model is used to characterize the correlation between the echo signal and the structural feature parameters of the target; The target image construction module is used to construct a target image based on the structural feature parameters.

[0017] An electronic device includes: a processor; and a memory for storing processor-executable instructions. The processor is configured to execute the instructions to implement the target image construction method described above.

[0018] A computer-readable storage medium, when the instructions in the storage medium are executed by the processor of a mobile terminal, enables the mobile terminal to perform the target image construction method described above.

[0019] The embodiments of this application have at least the following beneficial effects: In this embodiment, the echo signal obtained from radar detection of a target is acquired. The echo signal is then input into a preset target electromagnetic scattering parameter estimation model to obtain the target's structural feature parameters. The target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model. The iterative algorithm is the TSPN-ADMM algorithm based on the truncated Schatter-p norm alternating direction multiplier method. The property scattering center model is used to characterize the correlation between the echo signal and the target's structural feature parameters. A target image is constructed based on the structural feature parameters. This embodiment determines the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) based on the property scattering center model and the expanded TSPN-ADMM algorithm. This model can accurately analyze the sparse low-rank characteristics of echo signals under non-standard environments, uniformly solving the coupled problems of missing spectrum interference imaging and target structural feature parameter inversion. High-quality structural feature parameters can be obtained from the echo signal with missing spectrum through the target electromagnetic scattering parameter estimation model. Constructing a target image based on these structural feature parameters can improve the quality of the constructed target image. Attached Figure Description

[0020] Figure 1This is a flowchart illustrating the steps of a target image construction method provided in this application embodiment; Figure 2 This is a schematic diagram of an ISAR ultra-wideband echo signal provided in an embodiment of this application; Figure 3 This is a schematic diagram of a radar and a target provided in an embodiment of this application; Figure 4 This is a schematic diagram of a TSPN-ADMM Unrolling training process provided in an embodiment of this application; Figure 5 This is a schematic diagram of a target structure feature parameter inversion process based on TSPN-ADMM-NET provided in an embodiment of this application; Figure 6 This is a target simulation scene diagram of TSPN-ADMM-NET provided in the embodiments of this application; Figure 7 This is an inversion imaging map of the target structure feature parameters provided in the embodiments of this application; Figure 8 This is a schematic diagram illustrating the computational efficiency of iterative calculation and network model provided in the embodiments of this application; Figure 9 This is a schematic diagram of the structure of a target image construction device provided in the embodiments of this application; Figure 10 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0021] The embodiments of this application will be described below with reference to the accompanying drawings and preferred embodiments. Those skilled in the art can easily understand other advantages and effects of this application from the content disclosed in this specification. This application can also be implemented or applied through other different specific embodiments, and various details in this specification can also be modified or changed based on different viewpoints and applications without departing from the spirit of this application. It should be understood that the preferred embodiments are only for illustrating this application and are not intended to limit the scope of protection of this application.

[0022] In related technical solutions, in terms of estimating the structural characteristic parameters of targets, the structural characteristic parameter inversion method often encounters problems of insufficient accuracy and low computational efficiency when processing low-resolution echo signals obtained from radar-detected targets, which limits its application in real-time computing or processing.

[0023] With the rapid development of cloud computing and big data applications, the computing power provided by cloud computing infrastructure is being used more and more widely. In the cloud computing environment, especially when it comes to real-time image data processing, it can effectively provide high-burst computing services to meet the needs of real-time computing, thus providing a good basic computing environment for the real-time inversion of the structural feature parameters of the target.

[0024] In summary, by comprehensively utilizing phase information and multiple feature extraction techniques from target images, and based on cloud computing, the accuracy and computational efficiency of target recognition can be effectively improved, meeting the needs of real-time processing. These methods provide new perspectives and tools for estimating the structural feature parameters of targets, contributing to more reliable target recognition in complex environments. However, in terms of estimating the structural feature parameters of targets, the quality of target images constructed based on these parameters is low due to two problems: missing spectral interference imaging and target feature parameter (target structural feature parameter) inversion in the echo signals obtained by radar detection of targets.

[0025] To address the aforementioned issues, the purpose of this application is to propose a target structural feature parameter inversion method based on TSPN-ADMM-NET. Through a series of technical steps, it achieves sparse low-rank characteristic analysis of echo signals. Based on echo signal spectrum recovery and attribute scattering center model under discrete spectrum conditions, a target electromagnetic scattering parameter estimation model (i.e., TSPN-ADMM-NET) is established to improve the accuracy of target structural feature parameter estimation. Specifically, the core of this application lies in using TSPN-ADMM-NET (target electromagnetic scattering parameter estimation model) to invert the electromagnetic scattering feature parameters (i.e., structural feature parameters) of targets in non-standard environments based on the attribute scattering center model. An improved ADMM algorithm is used to uniformly solve the problems of missing spectrum interference imaging and target structural feature parameter inversion. The parameter calculation efficiency is improved by expanding the TSPN-ADMM iterative algorithm based on an unrolling network. Relying on the lightweight characteristics of the unrolling network model, it can be easily deployed in the cloud, facilitating efficient real-time processing.

[0026] Reference Figure 1 The flowchart illustrates the steps of a target image construction method provided in an embodiment of this application, specifically including the following steps: Step 101: Obtain the echo signal obtained by the radar detection target.

