CN114117912A - Sea clutter modeling and inhibiting method under data model dual drive - Google Patents
Sea clutter modeling and inhibiting method under data model dual drive Download PDFInfo
- Publication number
- CN114117912A CN114117912A CN202111419764.0A CN202111419764A CN114117912A CN 114117912 A CN114117912 A CN 114117912A CN 202111419764 A CN202111419764 A CN 202111419764A CN 114117912 A CN114117912 A CN 114117912A
- Authority
- CN
- China
- Prior art keywords
- sea clutter
- data
- model
- training
- discriminator
- 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.)
- Pending
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- 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
- G06N3/084—Backpropagation, e.g. using gradient descent
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Mathematical Physics (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a sea clutter modeling and inhibiting method under dual drive of a data model, and belongs to the technical field of radar and signal processing technology and artificial intelligence. The method comprises the following steps: establishing and generating a countermeasure network GAN as a driving model of a sea clutter modeling algorithm, and simultaneously adopting Gaussian white noise data and sea clutter data as driving data of the sea clutter modeling algorithm; model training of a generator and a discriminator in the GAN is carried out; building a sea clutter suppression model based on a Convolutional Neural Network (CNN), and performing model training on the CNN by using data generated by a GAN model as an input data set of the CNN; and measuring model performance, measuring the sea clutter modeling simulation effect of the GAN model by adopting mean square deviation MSD (maximum mean square deviation) test, and testing the sea clutter suppression effect of the CNN model by comparing the frequency spectrums of the radar before and after the CNN model when receiving signals. The method provided by the invention can effectively carry out simulation modeling of the sea clutter and can inhibit the sea clutter component in the radar receiving signal.
Description
Technical Field
The invention belongs to the technical field of radar and signal processing technology and artificial intelligence, and particularly relates to a sea clutter modeling and inhibiting method under double drive of a data model.
Background
The working environment of the marine radar is complex and changeable, and the received radar echo signals contain noise and sea clutter components besides the required target echo signals. Since the sea clutter has a high power level, the sea clutter component becomes one of the main factors affecting the operating performance of the marine radar. The sea clutter has non-Gaussian and non-stationary characteristics, so that the modeling simulation of the sea clutter is an important basis for further researching the inhibition of the sea clutter component in the radar receiving signal.
At present, the research and analysis on the sea clutter mainly comprises two analysis ideas based on statistical distribution and physical characteristics. Statistical distribution-based methods are mainly based on models, such as lognormal distribution, Weibull distribution, K distribution, and the like. The K distribution model can simultaneously consider the amplitude distribution characteristics and the correlation performance between pulses of the sea clutter, so that the statistical characteristics of the sea clutter can be more accurately reflected than other models. Conte and Marier respectively utilize a ball invariant random process SIRP and a zero memory nonlinear transformation method ZmNL to simulate the sea clutter, and the two methods are widely used sea clutter simulation methods at present. Ritchie calculates the cumulative amplitude distribution of the sea clutter to obtain the false alarm probability so as to evaluate the statistic of the sea clutter; watts analyzes the main characteristics of average power spectral density, Doppler spectrum extreme amplitude and the like by using the measurement result recorded by the Doppler spectrum, provides a sea clutter amplitude statistical model based on composite K distribution, and verifies through a large amount of sea clutter test data; weinberg provides a KK distribution sea clutter model with simpler calculation by utilizing mutually independent Gaussian vector weighted products and constructing in-phase orthogonal components. Methods based on physical properties are typically analyzed from chaotic or fractal characteristics of sea clutter. Haykin and other researches analyze the embedding dimension and Lyapunov index of the actually measured sea clutter, and a small target detection algorithm based on the combination of chaotic phase space reconstruction and back propagation BP neural network is provided; then, Leung detects a target signal by using a nonlinear prediction method based on the correlation dimension and a memory bank; the treble and the like predict the chaotic time sequence through a Support Vector Machine (SVM), realize target detection and are widely popularized. Unswoorth and the like find that sea clutter does not strictly have chaotic characteristics; hu and the like verify that sea clutter has fractal characteristics by using fractional Brownian motion, and perform offshore target detection by using Hurst indexes to obtain the recognition of most scholars, so that the Hu and the like become a follow-up research hotspot. In China, the Ronhan-type hybrid and the like respectively provide a multi-fractal target detection algorithm based on an SVM (support vector machine) and an attenuation fluctuation analysis by utilizing chaos and fractal characteristics of a sea clutter, and are continuously improved and perfected; the lie cycle and the like adopt a radial basis function neural network and space and time chaotic reconstruction to carry out small and weak target detection; liuning waves and the like are subjected to a great deal of analysis and research on frequency domain fractal characteristics and combination characteristics. The above documents separately analyze and research the sea clutter from two angles of statistical distribution and physical characteristics, and both have great reference values. However, the sea clutter modeling based on statistical distribution does not consider the physical characteristics of the sea clutter, and the inherent dynamic characteristics are difficult to reveal; the method based on physical characteristics realizes the detection of the small targets under the sea clutter background by utilizing the difference of the fractal parameters of the sea clutter with the small targets and the pure sea clutter, but fails to establish a sea clutter model under different sea conditions, and is inconvenient for simulation evaluation.
