CN109389058B - Sea clutter and noise signal classification method and system - Google Patents
Sea clutter and noise signal classification method and system Download PDFInfo
- Publication number
- CN109389058B CN109389058B CN201811113859.8A CN201811113859A CN109389058B CN 109389058 B CN109389058 B CN 109389058B CN 201811113859 A CN201811113859 A CN 201811113859A CN 109389058 B CN109389058 B CN 109389058B
- Authority
- CN
- China
- Prior art keywords
- training sample
- neural network
- convolutional neural
- radar
- training
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Images
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 method and a system for classifying sea clutter and noise signals, wherein the method comprises the following steps: inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network; and obtaining the classification result output by the trained convolutional neural network. According to the method and the system for classifying the sea clutter and the noise signals, the trained convolutional neural network is used for classifying the sea clutter and the noise signals, and the classification accuracy is high.
Description
Technical Field
The embodiment of the invention relates to the technical field of signal processing, in particular to a method and a system for classifying sea clutter and noise signals.
Background
The problems of sea clutter modeling and target detection in sea clutter are always hot and difficult problems of domestic and foreign research, and the statistical characteristic quantity for describing the sea clutter mainly comprises amplitude characteristics, frequency spectrum characteristics, time/space correlation characteristics and the like. Common models include a GIT model, a TSC model, a HYB model, an SIT model, an NRL model and the like according to the amplitude mean characteristic, and traditional probability distribution models include Rayleigh distribution, lognormal distribution, Weibull distribution, K distribution and the like according to the amplitude distribution characteristic; for the sea clutter doppler spectrum modeling research, common models include a Lee spectrum model, a Walker model, a Ward model and the like, and a series of Constant False Alarm Rate (CFAR) detectors are developed on the basis, however, actual sea clutter often deviates from an assumed statistical distribution model, the performance of the CFAR detectors is seriously reduced, and target detection is seriously influenced.
The classification of the sea clutter and the noise is completed by the core problem of target detection, so that the sea clutter and the noise signals are accurately classified, and a noise suppression method suitable for the background is automatically selected, so that powerful guarantee can be provided for improving the detection performance of the offshore targets.
Therefore, a method for classifying sea clutter and noise signals is needed to solve the above problems.
Disclosure of Invention
To solve the above problems, embodiments of the present invention provide a method and system for classifying sea clutter and noise signals, which overcome the above problems or at least partially solve the above problems.
In a first aspect, an embodiment of the present invention provides a method for classifying sea clutter and noise signals, including:
inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network;
and obtaining the classification result output by the trained convolutional neural network.
In a second aspect, an embodiment of the present invention provides a sea clutter and noise signal classification system, including:
the input module is used for inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network;
and the classification module is used for acquiring the classification result output by the trained convolutional neural network.
Third aspect an embodiment of the present invention provides an electronic device, including:
a processor, a memory, a communication interface, and a bus; the processor, the memory and the communication interface complete mutual communication through the bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform a method of sea clutter and noise signal classification as described above.
In a fourth aspect, embodiments of the present invention provide a non-transitory computer-readable storage medium storing computer instructions that cause the computer to perform the above sea clutter and noise signal classification method.
According to the method and the system for classifying the sea clutter and the noise signals, the trained convolutional neural network is used for classifying the sea clutter and the noise signals, and the classification accuracy is high.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below, and it is obvious that the drawings in the following description are some embodiments of the present invention, and those skilled in the art can also obtain other drawings according to the drawings without creative efforts.