[0027] In the embodiments of this application, the echo signal obtained by the radar through microwave detection of the target can be acquired. In some specific embodiments, the radar can be MWP (Microwave Wide Photon Radar). Specifically, MWP is a new type of radar, different from traditional radar. Its signal source uses an all-optical module, which can achieve a wide bandwidth and high carrier frequency signal, thereby obtaining high-resolution detection capability. At the same time, the radar can use ISAR (Inverse Synthetic Aperture Radar) to acquire the target echo signal. Specifically, ISAR is a signal processing technology for imaging moving targets, which is of great significance in various applications. ISAR can perform high-resolution imaging of targets such as aircraft, ships, and satellites, which helps in target classification and identification. The echo signal obtained by inserting the target through MWP can also be called the MWP signal.

[0028] In some embodiments of this application, ultra-wideband (UWB) signal technology can be used to obtain echo signals, which can effectively extract the feature structure parameters of the target from the complex-valued image (echo signal), thereby significantly improving the accuracy of target recognition. UWB signal technology can capture more subtle signal changes, providing richer information for image analysis and thus enhancing the performance of target recognition.

[0029] Step 102: Input the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural characteristic parameters of the target; wherein, the target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model; the iterative algorithm is the alternating direction multiplier method TSPN-ADMM algorithm based on the truncated Schatter-p norm; the property scattering center model is used to characterize the correlation between the echo signal and the structural characteristic parameters of the target.

[0030] Step 103: Construct the target image based on the structural feature parameters.

[0031] TSPN stands for Truncated Schatter-P Norm, a norm about the rank of a matrix. It's widely used in signal processing, machine learning, and optimization problems, especially when dealing with low-rank matrices. TSPN quantifies the size or strength of a matrix, particularly when dealing with singular values. ADMM stands for Alternate Direction Multiplier Method, a computational framework for solving decomposable convex optimization problems. It's particularly suitable for large-scale distributed optimization problems and also excels in handling fast-paced, high-convergence statistical learning and machine learning problems. ADMM decomposes a large global problem into multiple smaller local subproblems through a decomposition and coordination process, and coordinates the solutions of these subproblems to obtain the solution to the global problem. ADMM can be seen as a combination of dual decomposition and augmented Lagrange multiplier method, which ensures good convergence while maintaining decomposability.

[0032] The attribute scattering center model is a parameterized electromagnetic scattering model based on geometric diffraction theory and physical optics theory. It characterizes the electromagnetic scattering response of a complex target as a synthesis of multiple independent scattering centers, each with a set of structural feature parameters. The attribute scattering center model can be used to establish the correlation between the echo signal and the target's structural feature parameters. Thus, the target's structural feature parameters can be obtained by inversion using the attribute scattering center model, and a target image can be constructed based on these parameters. For example, the structural feature parameters at least include the range and azimuth coordinates of the scattering center in a relative coordinate system. Type parameters of geometric diffraction classification Length of distributed targets The angle between the distributed target and the normal to the coordinate azimuth direction. wait.

[0033] In this embodiment, the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) is obtained by expanding and solving the specified TSPN-ADMM algorithm and attribute scattering center model. After acquiring the echo signal, the echo signal can be input into the target electromagnetic scattering parameter estimation model to obtain the target's structural feature parameters. The target electromagnetic scattering parameter estimation model can be used to perform low-rank sparse processing on the echo signal to recover the missing spectral information, thereby obtaining complete and accurate structural feature parameters. Then, a high-quality target image can be generated by constructing the structural feature parameters.

[0034] In the above embodiments, the target electromagnetic scattering parameter estimation model is determined based on the attribute scattering center model and the expanded TSPN-ADMM algorithm. The target electromagnetic scattering parameter estimation model can accurately analyze the sparse low-rank characteristics of echo signals in non-standard environments, and uniformly solve the two coupled problems of missing spectrum interference imaging and target structural feature parameter inversion. High-quality structural feature parameters can be obtained from the echo signal with missing spectrum through the target electromagnetic scattering parameter estimation model. The target image can be constructed based on the structural feature parameters, which can improve the quality of the constructed target image.

[0035] In one embodiment of this application, after acquiring the echo signal obtained from radar detection of the target, the method further includes: The echo signal is preprocessed; wherein the preprocessing includes at least noise reduction, deskewing, and phase compensation.

[0036] In this embodiment of the application, after obtaining the echo signal detected by radar, before inputting the echo signal into the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET), the input echo signal needs to be preprocessed, including noise reduction, deskewing and phase compensation processing, to facilitate subsequent data processing.

[0037] In the above embodiments, before inputting the echo signal into the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET), the input echo signal needs to be preprocessed. This reduces interference factors in the echo signal input into the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) for structural feature parameter estimation, thereby improving the accuracy of the estimated structural feature parameters and thus improving the quality of the constructed target image.

[0038] In one embodiment of this application, before inputting the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural characteristic parameters of the target, the method may further include: A target electromagnetic scattering parameter estimation model is constructed based on the attribute scattering center model and the TSPN-ADMM algorithm. The target electromagnetic scattering parameter estimation model is developed using an unrolling network; The expanded target electromagnetic scattering parameter estimation model is trained using a preset loss function and training samples to obtain the predicted complete echo signal output by the expanded target electromagnetic scattering parameter estimation model for the missing real echo signal; the training samples include at least the missing real echo signal and the complete real echo signal corresponding to the missing real echo signal. The predicted complete echo signal and the complete real echo signal are input into the loss function to obtain the loss value output by the loss function; When the loss value output by the loss function satisfies the preset convergence condition, the trained target electromagnetic scattering parameter estimation model is obtained.

