CN109389058A - Sea clutter and noise signal classification method and system - Google Patents
Sea clutter and noise signal classification method and system Download PDFInfo
- Publication number
- CN109389058A CN109389058A CN201811113859.8A CN201811113859A CN109389058A CN 109389058 A CN109389058 A CN 109389058A CN 201811113859 A CN201811113859 A CN 201811113859A CN 109389058 A CN109389058 A CN 109389058A
- Authority
- CN
- China
- Prior art keywords
- training
- convolutional neural
- neural networks
- training sample
- radar
- 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.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biophysics (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Signal Processing (AREA)
- Radar Systems Or Details Thereof (AREA)
Abstract
The embodiment of the invention provides a kind of sea clutter and noise signal classification method and system, method includes: will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;The classification results of convolutional neural networks output after obtaining the training.A kind of sea clutter provided in an embodiment of the present invention and noise signal classification method and system, classify to sea clutter and noise signal by using the convolutional neural networks after training, have high classification accuracy.
Description
Technical field
The present embodiments relate to signal processing technology field more particularly to a kind of sea clutter and noise signal classification methods
And system.
Background technique
Sea clutter modeling and the target detection problems in sea clutter, are always the hot and difficult issue problem studied both at home and abroad,
The statistical characteristic value of description sea clutter mainly have amplitude characteristic, spectral characteristic and when/sky correlation properties etc..For its amplitude equalizing value
Characteristic, common model have GIT model, TSC model, HYB model, SIT model, NRL model etc., for its amplitude distribution characteristic,
Traditional probability Distribution Model has rayleigh distributed, logarithm normal distribution, Wei Buer distribution and K distribution etc.;It is how general for sea clutter
Spectrum Modeling Research is strangled, common model has Lee spectrum model, Walker model, Ward model etc., and has developed a system on this basis
Column constant false alarm (CFAR, Constant False Alarm Rate) detector, however practical sea clutter often deviates from the system of hypothesis
Distributed model is counted, the decline of constant false alarm detector performance is serious, seriously affects target detection.
The key problem of target detection is to complete the classification of sea clutter and noise, therefore, accurately by sea clutter and noise
Modulation recognition, and the miscellaneous suppressing method of making an uproar for being suitable for this background is chosen automatically, it can be provided for naval target detection performance
Strong guarantee.
Therefore, a kind of sea clutter is needed now with noise signal classification method to solve the above problems.
Summary of the invention
To solve the above-mentioned problems, the embodiment of the present invention provides one kind and overcomes the above problem or at least be partially solved
State the sea clutter and noise signal classification method and system of problem.
The first aspect embodiment of the present invention provides a kind of sea clutter and noise signal classification method, comprising:
It will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;
The classification results of convolutional neural networks output after obtaining the training.
The embodiment of the invention provides a kind of sea clutters and noise signal categorizing system for second aspect, comprising:
Input module, for will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;
Categorization module, for obtaining the classification results of the output of the convolutional neural networks after the training.
The embodiment of the invention provides a kind of electronic equipment for the third aspect, comprising:
Processor, memory, communication interface and bus;Wherein, the processor, memory, communication interface pass through described
Bus completes mutual communication;The memory is stored with the program instruction that can be executed by the processor, the processor
Described program instruction is called to be able to carry out a kind of sea clutter and noise signal classification method described above.
The embodiment of the invention provides a kind of non-transient computer readable storage medium, the non-transient calculating for fourth aspect
Machine readable storage medium storing program for executing stores computer instruction, and the computer instruction makes the computer execute above-mentioned sea clutter and noise letter
Number classification method.
A kind of sea clutter provided in an embodiment of the present invention and noise signal classification method and system, by using by training
Convolutional neural networks afterwards classify to sea clutter and noise signal, have high classification accuracy.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to be briefly described, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without creative efforts, can be with root
Other attached drawings are obtained according to these attached drawings.
Fig. 1 is a kind of sea clutter provided in an embodiment of the present invention and noise signal classification method flow diagram;
Fig. 2 is a kind of sea clutter provided in an embodiment of the present invention and noise signal categorizing system structure chart;
Fig. 3 is the structural block diagram of electronic equipment provided in an embodiment of the present invention.
Specific embodiment
In order to make the object, technical scheme and advantages of the embodiment of the invention clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical solution in the embodiment of the present invention is explicitly described, it is clear that described embodiment is the present invention
A part of the embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not having
Every other embodiment obtained under the premise of creative work is made, shall fall within the protection scope of the present invention.
