CN112949820B - Cognitive anti-interference target detection method based on generation of countermeasure network - Google Patents
Cognitive anti-interference target detection method based on generation of countermeasure network Download PDFInfo
- Publication number
- CN112949820B CN112949820B CN202110114075.2A CN202110114075A CN112949820B CN 112949820 B CN112949820 B CN 112949820B CN 202110114075 A CN202110114075 A CN 202110114075A CN 112949820 B CN112949820 B CN 112949820B
- Authority
- CN
- China
- Prior art keywords
- interference
- network
- data set
- target
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000001514 detection method Methods 0.000 title claims abstract description 122
- 230000001149 cognitive effect Effects 0.000 title claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 48
- 238000000034 method Methods 0.000 claims abstract description 35
- 230000001629 suppression Effects 0.000 claims abstract description 4
- 230000006870 function Effects 0.000 claims description 37
- 238000012360 testing method Methods 0.000 claims description 16
- 238000007781 pre-processing Methods 0.000 claims description 9
- 230000004913 activation Effects 0.000 claims description 8
- 238000013527 convolutional neural network Methods 0.000 claims description 8
- 230000008569 process Effects 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 6
- 239000013598 vector Substances 0.000 claims description 6
- 239000011159 matrix material Substances 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 2
- 230000008485 antagonism Effects 0.000 abstract description 2
- 238000004088 simulation Methods 0.000 description 13
- 238000010586 diagram Methods 0.000 description 11
- 230000008859 change Effects 0.000 description 5
- 238000013135 deep learning Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000005516 engineering process Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000003042 antagnostic effect Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000001364 causal effect Effects 0.000 description 1
- 230000002950 deficient Effects 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 230000006866 deterioration Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention discloses a cognitive anti-interference target detection method based on a generated countermeasure network, which comprises the following steps: constructing and generating an countermeasure network; generating a target data set and an interference data set by using the original small sample set; respectively training the generated countermeasure network by using the target data set and the interference data set to obtain a new echo sample; constructing an anti-interference detection network; training the anti-interference detection network by using a new echo sample; and performing target detection and interference suppression by using the trained anti-interference detection network. The invention provides a method for directly learning the distribution rule of data from the original small sample set by generating the antagonism network model, which overcomes the problem of insufficient generalization performance of the existing detection model when the training data is lacking, so that the method has better and more stable anti-interference detection performance under the condition of the small sample.
Description
Technical Field
The invention belongs to the technical field of radar anti-interference, and particularly relates to a cognitive anti-interference target detection method based on a generated countermeasure network.
Background
The main task of radar anti-interference target detection is to complete normal detection of targets in a complex interference scene. Along with the increasing deterioration of the electromagnetic environment where the radar is located, the traditional radar anti-interference processing process has the problems of inflexible working mode, low automation degree, poor adaptability and the like, and cannot meet the requirements of radar target detection.
In recent years, as the research and development of the deep learning algorithm are rapid, the problem of radar anti-interference detection can be solved by utilizing a correlation algorithm. Currently, there are methods for radar target detection using deep learning techniques. Liang Saiyuan in the published paper "study of anti-interference method of radar based on deep learning", a cognitive radar target detection method under a complex interference background is studied, and the method builds an end-to-end model from receiving target detection, so that anti-interference target detection is completed.
However, the above method requires relying on large-scale samples to accurately estimate the parameters of the detection model when training the target detection model. In the actual radar environment, the data sample resources are often very deficient, the detection model cannot be fully trained, the generalization performance is insufficient, the radar detection performance is severely reduced, and the application range of the deep learning technology in the aspect of radar anti-interference target detection is greatly limited.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a cognitive anti-interference target detection method based on generation of an antagonism network. The technical problems to be solved by the invention are realized by the following technical scheme:
a cognitive anti-interference target detection method based on a generated countermeasure network comprises the following steps:
s1: constructing and generating an countermeasure network;
s2: generating a target data set and an interference data set by using the original small sample set;
s3: respectively training the generated countermeasure network by using the target data set and the interference data set to obtain a new echo sample;
s4: constructing an anti-interference detection network;
s5: training the anti-interference detection network by using a new echo sample;
s6: and performing target detection and interference suppression by using the trained anti-interference detection network.