[0039] In one embodiment of this application, the method may further include: The particle swarm optimization algorithm is used to determine the optimized structural feature parameters for generating the simulated echo signal based on the predicted complete echo signal and the candidate structural feature parameters. The difference between the simulated echo signal generated based on the optimized structural feature parameters and the predicted complete echo signal satisfies the preset optimization objective.

[0040] Unrolling, or algorithm unrolling, is a technique that transforms iterative algorithms into deep neural networks. It represents each iteration of the algorithm as a layer in the network, "unrolling" the entire iterative process into a deep network. Unrolling improves model interpretability, generalization ability, and performance in fields such as signal processing. In this process, forward propagation through the network is equivalent to performing a finite number of iterations of the iterative algorithm. Simultaneously, algorithm parameters (e.g., model parameters and regularization coefficients) are converted into network parameters, which can be optimized by training the network. Algorithm unrolling reduces reliance on large amounts of high-quality training samples, which is particularly important in data-scarce scenarios. In this embodiment, the lightweight unrolling of the TSPN-ADMM-NET algorithm enables rapid training and computation, achieving efficient estimation of the target's structural feature parameters. Furthermore, cloud deployment enhances the real-time performance of the overall echo signal processing flow. Moreover, the adaptive learning and adjustment of algorithm parameters through the unrolling network avoids the limitations of frequent parameter adjustments in traditional algorithms, making the algorithm more suitable for real-time processing. Furthermore, adaptive parameter adjustment based on the Unrolling network enhances robustness under different probing conditions, improving the stability and reliability of the TSPN-ADMM-NET algorithm. In addition, TSPN-ADMM-NET requires processing large amounts of data; combined with cloud computing technology, it can easily manage the model from training to deployment. The elastic computing resources of cloud computing make model training and deployment more efficient, facilitating rapid response to different computing needs, while also reducing hardware costs and maintenance complexity.

[0041] In the above embodiments, the particle swarm optimization algorithm is used to iteratively optimize the candidate structural feature parameters so that the difference between the simulated echo signal generated based on the optimized structural parameters and the complete echo signal predicted by the target electromagnetic scattering parameter estimation model can reach the preset optimization target, such as minimization. This provides the inversion accuracy and reliability of the target's structural feature parameters, thereby enabling the construction of a high-quality target image based on the optimized structural feature parameters.

[0042] The training process of the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) is described below. (Refer to...) Figure 2 This is a schematic diagram of an ISAR ultra-wideband (Full-band) echo signal provided in an embodiment of this application. Regarding the frequency and samples of the echo signal, it is assumed that the total number of sampling bands in the frequency domain is... The designation of a continuous sub-band is denoted as . The length of the i-th subband is Meanwhile, the starting frequency and center frequency of the i-th sub-band are respectively... and Because ultra-wideband imaging radar has high resolution and rich target detail information, it is necessary to consider the electromagnetic scattering characteristics of the target. Based on the attribute scattering center model, the multi-subband echo model (echo signal) with electromagnetic scattering characteristics at a certain azimuth time can be expressed by the following formula (1): (1) In the above formula, This represents the echo signal corresponding to the i-th sub-band, where j represents the imaginary unit. For the target response magnitude, Indicates the i-th target. Represents the target range in the frequency domain. This represents the distance-frequency domain sampling interval, which is a known constant. This represents the starting frequency of the i-th sub-band. This represents the center frequency of the i-th sub-band. This indicates the operation of the natural exponent. The azimuth influence factor between the target and the received echo signal is indicated by reference. Figure 3 The diagram shown is a schematic representation of a radar and a target (UAV) according to an embodiment of this application. and The reference angle and signal response angle of the target in the azimuth direction. This represents the length of the target's distribution in the reference coordinate system. There exists a scenario where this length is 0, which is related to the echo response angle. This represents the distance from the target to the radar receiver, where... This represents the sampling point of the target in a two-dimensional scene, in a domain orthogonal to the frequency domain. c is the speed of light constant. This indicates that the target echo signal noise is negligible in the MWP signal. For ISAR, where... , The initial distance between the target and the radar. The rotational rate of the target, where Approaching the high frequency band of MWP-ISAR Therefore, this term can be ignored. For formula (1), ignoring the constant phase term and compensating for migration through resolution units, we obtain: (2) Ignoring the constant term, formula (2) simplifies to: (3) in, This represents the echo signal corresponding to the i-th sub-band after compensation and migration using resolution units. This represents the coordinates of the scattering center in the range direction. This represents the coordinates of the scattering center in the azimuth direction.

[0043] Since the ISAR imaging processing steps do not affect the echo spectrum, the received multi-subband signal ASC (Attribute Scattering Center) echo matrix (echo signal) can be expressed as: (4) Where matrix X represents the imaging scene, which is a two-dimensional matrix. This represents the noise matrix. The relationship between the imaging result (target image) and the echo signal can be expressed as the imaging scene multiplied by two Fourier transform matrices, i.e. They are respectively Figure 3 The Fourier transform matrices in the x and y directions. Treating the inter-subband spectrum as a signal-free response, the range-oriented ultra-wideband echo signal can be represented as: (5) in, This indicates an echo signal with missing spectrum. for The frequency band data echo that needs to be recovered between the two sub-bands, then The length of the vector is equal to the total length of the frequency samples. .right any pulse in The Hankel constructive transformation (a method for converting a one-dimensional signal or vector into a specific structured matrix) is performed as shown in formula (6): (6) The matrix in formula (6) above satisfies the relation. .

