ISAR coherent translation compensation and imaging method and device based on deep network

Abstract

The application provides a deep network-based ISAR coherent translational compensation and imaging method and device, and relates to the technical field of radar signal processing. The method comprises the following steps: pulse compression is performed on ISAR echoes to obtain a plurality of pulse-compressed echo signals; the plurality of pulse-compressed echo signals are sequentially input into a trained LSTM strategy network, so that the trained LSTM strategy network outputs a plurality of first distance grid offset estimation values; target translational parameters are obtained according to the plurality of first distance grid offset estimation values and the resolution of the distance dimension; the Adam optimizer is used to estimate the target translational parameters to obtain optimal target translational parameters; and the plurality of pulse-compressed echo signals are compensated and coherently accumulated according to the optimal target translational parameters to obtain a focused image. In this way, the compensation accuracy of the plurality of pulse-compressed echo signals is high, and the imaging result of the focused image is highly accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of radar signal processing, and particularly relates to an ISAR coherent translational compensation and imaging method and device based on a deep network. BACKGROUND

[0002] Inverse Synthetic Aperture Radar (ISAR) has all-weather, all-day, long-range and high-resolution imaging capabilities, and has important application value in situation awareness, prevention and early warning, civil aviation control and other fields. When ISAR imaging, the motion of a non-cooperative target can be equivalent to two processes, i.e., translational motion along the slant range and rotation around the center of the rotation platform. The translational motion along the slant range is meaningless for ISAR imaging, and will introduce envelope translation and phase error, so that the final imaging result is poor. The process of compensating for the translational motion is called translational compensation.

[0003] At present, the translational compensation is usually achieved by using a new parameterized ISAR motion compensation method based on a particle swarm optimizer to estimate polynomial coefficients, and Average Range Profile Energy (ARPE) and Entire Image Contrast (EIC) are selected as fitness functions for envelope alignment and phase auto-focusing, respectively. However, the envelope offset estimation result with ARPE as the target function often cannot meet the distance resolution precision requirement of focused imaging, and the envelope offset estimation error cannot be compensated in subsequent phase error estimation, thereby limiting the estimation precision of the entire algorithm, so that the precision of ISAR coherent translational compensation is low, and the accuracy of the imaging result is low. SUMMARY

[0004] The embodiment of the present application aims to provide an ISAR coherent translational compensation and imaging method and device based on a deep network, and solve the problem of low precision of ISAR coherent translational compensation and low accuracy of the imaging result.

[0005] To solve the above technical problems, the embodiment of the present application provides the following technical scheme:

[0006] The first aspect of the present application provides an ISAR coherent translational compensation and imaging method based on a deep network, and the method comprises the following steps:

[0007] Pulse compression is performed on the ISAR echo to obtain a plurality of pulse-compressed echo signals;

[0008] The plurality of pulse compressed echo signals are sequentially input into the trained LSTM strategy network, so that the trained LSTM strategy network outputs a plurality of first distance grid offset estimation values, the trained LSTM strategy network is a network obtained by training the LSTM strategy network using a training set, and the training set is a data set of the same signal form as the plurality of pulse compressed echo signals.

[0009] According to the plurality of first distance grid offset estimation values and the resolution of the distance dimension, the target translation parameter is obtained.

[0010] The target translation parameter is estimated using an Adam optimizer to obtain an optimal target translation parameter.

[0011] According to the optimal target translation parameter, the plurality of pulse compressed echo signals are compensated and coherently accumulated to obtain a focused image.

[0012] The second aspect of the application provides an ISAR coherent translation compensation and imaging device based on a deep network, the device comprising:

[0013] The pulse compression module is configured to pulse compress the ISAR echo to obtain a plurality of pulse compressed echo signals.

[0014] The input module is configured to sequentially input the plurality of pulse compressed echo signals into the trained LSTM strategy network, so that the trained LSTM strategy network outputs a plurality of first distance grid offset estimation values, the trained LSTM strategy network is a network obtained by training the LSTM strategy network using a training set, and the training set is a data set of the same signal form as the plurality of pulse compressed echo signals.

[0015] The acquisition module is configured to obtain the target translation parameter according to the plurality of first distance grid offset estimation values and the resolution of the distance dimension.

[0016] The estimation module is configured to estimate the target translation parameter using an Adam optimizer to obtain an optimal target translation parameter.

[0017] The compensation and coherent accumulation module is configured to compensate and coherently accumulate the plurality of pulse compressed echo signals according to the optimal target translation parameter to obtain a focused image.