Currently, flourished deep learning (Deep Learning) since 2006, convolutional neural networks
(Convolutional Neural Networks, CNN) has just shown wide application prospect in field of image processing,
The locality characteristic that data can be made full use of has the characteristics that translation invariant and powerful generalization ability, and shared using weight
Simplify network structure and achieves successful application in terms of the Object identifying in optical imagery field and recognition of face.In radar map
As processing aspect, for diversification data and feature that radar obtains, such as synthetic aperture radar (Synthetic Aperture
Radar, SAR) image, High Range Resolution (High Range Resolution Profiles, HRRP), micro-doppler
(Micro-Doppler) spectrogram, range Doppler (Range-Doppler, R-D) spectrogram etc., CNN is in SAR image target classification
Achievement more abundant is also achieved with the application studies such as identification, track segmentation, time-frequency two-dimensional figure target classification, but existing is ground
Study carefully main or the reflectogram or characteristic pattern that radar obtains are equivalent to a two dimensional image, then is handled.But radar return
Signal is one-dimensional signal, also rarely has research by CNN applied to the processing of one-dimensional radar echo signal at present.
For said circumstances, Fig. 1 is a kind of sea clutter provided in an embodiment of the present invention and noise signal classification method process
Schematic diagram, as shown in Figure 1, comprising:
It 101, will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;
102, the classification results of the convolutional neural networks output after obtaining the training.
It is understood that the embodiment of the present invention will carry out target acquisition background using convolutional neural networks CNN, also
It is the classification for realizing sea clutter signal and noise signal in the one-dimensional echo-signal of radar, to improve the standard of background radar judgement
Exactness, and then improve Methods for Target Detection Probability.It should be noted that the sea clutter and noise signal that provide through the embodiment of the present invention
Classification can be realized and meet/the judgement in down wind clutter region, noise region.
Specifically, in a step 101, the one-dimensional radar signal of radar to be sorted, that is, detections of radar naval target background is passed back
Signal, wherein contain sea clutter signal and noise signal, the embodiment of the present invention is inputted the convolutional Neural net after training
In network, the convolutional neural networks after training are that the embodiment of the present invention is trained in advance, and spy can be extracted from signal to be sorted
Sign, and identify the type of this feature, to be classified.Convolutional neural networks provided in an embodiment of the present invention generally by connecting entirely
Layer (outputting and inputting layer), convolutional layer, pond layer (down-sampling layer) composition, training process can be divided into propagated forward process substantially
And back-propagation process.The fundamental formular of the propagated forward of convolutional neural networks convolutional layer is as follows:
Wherein, f () is neuron excitation function, is linear function, in (1) formulaIt is i-th of neuron of kth layer
Output valve (or activation value),It is the input value of the neuron,For the parameter of the convolutional layer, the neuron of this layer it is public this
It is shared to carry out weight for a parameter.It is the bias term of different neurons.MiIt indicates in preceding layer (- 1 layer of kth) to kth convolutional layer
The input area of i-th of neuron,Indicate the value of neuron in input area.
In the down-sampling layer of convolutional neural networks, i-th of neuron of kth floor is the area n × n to the size of -1 floor of kth
Domain carries out pond, i.e. down-sampling operation is indicated with down (), calculates activation value by the activation primitive of the neuron, specifically
Formula is as follows:
Wherein, the equation left sideIndicate the output valve (or activation value) of l i-th of neuron of layer,InIt is
Input value, MiIt indicates in preceding layer (- 1 layer of kth) to the input area of i-th of neuron of kth convolutional layer,It is the neuron
Parameter,It is biasing.
In convolutional neural networks provided in an embodiment of the present invention, convolutional layer and down-sampling layer alternating, k-th convolutional layer
Next layer (+1 layer of kth) is down-sampling layer.When down-sampling layer carries out pond operation, a neuron can be with one piece of mind of convolutional layer
It is connected through member, (upsampling) is up-sampled to+1 convolutional layer of kth first, keeps its size identical with kth layer, then carry out
Gradient calculates.The operation for up-sampling layer can be realized with Kronecker product:
Wherein up () indicates up-sampling operation, and after above formula operation, a neuron becomes the two-dimensional surface of n × n
Neuronal layers, wherein the parameter of all neurons is identical, it is β, it can thus be concluded that convolutional layer residual error:
Wherein, Hadamard (Hadamard) product is indicated, vector is by element multiplication.
Cost function is further calculated to the gradient of parameter:
Wherein, J be network cost function as BP neural network,Indicate -1 layer of the kth region being convolved
In the output valve of coordinate (u, v),Indicate the residual values of the neuron of kth layer (u, v) coordinate position, which can be preceding
It is calculated into communication process.
Down-sampling layer is similar with convolutional layer, and gradient can be calculated in propagated forward, calculating formula are as follows:
Wherein, down () indicates down-sampling operation, to upper one layerThe down-sampled values of calculating, as l layers of input
Value.Gradient can be calculated later:
Wherein,It is gradient of l layers of (u, the v) coordinate position to parameter beta, l layers can be obtained to coordinate summation
Gradient.The residual values for indicating the neuron of kth layer (u, v) coordinate position can obtain to biasing coordinate summationLadder
Degree.After convolutional layer and the calculating of down-sampling layer gradient finish, using error backpropagation algorithm, neural network is solved.
So in a step 101 the one-dimensional echo-signal of radar to be sorted is inputted that the above process provides and instructed
In the convolutional neural networks perfected, the feature extraction of signal can be both completed.