In one embodiment of the present invention, step S1 includes:
1a) Building a generating countermeasure network structure comprising a generator and a discriminator;
1b) And setting parameters of the generator and parameters of the discriminator respectively to complete the construction of the generation countermeasure network.
In one embodiment of the present invention, step S2 includes:
2a) Preprocessing the small sample echo data collected in the external field test process to obtain an original sample set; wherein the original sample set comprises a plurality of target and interference mixed echo data;
2b) Randomly dividing the original sample set into two parts according to a proportion, putting one part into a training set, and putting the other part into a test set to obtain an original sample training set and an original sample test set;
2c) And cutting out target echo data and interference echo data in the original sample training set to obtain a target data set and an interference data set.
In one embodiment of the present invention, step S3 includes:
the step S3 comprises the following steps:
3a) Sampling from 100-dimensional random noise vectors, and inputting the sampled random noise vectors into the generator to obtain 500-dimensional generated data samples;
3b) Sequentially inputting the target data set and the 500-dimensional generated data sample into the discriminator to respectively obtain a loss function of the generator and a loss function of the discriminator;
3c) Learning and updating the generator and the discriminator by adopting an alternate training strategy according to the loss function of the generator and the discriminator;
3d) Updating the weight and the bias parameters of the generated countermeasure network by adopting an RMSProp algorithm, and cutting the weight and the bias parameters of the discriminator to a certain range after updating;
3e) Repeating the steps 3 a) to 3 d), and obtaining a trained generation countermeasure network when the loss function value of the discriminator tends to 0;
3f) Expanding the target data set by using a trained generation countermeasure network to obtain an expanded target data set;
3g) Replacing the target data set with the interference data set according to the method of the steps 3 a) to 3 e), retraining to generate an countermeasure network, and expanding the interference data set to obtain an expanded interference data set;
3h) And generating a new echo sample by using the expanded target data set and the expanded interference data set.
In one embodiment of the invention, the loss function of the generator is:
wherein z represents randomNoise, obeying the distribution p (z), g represents the generator, θ represents the parameters of the generator, g θ (z) represents the artifact generated by the generator;
the loss function of the discriminator is as follows:
wherein x represents a true signal, P r Representing the true signal distribution, f represents the arbiter, ω represents the parameters of the arbiter.
In one embodiment of the present invention, step S4 includes:
4a) Building an anti-interference detection model based on a convolutional neural network; the anti-interference detection model comprises an input convolution layer, a WaveNet layer and an output convolution layer;
4b) Setting parameters of each layer in the anti-interference detection model to complete construction of the anti-interference detection model.
In one embodiment of the invention, the WaveNet layer includes a gate activation unit, the function of which is expressed as:
Y＝tanh(W f,k *X)⊙σ(W g,k *X)；
wherein X represents the input of the unit, W represents a learnable convolution filter, f represents the filtering unit, g represents the gate unit, k represents the current network level,/represents the convolution operation,/represents the dot product operation,/represents the sigmoid activation function.
In one embodiment of the present invention, step S5 includes:
5a) Mixing the new echo sample with the original sample training set to obtain a training set with enhanced data;
5b) Preprocessing the training set after data enhancement, and taking a feature matrix obtained after preprocessing as input of the anti-interference detection network;
5c) And setting a cost function of the anti-interference detection network and performing minimization treatment on the cost function to obtain the anti-interference detection network matched with the original sample set.
In one embodiment of the present invention, the cost function of the tamper resistant detection network is expressed as:
L KL (P(Y|X),D(X))＝-P(Y|X)log(D w (X))-(1-P(Y|X))log(1-D w (X))；
wherein D is w Represents an anti-interference detection network, Y= [ Y ] 1 ,y 2 ,...,y N ]Labels indicating whether or not there is an object on each distance cell, P (y|x) = [ P (Y) 1 |X),P(y 2 |X),...,P(y N |X)]A probability set indicating whether or not an object exists on each distance cell, and X indicates an input radar echo signal.