Fig. 1 is a schematic flow chart of a method for classifying sea clutter and noise signals according to an embodiment of the present invention;
FIG. 2 is a block diagram of a sea clutter and noise signal classification system according to an embodiment of the present invention;
fig. 3 is a block diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly described below with reference to the drawings in the embodiments of the present invention, and it is obvious that the described embodiments are some embodiments, but not all embodiments, of the present invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
At present, since Deep Learning (Deep Learning) in 2006 is developed vigorously, a Convolutional Neural Network (CNN) shows a wide application prospect in the field of image processing, can fully utilize local characteristics of data, has the characteristics of translation invariance and strong generalization capability, can simplify a network structure by using weight sharing, and is successfully applied to the aspects of object recognition and face recognition in the field of optical images. In the aspect of Radar image processing, for diversified data and features obtained by a Radar, such as a Synthetic Aperture Radar (SAR) image, a High Resolution Range Profile (HRRP), a Micro-Doppler spectrogram, a Range-Doppler spectrogram (R-D) spectrogram, etc., a CNN also obtains relatively abundant results in application researches such as SAR image target classification and identification, track segmentation, time-frequency two-dimensional image target classification, etc., but the existing researches mainly make an echo diagram or a feature diagram obtained by the Radar equivalent to a two-dimensional image and then process the two-dimensional image. However, the radar echo signal is a one-dimensional signal, and there is a new research on applying CNN to one-dimensional radar echo signal processing.
In view of the above situation, fig. 1 is a schematic flow chart of a method for classifying sea clutter and noise signals according to an embodiment of the present invention, as shown in fig. 1, including:
101. inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network;
102. and obtaining the classification result output by the trained convolutional neural network.
It can be understood that, in the embodiment of the present invention, the convolutional neural network CNN is used to perform a target detection background, that is, the classification of the sea clutter signal and the noise signal in the radar one-dimensional echo signal is implemented, so that the accuracy of background radar judgment is improved, and the target discovery probability is further improved. It should be noted that, by classifying the sea clutter and the noise signals provided by the embodiment of the present invention, the determination of the windward/downwind clutter region and the noise region can be achieved.
Specifically, in step 101, a radar one-dimensional radar signal to be classified, that is, a signal returned by a radar detection sea target background, includes a sea clutter signal and a noise signal, and is input into a trained convolutional neural network, where the trained convolutional neural network is trained in advance in the embodiment of the present invention, and can extract features from the signal to be classified, and identify the type of the features, thereby classifying the signals. The convolutional neural network provided by the embodiment of the invention generally comprises a fully-connected layer (input and output layer), a convolutional layer and a pooling layer (down-sampling layer), and the training process can be basically divided into a forward propagation process and a backward propagation process. The basic formula for the forward propagation of convolutional neural network convolutional layers is as follows:
wherein f (-) is a neuron excitation function and is a linear function, and (1) is shown in the formula
Is the output value (or activation value) of the ith neuron of the kth layer,
is the input value to the neuron and,
as the parameter of the convolutional layer, the neurons of the layer share the parameter for weight sharing.
Are bias terms for different neurons. MiShows the input region to the ith neuron in the k-th convolutional layer in the previous layer (layer k-1),
representing the values of the neurons in the input region.
In a downsampling layer of a convolutional neural network, an ith neuron of a k-th layer pools an n × n region with the size of the k-1 layer, namely downsampling operation is expressed by down (·), and an activation value is calculated through an activation function of the neuron, wherein the specific formula is as follows:
wherein, to the left of the equation
Represents the output value (or activation value) of the ith neuron of the l-th layer,
in (1)
Is an input value, MiShows the input region to the ith neuron in the k-th convolutional layer in the previous layer (layer k-1),
is a parameter of the neuron that is,
is an offset.
In the convolutional neural network provided in the embodiment of the present invention, convolutional layers and downsampling layers alternate, and a layer (k + 1) next to a kth convolutional layer is a downsampling layer. When the down-sampling layer is used for carrying out pooling operation, one neuron is connected with one neuron of the convolution layer, firstly, the (k + 1) th convolution layer is up-sampled (upsampling) to make the size of the convolution layer be identical to that of the kth layer, and then, gradient calculation is carried out. The operation of the upsampling layer can be realized by a Kronecker product:
wherein up (·) represents the upsampling operation, and after the above formula operation, one neuron becomes an n × n two-dimensional planar neuron layer, where the parameters of all neurons are the same and are β, thereby obtaining convolutional layer residual error:
where, denotes the Hadamard product, the vectors are multiplied element by element.
The gradient of the cost function to the parameter is further calculated:
where J is the cost function of the network, as with the BP neural network,
an output value at coordinates (u, v) representing the convolved region of the (k-1) th layer,
residual values of neurons representing the kth layer (u, v) coordinate position, the gradient can be calculated during forward propagation.
