Optimization Method, Device, Equipment and Storage Medium for Radar Active Interference Recognition Model

Abstract

The present application relates to the technical field of radar signal processing, and in particular to a method, device, equipment and storage medium for optimizing a radar active interference recognition model. The above method is based on a data set. When the initial model does not meet the requirements of preset technical indicators, the network convolution kernel parameters in the initial model are optimized to obtain target parameters; based on the target parameters, the initial model parameters are replaced to obtain a model to be evaluated; when the model to be evaluated does not meet the requirements of preset hardware configuration, the weight channels of the model to be evaluated are measured and sorted, and then network pruning is performed to obtain a target model. The present application expands the receptive field of the target model by increasing the convolution kernel size in the initial model, increases the order of magnitude of model parameters, improves model performance, and controls the increase in model parameters. The radar active interference signal can be recognized through the target model, and the purpose of processing the radar active interference signal to obtain data is realized.

Description

Technical Field

[0001] The present application relates to the technical field of radar signal processing, and more particularly, to a method, device, equipment, and storage medium for optimizing a radar active interference recognition model. Background Art

[0002] With the rapid development of technologies such as solid-state circuits, very large scale integrated circuits (VLSIs), and digital radio frequency storage (DRFMs), active deceptive jamming exhibits characteristics such as high fidelity and high intelligence, posing a huge threat to the survival of radar systems. Therefore, the identification of active interference types is crucial for ensuring the performance and survival ability of radars in complex electromagnetic environments.

[0003] Currently, discriminative features and invariant features can be learned from data through a convolutional neural network module. However, in radar operating scenarios, especially when facing electronic countermeasures, it is very difficult to obtain data. Therefore, how to process radar active interference signals to obtain data is an urgent problem to be solved. Summary of the Invention

[0004] To solve the problem of how to process radar active interference signals in radar operating scenarios, the present application provides a method, device, equipment, and storage medium for optimizing a radar active interference recognition model.

[0005] The embodiments of the present application are implemented as follows:

[0006] The embodiments of the present application provide a method for optimizing a radar active interference recognition model, including:

[0007] Evaluating a preset initial model based on a data set; when the initial model does not meet the requirements of preset technical indicators, optimizing the network convolution kernel parameters in the initial model to obtain target parameters;

[0008] Replacing the initial model parameters with the target parameters to obtain a model to be evaluated;

[0009] Evaluating the model to be evaluated; when the model to be evaluated does not meet the requirements of preset hardware configuration, measuring and sorting the weight channels of the model to be evaluated, and then performing network pruning to obtain a target model.

[0010] In the above technical solution, when the initial model does not meet the requirements of preset technical indicators, optimizing the network convolution kernel parameters in the initial model to obtain target parameters includes:

[0011] Optimize the network convolution kernel parameters in the initial model through neural network architecture search and network pruning to obtain the target parameters.

[0012] In the above technical solution, the optimization of the network convolution kernel parameters in the initial model through neural network architecture search and network pruning includes:

[0013] Search for the network convolution kernel parameters in the initial model through neural network architecture search;

[0014] Screen the searched network convolution kernel parameters in the initial model through network pruning to obtain the target parameters.

[0015] In the above technical solution, the replacement of the initial model parameters based on the target parameters to obtain the model to be evaluated includes:

[0016] Replace the initial model parameters with the target parameters through convolution to obtain the model to be evaluated.

[0017] In the above technical solution, the method further includes:

[0018] Construct a radar target echo model and an active interference signal model, and obtain the radar active interference signal;

[0019] Obtain the data set based on the radar active interference signal.

[0020] In the above technical solution, the radar target echo model is:

[0021]

[0022]

[0023] where Rect A [] is the waveform gate function, A r is the echo signal amplitude, t is the time when the target echo is received, P(t) is the 0 and 1 states in the pseudo-code, τ A is the modulation pulse width, τ R is the missile-target distance time delay, ω0 is the carrier angular frequency, ω d is the Doppler angular frequency.

[0024] In the above technical solution, the obtaining of the data set based on the radar active interference signal includes:

[0025] Preprocess the radar active interference signal through time-frequency analysis;

[0026] Extract the features of the preprocessed radar active interference signal through the pseudo-Wigner transform model to obtain the data set.

[0027] An embodiment of the present application further provides a device for optimizing a radar active interference recognition model, including:

[0028] An evaluation module, configured to evaluate a preset initial model based on a data set; and also configured to evaluate the model to be evaluated;

[0029] An optimization module, configured to optimize the network convolution kernel parameters in the initial model to obtain target parameters when the initial model does not meet the requirements of preset technical indicators;

[0030] A replacement module, configured to replace the initial model parameters based on the target parameters to obtain a model to be evaluated;

[0031] A network pruning module, configured to measure and sort the weight channels of the model to be evaluated and then perform network pruning to obtain a target model when the model to be evaluated does not meet the requirements of preset technical indicators.