The invention has the beneficial effects that:
1. according to the cognitive anti-interference target detection method based on the generation of the countermeasure network, provided by the invention, the distribution rule of data is directly learned from the original small sample set by generating the countermeasure network model, so that the problem that the generalization performance of the existing detection model is insufficient when training data is lacking is solved, and the method has better and more robust anti-interference detection performance under the condition of the small sample;
2. according to the invention, the real echo data actually observed is taken as the training sample by the generation countermeasure network, and the distance between the training sample and the generation sample is measured by using the Wasserstein distance in the countermeasure training process, so that the high-quality generation sample which is closer to the real sample is obtained, and the application range is enlarged.
The present invention will be described in further detail with reference to the accompanying drawings and examples.
Drawings
Fig. 1 is a schematic flow chart of a cognitive anti-interference target detection method based on a generation countermeasure network according to an embodiment of the present invention;
FIGS. 2a-2b are schematic illustrations of the structure of a generated countermeasure network provided by embodiments of the present invention;
fig. 3 is a schematic structural diagram of an anti-interference detection network according to an embodiment of the present invention;
FIGS. 4a-4d are time-frequency diagrams of sample and real sample generated at different stages in the countermeasure process by the generator and the arbiter according to the embodiment of the present invention;
fig. 5 is a graph of the false alarm rate P at signal-to-noise ratio snr=14 dB fa ＝10 -4 When the method and the prior method are adopted, the detection probability of the detection network after learning under the test set is changed along with the interference signal ratio JSR;
fig. 6 is a graph of the false alarm rate P at the interference-to-signal ratio jsr=8 dB fa ＝10 -4 The method and the simulation graph of the detection probability of the detection network learned by the prior method under the test set along with the change of the SNR are adopted.
Detailed Description
The present invention will be described in further detail with reference to specific examples, but embodiments of the present invention are not limited thereto.
Example 1
Referring to fig. 1, fig. 1 is a flowchart of a cognitive anti-interference target detection method based on generating an anti-network according to an embodiment of the present invention, which includes the following steps:
s1: constructing a generated countermeasure network (WGAN), specifically comprising:
1a) A generation countermeasure network structure including a generator and a discriminator is built.
Referring to fig. 2a-2b, fig. 2a-2b are schematic structural diagrams of generating an countermeasure network according to an embodiment of the present invention, wherein fig. 2a is a schematic structural diagram of a generator, and fig. 2b is a schematic structural diagram of a discriminator. In this embodiment, both the generator and the arbiter include an input layer, a hidden layer (consisting of fully connected layers) and an output layer.
1b) And setting parameters of the generator and parameters of the discriminator respectively to complete the construction of the generation countermeasure network.
Specifically, for the generator, the number of input units thereof is set to 100, the number of units of the hidden layer is set to 128 and 256, the activation function is set to the linear rectification function ReLU, the number of output units is set to 500, and the activation function is set to tanh.
For the arbiter, the input unit number is set to 500, the unit number of the hidden layer is set to 256 and 128, the activation function is set to the linear rectification function ReLU, and the output unit number is set to 1.
S2: generating a target data set and an interference data set by using the original small sample set specifically comprises the following steps:
2a) Preprocessing the small sample echo data (namely an original small sample set) collected in the external field test process to obtain the original sample set; wherein the original sample set comprises a plurality of target and interference mixed echo data;
2b) And randomly dividing the original sample set into two parts according to a proportion, putting one part into a training set, and putting the other part into a test set to obtain an original sample training set and an original sample test set.
Specifically, in this embodiment, 30% of the original sample set is placed in the training set, 70% is placed in the test set, and the two data sets are not overlapped with each other and remain independent of each other.
2c) And intercepting target echo data and interference echo data in the original sample training set to obtain a target data set and an interference signal data set.
S3: and respectively training the generated countermeasure network by using the target data set and the interference data set to obtain a new echo sample.
The method specifically comprises the following steps:
3a) Sampling from 100-dimensional random noise vectors, and inputting the sampled random noise vectors into the generator to obtain 500-dimensional generated data samples.
