CN108318891A - It is a kind of that method is forced down based on the SAL data secondary lobes for improving SVA and CS - Google Patents
It is a kind of that method is forced down based on the SAL data secondary lobes for improving SVA and CS Download PDFInfo
- Publication number
- CN108318891A CN108318891A CN201711214945.3A CN201711214945A CN108318891A CN 108318891 A CN108318891 A CN 108318891A CN 201711214945 A CN201711214945 A CN 201711214945A CN 108318891 A CN108318891 A CN 108318891A
- Authority
- CN
- China
- Prior art keywords
- sal
- dimensional
- signal
- sva
- image
- 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
- 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
- G01S17/00—Systems using the reflection or reradiation of electromagnetic waves other than radio waves, e.g. lidar systems
- G01S17/88—Lidar systems specially adapted for specific applications
- G01S17/89—Lidar systems specially adapted for specific applications for mapping or imaging
- G01S17/90—Lidar systems specially adapted for specific applications for mapping or imaging using synthetic aperture techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/16—Matrix or vector computation, e.g. matrix-matrix or matrix-vector multiplication, matrix factorization
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Data Mining & Analysis (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Optimization (AREA)
- Mathematical Analysis (AREA)
- Pure & Applied Mathematics (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Databases & Information Systems (AREA)
- Software Systems (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Computer Networks & Wireless Communication (AREA)
- Electromagnetism (AREA)
- Radar, Positioning & Navigation (AREA)
- Remote Sensing (AREA)
- Image Processing (AREA)
Abstract
Method is forced down based on the SAL data secondary lobes for improving SVA and CS the invention discloses a kind of, solves that SAL image data secondary lobes are higher, and image quality is bad, the problem of image resolution ratio deficiency.Realize that step includes:Generate the primary data matrix of synthetic aperture laser radar imaging;Construction improves SVA algorithm models, is handled to using the model SAL data matrixes in distance;Construct sparse signal needed for CS and observation basic matrix;Solve compressed sensing underdetermined equation;Generate SAL image result matrixes, imaging;It completes to force down secondary lobe processing to SAL image datas, obtains high-definition picture.The present invention will improve SVA algorithms and CS is combined, it can be under the premise of keeping main lobe energy and image resolution ratio, suppressed sidelobes subtracts wide main lobe, reduces the operand and amount of storage of SAL image real time transfers, more efficiently forces down Synthetic Aperture Laser Radar data secondary lobe.For cutting down echo noise in synthetic aperture laser radar imaging, SAL image resolution ratios and picture quality are promoted.
Description
Technical field
The invention belongs to radar data processing technology field, the data secondary lobe being related in Synthetic Aperture Laser Radar forces down skill
Art, specifically a kind of Synthetic Aperture Laser Radar based on room for improvement apodization (SVA) and compressed sensing Reconstruction Method (CS)
(SAL) data secondary lobes forces down method.For cutting down echo noise in synthetic aperture laser radar imaging, SAL images point are promoted
Resolution and picture quality.
Background technology
Synthetic Aperture Laser Radar (SAL) is the high score using optical synthesis aperture technology and relevant heterodyne detection technology
Resolution imaging laser radar, application prospect are extensive;And in SAL imagings, return laser beam adulterates noise, and sidelobe level is caused to be lifted
Height, this will influence image resolution ratio, therefore to force down processing very necessary for secondary lobe;In addition, the high secondary lobe meeting of strong scattering point target
Neighbouring weak signal target is fallen into oblivion, follow-up SAL image objects detection is influenced, thus secondary lobe forces down the raising for SAL target acquisition quality
It is same essential.
For this higher problem of SAL image secondary lobes, conventional method in signal frequency domain windowing process by realizing secondary lobe
It forces down, but this method can lead to the broadening of main lobe and the reduction of resolution ratio except realizing that secondary lobe forces down, and influence the figure of SAL
Image quality amount;In addition, existing space apodization algorithm (SVA) is only effective to the signal of integral multiple nyquist sampling, while making master
Valve energy loses, it is ineffective to force down SAL image secondary lobes;And compressed sensing technology (CS) can also force down secondary lobe and make an uproar
Sound, but realize that the technology has certain degree of rarefication firstly the need of handled signal, this will be difficult to be suitable for target scattering signal
For under non-sparse SAL scenes.
In Synthetic Aperture Laser Radar field, objectively need one kind that can keep main lobe energy and image resolution ratio
Under the premise of, the more efficiently new algorithm of suppressed sidelobes.
Invention content
Processing this undesirable problem of imaging effect is forced down for SAL images secondary lobe in the prior art, the present invention proposes one
Kind can improve the synthesis hole based on room for improvement apodization (SVA) and compressed sensing Reconstruction Method (CS) of SAL image resolution ratios
Diameter laser radar (SAL) data secondary lobe forces down method.