[0032] An embodiment of the present application further provides a storage medium, in which at least one instruction is stored, and the instruction is loaded and executed by a processor to implement the radar active interference recognition model optimization method according to any one of the embodiments of the present application.

[0033] An embodiment of the present application further provides a device, the device includes a processor and a memory, and at least one instruction is stored in the memory, and the instruction is loaded and executed by the processor to implement the radar active interference recognition model optimization method according to any one of the embodiments of the present application.

[0034] An embodiment of the present application provides a method for optimizing a radar active interference recognition model. By evaluating a preset initial model based on a data set; when the initial model does not meet the requirements of preset hardware configuration, optimizing the network convolution kernel parameters in the initial model to obtain target parameters; replacing the initial model parameters based on the target parameters to obtain a model to be evaluated; evaluating the model to be evaluated, and when the model to be evaluated does not meet the requirements of preset technical indicators, measuring and sorting the weight channels of the model to be evaluated and then performing network pruning to obtain a target model. The present application optimizes and replaces the convolution kernel parameters in the initial model to increase the convolution kernel size in the initial model, so that the receptive field of the target model is expanded, the order of magnitude of the model parameters is increased, the purpose of improving the model performance and effectively controlling the increase of the model parameters is achieved, and thus the radar active interference signal can be recognized through the target model, and the purpose of processing the radar active interference signal to obtain data is realized. Description of the Drawings

[0035] To more clearly illustrate the technical solutions in the present application or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0036] Figure 1 One of the flow diagrams of the radar active interference recognition model optimization method provided in the embodiments of the present application;

[0037] Figure 2 Another flow diagram of the radar active interference recognition model optimization method provided in the embodiments of the present application;

[0038] Figure 3 Another flow diagram of the radar active interference recognition model optimization method provided in the embodiments of the present application;

[0039] Figure 4 The structural diagram of the radar active interference recognition model optimization device provided in the embodiments of the present application;

[0040] Figure 5 The entity structural diagram of a device provided in the embodiments of the present application. Detailed implementation manners

[0041] To make the objectives, technical solutions and advantages of the present application clearer, the following will clearly and completely describe the technical solutions in the present application with reference to the drawings in the present application. Obviously, the described embodiments are some, but not all, of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art without creative efforts based on the embodiments in the present application belong to the scope of protection of the present application.

[0042] Figure 1 One of the flow diagrams of the radar active interference recognition model optimization method provided in the embodiments of the present application, as Figure 1 shown, a radar active interference recognition model optimization method includes:

[0043] S101. Evaluate a preset initial model based on a data set; when the initial model does not meet the requirements of the preset technical indicators, optimize the network convolution kernel parameters in the initial model to obtain target parameters;

[0044] S102. Replace the initial model parameters based on the target parameters to obtain a model to be evaluated;

[0045] S103. Evaluate the model to be evaluated. When the model to be evaluated does not meet the preset hardware configuration requirements, measure and sort the weight channels of the model to be evaluated, and then perform network pruning to obtain the target model.

[0046] In some embodiments of the present application, when the initial model does not meet the preset technical index requirements in S101, optimize the network convolution kernel parameters in the initial model to obtain the target parameters, including:

[0047] Optimize the network convolution kernel parameters in the initial model through neural network architecture search and network pruning to obtain the target parameters.

[0048] In some embodiments of the present application, optimize the network convolution kernel parameters in the initial model through neural network architecture search and network pruning to obtain the target parameters, including:

[0049] S1011. Search for the network convolution kernel parameters in the initial model through neural network architecture search;

[0050] S1012. Screen out the network convolution kernel parameters in the searched initial model through network pruning to obtain the target parameters.

[0051] In some embodiments of the present application, S102 further includes:

[0052] Replace the initial model parameters with the target parameters through convolution to obtain the model to be evaluated.

[0053] It can be understood that in the embodiments of the present application, by optimizing and replacing the convolution kernel parameters in the initial model, the convolution kernel size in the initial model is increased, so that the receptive field of the target model is expanded, the order of magnitude of the model parameters is increased, the purpose of improving the model performance and effectively controlling the increase of the model parameters is achieved. In this way, the radar active interference signal can be identified through the target model, and the purpose of processing the radar active interference signal to obtain data can be realized.