Further, in a step 102, the classification results of the convolutional neural networks after being trained, i.e., according to step
The feature extracted in rapid 101 is that the one-dimensional echo-signal of radar to be sorted is classified, and exporting its classification results is sea clutter letter
Number either noise signal.According to the statistical result of test sample, method accuracy rate provided in an embodiment of the present invention is very high, energy
It is enough accurately to classify to sea clutter and noise signal.
A kind of sea clutter provided in an embodiment of the present invention and noise signal classification method, by using the volume after training
Product neural network classifies to sea clutter and noise signal, has high classification accuracy.
On the basis of the above embodiments, the convolutional neural networks are specially LeNet.
By the content of above-described embodiment it is found that the embodiment of the invention provides a kind of convolutional neural networks to carry out sea clutter
With noise classification task, it is preferred that convolutional neural networks provided in an embodiment of the present invention are LeNet.
LeNet is applied to one-dimensional Radar Return Sequences by the embodiment of the present invention, need to the network architecture parameters of LeNet into
Input, is adjusted to 1 × 400 sequence, by the reshape process in network structure, becomes 20 × 20 matrix by row modification,
And need to recalculate required network architecture parameters.First layer and third layer are convolutional layer, and convolution kernel size is 4 × 4,
16 and 32 convolution kernels are respectively provided with, full 0 is used and fills, this, which is just determined, does not change Output Size, only changes depth and (divides
Not Wei 16,32).The second layer and the 4th layer are down-sampling layer, and step-length is 2, therefore Output Size is made to reduce half, respectively 10
× 10 and 5 × 5.512) and layer 7 (input finally by full articulamentum, (the 6th layer, inputted as 5 × 5 × 32=800, exports as
512, export as classification number 2) obtain network output.It should be noted that in addition to the input substantially for meeting LeNet network structure is defeated
Out outside relationship, for parameters such as convolution kernel size, quantity, step-lengths, need to be adjusted according to actual training result, to this
Inventive embodiments are not especially limited.
On the basis of the above embodiments, the convolution mind by after the one-dimensional echo-signal input training of radar to be sorted
Through specifically including in network:
The data point of the one-dimensional echo-signal of radar to be sorted is rearranged into the two-dimensional matrix of n*n;
The two-dimensional matrix is inputted in the convolutional neural networks after the training.
It is understood that the embodiment of the present invention was input to neural network is the one-dimensional echo-signal of radar, it is one one
Dimensional signal, but in LeNet network structure, need the point by the one-dimensional radar echo signal of input to be rearranged into the two of a n × n
Matrix is tieed up, to carry out feature extraction by the sliding of LeNet network structure convolution kernel.
On the basis of the above embodiments, in the convolution by after the one-dimensional echo-signal input training of radar to be sorted
Before in neural network, the method also includes:
It obtains discrimination and is greater than the sample data of preset threshold as training sample set;
Preset convolutional neural networks are trained based on the training sample set.
As can be seen from the above embodiments, one-dimensional radar echo signal data satisfaction can use convolutional neural networks and be handled
Precondition, but segment space feature can be destroyed and be increased newly to rearrangement process, influence classification accuracy.Therefore, the present invention is implemented
The sample data that example needs that discrimination is selected to be greater than preset threshold is as training sample set, to reduce data to a certain extent
Classification results are influenced.Discrimination refers to the amplitude discrimination of sea clutter data and noise data in training sample.
On the basis of the above embodiments, preset convolutional neural networks are carried out based on the training sample set described
Before training, the method also includes:
Adjust data sequence length and/or the adjustment training sample set that the training sample concentrates each training sample
In each training sample time-frequency representation form.
By the content of above-described embodiment it is found that the embodiment of the present invention needs to obtain targeted training sample set to convolution
Neural network is trained.So in order to verify the best training sample of convolutional neural networks fitness, the embodiment of the present invention is adopted
It takes the method for adjustment and/or adjusts the mode of data sequence length to determine the optimum training mode of convolutional neural networks.
It is understood that the space characteristics discrimination that convolution kernel extracts is higher, classification accuracy is higher.Input convolution
Can the process that neural network carries out feature extraction and classifies, finally extract the significant feature of discrimination, be to guarantee that classification is quasi-
The key of true rate.Input before first convolutional layer between the two dimensional character figure of all categories being rearranged into have significant discrimination with
And it is of all categories itself have feature abundant, and be the premise that subsequent processing can continue to extract discrimination obvious characteristic.
It is found by verifying, the data sequence length for adjusting sample is trained and can cause larger shadow to classification accuracy
It rings, the embodiment of the present invention changes the sequence length of each training sample under the premise of n dimension square matrix can be rearranged by meeting,
It is down to it in 16 reduction process of sequence length from 4096 points of sequence length, has chosen the several representatives points 16 made a difference
(4×4)、64(8×8)、144(12×12)、256(16×16)、400(20×20)、576(24×24)、1024(32×32)、
1600 (40 × 40), 2304 (48 × 48), 2704 (52 × 52), 3600 (60 × 60), 4096 (64 × 64) are tested, and are obtained
The relationship of data sequence length and classification accuracy, as shown in table 1.