The downsampled layer is similar to the convolutional layer, and the gradient can be calculated in forward propagation by the following formula:
wherein down (-) denotes a down-sampling operation for the previous layer
The down-sampled value is calculated as the input value of the l-th layer. The gradient can then be calculated:
wherein the content of the first and second substances,
is the gradient of the coordinate position of the ith layer (u, v) to the parameter beta, and the gradient of the ith layer can be obtained by summing the coordinates.
Residual values of neurons representing the (u, v) th layer coordinate position, and summing the coordinates can yield a pair bias
Of the gradient of (c). After the gradient calculation of the convolutional layer and the downsampling layer is finished, the neural network is solved by using an error back propagation algorithm.
Then in step 101, the radar one-dimensional echo signal to be classified is input into the convolutional neural network provided by the above process and trained, so that the feature extraction of the signal can be completed.
Further, in step 102, a classification result of the trained convolutional neural network may be obtained, that is, the radar one-dimensional echo signal to be classified is classified according to the features extracted in step 101, and the classification result is output to be a sea clutter signal or a noise signal. According to the statistical result of the test sample, the method provided by the embodiment of the invention has very high accuracy and can accurately classify the sea clutter and the noise signal.
According to the method for classifying the sea clutter and the noise signals, the trained convolutional neural network is used for classifying the sea clutter and the noise signals, and the method has high classification accuracy.
On the basis of the above embodiment, the convolutional neural network is specifically LeNet.
From the content of the foregoing embodiments, the embodiment of the present invention provides a convolutional neural network to perform a task of classifying sea clutter and noise, and preferably, the convolutional neural network provided by the embodiment of the present invention is LeNet.
In the embodiment of the invention, LeNet is applied to a one-dimensional radar echo sequence, the network structure parameters of LeNet need to be modified, the input is adjusted to be a sequence of 1 × 400, the sequence is changed into a matrix of 20 × 20 through a reshape process in a network structure, and the required network structure parameters need to be recalculated. The first and third layers are convolutional layers, the convolutional kernel size is 4 × 4, 16 convolutional kernels and 32 convolutional kernels are respectively provided, all 0 padding is adopted, and therefore, the depth (16 and 32 respectively) is changed without changing the output size. The second and fourth layers are downsampled layers, each with a step size of 2, thus reducing the output size by half to 10 × 10 and 5 × 5, respectively. Finally, the network output is obtained by the full connection layer (layer 6, input is 800, output is 512, 5 × 5 × 32) and the seventh layer (input 512, output is class number 2). It should be noted that, in addition to meeting the basic input/output relationship of the LeNet network structure, parameters such as the size, the number, the step size, and the like of the convolution kernel need to be adjusted according to the actual training result, and the embodiment of the present invention is not particularly limited.
On the basis of the above embodiment, the inputting the radar one-dimensional echo signal to be classified into the trained convolutional neural network specifically includes:
rearranging data points of the radar one-dimensional echo signals to be classified into a two-dimensional matrix of n x n;
and inputting the two-dimensional matrix into the trained convolutional neural network.
It can be understood that the embodiment of the present invention inputs the radar one-dimensional echo signal into the neural network, which is a one-dimensional signal, but in the LeNet network structure, it is necessary to rearrange the points of the input one-dimensional radar echo signal into a two-dimensional matrix of n × n, so as to perform feature extraction through the convolution kernel sliding of the LeNet network structure.
On the basis of the above embodiment, before the radar one-dimensional echo signal to be classified is input into the trained convolutional neural network, the method further includes:
acquiring sample data with the discrimination degree larger than a preset threshold value as a training sample set;
and training a preset convolutional neural network based on the training sample set.
According to the embodiment, the one-dimensional radar echo signal data meet the precondition that the convolutional neural network can be used for processing, but the rearrangement process can destroy and newly add partial spatial features, so that the classification accuracy is influenced. Therefore, the embodiment of the invention needs to select the sample data with the discrimination degree greater than the preset threshold as the training sample set, thereby reducing the influence of the data on the classification result to a certain extent. The discrimination refers to amplitude discrimination of the sea clutter data and the noise data in the training sample.