[0018] Compared with the prior art, the ISAR coherent translational compensation and imaging method and device based on a deep network provided by the application perform pulse compression on ISAR echoes to obtain a plurality of pulse-compressed echo signals; the plurality of pulse-compressed echo signals are sequentially input into a trained LSTM strategy network, so that the trained LSTM strategy network outputs a plurality of first distance grid offset estimation values; target translational parameters are obtained according to the plurality of first distance grid offset estimation values and the resolution of the distance dimension; the Adam optimizer is used to estimate the target translational parameters to obtain optimal target translational parameters; and the plurality of pulse-compressed echo signals are compensated and coherently accumulated according to the optimal target translational parameters to obtain a focused image. In this way, the Adam optimizer is used to realize more stable non-convex problem solving, the target translational parameters can be accurately estimated, the influence of the target translational component on imaging is eliminated, the imaging result of the focused image is more clear and accurate, and the accuracy of the imaging result of the focused image is higher; the optimal target translational parameters are used to directly compensate and coherently accumulate the plurality of pulse-compressed echo signals, the transmission of different target translational parameter estimation errors in the cascaded compensation mode is avoided, the error accumulation of the final compensation result is avoided, the compensation accuracy of the plurality of pulse-compressed echo signals is higher, and the focusing degree and image scaling accuracy of the imaging result of the focused image are improved, so that the accuracy of the imaging result of the focused image is higher. BRIEF DESCRIPTION OF DRAWINGS

[0019] The above and other objects, features and advantages of the exemplary embodiments of the present application will be more apparent from the following detailed description read in conjunction with the accompanying drawings, in which several embodiments of the present application are shown by way of example, and wherein the same reference numerals refer to the same or similar components throughout the several views, in which:

[0020] Figure 1 A flowchart of the ISAR coherent translational compensation and imaging method based on a deep network is schematically shown;

[0021] Figure 2 An experimental target optical image is schematically shown;

[0022] Figure 3 A comparison diagram of the imaging result of the method of the application and the imaging result of a conventional method under different signal-to-noise ratios is schematically shown;

[0023] Figure 4 A curve diagram of a loss function changing with a signal-to-noise ratio is schematically shown;

[0024] Figure 5 A structural diagram of the ISAR coherent translational compensation and imaging device based on a deep network is schematically shown. DETAILED DESCRIPTION

[0025] Exemplary embodiments of the present application will be described herein below with reference to the accompanying drawings. While exemplary embodiments of the present application are shown in the drawings, it is understood that the present application can be embodied in various forms and should not be limited by the embodiments set forth herein. Rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the scope of the application to those skilled in the art.

[0026] It should be noted that unless otherwise stated, technical or scientific terms used herein should be understood in the ordinary sense employed by one of ordinary skill in the art to which the present application pertains.

[0027] The method in the embodiments of the present application will be described in detail below.

[0028] Figure 1 A flowchart of the deep network-based ISAR coherent translational compensation and imaging method in the embodiments of the present application is schematically shown, see Figure 1 As shown, the method can include:

[0029] S101, pulse compression is performed on the ISAR echo to obtain a plurality of pulse-compressed echo signals.

[0030] The ISAR echo is a large-bandwidth and long-time radar echo obtained by a radar receiver.

[0031] Specifically, pulse compression is performed on the ISAR echo to obtain a plurality of pulse-compressed echo signals, including:

[0032] Step A1: dividing the ISAR echo into a plurality of sub-apertures.

[0033] Step A2: performing distance compression processing on the plurality of sub-apertures to obtain corresponding one-dimensional range images, and transforming the one-dimensional range images to the frequency domain to obtain a plurality of frequency domain signals.

[0034] The linear frequency modulation technology is adopted to perform distance compression processing on the plurality of sub-apertures to obtain corresponding one-dimensional range images.

[0035] Step A3: according to the pulse compression method, constructing compensation factors of residual video (RVP) terms and envelope slant terms corresponding to the plurality of frequency domain signals.

[0036] Step A4: using the compensation factors of the residual video terms and the envelope slant terms, performing phase compensation on the peak frequency of each target scattering point in the plurality of frequency domain signals to obtain a plurality of pulse-compressed echo signals.

[0037] The plurality of pulse-compressed echo signals can also be referred to as a plurality of compensated echo signals.