The data model dual-drive modeling method is generally used for processing the condition that the mechanism of an object is not clear or is more complex, and a linear regression method, an SVM method, a neural network method and the like can be used. With the advent of generative countermeasure networks, more and more generative tasks began to adopt GAN expansion. At present, a large amount of actually measured sea clutter data are accumulated, and a reliable driving data basis is provided for modeling of sea clutter by using GAN.
Disclosure of Invention
The invention provides a sea clutter modeling and inhibiting method under double drive of a data model, aiming at solving the problem that the existing method can not fully represent sea clutter characteristics so as to seriously influence the sea clutter simulation modeling effect, and aiming at carrying out the modeling simulation of the sea clutter by fully utilizing the statistical distribution characteristics of sea clutter theoretical data and the physical characteristics of sea clutter actual measurement data and generating a large number of sea clutter data sets with expansibility.
In order to solve the above problems, the present invention adopts the following technical solutions.
A sea clutter modeling and inhibiting method under data model dual drive comprises the following steps:
building a GAN network as a driving model for sea clutter modeling, and adopting Gaussian white noise data, sea clutter simulation data which are generated by using a sphere invariant random process SIRP method and meet K distribution, and actually measured sea clutter data as driving data for sea clutter modeling;
secondly, performing model training of a generator and a discriminator of the GAN;
step three, building a CNN-based sea clutter suppression model, and using sea clutter data generated by the GAN model as an input data set of the CNN;
step four, performing CNN model training;
and fifthly, measuring model performance, measuring the sea clutter modeling simulation effect of the GAN model by adopting MSD (maximum-resolution) test, and testing the sea clutter suppression effect of the CNN model by comparing the frequency spectrums of the radar before and after the CNN model when receiving signals.
Further, the GAN network is built in the step one to serve as a driving model for sea clutter modeling, and Gaussian white noise data, sea clutter simulation data which are generated by a sphere invariant random process SIRP method and meet K distribution, and actually measured sea clutter data are used as driving data for sea clutter modeling;
the network overall structure of the driving model GAN is shown in fig. 2. The driving data needs to be preprocessed by using short-time Fourier transform (STFT), and the calculation formula of the STFT specifically comprises the following steps:
wherein y (u) is an initial signal, g (u) is a window function, superscript x is a complex conjugate, t represents a central position of the window function, u represents time domain time, and f represents frequency; STFTy(t, f) is the STFT of the initial signal y (u). Respectively performing STFT on the three data through a calculation formula to obtainAnd taking the three data sets as driving data of the text.
Further, the model training of the generator and the discriminator of the GAN is performed in the step two. The specific structure of the discriminator and the generator is shown in fig. 3, and the specific training mode of the GAN model is divided into the following two parts:
training 1: individual training discriminator
1) Obtaining a time frequency spectrum by using 500 parts of white noise sample data meeting Gaussian distribution and 500 parts of sea clutter simulation sample data meeting K distribution through STFT;
2) mixing the 1000 parts of time-frequency spectrum image samples, respectively constructing a training set and a test set according to the proportion of 8:2, setting an image label of Gaussian white noise to be 0, and setting an image label of sea clutter simulation data to be 1;
3) and taking the training set as the input R of the discriminator, inputting the training set into the discriminator to train the discriminator until the discriminator is converged, and then taking the test set as the input R of the discriminator to test the convergence effect of the discriminator.