3b) And sequentially inputting the target data set and the 500-dimensional generated data sample into the discriminator to respectively obtain the generator loss function and the loss function of the discriminator.
In this embodiment, the loss function of the generator is:
wherein z represents random noise, obeys the distribution p (z), g represents a generator (neural network), θ represents a parameter of the generator, g θ (z) represents the artifact generated by the generator;
the loss function of the discriminator is as follows:
wherein x represents a true signal, P r Representing the true signal distribution, f represents the arbiter (neural network), ω represents the parameters of the arbiter.
3c) And learning and updating the generator and the discriminator by adopting an alternate training strategy according to the loss functions of the generator and the discriminator, wherein the ratio of the updating times of the discriminator to the generator is 5:1.
3d) And updating the weight and the bias parameters of the generated countermeasure network by adopting an RMSProp algorithm, and cutting the weight and the bias parameters of the discriminator to a certain range after updating.
In this embodiment, the number of updates may be set to 100000 times, and the weights and bias parameters of the discriminators are cut to be between [ -0.01,0.01] after each update, and then the countermeasure network is updated.
3e) Repeating the steps 3 a) to 3 d), and obtaining the trained generated countermeasure network when the loss function value of the discriminator tends to 0.
3f) And expanding the target data set by using the trained generation countermeasure network to obtain an expanded target data set.
3g) According to the method of the steps 3 a) to 3 e), the target data set is replaced by the interference data set, and the interference data set is expanded by the retraining generation countermeasure network, so that the expanded interference data set is obtained.
3h) And generating a new echo sample by using the expanded target data set and the expanded interference data set.
In this embodiment, since the loss function value of the arbiter may be equal to the wasperstein distance, the present embodiment uses the real echo data actually observed by the generation countermeasure network as the training sample, and measures the distance between the training sample and the generated sample by using the wasperstein distance in the countermeasure training process, thereby obtaining a high-quality generated sample that is closer to the real sample, and expanding the application range.
S4: the method for constructing the anti-interference detection network specifically comprises the following steps:
4a) Building an anti-interference detection model based on a convolutional neural network; the anti-interference detection model comprises an input convolution layer, a WaveNet layer and an output convolution layer. Referring to fig. 3, fig. 3 is a schematic structural diagram of an anti-interference detection network according to an embodiment of the present invention.
4b) Setting parameters of each layer in the anti-interference detection model to complete construction of the anti-interference detection model.
Specifically, the number of input units of the anti-interference detection model is set to 250×250×6, 4 layers of convolution layers are included in the input convolution layers, the number of convolution kernels of each layer is set to 12, 18, 24 and 30, the size of the convolution kernels is set to 1×7, the wavenet layer is a 7-layer causal expansion convolution layer including a gate activation unit, the number of expansion factors is set to 1, 2, 4, 8, 16, 32, 64 and 128, the number of convolution kernels of each layer is set to 64, the number of convolution kernels of the output convolution layer is set to 64, and the size of the convolution kernels is set to 1×1.
S5: training the anti-interference detection network by using a new echo sample specifically comprises the following steps:
5a) Mixing the new echo sample with the original sample training set to obtain a training set with enhanced data;
5b) Preprocessing the training set after data enhancement, and taking a feature matrix obtained after preprocessing as input of the anti-interference detection network;
5c) And setting a cost function of the anti-interference detection network and performing minimization treatment on the cost function to obtain the anti-interference detection network matched with the original sample set. Wherein, the cost function of the anti-interference detection network is expressed as:
L KL (P(Y|X),D(X))＝-P(Y|X)log(D w (X))-(1-P(Y|X))log(1-D w (X))；
wherein D is w Represents an anti-interference detection network, Y= [ Y ] 1 ,y 2 ,…,y N ]Representing the distance unitsIf there is a tag of the target, P (y|x) = [ P (Y) 1 |X),P(y 2 |X),...,P(y N |X)]A probability set indicating whether or not an object exists on each distance cell, X indicating an input radar echo signal
S6: and performing target detection and interference suppression by using the trained anti-interference detection network.