The present invention is a kind of to force down method based on the SAL data secondary lobes for improving SVA and CS, which is characterized in that includes
Following steps:
Step 1, the primary data matrix of synthetic aperture laser radar imaging is generated:Input actual measurement SAL echo complex datas,
Generate M × N-dimensional SAL imaging primary data matrixes X;
Step 2, construction improves SVA algorithm models, is handled to using the model SAL data matrixes in distance:It is right
SAL primary datas matrix is handled with non-integral multiple Nyquist resamplings SVA and is had into the upward improvement SVA processing of row distance
There are M × N-dimensional SAL the first matrixs of consequence of image of signal sparsity;
Step 3, sparse signal needed for CS is constructed:Distance is tieed up to letter to every M of M × N-dimensional SAL the first matrixs of consequence of image
Rarefaction representation is carried out again after number normalized;
Step 4, basic matrix is observed needed for construction CS:Independent identically distributed gaussian random matrix is set as M × N-dimensional SAL
The observation basic matrix of every M dimensional signals in the first matrix of consequence of image later carries out every M dimensional signals using observation basic matrix
Compression processing obtains the every M of SAL images and ties up compression result signal;
Step 5, compressed sensing underdetermined equation is solved:A M every to SAL images ties up compression result in a manner of linear programming for solution
The CS underdetermined equations of signal are solved, and are obtained the every M of SAL images and are tieed up compressed sensing estimate vector, reconstruct SAL image originals letter
Number;
Step 6, SAL image result matrixes, imaging are generated:M × N is generated by every M dimension compressed sensing estimate vectors
SAL the second matrixs of consequence of image are tieed up, obtain forcing down result based on the SAL data secondary lobes for improving SVA and CS after carrying out imaging
Image.
The present invention compared with the conventional method, has the following advantages:
The present invention uses improved SVA algorithms, improvement SVA algorithms not only to inherit original SVA algorithms and can inhibit image first
The advantages of signal secondary lobe, and bandwidth signals main lobe can be subtracted, the energy loss of image will be further reduced in this, and SAL images is made to have
There is higher resolution;
The present invention is improved after SVA algorithm process SAL images and uses compressed sensing re-construction theory again, is appropriately utilized
Signal sparsity after the improved SVA algorithms of SAL image data matrixs, further forces down SAL picture signal secondary lobes, and to useful
Signal is restored, while the operand of data and amount of storage being made to reduce.
The present invention will creatively improve SVA algorithms and CS is combined, and can keep main lobe energy and image resolution ratio
Under the premise of, the operand and amount of storage of SAL image real time transfers are reduced, Synthetic Aperture Laser Radar number is more efficiently forced down
According to secondary lobe.
Description of the drawings
Fig. 1 forces down method flow diagram for the present invention's based on the SAL data secondary lobes for improving SVA and CS;
Fig. 2 is weighting function and signal sampling point and the graph of relation of peak point difference;
Fig. 3 is the imaging effect comparison diagram handled Chinese Academy of Sciences's actual measurement SAL data using the present invention, wherein
Fig. 3 (a) is original SAL echo datas direct imaging result figure, and Fig. 3 (b) is at improvement SVA progress data only through the invention
Design sketch after reason, Fig. 3 (c) are that the design sketch after data processing is all carried out according to flow of the present invention.
Specific implementation mode
It elaborates with reference to the accompanying drawings and detailed description to the present invention.
Embodiment 1
Include not only target information to be detected in the image that Synthetic Aperture Laser Radar echo is generated through signal processing, also mixes
Miscellaneous ambient noise, causes image data sidelobe level to raise, picture quality is bad, it is therefore necessary to affected SAL
Initial pictures carry out secondary lobe and force down processing, and to force down SAL image secondary lobes ineffective for the prior art, it is difficult to meet high-resolution
The requirement of SAL images, for this purpose, the present invention proposes a kind of forcing down based on the SAL data secondary lobes for improving SVA and CS by innovation
Method includes following steps referring to Fig. 1:
Step 1, the primary data matrix of synthetic aperture laser radar imaging is generated:For actual measurement SAL echoes, through number
According to conversion, converted SAL echo complex matrixs are inputted, generate SAL imaging primary data matrixes X.To wait for carrying out it
The algorithm process of subsequent step.
Step 2, construction improves SVA algorithm models, is handled to application enhancements SVA SAL data matrixes in distance:It is right
SAL primary datas matrix is handled with non-integral multiple Nyquist resamplings SVA and is had into the upward improvement SVA processing of row distance
M × N-dimensional SAL the first matrixs of consequence of image for having signal sparsity, with suppressed sidelobes, subtract wide main lobe, while signal being thinned out.
