Back-projection imaging method based on adam optimization, medium and device
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- XIDIAN UNIV
- Filing Date
- 2024-05-27
- Publication Date
- 2026-06-16
Smart Images
Figure CN118625319B_ABST
Abstract
Description
Technical Field
[0001] This invention belongs to the field of radar signal processing technology, specifically relating to a back projection imaging method, medium, and device based on Adam optimization. Background Technology
[0002] Inverse Synthetic Aperture Radar (ISAR) plays a crucial role in space target surveillance and space situational awareness due to its all-weather, all-day, and long-range imaging capabilities. ISAR continuously observes space targets over long periods and at wide angles by actively emitting broadband electromagnetic pulse signals to stably acquire high-resolution images. Theoretically, under the same observation conditions, the higher the frequency band and the larger the bandwidth of the signal emitted by the ISAR system, the higher the imaging resolution and the more accurate the detail depiction of the target. However, as the resolution of ISAR images continues to improve, the impact of target motion on imaging quality gradually increases. At this point, the high-order rotational components that are neglected in the low-frequency band and small bandwidth significantly affect the coherence between signal pulses, thus becoming a key factor affecting imaging quality and azimuth calibration accuracy.
[0003] To address the aforementioned issues, some researchers have proposed a rotation parameter estimation method called Average Range Profile Sharpness Maximization–Entire Image Sharpness Maximization (ARPSM-EISM). This method estimates the equivalent rotation velocity and the equivalent rotation center distance offset in two steps. First, based on the ARPSM criterion, the rotation velocity is estimated, and the distance is resampled according to the rotation velocity parameter. When the rotation velocity estimation accuracy is sufficiently high, the second-order range cell migration (RCM) is corrected, the image envelope is aligned, and the Average Range Profile Sharpness (ARPS) is maximized. Based on this, the phase error caused by the rotation center distance offset is iteratively compensated with the overall image sharpness (EIS) as the optimization objective, ultimately achieving ISAR imaging. However, in existing technologies, the coupling between ISAR echo range and azimuth height caused by the motion of maneuvering targets under large bandwidth and large rotation angle conditions results in high nonlinearity and complexity in the phase term compensation optimization problem, which is usually difficult to provide an accurate solution quickly. Some optimization algorithms are also at risk of getting trapped in local optima. In addition, with the increase in signal bandwidth, the increase in the size of the observed target, and the further increase in the imaging accumulation angle, the approximate model of the second-order expansion of the equivalent rotational motion of the target may also have the risk of mismatch.
[0004] Therefore, existing ISAR imaging methods suffer from poor imaging results. Summary of the Invention
[0005] To address the aforementioned problems in the prior art, this invention provides a back projection imaging method, medium, and apparatus based on Adam optimization.
[0006] The technical problem to be solved by this invention is achieved through the following technical solution:
[0007] In a first aspect, the present invention provides a back projection imaging method based on Adam optimization, comprising:
[0008] S101. Acquire ISAR echo signals and perform aperture division processing on the ISAR echo signals to obtain multiple sub-echo signals;
[0009] S102. Perform pulse compression and translation compensation processing on each sub-echo signal in sequence to obtain multiple focused imaging information;
[0010] S103. Obtain the current learning parameters for the current iteration, and use the current learning parameters to perform back projection BP imaging processing on multiple focusing imaging information to obtain the current imaging result.
[0011] S104. Calculate the image entropy of the current imaging result to obtain the current image entropy;
[0012] S105. Determine whether the current image entropy and the previous image entropy in the previous iteration meet the preset stopping condition.
[0013] S106. When the preset stopping condition of step S105 is not met, the current learning parameters are jointly optimized using the current image entropy and the Adam optimization algorithm to obtain optimized learning parameters, and the optimized learning parameters are used as the current learning parameters in step S103.
[0014] S107. Repeat steps S103-S106 until the preset stop condition of step S105 is met. Output the current imaging result obtained in the most recent execution of step S103 and use the current imaging result as the final projection imaging result.
[0015] Optionally, step S101 specifically includes:
[0016] Get the preset number of echoes;
[0017] The ISAR echo signal is divided into multiple sub-echo signals by aperture division according to the preset number of echoes.
