CN112612023A - Radar target identification method and computer readable storage medium - Google Patents
Radar target identification method and computer readable storage medium Download PDFInfo
- Publication number
- CN112612023A CN112612023A CN202011466447.XA CN202011466447A CN112612023A CN 112612023 A CN112612023 A CN 112612023A CN 202011466447 A CN202011466447 A CN 202011466447A CN 112612023 A CN112612023 A CN 112612023A
- Authority
- CN
- China
- Prior art keywords
- radar target
- target data
- real
- data
- simulated
- 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
- 238000000034 method Methods 0.000 title claims abstract description 36
- 238000004088 simulation Methods 0.000 claims abstract description 64
- 238000013145 classification model Methods 0.000 claims abstract description 48
- 238000013526 transfer learning Methods 0.000 claims abstract description 13
- 238000012549 training Methods 0.000 claims description 44
- 239000011159 matrix material Substances 0.000 claims description 25
- 238000005457 optimization Methods 0.000 claims description 20
- 238000000605 extraction Methods 0.000 claims description 13
- 238000005293 physical law Methods 0.000 claims description 10
- 230000007613 environmental effect Effects 0.000 claims description 9
- 238000003062 neural network model Methods 0.000 claims description 7
- 238000011176 pooling Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 claims description 3
- 238000004590 computer program Methods 0.000 claims description 3
- 238000013527 convolutional neural network Methods 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims 1
- 238000005516 engineering process Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 239000008186 active pharmaceutical agent Substances 0.000 description 3
- 230000008901 benefit Effects 0.000 description 3
- 238000001514 detection method Methods 0.000 description 2
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007123 defense Effects 0.000 description 1
- 239000000284 extract Substances 0.000 description 1
- 238000009432 framing Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000007774 longterm Effects 0.000 description 1
- 230000005012 migration Effects 0.000 description 1
- 238000013508 migration Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S13/00—Systems using the reflection or reradiation of radio waves, e.g. radar systems; Analogous systems using reflection or reradiation of waves whose nature or wavelength is irrelevant or unspecified
- G01S13/88—Radar or analogous systems specially adapted for specific applications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
- G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
- G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
-
- 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)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Artificial Intelligence (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Computer Networks & Wireless Communication (AREA)
- General Health & Medical Sciences (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Biomedical Technology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Electromagnetism (AREA)
- Medical Informatics (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The invention discloses a radar target identification method and a computer readable storage medium, which use real data as guidance to generate simulation data which is close to the real data and has random environment disturbance, and the simulation data is used as a source domain, the real data is used as a target domain, and the accuracy and the robustness of radar target classification model classification are improved through transfer learning, thereby realizing the target identification of various complex situations and high robustness in a real scene.
Description
Technical Field
The present invention relates to the field of radar detection technologies, and in particular, to a radar target identification method and a computer-readable storage medium.
Background
The radar target identification technology can automatically identify the category of the target, assist in estimating the threat of the radar target and improve the early warning detection performance. However, radar target identification is complex, the number of target samples is small, the number of various samples is unbalanced, and various defense means are applied, such as a stealth technology, a complex electromagnetic environment and the like, so that radar target identification becomes more and more difficult, and a perfect theoretical framework and a mature technology are not available so far to accurately identify radar targets.
Disclosure of Invention
The invention provides a radar target identification method and a computer readable storage medium, which are used for solving the problem that the prior art can not accurately identify a radar target.
In a first aspect, the present invention provides a radar target identification method, including:
generating simulated radar target data which is close to the real radar target data and has random environmental disturbance based on the real radar target data;
respectively extracting features of the real radar target data and the simulated radar target data, and constructing a real radar target sample library and a simulated radar target sample library;
pre-training through the simulation radar target sample library to obtain a pre-trained radar target classification model, and performing transfer learning on the pre-trained radar target classification model through the real radar target sample library to obtain a final radar target classification model;
and identifying the radar target to be detected through the final radar target classification model.
Optionally, the generating simulated radar target data with random environmental disturbance close to the real radar target data based on the real radar target data includes: and generating simulated radar target data which is close to the real radar target data and has random environment disturbance through a physical simulation model and a sample optimization model based on the real radar target data, wherein the sample optimization model is a model constructed based on generation of a countermeasure network and a physical law.
