CN115932770A - Method, system, equipment and terminal for precise and intelligent identification of radar radiation source individuals - Google Patents

Abstract

The invention belongs to the technical field of individual identification of radiation sources in accurate radar identification, and discloses an accurate intelligent identification method, system, equipment and terminal of individual radar radiation sources, which are used for solving a corresponding bispectrum for a received radar radiation source signal; extracting the characteristics of the bispectrum according to a Laplace-Gaussian operator; inputting the extracted features into a deep residual error network based on norm and dynamic learning rate for training to obtain a trained model; and realizing individual intelligent identification of radar radiation source signals by using the trained model. The accurate and intelligent identification method for the radar radiation source individuals realizes individual identification of the radiation source under the condition that the differences of the radar fingerprint characteristics are not obvious, improves the individual identification efficiency of the radiation source under the condition of ensuring the accuracy of the individual identification efficiency, improves the generalization and the robustness of a network model and the accuracy in use, effectively and deeply mines the difference characteristics among the individual radiation sources, and can achieve better identification effect under the condition that the radar fingerprint characteristics are similar.

Description

Method, system, equipment and terminal for precise and intelligent identification of radar radiation source individuals

technical field

The invention belongs to the technical field of individual radiation source identification in radar precise identification, and in particular relates to a precise and intelligent identification method, system, equipment and terminal for radar radiation source individuals.

Background technique

At present, the identification of radar emitters means that after a series of analysis and processing of the received individual signals of radar emitters, it is judged which step the received individual signals come from, the radar emitter and its position and parameters. The early radar was relatively simple, and the radar emitter identification was mostly processed and analyzed by artificially processing and analyzing the pulse-to-pulse parameters obtained from the measurement of the radar emitter, and then compared and analyzed with the individual information of each radar emitter in the known established database. However, in recent years, radar technology has advanced by leaps and bounds, and various new radars have emerged in an endless stream. The electromagnetic environment of the battlefield has become more complicated. The fingerprint information of each radar radiation source has a high degree of similarity, which makes it difficult to distinguish the radar signals intercepted by the radar receiving device. Therefore, finding a more accurate and quick identification method has become an urgent problem and the main development direction in the field of radar radiation sources.

The radiation source individual identification technology based on fingerprint information has started in the last century. The American scholar Boorstyn proposed the theory of "Specific Emitter Identification, SEI", which uses the intercepted specific radar emitter signal to extract the fingerprint characteristics of the signal, compares and analyzes it with the previously established information database, and classifies it according to the matching results. The radar radiation source that emits this signal is identified, but this method relies on expert judgment, and the human influence factor is large, which is prone to misjudgment. At the same time, the existing individual identification of radar emitters relies more on the difference between radar fingerprint information. If the difference is small, the identification result will be seriously affected. Based on this, a method of extracting fingerprint information by using the constellation diagram is proposed abroad, and input into the convolutional neural network for identification; ) and radar frequency (RF) and other statistics as the basis for classification and identification, the input network adopts naive Bayesian classifier, clustering, SVM and other methods; based on kernel principal component analysis (KPCA) predictive learning method; based on a method using Frechet The method of calculating the distance between signals, the pulse envelope or the instantaneous frequency based on the distance; based on the bispectrum + SURF (Speed-uprobust features) feature. However, although the accuracy rate is high by using some fingerprint features, the calculation amount is huge, and the accuracy rate of fingerprint features with a small amount of calculation is often low.

For China, there are methods based on phase observation model + long-short-term memory network, precise intelligent identification method of radar radiation source individual based on VMD decomposition, method of constructing data set based on pulse data stream, individual identification method based on fuzzy function, based on A precise intelligent identification method for radar emitter individuals based on Hilbert transform for feature extraction, and a classification method for emitter individuals based on deep belief network DBN and emitter signal envelope characteristics. However, some network models are overfitting to the training set, resulting in high accuracy during training but unsatisfactory results during testing. Therefore, it is urgent to find a fast and accurate radar fingerprint feature and highlight it to amplify its differences; at the same time, improve the network model to reduce its dependence on the training set and prevent its overfitting.