[0044] To accurately invert the structural feature parameters of a target in a complex-valued image (echo signal), it is necessary to establish a structural feature parameter model of the target. This application embodiment constructs its structural feature parameters based on the ASC electromagnetic scattering parameter estimation model, referring to... Figure 3 The structural characteristic parameters can be expressed as ,in, Represents the range and azimuth coordinates of the i-th scattering center in the relative coordinate system; Type parameters representing geometric diffraction classification; Indicates the length of the distributed target. This represents the angle by which the distributed target deviates from the normal to the coordinate azimuth direction. Therefore, the ASC electromagnetic scattering parameter estimation model for ultra-wideband echo signals can be expressed as the following optimized model: (7) in, This represents the complete echo signal. This represents the ultra-wideband echo signal generated based on the structural characteristic parameters derived from the target inversion. This is to accurately estimate the set of structural characteristic parameters. Establishing a reasonable metric for the difference between the simulated echo signal and the original echo signal is crucial. Let's assume we use the estimated set of structural characteristic parameters... The generated echo signal is Meanwhile, the approximate echo (predicting the complete echo signal) obtained from equation (1) is expected. The complete echo signal was recovered. Based on Euclidean distance, the metric expression for the echo signal or image matrix can be written as: (8) in, This represents the value of the metric corresponding to the difference. This indicates the prediction of the complete echo signal. Indicates analog echo signal, Denotes the square of the Frobenius norm. This represents the transformation from the image domain to the pulse compression domain.

[0045] In summary, based on multi-subband ISAR ultra-wideband echo signals, the problem of inverting the structural feature parameters of targets in complex images can be modeled as follows (9): that is, the target electromagnetic scattering parameter estimation model constructed based on the attribute scattering center model and the TSPN-ADMM algorithm can be: (9) in, This refers to the received signal data matrix (i.e., the missing echo signal obtained from the initial radar detection), where... This indicates the prediction of the complete echo signal. This indicates the truncated p-norm. Describing the L1 norm, Represents the L2 norm. This represents the simulated echo signal, which is generated based on the candidate structure characteristic parameters. This represents the transformation from the image domain to the pulse compression domain. This indicates echo sampling processing. Represents the imaging operator, Represents sparse constraint auxiliary variables. , Let represent the regularization constraint coefficient. For the convex approximation of the above model, we can obtain: (10) in, Represents the transformation from the image domain to the... The domain after echo preprocessing is generally the pulse compression domain. This indicates echo sampling processing. The imaging operator is represented by D, which is generally equivalent to a Fourier transform in the preprocessed data. D is an auxiliary variable. To solve the above model, this application proposes a TSPN-ADMM solution method based on the alternating direction multiplier method. The specific calculation process is as follows.

[0046] First, the augmented Lagrangian function expression for the above property scattering center model is given as follows: (11) In the above formula, Z and Y represent Lagrange multipliers. Indicates the penalty parameter. This represents the regularization constraint coefficient. For The problem can be represented as: (12) By further combining the quadratic terms, we can obtain: (13) The above formula can be solved using the GST (Generalized Soft Threshold) algorithm. This is achieved using known parameters. and The threshold can be obtained as follows: (14) The corresponding generalized threshold shrinkage equation can be written as: (15) In the above formula, Let i be the i-th singular value after singular value decomposition. Let be the i-th singular value of the low-rank matrix to be found. Therefore, The specific solution process is as follows: enter:

[0047] calculate :

[0048] calculate:

[0049] In the above process, This represents the matrix conjugate transpose operation. For In contrast to the generalized soft thresholding algorithm, the corresponding analytical solution is: (16) for Its iterative analytical solution can be written as: (17) Based on equation (8), in Dis Let from The prior parameter set (prior information) of the target space information obtained from the echo is The set of structural characteristic parameters of the ASC model echo that needs to be solved is: The ASC model echo can be represented as... .based on This prior information is used to estimate the parameters of the ASC model using the PSO (Particle Swarm Optimization) algorithm, and its expression is as follows: (18) However, solving this objective function presents significant challenges because, if the target scenario has P targets, the size of the parameter matrix to be estimated is... Estimating so many parameters at once would significantly reduce the robustness of the algorithm. Therefore, in this embodiment, the estimation strategy is broken down into estimating each scattering center individually, thus transforming a high-dimensional problem into multiple low-dimensional problems.

[0050] The problem of solving for the structural characteristic parameters of the p-th scattering center can be viewed as a five-dimensional single-objective optimization problem. The PSO algorithm based on dynamic weights, in a single iteration... The update iterative formula can be written as: (19)

[0051]

[0052] in, The velocity factor representing the structural characteristic parameter of the p-th scattering center in the t-th iteration. It is the inertia factor of the iterative algorithm. and Represents the particle learning factor, typically taking values ​​in the range of... between. and Let represent the individual optimal solution (individual optimal structural characteristic parameters) and the global optimal solution (global optimal structural characteristic parameters) for the p-th particle in the current iteration, respectively. Inertia factor. The larger the particle size, the stronger its global search capability and the weaker its local search capability.

[0053] By extending the structural feature parameter estimation strategy for the P-scattering center during the single-objective parameter optimization process, the center of the entire scene can be obtained: Step 1: Utilize cosine similarity to solve for the prior range of parameter optimization, reducing the optimization space. Set the maximum number of iterations to T, the error threshold e, and the number of scattering centers P, while letting p=1 and t=1.

[0054] Step 2: Generate intermediate parameters, that is, generate the corresponding echo based on the parameter particle set:

[0055] Step 3: Generate the set of structural feature parameters for the p-th scattering center. And generate the corresponding echo. ; Step 4: According to the formula The population fitness of the p-th scattering center in the t-th iteration is calculated iteratively. After T iterations, the optimal parameter set of the p-th scattering center is obtained. Step 5: Subtract the currently estimated scattering center echo from the filled spectrum transform echo. , get ; Step 6: Determine if p equals P. If not, p = p + 1, go to Step 2; if yes, output the estimated set of optimal parameters (optimal structural characteristic parameters) for the P scattering centers. For Lagrange multipliers Y and Z, we have: (20) In the derivation of the above sub-problem models, a low-rank solver based on the truncated p-norm is used for the spectrum recovery problem. To better utilize the high-resolution characteristics of MWB-ISAR, a sparse regularization term in the image domain is introduced to improve the model's solution performance. For the problem of estimating the electromagnetic scattering features of the scattering center, a multi-parameter optimization solution based on the particle swarm optimization algorithm is used. Considering the coupling between multiple sub-problems and the inconsistent impact on the final target, a multi-problem balancing factor is introduced to address the issue. Finally, based on the ADMM multi-task algorithm, a target structure feature parameter inversion algorithm is proposed, the process of which is summarized as follows: Input: Sampled multi-subband echo (Echo signal obtained from initial radar detection), maximum number of ADMM iterations T, intermediate parameter set of the algorithm .

[0056] initialization: Generate using the initial prior parameter set obtained from cosine similarity .

[0057] Step 1: Update based on the derived generalized soft threshold algorithm .

[0058] Step 2: Update , .

[0059] Step 3: Update , .

[0060] Step 4: Calculation and judge Whether it is satisfied or not, if it is not satisfied, then... ; Step 5: When Update according to the dynamic weight PSO algorithm flow designed in this chapter. This yields the optimal estimate of the scattering characteristic parameters for the current state; if , echo corresponding to the prior parameter set .

[0061] Step 6: Update , .

[0062] Step 7: Update , .

[0063] Step 8: Update Determine if the number of iterations has reached the maximum; if not, proceed to Step 1. Step 9: Output and parameter set .

[0064] Based on the TSPN-ADMM method described above, and to reduce the need for algorithm parameter tuning, referencing the Unrolling network, we assume the training samples are... The parameters that need to be learned can be expressed as The loss function of the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) expanded by the iterative algorithm can be expressed as follows: (twenty one) in, Indicates the loss value. Indicates training samples, This represents the predicted complete echo signal corresponding to the i-th missing true echo signal in the training samples. This represents the i-th complete true echo signal in the training samples. This represents the square of the Frobenius norm.

[0065] Finally, based on the output predicted echo signal and the actual echo signal and structural feature parameters This allows us to obtain the target super-resolution reconstructed echo and structural characteristic parameters.

[0066] Reference Figure 4 This document presents a schematic diagram of a TSPN-ADMM Unrolling training process provided in an embodiment of this application. First, the relationship between the Unrolling network and the ADMM algorithm is explained: The ADMM algorithm is the "logical prototype" of the Unrolling network. The ADMM algorithm solves optimization problems through fixed iterative steps, such as variable updates and multiplier updates. The Unrolling network directly maps each iteration to a layer of the network, maintaining mathematical consistency in the iterations. It naturally possesses the network advantages of iterative problem-solving algorithms. Unrolling makes ADMM parameters "learnable": In traditional ADMM, thresholds, penalty coefficients, etc., are manually set fixed values. The Unrolling network transforms these parameters into learnable parameters for data-driven optimization. The network structure adapts to the iterative characteristics of ADMM: The number of layers in the Unrolling network corresponds to the number of ADMM iterations. The computational modules of each layer (such as transformations and threshold operations) perfectly match the iterative steps of ADMM, eliminating the need for additional network topology design.

[0067] Next, the training process of the ADMM algorithm's unrolling network will be explained. The specific training process is as follows: 1. Initialization and raw data input: The input is the sampled multi-subband echo. , Initialize process parameters algorithm initialization steps. Among them, the initial parameters related to the target characteristics are obtained using formula (18) and used as regression reference. 2. Iterative layer training (core: Layer 1~Layer K): Each layer (Layer k) corresponds to one ADMM iteration (Step 1~Step 8), the process is as follows: First, solve based on the wide-area soft threshold and initial values ​​(formulas 12~16). Then, the iteration of the current layer is obtained based on equations (17) and (18). Finally, the calculation is based on equation (20). .

[0068] The above process represents the calculation of each variable in a single-layer network structure. After the single-layer network model is calculated, the learnable parameters of the layer are recorded for calculating the single-layer loss function. For details, please refer to formula (21) and the gradient of the entire network layer. Among them, the number of layers K is a variable set by the user and can be set based on the convergence graph of the iterative algorithm. After all K layers are solved, it is considered that the training of the single-instance unrolling network is completed. The number of training rounds is automatically determined based on the loss of the loss function to determine whether to continue training. After the training is completed, the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) can be obtained.

[0069] After obtaining the trained target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET), the structural characteristic parameters of the target can be estimated using the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET).

[0070] Reference Figure 5 This is a schematic diagram of a target structure feature parameter inversion process based on TSPN-ADMM-NET provided in an embodiment of this application. The specific inversion process based on the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) is as follows: 1. Data preprocessing: First, the input echo signal is preprocessed, including noise reduction, deskewing, and phase compensation, to facilitate subsequent data processing.

[0071] 2. Low-rank sparsity processing: TSPN-ADMM-NET is used to perform low-rank sparsity processing on the preprocessed echo signal to recover missing information. This step fully utilizes the low-rank and sparsity characteristics of the signal and improves the efficiency and accuracy of spectrum recovery through algorithm optimization, while ensuring the reliability of phase recovery.

[0072] 3. Target Feature Parameter Estimation: Based on the spectral information recovery, the structural feature parameters of the target, including location and structure type, are estimated using the attribute scattering center model. This step further improves the accuracy of the target's structural feature parameter estimation through parameter optimization.

[0073] 4. Adaptive Parameter Adjustment: An unrolling network is used for adaptive parameter optimization, automatically adjusting the algorithm parameters during the inversion process to adapt to different image characteristics and target types. This step improves the robustness of TSPN-ADMM-NET under different conditions for estimating different target feature parameters. Different conditions can include target projection at different angles, scenes with different signal-to-noise ratios, etc.

[0074] 5. Output results: Output the structural feature parameters of the target, and the target image reconstructed based on the structural feature parameters of the target.

[0075] For example, refer to Figure 6 This is a target simulation scene diagram of TSPN-ADMM-NET provided in the embodiments of this application. Figure 6 Element 1 to 5 represent different types of target distributions. The structural feature parameters of the five targets obtained through the inversion process are shown in Table 1 below. The scene detection parameters required to obtain the target structural feature parameters in the inversion process are shown in Table 2 below. Table 1:

[0076] Table 2:

[0077] Reference Figure 7 The above are target structure feature parameter inversion imaging images provided in the embodiments of this application, wherein (a) is an imaging image (target image) constructed by target structure feature parameters inverted by conventional methods; (b) is an imaging image constructed by target structure feature parameters inverted by reference method (PSO-OMP); and (c) is an imaging image constructed by target structure feature parameters inverted by TSPN-ADMM-NET of this application. It can be seen from the comparison that the image quality of the inversion imaging image by TSPN-ADMM-NET of this application is higher.

[0078] Referring to Table 3, the target structure characteristic parameters are obtained by inversion using the conventional method, the reference method, and the TSPN-ADMM-NETs method of this application: Table 3:

[0079] pass Figure 6As can be seen from Table 3, this application is more accurate than traditional methods and reference methods in retrieving target structural feature parameters. Figure 8 This is a schematic diagram illustrating the computational efficiency of iterative calculation and network model calculation provided in an embodiment of this application. Figure 8 It can be seen that the TSPN-ADMM-NET, which utilizes the Unrolling network, is more efficient in computation than the traditional TSPN-ADMM algorithm.

[0080] In summary, this application discloses a target structural feature parameter inversion method based on TSPN-ADMM-NET. Leveraging the advantages of cloud computing resource scheduling and the lightweight nature of the unrolling network model, it can be easily deployed in the cloud, aiming to provide an efficient, high-precision, and real-time strategy for complex-valued image target feature inversion. The main objectives of this method are as follows: 1. Improve the accuracy of target structural feature parameter inversion: By utilizing the low-rank and sparse characteristics of the signal, based on the attribute scattering center model, and combined with the improved TSPN-ADMM-NET algorithm, the estimation accuracy of target structural feature parameters is improved, expanding the dimension of target structural feature parameter inversion. 2. Optimize the efficiency of target image parameter inversion: Based on deep learning technology, the TSPN-ADMM-NET algorithm is expanded into a multi-layer network model using an unrolling network, and adaptive training is used to improve the efficiency of target image parameter inversion. 3. Optimize the real-time processing flow: Relying on the lightweight expansion of traditional iterative algorithms using the unrolling network, rapid training and computation are achieved, meeting the requirements for efficient estimation of target feature parameters. Deployment in the cloud further improves the real-time performance of the overall processing flow. 4. Suitable for efficient cloud computing deployment: Because the algorithm model is a lightweight network with fewer dependent parameters, it is easy to migrate and deploy in the cloud.

[0081] It should be noted that, for the sake of simplicity, the method embodiments are all described as a series of actions. However, those skilled in the art should understand that the embodiments of this application are not limited to the described order of actions, because according to the embodiments of this application, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also understand that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily necessary for the embodiments of this application.

[0082] Reference Figure 9 The diagram shows a structural block diagram of a target image construction device provided in an embodiment of this application. The device may specifically include the following modules: The echo signal acquisition module 901 is used to acquire the echo signal obtained by radar detection of the target; The structural feature parameter determination module 902 is used to input the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural feature parameters of the target; wherein, the target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model; the iterative algorithm is the TSPN-ADMM algorithm based on the truncated Schatter-p norm; the property scattering center model is used to characterize the correlation between the echo signal and the structural feature parameters of the target; The target image construction module 903 is used to construct a target image based on the structural feature parameters.

[0083] In one embodiment of this application, the apparatus further includes: a model training module, used for: A target electromagnetic scattering parameter estimation model is constructed based on the attribute scattering center model and the TSPN-ADMM algorithm. The target electromagnetic scattering parameter estimation model is developed using an unrolling network; The expanded target electromagnetic scattering parameter estimation model is trained using a preset loss function and training samples to obtain the predicted complete echo signal output by the expanded target electromagnetic scattering parameter estimation model for the missing real echo signal; the training samples include at least the missing real echo signal and the complete real echo signal corresponding to the missing real echo signal. The predicted complete echo signal and the complete real echo signal are input into the loss function to obtain the loss value output by the loss function; When the loss value output by the loss function satisfies the preset convergence condition, the trained target electromagnetic scattering parameter estimation model is obtained.