The relationship of table 1 data sequence length and classification accuracy
Sequence length is longer it can be seen from 1 content of table, and loss function value loss when training to 1000 step is lower,
Accuracy rate is higher, and train epochs needed for reaching convergence are fewer, and final classification accuracy rate is higher.And it can be found that guaranteeing
Under the premise of train epochs are enough, when sequence length is more than or equal to 64, very high classification accuracy can reach.It is long in sequence
Under the premise of degree meets above-mentioned condition, when miscellaneous noise ratio is higher than 1.5dB, sea clutter and noise can be stablized area using LeNet
It separates, as miscellaneous noise ratio gradually reduces, classification accuracy can also be gradually reduced.
It is understood that the embodiment of the present invention can determine convolutional Neural by adjusting the mode of data sequence length
The optimum training mode of network.And as long as sequence length, when being more than or equal to 64, training effect is very good.
On the basis of the above embodiments, the adjustment training sample concentrates the time-frequency representation shape of each training sample
Formula specifically includes:
Each training sample is concentrated to carry out Fast Fourier Transform (FFT) the training sample, by each training sample by time domain
Data are transformed to frequency domain data.
It is understood that it is generally time domain data for the sea clutter signal of training and the form of expression of noise signal,
But the discrimination of two kinds of signals is not obvious under time domain data.It can not be restrained so as to cause training, to cannot complete to classify
Task.
For said circumstances, the embodiment of the present invention provides a kind of mode of Signal Pretreatment, i.e., by each training sample into
Each training sample is transformed to frequency domain data by time domain data by row Fast Fourier Transform (FFT), to by sea clutter signal and make an uproar
The discrimination of acoustical signal improves, and table 2 is pretreatment training result provided in an embodiment of the present invention and does not pre-process training result pair
Compare table.
Table 2 pre-processes training result and does not pre-process training result contrast table
As shown in table 2, data prediction is affected to the classification accuracy of LeNet.Although without pretreated number
According to sample, in the sufficiently long situation of single sequence length, is adjusted by certain parameter, also can guarantee that certain classification is accurate
Rate, but under the premise of being changed without network structure, not significantly improving sequence length, classification accuracy is improved, then must need to use this
The data prediction means that inventive embodiments provide are to improve the discrimination of different classes of signal.
There is provided through the embodiment of the present invention Fast Fourier Transform (FFT) pre-processes data, significantly improve classification
Accuracy rate.
On the basis of the above embodiments, described that preset convolutional neural networks are instructed based on the training sample set
Practice, specifically include:
The preset convolutional neural networks are trained using control variate method, with the determination convolutional neural networks
In each network architecture parameters.
It is understood that there are many network architecture parameters to network for meeting in the training process to convolutional neural networks
Training result impact, common network architecture parameters for example: convolution kernel size, step-length, first layer convolution nucleus number, second
Layer convolution nucleus number and implicit number of nodes etc..And the Judging index of training result usually uses: step number, most needed for reaching convergent
Whole accuracy rate, Loss decline process and training speed etc..So in order to determine, each network structure is joined in convolutional neural networks
Several selections or selective rules, the embodiment of the present invention use control variate method and are trained to it.
Control variate method refers under conditions of input is same sea clutter data and noise data, keeps other ginsengs
Number is constant, only changes a parameter and is repeatedly trained, obtains crowd size (batchsize), convolution kernel size (kernal
Size), the ginseng such as step-length (stride), first layer convolution nucleus number (conv1), second layer convolution nucleus number (conv2), implicit number of nodes
Several pairs of training speeds (under same step number accuracy rate and loss measure) and category of model accuracy rate (reach standard when convergence
True rate) influence.
Table 3 is that control variate method provided in an embodiment of the present invention determines that batch size influences result to training effect.
3 control variate method of table determines that batch size influences result to training effect
As shown in table 3, in the case where guaranteeing that remaining 4 network architecture parameters is constant, only change batch size, thus
It was found that accuracy rate improves after batch size reduction and training speed is very fast, it was demonstrated that batch size reduction is to play positive shadow to training effect
Loud.Remaining several network architecture parameters can be determined similar to the above method using control variate method provided in an embodiment of the present invention
Connection between training result, so that most suitable network architecture parameters be selected to be trained.
Fig. 2 is a kind of sea clutter provided in an embodiment of the present invention and noise signal categorizing system structure chart, as shown in Fig. 2,
The system comprises: input module 201 and categorization module 202, in which:
Input module 201 is used for will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;
Categorization module 202 is used to obtain the classification results of the convolutional neural networks output after the training.
It is specific how sea clutter and noise signal classification to can be used for holding by input module 201 and categorization module 202
The technical solution of row sea clutter and noise signal classification method embodiment shown in FIG. 1, it is similar that the realization principle and technical effect are similar,
Details are not described herein again.
A kind of sea clutter provided in an embodiment of the present invention and noise signal categorizing system, by using the volume after training
Product neural network classifies to sea clutter and noise signal, has high classification accuracy.
On the basis of the above embodiments, the convolutional neural networks are specially LeNet.