Optionally, generating simulated radar target data with random environmental disturbance close to the real radar target data through a physical simulation model and a sample optimization model based on the real radar target data, including: generating a series of simulated radar target data under ideal conditions based on the physical attributes of the real radar target data through the physical simulation model; and generating simulated radar target data which is close to the real radar target data and has environment random disturbance through the sample optimization model.
Optionally, generating simulated radar target data based on the physical simulation model, that is, generating a series of simulated radar target data under an ideal condition with the physical attribute of the real radar target data as guidance, then generating simulated radar target data which is close to the real radar target data and has random environmental disturbance through the sample optimization model, and modifying the simulated radar target data by the sample optimization model according to a judgment result fed back by a discriminator until the discriminator cannot distinguish the real radar target data from the simulated radar target data;
the discriminator is used for judging whether the input radar target data is real radar target data or simulated radar target data and feeding back the judgment result to the generator.
Optionally, respectively performing feature extraction on the real radar target data and the simulated radar target data, and constructing a simulated radar target sample library and a real radar target sample library, including: performing feature extraction on the real radar target data to generate two-dimensional feature matrices, wherein each two-dimensional feature matrix and the category of the corresponding real radar target data form a training sample, and all training samples under the real radar target data form a real radar target sample library; and performing feature extraction on the simulated radar target data to generate two-dimensional feature matrices, wherein each two-dimensional feature matrix and the category of the simulated radar target data corresponding to the two-dimensional feature matrix form a training sample, and all the training samples under the simulated radar target data form the simulated radar target sample library.
Optionally, the pre-training through the simulated radar target sample library to obtain a pre-trained radar target classification model includes: and pre-training a preset full convolution depth network model by using a simulation radar target sample library, inputting a two-dimensional characteristic matrix and a category of the simulation radar target sample library into the full convolution neural network model for pre-training, obtaining parameters of each layer of the network, and obtaining a pre-trained radar target classification model.
Optionally, each convolution layer of the full convolution neural network model uses 3 × 3 small convolution kernels, and a pooling layer is provided after a preset number of convolution layers.
Optionally, the fully convolutional neural network model comprises a deep neural network of 10 convolutional layers, 4 pooling layers, and three fully-connected layers.
Optionally, the performing transfer learning on the pre-trained radar target classification model through the real radar target sample library to obtain a final radar target classification model includes:
and training the pre-trained radar target classification model by using the two-dimensional characteristic matrix and the category of the real radar target sample library until the pre-trained radar target classification model is converged to obtain a final radar target classification model.
In a second aspect, the present invention provides a computer-readable storage medium storing a signal-mapped computer program which, when executed by at least one processor, implements a radar target identification method as described in any one of the above.
The invention has the following beneficial effects:
the method uses real data as guidance to generate simulation data which is close to the real data and has random environment disturbance, uses the simulation data as a source domain and the real data as a target domain, improves the accuracy and robustness of classification of the radar target classification model through transfer learning, and realizes target identification under various complex conditions and high robustness in a real scene.
The foregoing description is only an overview of the technical solutions of the present invention, and the embodiments of the present invention are described below in order to make the technical means of the present invention more clearly understood and to make the above and other objects, features, and advantages of the present invention more clearly understandable.
Drawings
Various other advantages and benefits will become apparent to those of ordinary skill in the art upon reading the following detailed description of the preferred embodiments. The drawings are only for purposes of illustrating the preferred embodiments and are not to be construed as limiting the invention. Also, like reference numerals are used to refer to like parts throughout the drawings. In the drawings:
fig. 1 is a schematic flow chart of a radar target identification method according to a first embodiment of the present invention;
FIG. 2 is a schematic flowchart of a radar target classification model training process according to a first embodiment of the present invention;
fig. 3 is a schematic flowchart of radar target data generation and expansion according to a first embodiment of the present invention.
Detailed Description
The embodiment of the invention provides a radar target identification method based on transfer learning, aiming at the problems of low identification precision and poor robustness caused by the small number of radar target samples and the unbalanced number of various samples, wherein the method uses real data as guidance to generate simulation data which is close to the real data and has random environment disturbance; the method comprises the steps of constructing a deep network model as a classifier, taking simulation data as a source domain and real data as a target domain, improving the accuracy and robustness of deep network model target classification through transfer learning, and realizing target identification under various complex conditions and high robustness in a real scene. The present invention will be described in further detail below with reference to the drawings and examples. It should be understood that the specific embodiments described herein are merely illustrative of the invention and do not limit the invention.