And the trained anti-interference detection network is utilized to realize the anti-interference target detection task under the condition of a small sample.
According to the cognitive anti-interference target detection method based on the generation of the countermeasure network, provided by the invention, the distribution rule of the data is directly learned from the original small sample set by generating the countermeasure network model, so that the problem that the generalization performance of the existing detection model is insufficient when training data is lacking is solved, and the method has better and more robust anti-interference detection performance under the condition of the small sample.
Example two
On the basis of the first embodiment, the effects of the present invention will be further described with reference to simulation experiments.
1. Simulation experiment conditions:
the hardware test platform of the simulation experiment of the invention is: the processor is Intel (R) Core ((TM) i 7-4790K), the main frequency is 4.0GHz, the memory is 16GB, and the software platform is Ubuntu 16.04LTS, 64-bit operating system and Python 2.7.
2. Simulation content and simulation result analysis:
in order to prove that the cognitive anti-interference target detection method based on the generation of the anti-network WGAN can obtain better detection performance under the condition of a small sample, the simulation experiment of the invention adopts the cognitive anti-interference target detection method based on the generation of the anti-network WGAN and the anti-interference detection method based on the convolutional neural network in the prior art.
The anti-interference detection method based on the convolutional neural network in the prior art is a radar anti-interference target detection method based on the convolutional neural network, which is proposed in an engineering and major paper of the university of electronic science and technology Liang Saiyuan of western security, namely, a research on a radar anti-interference method based on deep learning.
In the simulation experiment, the initial sample training set is used for performing countermeasure training on the generated countermeasure network WGAN, and after the generated countermeasure network WGAN converges, the sample set can be expanded. In the simulation use of the invention, the detection model is trained by adopting different numbers of expansion sample sets, wherein the expansion sample numbers are respectively 1000, 2000, 4000, 6000 and 8000.
Referring to fig. 4a-4d, fig. 4a-4d are time-frequency diagrams of a sample and a real sample generated at different stages in the countermeasure process by the generator and the arbiter according to the embodiment of the present invention, wherein the x-axis represents time and the y-axis represents frequency. Fig. 4a is a time-frequency diagram of a sample generated after 300 rounds of challenge training, fig. 4b is a time-frequency diagram of a sample generated after 1000 rounds of challenge training, fig. 4c is a time-frequency diagram of a sample generated after 3000 rounds of challenge training, and fig. 4d is a time-frequency diagram of a real sample. Comparing the results of fig. 4a-4d, it can be found that, with the increasing number of iterations, the quality of the generated samples is gradually improved, after 3000 rounds of training, the generated model has effectively learned the distribution characteristics of the real samples, which proves that the generated countermeasure network WGAN technology adopted by the invention can generate samples similar to the original samples, and achieves the purpose of sample expansion.
Further, given signal-to-noise ratio snr=14 dB, false alarm rate P fa ＝10 -4 Simulation tests are carried out on the average detection probability variation along with the change of the interference signal ratio JSR under different sample numbers, and the results are shown in table 1.
According to comparison, as the number of samples is continuously increased, the anti-interference detection performance is gradually improved, and the cognitive anti-interference target detection method based on the generation of the anti-network WGAN can improve the detection capability of a detection network by providing large-scale samples, so that the detection performance degradation caused by the lack of data samples in the prior art is overcome.
TABLE 1
In order to further verify the simulation experiment effect of the present invention, the detection results of the two methods are compared by 4000 monte carlo results using the prior art method and the method of the present invention, respectively.