On the basis of the above embodiment, before the training a preset convolutional neural network based on the training sample set, the method further includes:
and adjusting the length of the data sequence of each training sample in the training sample set and/or adjusting the time-frequency representation form of each training sample in the training sample set.
It can be known from the content of the above embodiment that the embodiment of the present invention needs to acquire a targeted training sample set to train the convolutional neural network. Then, in order to verify the training sample with the best fitness of the convolutional neural network, the embodiment of the present invention determines the best training mode of the convolutional neural network by using an adjustment method and/or an adjustment method for adjusting the length of the data sequence.
It can be understood that the higher the spatial feature discrimination extracted by the convolution kernel, the higher the classification accuracy. The process of inputting the convolutional neural network for feature extraction and classification is the key for ensuring the classification accuracy, and finally whether the features with obvious discrimination can be extracted or not. The two-dimensional feature maps of each category rearranged before the first convolution layer is input have obvious discrimination and each category has rich features, and the precondition that the subsequent processing can continue to extract the features with obvious discrimination is also provided.
According to verification, on the premise that n-dimensional square matrix rearrangement is achieved, the training by adjusting the data sequence length of the samples greatly affects the classification accuracy, in the process of reducing the sequence length of each training sample from 4096 points of the sequence length to 16 points of the sequence length, the embodiment of the invention selects several important representative points 16(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) and 4096(64 × 64) for testing, and obtains the relationship between the data sequence length and the classification accuracy, as shown in table 1.
TABLE 1 relationship of data sequence length to Classification accuracy
It can be seen from table 1 that the longer the sequence length is, the lower the loss function value loss when the sequence is trained to 1000 steps is, the higher the accuracy is, the fewer training steps are required to achieve convergence, and the higher the final classification accuracy is. And can find that under the premise of ensuring that the training steps are enough, when the sequence length is more than or equal to 64, the high classification accuracy can be achieved. On the premise that the sequence length meets the conditions, when the noise-to-noise ratio is higher than 1.5dB, the sea clutter and the noise can be stably distinguished by using LeNet, and the classification accuracy can be gradually reduced along with the gradual reduction of the noise-to-noise ratio.
It can be appreciated that the embodiment of the present invention can determine the optimal training mode of the convolutional neural network by adjusting the length of the data sequence. And the training effect is very good as long as the sequence length is greater than or equal to 64.
On the basis of the foregoing embodiment, the adjusting the time-frequency representation form of each training sample in the training sample set specifically includes:
and performing fast Fourier transform on each training sample in the training sample set, and transforming time domain data of each training sample into frequency domain data.
It will be appreciated that the representation of the sea clutter signal and the noise signal used for training is typically time domain data, but the discrimination between the two signals is not apparent in the time domain data. Resulting in the inability of the training to converge and thus the inability to complete the classification task.
In view of the above situation, an embodiment of the present invention provides a signal preprocessing method, that is, each training sample is subjected to fast fourier transform, and each training sample is transformed from time domain data to frequency domain data, so as to improve the discrimination between a sea clutter signal and a noise signal, and table 2 is a comparison table of a preprocessing training result and a non-preprocessing training result provided in the embodiment of the present invention.
TABLE 2 comparison of pre-treatment training results and non-pre-treatment training results
As shown in table 2, the data preprocessing has a large influence on the classification accuracy of LeNet. Although a certain classification accuracy can be ensured by adjusting certain parameters under the condition that the length of a single sequence is long enough without preprocessing a data sample, the classification accuracy is improved on the premise that the network structure is not changed and the sequence length is not remarkably improved, the data preprocessing means provided by the embodiment of the invention needs to be adopted to improve the discrimination of different types of signals.
The embodiment of the invention provides fast Fourier transform to preprocess data, thereby remarkably improving the classification accuracy.
On the basis of the above embodiment, the training of the preset convolutional neural network based on the training sample set specifically includes:
and training the preset convolutional neural network by adopting a control variable method to determine each network structure parameter in the convolutional neural network.