A first embodiment of the present invention provides a radar target identification method, and referring to fig. 1, the method includes:
s101, generating simulated radar target data which are close to real radar target data and have random environmental disturbance based on the real radar target data;
specifically, in the embodiment of the invention, the real radar target data is used as a basis, and the physical simulation model and the sample optimization model are used for generating simulation radar target data which is close to the real radar target data and has random environment disturbance, or the simulation radar target data can be also referred to as radar simulation data for short, and the real radar target data in the embodiment of the invention can also be referred to as radar real data for short;
in specific implementation, the embodiment of the invention generates a series of simulated radar target data under ideal conditions based on the physical attributes of the real radar target data through the physical simulation model; generating simulated radar target data which are close to the real radar target data and have environment random disturbance through the sample optimization model;
it should be noted that the sample optimization model according to the embodiment of the present invention is a model constructed based on the generation of the countermeasure network and the physical law.
S102, respectively extracting characteristics of the real radar target data and the simulated radar target data, and constructing a simulated radar target sample library and a real radar target sample library;
namely, the embodiment of the invention extracts the characteristics of the simulation data and the real data, and constructs the simulation sample library and the real sample library for training.
Specifically, the embodiment of the invention performs feature extraction on the real radar target data to generate two-dimensional feature matrices, each two-dimensional feature matrix and the category of the corresponding real radar target data form a training sample, and all the training samples under the real radar target data form the real radar target sample library; and performing feature extraction on the simulated radar target data to generate two-dimensional feature matrices, wherein each two-dimensional feature matrix and the category of the simulated radar target data corresponding to the two-dimensional feature matrices form a training sample, and all training samples under the simulated radar target data form the simulated radar target sample library.
In practical implementation, any sample in the embodiments of the present invention is in the form of (data, label), and the two-dimensional feature in the embodiments of the present invention is data in the sample, and the category is the label in the sample.
S103, pre-training the simulation radar target sample library to obtain a pre-trained radar target classification model, and performing transfer learning on the pre-trained radar target classification model through the real radar target sample library to obtain a final radar target classification model;
that is to say, in the embodiment of the invention, a simulation radar target sample library is used for pre-training a preset full convolution depth network model, a two-dimensional characteristic matrix and a category of the simulation radar target sample library are input into the full convolution neural network model for pre-training, parameters of each layer of the network are obtained, and a pre-trained radar target classification model is obtained. And then, training the pre-trained radar target classification model by using the two-dimensional characteristic matrix and the category of the real radar target sample library until the pre-trained radar target classification model is converged to obtain a final radar target classification model.
And S104, identifying the radar target to be detected through the final radar target classification model.
Specifically, the embodiment of the invention is based on the training of a radar target classification model of transfer learning, a depth network model is constructed as the radar target classification model, a self-adaptive cross-domain learning method is introduced, a large number of cross-domain learning experiences are utilized, a large number of simulation samples (namely, the simulation radar target data can be also referred to as simulation data for short) are firstly used for pre-training the radar target classification model, then real samples (namely, the real radar target data can be also referred to as real data for short) are used for fine tuning the model, and a cross-domain learning strategy on target recognition is self-adaptively adjusted, so that the radar target classification model with high accuracy and strong robustness is obtained.
In general, the embodiment of the invention uses a limited real sample as an example, generates a large number of simulation samples under uncertain environments, classifies radar targets by constructing a deep network model, and improves the accuracy and robustness of model classification by deep migration learning.
The training process of the radar target classification model according to the embodiment of the invention is shown in fig. 2, and comprises three steps of radar target data generation and expansion based on a limited example, generation of a simulation radar target sample library and a real radar target sample library (hereinafter, referred to as a simulation sample library and a real sample library respectively) and radar target classification model training based on transfer learning.
Specifically, the embodiment of the invention is based on the generation and the expansion of radar target data of a limited example, and aims to generate simulation data close to real data with environment random disturbance; generating a simulation sample library and a real sample library, wherein the purpose of the generation is to extract the characteristics of simulation data and real data and construct the simulation sample library and the real sample library for training; the method comprises the steps of training a radar target classification model based on transfer learning, constructing a depth network model as the radar target classification model, introducing an adaptive cross-domain learning method, utilizing a large amount of cross-domain learning experiences, pre-training the radar target classification model by using a large amount of simulation samples, fine-tuning the model by using real samples, and adaptively adjusting a cross-domain learning strategy on target recognition.