Referring to fig. 5, fig. 5 shows the false alarm rate P when the snr=14 dB fa ＝10 -4 The method and the simulation graph of the detection probability of the detection network learned by the prior art under the test set along with the change of the interference signal ratio JSR are adopted, wherein the x-axis represents the interference signal ratio JSR, the y-axis represents the detection probability, the curve represented by the solid line "- - -" is the change curve of the detection probability of the prior art under different interference signal ratios, and the curve represented by the dotted line "-" is the change curve of the detection probability of the detection method based on the cognitive anti-interference target for generating the anti-network WGAN under different interference signal ratios. Fig. 6 is a graph of the false alarm rate P at the interference-to-signal ratio jsr=8 dB fa ＝10 -4 The method and the simulation graph of the variation of the detection probability of the detection network under the test set, which is learned by the prior art, along with the SNR are adopted, wherein the x-axis represents the SNR, the y-axis represents the detection probability, the curve represented by the solid line "- - -" is the variation curve of the detection probability of the prior art under different SNRs, and the curve represented by the dotted line "-" is the variation curve of the detection probability of the detection method under different SNRs based on the cognitive anti-interference target for generating the anti-network WGAN. As can be found from fig. 5 and fig. 6, compared with the existing radar anti-interference target detection method based on the convolutional neural network, the method for detecting the cognitive anti-interference target based on the generating anti-interference network WGAN has an obvious improvement effect on the detection probability, and proves that the method can provide a large number of training samples under the condition that the number of the original samples is limited, and the driving depth network obtains better anti-interference detection performance.
In combination with table 1, fig. 5 and fig. 6, it can be concluded that: compared with the radar anti-interference target detection method based on the convolutional neural network in the prior art, the detection effect of the detection method based on the generating anti-interference target detection method based on the WGAN of the antagonistic network can be achieved under the condition of small samples. According to the method, the original training set can be expanded by learning the data distribution of the real samples and generating the samples, the problem of poor detection performance caused by lack of sufficient samples in the prior art is solved, and the target detection effect is improved.
The foregoing is a further detailed description of the invention in connection with the preferred embodiments, and it is not intended that the invention be limited to the specific embodiments described. It will be apparent to those skilled in the art that several simple deductions or substitutions may be made without departing from the spirit of the invention, and these should be considered to be within the scope of the invention.
Claims (6)
1. The cognitive anti-interference target detection method based on the generation of the countermeasure network is characterized by comprising the following steps of:
s1: constructing and generating an countermeasure network;
s2: generating a target data set and an interference data set by using the original small sample set;
s3: respectively training the generated countermeasure network by using the target data set and the interference data set to obtain a new echo sample;
3a) Sampling from 100-dimensional random noise vectors, and inputting the sampled random noise vectors into a generator to obtain 500-dimensional generated data samples;
3b) Sequentially inputting the target data set and the 500-dimensional generated data sample into a discriminator to respectively obtain a loss function of the generator and a loss function of the discriminator;
3c) Learning and updating the generator and the discriminator by adopting an alternate training strategy according to the loss function of the generator and the discriminator;
3d) Updating the weight and the bias parameters of the generated countermeasure network by adopting an RMSProp algorithm, and cutting the weight and the bias parameters of the discriminator to a certain range after updating;
3e) Repeating the steps 3 a) to 3 d), and obtaining a trained generation countermeasure network when the loss function value of the discriminator tends to 0;
3f) Expanding the target data set by using a trained generation countermeasure network to obtain an expanded target data set;
3g) Replacing the target data set with the interference data set according to the method of the steps 3 a) to 3 e), retraining to generate an countermeasure network, and expanding the interference data set to obtain an expanded interference data set;
3h) Generating a new echo sample by using the expanded target data set and the expanded interference data set;
s4: constructing an anti-interference detection network;
4a) Building an anti-interference detection model based on a convolutional neural network; the anti-interference detection model comprises an input convolution layer, a WaveNet layer and an output convolution layer;
the WaveNet layer includes a gate activation unit whose function is expressed as:
Y＝tanh(W f,k *X)⊙σ(W g,k *X)；
wherein X represents the input of the unit, W represents a learnable convolution filter, f represents the filtering unit, g represents the gate unit, k represents the current network level,/represents the convolution operation,/represents the dot product operation,/represents the sigmoid activation function;
4b) Setting parameters of each layer in the anti-interference detection model to complete construction of the anti-interference detection model;
s5: training the anti-interference detection network by using a new echo sample;
s6: and performing target detection and interference suppression by using the trained anti-interference detection network.