Step 3, sparse signal needed for CS is constructed:Distance is tieed up to letter to every M of M × N-dimensional SAL the first matrixs of consequence of image
Rarefaction representation is carried out again after number normalized, is carried out the calculating of CS reconstruction signals to be follow-up and is prepared.The present invention is to M × N-dimensional
Each column signal of the first matrix of consequence of SAL images is handled.
Step 4, basic matrix is observed needed for construction CS:Independent identically distributed gaussian random matrix is set as M × N-dimensional SAL
The observation basic matrix of every M dimensional signals in the first matrix of consequence of image later carries out every M dimensional signals using observation basic matrix
Compression processing obtains the every M of SAL images and ties up compression result signal, and pickup reduces number for the useful information needed for SAL images
It is also to prepare for follow-up progress CS signal reconstructions calculating here according to amount of storage.
Referring to Fig.1, the observation basic matrix design process in step 3 and step 4 of the present invention can execute parallel, can not also
It is successively serial to execute, it prepares jointly for the every M dimensional signal compression processings of subsequent M × N-dimensional SAL images.
Step 5, compressed sensing underdetermined equation is solved:A M every to SAL images ties up compression result in a manner of linear programming for solution
The CS underdetermined equations of signal are solved, and SAL image original signals are reconstructed, obtain SAL images every M dimension compressed sensing estimate to
Amount, that is, complete the compressed sensing estimation of N number of M dimensional signals vector.
Compressed sensing technology needs based on sparse signal, and needs one observation basic matrix of construction, therefore the present invention
It is to be used as place mat to solve CS underdetermined equations, reconstructing SAL image original signals in step 3 and step 4.
Step 6, SAL image result matrixes, imaging are generated:M is directly generated by every M dimension compressed sensing estimate vectors
× N-dimensional SAL the second matrixs of consequence of image carry out imaging to M × N-dimensional SAL the second matrixs of consequence of image, obtain based on improvement
The SAL data secondary lobes of SVA and CS force down result images.
In the prior art, have and individually carry out forcing down secondary lobe processing using SVA, also have and individually use compressed sensing technology CS
It carrying out forcing down secondary lobe processing, the present invention is imaged primary data matrix X for SAL, will be applied thereon after SVA algorithm improvements,
The SAL images after secondary lobe tentatively forces down are obtained, next again in the SAL images by CS theoretical origins after secondary lobe tentatively forces down,
SAL image of the secondary lobe through further forcing down is obtained, to more efficiently force down Synthetic Aperture Laser Radar image data secondary lobe.
Embodiment 2
Method is forced down based on the SAL data secondary lobes for improving SVA and CS with embodiment 1, acquisition M × N-dimensional in step 2
The process of the first matrix of consequence of SAL images includes:
2a) the improvement of SVA algorithms:
To M × N-dimensional data matrix X of initial SAL images into row distance to improvement SVA algorithm process, to each columns
According to real and imaginary parts carry out identical improvement SVA algorithm process respectively and be denoted as the real part of each column data by taking real part as an example
M dimensional signal vectors I, M are positive integer, and carrying out non-integral multiple Nyquist resamplings to signal vector I obtains output signal vector
I0:
I0(n)=I (n)+α (n) * (I (n-1/Ns)+I(n+1/Ns))
Wherein, I (n) indicates that n-th of sample point data of signal vector I, n are the serial numbers of X line numbers, and 1≤n≤M, n are just
Integer, 1/NsFor non-integral multiple Nyquist sample rates, 0 < NsEqual interval sampling point data item sum adds before and after < 1, α (n) are n
Weight function, I0(n) it is output of n-th of sample point data I (n) after non-integral multiple Nyquist resamplings in signal vector I
Value.For n 1 to the corresponding output valve in the sections M, overall structure output signal is vectorial.
The present invention improves in original SVA algorithms in the way of the progress resampling of integral multiple Nyquist sample rates, and use is non-
Integral multiple Nyquist sample rates carry out resampling to obtain output signal vector, so that main lobe is become while low signal secondary lobe
It is narrow, reduce main lobe energy loss.
It is 2b) that signal is made to export side-lobe energy minimum, needs to make signal real and imaginary parts energy while reaching minimum, with reality
For portion, the weighting function α (n) of data item sum should meet calculation formula at equal intervals before and after n:
Wherein, derivative operation is sought in d expressions, | I0(n)2For signal real part energy.
