CN115097391A - Synthetic aperture radar interference suppression method based on automatic encoder - Google Patents

Synthetic aperture radar interference suppression method based on automatic encoder Download PDF

Info

Publication number
CN115097391A
CN115097391A CN202210697543.8A CN202210697543A CN115097391A CN 115097391 A CN115097391 A CN 115097391A CN 202210697543 A CN202210697543 A CN 202210697543A CN 115097391 A CN115097391 A CN 115097391A
Authority
CN
China
Prior art keywords
interference
time
synthetic aperture
aperture radar
signal
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210697543.8A
Other languages
Chinese (zh)
Inventor
黄岩
王韵旋
刘江
陈筠力
刘艳阳
毛源
余旭涛
洪伟
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Southeast University
Shanghai Institute of Satellite Engineering
Original Assignee
Southeast University
Shanghai Institute of Satellite Engineering
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Southeast University, Shanghai Institute of Satellite Engineering filed Critical Southeast University
Priority to CN202210697543.8A priority Critical patent/CN115097391A/en
Publication of CN115097391A publication Critical patent/CN115097391A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/28—Details of pulse systems
    • G01S7/285—Receivers
    • G01S7/292—Extracting wanted echo-signals
    • G01S7/2923—Extracting wanted echo-signals based on data belonging to a number of consecutive radar periods
    • G01S7/2928—Random or non-synchronous interference pulse cancellers
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/35—Details of non-pulse systems
    • G01S7/352—Receivers
    • G01S7/354—Extracting wanted echo-signals
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01S—RADIO DIRECTION-FINDING; RADIO NAVIGATION; DETERMINING DISTANCE OR VELOCITY BY USE OF RADIO WAVES; LOCATING OR PRESENCE-DETECTING BY USE OF THE REFLECTION OR RERADIATION OF RADIO WAVES; ANALOGOUS ARRANGEMENTS USING OTHER WAVES
    • G01S7/00—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00
    • G01S7/02—Details of systems according to groups G01S13/00, G01S15/00, G01S17/00 of systems according to group G01S13/00
    • G01S7/36—Means for anti-jamming, e.g. ECCM, i.e. electronic counter-counter measures
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V10/00—Arrangements for image or video recognition or understanding
    • G06V10/40—Extraction of image or video features
    • G06V10/54—Extraction of image or video features relating to texture

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Radar, Positioning & Navigation (AREA)
  • Remote Sensing (AREA)
  • Theoretical Computer Science (AREA)
  • Computer Networks & Wireless Communication (AREA)
  • Artificial Intelligence (AREA)
  • Data Mining & Analysis (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Multimedia (AREA)
  • Biomedical Technology (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Health & Medical Sciences (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Radar Systems Or Details Thereof (AREA)

Abstract

The invention discloses a synthetic aperture radar interference suppression method based on an automatic encoder, which builds and trains an interference suppression algorithm based on the synthetic aperture radar signal polluted by narrow-band interference and broadband interference. The input of the IMN is a time-frequency spectrogram with interference components, useful information of a target signal is extracted through a self-encoder network structure, and the time-frequency spectrogram of a target echo signal is reconstructed. The invention innovatively uses the automatic encoder in the machine learning method, has the capability of automatically inhibiting the narrow-band interference and the wide-band interference in the synthetic aperture radar, and has higher similarity between the recovered signal and the undisturbed signal.

Description

Synthetic aperture radar interference suppression method based on automatic encoder
Technical Field
The invention belongs to the field of synthetic aperture radar interference suppression, and particularly relates to a synthetic aperture radar interference suppression method based on an automatic encoder.
Background
With the development of electronic technology and mobile wireless technology, the pursuit of greater bandwidth by radio devices is increasing. Radar systems, particularly imaging radar systems represented by synthetic aperture radars, require a large bandwidth to achieve good range resolution. The larger bandwidth means a larger possibility of being affected by interference, therefore, radio frequency interference becomes a main problem of accurate remote sensing of the synthetic aperture radar system, and greatly influences the synthetic aperture radar imaging and subsequent interpretation processes.
During the last decades researchers have been working on how to effectively suppress radio frequency interference in synthetic aperture radar data. Interference suppression methods can be roughly classified into three categories, namely nonparametric methods, parametric methods and semi-parametric methods. The nonparametric method usually projects the original data onto a signal subspace, and then filters out interference signals by using a filter method. Parametric methods typically estimate parameters of narrowband and wideband radio frequency interference models. The semi-parametric method uses machine learning models (e.g., sparse recovery and low rank recovery methods) to suppress radio frequency interference.
In recent years, deep learning tools have enjoyed excellent performance in almost all tasks in the fields of computer vision and natural language processing. The supervised learning framework can extract the hierarchical features of the targets in the image and further be used for classifying different targets. Deep convolutional neural networks can be used to mitigate narrowband and wideband interference in synthetic aperture radar images. While for unsupervised learning frameworks such as autoencoders and generative warfare networks, they can naturally discover the intrinsic properties of the data without any pre-assigned labels and are widely used in the field of image denoising. Furthermore, unsupervised learning methods based on Principal Component Analysis (PCA) are widely used to mitigate radio frequency interference in corrupted synthetic aperture radar data.
Disclosure of Invention
The purpose of the invention is as follows: in view of the prior art, an interference suppression method for a synthetic aperture radar based on an automatic encoder is provided, which is characterized in that a time-frequency spectrogram with interference components is used as input, and useful information of a target signal is extracted through a network structure of the automatic encoder, so that the time-frequency spectrogram of a target echo signal is reconstructed.
The invention relates to a synthetic aperture radar interference suppression method based on an automatic encoder, which comprises the following steps:
step 1: acquiring time-frequency characteristic diagrams of polluted and undisturbed synthetic aperture radar signals, and taking the time-frequency characteristic diagrams as data sets, wherein the data sets are divided into training sets and test sets;
step 2: constructing a depth separable convolution neural network based on an automatic encoder, wherein the depth separable convolution neural network is used for interference suppression and useful signal recovery, and automatically extracting and selecting textural features and spatial information in an image; and training the deep separable convolutional neural network by adopting a training set, optimizing parameters and a network structure to obtain a high-accuracy model, taking the time-frequency characteristic diagram polluted by interference as input, taking the corresponding time-frequency domain characteristic diagram not polluted by interference as expected output, extracting useful information of a target signal, reconstructing the time-frequency characteristic diagram of the target echo signal, and obtaining the trained deep separable convolutional neural network.
And step 3: and converting the time-frequency spectrogram of the reconstructed target echo signal from a short-time Fourier transform domain back to a time domain, and then reconnecting the time-domain signals together along a distance dimension. Fourier transformation and inverse Fourier transformation are respectively carried out along the fast time dimension and the slow time dimension to obtain a restored synthetic aperture radar image;
and 4, step 4: and (3) inputting the interfered synthetic aperture radar signals serving as the test set into the trained deep separable convolutional neural network to obtain a time-frequency spectrogram of the recovered synthetic aperture radar signals, and repeating the step (3) to obtain the recovered synthetic aperture radar image.
Further, in step 1, a time-frequency characteristic diagram of a polluted and undisturbed synthetic aperture radar signal is obtained and used as a data set, and the specific process is as follows:
selecting a synthetic aperture radar signal which is not polluted by interference, and adding narrow-band interference and broadband interference to the radar signal in a fast time domain to form a synthetic aperture radar signal polluted by interference;
dividing the synthetic aperture radar signals polluted by interference and the synthetic aperture radar signals not polluted by the interference into a plurality of blocks in a slow time dimension along the distance dimension; and performing short-time Fourier transform along the distance dimension to obtain a filtered time-frequency characteristic diagram of the polluted and undisturbed synthetic aperture radar signal, and taking the time-frequency characteristic diagram as a data set.
Further, the narrowband interference is multi-frequency point narrowband interference, and the broadband interference is frequency modulation continuous wave interference.
Further, the deep separable convolutional neural network comprises an encoder and a decoder, and the encoder and the decoder are symmetrical in structure;
the encoder comprises 9 depth separable convolution layers (partitioned _ conv2d layers) and 8 batch normalization layers (batch _ normalization layers), the sizes of convolution kernels are all 3 x 3, and the step size is 1;
the 9 convolutional layers perform a deep convolution operation on each channel of the input data, and then linearly connect the output of the deep convolution using a point convolution. The network structure can greatly reduce the parameter quantity and the calculated quantity of the model, thereby improving the detection rate under the condition that the detection precision is not obviously changed.
Further, a loss function of the depth separable convolutional neural network based on the automatic encoder is expressed by Mean Square Error (MSE), which reflects a difference degree between a radar echo time-frequency spectrum characteristic diagram without interference components and a radar echo time-frequency spectrum diagram after interference suppression, and the loss function can be expressed as:
Figure BDA0003702721400000031
where M and N represent the dimensions of the image, I ori (m, n) represents the corresponding gray value of the radar echo time-frequency spectrogram without interference component on (m, n), G IMN (I inp (m, n)) is the result of IMN of the corresponding gray value of the radar echo time-frequency spectrogram with the interference component on (m, n).
Has the advantages that:
1. the invention uses an automatic encoder to realize automatic suppression of interference.
2. The invention uses the normalization layer to improve the learning speed and stability of the model
3. The present invention uses the mean square error as a loss function to achieve the recovery of the interfered signal.
4. The invention uses a depth separable convolutional neural network to automatically extract and select textural features and spatial information in an image.
Drawings
FIG. 1 is a synthetic aperture radar image uncontaminated by interference;
FIG. 2 is a synthetic aperture radar image contaminated with narrow-band interference;
FIG. 3 is a synthetic aperture radar image contaminated with broadband interference;
FIG. 4 is a time-frequency spectrum of a synthetic aperture radar signal contaminated with narrowband interference;
FIG. 5 is a time-frequency spectrum of a synthetic aperture radar signal contaminated with broadband interference;
FIG. 6 is a schematic diagram of a deep separable convolutional neural network architecture based on an auto-encoder;
FIG. 7 is a time-frequency spectrum of a recovered synthetic aperture radar signal contaminated with narrow-band interference;
FIG. 8 is a time-frequency spectrum of a recovered synthetic aperture radar signal contaminated with broadband interference;
FIG. 9 is a recovered synthetic aperture radar image contaminated with narrow-band interference;
FIG. 10 is a recovered synthetic aperture radar image contaminated with broadband interference;
Detailed Description
The invention is further explained below with reference to the drawings.
Carefully select a synthetic aperture radar signal that is not contaminated by interference, with a distance dimension of 1024 and an azimuth dimension of 3000, and plot the synthetic aperture radar image as shown in fig. 1. In the fast time domain, multi-frequency point narrowband interference is added to the radar signal as narrowband interference, frequency modulated continuous wave interference is added as wideband interference, and a synthetic aperture radar image contaminated by narrowband interference as shown in fig. 2 and a synthetic aperture radar image contaminated by wideband interference as shown in fig. 3 are drawn. Then, the synthetic aperture radar signals which are polluted by interference and are not polluted by interference are divided into 8 segments along the distance dimension, each segment is 128 sampling points, and short-time Fourier transform is carried out along the distance dimension to obtain a time-frequency characteristic diagram of the filtered polluted synthetic aperture radar signals with the size of 128 multiplied by 128. As shown in fig. 4, the time-frequency characteristic diagram of the filtered polluted synthetic aperture radar signal with size of 128 × 128 is obtained by performing short-time fourier transform along the distance dimension after the synthetic aperture radar signal polluted by the narrow-band interference is cut; as shown in fig. 5, a time-frequency characteristic diagram of a 128 × 128 filtered contaminated synthetic aperture radar signal is obtained by performing short-time fourier transform along the distance dimension after being cut by the synthetic aperture radar signal contaminated by the broadband interference.
And obtaining a synthetic aperture radar signal data set with interference pollution and without interference pollution in the time-frequency domain through the division and time-frequency domain processing of all synthetic aperture radar signals. An autoencoder-based depth separable convolutional neural network as shown in fig. 6 is constructed, which includes 9 depth separable convolutional layers (partitioned _ conv2d layers) and 8 batch normalization layers (batch _ normalization layers), and the sizes of the convolutional cores are 3 × 3 and the step size is 1. The loss function of the network is expressed by Mean Square Error (MSE), which reflects the difference degree between the radar echo time-frequency spectrogram without interference components and the radar echo time-frequency spectrogram after interference suppression. The loss function of an interference suppression network can be expressed as:
Figure BDA0003702721400000041
where M and N represent the dimensions of the image, I ori (m, n) represents the corresponding gray value of the radar echo time-frequency spectrogram without interference component on (m, n), G IMN (I inp (m, n)) is the result of IMN of the corresponding gray value of the radar echo time-frequency spectrogram with the interference component on (m, n).
And taking the time-frequency spectrogram with interference components as input, taking the time-frequency spectrogram without interference pollution as expected output, obtaining a synthetic aperture radar interference suppression and useful signal recovery model based on an automatic encoder, and optimizing parameters and a network structure to obtain a high-accuracy model. And extracting useful information of the target signal through a network structure of the self-encoder, and reconstructing a time-frequency spectrogram of the target echo signal. FIG. 7 is a time-frequency spectrum of a synthetic aperture radar signal contaminated by narrow-band interference after recovery by an automatic encoder's deep separable convolutional neural network; figure 8 is a time-frequency spectrum of a synthetic aperture radar signal contaminated with broadband interference after recovery by an automatic encoder's deep separable convolutional neural network.
And converting the time-frequency spectrogram of the reconstructed target echo signal from a short-time Fourier transform domain back to a time domain, and then reconnecting the time-domain signals together along a distance dimension. And respectively performing Fourier transform and inverse Fourier transform along the fast time dimension and the slow time dimension to obtain the restored synthetic aperture radar image. FIG. 9 is an imaging result of a synthetic aperture radar signal contaminated by narrow-band interference recovered by a deep separable convolutional neural network of an auto-encoder; figure 10 is an imaging result of a synthetic aperture radar signal contaminated with broadband interference recovered by a deep separable convolutional neural network of an auto-encoder.
The foregoing is only a preferred embodiment of the present invention, and it should be noted that, for those skilled in the art, various modifications and decorations can be made without departing from the principle of the present invention, and these modifications and decorations should also be regarded as the protection scope of the present invention.

Claims (5)

1. A synthetic aperture radar interference suppression method based on an automatic encoder is characterized by comprising the following steps:
step 1: acquiring time-frequency characteristic diagrams of polluted and undisturbed synthetic aperture radar signals, and taking the time-frequency characteristic diagrams as data sets, wherein the data sets are divided into training sets and test sets;
step 2: building a depth separable convolutional neural network based on an automatic encoder, wherein the depth separable convolutional neural network is used for interference suppression and useful signal recovery; and training the deep separable convolutional neural network by adopting a training set, taking the time-frequency characteristic graph polluted by interference as input, taking the corresponding time-frequency characteristic graph not polluted by interference as expected output, extracting useful information of a target signal, and reconstructing a time-frequency characteristic graph of a target echo signal to obtain the trained deep separable convolutional neural network.
And step 3: and converting the time-frequency spectrogram of the reconstructed target echo signal from a short-time Fourier transform domain back to a time domain, and then reconnecting the time-domain signals together along a distance dimension. Fourier transformation and inverse Fourier transformation are respectively carried out along the fast time dimension and the slow time dimension to obtain a restored synthetic aperture radar image;
and 4, step 4: and (3) inputting the interfered synthetic aperture radar signals serving as the test set into the trained deep separable convolutional neural network to obtain a time-frequency spectrogram of the recovered synthetic aperture radar signals, and repeating the step (3) to obtain the recovered synthetic aperture radar image.
2. The method for suppressing interference of synthetic aperture radar based on automatic encoder as claimed in claim 1, wherein the step 1 obtains time-frequency characteristic diagram of the polluted and undisturbed synthetic aperture radar signal as data set, and the specific process is as follows:
selecting a synthetic aperture radar signal which is not polluted by interference, and adding narrow-band interference and broadband interference to the radar signal in a fast time domain to form a synthetic aperture radar signal polluted by interference;
respectively dividing the synthetic aperture radar signal polluted by interference and the synthetic aperture radar signal not polluted by interference into a plurality of blocks in a slow time dimension along a distance dimension; and performing short-time Fourier transform along the distance dimension to obtain a filtered time-frequency characteristic diagram of the polluted and undisturbed synthetic aperture radar signal, and taking the time-frequency characteristic diagram as a data set.
3. The method according to claim 2, wherein the narrowband interference is multi-frequency narrowband interference, and the wideband interference is frequency modulated continuous wave interference.
4. The method of claim 1, wherein the deep separable convolutional neural network comprises an encoder and a decoder, and the encoder and the decoder are structurally symmetric;
the encoder comprises 9 depth-separable convolutional layers and 8 batch normalization layers; the 9 convolutional layers perform a deep convolution operation on each channel of the input data, and then linearly connect the output of the deep convolution using a point convolution.
5. The method for interference suppression of synthetic aperture radar based on automatic encoder as claimed in claim 2, wherein the loss function of the depth separable convolutional neural network based on automatic encoder is expressed by Mean Square Error (MSE), and the loss function can be expressed as:
Figure FDA0003702721390000021
where M and N represent the dimensions of the image, I ori (m, n) represents the corresponding gray value of the radar echo time-frequency spectrogram without interference component on (m, n), G IMN (I inp (m, n)) is the result of IMN of the corresponding gray value of the radar echo time-frequency spectrogram with the interference component on (m, n).
CN202210697543.8A 2022-06-20 2022-06-20 Synthetic aperture radar interference suppression method based on automatic encoder Pending CN115097391A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210697543.8A CN115097391A (en) 2022-06-20 2022-06-20 Synthetic aperture radar interference suppression method based on automatic encoder

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210697543.8A CN115097391A (en) 2022-06-20 2022-06-20 Synthetic aperture radar interference suppression method based on automatic encoder

Publications (1)

Publication Number Publication Date
CN115097391A true CN115097391A (en) 2022-09-23

Family

ID=83291037

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210697543.8A Pending CN115097391A (en) 2022-06-20 2022-06-20 Synthetic aperture radar interference suppression method based on automatic encoder

Country Status (1)

Country Link
CN (1) CN115097391A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116540190A (en) * 2023-07-06 2023-08-04 西安电子科技大学 End-to-end self-supervision intelligent interference suppression method and device and electronic equipment

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN116540190A (en) * 2023-07-06 2023-08-04 西安电子科技大学 End-to-end self-supervision intelligent interference suppression method and device and electronic equipment

Similar Documents

Publication Publication Date Title
CN110927706A (en) Convolutional neural network-based radar interference detection and identification method
CN111239697B (en) Multidimensional domain combined SAR broadband interference suppression method based on low-rank matrix decomposition
CN111582320B (en) Dynamic individual identification method based on semi-supervised learning
CN112327259B (en) Method and device for eliminating interference signals in SAR image
CN111222442A (en) Electromagnetic signal classification method and device
CN115097391A (en) Synthetic aperture radar interference suppression method based on automatic encoder
CN108428221A (en) A kind of neighborhood bivariate shrinkage function denoising method based on shearlet transformation
Gongwen et al. On medical image segmentation based on wavelet transform
CN112926555B (en) Small sample passive behavior sensing method based on self-encoder data enhancement
CN107123431A (en) A kind of underwater sound signal noise-reduction method
CN101430788A (en) Lunette local Wiener filtering method based on second generation Curvelet transformation
CN113205564B (en) SAR intelligent target edge reconstruction method
CN113238193B (en) SAR echo broadband interference suppression method based on multi-component joint reconstruction
CN116405100B (en) Distortion signal restoration method based on priori knowledge
CN109190693A (en) Variant target high Resolution Range Profile Identification method based on block management loading
CN112929141A (en) Unmanned aerial vehicle detection and identification method and system based on graph-borne signal matching
CN102509268B (en) Immune-clonal-selection-based nonsubsampled contourlet domain image denoising method
CN112927169B (en) Remote sensing image denoising method based on wavelet transformation and improved weighted kernel norm minimization
CN112197964B (en) Local impact signal positioning method based on spectrum pq-mean
CN115828069A (en) End-to-end magnetic anomaly signal noise reduction method based on deep learning
CN114881215A (en) Millimeter wave radar data screening neural network processing method
CN113848532A (en) FMCW radar signal noise reduction system and method based on noise reduction model
CN114674410A (en) Component number time-varying underwater acoustic signal instantaneous frequency estimation method
CN107315713B (en) One-dimensional signal denoising and enhancing method based on non-local similarity
CN112213698A (en) Deception jamming identification method based on sparse representation classification

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