Step S101 of the embodiment of the present invention is specifically to generate a series of simulated radar target data under ideal conditions by using the simulated radar target data generated by the generator based on the physical simulation model and using the physical attributes of the real radar target data as guidance. Generating simulated radar target data which are close to the real radar target data and have environment random disturbance through the sample optimization model, and correcting the simulated radar target data according to a judgment result fed back by a discriminator in the process until the discriminator cannot distinguish the real radar target data from the simulated radar target data; the discriminator is used for judging whether the input radar target data is real radar target data or simulated radar target data and feeding back the judgment result to the generator.
Specifically, as shown in fig. 3, the embodiment of the present invention generates simulation data with reality by using a physical simulation model and a sample optimization model based on a generated countermeasure network and a physical law based on real data. Specifically, the data generation in the embodiment of the present invention includes two steps: firstly, a physical simulation model is used, and a series of radar target simulation data under ideal conditions are generated based on data physical attributes. And then constructing a sample optimization model based on the generation of the countermeasure network and the physical law, and generating simulation data which is close to the real data of the radar target and has environment random disturbance. The input of the sample optimization model is the simulated radar target data, the physical law (including the physical laws of the data physical attributes and the environmental attributes) and the real radar target data of the known category in step S101, and the simulated data with authenticity is output.
In specific implementation, the sample optimization model in the embodiment of the invention is composed of a generator and a discriminator, wherein the generator inputs a physical law and simulation data in the step S101, the physical law restrains the generation process, and the generator can output high-quality simulation data which not only accords with the actual physical law, but also is similar to real data. The ultimate goal of the generator is to make it impossible for the arbiter to tell whether the input data is real data or simulated data.
The input of the discriminator in the embodiment of the invention is simulation data generated by the generator and real data of known types, and the discriminator judges whether the input data is the real data or the simulation data, and the main aim of the discriminator is to accurately distinguish whether the input data is the real data or the simulation data.
The training process for generating the confrontation network model in the embodiment of the invention is as follows:
step one, a generator is fixed, and the discriminator is trained, so that the discriminator can more accurately discriminate whether input data is real data or simulation data.
And step two, fixing a discriminator to train the generator so that the data distribution of the generated data is close to the real data.
And alternately performing the first step and the second step until the generated data distribution and the real data distribution are basically consistent, and judging whether the input data is simulation data or real data if the model is in Nash equilibrium.
In the embodiment of the invention, the simulation sample library and the real sample library in the step S102 mainly perform feature extraction on original data to generate a two-dimensional feature matrix, one two-dimensional feature matrix and a corresponding class thereof form a training sample, and a large number of training samples form the sample library. And performing feature extraction on the simulation data and the real data by using the same method to generate a simulation data and real data sample library.
Taking the RCS data as an example, framing, feature extraction, and feature combination are required to be performed on the data to generate a feature sample for training. Each radar target comprises a plurality of RCS data with equal length, firstly, each RCS data is framed, namely, each RCS data is intercepted for one frame every t seconds, the overlapping part between the frames is delta t seconds, and then, position characteristic parameters (such as mean value, maximum value, minimum value and median), spread characteristic parameters (such as range, variance, standard mean deviation, variation coefficient and the like) and other characteristic parameters (such as k-order origin moment, k-order center moment and the like) and other characteristics are extracted. For a sample containing m RCS data with equal length, after feature extraction, a feature matrix of corresponding features is obtained, and features are carried out according to the mean value
For example, the final feature matrix is:
wherein
j is the number of frames divided per RCS data. Thus, for the plurality of features described above, a plurality of two-dimensional feature matrices will be generated. When the characteristic combination is carried out, corresponding rows of different characteristic matrixes are connected to form a two-dimensional characteristic matrix of combined characteristics. Feature matrix by mean
Range feature matrix
The standard deviation feature matrix S is taken as an example, and the combined feature matrix F is:
finally, each training sample consists of a feature matrix F and a corresponding target class. And finally obtaining a simulation radar target sample library and a real radar target sample library by carrying out feature extraction on all RCS simulation data and RCS real data.