Sea clutter provided in an embodiment of the present invention can make with convolutional neural networks used in noise signal categorizing system
With the network LeNet, through simulation results show, which being capable of good completion sea clutter and noise signal classification task.
On the basis of the above embodiments, input module 201 specifically includes:
Rearrangement units, the rearrangement units are used to the data point of the one-dimensional echo-signal of radar to be sorted being rearranged into n*n
Two-dimensional matrix;
Input unit, the two-dimensional matrix for being inputted the convolutional neural networks after the training by the input unit
In.
On the basis of the above embodiments, the system also includes:
Module is obtained, the module that obtains is used to obtain sample data of the discrimination greater than preset threshold as training sample
Collection;
Training module, the training module are used to instruct preset convolutional neural networks based on the training sample set
Practice.
On the basis of the above embodiments, the system also includes:
Module is adjusted, the adjustment module concentrates the data sequence of each training sample long for adjusting the training sample
Spend and/or adjust the time-frequency representation form that the training sample concentrates each training sample.
On the basis of the above embodiments, the adjustment module specifically includes:
Pretreatment unit, the pretreatment unit are used to concentrate each training sample to carry out quick Fu the training sample
In leaf transformation, each training sample is transformed to frequency domain data by time domain data.
Pretreatment unit is provided through the embodiment of the present invention, training sample frequency domain can be converted to by time domain, improve sample
Discrimination between this, to improve training effect.
On the basis of the above embodiments, the training module specifically includes:
Structural parameters determination unit, the structural parameters determination unit are used for using control variate method to the preset volume
Product neural network is trained, with each network architecture parameters in the determination convolutional neural networks.
The control variate method that the structural parameters determination unit provided through the embodiment of the present invention provides can determine convolution mind
Through the structural parameters in network training process to the influence degree of classification accuracy and its influence degree to classification results.
Fig. 3 is the structural block diagram of electronic equipment provided in an embodiment of the present invention, referring to Fig. 3, the electronic equipment, comprising:
Processor (processor) 310, communication interface (Communications Interface) 320, memory (memory) 330
With bus 340, wherein processor 310, communication interface 320, memory 330 complete mutual communication by bus 340.Place
Reason device 310 can call the logical order in memory 330, to execute following method: the one-dimensional echo of radar to be sorted is believed
Number input training after convolutional neural networks in;The classification results of convolutional neural networks output after obtaining the training.
The embodiment of the present invention discloses a kind of computer program product, and the computer program product is non-transient including being stored in
Computer program on computer readable storage medium, the computer program include program instruction, when described program instructs quilt
When computer executes, computer is able to carry out method provided by above-mentioned each method embodiment, for example, by thunder to be sorted
Up in the convolutional neural networks after the input training of one-dimensional echo-signal;Point of convolutional neural networks output after obtaining the training
Class result.
The embodiment of the present invention provides a kind of non-transient computer readable storage medium, the non-transient computer readable storage
Medium storing computer instruction, the computer instruction make the computer execute side provided by above-mentioned each method embodiment
Method, for example, will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;Obtain the training
The classification results of convolutional neural networks output afterwards.
Those of ordinary skill in the art will appreciate that: realize that all or part of the steps of above method embodiment can pass through
The relevant hardware of program instruction is completed, and program above-mentioned can be stored in a computer readable storage medium, the program
When being executed, step including the steps of the foregoing method embodiments is executed;And storage medium above-mentioned includes: ROM, RAM, magnetic disk or light
The various media that can store program code such as disk.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations;Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify the technical solutions described in the foregoing embodiments or equivalent replacement of some of the technical features;
And these are modified or replaceed, technical solution of various embodiments of the present invention that it does not separate the essence of the corresponding technical solution spirit and
Range.
Claims (10)
1. a kind of sea clutter and noise signal classification method characterized by comprising
It will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;
The classification results of convolutional neural networks output after obtaining the training.
2. the method according to claim 1, wherein the convolutional neural networks are specially LeNet.
3. the method according to claim 1, wherein described by the one-dimensional echo-signal input instruction of radar to be sorted
In convolutional neural networks after white silk, specifically include:
The data point of the one-dimensional echo-signal of radar to be sorted is rearranged into the two-dimensional matrix of n*n;
The two-dimensional matrix is inputted in the convolutional neural networks after the training.
4. the method according to claim 1, wherein the one-dimensional echo-signal of radar to be sorted is inputted described
Before in convolutional neural networks after training, the method also includes:
It obtains discrimination and is greater than the sample data of preset threshold as training sample set;
Preset convolutional neural networks are trained based on the training sample set.
5. according to the method described in claim 4, it is characterized in that, being based on the training sample set to preset convolution described
Before neural network is trained, the method also includes:
Adjusting the training sample concentrates the data sequence length of each training sample and/or the adjustment training sample to concentrate often
The time-frequency representation form of a training sample.
6. according to the method described in claim 5, it is characterized in that, the adjustment training sample concentrates each training sample
Time-frequency representation form, specifically include:
Each training sample is concentrated to carry out Fast Fourier Transform (FFT) the training sample, by each training sample by time domain data
It is transformed to frequency domain data.