2. The cognitive anti-interference target detection method based on generation of an countermeasure network according to claim 1, wherein step S1 includes:
1a) Building a generating countermeasure network structure comprising a generator and a discriminator;
1b) And setting parameters of the generator and parameters of the discriminator respectively to complete the construction of the generation countermeasure network.
3. The cognitive anti-interference target detection method based on generation of an countermeasure network according to claim 1, wherein step S2 includes:
2a) Preprocessing the small sample echo data collected in the external field test process to obtain an original sample set; wherein the original sample set comprises a plurality of target and interference mixed echo data;
2b) Randomly dividing the original sample set into two parts according to a proportion, putting one part into a training set, and putting the other part into a test set to obtain an original sample training set and an original sample test set;
2c) And cutting out target echo data and interference echo data in the original sample training set to obtain a target data set and an interference data set.
4. The cognitive anti-interference target detection method based on generation of an countermeasure network of claim 1, wherein the loss function of the generator is:
wherein z represents random noise, obeys the distribution p (z), g represents a generator, θ represents a parameter of the generator, g θ (z) represents the artifact generated by the generator;
the loss function of the discriminator is as follows:
wherein x represents a true signal, P r Representing the true signal distribution, f represents the arbiter, ω represents the parameters of the arbiter.
5. The cognitive anti-interference target detection method based on generation of an countermeasure network according to claim 3, wherein step S5 includes:
5a) Mixing the new echo sample with the original sample training set to obtain a training set with enhanced data;
5b) Preprocessing the training set after data enhancement, and taking a feature matrix obtained after preprocessing as input of the anti-interference detection network;
5c) And setting a cost function of the anti-interference detection network and performing minimization treatment on the cost function to obtain the anti-interference detection network matched with the original sample set.
6. The cognitive anti-interference target detection method based on generation of an anti-interference network according to claim 5, wherein the cost function of the anti-interference detection network is expressed as:
L KL (P(Y|X),D(X))＝-P(Y|X)log(D w (X))-(1-P(Y|X))log(1-D w (X))；
wherein D is w Represents an anti-interference detection network, Y= [ Y ] 1 ,y 2 ,...,y N ]Labels indicating whether or not there is an object on each distance cell, P (y|x) = [ P (Y) 1 |X),P(y 2 |X),…,P(y N |X)]A probability set indicating whether or not an object exists on each distance cell, and X indicates an input radar echo signal.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110114075.2A CN112949820B (en) | 2021-01-27 | 2021-01-27 | Cognitive anti-interference target detection method based on generation of countermeasure network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110114075.2A CN112949820B (en) | 2021-01-27 | 2021-01-27 | Cognitive anti-interference target detection method based on generation of countermeasure network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112949820A CN112949820A (en) | 2021-06-11 |
| CN112949820B true CN112949820B (en) | 2024-02-02 |
Family
ID=76238203
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110114075.2A Active CN112949820B (en) | 2021-01-27 | 2021-01-27 | Cognitive anti-interference target detection method based on generation of countermeasure network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112949820B (en) |
Families Citing this family (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114301749B (en) * | 2021-12-27 | 2024-02-13 | 杭州电子科技大学 | Signal anti-interference method based on generation of countermeasure network |
| CN114330924B (en) * | 2022-01-10 | 2023-04-18 | 中国矿业大学 | Complex product change strength prediction method based on generating type countermeasure network |
| CN114509731B (en) * | 2022-01-24 | 2023-09-05 | 电子科技大学 | Radar main lobe anti-interference method based on double-stage depth network |
| CN114818777B (en) * | 2022-03-18 | 2023-07-21 | 北京遥感设备研究所 | Training method and device for active angle deception jamming recognition model |
| CN115100233B (en) * | 2022-06-15 | 2024-10-18 | 西安电子科技大学 | Radar target tracking method based on generation of anti-network resampling particle filtering |