It can be understood that in the training process of the convolutional neural network, many network structure parameters affect the training result of the network, and common network structure parameters are as follows: the convolution kernel size, the step size, the number of the first layer of convolution kernels, the number of the second layer of convolution kernels, the number of the hidden nodes and the like. The judgment indexes of the training result often adopt: the number of steps required to achieve convergence, the final accuracy, the Loss reduction process, the training speed, and the like. In order to determine the selection or the selection rule of each network structure parameter in the convolutional neural network, the embodiment of the invention trains the network structure parameter by using a control variable method.
The control variable method is characterized in that under the condition that the same sea clutter data and noise data are input, other parameters are kept unchanged, only one parameter is changed for training for multiple times, and the influence of parameters such as batch size (batch size), convolution kernel size (kernel size), step size (stride), first layer convolution kernel number (conv1), second layer convolution kernel number (conv2), hidden node number on training speed (measured by accuracy and loss under the same step number) and model classification accuracy (accuracy when convergence is achieved) is obtained.
Table 3 shows the effect of the batch size determination on the training effect by the controlled variable method provided in the embodiment of the present invention.
TABLE 3 results of the effect of batch size determination on training results by controlled variables
As shown in table 3, under the condition that the remaining 4 network structure parameters are not changed, only the batch size is changed, so that the accuracy is improved and the training speed is faster after the batch size is reduced, and the reduction of the batch size has a positive effect on the training effect. The control variable method provided by the embodiment of the invention can determine the relation between the rest network structure parameters and the training result similarly to the method, thereby selecting the most suitable network structure parameter for training.
Fig. 2 is a structural diagram of a sea clutter and noise signal classification system according to an embodiment of the present invention, as shown in fig. 2, the system includes: an input module 201 and a classification module 202, wherein:
the input module 201 is configured to input a radar one-dimensional echo signal to be classified into a trained convolutional neural network;
the classification module 202 is configured to obtain a classification result output by the trained convolutional neural network.
Specifically, how to classify the sea clutter and the noise signals through the input module 201 and the classification module 202 may be used to execute the technical scheme of the embodiment of the sea clutter and noise signal classification method shown in fig. 1, and the implementation principle and the technical effect are similar, and are not described herein again.
According to the sea clutter and noise signal classification system provided by the embodiment of the invention, the trained convolutional neural network is used for classifying the sea clutter and noise signals, so that the classification accuracy is very high.
On the basis of the above embodiment, the convolutional neural network is specifically LeNet.
The convolutional neural network used in the sea clutter and noise signal classification system provided by the embodiment of the invention can use a LeNet network, and simulation results prove that the network can well complete the task of classifying the sea clutter and noise signals.
On the basis of the above embodiment, the input module 201 specifically includes:
the rearrangement unit is used for rearranging data points of the radar one-dimensional echo signals to be classified into a two-dimensional matrix of n x n;
an input unit for inputting the two-dimensional matrix into the trained convolutional neural network.
On the basis of the above embodiment, the system further includes:
the acquisition module is used for acquiring sample data with the discrimination degree larger than a preset threshold value as a training sample set;
and the training module is used for training a preset convolutional neural network based on the training sample set.
On the basis of the above embodiment, the system further includes:
and the adjusting module is used for adjusting the length of the data sequence of each training sample in the training sample set and/or adjusting the time-frequency representation form of each training sample in the training sample set.
On the basis of the above embodiment, the adjusting module specifically includes:
and the preprocessing unit is used for performing fast Fourier transform on each training sample in the training sample set and transforming each training sample from time domain data into frequency domain data.
The preprocessing unit provided by the embodiment of the invention can convert the training samples from time domain to frequency domain, and improve the discrimination between the samples, thereby improving the training effect.
On the basis of the above embodiment, the training module specifically includes:
and the structural parameter determining unit is used for training the preset convolutional neural network by adopting a control variable method so as to determine each network structural parameter in the convolutional neural network.
The control variable method provided by the structural parameter determining unit provided by the embodiment of the invention can determine the influence degree of the structural parameters on the classification accuracy and the influence degree of the structural parameters on the classification result in the convolutional neural network training process.
Fig. 3 is a block diagram of an electronic device according to an embodiment of the present invention, and referring to fig. 3, the electronic device includes: a processor (processor)310, a communication Interface (communication Interface)320, a memory (memory)330 and a bus 340, wherein the processor 310, the communication Interface 320 and the memory 330 complete communication with each other through the bus 340. The processor 310 may call logic instructions in the memory 330 to perform the following method: inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network; and obtaining the classification result output by the trained convolutional neural network.