Through the above analysis, the problems and defects in the prior art are:

(1) With the development of radar technology, the complexity of signals and the complexity of electromagnetic environment are increasing day by day, making it difficult for inter-pulse parameters to meet the requirements of recognition accuracy, and it is necessary to use intra-pulse modulation information. However, the existing individual identification of radar emitters using intrapulse information is more dependent on the difference between the fingerprint information of radar emitters. If the difference between the fingerprint information of emitters to be identified is small, the recognition accuracy will be seriously reduced. .

(2) The existing individual radiation source identification technology based on fingerprint information relies on expert judgment, and the error in the recognition rate is relatively large when using artificial intelligence. It is necessary to use artificial intelligence methods to mine deeper abstract features for identification.

(3) When the existing artificial intelligence method is used to identify individual radar emitters, the shallow network often does not get a good recognition effect, while the deep network will greatly increase the time consumption, and there are problems of overfitting and network degradation. question.

Contents of the invention

Aiming at the problems existing in the prior art, the present invention provides a precise and intelligent identification method, system, device and terminal of a radar radiation source individual, and particularly relates to a precise and intelligent identification method, system and medium of a radar radiation source individual when the fingerprint features are similar , equipment and terminals.

The present invention is achieved in this way, a method for precise and intelligent identification of radar radiation source individuals. The precise and intelligent identification method for radar radiation source individuals includes: obtaining the bispectrum corresponding to the radar radiation source signal and performing Laplacian-Gauss calculation Sub-feature extraction; use the extracted features to build a trained model, and use the trained model to realize the precise and intelligent identification of individual radar emitters.

Further, the method for precise and intelligent identification of radar radiation source individuals includes the following steps:

Step 1, obtain the corresponding bispectrum for the received radar emitter signal, and mine the deep features of the radar emitter signal;

Step 2, extracting the features of the bispectrum according to the Laplacian-Gaussian operator, further highlighting the edge difference information between individual radar emitters;

Step 3, inputting the extracted features into a deep residual network based on norm and dynamic learning rate for training, to obtain a trained model, and to achieve an ideal recognition effect with less consumption;

Step 4, using the trained model to realize individual intelligent recognition of radar emitter signals.

Further, obtaining the corresponding bispectrum for the received radar emitter signal in the step 1 includes:

The third-order cumulant is calculated for the radar emitter signal x(t), and the expression is as follows:

C _3s (τ ₁ ,τ ₂ )=E{s ^* (t)x(t+τ ₁ )x(t+τ ₂ )};

Among them, x is the received signal, τ is the time delay, s ^* (t) is the conjugate signal; E is the mathematical expectation for calculating the corresponding value, and the obtained C _3s (τ ₁ , τ ₂ ) is the corresponding third-order cumulant.

The bispectral transformation is calculated for the third-order cumulant C _3s (τ ₁ , τ ₂ ), the expression is as follows:

为求得三阶累积量的二维傅里叶变换，所得B_s(ω₁，ω₂)为双谱。Among them, C _3s (τ ₁ , τ ₂ ) is the obtained corresponding third-order cumulant,

In order to obtain the two-dimensional Fourier transform of the third-order cumulant, the obtained B _s (ω ₁ , ω ₂ ) is a bispectrum.

Further, performing feature extraction of the bispectrum according to the Laplacian-Gaussian operator in the step 2 includes:

The Laplacian-Gaussian operator with 0 as the center and σ as the Gaussian standard deviation is expressed as follows:

Among them, G _σ (x, y) is a second-order Gaussian function, the expression is as follows:

Among them, σ is the Gaussian standard deviation.

Then perform Laplacian-Gaussian feature extraction on the bispectrum:

LoGB _s (ω ₁ , ω ₂ ) = LoG*B _s (ω ₁ , ω ₂ );

Among them, * represents the convolution operation, and the operator is used to perform the convolution operation on the bispectrum.

Further, in the step 3, the extracted features are input into the deep residual network based on norm and dynamic learning rate for training, and the trained model includes:

When the network calculates the parameters of each layer, the L2 norm is introduced, and the square root of each feature vector element is calculated to make the network more sparse and smooth.

||x|| ₂ = (|x ₁ | ² +|x ₁ | ² +|x ₁ | ² +...+|x _n | ² ) ^1/2 ;

The dynamic learning rate used in each round of learning is as follows:

lr(n)=lr*0.2 ^[n/10] ;

Among them, lr(n) is the current learning rate, n is the batch, and lr is the preset learning rate. After every 10 batches, the learning rate becomes 0.2 times the current one, and the learning rate gradually decreases with the round change.

The deep residual network is composed of various residual blocks, and the fast learning rule of each residual is:

F(x)=H(x)-x;

Among them, H(x) is the backpropagation function, F(x) is the forward propagation function, and x is the network input.

Input the bispectrum extracted by the Laplacian-Gaussian operator into a single channel to generate H×W×1 features, where H and W are the length and width of the feature map, and 1 is the number of channels; the features are input into The convolution operation is performed in the residual block 1, and the result is input into the residual block 2 until the operation of 4 residual blocks; finally, the final recognition result is obtained through the pooling layer and the fully connected layer.

Further, using the trained model in the step 4 to realize the individual intelligent identification of the radar emitter signal includes:

After receiving the unclassified preprocessed signal of the trained radar emitter individual, the corresponding bispectrum is obtained, and the bispectrum is extracted according to the Laplacian-Gaussian operator, and the trained network model is loaded; the feature Input it into the network to get the recognition result; through the Softmax layer, the probability matrix of each radar radiation source individual is obtained, and the one with the highest probability is the recognition result. The Softmax expression is as follows:

M=max(z);

Among them, z is the result vector, max is the maximum value of z, z _i is the ith result.

Another object of the present invention is to provide a system for precise and intelligent identification of individual radar radiation sources using the method for precise and intelligent identification of individual radar radiation sources. The system for precise and intelligent identification of individual radar radiation sources includes:

The bispectrum obtaining module is used to obtain the corresponding bispectrum for the received radar emitter signal;

A feature extraction module is used to extract the features of the bispectrum according to the Laplacian-Gaussian operator;

A feature training module, configured to input the extracted features into a deep residual network based on norm and dynamic learning rate for training to obtain a trained model;

The individual identification module is used to realize the intelligent identification of the individual of the radar radiation source signal by using the trained model.

Another object of the present invention is to provide a computer device, the computer device includes a memory and a processor, the memory stores a computer program, when the computer program is executed by the processor, the processor executes the The steps of the method for precise and intelligent identification of radar radiation source individuals are described.

Another object of the present invention is to provide a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the processor executes the steps of the method for precise and intelligent identification of radar radiation source individuals. .

Another object of the present invention is to provide an information data processing terminal, which is used to realize the precise intelligent identification system for individual radar radiation sources.

Combining the above-mentioned technical solutions and technical problems to be solved, the advantages and positive effects of the technical solutions to be protected in the present invention are as follows:

First, aiming at the technical problems existing in the above-mentioned prior art and the difficulty of solving the problems, closely combining the technical solution to be protected in the present invention and the results and data in the research and development process, etc., a detailed and profound analysis of how the technical solution of the present invention is solved Technical problems, some creative technical effects brought about after solving the problems. The specific description is as follows:

In the radar radiation source individual accurate intelligent identification method provided by the present invention, at first obtain its corresponding bispectrum to the received radar radiation source signal; ) to perform feature extraction; finally, input the extracted features into the deep residual network (Resnet) based on norm and dynamic learning rate for training to obtain the trained model, and use this model to realize the individual intelligence of the radar emitter signal identify. The invention effectively digs deeply into the difference characteristics among radiation source individuals, and can achieve better recognition effect under the condition of similar radar fingerprint characteristics.

The invention can effectively realize individual radiation source identification in the case of radiation sources with similar radar fingerprint features, and has better performance compared with other methods. In addition, the radar radiation source individual identification method provided by the present invention is also applicable to the radiation source individual identification when the radar fingerprint features have large differences.

Second, regarding the technical solution as a whole or from the perspective of a product, the technical effects and advantages of the technical solution to be protected by the present invention are specifically described as follows:

The precise and intelligent identification method for radar radiation source individuals provided by the present invention realizes the identification of radiation source individuals in complex electromagnetic environments where the extraction of radar fingerprint features is insufficient and the differences between them are not obvious, while improving the network to ensure its accuracy In the case of high efficiency, the identification efficiency of individual radiation sources is improved, the generalization and robustness of the network model are improved, and the accuracy of its use is improved.

Third, as an auxiliary evidence of the inventiveness of the claims of the present invention, it is also reflected in the following important aspects:

The precise and intelligent identification method for radar radiation source individuals provided by the present invention combines neural network, radar intrapulse signal processing technology and pattern recognition theory enhancement.

(1) The commercial value after the conversion of the technical solution of the present invention is:

The present invention truly applies the radar radiation source individual identification technology to the actual radar radiation source individual signal identification, not only in the identification of theoretical simulation signals.

(2) Whether the technical solution of the present invention solves the technical problem that people have always been eager to solve but have not been able to achieve success:

In today's complex electromagnetic environment, it can effectively further extract the characteristics of radar radiation sources, amplify the differences between radar radiation sources, and provide higher recognition accuracy when the fingerprints of radar radiation sources are similar. At the same time, the improved network prevents the phenomenon of overfitting while ensuring the correct rate of training. There is no need to stack the number of network layers during training, which reduces the cost and maintains a high correct rate during recognition.

Description of drawings

In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the drawings that need to be used in the embodiments of the present invention. Obviously, the drawings described below are only some embodiments of the present invention. For Those of ordinary skill in the art can also obtain other drawings based on these drawings without making creative efforts.

Fig. 1 is a flowchart of a method for precise and intelligent identification of radar radiation source individuals provided by an embodiment of the present invention;

Fig. 2(a) is a schematic diagram of the accuracy rate and loss function of individual radiation source identification using the precise intelligent identification method for individual radar radiation sources provided by an embodiment of the present invention;

Fig. 2(b) is a schematic diagram of the correct rate and loss function of individual radiation source identification using the bispectrum method provided by the embodiment of the present invention.

Detailed ways

In order to make the object, technical solution and advantages of the present invention more clear, the present invention will be further described in detail below in conjunction with the examples. It should be understood that the specific embodiments described here are only used to explain the present invention, not to limit the present invention.

Aiming at the problems existing in the prior art, the present invention provides a precise and intelligent identification method, system, equipment and terminal for individual radar radiation sources. The present invention will be described in detail below with reference to the accompanying drawings.

1. Explain the embodiment. In order to make those skilled in the art fully understand how to implement the present invention, this part is an explanatory embodiment for explaining the technical solution of the claims.

As shown in Figure 1, the precise and intelligent identification method for individual radar emitters provided by the embodiment of the present invention includes the following steps:

S101. Obtain a corresponding bispectrum for the received radar emitter signal;

S102, performing feature extraction on the bispectrum according to the Laplacian-Gaussian operator;

S103, inputting the extracted features into a deep residual network based on a norm and a dynamic learning rate for training to obtain a trained model;

S104, using the trained model to realize individual intelligent identification of radar emitter signals.

As a preferred embodiment, the embodiment of the present invention provides a method for precise and intelligent identification of radar radiation source individuals when the fingerprint features are similar, specifically including the following steps:

The first step is to obtain the corresponding bispectrum of the received radar emitter signal, and extract the features of the bispectrum according to the Laplacian of Gaussian (LoG) operator. The specific implementation process is as follows: :

The third-order cumulant is calculated for the radar emitter signal x(t), and the expression is as follows:

C _3s (τ ₁ ,τ ₂ )=E{s ^* (t)x(t+τ ₁ )x(t+τ ₂ )}

Among them, x is the received signal, τ is the time delay, s ^* (t) is its conjugate signal, E calculates the mathematical expectation of its corresponding value, and the obtained C _3s (τ ₁ , τ ₂ ) is its corresponding third-order cumulant .

The bispectral transformation is calculated for the third-order cumulant C _3s (τ ₁ , τ ₂ ), the expression is as follows:

即为求得三阶累积量的二维傅里叶变换，所得B_s(ω₁，ω₂)即为双谱。Among them, C _3s (τ ₁ , τ ₂ ) is the obtained corresponding third-order cumulant,

That is to obtain the two-dimensional Fourier transform of the third-order cumulant, and the obtained B _s (ω ₁ , ω ₂ ) is the bispectrum.

The Laplacian-Gaussian operator expression with 0 as the center and σ as the Gaussian standard deviation is as follows:

Among them, G _σ (x, y) is a second-order Gaussian function, as follows:

Among them, σ is the Gaussian standard deviation.

Then, the Laplacian-Gaussian operator feature extraction is further performed on the bispectrum as follows:

LoGB _s (ω ₁ , ω ₂ ) = LoG*B _s (ω ₁ , ω ₂ )

Among them, * represents a convolution operation, that is, an operator is used to perform a convolution operation on the bispectrum.

The second step is to input the extracted features into the deep residual network (Resnet) based on norm and dynamic learning rate for training to obtain the trained model. The specific implementation process is as follows:

When the network calculates the parameters of each layer, the L2 norm is introduced, that is, the square sum of the elements of each feature vector is then calculated to make the network more sparse and smooth.

||x|| ₂ ＝(|x ₁ | ² +|x ₁ | ² +|x ₁ | ² +...+|x _n | ² ) ^1/2

The dynamic learning rate used in each round of learning is as follows:

lr(n)=lr*0.2 ^[n/10]

Among them, lr(n) is the current learning rate, n is the batch, and lr is the preset learning rate, that is, the learning rate becomes 0.2 times the current after every 10 batches, that is, the learning rate gradually decreases with the round transformation .

The deep residual network is composed of various residual blocks, and the fast learning rule of each residual is:

F(x)=H(x)-x

Among them, H(x) is the backpropagation function, F(x) is the forward propagation function, and x is the network input.

Input the bispectrum extracted by the Laplacian-Gaussian operator into a single channel to generate H×W×1 features, where H and W are the length and width of the feature map, 1 is the number of channels, and then the feature Input to the residual block 1 for convolution operation, and input the result to the residual block 2 until the operation of 4 residual blocks, and finally through the pooling layer and the fully connected layer to obtain the final recognition result.

The third step is to use the model to realize the individual intelligent identification of the radar emitter signal. The specific implementation process is as follows:

After receiving the unclassified preprocessed signal of the trained radar emitter individual, the corresponding bispectrum is obtained, and the feature extraction of the bispectrum is performed according to the Laplacian of Gaussian operator (Laplacian of gaussain, LoG). , and load the trained network model. Input this feature into the network to obtain the recognition result, and obtain the probability matrix of each radar emitter individual through the Softmax layer. The one with the highest probability is the recognition result. The Softmax expression is as follows:

M=max(z)

Among them, z is the result vector, max is the maximum value of z, z _i is the ith result.

The radar radiation source individual accurate intelligent identification system provided by the embodiment of the present invention includes:

The bispectrum obtaining module is used to obtain the corresponding bispectrum for the received radar emitter signal;

The feature extraction module is used to extract the features of the bispectrum according to the Laplacian-Gaussian operator;

The feature training module is used to input the extracted features into the deep residual network based on norm and dynamic learning rate for training to obtain the trained model;

The individual identification module is used to realize the intelligent identification of the individual of the radar radiation source signal by using the trained model.

2. Application examples. In order to prove the creativity and technical value of the technical solution of the present invention, this part is the application example of the claimed technical solution on specific products or related technologies.

After the present invention intercepts the individual signal of the radar radiation source, the signal is passed into the system after the receiver is preprocessed, and the bispectrum is extracted from the signal through this method, and the further extraction of the feature is carried out with the Gaussian Laplacian operator to form a training set and Validation set. During training, load the improved model, use the training set to get the trained model and use the validation set to verify the effect. After subsequent interception of the individual signal of the radar radiation source, the signal is transmitted to the system after preprocessing by the receiver, and the bispectrum is extracted from the signal by this method, and the feature is further extracted by the Gaussian Laplacian operator, and the trained model is used It is judged which radar emitter individual emits the signal, so as to realize the identification of the radar emitter individual.

3. Evidence of the relevant effects of the embodiment. The embodiment of the present invention has achieved some positive effects in the process of research and development or use, and indeed has great advantages compared with the prior art. The following content is described in conjunction with the data and charts of the test process.

The simulation experiment provided by the embodiment of the present invention uses the signals of 5 different radar radiation source individuals, the channel environment is Gaussian white noise, the signal-to-noise ratio is set to 10dB, and each individual has 1500 sample data for network training, 500 samples The data is used for testing, so the total number of training samples is 7500, and the total number of test set samples is 2500. During training, the control group used bispectral feature extraction for all signals in the sample set, and input the deep residual network for training. During the training process, the SGD optimization method was used, and the loss function was the cross-entropy loss function. The amount of data is set to 32, and a total of 100 training batches are set. In addition, this method is used to extract bispectral features from all signals in the sample set, and the Laplacian-Gaussian operator is used to further extract features, and the improved deep residual network is input for training. The SGD optimization method is used in the training process, and the loss function is the cross-entropy loss function, the sample data size of each batch is set to 32 during the training process, and a total of 100 training batches are set. As shown in Figure 2(a)~(b), the horizontal axis is the training rounds, and the vertical axis is the correct rate and loss function. Through the comparison of the training set and the verification set of the two methods, it can be seen that in the training set and the verification set In summary, compared with the bispectrum method, the accurate intelligent identification method for radar radiation source individuals provided by the embodiment of the present invention has an accuracy rate increased from 67% to 80%, and the loss function is greatly reduced, which proves that the method of the present invention can be used in the case of similar radar fingerprint information. Under the feasibility, better results have been achieved.

It should be noted that the embodiments of the present invention can be realized by hardware, software, or a combination of software and hardware. The hardware part can be implemented using dedicated logic; the software part can be stored in memory and executed by a suitable instruction execution system such as a microprocessor or specially designed hardware. Those of ordinary skill in the art will understand that the above-described devices and methods can be implemented using computer-executable instructions and/or contained in processor control code, for example, on a carrier medium such as a magnetic disk, CD or DVD-ROM, such as a read-only memory Such code is provided on a programmable memory (firmware) or on a data carrier such as an optical or electronic signal carrier. The device and its modules of the present invention may be implemented by hardware circuits such as VLSI or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., It can also be realized by software executed by various types of processors, or by a combination of the above-mentioned hardware circuits and software such as firmware.

The above is only a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Anyone familiar with the technical field within the technical scope disclosed in the present invention, whoever is within the spirit and principles of the present invention Any modifications, equivalent replacements and improvements made within shall fall within the protection scope of the present invention.

Claims (10)

1. The method for accurately and intelligently identifying the radar radiation source individuals is characterized by comprising the following steps of: obtaining a bispectrum corresponding to a radar radiation source signal and extracting features of a Laplace-Gaussian operator; and constructing a trained model by using the extracted features, and realizing accurate and intelligent identification of the radar radiation source individuals by using the trained model.

2. The method for accurately and intelligently identifying individuals with radar radiation sources according to claim 1, wherein the method for accurately and intelligently identifying individuals with radar radiation sources comprises the following steps:

step one, solving a corresponding bispectrum for a received radar radiation source signal;

secondly, extracting the characteristics of the bispectrum according to a Laplace-Gaussian operator;

inputting the extracted features into a deep residual error network based on norm and dynamic learning rate for training to obtain a trained model;

and fourthly, utilizing the trained model to realize the individual intelligent identification of the radar radiation source signal.

3. The method for accurately and intelligently identifying individuals as radar radiation sources according to claim 2, wherein the obtaining of the corresponding bispectrum of the received radar radiation source signals in the first step comprises:

calculating a third-order cumulant of a radar radiation source signal x (t), wherein the expression is as follows:

C _3s (τ ₁ ,τ ₂ )＝E{s ^* (t)x(t+τ ₁ )x(t+τ ₂ )}；

where x is the received signal, τ is the time delay, s ^* (t) is a conjugate signal; e is the mathematical expectation of the corresponding value, resulting in C _3s (τ ₁ ,τ ₂ ) Is the corresponding third order cumulant;

for third order cumulant C _3s (τ ₁ ,τ ₂ ) And (3) solving a bispectrum transformation, wherein the expression is as follows:

wherein, C _3s (τ ₁ ,τ ₂ ) For the found corresponding third order cumulant,

two-dimensional Fourier transform to obtain third-order cumulant, resulting in B _s (ω ₁ ,ω ₂ ) Is a bispectrum.

4. The method for accurately and intelligently identifying individuals as radar radiation sources according to claim 2, wherein the step two of extracting features of the bispectrum according to a laplacian-gaussian operator comprises the following steps:

the laplacian-gaussian operator with 0 as the center and σ as the gaussian standard deviation has the following expression:

wherein G is _σ (x, y) is a second order Gaussian function, and the expression is as follows:

wherein σ is a Gaussian standard deviation;

performing Laplace-Gaussian operator feature extraction on the bispectrum:

LoGB _s (ω ₁ ,ω ₂ )＝LoG＊B _s (ω ₁ ,ω ₂ )；

wherein, the star represents convolution operation, and the operator is used for carrying out convolution operation on the bispectrum.

5. The method for accurately and intelligently identifying individuals with radar radiation sources according to claim 2, wherein the step three of inputting the extracted features into a deep residual error network based on norm and dynamic learning rate for training, and obtaining the trained model comprises:

when the network calculates each layer of parameters, an L2 norm is introduced, and the square root is obtained after the square sum of each feature vector element, so that the network is more sparse and smooth;

||x|| ₂ ＝(|x ₁ | ² +|x ₁ | ² +|x ₁ | ² +…+|x _n | ² ) ^1/2 ；

the dynamic learning rate is adopted in each round of learning as follows:

lr(n)＝lr*0.2 ^[n/10] ；

wherein lr (n) is the current learning rate, n is the batch, lr is the preset learning rate, the learning rate becomes 0.2 times of the current learning rate every 10 batches, and the learning rate is gradually reduced along with the round conversion;

the deep residual error network is composed of various residual error blocks, wherein each residual error fast learning rule is as follows:

F(x)＝H(x)-x；

wherein H (x) is a reverse propagation function, F (x) is a forward propagation function, and x is a network input;

inputting the bispectrum extracted by the Laplace-Gaussian operator into a single channel to generate H multiplied by W multiplied by 1 characteristics, wherein H and W are the length and width of the characteristic diagram respectively, and 1 is the number of channels; inputting the characteristics into a residual block 1 for convolution operation, and inputting the result into a residual block 2 until 4 residual blocks are operated; and finally, obtaining a final recognition result through the pooling layer and the full connection layer.

6. The method for accurately and intelligently identifying individuals as radar radiation sources according to claim 2, wherein the step four of utilizing the trained model to perform intelligent identification of individuals as radar radiation source signals comprises:

obtaining a corresponding bispectrum after receiving the unclassified preprocessed signals of the trained radar radiation source individuals, extracting the characteristics of the bispectrum according to a Laplace-Gaussian operator, and loading a trained network model; inputting the characteristics into a network to obtain an identification result; obtaining probability matrixes of each radar radiation source individual through a Softmax layer, wherein the highest probability is an identification result, and the Softmax expression is as follows:

M＝max(z)；

wherein z is the result vector, max is the maximum value of z, z _i Is the ith result.

7. The system for accurately and intelligently identifying the radar radiation source individuals by applying the method for accurately and intelligently identifying the radar radiation source individuals according to any one of claims 1 to 6, wherein the system for accurately and intelligently identifying the radar radiation source individuals comprises the following components:

the double-spectrum calculating module is used for calculating the corresponding double spectrum of the received radar radiation source signal;

the characteristic extraction module is used for extracting the characteristics of the bispectrum according to a Laplace-Gaussian operator;

the feature training module is used for inputting the extracted features into a deep residual error network based on norm and dynamic learning rate for training to obtain a trained model;

and the individual identification module is used for realizing the individual intelligent identification of the radar radiation source signal by utilizing the trained model.

8. A computer arrangement, characterized in that the computer arrangement comprises a memory and a processor, the memory storing a computer program which, when executed by the processor, causes the processor to carry out the steps of the method for accurate intelligent identification of an individual of a radar radiation source according to any one of claims 1 to 6.

9. A computer-readable storage medium, storing a computer program which, when executed by a processor, causes the processor to carry out the steps of the method for accurate intelligent identification of an individual of a radar radiation source as claimed in any one of claims 1 to 6.

10. An information data processing terminal, characterized in that the information data processing terminal is used for implementing the radar radiation source individual precise intelligent identification system according to claim 7.

Images