The radar target classification model according to the embodiment of the present invention will be explained and illustrated in detail below:
the method comprises the steps of pre-training a preset full-convolution depth network model by using a simulated radar target sample library, inputting a two-dimensional characteristic matrix and a category of the simulated radar target sample library into the full-convolution neural network model for pre-training, obtaining parameters of each layer of the network, and obtaining a pre-trained radar target classification model.
Specifically, a deep full convolution model is designed in the radar target classification model in the embodiment of the invention, the input of the model is two-dimensional characteristic data, and the output is a predicted radar target category. The model uses 3 × 3 small convolution kernels for each convolution layer, and adds pooling layers after a plurality of convolution layers, thereby ensuring that the deep convolutional neural network can excellently express long-term correlation of features by accumulating a very large number of convolution pooling layer pairs. The deep fully-convolutional model structure is a deep neural network comprising 10 convolutional layers, 4 pooling layers and three fully-connected layers.
In addition, the radar target classification model in the embodiment of the invention is trained by adopting a cross-domain transfer learning method, and a radar target simulation sample is used as a source sample domain DSThe type of the simulation sample is the corresponding task TS(ii) a Taking radar target real samples as target sample domain DTThe type of the real sample is a learning task TTAt DS≠DTOr TS≠TTBy using DSAnd TSThe learned knowledge, such as the determined corresponding characteristic parameters or weights, etc., to improve the target prediction function in the target sample domain DTThe predictive power of (1).
Specifically, a full convolution depth network model is pre-trained by using a radar target simulation sample library, and D is obtainedSThe two-dimensional features and the categories are input into the deep convolution neural network model for pre-training to obtain parameters of each layer of the network, and a pre-trained classification model is obtained. And then setting the parameters of the lower network to be fixed. And finally, training the rest layers by using a radar target real sample library until the model is converged to obtain a radar target classification model.
In general, compared with the existing radar target identification technology, the embodiment of the invention takes real data as guidance, and uses the generated countermeasure network to generate simulation data with authenticity, so that the difference between the simulation data and the real data can be effectively reduced, and the accuracy of radar target identification is further improved.
A second embodiment of the present invention provides a computer-readable storage medium storing a signal-mapped computer program which, when executed by at least one processor, implements the radar target identification method of any one of the first embodiments of the present invention.
The relevant content of the embodiments of the present invention can be understood by referring to the first embodiment of the present invention, and will not be discussed in detail herein.
Although the preferred embodiments of the present invention have been disclosed for illustrative purposes, those skilled in the art will appreciate that various modifications, additions and substitutions are possible, and the scope of the invention should not be limited to the embodiments described above.
Claims (10)
1. A radar target identification method is characterized by comprising the following steps:
generating simulated radar target data which is close to the real radar target data and has random environmental disturbance based on the real radar target data;
respectively extracting features of the real radar target data and the simulated radar target data, and constructing a simulated radar target sample library and a real radar target sample library;
pre-training through the simulation radar target sample library to obtain a pre-trained radar target classification model, and performing transfer learning on the pre-trained radar target classification model through the real radar target sample library to obtain a final radar target classification model;
and identifying the radar target to be detected through the final radar target classification model.
2. The method of claim 1, wherein generating simulated radar target data with random environmental perturbations proximate to the real radar target data based on the real radar target data comprises:
and generating simulated radar target data which is close to the real radar target data and has random environment disturbance through a physical simulation model and a sample optimization model based on the real radar target data, wherein the sample optimization model is a model constructed based on generation of a countermeasure network and a physical law.
3. The method of claim 2, wherein generating simulated radar target data with random environmental perturbations proximate to the real radar target data based on the real radar target data via a physical simulation model and a sample optimization model comprises:
generating a series of simulated radar target data under ideal conditions based on the physical attributes of the real radar target data through the physical simulation model;
and generating simulated radar target data which is close to the real radar target data and has environment random disturbance through the sample optimization model.
4. The method of claim 3,
generating simulated radar target data through the physical simulation model, namely generating a series of simulated radar target data under an ideal condition by taking the physical attribute of the real radar target data as guidance, then generating simulated radar target data which is provided with environment random disturbance and is close to the real radar target data through the sample optimization model, and correcting the simulated radar target data by a generator of the sample optimization model according to a judgment result fed back by a discriminator until the discriminator cannot distinguish the real radar target data from the simulated radar target data;
the discriminator is used for judging whether the input radar target data is real radar target data or simulated radar target data and feeding back the judgment result to the generator.