Training 2: simultaneous training generator and discriminator
1) Obtaining a time frequency spectrum from 500 parts of actually measured sea clutter data through STFT;
2) constructing the 500 time-frequency spectrum image samples into a data set, and setting an image label as 1;
3) taking the data set as an input R of a discriminator, and taking a random sequence satisfying Gaussian distribution as an input Z of a generator;
4) training iteration of the discriminator and the generator is alternately carried out according to the sequence of firstly training the discriminator for 1 time and then training the generator for 1 time, and after training is finished, subsequent comparison simulation is carried out on the time-frequency spectrum image of the sea clutter generated by the generator and the actually-measured time-frequency spectrum image of the sea clutter.
Further, a sea clutter suppression model based on the CNN is built in the step three, and the sea clutter data generated by the GAN model is used as an input data set of the CNN. The overall network structure of the CNN is shown in fig. 4.
Further, the model training of the CNN is performed in step four, a network of the CNN is specifically structured as shown in fig. 5, and the specific training mode is as follows:
1) adding 500 parts of sea clutter time-frequency spectrum images generated by a generator of the GAN with time-frequency spectrums of target echoes of three different Doppler frequencies to obtain data sets of radar receiving signals under the three different Doppler frequencies, respectively constructing a training set and a testing set for each data set according to the proportion of 8:2, and setting labels as the time-frequency spectrums of the target echoes of the Doppler frequencies corresponding to each data set;
2) inputting the training set into the CNN for iterative training, and after the training is finished, inputting the test set to obtain a time-frequency spectrum of the radar receiving signal after the sea clutter component is suppressed.
Further, the measurement of the model performance in the fifth step is carried out, the sea clutter modeling simulation effect of the GAN model is measured by adopting MSD (maximum data rate) test, and the sea clutter suppression effect of the CNN model is tested by comparing frequency spectrums before and after the CNN model is used when the radar receives signals;
the formula for the MSD test is defined as
Wherein p ise(xk) Probability distribution function PDF, p for actually measured sea clutter datat(xk) The PDF of the sea clutter model generated for the algorithm, n, represents the length of the selected data sequence. The smaller the MSD value is, the greater the similarity of the two distributions is, and the better the prediction effect of the algorithm model is.
Advantageous effects
The sea clutter modeling and inhibiting method under the dual drive of the data model has the following advantages:
1. the driving data used for sea clutter modeling simulation is carried out, and the statistical distribution characteristic of sea clutter theoretical data and the physical characteristic of sea clutter actual measurement data are fully utilized;
2. the sea clutter simulation modeling method has better fitting degree in the aspect of fitting the measured data compared with a sea clutter simulation algorithm based on the measured data, a sea clutter simulation algorithm based on a back propagation neural network and a sea clutter simulation algorithm based on a gated cyclic neural network, and shows that the sea clutter data generated by the method has better expansibility;
3. the method can generate a large number of sea clutter data sets with expansibility, and provides reliable and massive sea clutter basic data sets for subsequent research directions of sea clutter suppression, target detection and the like.
Drawings
FIG. 1 is a flow chart of a method for modeling and suppressing sea clutter under dual drive of a data model according to the present invention;
FIG. 2 is a network overall structure of a driving model GAN of the present invention;
FIG. 3 is a detailed structure of the discriminator and generator of the present invention;
fig. 4 is a network overall structure of the CNN of the present invention;
fig. 5 is a network specific structure of CNN of the present invention;
FIG. 6 is a loss function of the discriminator of the present invention;
FIG. 7 is a comparison of the outputs of the generators for different training runs in accordance with the present invention;
FIG. 8 is a loss function of the CNN of the present invention;
FIG. 9 is a time-frequency spectrum comparison before and after sea clutter suppression according to the present invention.