[0038] S102. Inputting the echo signals after multiple pulse compression into the trained LSTM strategy network in sequence, so that the trained LSTM strategy network outputs multiple first distance grid offset estimation values.

[0039] The trained LSTM strategy network is a network obtained by training the LSTM strategy network using a training set, and the training set is a data set having the same signal form as the echo signals after compression of a plurality of pulses.

[0040] Specifically, the echo signals after multiple pulse compression are sequentially input into the trained LSTM strategy network, so that the hidden nodes in the trained LSTM strategy network obtain multiple first distance grid offset estimation values ​​through a fully connected layer with an output dimension of 1.

[0041] The multiple first range grid offset estimation values ​​are range grid offset estimation values ​​corresponding to the center position of the echo target.

[0042] Specifically, the LSTM policy network is trained using the training set, including:

[0043] Step B1: Use the predicted distance parameters in the training set as the initial state s0 of the LSTM policy network.

[0044] The predicted distance parameters include the predicted speed, the predicted acceleration, and the predicted jerk.

[0045] Step B2: Based on the initial state s0, feedback the training action a0 of the LSTM policy network.

[0046] Step B3: Acquire the actual state s1 according to the actual distance parameters of the echo signals after multiple pulse compression.

[0047] Step B4: Calculate the gap loss between the initial state and the actual state.

[0048] The gap loss can be expressed as c0(s0,s1).

[0049] Step B5: Update the initial state according to the training action, gap loss and learning rate, and return to the step of feeding back the training action of the LSTM policy network according to the initial state until the total number of training times reaches the preset total number of times, and then stop training to obtain the trained LSTM policy network.

[0050] According to the training action, gap loss and learning rate, the initial state is updated and the process returns to step B2 until the total number of training times reaches the preset total number of times to obtain the trained LSTM policy network.

[0051] The preset total number of times may be 200 times, and there may be multiple preset total numbers, which are not limited here.

[0052] In addition to the LSTM policy network, the application can also use other time-dependent recurrent neural network structures, such as Gated Recurrent Unit (GRU) network, Recurrent Neural Network (RNN), Bidirectional Long Short Term Memory (Bi LSTM) network, Stacked Long Short Term Memory (Stacked LSTM), etc. to extract features to estimate the target translational parameters.

[0053] The deep network, i.e., the LSTM policy network, is used to coarsely estimate the predicted distance parameters, which avoids the problem that the parameterized initial state setting cannot be self-adaptive, and improves the overall operation efficiency.

[0054] S103, obtaining the target translational parameters according to the plurality of first distance grid offset estimation values and the resolution of the distance dimension.

[0055] The target translational parameters include target velocity, target acceleration and target jerk.

[0056] Specifically, the target translational parameters are obtained according to the plurality of first distance grid offset estimation values and the resolution of the distance dimension, including:

[0057] Step C1: multiplying the plurality of first distance grid offset estimation values and the resolution of the distance dimension to obtain a plurality of second distance grid offset estimation values.

[0058] The plurality of second distance grid offset estimation values is a coarse estimation value.

[0059] Step C2: fitting the plurality of second distance grid offset estimation values to obtain a corresponding fitting curve.

[0060] Step C3: obtaining the target translational parameters through the fitting curve.

[0061] The target translational parameters include target velocity v, target acceleration a and target jerk a r .

[0062] S104, estimating the target translational parameters by using the Adam optimizer to obtain optimal target translational parameters.

[0063] In addition to the Adam optimizer, the application can also use other gradient descent type optimizers, such as the Root Mean Square Propagation (RMSprop) optimizer, the Adagrad optimizer, and the like.

[0064] Specifically, the target translational parameter is estimated by using the Adam optimizer to obtain an optimal target translational parameter, which includes:

[0065] Step D1: According to the target translational parameter, the multiple pulse compressed echo signals are compensated and imaged to obtain the imaging result of the current round.

[0066] Specifically, according to the target translational parameter, the multiple pulse compressed echo signals are compensated and imaged to obtain the imaging result of the current round, which includes:

[0067] Step D11: The target translational parameter (v, a, a r ) is taken as the learning parameter of the imaging network to obtain an imaging parameter estimation value

[0068] Step D12: According to the imaging parameter estimation value, a first translational compensation term corresponding to the multiple pulse compressed echo signals is constructed.

[0069] Step D13: According to the first translational compensation term, the multiple pulse compressed echo signals are compensated to obtain multiple compensated echo signals.

[0070] Step D14: The multiple compensated echo signals are coherently accumulated to obtain the imaging result of the current round.