Obtain the solution expression formula of weighting function α (n):
α (n)=- I (n)/(I (n-1/Ns)+I(n+1/Ns)) (2)
When with n0As signal peak value point, it is contemplated that the expression formula of range value at signal sampling point:
I (n)=sinc [Ns(n-n0)], it substitutes this expression into formula (2), obtains α (n) and expressed through deformed solution
Formula:
The expression formula contains the value of non-integral multiple Nyquist resamplings weighting function between sampled point and peak point
The variation of distance and the information changed, when | n-n0| when < 1, sampled point should retain sample point data, at this time within the scope of main lobe
α (n) < 0;When | n-n0| when >=1, sampled point should give up sample point data, at this time 0≤α (n)≤0.5 in secondary lobe region, this
It is equal with the original SVA Algorithm Analysis conclusion of integral multiple Nyquist resamplings.
2c) construction improves SVA algorithm models:
Further to force down secondary lobe, while main lobe being inhibited to broaden, resampling is carried out with non-integral multiple Nyquist sample rates
It improves SVA algorithms the weighting function value condition that signal exports is optimized, the difference of foundation weighting function α (n) value,
Any sample point data I (n) is obtained in signal vector I after non-integral multiple Nyquist resamplings improve SVA algorithm process not
With output valve I0m(n):
With reference to Fig. 2, the graph of relation of Fig. 2 differences between weighting coefficient and signal sampling point and peak point, to make main lobe
Width reduces, and image resolution ratio improves, and enables γminLess than zero, this makes distance between sampled point and peak point be less than 1/Ns;To make
Weak signal target main lobe is effectively recovered from the secondary lobe of multiple strong targets, true SAL scenes is restored, enables γmaxMore than 1/2.
The imaginary part of the N column signal vectors of M × N-dimensional data matrix X of initial SAL images is carried out identical non-integral multiple
Nyquist resamplings improve SVA algorithm process, and generate M × N-dimensional SAL the first matrixs of consequence of image, and M indicates M × N-dimensional SAL figures
As the line number of the first matrix of consequence, N indicates the columns of M × N-dimensional SAL the first matrixs of consequence of image.
The improvement SVA algorithm models that the present invention is built reflect comprehensively urgently promotes wanting for image resolution ratio under SAL scenes
It asks, describes the novel non-linearity suppressing method of SAL image data secondary lobes.
Improvement SVA algorithms used in the present invention, which not only inherit original SVA algorithms, can make signal secondary lobe be inhibited
The advantages of, and bandwidth signals main lobe can be subtracted, it is further reduced energy loss, makes SAL images that there is more good resolution ratio,
Make SAL image datas that there is signal sparsity simultaneously, this creates condition for the processing of next compressed sensing.
Embodiment 3
Method is forced down based on the SAL data secondary lobes for improving SVA and CS with embodiment 1-2, in step 3 the construction
The process of sparse signal needed for CS includes:
Whole M dimensional signals vectors of M × N-dimensional SAL the first matrixs of consequence of image 3a) are subjected to identical normalized,
Remember that one of whole M dimensional signals M dimensional signal vectors are I', is to the expression formula after signal vector I' normalizeds:
I "=- orth (I')
Wherein, orth indicates that the normalization operation to vector, I " indicate to tie up letter to the M after signal vector I' normalizeds
Number vector, | I " |=1;
3b) signal indicates rarefaction, and it is that M × M ties up unit matrix that setting M × M, which ties up sparse basis array ψ,
M dimensional signal vectors I " is expressed as:
I "=ψ * s
Wherein, ψ indicates that the sparse basis array of M dimensional signal vectors I ", s indicate sparse coefficients of the I " on ψ, for M × 1 tie up to
Amount, in the present embodiment, s=I ".
The real and imaginary parts of SAL image datas are pressed row and carry out rarefaction representation by the present invention, are convenient data processing, will be believed
After number vector I' normalizeds, setting M × M ties up sparse basis arrays of the unit matrix ψ as sparse signal vector I ".
Embodiment 4
Method is forced down based on the SAL data secondary lobes for improving SVA and CS with embodiment 1-3, it is described to M in step 4
The processing procedure of observation basic matrix needed for every M dimensional signals vector I " the constructions CS of × N-dimensional SAL the first matrixs of consequence of image is identical,
Include:
4a) determine the reasonable dimension of observation basic matrix:
Note observation basic matrix is the line number that A × B ties up matrix φ, A expression φ, and B indicates the columns of φ, and B should be with signal vector
The dimension M of I " is equal, and A should meet calculation expression:
A=fix (K*ln (M/K) * β)
Wherein, K indicates that the number of the nonzero element contained by M dimensional signal vectors I ", ln expressions seek natural logrithm, fix
Indicating that truncation floor operation, β are observed differential, the value of β should make K < A < < M, in the present embodiment, β=1.For M dimensional signals
Vectorial I ", it is to be substituted into obtained by above formula by the K mean values of N number of M dimensional signals vector I " that the line number A for observing basic matrix φ, which is fixed value,
End value.