Detailed Description
In order to better understand the purpose, structure and function of the present invention, the sea clutter modeling and suppressing method under data model dual drive according to the present invention is further described in detail below with reference to the accompanying drawings.
Example 1
The embodiment provides a method for modeling and suppressing sea clutter under dual drive of a data model, as shown in fig. 1, fig. 2, fig. 3, fig. 4 and fig. 5, comprising the following steps:
building a GAN network as a driving model for sea clutter modeling, and adopting Gaussian white noise data, sea clutter simulation data which are generated by using a sphere invariant random process SIRP method and meet K distribution, and actually measured sea clutter data as driving data for sea clutter modeling;
the network overall structure of the driving model GAN is shown in fig. 2. The driving data needs to be preprocessed by using short-time Fourier transform (STFT), and the calculation formula of the STFT specifically comprises the following steps:
wherein y (u) is an initial signal, g (u) is a window function, superscript x is a complex conjugate, t represents a central position of the window function, u represents time domain time, and f represents frequency; STFTy(t, f) is the STFT of the initial signal y (u). And respectively performing STFT on the three data through a calculation formula to obtain respective time-frequency spectrum data sets, and taking the three data sets as driving data of the text.
Secondly, performing model training of a generator and a discriminator of the GAN;
the specific structure of the discriminator and the generator is shown in fig. 3. In each layer network, I represents the input data dimension, O represents the output data dimension, k represents the size of the convolution kernel, and s represents the moving step of the convolution kernel. The specific training mode of the GAN model is divided into the following two parts:
training 1: individual training discriminator
1) Obtaining a time frequency spectrum by using 500 parts of white noise sample data meeting Gaussian distribution and 500 parts of sea clutter simulation sample data meeting K distribution through STFT;
2) mixing the 1000 parts of time-frequency spectrum image samples, respectively constructing a training set and a test set according to the proportion of 8:2, setting an image label of Gaussian white noise to be 0, and setting an image label of sea clutter simulation data to be 1;
3) and taking the training set as the input R of the discriminator, inputting the training set into the discriminator to train the discriminator until the discriminator is converged, and then taking the test set as the input R of the discriminator to test the convergence effect of the discriminator.
Training 2: simultaneous training generator and discriminator
1) Obtaining a time frequency spectrum from 500 parts of actually measured sea clutter data through STFT;
2) constructing the 500 time-frequency spectrum image samples into a data set, and setting an image label as 1;
3) taking the data set as an input R of a discriminator, and taking a random sequence satisfying Gaussian distribution as an input Z of a generator;
4) training iteration of the discriminator and the generator is alternately carried out according to the sequence of firstly training the discriminator for 1 time and then training the generator for 1 time, and after training is finished, subsequent comparison simulation is carried out on the time-frequency spectrum image of the sea clutter generated by the generator and the actually-measured time-frequency spectrum image of the sea clutter.
Step three, building a CNN-based sea clutter suppression model, and using sea clutter data generated by the GAN model as an input data set of the CNN; the overall network structure of the CNN is shown in fig. 4.
Step four, performing CNN model training;
the specific network structure of CNN is shown in fig. 5, and the specific training mode is as follows:
1) adding 500 parts of sea clutter time-frequency spectrum images generated by a generator of the GAN with time-frequency spectrums of target echoes of three different Doppler frequencies to obtain data sets of radar receiving signals under the three different Doppler frequencies, respectively constructing a training set and a testing set for each data set according to the proportion of 8:2, and setting labels as the time-frequency spectrums of the target echoes of the Doppler frequencies corresponding to each data set;
2) inputting the training set into the CNN for iterative training, and after the training is finished, inputting the test set to obtain a time-frequency spectrum of the radar receiving signal after the sea clutter component is suppressed.
And fifthly, measuring model performance, measuring the sea clutter modeling simulation effect of the GAN model by adopting MSD (maximum-resolution) test, and testing the sea clutter suppression effect of the CNN model by comparing the frequency spectrums of the radar before and after the CNN model when receiving signals.
The formula for the MSD test is defined as
Wherein p ise(xk) Probability distribution function PDF, p for actually measured sea clutter datat(xk) The PDF of the sea clutter model generated for the algorithm, n, represents the length of the selected data sequence. The smaller the MSD value is, the greater the similarity of the two distributions is, and the better the prediction effect of the algorithm model is.
First, the GAN model is subjected to simulation training of training 1, and the convergence of the loss function of the discriminator is shown in fig. 6. It can be seen that the arbiter has converged.
After the discriminator is converged, the GAN model is subjected to simulation training of training 2. When the values of the Epoch in the training round are {1,10,20,30,40,50}, respectively, a time-frequency diagram of the sea clutter data generated by the generator is shown in fig. 7. It can be found that the feature learning and expression capability of the generator for the sea clutter is continuously improved.
The text is used for training 500 parts of measured data of the GAN, and meanwhile, the text is used as input data of three sea clutter simulation algorithms, namely a sea clutter simulation algorithm based on the measured data, a sea clutter simulation algorithm based on a Back Propagation Neural Network (BPNN) and a sea clutter simulation algorithm based on a Gated Recurrent Neural Network (GRNN), so that 500 parts of sea clutter simulation data are generated.
500 measured data are again taken as a control group for comparative analysis of the MSD test results of the algorithm and the three algorithms. The average results of the MSD test of these four algorithms with the control are shown in table 1. It can be seen that the MSD mean of the algorithm herein is the smallest, indicating that the sea clutter simulation data generated by the algorithm herein fits the measured data the highest. Since the sea clutter data generated by the four algorithms are not generated by the control group, the sea clutter simulation algorithm can be used for fitting more actually measured sea clutter data, and has the best expansibility.
TABLE 1 MSD comparison of sea clutter simulation algorithm to control group
| Algorithm | Text algorithm | Measured data algorithm | BPNN algorithm | GRNN algorithm |
| MSD | 1.28×10-5 | 5.35×10-4 | 4.49×10-5 | 1.62×10-5 |
The sea clutter data generated by the algorithm is used for constructing a radar receiving signal data set, simulation training is carried out on CNN, and FIG. 8 shows the condition of loss function convergence during CNN training under three different Doppler frequencies.
After the network training of the CNN is completed, the input test set can suppress the sea clutter component in the frequency spectrum when the radar receives the signal, and fig. 9 is a comparison graph before and after the suppression of the sea clutter component obtained under three different doppler frequencies. It can be found that the CNN model herein can effectively suppress the sea clutter component in the radar reception signal.
The examples described herein are merely illustrative of the preferred embodiments of the present invention and do not limit the spirit and scope of the present invention, and various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the design concept of the present invention shall fall within the protection scope of the present invention.
Claims (6)
1. A sea clutter modeling and inhibiting method under data model dual drive is characterized by comprising the following steps:
building a GAN network as a driving model for sea clutter modeling, and adopting Gaussian white noise data, sea clutter simulation data which are generated by using a sphere invariant random process SIRP method and meet K distribution, and actually measured sea clutter data as driving data for sea clutter modeling;
secondly, performing model training of a generator and a discriminator of the GAN;
step three, building a CNN-based sea clutter suppression model, and using sea clutter data generated by the GAN model as an input data set of the CNN;
step four, performing CNN model training;
and fifthly, measuring model performance, measuring the sea clutter modeling simulation effect of the GAN model by adopting MSD (maximum-resolution) test, and testing the sea clutter suppression effect of the CNN model by comparing the frequency spectrums of the radar before and after the CNN model when receiving signals.
2. The method for modeling and suppressing the sea clutter under the dual drive of the data model according to claim 1, wherein the overall structure of the driving model GAN network in the step one comprises a generator G and a discriminator D, wherein the input of the generator G is Z, and after Z is input into the generator G, sea clutter data G (Z) which do not exist in the real world are generated, and then the sea clutter data true and false are judged through the discriminator D; through continuous network iterative training, the generator G and the discriminator D respectively update the network parameters thereof, so that the loss function is minimum, and finally a Nash equilibrium state is achieved, the GAN model at the moment reaches an optimal state, and the sea clutter data generated by the generator G is closest to the real sea clutter data; the driving data needs to be preprocessed by using short-time Fourier transform (STFT), and the calculation formula of the STFT specifically comprises the following steps:
wherein y (u) is an initial signal, g (u) is a window function, superscript x is a complex conjugate, t represents a central position of the window function, u represents time domain time, and f represents frequency; STFTy(t, f) is the STFT of the initial signal y (u); and respectively performing STFT on the three driving data through a calculation formula to obtain respective time-frequency spectrum data sets, and taking the three driving data sets as driving data for sea clutter modeling.
3. The method according to claim 1, wherein the discriminator in the second step includes a convolution layer and a full-link layer, wherein a Dropout layer for preventing overfitting is added, and the generator includes a full-link layer and an anti-convolution layer; the specific training mode of the GAN model is divided into the following two parts:
training 1: training a discriminator independently;
1) obtaining a time frequency spectrum by using 500 parts of white noise sample data meeting Gaussian distribution and 500 parts of sea clutter simulation sample data meeting K distribution through STFT;
2) mixing the 1000 parts of time-frequency spectrum image samples, respectively constructing a training set and a test set according to the proportion of 8:2, setting an image label of Gaussian white noise to be 0, and setting an image label of sea clutter simulation data to be 1;
3) inputting the training set as the input R of a discriminator to the discriminator for discriminant training until the discriminator is converged, and then testing the convergence effect of the discriminator by using the test set as the input R of the discriminator;
training 2: training a generator and a discriminator at the same time;
1) obtaining a time frequency spectrum from 500 parts of actually measured sea clutter data through STFT;
2) constructing the 500 time-frequency spectrum image samples into a data set, and setting an image label as 1;
3) taking the data set as an input R of a discriminator, and taking a random sequence satisfying Gaussian distribution as an input Z of a generator;
4) training iteration of the discriminator and the generator is alternately carried out according to the sequence of firstly training the discriminator for 1 time and then training the generator for 1 time, and after training is finished, subsequent comparison simulation is carried out on the time-frequency spectrum image of the sea clutter generated by the generator and the actually-measured time-frequency spectrum image of the sea clutter.
4. The method according to claim 1, wherein the CNN network in step three includes a data input layer, a convolutional layer and an anti-convolutional layer; the input data set is subjected to convolution layer, data characteristics in the time frequency spectrum are extracted, and then the time frequency spectrum is subjected to characteristic reconstruction through the deconvolution layer.
5. The method according to claim 1, wherein the model training of the CNN is performed in step four, the CNN includes two convolutional layers and three anti-convolutional layers, and in order to accelerate the training of the network, a batch normalization layer is added after both the convolutional layers and the anti-convolutional layers; the specific training mode is as follows:
1) adding 500 parts of sea clutter time-frequency spectrum images generated by a generator of the GAN with time-frequency spectrums of target echoes of three different Doppler frequencies to obtain data sets of radar receiving signals under the three different Doppler frequencies, respectively constructing a training set and a testing set for each data set according to the proportion of 8:2, and setting labels as the time-frequency spectrums of the target echoes of the Doppler frequencies corresponding to each data set;
2) inputting the training set into the CNN for iterative training, and after the training is finished, inputting the test set to obtain a time-frequency spectrum of the radar receiving signal after the sea clutter component is suppressed.