| CN115310361B (en) * | 2022-08-16 | 2023-09-15 | 中国矿业大学 | Underground coal mine dust concentration prediction method and system based on WGAN-CNN |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766835A (en) * | 2019-01-11 | 2019-05-17 | 西安电子科技大学 | The SAR target identification method of confrontation network is generated based on multi-parameters optimization |
| CN110575141A (en) * | 2019-09-09 | 2019-12-17 | 西北大学 | Epilepsy detection method based on generation countermeasure network |
| CN111612717A (en) * | 2020-05-20 | 2020-09-01 | 上海杉达学院 | Water surface image reflection removing and repairing method and device based on countermeasure network |
| WO2020233207A1 (en) * | 2019-05-20 | 2020-11-26 | 广东省智能制造研究所 | Hyperspectral data analysis method based on semi-supervised learning strategy |
-
2021
- 2021-01-27 CN CN202110114075.2A patent/CN112949820B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109766835A (en) * | 2019-01-11 | 2019-05-17 | 西安电子科技大学 | The SAR target identification method of confrontation network is generated based on multi-parameters optimization |
| WO2020233207A1 (en) * | 2019-05-20 | 2020-11-26 | 广东省智能制造研究所 | Hyperspectral data analysis method based on semi-supervised learning strategy |
| CN110575141A (en) * | 2019-09-09 | 2019-12-17 | 西北大学 | Epilepsy detection method based on generation countermeasure network |
| CN111612717A (en) * | 2020-05-20 | 2020-09-01 | 上海杉达学院 | Water surface image reflection removing and repairing method and device based on countermeasure network |
Non-Patent Citations (1)
| Title |
|---|
| 王德文 ; 杨凯华 ; .基于生成式对抗网络的窃电检测数据生成方法.电网技术.2020,(第02期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112949820A (en) | 2021-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112949820B (en) | Cognitive anti-interference target detection method based on generation of countermeasure network | |
| CN112949387B (en) | Intelligent anti-interference target detection method based on transfer learning | |
| CN111126471A (en) | Microseism event detection method and system | |
| CN112308008B (en) | Radar radiation source individual identification method based on working mode open set of transfer learning | |
| CN114595732B (en) | Radar radiation source sorting method based on depth clustering | |
| Sun et al. | Cluster guide particle swarm optimization (CGPSO) for underdetermined blind source separation with advanced conditions | |
| CN113673312B (en) | Deep learning-based radar signal intra-pulse modulation identification method | |
| CN111880158A (en) | Radar target detection method and system based on convolutional neural network sequence classification | |
| CN113109782B (en) | Classification method directly applied to radar radiation source amplitude sequence | |
| CN113315593A (en) | Frequency spectrum sensing algorithm based on FLOM covariance matrix and LSTM neural network | |
| CN104392086B (en) | A kind of signal deteching circuit and method based on Pearson came rand variate coefficient correlation | |
| CN109067427A (en) | A kind of frequency hop sequences prediction technique based on Optimization-type wavelet neural network | |
| CN110932807A (en) | Spectrum sensing method of MIMO (multiple input multiple output) system under non-Gaussian noise | |
| CN114117912A (en) | Sea clutter modeling and inhibiting method under data model dual drive | |
| CN114943245A (en) | Automatic modulation recognition method and device based on data enhancement and feature embedding | |
| CN113539386A (en) | CLMVO-ELM-based dissolved oxygen concentration prediction method, device, equipment and storage medium | |
| Su et al. | Statistical detection of weak pulse signal under chaotic noise based on Elman neural network | |
| CN112836788B (en) | Low-power-consumption deep learning network method for interference type identification | |
| Wang et al. | A structurally re-parameterized convolution neural network-based method for gearbox fault diagnosis in edge computing scenarios | |
| CN116863959B (en) | Dolphin sound generating method based on generating countermeasure network | |
| Chen et al. | The adaptive constant false alarm rate for sonar target detection based on back propagation neural network access | |
| CN115422977A (en) | Radar radiation source signal identification method based on CNN-BLS network | |
| CN114997378A (en) | Inductive graph neural network pruning method, system, device and storage medium | |
| Huang et al. | A method for extracting fingerprint feature of communication satellite signal | |
| CN113359091B (en) | Deep learning-based multi-kernel function aliasing radar radiation source identification method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |