Joint optimization SAR imaging method combining learnable emission mode and imaging algorithm

Abstract

The invention provides a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm, and relates to the field of SAR imaging, and the method comprises the steps: taking a random part in a non-uniform pulse emission moment as a learnable variable, carrying out the simulation to obtain a non-uniform echo signal in a current emission mode, and carrying out the calculation of the learnable emission mode; and obtaining an SAR image through a deep expansion SAR imaging algorithm based on approximate observation and neural network sparse domain mapping, taking the SAR image as an optimization target, and optimizing parameters in the pulse emission mode and the imaging algorithm at the same time to obtain a non-uniform pulse emission mode sequence with optimal anti-aliasing performance and a matched SAR imaging algorithm. According to the method, the learnable pulse emission mode and the SAR imaging algorithm are combined, and the final SAR imaging result is taken as an optimization target, so that the imaging algorithm fully utilizes the advantages of the non-uniform emission mode, and the imaging quality is improved as much as possible while the breadth is expanded.

Description

Technical Field

[0001] The present application relates to the field of synthetic aperture radar imaging, and in particular to a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm. Background Art

[0002] Most existing synthetic aperture radar (SAR) imaging algorithms based on compressed sensing use random sampling to obtain downsampled echo signals, which only reduces the amount of data but does not change the minimum interval between adjacent pulses, thus failing to expand the mapping bandwidth. SAR imaging algorithms that achieve wide mapping bands often decouple the emission mode and the imaging algorithm and design them separately, resulting in the imaging algorithm being unable to fully utilize the anti-aliasing advantage of the non-uniform emission mode, thus affecting the imaging quality. Summary of the invention

[0003] The present application aims to solve one of the technical problems in the related art at least to some extent.

[0004] To this end, the first purpose of the present application is to propose a joint optimization SAR imaging method that combines a learnable emission pattern and an imaging algorithm, combining a learnable pulse emission pattern and a SAR imaging algorithm, taking the final SAR imaging result as the optimization target, so that the imaging algorithm can make full use of the advantages of the non-uniform emission pattern and improve the imaging quality as much as possible while achieving expanded bandwidth.

[0005] The second objective of the present application is to provide an electronic device.

[0006] The third object of the present application is to provide a computer-readable storage medium.

[0007] A fourth object of the present application is to provide a computer program product.

[0008] To achieve the above-mentioned purpose, the first embodiment of the present application proposes a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm, including:

[0009] S1, generating a non-uniform pseudo echo signal according to SAR system parameters and simulation scenarios as an input signal of an imaging iteration process, and initializing an imaging result matrix to a zero matrix as an initial state of the imaging iteration process;

[0010] S2. Based on the current imaging result, an approximate echo is first generated by an inverse imaging operator, the residual between the approximate echo and the simulated echo is calculated, and then the imaging operator is used to obtain the imaging result of the residual, and the imaging result of the residual is used to linearly update the imaging result of the previous round;

[0011] S3, mapping the updated linear imaging result to the sparse domain, and using the soft threshold iterative function to filter out the noise to obtain the nonlinear imaging result;

[0012] S4, repeating the linear and nonlinear updating steps until the set number of iterations is reached to obtain the final imaging result;

[0013] S5. After the iteration, the loss function and gradient are calculated, and a random parameter freezing strategy is used for the pulse emission time, and a part of the pulse emission time is randomly selected to freeze the gradient of the parameters;

[0014] S6, back-propagating and updating the pulse emission mode and imaging algorithm parameters according to the gradient updated by the random parameter freezing strategy;

[0015] S7. Repeat steps S2-S6 until convergence, and obtain the final optimized pulse emission mode and matching imaging algorithm.

[0016] Optionally, the generating of a non-uniform pseudo echo signal according to the SAR system parameters and the simulation scene as an input signal of the imaging iteration process, and initializing the imaging result matrix to a zero matrix as an initial state of the imaging iteration process, includes:

[0017] According to the SAR system parameter Θ and the simulation scene X, a simulated azimuth non-uniform echo signal Y is generated as the input signal of the imaging iteration process;

[0018] Initialize imaging result X 0 is a zero matrix, which serves as the initial state of the imaging iteration process.