4b) the observation basic matrix of design M dimensional signal vectors I ":
The design of the observation basic matrix of M dimensional signal vectors I ", should make observation basic matrix φ uncorrelated to sparse basis array ψ,
And keeping the matrix being made of arbitrary A column vector in observation basic matrix φ nonsingular, setting observation basic matrix is that A × B dimensions are independent
With the gaussian random matrix of distribution, the observation basic matrix expression formula of M dimensional signal vectors I " is:
φ=randn (A, B)+i*randn (A, B)
Wherein, randn (A, B) expressions are sought in mean value being 0, and variance is that A × B of 1 normal distribution ties up random matrix, i tables
Show imaginary unit.
4c) compression processing M dimensional signals vector I ":
Compression processing is carried out to M dimensional signal vector I " signals using observation basic matrix, compression processing expression formula is:
Y=φ * I "
Wherein, Y indicates compressed treated the M dimension compression result vectors of M dimensional signal vectors I ".
Present invention determine that the reasonable dimension of observation basic matrix needed for CS, and independent identically distributed gaussian random matrix is set and is made
For the observation basic matrix of every M dimensional signals in M × N-dimensional SAL the first matrixs of consequence of image, meets and observe base in compressive sensing theory
The design requirement of matrix.
Embodiment 5
Method is forced down based on the SAL data secondary lobes for improving SVA and CS with embodiment 1-4, in steps of 5 the solution
The process of compressed sensing underdetermined equation includes:
5a) compressed sensing underdetermined equation mathematic(al) representation is:
Y=φ * x
Wherein, x indicates that sparse signal, φ indicate that observation basic matrix, y indicate compression result vector;
The theoretically reconstruct of original signal, by l0Signal reconstruction optimization problem is solved under norm, reconstruct expression formula is:
Wherein, α indicates that sparse coefficients of the x on ψ, x=ψ * α, the meaning for reconstructing expression formula are, y=φ * ψ * α about
Under the conditions of beam, seek making the optimum estimation value α ' when nonzero element number minimum in vectorial α.
5b) reconstruct SAL image original signals:
X=I ", α=s, y=Y are enabled, sparse coefficient s, M of M dimensional signal vectors I ", I " on its sparse basis array ψ is tieed up
Compression result vector Y substitutes into compressed sensing underdetermined equation and reconstruct expression formula;
Convert the optimization problem of theoretically signal reconstruction to linear programming for solution problem, expression formula is:
Thus to obtain the estimation s' of sparse coefficient s;
The compressed sensing estimated value expression formula of any M dimensional signals vector I " of M × N-dimensional SAL the first matrixs of consequence of image is:
I " '=ψ * s'
Wherein, ψ indicates that sparse basis array, I " ' indicate the compressed sensing estimated value of original signal I ", and I " ' is M dimensional vectors.
The present invention is in the reconstruct original signal step for carrying out compressed sensing processing to M × N-dimensional SAL the first matrixs of consequence of image
In, it uses linear programming mode and solves underdetermined equation, be the reasonable method for transformation of signal reconstruction optimization problem.
A more detailed example is given below, in conjunction with attached drawing, the present invention is further described.
Embodiment 6
Method is forced down with embodiment 1-5 based on the SAL data secondary lobes for improving SVA and CS, is of the invention referring to Fig.1
It is a kind of that method flow diagram is forced down based on the SAL data secondary lobes for improving SVA and CS;Based on the SAL data secondary lobes for improving SVA and CS
Force down method, include the following steps:
Step 1, the primary data matrix of synthetic aperture laser radar imaging is generated:Input actual measurement SAL echo complex matrixs,
It generates SAL and is imaged primary data matrix X, to wait for carrying out algorithm process to it to implement subsequent step.
Specifically, the initial SAL image data matrixs X generating modes are as follows:It is multiple from Chinese Academy of Sciences's actual measurement SAL echoes
Interception obtains M × N-dimensional complex matrix corresponding to area-of-interest in matrix number, and longitudinal direction is distance to lateral is orientation, will
The M × N-dimensional complex matrix is denoted as initial SAL image data matrixs X, and M indicates the line number of initial SAL image data matrixs X, N
Indicate the columns of initial SAL image data matrixs X, M and N are respectively positive integer.
It is initial SAL image data matrixs X direct imaging result figures with reference to Fig. 3 (a).
Step 2, SAL data matrixes are handled to application enhancements SVA in distance:SAL primary data matrixes are carried out
Apart from upward improvement SVA processing, is handled with non-integral multiple Nyquist resamplings SVA and obtain M × N with signal sparsity
SAL the first matrixs of consequence of image are tieed up, with suppressed sidelobes, subtract wide main lobe, while signal being thinned out.
Initial SAL image data matrixs X is operated into row distance to zero padding 2a), obtains the M' after initial SAL images zero padding
× N' ties up statistical matrix X', and longitudinal direction is distance to lateral is orientation, and M' indicates that M' × N' after initial SAL images zero padding is tieed up
The line number of statistical matrix X', N' indicate the columns of the dimension statistical matrixs of M' × N' after initial SAL images zero padding X', M'=M+L, N'
=N, 1 < L < M, L are the number for the row progress zero padding that statistical matrix X' is tieed up to M' × N' after initial SAL images zero padding, L, M'
It is respectively positive integer with N'.
To M × N-dimensional data matrix X of initial SAL images into row distance to improvement SVA algorithm process, to each columns
According to real and imaginary parts carry out identical improvement SVA algorithm process respectively and be denoted as the real part of each column data by taking real part as an example
M dimensional signal vectors I, M are positive integer, and carrying out non-integral multiple Nyquist resamplings to signal vector I obtains output signal vector
I0:
I0(n)=I (n)+α (n) * (I (n-1/Ns)+I(n+1/Ns))
Wherein, I (n) indicates that n-th of sample point data of signal vector I, n are the serial numbers of X line numbers, and 1≤n≤M, n are just
Integer, 1/NsFor non-integral multiple Nyquist sample rates, 0 < Ns< 1, takes N in the present embodiments=1/L, α (n) are front and back etc. for n
The weighting function of interval sampling point data item sum, I0(n) be signal vector I in n-th of sample point data I (n) through non-integral multiple
Output valve after Nyquist resamplings.For n 1 to the corresponding output valve in the sections M, overall structure output signal is vectorial.
It is 2b) that signal is made to export side-lobe energy minimum, needs to make signal real and imaginary parts energy while reaching minimum, with reality
For portion, the weighting function α (n) of data item sum should meet calculation formula at equal intervals before and after n:
Wherein, derivative operation is sought in d expressions, | I0(n)|2For signal real part energy.
Obtain the solution expression formula of weighting function α (n):
α (n)=- I (n)/(I (n-1/Ns)+I(n+1/Ns)) (2)
When with n0As signal peak value point, it is contemplated that the expression formula of range value at signal sampling point:I (n)=sinc [Ns(n-
n0)], it substitutes this expression into formula (2), obtains α (n) through deformed solution expression formula:
2c) construction improves SVA algorithm models:
Further to force down secondary lobe, while main lobe being inhibited to broaden, resampling is carried out with non-integral multiple Nyquist sample rates
It improves SVA algorithms the weighting function value condition that signal exports is optimized, the difference of foundation weighting function α (n) value,
Any sample point data I (n) is obtained in signal vector I after non-integral multiple Nyquist resamplings improve SVA algorithm process not
With output valve I0m(n):
With reference to Fig. 2, the graph of relation of Fig. 2 differences between weighting coefficient and signal sampling point and peak point, to make main lobe
Width reduces, and image resolution ratio improves, and enables γminLess than zero, this makes distance between sampled point and peak point be less than 1/Ns;To make
Weak signal target main lobe is effectively recovered from the secondary lobe of multiple strong targets, true SAL scenes is restored, enables γmaxMore than 1/2.At this
In embodiment, γ is enabledmax=-γmin=0.89.
The imaginary part of the N column signal vectors of M × N-dimensional data matrix X of initial SAL images is carried out identical non-integral multiple
Nyquist resamplings improve SVA algorithm process, and generate M × N-dimensional SAL the first matrixs of consequence of image, and M indicates M × N-dimensional SAL figures
As the line number of the first matrix of consequence, N indicates the columns of M × N-dimensional SAL the first matrixs of consequence of image.
With reference to Fig. 3 (b), the design sketch after data processing is carried out to SAL images for the SVA algorithms that improve through the invention.
Step 3, sparse signal needed for CS is constructed:Distance is tieed up to letter to every M of M × N-dimensional SAL the first matrixs of consequence of image
Rarefaction representation is carried out again after number normalized, is carried out the calculating of CS reconstruction signals to be follow-up and is prepared;
3a) for the sake of convenient data processing, by whole M dimensional signals vectors of M × N-dimensional SAL the first matrixs of consequence of image into
The identical normalized of row, one of note whole M dimensional signals M dimensional signal vectors are I', at signal vector I' normalization
Expression formula after reason is:
I "=- orth (I')
Wherein, orth indicates that the normalization operation to vector, I " indicate to tie up letter to the M after signal vector I' normalizeds
Number vector, | I " |=1;
3b) signal indicates rarefaction, and it is that M × M ties up unit matrix that setting M × M, which ties up sparse basis array ψ, the reason is that, this reality
Apply initial SAL image data matrixs X in example through distance to improvement SVA algorithm process after gained M × N-dimensional SAL the first knots of image
Any M dimensional signals vector of fruit matrix is sparse signal vector, and M dimensional signal vector I " rarefactions are expressed as:
I "=ψ * s
Wherein, ψ indicates that the sparse basis array of M dimensional signal vectors I ", s indicate sparse coefficients of the I " on ψ, for M × 1 tie up to
Amount, in the present embodiment, s=I ".