[0018] Optionally, step S102 specifically includes:
[0019] Each sub-echo signal is pulse-compressed using delinear frequency modulation or matched filtering techniques to obtain multiple one-dimensional range profiles.
[0020] Translational compensation is performed on multiple one-dimensional distance images to obtain multiple focused imaging information.
[0021] Optionally, translational compensation processing is performed on multiple one-dimensional range images to obtain multiple focused imaging information, including:
[0022] Multiple one-dimensional distance images are envelope aligned using the neighbor correlation method to obtain aligned one-dimensional distance images.
[0023] The minimum entropy autofocus algorithm is used to perform phase compensation processing on the aligned one-dimensional range image to obtain multiple focused imaging information.
[0024] Optionally, step S103 specifically includes:
[0025] Get the current learning parameters for the current iteration;
[0026] The current learning parameters are used to divide multiple focused imaging information into grids to obtain the focused imaging grid;
[0027] The projection echo information is obtained by projecting the focused imaging grid using an interpolation algorithm.
[0028] Back-projection back-back imaging is performed on the projected echo information based on the current learning parameters to complete the echo compensation and coherent accumulation processing of the projected echo information, and obtain the current imaging result.
[0029] Optionally, the current learning parameters include: the current equivalent rotation center distance offset, the current equivalent rotation angular velocity, and the current equivalent rotation angular acceleration.
[0030] Optionally, the optimized learning parameters are represented as:
[0031]
[0032] Indicates the current learning parameters. This represents optimizing the learning parameters, where η represents the learning rate. The second moment is represented after bias correction, and ε represents a constant. The first moment after deviation correction;
[0033]
[0034]
[0035] m l Let v represent the first moment.l Let β1 represent the decay rate controlling the first-order moment estimate, β2 represent the decay rate controlling the second-order moment estimate, and l represent the iteration number. Let β1 represent the value of β1 in the l-th iteration. Let β2 represent the value of β2 in the l-th iteration.
[0036] Optionally, the preset stop conditions include:
[0037] The difference between the current image entropy and the previous image entropy is less than a preset threshold.
[0038] In a second aspect, the present invention provides a storage medium storing a computer program, which, when run by a processor, executes the steps of the Adam-optimized back projection imaging method described in the first aspect above.
[0039] Thirdly, the present invention provides an Adam-optimized rear projection imaging device, comprising: a processor, a storage medium and a bus, wherein the storage medium stores machine-readable instructions executable by the processor, and when the Adam-optimized rear projection imaging device is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the Adam-optimized rear projection imaging method described in the first aspect above.
[0040] This invention provides a back projection imaging method, medium, and device based on Adam optimization. The back projection imaging method based on Adam optimization includes: S101, acquiring ISAR echo signals and performing aperture segmentation processing on the ISAR echo signals to obtain multiple sub-echo signals; S102, sequentially performing pulse compression and translational compensation processing on each sub-echo signal to obtain multiple focused imaging information; S103, acquiring the current learning parameters for the current iteration, and using the current learning parameters to perform back projection (BP) imaging processing on the multiple focused imaging information to obtain the current imaging result; S104, calculating the image entropy of the current imaging result to obtain the current image entropy; S105, determining... S106. Does the current image entropy meet the preset stopping condition compared to the previous image entropy in the last iteration? S107. If the preset stopping condition in step S105 is not met, use the current image entropy and the Adam optimization algorithm to perform joint parameter optimization on the current learning parameters to obtain optimized learning parameters, and use the optimized learning parameters as the current learning parameters in step S103. S108. Repeat steps S103-S106 until the preset stopping condition in step S105 is met, output the current imaging result obtained in the most recent execution of step S103, and use the current imaging result as the final projection imaging result. In this invention, by performing aperture segmentation processing on the ISAR echo signal and acquiring the current imaging result based on the aperture segmentation processing result and the BP imaging processing method, the model mismatch problem caused by Taylor expansion of the equivalent rotation angle under large bandwidth and large rotation angle conditions in the prior art is avoided. Secondly, by performing joint parameter optimization processing on the current learning parameters, the error accumulation of the final optimization result caused by the propagation of different parameter estimation errors in the cascade optimization compensation method is avoided, thereby improving the focusing accuracy and calibration accuracy of the final projection imaging result. In addition, the Adam optimization algorithm is used for non-convex optimization problems, which improves the convergence speed. Finally, by using image entropy as the objective function of iterative optimization, the energy gain brought by the coherent accumulation of the range and azimuth dimensions can be obtained simultaneously, which improves the noise robustness of the method execution. The above improvements together improve the imaging effect of the ISAR echo signal.