[0054] Figure 2 This is the second flow diagram of the radar active interference recognition model optimization method provided in the embodiments of the present application. As Figure 2 shown, in some embodiments of the present application, the method further includes:

[0055] S104. Construct a radar target echo model and an active interference signal model to obtain a radar active interference signal;

[0056] In some embodiments of the present application, the radar target echo model is shown in Formula 1-1:

[0057]

[0058] Among them, Rect A [] is a waveform gate function, A r is the amplitude of the echo signal, t is the time when the target echo is received, P(t) is the state of 0 and 1 in the pseudo-code, τ A is the modulation pulse width, τ R is the time delay of the missile-target distance, ω0 is the carrier angular frequency, ω d is the Doppler angular frequency.

[0059] In some embodiments of the present application, let the random code element width T c = 50ns, the code length P = 31, so the pulse width T = 1us, the pulse repetition period T R = 5us, the carrier frequency f0 = 220MHZ, the missile-target distance R t = 60m, and the signal-to-noise ratio is 10dB. When there is no interference signal, the radar target echo signal is processed to obtain an intermediate frequency signal, which is output after correlation with the local oscillator signal.

[0060] S105. Obtain a data set based on the radar active interference signal.

[0061] In some embodiments of the present application, the active interference signal model includes at least one of noise amplitude modulation interference, noise frequency modulation interference, noise phase modulation interference, intermittent sampling direct retransmission interference, intermittent sampling repeated retransmission interference, and range deception interference.

[0062] In some embodiments of the present application, the noise amplitude modulation interference is as shown in Formula 1-2:

[0063] J A (t) = [A0 + n(t)]cosω j t 1-2

[0064] Among them, A0 is the carrier amplitude; t is the time when the target echo is received; ω j is the carrier angular frequency; n(t) is a generalized stationary noise with a mean of 0 and a variance of σ 2 .

[0065] In some embodiments of the present application, the noise frequency modulation interference is as shown in Formula 1-3:

[0066]

[0067] Among them, U j is the amplitude of the modulation signal, t is the time when the target echo is received, f j is the carrier frequency of the modulation signal, K FM is the frequency modulation slope of the noise frequency modulation signal, u n (t) is a generalized stationary random process with a mean of zero and a variance of , that is, modulation noise, is the initial phase uniformly distributed on [0, 2π], and is independent of u n (t) satisfies the independent distribution relationship.

[0068] In some embodiments of the present application, the noise phase modulation interference is shown in Formula 1-4:

[0069]

[0070] where, U j is the amplitude of the modulation signal, t is the time when the target echo is received, f j is the carrier frequency of the modulation signal, is the phase modulation slope of the noise phase modulation signal, u n (t) represents the modulation noise, that is, a generalized stationary random process with a mean of zero and a variance of and is the initial phase uniformly distributed on [0, 2π], and is independent of u n (t) satisfies the independent distribution relationship.

[0071] In some embodiments of the present application, the intermittent sampling direct retransmission interference is shown in Formula 1-5:

[0072]

[0073] where, Rect A [] is the waveform gate function, A j is the amplitude of the interference signal received by the receiver, t is the time when the target echo is received, P(t) is the pseudo-random code signal identical to the target signal, τ j is the total interference delay (including the inherent delay of the jammer and the interference delay set by the interfering party), τ A is the modulation pulse width, ω0 is the carrier angular frequency.

[0074] In some embodiments of the present application, the intermittent sampling repeat retransmission interference is shown in Formula 1-6:

[0075]

[0076] where, rect() is the waveform gate function, t is the time when the target echo is received, N is the number of interference slices, T J is the interference slice width, K r is the modulation frequency of the transmitted signal, τ is the delay introduced by the distance from the jammer to the radar, and N is the number of interference slices.

[0077] In some embodiments of the present application, the range deception interference is shown in Formula 1-7:

[0078]

[0079] Among them, rect() is the waveform gate function, t is the time when the target echo is received, τ is the delay introduced by the distance from the jammer to the radar, and K r is the frequency modulation rate of the transmitted signal, M is the number of times each slice is forwarded, N is the number of interference slices, and T n =(M + 1)·T J is the time interval for the jammer to intercept the signal.

[0080] In some embodiments of the present application, the parameters of noise amplitude modulation interference, noise frequency modulation interference, noise phase modulation interference, intermittent sampling direct retransmission interference, intermittent sampling repeated retransmission interference, and range deception interference can be set as shown in Table 1-1.