[0071] Step D2: According to the imaging result of the current round, a loss function corresponding to the current round is determined.

[0072] The loss function of the current round is the image entropy of the imaging result of the current round.

[0073] In addition to the image entropy of the imaging result of the current round, the loss function can also be other image indicators, such as mean, sharpness, and the like.

[0074] Specifically, the expression of the loss function of the current round is:

[0075]

[0076] wherein, Entropy(·) is the loss function of the current round, i.e., the image entropy of the imaging result of the current round, is the average energy value of the imaging result of the current round, I(y, x) is the imaging result of the current round, X is the length of the image azimuth direction, Y is the length of the image distance direction, x is the image azimuth direction index, and y is the image distance direction index.

[0077] Step D3: solving the target translation parameter of the next round by using the Adam optimizer according to the target translation parameter and the loss function of the current round.

[0078] Specifically, the target translation parameter of the next round is solved by using the Adam optimizer according to the target translation parameter and the loss function of the current round, including:

[0079] Step D31: taking the target translation parameter (v, a, a r ) as the optimization parameter of the current round and initializing the first moment m l and the second moment v l of the Adam optimizer.

[0080] The first moment m l and the second moment v l of the Adam optimizer are initialized to 0.

[0081] Step D32: calculating the gradient of the loss function of the current round according to the loss function of the current round and the optimization parameter of the current round.

[0082] The expression of the gradient of the loss function of the current round is:

[0083]

[0084] wherein g l is the gradient of the loss function of the current round, Entropy(I(y, x)) is the loss function of the current round, is the optimization parameter of the current round, I(y, x) is the imaging result of the current round, x is the image azimuth direction index, and y is the image distance direction index.

[0085] Step D33: updating the initialized first moment according to the gradient, the initialized first moment, and the first decay rate to obtain an updated first moment.

[0086] The expression of the updated first moment is:

[0087]

[0088] wherein m l ′ is the updated first moment, is the first decay rate, m l is the initialized first moment, and g l is the gradient.

[0089] the first decay rate to control the decay rate of the initialized first moment, the first decay rate β1 l close to 1, for example,

[0090] Step D34: updating the initialized second moment according to the gradient, the initialized second moment and the second decay rate, to obtain an updated second moment.

[0091] The expression of the updated second moment is:

[0092]

[0093] wherein v l is the updated second moment, is the second decay rate, v l is the initialized second moment, g l is the gradient.

[0094] the second decay rate to control the decay rate of the initialized second moment, the second decay rate β2 close to 1, for example,

[0095] Step D35: performing bias correction on the updated first moment and the updated second moment, to obtain a corrected first moment and a corrected second moment.

[0096] The expression of the corrected first moment is:

[0097]

[0098] wherein, is the corrected first moment, m l is the updated first moment, is the first decay rate.

[0099] The expression of the corrected second moment is:

[0100]

[0101] wherein, is the corrected second moment, is the updated second moment, β2 l is the second decay rate.

[0102] Step D36: obtaining the optimization parameter of the next round according to the optimization parameter of the current round, the corrected first moment, the corrected second moment and the learning rate, the optimization parameter of the next round being the target translation parameter of the next round.

[0103] The expression of the optimization parameter of the next round is:

[0104]

[0105] wherein, is the optimization parameter of the next round, is the optimization parameter of the current round, is the first moment after correction, is the second moment after correction, η is a learning rate, ε is a non-zero real number, which is a very small number, and is usually set to 10 -8 .

[0106] Step D4: determining whether the difference between the loss function of the current round and the loss function of the last round is less than a preset value, if yes, the loss function of the current round converges, and the target translation parameter of the next round is determined as the optimal target translation parameter; if no, the target translation parameter of the next round is taken as the target translation parameter, and the step of performing translation compensation and imaging on the multiple pulse compressed echo signals according to the target translation parameter is returned to obtain the imaging result of the current round, until the difference between the loss function of the next round and the loss function of the current round is less than the preset value, and the iteration is stopped.

[0107] The preset value can be 10 -9 .

[0108] Specifically, it is determined whether the difference between the loss function of the current round and the loss function of the last round is less than 10 -9 , if yes, the loss function of the current round converges, indicating that the optimal target translation parameter has been obtained, and the target translation parameter of the next round is determined as the optimal target translation parameter. If no, the target translation parameter of the next round is taken as the target translation parameter, and the step D1 is returned until the difference between the loss function of the next round and the loss function of the current round is less than 10 -9 , and the iteration is stopped to obtain the optimal target translation parameter.