[0041] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description
[0042] Figure 1 A schematic flowchart of a back projection imaging method based on Adam optimization provided in an embodiment of the present invention;
[0043] Figure 2 A schematic diagram of the overall execution flow of the Adam-optimized back projection imaging method provided in an embodiment of the present invention;
[0044] Figure 3The simulation target model provided in the embodiments of the present invention;
[0045] Figure 4 The imaging results provided by the embodiments of the present invention are for each imaging method when the accumulated rotation angle is 8.5944°.
[0046] Figure 5 The imaging results provided by the embodiments of the present invention are for each imaging method when the accumulated rotation angle is 25.7831°;
[0047] Figure 6 Image entropy change curves of the conventional method and the method of the present invention under different accumulation angles provided in the embodiments of the present invention;
[0048] Figure 7 This is a schematic diagram of a back projection imaging device based on Adam optimization, provided as an embodiment of the present invention. Detailed Implementation
[0049] The present invention will be further described in detail below with reference to specific embodiments, but the implementation of the present invention is not limited thereto.
[0050] To improve the imaging effect of ISAR echo signals, this invention provides a back projection imaging method based on Adam optimization. Figure 1 This is a schematic flowchart illustrating a back projection imaging method based on Adam optimization, provided as an embodiment of the present invention. Figure 1 As shown, the method includes:
[0051] S101. Acquire the ISAR echo signal and perform aperture division processing on the ISAR echo signal to obtain multiple sub-echo signals.
[0052] Optionally, step S101 specifically includes:
[0053] Get the preset number of echoes;
[0054] The ISAR echo signal is divided into multiple sub-echo signals by aperture division according to the preset number of echoes.
[0055] S102. Perform pulse compression and translation compensation processing on each sub-echo signal in sequence to obtain multiple focused imaging information.
[0056] Optionally, step S102 specifically includes:
[0057] Each sub-echo signal is pulse-compressed using delinear frequency modulation or matched filtering techniques to obtain multiple one-dimensional range profiles.
[0058] Translational compensation is performed on multiple one-dimensional distance images to obtain multiple focused imaging information.
[0059] Optionally, translational compensation processing is performed on multiple one-dimensional range images to obtain multiple focused imaging information, including:
[0060] Multiple one-dimensional distance images are envelope aligned using the neighbor correlation method to obtain aligned one-dimensional distance images.
[0061] The minimum entropy autofocus algorithm is used to perform phase compensation processing on the aligned one-dimensional range image to obtain multiple focused imaging information.
[0062] Optionally, in this embodiment of the invention, the phase compensation processing of the aligned one-dimensional range image is performed using the minimum entropy autofocus algorithm, which can compensate for the phase error caused by translation and radar system noise, so as to obtain focused imaging information with better focusing effect.
[0063] It should be noted that, in this embodiment of the invention, envelope alignment processing of the one-dimensional range image can correct the envelope offset phenomenon caused by the translation of the target in the one-dimensional range image, so that the same scattering point is in the same range cell in different echoes, thereby improving the final imaging effect.
[0064] S103. Obtain the current learning parameters for the current iteration, and use the current learning parameters to perform back projection BP imaging processing on multiple focusing imaging information to obtain the current imaging result.
[0065] It should be noted that in this embodiment, the initial learning parameters are manually initialized. The initial learning parameters generally include: the initial equivalent rotation center distance offset Δr, the initial equivalent rotation angular velocity ω, and the initial equivalent rotation angular acceleration a. Δr, ω, and a are generally set to 0 in the manual initialization settings.