[0081] Table 1-1

[0082]

[0083]

[0084] In some embodiments of the present application, S105 includes:

[0085] S1051. Preprocess the radar active interference signal through time-frequency analysis;

[0086] S1052. Extract the features of the preprocessed radar active interference signal through the pseudo-Wigner transform model to obtain a data set.

[0087] In some embodiments of the present application, S104 and S105 can be implemented using MATLAB 2021b.

[0088] It can be understood that in the embodiments of the present application, by preprocessing the radar active interference signal based on the transform domain through time-frequency analysis, high-dimensional features in the radar active interference signal can be extracted, and the time-frequency aggregation of the pseudo-Wigner transform model is strong and the cross-term suppression effect is good. In this way, the obtained data set can better represent signals of different modulation types, and the target model obtained based on the data set can better identify the radar active interference signal, achieving the purpose of processing the radar active interference signal to obtain data.

[0089] Figure 3 This is the third flow chart of the radar active interference recognition model optimization method provided in the embodiments of the present application. As Figure 3 shown, the embodiments of the present application also provide a radar active interference recognition model optimization method, including:

[0090] S201. Input the radar active interference signal data set (data set) into the initial model;

[0091] S202. Evaluate the initial model;

[0092] S203. Determine whether the initial model meets the technical indicators; if it meets, proceed to S204; if it does not meet, proceed to S205;

[0093] S204. Determine whether the initial model meets the hardware configuration; if it meets, obtain the target model; if it does not meet, proceed to S209;

[0094] S205. Optimize the network convolution kernel parameters in the initial model through neural network architecture search and network pruning, and use convolution to replace the model parameters to obtain the first model (model to be evaluated);

[0095] S206. Evaluate the first model;

[0096] S207. Determine whether the first model meets the technical indicators; if it meets, proceed to S208; if it does not meet, proceed to S205;

[0097] S208. Determine whether the first model meets the hardware configuration; if it meets, obtain the target model; if it does not meet, proceed to S209;

[0098] S209. Measure and sort the weight channels of the model to be evaluated, and perform network pruning according to the requirements and constraints to obtain the second model (model to be evaluated);

[0099] S210. Evaluate the second evaluated model;

[0100] S211. Determine whether the second model to be evaluated meets the hardware configuration; if it meets, obtain the target model; if it does not meet, proceed to S209.

[0101] Figure 4 It is a schematic structural diagram of a radar active interference recognition model optimization device provided in an embodiment of the present application. As Figure 4 shown, an embodiment of the present application also provides a radar active interference recognition model optimization device, including:

[0102] An evaluation module 41, configured to evaluate a preset initial model based on a data set; and also configured to evaluate the model to be evaluated;

[0103] An optimization module 42, configured to optimize the network convolution kernel parameters in the initial model to obtain target parameters when the initial model does not meet the preset technical indicator requirements;

[0104] A replacement module 43, configured to replace the initial model parameters based on the target parameters to obtain a model to be evaluated;

[0105] A network pruning module 44, configured to measure and sort weight channels of the to-be-evaluated model when the to-be-evaluated model does not meet the requirements of preset technical indicators, and then perform network pruning to obtain a target model.

[0106] In some embodiments of the present application, the optimization module 42 is further configured to optimize network convolution kernel parameters in the initial model through neural network architecture search and network pruning to obtain the target parameters.

[0107] In some embodiments of the present application, the optimization module 42 is further configured to search for network convolution kernel parameters in the initial model through neural network architecture search; and is further configured to screen out the network convolution kernel parameters in the searched initial model through network pruning to obtain the target parameters.

[0108] In some embodiments of the present application, the replacement module 43 is further configured to replace the target parameters with the initial model parameters through convolution to obtain a to-be-evaluated model.

[0109] In some embodiments of the present application, the device further includes a model construction module 45 and an acquisition module 46; wherein,

[0110] The model construction module 45 is further configured to construct a radar target echo model and an active interference signal model, and acquire a radar active interference signal;

[0111] The acquisition module 46 is further configured to acquire the data set based on the radar active interference signal.

[0112] In some embodiments of the present application, the acquisition module 46 is further configured to preprocess the radar active interference signal through time-frequency analysis; and is further configured to extract features from the preprocessed radar active interference signal through a pseudo-Wigner transform model to obtain the data set.

[0113] The present application embodiment further provides a storage medium, including at least one instruction stored in the storage medium, and the instruction is loaded and executed by a processor to implement the radar active interference recognition model optimization method described in any of the above embodiments.