[0109] The Adam optimizer is used to realize more stable solution of non-convex problems, and then the target translation parameter is accurately estimated to eliminate the influence of the target translation component on imaging.

[0110] S105, compensating and coherently accumulating the multiple pulse compressed echo signals according to the optimal target translation parameter to obtain a focused image.

[0111] Specifically, the multiple pulse compressed echo signals are compensated and coherently accumulated according to the optimal target translation parameter to obtain a focused image, including:

[0112] According to the optimal target translational parameter, a plurality of second translational compensation terms corresponding to the plurality of pulse compressed echo signals are constructed; the plurality of pulse compressed echo signals are compensated according to the second translational compensation terms to obtain a plurality of final compensated echo signals; and the plurality of final compensated echo signals are coherently accumulated to perform two-dimensional imaging to obtain a focused image.

[0113] The target translational parameter estimation process aims at the loss function of the final imaging result, i.e., image entropy, and directly constructs a translational compensation term to jointly compensate the envelope error and phase error caused by the target translational parameter, thereby avoiding the transmission of different target translational parameter estimation errors in the cascaded compensation mode, resulting in error accumulation of the final compensation result, improving the focusing degree of the imaging result of the focused image and the image calibration precision. The Adam optimizer is used for non-convex optimization problems, and the convergence speed is better than other gradient descent algorithms, and the target translational parameter update is not affected by the scaling transformation of the gradient, can automatically adjust the learning rate, and is suitable for unstable target functions. At the same time, taking the image entropy as the optimization objective function can obtain the energy gain brought by the two-dimensional coherent accumulation of distance and azimuth, ensuring that the algorithm has high noise robustness.

[0114] Figure 2 An experimental target optical image is schematically shown, Figure 3 A comparison chart of the imaging results of the method of the present application and the imaging results of the traditional method under different signal-to-noise ratios is schematically shown, Figure 3 is the result of imaging Figure 2 , see Figure 3 , Figure 3 The horizontal coordinate of all the pictures in Figure 3 is the azimuth unit number, and the vertical coordinate is the distance unit number. The two pictures in the first row are the imaging results of the traditional method and the imaging results of the method of the present application under the condition of 15 dB signal-to-noise ratio, the two pictures in the second row are the imaging results of the traditional method and the imaging results of the method of the present application under the condition of 10 dB signal-to-noise ratio, the two pictures in the third row are the imaging results of the traditional method and the imaging results of the method of the present application under the condition of 5 dB signal-to-noise ratio, and the two pictures in the last row are the imaging results of the traditional method and the imaging results of the method of the present application under the condition of 0 dB signal-to-noise ratio. Figure 3 In , for electromagnetic data, the imaging results of the traditional method have obvious defocusing under any signal-to-noise ratio, while the method of the present application can finally estimate the corresponding optimal parameters according to different test data, and then obtain better focused imaging results, verifying the effectiveness and noise resistance of the method of the present application.

[0115] Figure 4 A curve graph of the loss function with respect to the signal-to-noise ratio is schematically shown, see Figure 4As shown in the figure, the abscissa is the signal-to-noise ratio, the ordinate is the image entropy value, that is, the value of the loss function, the blue line is the image entropy value of the imaging result of the traditional method, and the orange line is the image entropy value of the imaging result of the method of the application. It can be seen intuitively that when the signal-to-noise ratio is the same, the image entropy value of the imaging result of the method of the application is significantly lower than that of the imaging result of the traditional method, that is, the image focusing effect of the method of the application is better, and the algorithm compensation capability is more accurate.

[0116] Based on the above Figure 1 It can be seen from the implementation mode that the ISAR coherent translational compensation and imaging method based on a deep network includes pulse compression on ISAR echoes to obtain a plurality of pulse-compressed echo signals; the plurality of pulse-compressed echo signals are sequentially input into a trained LSTM strategy network to enable the trained LSTM strategy network to output a plurality of first distance grid offset estimation values; target translational parameters are obtained according to the plurality of first distance grid offset estimation values and the resolution of the distance dimension; the Adam optimizer is used to estimate the target translational parameters to obtain optimal target translational parameters; and the plurality of pulse-compressed echo signals are compensated and coherently accumulated according to the optimal target translational parameters to obtain a focused image. In this way, the Adam optimizer is used to realize more stable non-convex problem solving, the target translational parameters can be accurately estimated, the influence of the target translational component on imaging is eliminated, the imaging result of the focused image is more clear and accurate, and the accuracy of the imaging result of the focused image is higher. The optimal target translational parameters are used to directly compensate and coherently accumulate the plurality of pulse-compressed echo signals, avoiding the transmission of different target translational parameter estimation errors in the cascaded compensation mode, resulting in error accumulation of the final compensation result. The compensation accuracy of the plurality of pulse-compressed echo signals is higher, and the focusing degree and image scaling accuracy of the imaging result of the focused image are further improved, so that the accuracy of the imaging result of the image is higher.