5. The method of claim 1, wherein the performing feature extraction on the real radar target data and the simulated radar target data respectively to construct a real radar target sample library and a simulated radar target sample library comprises:
performing feature extraction on the real radar target data to generate two-dimensional feature matrices, wherein each two-dimensional feature matrix and the category of the corresponding real radar target data form a training sample, and all training samples under the real radar target data form a real radar target sample library;
and performing feature extraction on the simulated radar target data to generate two-dimensional feature matrices, wherein each two-dimensional feature matrix and the category of the simulated radar target data corresponding to the two-dimensional feature matrix form a training sample, and all the training samples under the simulated radar target data form the simulated radar target sample library.
6. The method of claim 1, wherein the pre-training through the simulated radar target sample library to obtain a pre-trained radar target classification model comprises:
and pre-training a preset full convolution depth network model by using a simulation radar target sample library, inputting a two-dimensional characteristic matrix and a category of the simulation radar target sample library into the full convolution neural network model for pre-training, obtaining parameters of each layer of the network, and obtaining a pre-trained radar target classification model.
7. The method of claim 6,
each convolution layer of the full convolution neural network model uses 3 x 3 small convolution kernels, and a pooling layer is provided after a preset number of convolution layers.
8. The method of claim 7,
the fully convolutional neural network model comprises a deep neural network of 10 convolutional layers, 4 pooling layers and three fully-connected layers.
9. The method according to claim 6, wherein the performing transfer learning on the pre-trained radar target classification model through the real radar target sample library to obtain a final radar target classification model comprises:
and training the pre-trained radar target classification model by using the two-dimensional characteristic matrix and the category of the real radar target sample library until the pre-trained radar target classification model is converged to obtain a final radar target classification model.
10. A computer-readable storage medium, characterized in that it stores a computer program of signal mapping, which when executed by at least one processor, implements the radar target recognition method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011466447.XA CN112612023A (en) | 2020-12-14 | 2020-12-14 | Radar target identification method and computer readable storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202011466447.XA CN112612023A (en) | 2020-12-14 | 2020-12-14 | Radar target identification method and computer readable storage medium |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN112612023A true CN112612023A (en) | 2021-04-06 |
Family
ID=75233643
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202011466447.XA Pending CN112612023A (en) | 2020-12-14 | 2020-12-14 | Radar target identification method and computer readable storage medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112612023A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822892A (en) * | 2021-11-24 | 2021-12-21 | 腾讯科技（深圳）有限公司 | Evaluation method, device and equipment of simulated radar and computer program product |
| CN117237743A (en) * | 2023-11-09 | 2023-12-15 | 深圳爱莫科技有限公司 | Small sample quick-elimination product identification method, storage medium and processing equipment |
Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101794396A (en) * | 2010-03-25 | 2010-08-04 | 西安电子科技大学 | System and method for recognizing remote sensing image target based on migration network learning |
| CN108985268A (en) * | 2018-08-16 | 2018-12-11 | 厦门大学 | Conclusion formula high Resolution Range Profile Identification of Radar method based on depth migration study |
| CN109117793A (en) * | 2018-08-16 | 2019-01-01 | 厦门大学 | Direct-push high Resolution Range Profile Identification of Radar method based on depth migration study |
| CN109254274A (en) * | 2018-11-23 | 2019-01-22 | 哈尔滨工程大学 | A kind of Radar emitter discrimination method based on Fusion Features |
| CN110472627A (en) * | 2019-07-02 | 2019-11-19 | 五邑大学 | One kind SAR image recognition methods end to end, device and storage medium |
| CN110610207A (en) * | 2019-09-10 | 2019-12-24 | 重庆邮电大学 | Small sample SAR image ship classification method based on transfer learning |
| CN110824450A (en) * | 2019-10-15 | 2020-02-21 | 中国人民解放军国防科技大学 | Radar target HRRP robust identification method in noise environment |