An embodiment of the present invention discloses a computer program product, which includes a computer program stored on a non-transitory computer readable storage medium, the computer program including program instructions, when the program instructions are executed by a computer, the computer can execute the methods provided by the above method embodiments, for example, the method includes: inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network; and obtaining the classification result output by the trained convolutional neural network.
Embodiments of the present invention provide a non-transitory computer-readable storage medium, which stores computer instructions, where the computer instructions cause the computer to perform the methods provided by the above method embodiments, for example, the methods include: inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network; and obtaining the classification result output by the trained convolutional neural network.
Those of ordinary skill in the art will understand that: all or part of the steps for implementing the method embodiments may be implemented by hardware related to program instructions, and the program may be stored in a computer readable storage medium, and when executed, the program performs the steps including the method embodiments; and the aforementioned storage medium includes: various media that can store program codes, such as ROM, RAM, magnetic or optical disks.
Finally, it should be noted that: the above examples are only intended to illustrate the technical solution of the present invention, but not to limit it; although the present invention has been described in detail with reference to the foregoing embodiments, it will be understood by those of ordinary skill in the art that: the technical solutions described in the foregoing embodiments may still be modified, or some technical features may be equivalently replaced; and such modifications or substitutions do not depart from the spirit and scope of the corresponding technical solutions of the embodiments of the present invention.
Claims (6)
1. A method for classifying sea clutter and noise signals is characterized by comprising the following steps:
inputting a radar one-dimensional echo signal to be classified into a trained convolutional neural network, wherein the radar one-dimensional echo signal is a signal returned by a radar monitoring sea target background and comprises a sea clutter signal and a noise signal; the method for inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network specifically comprises the following steps: rearranging data points of the radar one-dimensional echo signals to be classified into a two-dimensional matrix of n x n; inputting the two-dimensional matrix into the trained convolutional neural network;
obtaining a classification result output by the trained convolutional neural network;
before the inputting the radar one-dimensional echo signal to be classified into the trained convolutional neural network, the method further includes:
acquiring sample data with the discrimination degree larger than a preset threshold value as a training sample set;
adjusting the length of a data sequence of each training sample in the training sample set and/or adjusting the time-frequency representation form of each training sample in the training sample set; the adjusting the time-frequency representation form of each training sample in the training sample set specifically includes: performing fast Fourier transform on each training sample in the training sample set, and transforming time domain data of each training sample into frequency domain data;
training a preset convolutional neural network based on the training sample set;
and the discrimination is amplitude discrimination of the sea clutter data and the noise data in the training sample.
2. The method according to claim 1, characterized in that the convolutional neural network is in particular a LeNet.
3. The method according to claim 1, wherein the training a predetermined convolutional neural network based on the training sample set specifically comprises:
and training the preset convolutional neural network by adopting a control variable method to determine each network structure parameter in the convolutional neural network.
4. A sea clutter and noise signal classification system, comprising:
the system comprises an input module, a convolutional neural network and a data processing module, wherein the input module is used for inputting radar one-dimensional echo signals to be classified into the trained convolutional neural network, and the radar one-dimensional echo signals are signals returned by a radar monitoring sea target background and comprise sea clutter signals and noise signals; the method for inputting the radar one-dimensional echo signals to be classified into the trained convolutional neural network specifically comprises the following steps: rearranging data points of the radar one-dimensional echo signals to be classified into a two-dimensional matrix of n x n; inputting the two-dimensional matrix into the trained convolutional neural network;
the classification module is used for acquiring a classification result output by the trained convolutional neural network;
before the radar one-dimensional echo signal to be classified is input into the trained convolutional neural network, the method further comprises the following steps:
acquiring sample data with the discrimination degree larger than a preset threshold value as a training sample set;
adjusting the length of a data sequence of each training sample in the training sample set and/or adjusting the time-frequency representation form of each training sample in the training sample set; the adjusting the time-frequency representation form of each training sample in the training sample set specifically includes: performing fast Fourier transform on each training sample in the training sample set, and transforming time domain data of each training sample into frequency domain data;
training a preset convolutional neural network based on the training sample set;
and the discrimination is amplitude discrimination of the sea clutter data and the noise data in the training sample.