7. according to the method described in claim 4, it is characterized in that, described refreshing to preset convolution based on the training sample set
It is trained, specifically includes through network:
The preset convolutional neural networks are trained using control variate method, in the determination convolutional neural networks
Each network architecture parameters.
8. a kind of sea clutter and noise signal categorizing system characterized by comprising
Input module, for will be in the convolutional neural networks after the one-dimensional echo-signal input training of radar to be sorted;
Categorization module, for obtaining the classification results of the output of the convolutional neural networks after the training.
9. a kind of electronic equipment, which is characterized in that including memory and processor, the processor and the memory pass through always
Line completes mutual communication;The memory is stored with the program instruction that can be executed by the processor, the processor tune
The method as described in claim 1 to 7 is any is able to carry out with described program instruction.
10. a kind of non-transient computer readable storage medium, which is characterized in that the non-transient computer readable storage medium is deposited
Computer instruction is stored up, the computer instruction makes the computer execute method as described in any one of claim 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811113859.8A CN109389058B (en) | 2018-09-25 | 2018-09-25 | Sea clutter and noise signal classification method and system |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811113859.8A CN109389058B (en) | 2018-09-25 | 2018-09-25 | Sea clutter and noise signal classification method and system |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109389058A true CN109389058A (en) | 2019-02-26 |
CN109389058B CN109389058B (en) | 2021-03-23 |
Family
ID=65417782
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811113859.8A Active CN109389058B (en) | 2018-09-25 | 2018-09-25 | Sea clutter and noise signal classification method and system |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109389058B (en) |
Cited By (20)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109901129A (en) * | 2019-03-06 | 2019-06-18 | 中国人民解放军海军航空大学 | Object detection method and system in a kind of sea clutter |
CN110146855A (en) * | 2019-06-11 | 2019-08-20 | 北京无线电测量研究所 | Radar Intermittent AF panel thresholding calculation method and device |
CN110232342A (en) * | 2019-06-03 | 2019-09-13 | 中国人民解放军海军航空大学 | Sea situation level determination method and device based on convolutional neural networks |
CN110658508A (en) * | 2019-10-17 | 2020-01-07 | 中国人民解放军火箭军工程大学 | K distribution sea clutter parameter estimation method based on characteristic quantity |
CN110659684A (en) * | 2019-09-23 | 2020-01-07 | 中国人民解放军海军航空大学 | Convolutional neural network-based STBC signal identification method |
CN110956197A (en) * | 2019-10-28 | 2020-04-03 | 新绎健康科技有限公司 | Method and system for establishing pulse wave noise signal identification model based on convolutional neural network |
CN111310700A (en) * | 2020-02-27 | 2020-06-19 | 电子科技大学 | Intermediate frequency sampling sequence processing method for radiation source fingerprint feature identification |
RU2731332C1 (en) * | 2020-01-16 | 2020-09-01 | Акционерное общество "Концерн "Созвездие" | Method for formation of decisionmaking device threshold based on neuroregulator |
CN111693953A (en) * | 2020-05-11 | 2020-09-22 | 中山大学 | Target classification recognition model, method, system and device based on micro Doppler |
CN111812598A (en) * | 2020-07-30 | 2020-10-23 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Time domain and frequency domain multi-feature-based ground and sea clutter classification method |
CN111880158A (en) * | 2020-08-06 | 2020-11-03 | 中国人民解放军海军航空大学 | Radar target detection method and system based on convolutional neural network sequence classification |
CN111950198A (en) * | 2020-08-10 | 2020-11-17 | 北京环境特性研究所 | Ground clutter simulation method based on neural network |
CN112068085A (en) * | 2020-10-16 | 2020-12-11 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Radar sea clutter original data rapid preprocessing method based on deep learning |
CN112099014A (en) * | 2020-08-24 | 2020-12-18 | 广东工业大学 | Road millimeter wave noise model detection and estimation method based on deep learning |
CN112711032A (en) * | 2020-12-07 | 2021-04-27 | 中国人民解放军海军航空大学 | Radar target detection method and system based on graph data and GCN |
CN112859034A (en) * | 2021-04-26 | 2021-05-28 | 中国人民解放军国防科技大学 | Natural environment radar echo amplitude model classification method and device |
CN113189554A (en) * | 2021-04-12 | 2021-07-30 | 中国人民解放军海军航空大学航空作战勤务学院 | Processing method of radar measured echo data, electronic equipment and storage medium |
CN113505697A (en) * | 2021-07-09 | 2021-10-15 | 哈尔滨工业大学 | High-frequency ground wave radar clutter classification method based on U-Net network |