[0117] Based on the same inventive concept, as an implementation of the above-mentioned ISAR coherent translational compensation and imaging method based on a deep network, the embodiment of the application further provides an ISAR coherent translational compensation and imaging device based on a deep network. Figure 5 The structure diagram of the ISAR coherent translational compensation and imaging device based on a deep network in the embodiment of the application is shown in FIG. 5. Figure 5 As shown in the figure, the device can include:

[0118] The pulse compression module 501 is configured to pulse compress ISAR echoes to obtain a plurality of pulse-compressed echo signals.

[0119] An input module 502 is configured to sequentially input the plurality of pulse-compressed echo signals into a trained LSTM strategy network, so that the trained LSTM strategy network outputs a plurality of first distance grid offset estimates, wherein the trained LSTM strategy network is a network obtained by training the LSTM strategy network using a training set, wherein the training set is a data set having the same signal form as the plurality of pulse-compressed echo signals;

[0120] An acquisition module 503 is configured to acquire target translation parameters based on a plurality of first range grid offset estimation values ​​and a resolution of the range dimension;

[0121] An estimation module 504 is used to estimate the target translation parameters using an Adam optimizer to obtain the optimal target translation parameters;

[0122] The compensation and coherent accumulation module 505 is used to compensate and coherently accumulate the echo signals after multiple pulse compression according to the optimal target translation parameters to obtain a focused image.

[0123] The pulse compression module 501 is specifically used to divide the ISAR echo into multiple sub-apertures; perform range compression processing on the multiple sub-apertures to obtain corresponding one-dimensional range images, and transform the one-dimensional range images into the frequency domain to obtain multiple frequency domain signals; construct compensation factors for the residual video items and envelope tilt items corresponding to the multiple frequency domain signals based on the pulse compression method; use the compensation factors of the residual video items and envelope tilt items to perform phase compensation on the peak frequency of each target scattering point in the multiple frequency domain signals to obtain multiple echo signals after pulse compression.

[0124] The acquisition module 503 is specifically configured to multiply the multiple first distance grid offset estimation values ​​and the resolution of the distance dimension to obtain multiple second distance grid offset estimation values; fit the multiple second distance grid offset estimation values ​​to obtain corresponding fitting curves; and obtain the target translation parameters through the fitting curves.

[0125] The estimation module 504 is specifically configured to perform translational compensation and imaging on the plurality of pulse-compressed echo signals according to the target translational parameter to obtain an imaging result of a current round; determine a loss function of the corresponding current round according to the imaging result of the current round; solve a target translational parameter of a next round by using an Adam optimizer according to the target translational parameter and the loss function of the current round; determine whether a difference between the loss function of the current round and a loss function of a previous round is less than a preset value, if yes, the loss function of the current round converges, and the target translational parameter of the next round is determined as an optimal target translational parameter, if not, the target translational parameter of the next round is taken as the target translational parameter, and the step of obtaining the imaging result of the current round by performing translational compensation and imaging on the plurality of pulse-compressed echo signals according to the target translational parameter is returned until the difference between the loss function of the next round and the loss function of the current round is less than the preset value, and the iteration is stopped.

[0126] The estimation module 504 performs translational compensation and imaging on the plurality of pulse-compressed echo signals according to the target translational parameter to obtain an imaging result of a current round, and the method comprises the following steps: taking the target translational parameter as a learning parameter of an imaging network to obtain an imaging parameter estimation value; constructing a first translational compensation term corresponding to the plurality of pulse-compressed echo signals according to the imaging parameter estimation value; compensating the plurality of pulse-compressed echo signals according to the first translational compensation term to obtain a plurality of compensated echo signals; and performing coherent accumulation on the plurality of compensated echo signals to obtain the imaging result of the current round.

[0127] The estimation module 504, the expression of the loss function of the current round is:

[0128]

[0129] Wherein, Entropy(·) is the loss function of the current round, I(y,x) is an average energy value of the imaging result of the current round, I(y,x) is the imaging result of the current round, X is an image azimuth length, Y is an image range length, x is an image azimuth index, and y is an image range index.

[0130] The estimation module 504 solves the target translational parameter of the next round according to the target translational parameter and the loss function of the current round by using the Adam optimizer, including: taking the target translational parameter as the optimization parameter of the current round, and initializing the first moment and the second moment of the Adam optimizer; calculating the gradient of the loss function of the current round according to the loss function of the current round and the optimization parameter of the current round; updating the initialized first moment according to the gradient, the initialized first moment and the first decay rate, to obtain the updated first moment; updating the initialized second moment according to the gradient, the initialized second moment and the second decay rate, to obtain the updated second moment; performing bias correction on the updated first moment and the updated second moment to obtain the corrected first moment and the corrected second moment; obtaining the optimization parameter of the next round according to the optimization parameter of the current round, the corrected first moment, the corrected second moment and the learning rate, and the optimization parameter of the next round is the target translational parameter of the next round.

[0131] The target translational parameter includes a target velocity, a target acceleration and a target jerk.

[0132] It should be pointed out here that the above description of the ISAR phase coherent translational compensation and imaging device based on the deep network is similar to the description of the above-mentioned ISAR phase coherent translational compensation and imaging method based on the deep network, and has the same beneficial effects as the ISAR phase coherent translational compensation and imaging method based on the deep network. For technical details of the ISAR phase coherent translational compensation and imaging device based on the deep network of the embodiments of the present application, please refer to the description of the ISAR phase coherent translational compensation and imaging method based on the deep network of the present application.

[0133] The above is only a specific embodiment of the present application, but the protection scope of the present application is not limited thereto, any person skilled in the art can easily think of changes or replacements within the technical range disclosed by the present application, which should be covered within the protection scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A deep network-based ISAR coherent translation compensation and imaging method, characterized by: include: Perform pulse compression on the ISAR echo to obtain multiple pulse compressed echo signals; sequentially inputting the multiple pulse-compressed echo signals into a trained LSTM strategy network so that the trained LSTM strategy network outputs multiple first distance grid offset estimates, wherein the trained LSTM strategy network is a network obtained by training the LSTM strategy network using a training set, wherein the training set is a data set having the same signal form as the multiple pulse-compressed echo signals; Obtaining target translation parameters according to the plurality of first range grid offset estimation values ​​and the resolution of the range dimension; The target translation parameters are estimated using the Adam optimizer to obtain the optimal target translation parameters; According to the optimal target translation parameters, compensation and coherent accumulation are performed on the echo signals after the plurality of pulse compressions to obtain a focused image.

2. The ISAR coherent translation compensation and imaging method based on deep network according to claim 1, characterized in that: The pulse compressing the ISAR echo to obtain a plurality of pulse-compressed echo signals includes: Dividing the ISAR echo into a plurality of sub-apertures; performing range compression processing on the multiple sub-apertures to obtain corresponding one-dimensional range images, and transforming the one-dimensional range images into a frequency domain to obtain multiple frequency domain signals; constructing compensation factors for residual video items and envelope ramp items corresponding to the plurality of frequency domain signals according to a pulse compression method; Phase compensation is performed on the peak frequency of each target scattering point in the multiple frequency domain signals using the compensation factors of the residual video item and the envelope ramp item to obtain the multiple pulse compressed echo signals.

3. The ISAR coherent translation compensation and imaging method based on deep network according to claim 1, characterized in that: The method of training the LSTM strategy network using the training set includes: Using the predicted distance parameters in the training set as the initial state of the LSTM strategy network, the predicted distance parameters are parameters including predicted speed, predicted acceleration, and predicted jerk; Feedback the training action of the LSTM strategy network based on the initial state; Acquiring an actual state according to actual distance parameters of the echo signals after the plurality of pulse compressions; Calculating the difference loss between the initial state and the actual state; According to the training action, the gap loss and the learning rate, the initial state is updated, and the step of feeding back the training action of the LSTM policy network according to the initial state is returned until the total number of training times reaches a preset total number of times, so as to obtain the trained LSTM policy network.