5. An electronic device, comprising a memory and a processor, wherein the processor and the memory communicate with each other via a bus; the memory stores program instructions executable by the processor, the processor invoking the program instructions to perform the method of any of claims 1 to 3.
6. A non-transitory computer-readable storage medium storing computer instructions that cause a computer to perform the method of any one of claims 1 to 3.
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 CN109389058A (en) | 2019-02-26 |
| CN109389058B true 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) |
Families Citing this family (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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| 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 |
| CN112711032B (en) * | 2020-12-07 | 2022-05-13 | 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 |
| CN113189554B (en) * | 2021-04-12 | 2022-06-07 | 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 |
| CN112859034B (en) * | 2021-04-26 | 2021-07-16 | 中国人民解放军国防科技大学 | Natural environment radar echo amplitude model classification method and device |
| 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 |
| CN113759336B (en) * | 2021-08-31 | 2024-04-30 | 东南大学 | Sea clutter suppression method under graph feature learning |
Citations (4)
| 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 |
| 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 |
| 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 |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107203692B (en) * | 2017-05-09 | 2020-05-05 | 哈尔滨工业大学（威海） | Electrocardio data digital signal processing method based on deep convolutional neural network |
| CN107220606B (en) * | 2017-05-22 | 2020-05-19 | 西安电子科技大学 | Radar radiation source signal identification method based on one-dimensional convolutional neural network |
| CN108229404B (en) * | 2018-01-09 | 2022-03-08 | 东南大学 | Radar echo signal target identification method based on deep learning |
-
2018
- 2018-09-25 CN CN201811113859.8A patent/CN109389058B/en active Active
Patent Citations (4)
| 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 |
| 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 |
| 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 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109389058A (en) | 2019-02-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109389058B (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 | |
| CN109522857B (en) | People number estimation method based on generation type confrontation network model | |
| CN107316013B (en) | Hyperspectral image classification method based on NSCT (non-subsampled Contourlet transform) and DCNN (data-to-neural network) | |
| CN110361778B (en) | Seismic data reconstruction method based on generation countermeasure network | |
| CN112001270B (en) | Ground radar automatic target classification and identification method based on one-dimensional convolutional neural network | |
| CN113657491A (en) | Neural network design method for signal modulation type recognition | |
| CN110211127A (en) | Image partition method based on bicoherence network | |
| CN113571067A (en) | Voiceprint recognition countermeasure sample generation method based on boundary attack | |
| CN112560710B (en) | Method for constructing finger vein recognition system and finger vein recognition system | |
| CN111880158A (en) | Radar target detection method and system based on convolutional neural network sequence classification | |
| CN102034111A (en) | Method for identifying and detecting aircraft structural damage conditions in diversified way | |
| CN114117912A (en) | Sea clutter modeling and inhibiting method under data model dual drive | |
| CN115661627A (en) | Single-beam underwater target identification method based on GAF-D3Net | |
| Wang et al. | Research on feature extraction and recognition method of underwater acoustic target based on deep convolutional network | |
| Yi et al. | A bearing fault diagnosis method based on an improved depth residual network | |
| CN115564983A (en) | Target detection method and device, electronic equipment, storage medium and application thereof | |
| Liu et al. | Dim and small target detection in multi-frame sequence using Bi-Conv-LSTM and 3D-Conv structure | |
| Cinelli et al. | Vessel classification through convolutional neural networks using passive sonar spectrogram images | |
| CN109784311B (en) | Target identification method based on linear frequency modulation wavelet atomic network | |
| CN114492540B (en) | Training method and device of target detection model, computer equipment and storage medium | |
| CN116092005A (en) | Fish identification method, system, computer equipment and storage medium | |
| CN112346056B (en) | Resolution characteristic fusion extraction method and identification method of multi-pulse radar signals | |
| CN115169407B (en) | Weak transient signal detection method and system based on time domain depth feature learning | |
| CN112990279A (en) | Radar high-resolution range profile library outside target rejection method based on automatic encoder |
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 |