CN113687325A (en) * | 2021-07-08 | 2021-11-23 | 西安电子科技大学 | Shielded small target detection method based on LP and HRRP models |
CN113759336A (en) * | 2021-08-31 | 2021-12-07 | 东南大学 | Sea clutter suppression method under graph feature learning |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107169435A (en) * | 2017-05-10 | 2017-09-15 | 天津大学 | A kind of convolutional neural networks human action sorting technique based on radar simulation image |
CN107203692A (en) * | 2017-05-09 | 2017-09-26 | 哈尔滨工业大学(威海) | The implementation method of atrial fibrillation detection based on depth convolutional neural networks |
CN107220606A (en) * | 2017-05-22 | 2017-09-29 | 西安电子科技大学 | The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks |
CN107884768A (en) * | 2017-10-12 | 2018-04-06 | 上海无线电设备研究所 | The adaptive active radar system of complex environment based on big data analytical technology |
CN108226889A (en) * | 2018-01-19 | 2018-06-29 | 中国人民解放军陆军装甲兵学院 | A kind of sorter model training method of radar target recognition |
CN108229404A (en) * | 2018-01-09 | 2018-06-29 | 东南大学 | A kind of radar echo signal target identification method based on deep learning |
CN108564115A (en) * | 2018-03-30 | 2018-09-21 | 西安电子科技大学 | Semi-supervised polarization SAR terrain classification method based on full convolution GAN |
-
2018
- 2018-09-25 CN CN201811113859.8A patent/CN109389058B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107203692A (en) * | 2017-05-09 | 2017-09-26 | 哈尔滨工业大学(威海) | The implementation method of atrial fibrillation detection based on depth convolutional neural networks |
CN107169435A (en) * | 2017-05-10 | 2017-09-15 | 天津大学 | A kind of convolutional neural networks human action sorting technique based on radar simulation image |
CN107220606A (en) * | 2017-05-22 | 2017-09-29 | 西安电子科技大学 | The recognition methods of radar emitter signal based on one-dimensional convolutional neural networks |
CN107884768A (en) * | 2017-10-12 | 2018-04-06 | 上海无线电设备研究所 | The adaptive active radar system of complex environment based on big data analytical technology |
CN108229404A (en) * | 2018-01-09 | 2018-06-29 | 东南大学 | A kind of radar echo signal target identification method based on deep learning |
CN108226889A (en) * | 2018-01-19 | 2018-06-29 | 中国人民解放军陆军装甲兵学院 | A kind of sorter model training method of radar target recognition |
CN108564115A (en) * | 2018-03-30 | 2018-09-21 | 西安电子科技大学 | Semi-supervised polarization SAR terrain classification method based on full convolution GAN |
Non-Patent Citations (2)
Title |
---|
EMRE ARDIÇ ET AL.: "Classification of 1D signals using deep neural networks", 《2018 26TH SIGNAL PROCESSING AND COMMUNICATIONS APPLICATIONS CONFERENCE (SIU)》 * |
井博军: "基于深度学习的雷达辐射源识别技术研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (31)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109901129A (en) * | 2019-03-06 | 2019-06-18 | 中国人民解放军海军航空大学 | Object detection method and system in a kind of sea clutter |
CN110232342A (en) * | 2019-06-03 | 2019-09-13 | 中国人民解放军海军航空大学 | Sea situation level determination method and device based on convolutional neural networks |
CN110146855B (en) * | 2019-06-11 | 2020-10-23 | 北京无线电测量研究所 | Radar intermittent interference suppression threshold calculation method and device |
CN110146855A (en) * | 2019-06-11 | 2019-08-20 | 北京无线电测量研究所 | Radar Intermittent AF panel thresholding calculation method and device |
CN110659684A (en) * | 2019-09-23 | 2020-01-07 | 中国人民解放军海军航空大学 | Convolutional neural network-based STBC signal identification method |
CN110658508A (en) * | 2019-10-17 | 2020-01-07 | 中国人民解放军火箭军工程大学 | K distribution sea clutter parameter estimation method based on characteristic quantity |
CN110658508B (en) * | 2019-10-17 | 2023-03-10 | 中国人民解放军火箭军工程大学 | K distribution sea clutter parameter estimation method based on characteristic quantity |
CN110956197A (en) * | 2019-10-28 | 2020-04-03 | 新绎健康科技有限公司 | Method and system for establishing pulse wave noise signal identification model based on convolutional neural network |
RU2731332C1 (en) * | 2020-01-16 | 2020-09-01 | Акционерное общество "Концерн "Созвездие" | Method for formation of decisionmaking device threshold based on neuroregulator |
CN111310700A (en) * | 2020-02-27 | 2020-06-19 | 电子科技大学 | Intermediate frequency sampling sequence processing method for radiation source fingerprint feature identification |
CN111310700B (en) * | 2020-02-27 | 2024-02-13 | 电子科技大学 | Intermediate frequency sampling sequence processing method for radiation source fingerprint feature recognition |
CN111693953B (en) * | 2020-05-11 | 2023-12-05 | 中山大学 | micro-Doppler-based target classification recognition model, method, system and device |
CN111693953A (en) * | 2020-05-11 | 2020-09-22 | 中山大学 | Target classification recognition model, method, system and device based on micro Doppler |
CN111812598B (en) * | 2020-07-30 | 2022-03-22 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Time domain and frequency domain multi-feature-based ground and sea clutter classification method |