Step 4, basic matrix is observed needed for construction CS:Independent identically distributed gaussian random matrix is set as M × N-dimensional SAL
The observation basic matrix of every M dimensional signals in the first matrix of consequence of image later carries out every M dimensional signals using observation basic matrix
Compression processing obtains the every M of SAL images and ties up compression result signal, and pickup reduces number for the useful information needed for SAL images
It is also to prepare for follow-up progress CS signal reconstructions calculating here according to amount of storage.
4a) determine the reasonable dimension of the observation basic matrix;
Specifically, note observation basic matrix is the line number that A × B ties up matrix φ, A expression φ, and B indicates the columns of φ, B Ying Yuxin
The dimension M of number vector I " is equal, and A should meet calculation expression:
A=fix (K*ln (M/K) * β)
Wherein, K indicates that the number of the nonzero element contained by M dimensional signal vectors I ", ln expressions seek natural logrithm, fix
Indicate truncation floor operation, β is observed differential, and the value of β should make K < A < < M that will observe base for M dimensional signal vector I "
It is the end value substituted by the K mean values of N number of M dimensional signals vector I " obtained by above formula that the line number A of matrix φ, which fixes value,;
In the present embodiment, β=1.43, therefore the calculation expression can also be:
A=fix (K*ln (M/K) * 1.43)
4b) the observation basic matrix of design M dimensional signal vectors I ":
The design method of the observation basic matrix of M dimensional signal vectors I " is have away from property for guarantee is equivalent, should make observation group moment
Battle array φ is uncorrelated to sparse basis array ψ, and keeps the matrix being made of arbitrary A column vector in observation basic matrix φ nonsingular, this
Setting observation basic matrix is that A × B ties up independent identically distributed gaussian random matrix, the observation base of M dimensional signal vectors I " in embodiment
Matrix expression is:
φ=randn (A, B)+i*randn (A, B)
Wherein, randn (A, B) expressions are sought in mean value being 0, and variance is that A × B of 1 normal distribution ties up random matrix, i tables
Show imaginary unit.
4c) compression processing M dimensional signals vector I ":
Compression processing is carried out to M dimensional signal vector I " signals using observation basic matrix, compression processing expression formula is:
Y=φ * I "
Wherein, Y indicates compressed treated the M dimension compression result vectors of M dimensional signal vectors I ".
Step 5, compressed sensing underdetermined equation is solved:A M every to SAL images ties up compression result in a manner of linear programming for solution
The CS underdetermined equations of signal are solved, and SAL image original signals are reconstructed, obtain SAL images every M dimension compressed sensing estimate to
Amount, compressed sensing technology need based on sparse signal, and need one observation basic matrix of construction, therefore the present invention is in step
3 and step 4 be for solve CS underdetermined equations, reconstruct SAL image original signals be used as place mat.
5a) compressed sensing underdetermined equation mathematic(al) representation is:
Y=φ * x
Wherein, x indicates that sparse signal, φ indicate that observation basic matrix, y indicate compression result vector;
The theoretically reconstruct of original signal, by l0Signal reconstruction optimization problem is solved under norm, reconstruct expression formula is:
Wherein, α indicates that sparse coefficients of the x on ψ, x=ψ * α, the meaning for reconstructing expression formula are, y=φ * ψ * α about
Under the conditions of beam, seek making the optimum estimation value α ' when nonzero element number minimum in vectorial α.
5b) reconstruct SAL image original signals:
In the present embodiment, x=I ", α=s, y=Y are enabled, by M dimensional signal vectors I ", I " on its sparse basis array ψ
Sparse coefficient s, M ties up compression result vector Y and substitutes into compressed sensing underdetermined equation and reconstruct expression formula, the reconstruct of SAL image original signals
Process is:
Implement convex optimized algorithm, the meaning of the convex optimized algorithm is, by the l0Norm minimum require relax for
With its l with equivalence1Norm minimum requirement, thus converts the optimization problem of theoretically signal reconstruction to linear gauge
Solve problems are drawn, expression formula is:
Thus to obtain the estimation s' of sparse coefficient s, the compressed sensing estimated value expression formula of original signal I " is:
I " '=ψ * s'
Wherein, ψ indicates that sparse basis array, I " ' indicate the compressed sensing estimated value of original signal I ", in the present embodiment, I " '
=s', I " ' are M dimensional vectors.
Step 6, SAL image result matrixes, imaging are generated:M × N is generated by every M dimension compressed sensing estimate vectors
SAL the second matrixs of consequence of image are tieed up, obtain forcing down result based on the SAL data secondary lobes for improving SVA and CS after carrying out imaging
Image.