[0019] Optionally, the calculating, based on the current imaging result, a residual between the approximate echo and the simulated echo, and using the residual to linearly update the imaging result includes:

[0020] Based on the previous round of imaging results X k-1 , according to the approximate observation inverse imaging operator, the approximate echo matrix G(X k-1 ), where G() represents the inverse imaging operator;

[0021] Calculate the approximate echo matrix G(X k-1 ) and the residual of the simulated echo Y, and obtain the approximate imaging result H(YG(X) of the residual according to the imaging operator k-1 )), where H() represents the imaging operator;

[0022] The residual coefficients and the approximate imaging results H(YG(X k-1 )) The product is added to the imaging result X of the previous round k-1 On the above, we get the updated linear imaging result, which is expressed as:

[0023] Z k =X k-1 +α k ·H(YG(X k-1 ))

[0024] Among them, Z k is the linear imaging result after the kth iteration update.

[0025] Optionally, mapping the updated linear imaging result to a sparse domain and filtering out noise using a soft threshold iterative function to obtain a nonlinear imaging result includes:

[0026] Separate the linear imaging result Z as a complex matrix k The real and imaginary parts are converted into real matrix D(Z k );

[0027] Through three layers of convolution Conv, the real matrix D(Z k ) is mapped to the sparse domain, and the noise is filtered out using the soft threshold iterative function to obtain the result after the noise is filtered out. Among them, F k Represents the processing process of three-layer convolution Conv, represents the soft threshold iterative function;

[0028] The result after filtering out the noise is then mapped back to the image domain through three layers of convolution Conv and used as the residual and the initial complex matrix Z k If we want to add, we can get the updated nonlinear imaging result, the expression is:

[0029]

[0030] in, It represents the process of three-layer convolution Conv mapping the result back to the image domain, C is the constant coefficient of residual adjustment, X k is the nonlinear imaging result after the kth iteration update.

[0031] Optionally, the calculating loss function includes:

[0032] Calculate the less than zero penalty term for pulse emission mode

[0033] Penalty term of computational imaging algorithm in mapping between sparse domain and image domain

[0034] Calculate the mean square error (MSE) difference between the final imaging function and the target image

[0035] Calculate the loss function, the expression is:

[0036]

[0037] Among them, L(Γ Total ) is the loss value of the loss function, γ1, γ2, γ3 are the weight coefficients of different loss terms in the loss function, is the total number of pulses in pulse transmission mode, For nth a The emission parameters of the pulses, J is the total number of imaging images, j is the number of imaging images, K is the set number of iterations, GT is the target image, X K For the final imaging result, represents the L2 norm.

[0038] Optionally, the random parameter freezing strategy is used for the pulse emission moments to randomly select a portion of the pulse emission moments to freeze the gradient of the parameters, including:

[0039] Applying the random parameter freezing strategy, a set of mask sequences with the same length as the pulse emission time is randomly generated, and the expression is:

[0040]

[0041] Among them, each element in the mask sequence is a random variable that obeys the Bernoulli distribution;

[0042] The mask sequence and the loss function are used to compare the pulse emission pattern d η The gradient of is multiplied to achieve the purpose of random zeroing. The expression is:

[0043]

[0044] in, Indicates the pulse transmission mode d η The loss function Perform gradient calculation to obtain the corresponding gradient vector; It means multiplying the mask sequence m by the gradient vector term by term.

[0045] To achieve the above-mentioned purpose, a second aspect of the present application provides an electronic device, including: a processor, and a memory communicatively connected to the processor;

[0046] The memory stores computer-executable instructions;

[0047] The processor executes the computer-executable instructions stored in the memory to implement the method as described in any one of the first aspects.

[0048] To achieve the above-mentioned purpose, the third aspect embodiment of the present application proposes a computer-readable storage medium, in which computer-readable storage medium is stored computer-executable instructions, and when the computer-executable instructions are executed by a processor, they are used to implement the method as described in any one of the first aspects.

[0049] To achieve the above-mentioned purpose, the fourth aspect of the present application provides a computer program product, which implements any method in the first aspect when executed by a processor.

[0050] The technical solution provided by the embodiments of the present application brings at least the following beneficial effects:

[0051] By combining the learnable pulse emission pattern with the SAR imaging algorithm, the final imaging quality is optimized, so that the imaging algorithm can fully utilize the anti-aliasing advantage of the non-uniform emission pattern and successfully achieve the expansion of the wide mapping swath. The depth expansion method based on approximate observation is adopted to significantly reduce the computational complexity. By introducing a random parameter freezing strategy, the randomness and robustness of the pulse emission pattern are improved, thereby effectively improving the resolution and quality of SAR imaging and meeting the imaging requirements of high-resolution wide mapping swaths.

[0052] Additional aspects and advantages of the present application will be given in part in the description below, and in part will become apparent from the description below, or will be learned through the practice of the present application. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The above and / or additional aspects and advantages of the present application will become apparent and easily understood from the following description of the embodiments in conjunction with the accompanying drawings, in which:

[0054] Figure 1 A flowchart of a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm provided in an embodiment of the present application;

[0055] Figure 2 A flowchart of a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm provided in an embodiment of the present application;

[0056] Figure 3 A flowchart of a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm provided in an embodiment of the present application. DETAILED DESCRIPTION

[0057] Embodiments of the present application are described in detail below, and examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are intended to be used to explain the present application, and should not be construed as limiting the present application.

[0058] Most existing SAR imaging algorithms based on compressed sensing use random sampling to obtain downsampled echo signals, which only reduces the amount of data but does not change the minimum interval between adjacent pulses, thus failing to expand the mapping bandwidth. SAR imaging algorithms that achieve wide mapping bands often decouple the emission mode and the imaging algorithm and design them separately, resulting in the imaging algorithm being unable to fully utilize the anti-aliasing advantage of the non-uniform emission mode, thus affecting the imaging quality.

[0059] In order to overcome the problem that the existing wide-band SAR imaging algorithm cannot fully remove the aliasing caused by non-uniform emission, thereby affecting the imaging quality, the embodiment of the present application provides a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm. Figure 1 , 2 3 are flowchart diagrams of a joint optimization SAR imaging method combining a learnable emission mode and an imaging algorithm provided in an embodiment of the present application.

[0060] It should be noted that before the process starts, the following input parameters need to be set:

[0061] SAR simulation scene X: used to simulate the real scene of SAR imaging and provide the algorithm with the expected imaging target.

[0062] SAR system parameters Θ: important parameters that control the characteristics of the SAR system, including transmission frequency, bandwidth, sampling rate, etc., which directly affect the characteristics of the generated echo signal.

[0063] Iteration number K: controls the update number of the imaging algorithm and is used to adjust the accuracy of the optimization process (set to 4 times in the present invention).

[0064] Loss function coefficients γ1, γ2, γ3: respectively control the weights of different parts of the loss function to balance the imaging quality, anti-aliasing performance and emission mode limitations.

[0065] like Figure 1 As shown, the method comprises the following steps:

[0066] Step S1, generating a non-uniform pseudo echo signal according to SAR system parameters and a simulation scenario as an input signal of an imaging iteration process, and initializing an imaging result matrix to a zero matrix as an initial state of the imaging iteration process.

[0067] In the embodiment of the present application, firstly, a simulated azimuthally non-uniform echo signal Y is generated according to the SAR system parameter Θ and the simulation scene X as an input signal of the imaging iteration process.

[0068] It can be understood that the generated simulation signal Y is the input signal of the entire imaging process. It simulates the real SAR data and provides a basic reference for the subsequent imaging algorithm. The generated simulation signal not only provides actual observation data for the subsequent steps, but also defines the imaging bandwidth requirements and signal distribution of the system, and provides an optimization target for the imaging algorithm of step S2 and step S3.

[0069] In addition, the imaging result X needs to be initialized 0 is a zero matrix, which serves as the initial state of the imaging iteration process.

[0070] It can be understood that the zero matrix initialization provides an unbiased starting point, so that the imaging process starts from a consistent base state. This initialization step ensures that subsequent iterative updates can be optimized without initial bias, thus ensuring the stability of the imaging algorithm.

[0071] In addition, the initial imaging result matrix will be directly passed as input to the linear update part in step S2 to start the gradual optimization of imaging.

[0072] Step S2: Based on the current imaging result, an approximate echo is first generated by an inverse imaging operator, the residual between the approximate echo and the simulated echo is calculated, and then the imaging operator is used to obtain the imaging result of the residual, and the imaging result of the residual is used to linearly update the imaging result of the previous round.

[0073] In the embodiment of the present application, in each iteration, linear updating is used to gradually approach the target imaging result.

[0074] Specifically, in each update, the inverse imaging operator G() is used to image the previous round of results X k-1 Perform reverse projection to obtain the approximate echo matrix G(X k-1 ), the inverse imaging operator is used to simulate the process of returning from the image domain to the frequency domain during the imaging process in order to generate the observation signal corresponding to the current imaging result.

[0075] Then, calculate the approximate echo matrix G(X k-1 ) and the difference between the simulated echo Y to generate a residual matrix, which represents the deviation between the current imaging result and the real observation signal.

[0076] Then, the residual matrix is ​​projected according to the imaging operator to obtain the correction information mapped from the observation domain back to the image domain, that is, the approximate imaging result H(YG(X k-1 )), where H() represents the imaging operator. The imaging operator projects the frequency domain residual information back to the image domain, thereby applying a correction to the image representation of the current imaging result.

[0077] Finally, the residual coefficient α is used to approximate the imaging result H(YG(Xk-1 )) is weighted and superimposed on the imaging result X of the previous round k-1 On the above, we get the updated linear imaging result, which is expressed as:

[0078] Z k =X k-1 +α k ·H(YG(X k-1 ))

[0079] Among them, Z k is the linear imaging result after the kth iteration update.

[0080] Different from the traditional observation matrix method, the embodiment of the present application uses the inverse imaging operator and the imaging operator in combination, so that the imaging process can fully utilize the information in the frequency domain and the image domain in each iteration, and can convert the multiplication operation of the large matrix into multiple small matrix multiplications, greatly reducing the time complexity. This process not only ensures the effective correction between the observation signal and the imaging result, but also enables the algorithm to gradually approach the real imaging result with low computational complexity.

[0081] Step S3: Map the updated linear imaging result to the sparse domain, and use the soft threshold iterative function to filter out noise to obtain a nonlinear imaging result.

[0082] In the embodiment of the present application, in each iteration, the image may be nonlinearly updated to further refine the imaging result and enhance the image details.

[0083] Specifically, we first separate the linear imaging result Z as a complex matrix k The real and imaginary parts are converted into real matrix D(Z k ).

[0084] Then, the real matrix D(Z k ) is mapped to a sparse domain to remove noise and enhance the sparsity of the image.

[0085] Furthermore, the soft threshold iterative function is used to filter out noise and retain key image features to obtain the result after noise filtering. Among them, F k Represents the processing process of three-layer convolution Conv, Represents the soft threshold iterative function.

[0086] Finally, the result after filtering out the noise is mapped back to the image domain through three layers of convolution Conv and used as the residual and the initial complex matrix Z k If we want to add, we can get the updated nonlinear imaging result, the expression is:

[0087]

[0088] in, It represents the process of three-layer convolution Conv mapping the result back to the image domain, C is the constant coefficient of residual adjustment, X k is the nonlinear imaging result after the kth iteration update.

[0089] It should be noted that the nonlinear update effectively removes residual noise and enhances image details through sparse mapping and soft threshold iteration, so that the imaging results are further optimized in each round of iteration. The introduction of sparse mapping expands the application of the algorithm in non-sparse scenarios and ensures higher imaging quality.

[0090] Step S4: Repeat the linear and nonlinear updating steps until the set number of iterations is reached to obtain the final imaging result.

[0091] In the embodiment of the present application, step S2 and step S3 are repeated to alternately perform linear and nonlinear updates until a preset number of iterations K is reached, and then the iteration process is stopped to obtain the final imaging result X. K .

[0092] Through alternating optimization of linear and nonlinear updates, the imaging results gradually approach the target image in each round, and the optimization results are continuously accumulated, ultimately achieving high-precision imaging. The iterative optimization process not only reduces the error, but also effectively handles noise and details, thereby gradually improving the imaging quality.

[0093] Step S5: After the iteration is completed, the loss function and the gradient are calculated, and a random parameter freezing strategy is used for the pulse emission time, and a part of the pulse emission time is randomly selected to freeze the gradient of the parameters.

[0094] After the iteration is completed, the error of the current imaging result is evaluated by calculating the loss function and providing feedback for subsequent optimization.

[0095] In the embodiment of the present application, the loss function mainly includes three parts, namely:

[0096] (1) Less than zero penalty term for pulse emission mode Used to limit the parameter range in the emission mode to ensure that the parameters are reasonable;

[0097] (2) Penalty term of imaging algorithm in mapping between sparse domain and image domain Used to constrain the error of imaging algorithm in the process of grating domain and image mapping;

[0098] (3) The mean square error (MSE) gap between the final imaging function and the target image Measures the error between the final imaging result and the target image.

[0099] In the embodiment of the present application, the expression of the loss function is:

[0100]

[0101] Among them, L(Γ Total ) is the loss value of the loss function, γ1, γ2, γ3 are the weight coefficients of different loss terms in the loss function, is the total number of pulses in pulse transmission mode, For the na The emission parameters of the pulses, J is the total number of imaging images, j is the number of imaging images, K is the set number of iterations, GT is the target image, X K For the final imaging result, represents the L2 norm.

[0102] It should be noted that the loss function calculation results provide quantitative feedback for the optimization process and guide the adjustment direction of the pulse emission mode and imaging parameters in the next iteration. By minimizing the loss function, the imaging algorithm can gradually adjust various parameters to make the imaging results approach the optimal solution, thereby improving the imaging accuracy and anti-aliasing capability.

[0103] In addition, the embodiment of the present application also introduces a random parameter freezing strategy in the back-propagation process to enhance the randomness and robustness of the pulse emission pattern.

[0104] Specifically, we first apply the random parameter freezing strategy to randomly generate a set of mask sequences with the same length as the pulse emission time, expressed as:

[0105]

[0106] Each element in the mask sequence is a random variable that obeys the Bernoulli distribution, that is, each element takes the value of 1 with a probability of 70%.

[0107] Then, the mask sequence is compared with the loss function for the pulse emission pattern d η Multiply the gradient of , and randomly freeze 30% of the elements in the gradient to achieve the purpose of random zeroing. The expression is:

[0108]

[0109] in, Indicates the pulse transmission mode d η The loss function Perform gradient calculation to obtain the corresponding gradient vector; It means multiplying the mask sequence m by the gradient vector term by term.

[0110] It should be noted that the random part of the pulse emission pattern is used as a learnable variable, and the gradient of some sequence elements is set to zero during the update process through the random parameter freezing strategy, which improves the anti-interference ability of the emission pattern and makes the imaging algorithm more robust in complex environments. In addition, the freezing strategy prevents overfitting of the model during the optimization process, allowing the algorithm to maintain better generalization and stability.

[0111] Step S6: back-propagate and update the pulse emission mode and imaging algorithm parameters according to the gradient updated by the random parameter freezing strategy.

[0112] In the embodiment of the present application, through the random parameter freezing strategy, the optimized gradient makes the pulse emission mode more random and robust, reducing the performance degradation caused by overfitting. In addition, the strategy of freezing part of the gradient enhances the anti-aliasing capability and makes the emission mode more adaptable in complex environments. At the same time, the parameters of the imaging algorithm are dynamically optimized through back propagation, further improving the resolution and quality of the imaging results.

[0113] Step S7, continuously repeating steps S2-S6 until convergence, to obtain the final optimized pulse emission mode and matching imaging algorithm.

[0114] In the embodiment of the present application, the optimization steps S2-S6 are repeated until convergence, and finally the optimized pulse emission mode d is obtained. η and imaging algorithm parameters to achieve the goal of high-resolution and wide-swath imaging. In the wide-swath scenario, the optimized pulse emission mode has stronger anti-aliasing capability, which can achieve wide-swath imaging under the premise of high resolution; the matching imaging algorithm significantly improves the imaging clarity and detail expression while maintaining high computational efficiency, meeting the needs of high-resolution imaging.

[0115] In an actual SAR system, pulse emission and echo data collection can be performed according to the trained pulse emission time, and then the trained imaging algorithm is used to obtain the imaging result.

[0116] In terms of the effect of SAR imaging, in order to verify the superiority of the method in this application, the method in this application is compared with other literature methods, and the comparison imaging result indicators are shown in Table 1.

[0117] Table 1

[0118] method TBR ENT Other methods in the literature 36.39 4.912 The method proposed by the present invention 47.49 4.069

[0119] As shown in Table 1, the TBR value of this application is 47.49, which is significantly higher than 36.39 in other literature methods, indicating that this application has significant advantages in the utilization efficiency of time-frequency bandwidth. The improvement of TBR means that under the same bandwidth, this application can better meet the imaging requirements of wide mapping bands.

[0120] In addition, ENT is another indicator of imaging quality, which is used to measure the richness of image information. The lower the ENT value, the clearer the image and the less redundant information. The ENT value of this application is 4.069, which is significantly lower than the 4.912 of other methods in the literature, indicating that the method of this application also has advantages in imaging clarity.

[0121] The method of the present invention achieves higher TBR and lower ENT by jointly optimizing the pulse emission mode and the SAR imaging algorithm. Compared with the methods in other literatures, the method of the present invention significantly improves the imaging resolution and quality while achieving a wide mapping swath.

[0122] In order to implement the above embodiments, the present application also proposes an electronic device, comprising: a processor, and a memory communicatively connected to the processor; the memory stores computer-executable instructions; the processor executes the computer-executable instructions stored in the memory to implement the method provided by the above embodiments.

[0123] In order to implement the above embodiments, the present application also proposes a computer-readable storage medium, in which computer-executable instructions are stored. When the computer-executable instructions are executed by a processor, they are used to implement the methods provided by the above embodiments.

[0124] In order to implement the above embodiments, the present application also proposes a computer program product, including a computer program, which implements the methods provided by the above embodiments when executed by a processor.

[0125] The collection, storage, use, processing, transmission, provision and disclosure of user personal information involved in this application are in compliance with relevant laws and regulations and do not violate public order and good morals.

[0126] It should be noted that personal information from users should be collected for legitimate and reasonable purposes and should not be shared or sold outside of these legitimate uses. In addition, such collection / sharing should be carried out after receiving the user's informed consent, including but not limited to notifying the user to read the user agreement / user notice and sign the agreement / authorization including authorization of relevant user information before the user uses the function. In addition, any necessary steps should be taken to protect and safeguard access to such personal information data and ensure that others who have access to personal information data comply with its privacy policy and procedures.

[0127] The present application is expected to provide an implementation scheme for users to selectively block the use or access of personal information data. That is, the present disclosure is expected to provide hardware and / or software to prevent or block access to such personal information data. Once the personal information data is no longer needed, the risk can be minimized by limiting data collection and deleting the data. In addition, when applicable, such personal information is de-identified to protect the privacy of the user.

[0128] In the description of the aforementioned embodiments, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" etc. means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine the different embodiments or examples described in this specification and the features of the different embodiments or examples, without contradiction.

[0129] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "plurality" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined.

[0130] Any process or method description in a flowchart or otherwise described herein may be understood to represent a module, fragment or portion of code comprising one or more executable instructions for implementing the steps of a custom logical function or process, and the scope of the preferred embodiments of the present application includes alternative implementations in which functions may not be performed in the order shown or discussed, including performing functions in a substantially simultaneous manner or in the reverse order depending on the functions involved, which should be understood by technicians in the technical field to which the embodiments of the present application belong.

[0131] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by an instruction execution system, device or apparatus (such as a computer-based system, a system including a processor, or other system that can fetch instructions from an instruction execution system, device or apparatus and execute the instructions), or in combination with these instruction execution systems, devices or apparatuses. For the purpose of this specification, "computer-readable medium" can be any device that can contain, store, communicate, propagate or transmit a program for use by an instruction execution system, device or apparatus, or in combination with these instruction execution systems, devices or apparatuses. More specific examples of computer-readable media (a non-exhaustive list) include the following: an electrical connection with one or more wires (electronic device), a portable computer disk box (magnetic device), a random access memory (RAM), a read-only memory (ROM), an erasable and programmable read-only memory (EPROM or flash memory), a fiber optic device, and a portable compact disk read-only memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program is printed, since the program may be obtained electronically, for example, by optically scanning the paper or other medium and then editing, interpreting or processing in other suitable ways if necessary, and then stored in a computer memory.

[0132] It should be understood that the various parts of the present application can be implemented by hardware, software, firmware or a combination thereof. In the above-mentioned embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented by hardware, as in another embodiment, it can be implemented by any one of the following technologies known in the art or their combination: a discrete logic circuit having a logic gate circuit for implementing a logic function for a data signal, a dedicated integrated circuit having a suitable combination of logic gate circuits, a programmable gate array (PGA), a field programmable gate array (FPGA), etc.

[0133] A person skilled in the art may understand that all or part of the steps in the method for implementing the above-mentioned embodiment may be completed by instructing related hardware through a program, and the program may be stored in a computer-readable storage medium, which, when executed, includes one or a combination of the steps of the method embodiment.

[0134] In addition, each functional unit in each embodiment of the present application may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium.

[0135] The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, etc. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limiting the present application. A person of ordinary skill in the art may change, modify, replace and modify the above embodiments within the scope of the present application.

[0136] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this application can be executed in parallel, sequentially or in different orders, as long as the expected results of the technical solution of this application can be achieved, and this document is not limited here.

[0137] The above specific implementations do not constitute a limitation on the protection scope of this application. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of this application should be included in the protection scope of this application.

Claims

1. A joint optimization SAR imaging method combining a learnable emission pattern and an imaging algorithm, characterized in that: The following steps are involved: S1, generating a non-uniform pseudo echo signal according to SAR system parameters and simulation scenarios as an input signal of an imaging iteration process, and initializing an imaging result matrix to a zero matrix as an initial state of the imaging iteration process; S2. Based on the current imaging result, an approximate echo is first generated by an inverse imaging operator, the residual between the approximate echo and the simulated echo is calculated, and then the imaging operator is used to obtain the imaging result of the residual, and the imaging result of the residual is used to linearly update the imaging result of the previous round; S3, mapping the updated linear imaging result to the sparse domain, and using the soft threshold iterative function to filter out the noise to obtain the nonlinear imaging result; S4, repeating the linear and nonlinear updating steps until the set number of iterations is reached to obtain the final imaging result; S5. After the iteration, the loss function and gradient are calculated, and a random parameter freezing strategy is used for the pulse emission time, and a part of the pulse emission time is randomly selected to freeze the gradient of the parameters; S6, back-propagating and updating the pulse emission mode and imaging algorithm parameters according to the gradient updated by the random parameter freezing strategy; S7. Repeat steps S2-S6 until convergence, and obtain the final optimized pulse emission mode and matching imaging algorithm.

2. The method according to claim 1, characterized in that The method of generating a non-uniform pseudo echo signal according to the SAR system parameters and the simulation scene as an input signal of the imaging iteration process, and initializing the imaging result matrix to a zero matrix as an initial state of the imaging iteration process, includes: According to the SAR system parameter Θ and the simulation scene X, a simulated azimuth non-uniform echo signal Y is generated as the input signal of the imaging iteration process; Initialize imaging result X 0 is a zero matrix, which serves as the initial state of the imaging iteration process.

3. The method according to claim 2, characterized in that Based on the current imaging result, the inverse imaging operator is used to generate an approximate echo, the residual between the approximate echo and the simulated echo is calculated, and then the imaging operator is used to obtain the imaging result of the residual, and the imaging result of the residual is used to linearly update the imaging result of the previous round, including: Based on the previous round of imaging results X k-1 , according to the approximate observation inverse imaging operator, the approximate echo matrix G(X k-1 ), where G(·) represents the inverse imaging operator; Calculate the approximate echo matrix G(X k-1 ) and the residual of the simulated echo Y, and obtain the approximate imaging result H(YG(X) of the residual according to the imaging operator k-1 )), where H(·) represents the imaging operator; The residual coefficients and the approximate imaging results H(YG(X k-1 )) The product is added to the imaging result X of the previous round k-1 On the above, we get the updated linear imaging result, which is expressed as: Z k =X k-1 +α k ·H(Y-G(X k-1 )) Among them, Z k is the linear imaging result after the kth iteration update.

4. The method according to claim 3, characterized in that The updated linear imaging result is mapped to a sparse domain, and the noise is filtered out using a soft threshold iterative function to obtain a nonlinear imaging result, including: Separate the linear imaging result Z as a complex matrix k The real and imaginary parts of k ); Through three layers of convolution Conv, the real matrix D(Z k ) is mapped to the sparse domain, and the noise is filtered out using the soft threshold iterative function to obtain the result after the noise is filtered out. Among them, F k Represents the processing process of three-layer convolution Conv, represents the soft threshold iterative function; The result after filtering out the noise is then mapped back to the image domain through three layers of convolution Conv and used as the residual and the initial complex matrix Z k If we add, we get the updated nonlinear imaging result, the expression is: in, It represents the process of three-layer convolution Conv mapping the result back to the image domain, C is the constant coefficient of residual adjustment, X k is the nonlinear imaging result after the kth iteration update.

5. The method according to claim 4, characterized in that The calculation loss function includes: Calculate the less than zero penalty term for pulse emission mode Penalty term of computational imaging algorithm in mapping between sparse domain and image domain Calculate the mean square error (MSE) difference between the final imaging function and the target image Calculate the loss function, the expression is: Among them, L(Γ Total ) is the loss value of the loss function, γ1, γ2, γ3 are the weight coefficients of different loss terms in the loss function, is the total number of pulses in pulse transmission mode, For nth a The emission parameters of the pulses, J is the total number of imaging images, j is the number of imaging images, K is the set number of iterations, GT is the target image, X K The final imaging result is represents the L2 norm.

6. The method according to claim 5, characterized in that The random parameter freezing strategy is used for the pulse emission time, and a portion of the pulse emission time is randomly selected to freeze the gradient of the parameter, including: Applying the random parameter freezing strategy, a set of mask sequences with the same length as the pulse emission time is randomly generated, and the expression is: Among them, each element in the mask sequence is a random variable that obeys the Bernoulli distribution; The mask sequence and the loss function are used to compare the pulse emission pattern d η The gradient of is multiplied to achieve the purpose of random zeroing. The expression is: in, Indicates the pulse transmission mode d η The loss function Perform gradient calculation to obtain the corresponding gradient vector; Indicates that the mask sequence m Multiply by the gradient vector term by term.

7. An electronic device, characterized in that: include: A processor, and a memory communicatively connected to the processor; The memory stores computer-executable instructions; The processor executes the computer-executable instructions stored in the memory to implement the method according to any one of claims 1 to 6.

8. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, which are used to implement the method according to any one of claims 1 to 6 when executed by a processor.

9. A computer program product, characterized in that The invention comprises a computer program, which implements the method according to any one of claims 1 to 6 when being executed by a processor.

Images