CN111812598A (en) * | 2020-07-30 | 2020-10-23 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Time domain and frequency domain multi-feature-based ground and sea clutter classification method |
CN111880158A (en) * | 2020-08-06 | 2020-11-03 | 中国人民解放军海军航空大学 | Radar target detection method and system based on convolutional neural network sequence classification |
CN111950198B (en) * | 2020-08-10 | 2024-02-02 | 北京环境特性研究所 | Ground clutter simulation method based on neural network |
CN111950198A (en) * | 2020-08-10 | 2020-11-17 | 北京环境特性研究所 | Ground clutter simulation method based on neural network |
CN112099014A (en) * | 2020-08-24 | 2020-12-18 | 广东工业大学 | Road millimeter wave noise model detection and estimation method based on deep learning |
CN112099014B (en) * | 2020-08-24 | 2023-08-22 | 广东工业大学 | Road millimeter wave noise model detection estimation method based on deep learning |
CN112068085B (en) * | 2020-10-16 | 2022-05-06 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Radar sea clutter original data rapid preprocessing method based on deep learning |
CN112068085A (en) * | 2020-10-16 | 2020-12-11 | 中国电波传播研究所(中国电子科技集团公司第二十二研究所) | Radar sea clutter original data rapid preprocessing method based on deep learning |
CN112711032A (en) * | 2020-12-07 | 2021-04-27 | 中国人民解放军海军航空大学 | Radar target detection method and system based on graph data and GCN |
CN113189554A (en) * | 2021-04-12 | 2021-07-30 | 中国人民解放军海军航空大学航空作战勤务学院 | Processing method of radar measured echo data, electronic equipment and storage medium |
CN112859034A (en) * | 2021-04-26 | 2021-05-28 | 中国人民解放军国防科技大学 | Natural environment radar echo amplitude model classification method and device |
CN113687325A (en) * | 2021-07-08 | 2021-11-23 | 西安电子科技大学 | Shielded small target detection method based on LP and HRRP models |
CN113687325B (en) * | 2021-07-08 | 2024-02-06 | 西安电子科技大学 | Method for detecting shielding small target based on LP and HRRP models |
CN113505697B (en) * | 2021-07-09 | 2022-07-29 | 哈尔滨工业大学 | High-frequency ground wave radar clutter classification method based on U-Net network |
CN113505697A (en) * | 2021-07-09 | 2021-10-15 | 哈尔滨工业大学 | High-frequency ground wave radar clutter classification method based on U-Net network |
CN113759336A (en) * | 2021-08-31 | 2021-12-07 | 东南大学 | Sea clutter suppression method under graph feature learning |
CN113759336B (en) * | 2021-08-31 | 2024-04-30 | 东南大学 | Sea clutter suppression method under graph feature learning |
Also Published As
Publication number | Publication date |
---|---|
CN109389058B (en) | 2021-03-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109389058A (en) | Sea clutter and noise signal classification method and system | |
CN109086700B (en) | Radar one-dimensional range profile target identification method based on deep convolutional neural network | |
CN110334741B (en) | Radar one-dimensional range profile identification method based on cyclic neural network | |
CN110363151A (en) | Based on the controllable radar target detection method of binary channels convolutional neural networks false-alarm | |
CN109188414A (en) | A kind of gesture motion detection method based on millimetre-wave radar | |
CN106529428A (en) | Underwater target recognition method based on deep learning | |
CN110175560A (en) | A kind of radar signal intra-pulse modulation recognition methods | |
CN113571067B (en) | Voiceprint recognition countermeasure sample generation method based on boundary attack | |
CN107315996A (en) | A kind of noise characteristic extracting method of ships under water based on IMF Energy-Entropies and PCA | |
CN110348356A (en) | A kind of successive frame RD images steganalysis method based on depth light stream network | |
CN108399430B (en) | A kind of SAR image Ship Target Detection method based on super-pixel and random forest | |
CN109901129A (en) | Object detection method and system in a kind of sea clutter | |
CN113657491A (en) | Neural network design method for signal modulation type recognition | |
CN110459225A (en) | A kind of speaker identification system based on CNN fusion feature | |
CN105913081A (en) | Improved PCAnet-based SAR image classification method | |
Guo et al. | Radar target recognition based on feature Pyramid fusion lightweight CNN | |
CN114117912A (en) | Sea clutter modeling and inhibiting method under data model dual drive | |
Principe et al. | Target discrimination in synthetic aperture radar using artificial neural networks | |
Ye et al. | Recognition algorithm of emitter signals based on PCA+ CNN | |
Chu et al. | Attention enhanced spatial temporal neural network for HRRP recognition | |
Cinelli et al. | Vessel classification through convolutional neural networks using passive sonar spectrogram images | |
Ruan et al. | Automatic recognition of radar signal types based on CNN-LSTM | |
CN109784311B (en) | Target identification method based on linear frequency modulation wavelet atomic network | |
KR102407834B1 (en) | Method and apparatus for property-based classification of long-pulse radar signals | |
CN116430317A (en) | Radiation source modulation pattern and individual 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 |