4. The ISAR coherent translation compensation and imaging method based on deep network according to claim 1, characterized in that: The acquiring target translation parameters according to the plurality of first range grid offset estimation values ​​and the resolution of the range dimension includes: multiplying the plurality of first distance grid offset estimation values ​​by the resolution of the distance dimension to obtain a plurality of second distance grid offset estimation values; Fitting the plurality of second distance grid offset estimation values ​​to obtain a corresponding fitting curve; The target translation parameters are obtained through the fitting curve.

5. The ISAR coherent translation compensation and imaging method based on deep network according to claim 1, characterized in that: The method of estimating the target translation parameters using the Adam optimizer to obtain the optimal target translation parameters includes: performing translation compensation and imaging on the echo signals after the plurality of pulse compressions according to the target translation parameters to obtain an imaging result of the current round; Determining a corresponding loss function for the current round according to the imaging result of the current round; Solve the target translation parameters for the next round using the Adam optimizer according to the target translation parameters and the loss function of the current round; Determine whether the difference between the loss function of the current round and the loss function of the previous round is less than a preset value. If so, the loss function of the current round converges, and the target translation parameters of the next round are determined as the optimal target translation parameters. If not, the target translation parameters of the next round are used as the target translation parameters, and return to the step of performing translation compensation and imaging on the echo signals after compression of the multiple pulses according to the target translation parameters to obtain the imaging results of the current round, and stop iteration until the difference between the loss function of the next round and the loss function of the current round is less than the preset value.

6. The ISAR coherent translation compensation and imaging method based on deep network according to claim 5, characterized in that: The step of performing translation compensation and imaging on the echo signals after the plurality of pulse compressions according to the target translation parameters to obtain the imaging results of the current round includes: Using the target translation parameters as learning parameters of an imaging network to obtain imaging parameter estimates; constructing a first translation compensation term corresponding to the echo signals after the plurality of pulse compressions according to the imaging parameter estimation value; compensating the plurality of pulse-compressed echo signals according to the first translation compensation term to obtain a plurality of compensated echo signals; Coherently accumulate the multiple compensated echo signals to obtain the imaging result of the current round.

7. The ISAR coherent translation compensation and imaging method based on deep network according to claim 5, characterized in that: The expression of the loss function of the current round is: Wherein, Entropy(·) is the loss function of the current round, is the average energy value of the imaging result of the current round, I(y,x) is the imaging result of the current round, X is the image azimuth length, Y is the image distance length, x is the image azimuth index, and y is the image distance index.

8. The ISAR coherent translation compensation and imaging method based on deep network according to claim 5, characterized in that: The step of solving the target translation parameters for the next round using the Adam optimizer according to the target translation parameters and the loss function of the current round includes: The target translation parameters are used as optimization parameters for the current round, and the first-order moment and second-order moment of the Adam optimizer are initialized; Calculating the gradient of the loss function of the current round according to the loss function of the current round and the optimization parameters of the current round; updating the initialized first-order moment according to the gradient, the initialized first-order moment, and the first decay rate to obtain an updated first-order moment; updating the initialized second-order moment according to the gradient, the initialized second-order moment, and the second decay rate to obtain an updated second-order moment; performing deviation correction on the updated first-order moment and the updated second-order moment to obtain a corrected first-order moment and a corrected second-order moment; The optimization parameters for the next round are obtained according to the optimization parameters of the current round, the corrected first-order moment, the corrected second-order moment, and the learning rate. The optimization parameters for the next round are the target translation parameters for the next round.

9. The ISAR coherent translation compensation and imaging method based on deep network according to claim 1, characterized in that: The target translation parameters include target velocity, target acceleration and target jerk.

10. An ISAR coherent translation compensation and imaging device based on a deep network, characterized in that: include: A pulse compression module is used to perform pulse compression on the ISAR echo to obtain multiple pulse-compressed echo signals; an input module, configured to sequentially input the plurality of pulse-compressed echo signals into a trained LSTM strategy network, so that the trained LSTM strategy network outputs a plurality of first distance grid offset estimates, wherein the trained LSTM strategy network is a network obtained by training the LSTM strategy network using a training set, wherein the training set is a data set having the same signal form as the plurality of pulse-compressed echo signals; an acquisition module, configured to acquire target translation parameters according to the plurality of first range grid offset estimation values ​​and the resolution of the range dimension; An estimation module, configured to estimate the target translation parameters using an Adam optimizer to obtain optimal target translation parameters; The compensation and coherent accumulation module is used to compensate and coherently accumulate the echo signals after the multiple pulse compressions according to the optimal target translation parameters to obtain a focused image.

Images