[0084] In one embodiment of this application, the loss function is:

[0085] in, Indicates the loss value. Indicates training samples, This represents the predicted complete echo signal corresponding to the i-th missing true echo signal in the training samples. This represents the i-th complete true echo signal in the training samples. This represents the square of the Frobenius norm.

[0086] In one embodiment of this application, the target electromagnetic scattering parameter estimation model is as follows:

[0087] in, This indicates the prediction of the complete echo signal. This indicates the truncated p-norm. Describing the L1 norm, Represents the L2 norm. This indicates a missing echo signal. This represents a simulated echo signal, which is generated based on candidate structure feature parameters. This represents the transformation from the image domain to the pulse compression domain. This indicates echo sampling processing. Represents the imaging operator, Represents sparse constraint auxiliary variables. , is the regularization constraint coefficient.

[0088] In one embodiment of this application, the apparatus further includes: an optimization module, configured to: The particle swarm optimization algorithm is used to determine the optimized structural feature parameters for generating the simulated echo signal based on the predicted complete echo signal and the candidate structural feature parameters. The difference between the simulated echo signal generated based on the optimized structural feature parameters and the predicted complete echo signal satisfies the preset optimization objective.

[0089] In one embodiment of this application, the structural feature parameters include at least the range and azimuth coordinates of the scattering center in the relative coordinate system, the type parameter of geometric diffraction classification, the length of the distributed target, and the angle at which the distributed target deviates from the normal of the coordinate azimuth direction.

[0090] In one embodiment of this application, the apparatus further includes: a preprocessing module, configured to: The echo signal is preprocessed; wherein the preprocessing includes at least noise reduction, deskewing, and phase compensation.

[0091] In this embodiment, the echo signal obtained from radar detection of a target is acquired. The echo signal is then input into a preset target electromagnetic scattering parameter estimation model to obtain the target's structural feature parameters. The target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model. The iterative algorithm is the TSPN-ADMM algorithm based on the truncated Schatter-p norm alternating direction multiplier method. The property scattering center model is used to characterize the correlation between the echo signal and the target's structural feature parameters. A target image is constructed based on the structural feature parameters. This embodiment determines the target electromagnetic scattering parameter estimation model (TSPN-ADMM-NET) based on the property scattering center model and the expanded TSPN-ADMM algorithm. This model can accurately analyze the sparse low-rank characteristics of echo signals under non-standard environments, uniformly solving the coupled problems of missing spectrum interference imaging and target structural feature parameter inversion. High-quality structural feature parameters can be obtained from the echo signal with missing spectrum through the target electromagnetic scattering parameter estimation model. Constructing a target image based on these structural feature parameters can improve the quality of the constructed target image.

[0092] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment. This application also provides an electronic device, such as... Figure 10 As shown, it includes a processor 1001, a device interface 1002, a memory 1003, and a bus 1004; Memory 1003 is used to store computer programs; The processor 1001 executes the above steps when executing the program stored in the memory 1003.

[0093] The bus mentioned in the above terminal can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into address bus, data bus, control bus, etc. For ease of illustration, only one thick line is used to represent it in the diagram, but this does not mean that there is only one bus or one type of bus.

[0094] The memory may include random access memory (RAM) or non-volatile memory, such as at least one disk storage device. Optionally, the memory may also be at least one storage device located remotely from the aforementioned processor.

[0095] The processors mentioned above can be general-purpose processors, including central processing units (CPUs), network processors (NPs), etc.; they can also be digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.

[0096] This application also provides a storage medium that, when the instructions in the storage medium are executed by the processor of an electronic device, enables the electronic device to perform the target image construction method of the foregoing embodiments.

[0097] As the device embodiment is basically similar to the method embodiment, the description is relatively simple, and relevant parts can be found in the description of the method embodiment.

[0098] The algorithms and displays provided herein are not inherently related to any particular computer, virtual device, or other equipment. The structure required to construct such a device is obvious from the above description. Furthermore, this application is not directed to any particular programming language. It should be understood that the content of this application described herein can be implemented using various programming languages, and the above description of specific languages ​​is for the purpose of disclosing the best mode of implementation of this application.

[0099] Numerous specific details are set forth in the specification provided herein. However, it will be understood that embodiments of this application may be practiced without these specific details. In some instances, well-known methods, structures, and techniques have not been shown in detail so as not to obscure the understanding of this specification.

[0100] Similarly, it should be understood that, in order to simplify this application and aid in understanding one or more of the various inventive aspects, in the above description of exemplary embodiments of this application, various features of this application are sometimes grouped together into a single embodiment, figure, or description thereof. However, this method of disclosure should not be construed as reflecting an intention that the claimed application requires more features than are expressly recited in each claim. Rather, as reflected in the following claims, inventive aspects lie in fewer than all features of a single foregoing disclosed embodiment. Therefore, the claims following the detailed description are hereby expressly incorporated into that detailed description, wherein each claim itself is a separate embodiment of this application.

[0101] Those skilled in the art will understand that modules in the device of the embodiments can be adaptively changed and placed in one or more devices different from that embodiment. Modules, units, or components in the embodiments can be combined into a single module, unit, or component, and further, they can be divided into multiple sub-modules, sub-units, or sub-components. Except where at least some of such features and / or processes or units are mutually exclusive, any combination can be used to combine all features disclosed in this specification (including the accompanying claims, abstract, and drawings) and all processes or units of any method or device so disclosed. Unless expressly stated otherwise, each feature disclosed in this specification (including the accompanying claims, abstract, and drawings) may be replaced by an alternative feature that serves the same, equivalent, or similar purpose.