[0066] Optionally, step S103 specifically includes:
[0067] Get the current learning parameters for the current iteration;
[0068] The current learning parameters are used to divide multiple focused imaging information into grids to obtain the focused imaging grid;
[0069] The projection echo information is obtained by projecting the focused imaging grid using an interpolation algorithm.
[0070] Back-projection back-back imaging is performed on the projected echo information based on the current learning parameters to complete the echo compensation and coherent accumulation processing of the projected echo information, and obtain the current imaging result.
[0071] In this embodiment, the x-coordinate and y-coordinate of the focused imaging grid can be represented as:
[0072] x = [x Sta :ρa :x End ];
[0073] y = [y Sta :ρ r :y End ];
[0074] Where, x Sta x End y Sta and y End These represent the initial dimension of the imaging plane in the azimuth direction, the final dimension of the imaging plane in the azimuth direction, the initial dimension of the imaging plane in the range direction, and the final dimension of the imaging plane in the range direction, respectively. ρ a Represents the azimuth resolution unit, ρ r This indicates a range-resolved unit.
[0075]
[0076]
[0077] Where λ is the wavelength, θ end To observe the initial rotation angle of the sub-aperture, θ sta To observe the final rotation angle of the sub-aperture, C is the speed of light, γ is the modulation frequency, and T... p Indicates the pulse width.
[0078] Additionally, it should be noted that in this embodiment, Where ω' represents the current equivalent rotational angular velocity in the current learning parameters, a' represents the current equivalent rotational angular acceleration in the current learning parameters, and t CPI This indicates the imaging coherence accumulation time.
[0079] Optionally, the current learning parameters include: the current equivalent rotation center distance offset, the current equivalent rotation angular velocity, and the current equivalent rotation angular acceleration.
[0080] S104. Calculate the image entropy of the current imaging result to obtain the current image entropy.
[0081] In this embodiment, the current image entropy can be expressed as:
[0082]
[0083]
[0084] Where Entropy(·) is the image entropy function, and I(y,x) represents the current imaging result. Let X represent the azimuth length of the current imaging result and Y represent the range length of the current imaging result.
[0085] S105. Determine whether the current image entropy and the previous image entropy in the previous iteration meet the preset stopping condition.
[0086] S106. When the preset stopping condition in step S105 is not met, the current learning parameters are jointly optimized using the current image entropy and the Adam optimization algorithm to obtain optimized learning parameters, and the optimized learning parameters are used as the current learning parameters in step S103.
[0087] Optionally, in this embodiment, other gradient descent optimization methods (such as RMSprop, Adagrad, etc.) besides the Adam optimization algorithm can also be used to search for and optimize the learning parameters. Alternatively, other image metrics (such as mean, sharpness, etc.) besides image entropy can be used to construct a function for solving the learning parameters, and the result of the solution function can be used as a criterion for whether optimization stops.
[0088] Optionally, the optimized learning parameters are represented as:
[0089]
[0090] Indicates the current learning parameters. This represents optimizing the learning parameters, where η represents the learning rate. The second moment is represented after bias correction, and ε represents a constant. The first moment after deviation correction;
[0091]
[0092]
[0093] m l Let v represent the first moment. l Let β1 represent the decay rate controlling the first-order moment estimate, β2 represent the decay rate controlling the second-order moment estimate, and l represent the iteration number. Let β1 represent the value of β1 in the l-th iteration. Let β2 represent the value of β2 in the l-th iteration.
[0094] In addition, the first moment m l It can be represented as:
[0095] m l =β1m l-1 +(1-β1)g l ;
[0096] Where, m l-1 Let g represent the first moment at iteration number l-1. l The gradient represents the current image entropy.
[0097] Second moment v l Represented as:
[0098]
[0099] v l-1 Denotes the second moment at iteration number l-1.
[0100] This represents the current learned parameters at iteration number l. This indicates gradient calculation.
[0101] Additionally, it should be noted that β1 and β2 are usually close to 1, for example, β1 = 0.9 and β2 = 0.999. ε is a very small number to avoid division by zero (it is usually set to 10). -8 ).