[0114] Figure 5 This is a schematic physical structure diagram of a device provided by an embodiment of the present application, as Figure 5As shown, the electronic device may include: a processor 810, a communications interface 820, a memory 830, and a communication bus 840. Among them, the processor 810, the communications interface 820, and the memory 830 communicate with each other through the communication bus 840. The processor 810 may call the logical instructions in the memory 830 to execute the radar active interference recognition model optimization method described in any of the above embodiments.

[0115] In addition, when the logical instructions in the above-mentioned memory 830 are implemented in the form of software functional units and sold or used as independent products, they may be stored in a computer-readable storage medium. Based on such an understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of this technical solution, may be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in various embodiments of this application. The foregoing storage medium includes: various media such as USB flash drives, mobile hard disks, read-only memories (ROMs), random access memories (RAMs), magnetic disks, or optical discs that can store program codes.

[0116] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place, or may be distributed to multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art can understand and implement it without creative labor.

[0117] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, and of course, it can also be implemented by hardware. Based on such an understanding, the technical solution, in essence, or the part that contributes to the prior art, may be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disc, etc., and includes several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments.

[0118] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than limiting them; although the present application has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some of the technical features; and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the various embodiments of the present application.

Claims

1. An optimization method for a radar active interference recognition model, characterized in that, Including: Construct the radar target echo model and the active interference signal model, obtain the radar active interference signal, and based on the radar active interference signal, obtain a data set; Evaluate a preset initial model based on the data set; in the case that the initial model does not meet the requirements of the preset technical indicators, optimize the network convolution kernel parameters in the initial model to obtain target parameters; Based on the target parameters, replace the initial model parameters to obtain a model to be evaluated; Evaluate the model to be evaluated; in the case that the model to be evaluated does not meet the requirements of the preset hardware configuration, measure and sort the weight channels of the model to be evaluated, and then perform network pruning to obtain a target model; Wherein, the radar target echo model is: ； Among them, is a waveform gate function, , where \(t\) is the time when the target echo is received, is the 0 and 1 states in the pseudo-code, is the modulation pulse width, is the time delay of the missile-target distance, is the carrier angular frequency, is the Doppler angular frequency.

2. The method for optimizing the radar active interference recognition model according to claim 1, wherein In the case that the initial model does not meet the requirements of the preset technical indicators, optimizing the network convolution kernel parameters in the initial model to obtain target parameters includes: Optimizing the network convolution kernel parameters in the initial model through neural network architecture search and network pruning to obtain the target parameters.

3. The method for optimizing the radar active interference recognition model according to claim 2, wherein The optimizing the network convolution kernel parameters in the initial model through neural network architecture search and network pruning includes: Search for the network convolution kernel parameters in the initial model through neural network architecture search; Screen out the network convolution kernel parameters in the initial model found through network pruning to obtain the target parameters.

4. The method for optimizing the radar active interference recognition model according to claim 1, wherein The replacing the initial model parameters based on the target parameters to obtain a model to be evaluated includes: Replace the initial model parameters with the target parameters through convolution to obtain a model to be evaluated.

5. The method for optimizing the radar active interference recognition model according to claim 1, wherein The obtaining a data set based on the radar active interference signal includes: Preprocess the radar active interference signal through time-frequency analysis; Extract features from the preprocessed radar active interference signal through the pseudo-Wigner transform model to obtain the data set.

6. An optimization device for a radar active interference recognition model, characterized in that Including: An evaluation module, configured to construct the radar target echo model and the active interference signal model, obtain the radar active interference signal, and based on the radar active interference signal, obtain a data set; evaluate a preset initial model based on the data set; and is also configured to evaluate the model to be evaluated; wherein, the radar target echo model is: ； Among them, is a waveform gate function, , where t is the time when the target echo is received, is the 0 and 1 states in the pseudo-code, is the modulation pulse width, is the time delay of the missile-target distance, is the carrier angular frequency, is the Doppler angular frequency; An optimization module, configured to optimize the network convolution kernel parameters in the initial model to obtain target parameters in the case that the initial model does not meet the requirements of the preset technical indicators; A replacement module, configured to replace the initial model parameters based on the target parameters to obtain a model to be evaluated; A network pruning module, configured to measure and sort the weight channels of the model to be evaluated in the case that the model to be evaluated does not meet the requirements of the preset technical indicators, and then perform network pruning to obtain a target model.

7. A storage medium, characterized in that, At least one instruction is stored in the storage medium, and the instruction is loaded and executed by a processor to implement the method for optimizing a radar active interference recognition model according to any one of claims 1-5.

8. A device, characterized in that, The device includes a processor and a memory, and at least one instruction is stored in the memory. The instruction is loaded and executed by the processor to implement the method for optimizing the radar active interference recognition model according to any one of claims 1-5.

Images