[0102] The various component embodiments of this application can be implemented in hardware, or as software modules running on one or more processors, or a combination thereof. Those skilled in the art will understand that microprocessors or digital signal processors (DSPs) can be used in practice to implement some or all of the functions of some or all of the components in the sequencing device according to this application. This application can also be implemented as a device or apparatus program for performing part or all of the methods described herein. Such an implementation of this application can be stored on a computer-readable medium, or can take the form of one or more signals. Such signals can be downloaded from an Internet website, provided on a carrier signal, or provided in any other form.

[0103] It should be noted that the above embodiments are illustrative of this application and not restrictive, and that those skilled in the art can devise alternative embodiments without departing from the scope of the appended claims. In the claims, any reference signs placed between parentheses should not be construed as limiting the claims. The word "comprising" does not exclude the presence of elements or steps not listed in the claims. The word "a" or "an" preceding an element does not exclude the presence of a plurality of such elements. This application can be implemented by means of hardware comprising several different elements and by means of a suitably programmed computer. In the unit claims enumerating several means, several of these means may be embodied by the same item of hardware. The use of the words first, second, and third, etc., does not indicate any order. These words can be interpreted as names.

[0104] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the devices, apparatuses, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0105] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

[0106] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

[0107] It should be noted that the various data-related processes in the embodiments of this application are carried out in compliance with the relevant data protection laws and policies of the country where the location is located, and with the authorization granted by the owner of the corresponding device.

Claims

1. A method for constructing a target image, characterized in that, The method includes: Acquire echo signals obtained from radar detection of targets; The echo signal is input into a preset target electromagnetic scattering parameter estimation model to obtain the structural characteristic parameters of the target; wherein, the target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model; the iterative algorithm is the alternating direction multiplier method TSPN-ADMM algorithm based on the truncated Schatter-p norm; the property scattering center model is used to characterize the correlation between the echo signal and the structural characteristic parameters of the target; The target image is constructed based on the structural feature parameters.

2. The method according to claim 1, characterized in that, Before inputting the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural characteristic parameters of the target, the method further includes: A target electromagnetic scattering parameter estimation model is constructed based on the attribute scattering center model and the TSPN-ADMM algorithm. The target electromagnetic scattering parameter estimation model is developed using an unrolling network; The expanded target electromagnetic scattering parameter estimation model is trained using a preset loss function and training samples to obtain the predicted complete echo signal output by the expanded target electromagnetic scattering parameter estimation model for the missing real echo signal; the training samples include at least the missing real echo signal and the complete real echo signal corresponding to the missing real echo signal. The predicted complete echo signal and the complete real echo signal are input into the loss function to obtain the loss value output by the loss function; When the loss value output by the loss function satisfies the preset convergence condition, the trained target electromagnetic scattering parameter estimation model is obtained.

3. The method according to claim 2, characterized in that, The loss function is: in, Indicates the loss value. Indicates training samples, This represents the predicted complete echo signal corresponding to the i-th missing true echo signal in the training samples. This represents the i-th complete true echo signal in the training samples. This represents the square of the Frobenius norm.

4. The method according to claim 3, characterized in that, The target electromagnetic scattering parameter estimation model is as follows: in, This indicates the prediction of the complete echo signal. This indicates the truncated p-norm. Describing the L1 norm, Describing the L2 norm, This indicates a missing echo signal. This represents a simulated echo signal, which is generated based on candidate structure feature parameters. This represents the transformation from the image domain to the pulse compression domain. This indicates echo sampling processing. Represents the imaging operator, Represents sparse constraint auxiliary variables. , is the regularization constraint coefficient.

5. The method according to claim 4, characterized in that, The method further includes: The particle swarm optimization algorithm is used to determine the optimized structural feature parameters for generating the simulated echo signal based on the predicted complete echo signal and the candidate structural feature parameters. The difference between the simulated echo signal generated based on the optimized structural feature parameters and the predicted complete echo signal satisfies the preset optimization objective.

6. The method according to claim 1, characterized in that, The structural feature parameters include at least the range and azimuth coordinates of the scattering center in the relative coordinate system, the type parameter of geometric diffraction classification, the length of the distributed target, and the angle at which the distributed target deviates from the normal of the coordinate azimuth.

7. The method according to claim 1, characterized in that, After acquiring the echo signal obtained from radar detection of the target, the method further includes: The echo signal is preprocessed; wherein the preprocessing includes at least noise reduction, deskewing, and phase compensation.

8. A target image construction apparatus, characterized in that, The device includes: The echo signal acquisition module is used to acquire the echo signal obtained by radar detection of the target; The structural feature parameter determination module is used to input the echo signal into a preset target electromagnetic scattering parameter estimation model to obtain the structural feature parameters of the target; wherein, the target electromagnetic scattering parameter estimation model is obtained by expanding and solving a specified iterative algorithm and a property scattering center model; the iterative algorithm is the TSPN-ADMM algorithm based on the truncated Schatter-p norm; the property scattering center model is used to characterize the correlation between the echo signal and the structural feature parameters of the target; The target image construction module is used to construct a target image based on the structural feature parameters.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to execute the instructions to implement the target image construction method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, When the instructions in the storage medium are executed by the processor of the mobile terminal, the mobile terminal is able to perform the target image construction method as described in any one of claims 1 to 7.

Images