| CN112069898A (en) * | 2020-08-05 | 2020-12-11 | 中国电子科技集团公司电子科学研究院 | Method and device for recognizing human face group attribute based on transfer learning |
-
2020
- 2020-12-14 CN CN202011466447.XA patent/CN112612023A/en active Pending
Patent Citations (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101794396A (en) * | 2010-03-25 | 2010-08-04 | 西安电子科技大学 | System and method for recognizing remote sensing image target based on migration network learning |
| CN108985268A (en) * | 2018-08-16 | 2018-12-11 | 厦门大学 | Conclusion formula high Resolution Range Profile Identification of Radar method based on depth migration study |
| CN109117793A (en) * | 2018-08-16 | 2019-01-01 | 厦门大学 | Direct-push high Resolution Range Profile Identification of Radar method based on depth migration study |
| CN109254274A (en) * | 2018-11-23 | 2019-01-22 | 哈尔滨工程大学 | A kind of Radar emitter discrimination method based on Fusion Features |
| CN110472627A (en) * | 2019-07-02 | 2019-11-19 | 五邑大学 | One kind SAR image recognition methods end to end, device and storage medium |
| CN110610207A (en) * | 2019-09-10 | 2019-12-24 | 重庆邮电大学 | Small sample SAR image ship classification method based on transfer learning |
| CN110824450A (en) * | 2019-10-15 | 2020-02-21 | 中国人民解放军国防科技大学 | Radar target HRRP robust identification method in noise environment |
| CN112069898A (en) * | 2020-08-05 | 2020-12-11 | 中国电子科技集团公司电子科学研究院 | Method and device for recognizing human face group attribute based on transfer learning |
Non-Patent Citations (2)
| Title |
|---|
| 潘宗序: "基于深度学习的雷达图像目标识别研究进展", 《中国科学:信息科学》, vol. 49, no. 12, pages 1626 - 1639 * |
| 郭继光: "基于迁移学习的小样本弹道导弹目标识别方法", 《中国电子科学研究院学报》, vol. 17, no. 07, pages 626 - 634 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822892A (en) * | 2021-11-24 | 2021-12-21 | 腾讯科技（深圳）有限公司 | Evaluation method, device and equipment of simulated radar and computer program product |
| CN117237743A (en) * | 2023-11-09 | 2023-12-15 | 深圳爱莫科技有限公司 | Small sample quick-elimination product identification method, storage medium and processing equipment |
| CN117237743B (en) * | 2023-11-09 | 2024-02-27 | 深圳爱莫科技有限公司 | Small sample quick-elimination product identification method, storage medium and processing equipment |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111753881B (en) | Concept sensitivity-based quantitative recognition defending method against attacks | |
| CN113033520B (en) | Tree nematode disease wood identification method and system based on deep learning | |
| CN110866287B (en) | Point attack method for generating countercheck sample based on weight spectrum | |
| CN110276377A (en) | A kind of confrontation sample generating method based on Bayes's optimization | |
| CN103839073B (en) | Polarization SAR image classification method based on polarization features and affinity propagation clustering | |
| CN111754519B (en) | Class activation mapping-based countermeasure method | |
| CN111311702B (en) | Image generation and identification module and method based on BlockGAN | |
| CN112612023A (en) | Radar target identification method and computer readable storage medium | |
| CN112446357B (en) | SAR automatic target recognition method based on capsule network | |
| CN106815323A (en) | A kind of cross-domain vision search method based on conspicuousness detection | |
| CN118230175B (en) | Real estate mapping data processing method and system based on artificial intelligence | |
| Xiaorui et al. | A preliminary study on targets association algorithm of radar and AIS using BP neural network | |
| CN113487600A (en) | Characteristic enhancement scale self-adaptive sensing ship detection method | |
| CN109101984B (en) | Image identification method and device based on convolutional neural network | |
| CN103971362B (en) | SAR image change-detection based on rectangular histogram and elite genetic algorithm for clustering | |
| CN115861595B (en) | Multi-scale domain self-adaptive heterogeneous image matching method based on deep learning | |
| CN112560034A (en) | Malicious code sample synthesis method and device based on feedback type deep countermeasure network | |
| CN116503685A (en) | optical-ISAR image conversion method and device based on hierarchical domain prior | |
| CN113553708B (en) | Method and device for tracing key influence factors of simulation model | |
| CN113011446B (en) | Intelligent target recognition method based on multi-source heterogeneous data learning | |
| CN113971442A (en) | Method and system for generating universal countermeasure disturbance based on self-walking learning | |
| CN113449631A (en) | Image classification method and system | |
| CN117689890B (en) | Semantic segmentation method, device and storage medium based on fine and fog scene | |
| CN116150038B (en) | Neuron sensitivity-based white-box test sample generation method | |
| CN116664876B (en) | Method and system for determining feature sample priority based on significant feature variation |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210406 |