6. The method according to claim 1, wherein the formula of the MSD test in step five is defined as
Wherein p ise(xk) Probability distribution function PDF, p for actually measured sea clutter datat(xk) A PDF of the sea clutter model generated for the algorithm, n representing the length of the selected data sequence; the smaller the MSD value is, the greater the similarity of the two distributions is, and the better the prediction effect of the algorithm model is.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111419764.0A CN114117912A (en) | 2021-11-26 | 2021-11-26 | Sea clutter modeling and inhibiting method under data model dual drive |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111419764.0A CN114117912A (en) | 2021-11-26 | 2021-11-26 | Sea clutter modeling and inhibiting method under data model dual drive |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN114117912A true CN114117912A (en) | 2022-03-01 |
Family
ID=80370017
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111419764.0A Pending CN114117912A (en) | 2021-11-26 | 2021-11-26 | Sea clutter modeling and inhibiting method under data model dual drive |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114117912A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114970378A (en) * | 2022-08-01 | 2022-08-30 | 青岛国数信息科技有限公司 | Sea clutter sample library construction method based on GAN network |
| CN115372960A (en) * | 2022-07-11 | 2022-11-22 | 西北工业大学 | Sky wave radar ground and sea clutter data enhancement method for improving generation of countermeasure network |
| CN117217103A (en) * | 2023-11-09 | 2023-12-12 | 南京航空航天大学 | Satellite-borne SAR sea clutter generation method and system based on multi-scale attention mechanism |
-
2021
- 2021-11-26 CN CN202111419764.0A patent/CN114117912A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115372960A (en) * | 2022-07-11 | 2022-11-22 | 西北工业大学 | Sky wave radar ground and sea clutter data enhancement method for improving generation of countermeasure network |
| CN115372960B (en) * | 2022-07-11 | 2024-05-10 | 西北工业大学 | Method for enhancing sky-wave radar land-sea clutter data of improved generation countermeasure network |
| CN114970378A (en) * | 2022-08-01 | 2022-08-30 | 青岛国数信息科技有限公司 | Sea clutter sample library construction method based on GAN network |
| CN114970378B (en) * | 2022-08-01 | 2022-10-25 | 青岛国数信息科技有限公司 | Sea clutter sample library construction method based on GAN network |
| CN117217103A (en) * | 2023-11-09 | 2023-12-12 | 南京航空航天大学 | Satellite-borne SAR sea clutter generation method and system based on multi-scale attention mechanism |
| CN117217103B (en) * | 2023-11-09 | 2024-03-15 | 南京航空航天大学 | Satellite-borne SAR sea clutter generation method and system based on multi-scale attention mechanism |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Shao et al. | Deep fusion for radar jamming signal classification based on CNN | |
| CN114117912A (en) | Sea clutter modeling and inhibiting method under data model dual drive | |
| Yan et al. | Robust target detection within sea clutter based on graphs | |
| CN108008385B (en) | Interference environment ISAR high-resolution imaging method based on management loading | |
| CN111965632B (en) | Radar target detection method based on Riemann manifold dimensionality reduction | |
| CN111722199A (en) | Radar signal detection method based on convolutional neural network | |
| CN112859014A (en) | Radar interference suppression method, device and medium based on radar signal sorting | |
| CN109934101A (en) | Radar clutter recognition method based on convolutional neural networks | |
| CN114114166A (en) | Radar pulse de-interlacing method based on DTM algorithm | |
| KR102407835B1 (en) | Method and apparatus for classifying pulse radar signal properties based on machine learning | |
| CN116047427A (en) | Small sample radar active interference identification method | |
| Wang et al. | A robust constant false alarm rate detector based on the Bayesian estimator for the non-homogeneous Weibull clutter in HFSWR | |
| CN117368877A (en) | Radar image clutter suppression and target detection method based on generation countermeasure learning | |
| KR102407834B1 (en) | Method and apparatus for property-based classification of long-pulse radar signals | |
| Wang et al. | Through-wall human motion representation via autoencoder-self organized mapping network | |
| CN110850421A (en) | Underwater target detection method based on space-time adaptive processing of reverberation symmetric spectrum | |
| Ma et al. | Radar working mode recognition based on hierarchical feature representation and clustering | |
| Fan et al. | Weak target detection based on deep neural network under sea clutter background | |
| Guo et al. | Target depth estimation by deep neural network based on acoustic interference structure in deep water | |
| Zhang et al. | A Fast IAA–Based SR–STAP Method for Airborne Radar | |
| Tian et al. | Multiresolution jamming recognition with few-shot learning | |
| CN110208770B (en) | Deep neural network-based radar waveform design method | |
| Chen et al. | Single-channel Blind Source Separation Algorithm Based on Water Area Noise Characteristics | |
| Zhang et al. | Radar target recognition based on complex HRRP using convolutional neural network | |
| Wang et al. | Weather Radar High-Resolution Spectral Moment Estimation Using Bidirectional Extreme Learning Machine |
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 |