[0102] S107. Repeat steps S103-S106 until the preset stop condition of step S105 is met. Output the current imaging result obtained in the most recent execution of step S103 and use the current imaging result as the final projection imaging result.
[0103] When the difference between the current image entropy and the previous image entropy is less than 1e-9, the target loss function is considered to have converged, and the optimal current learning parameters have been obtained. These optimal current learning parameters are then saved. This represents the optimal current equivalent rotation center distance offset. Indicates the optimal current equivalent rotational angular velocity, This represents the optimal current equivalent rotational angular acceleration.
[0104] Based on this Back-projection (BP) imaging is performed on multiple focused imaging data to obtain the final projection imaging result.
[0105] For example, in an embodiment of this application, when the current learning parameter is The final projection imaging result I(y,x) obtained when executing S103 can be expressed as:
[0106]
[0107] I(y,x) represents the current imaging result as the final projection imaging result, M represents the echo number, m=[1:M] represents the m-th echo, σ p Δθ represents the backscattering intensity corresponding to the p-th scattering point. m Let x be the angular difference between the m-th echo and the (m-1)-th echo. pLet x and y represent the theoretical values of the p-th scattering point. p Let y and θ represent the theoretical values of the p-th scattering point. m This represents the equivalent rotation angle of the m-th echo. t m Let j represent the slow time of the m-th echo, and j represent an imaginary number.
[0108] sinc(.) represents the sinc function;
[0109]
[0110] sin(t) represents the sine function, and t represents the independent variable.
[0111] To illustrate the overall execution steps of the Adam-optimized back projection imaging method provided in this embodiment of the invention, Figure 2 This is a schematic diagram illustrating the overall execution flow of the Adam-optimized back projection imaging method provided in an embodiment of the present invention. Figure 2 As shown, after receiving the signal, pulse compression and translational compensation are performed to obtain a blurred image. An FFT transform is then performed on the blurred image to obtain the azimuth frequency domain signal. In the right-hand step, range-resolved units and azimuth-resolved units are first defined based on the initial equivalent rotation parameters, resulting in a new two-dimensional imaging grid. Back projection is then performed based on this new two-dimensional imaging grid and the azimuth frequency domain signal to obtain the focused image (the current imaging result). (The back projection process specifically involves searching for the phase error term in the current imaging result using the back projection algorithm and performing rotational phase error compensation). The image entropy (entropy difference) of this focused image is used to determine whether the iterative optimization process has ended. If it has ended, the final focused image is obtained; otherwise, the equivalent rotation parameters are updated and the next iteration begins.
[0112] This invention provides a back projection imaging method based on Adam optimization, comprising: S101, acquiring ISAR echo signals and performing aperture segmentation processing on the ISAR echo signals to obtain multiple sub-echo signals; S102, sequentially performing pulse compression and translational compensation processing on each sub-echo signal to obtain multiple focused imaging information; S103, acquiring the current learning parameters for the current iteration, and using the current learning parameters to perform back projection (BP) imaging processing on the multiple focused imaging information to obtain the current imaging result; S104, calculating the image entropy of the current imaging result to obtain the current image entropy; S105. 1. Determine whether the current image entropy and the previous image entropy in the previous iteration meet the preset stopping condition; S106. When the preset stopping condition in step S105 is not met, use the current image entropy and the Adam optimization algorithm to perform joint parameter optimization processing on the current learning parameters to obtain optimized learning parameters, and use the optimized learning parameters as the current learning parameters in step S103; S107. Repeat steps S103-S106 until the preset stopping condition in step S105 is met, output the current imaging result obtained in the most recent execution of step S103, and use the current imaging result as the final projection imaging result. In this invention, by performing aperture segmentation processing on the ISAR echo signal and acquiring the current imaging result based on the aperture segmentation processing result and the BP imaging processing method, the model mismatch problem caused by Taylor expansion of the equivalent rotation angle under large bandwidth and large rotation angle conditions in the prior art is avoided. Secondly, by performing joint parameter optimization processing on the current learning parameters, the error accumulation of the final optimization result caused by the propagation of different parameter estimation errors in the cascade optimization compensation method is avoided, thereby improving the focusing accuracy and calibration accuracy of the final projection imaging result. In addition, the Adam optimization algorithm is used for non-convex optimization problems, which improves the convergence speed. Finally, by using image entropy as the objective function of iterative optimization, the energy gain brought by the coherent accumulation of the range and azimuth dimensions can be obtained simultaneously, which improves the noise robustness of the method execution. The above improvements together improve the imaging effect of the ISAR echo signal.
[0113] Optionally, the preset stop conditions include:
[0114] The difference between the current image entropy and the previous image entropy is less than a preset threshold.
[0115] To demonstrate the effectiveness of the Adam-optimized back projection imaging method provided by this invention, simulation verification was also performed in the embodiments of this invention. Figure 3 The simulation target model provided for the embodiments of the present invention. Figure 4 The imaging results provided by the embodiments of the present invention are for each imaging method when the accumulated rotation angle is 8.5944°. Figure 5The images provided in this embodiment of the invention show the imaging results corresponding to various imaging methods when the accumulated rotation angle is 25.7831°. For example... Figure 4-5 As shown, Figure 4 and Figure 5 Figure (a) represents the traditional RD method. Figure 4 and Figure 5 Figure (b) represents the Keystone transformation method. Figure 4 and Figure 5 Figure (c) illustrates the method of the present invention. From... Figure 4 and Figure 5 It can be seen that as the accumulated rotation angle increases, the imaging results of the traditional RD method have obvious defocus. Even after using the Keystone transform to compensate for the error caused by rotation, the imaging results still have obvious defocus in the azimuth dimension. However, the image obtained by the method of this invention has a good focusing effect on any scattering point.
[0116] Figure 6 Image entropy change curves of the conventional method and the method of the present invention under different accumulation angles, as provided in embodiments of the present invention. Figure 6 It can be seen that when the accumulation angle is the same, the image entropy of the imaging results based on the method of the present invention is significantly lower than that of the traditional method, which means that the image focusing effect of the method of the present invention is better and the algorithm compensation capability is more accurate.
[0117] The method provided in this invention can be applied to electronic devices. Specifically, the electronic device can be a desktop computer, a portable computer, a smart mobile terminal, a server, etc. No limitation is made herein; any electronic device that can implement this invention falls within the protection scope of this invention.
[0118] Based on the same inventive concept, embodiments of the present invention also provide a rear projection imaging device based on Adam optimization. Figure 7 A schematic diagram of a Adam-optimized rear projection imaging device provided in an embodiment of the present invention includes: a processor 710, a storage medium 720, and a bus 730. The storage medium 720 stores machine-readable instructions executable by the processor 710. When the Adam-optimized rear projection imaging device is running, the processor 710 communicates with the storage medium 720 via the bus 730, and the processor 710 executes the machine-readable instructions to perform the steps of the above-described method embodiment. Specific implementation methods and technical effects are similar and will not be repeated here.
[0119] The storage medium may include random access memory (RAM) or non-volatile memory (NVM), such as at least one disk storage device. Optionally, the storage medium may also be at least one storage device located remotely from the aforementioned processor.
[0120] 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.
[0121] This invention also provides a computer-readable storage medium. The computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of any of the aforementioned Adam-optimized back projection imaging methods.
[0122] It should be noted that the terms "first," "second," etc., are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. Rather, they are merely examples of apparatuses and methods consistent with some aspects of the invention.
[0123] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Furthermore, those skilled in the art can combine and integrate the different embodiments or examples described in this specification.
[0124] Although the invention has been described herein in conjunction with various embodiments, those skilled in the art will understand and implement other variations of the disclosed embodiments by reviewing the accompanying drawings and the disclosure in carrying out the claimed invention. In the description of the invention, the word "comprising" does not exclude other components or steps, "a" or "an" does not exclude a plurality, and "a plurality" means two or more, unless otherwise explicitly specified. Furthermore, while different embodiments may describe certain measures, this does not mean that these measures cannot be combined to produce good results.
[0125] The above description, in conjunction with specific preferred embodiments, provides a further detailed explanation of the present invention. It should not be construed that the specific implementation of the present invention is limited to these descriptions. For those skilled in the art, various simple deductions or substitutions can be made without departing from the concept of the present invention, and all such modifications and substitutions should be considered within the scope of protection of the present invention.
Claims
1. A back projection imaging method based on Adam optimization, characterized in that, include: S101. Acquire the ISAR echo signal and perform aperture division processing on the ISAR echo signal to obtain multiple sub-echo signals; S102. Perform pulse compression and translation compensation processing on each of the sub-echo signals in sequence to obtain multiple focused imaging information; S103. Obtain the current learning parameters for the current iteration, and use the current learning parameters to perform back projection BP imaging processing on the multiple focused imaging information to obtain the current imaging result. S104. Calculate the image entropy of the current imaging result to obtain the current image entropy; S105. Determine whether the current image entropy and the previous image entropy of the previous iteration meet the preset stopping condition; S106. When the preset stopping condition in step S105 is not met, the current learning parameters are jointly optimized using the current image entropy and the Adam optimization algorithm to obtain optimized learning parameters, and the optimized learning parameters are used as the current learning parameters in step S103. S107. Repeat steps S103-S106 until the preset stop condition of step S105 is met, output the current imaging result obtained in the most recent execution of step S103, and use the current imaging result as the final projection imaging result. Step S103 specifically includes: Obtain the current learning parameters for the current iteration; The multiple focused imaging information are divided into grids using the current learning parameters to obtain a focused imaging grid. The focused imaging grid is projected using an interpolation algorithm to obtain the projected echo information. Back projection back-projection (BP) imaging is performed on the projected echo information based on the current learning parameters to complete echo compensation and coherent accumulation processing of the projected echo information, thereby obtaining the current imaging result. The current learning parameters include: current equivalent rotation center distance offset, current equivalent rotation angular velocity, and current equivalent rotation angular acceleration; The optimized learning parameters are represented as follows: ； Indicates the current learning parameters. This indicates optimizing the learning parameters. Indicates the learning rate. This represents the second moment after deviation correction. Represents a constant. The first moment after deviation correction; ； ； Describing the first moment, Describing the second moment, This represents the decay rate controlling the first-order moment estimate. This represents the decay rate controlling the second-order moment estimate. Indicates the number of iterations. Indicates the first under the number of iterations , Indicates the first under the number of iterations .
2. The back projection imaging method based on Adam optimization according to claim 1, characterized in that, Step S101 specifically includes: Get the preset number of echoes; The ISAR echo signal is divided into multiple sub-echo signals by aperture division according to the preset number of echoes.
3. The back projection imaging method based on Adam optimization according to claim 1, characterized in that, Step S102 specifically includes: Each sub-echo signal is pulse-compressed using delinear frequency modulation or matched filtering techniques to obtain multiple one-dimensional range profiles. Translational compensation processing is performed on the multiple one-dimensional distance images to obtain the multiple focused imaging information.
4. The back projection imaging method based on Adam optimization according to claim 3, characterized in that, The translational compensation processing of the plurality of one-dimensional range images to obtain the plurality of focused imaging information includes: The multiple one-dimensional distance images are envelope aligned based on the neighbor correlation method to obtain aligned one-dimensional distance images. The phase compensation process is performed on the aligned one-dimensional range image using the minimum entropy autofocus algorithm to obtain the multiple focused imaging information.
5. The back projection imaging method based on Adam optimization according to claim 1, characterized in that, The preset stop conditions include: The difference between the current image entropy and the previous image entropy is less than a preset threshold.
6. A storage medium, characterized in that, The storage medium stores a computer program, which, when executed by a processor, performs the steps of the Adam-optimized back projection imaging method as described in any one of claims 1-5.
7. A back projection imaging device based on Adam optimization, characterized in that, include: The device includes a processor, a storage medium, and a bus. The storage medium stores machine-readable instructions executable by the processor. When the Adam-optimized back projection imaging apparatus is running, the processor communicates with the storage medium via the bus, and the processor executes the machine-readable instructions to perform the steps of the Adam-optimized back projection imaging method as described in any one of claims 1-5.