With reference to 3 (c), on the basis of step 2 gained Synthetic Aperture Laser Radar result figure, to continue with CS algorithms pair
SAL images carry out the design sketch after data processing, this is also all to carry out being based on changing to initial SAL images according to flow of the present invention
Into the data process effects figure of SVA and CS.
The present invention mentality of designing be:Firstly generate initial SAL image data matrixs, with wait for carrying out it algorithm process with
Implement subsequent step;SVA processing is improved to data to synthetic aperture laser radar imaging matrix distance again, to inhibit other
Valve subtracts wide main lobe, to which signal to be thinned out, obtains the first matrix of consequence of M × N-dimensional image;So far it realizes using SVA to original
Image data forces down secondary lobe and LS-SVM sparseness;Later to any M dimensional signals of first matrix of consequence of M × N-dimensional image
Vector carries out rarefaction representation;Next the observation of any M dimensional signals vector of first matrix of consequence of M × N-dimensional image is constructed
Basic matrix, with compressed signal, pickup useful information, to reduce data storage capacity;Then compressed sensing underdetermined equation, weight are solved
Structure original signal;So far it realizes and utilizes recoveries of the CS to sparse SAL images;It finally presses each row compressed sensing estimated value and generates M × N
The second matrix of consequence of image is tieed up, obtaining Synthetic Aperture Laser Radar data secondary lobe after progress imaging forces down result images;Extremely
This completes to force down secondary lobe processing to SAL image datas, obtains high-definition picture.
In brief, disclosed by the invention a kind of based on room for improvement apodization (SVA) and compressed sensing Reconstruction Method (CS)
Synthetic Aperture Laser Radar (SAL) data secondary lobe force down method, it is higher that the present invention solves SAL image data secondary lobes, at
Image quality amount is bad, the problem of image resolution ratio deficiency.Realize that step includes:Generate the initial number of synthetic aperture laser radar imaging
According to matrix;Construction improves SVA algorithm models, is handled to using the model SAL data matrixes in distance;It constructs needed for CS
Sparse signal and observation basic matrix;Solve compressed sensing underdetermined equation;Generate SAL image result matrixes, imaging;Completion pair
SAL image datas force down secondary lobe processing, obtain high-definition picture.The present invention will improve SVA algorithms and CS is combined, can
Under the premise of keeping main lobe energy and image resolution ratio, suppressed sidelobes subtracts wide main lobe, reduces the fortune of SAL image real time transfers
Calculation amount and amount of storage more efficiently force down Synthetic Aperture Laser Radar data secondary lobe.For synthetic aperture laser radar imaging
Middle reduction echo noise promotes SAL image resolution ratios and picture quality.
Obviously, various changes and modifications can be made to the invention without departing from essence of the invention by those skilled in the art
God and range;In this way, if these modifications and changes of the present invention belongs to the range of the claims in the present invention and its equivalent technologies
Within, then the present invention is also intended to include these modifications and variations.
Claims (5)
1. a kind of forcing down method based on the SAL data secondary lobes for improving SVA and CS, which is characterized in that include following steps:
Step 1, the primary data matrix of synthetic aperture laser radar imaging is generated:Input actual measurement SAL echo complex datas, generate
M × N-dimensional SAL imaging primary data matrixes X;
Step 2, construction improves SVA algorithm models, is handled to using the model SAL data matrixes in distance:To at the beginning of SAL
For beginning data matrix into the upward improvement SVA processing of row distance, handling to obtain with non-integral multiple Nyquist resamplings SVA has letter
M × N-dimensional SAL the first matrixs of consequence of image of number sparsity;
Step 3, sparse signal needed for CS is constructed:Every M dimension distances of M × N-dimensional SAL the first matrixs of consequence of image are returned to signal
Rarefaction representation is carried out again after one change processing;
Step 4, basic matrix is observed needed for construction CS:Independent identically distributed gaussian random matrix is set as M × N-dimensional SAL images
The observation basic matrix of every M dimensional signals in first matrix of consequence later compresses every M dimensional signals using observation basic matrix
Processing obtains the every M of SAL images and ties up compression result signal;
Step 5, compressed sensing underdetermined equation is solved:A M every to SAL images ties up compression result signal in a manner of linear programming for solution
CS underdetermined equations solved, obtain the every M of SAL images and tie up compressed sensing estimate vector, reconstruct SAL image original signals;
Step 6, SAL image result matrixes, imaging are generated:M × N-dimensional SAL is generated by every M dimension compressed sensing estimate vectors
The second matrix of consequence of image obtains forcing down result images based on the SAL data secondary lobes for improving SVA and CS after carrying out imaging.
2. a kind of SAL data secondary lobes based on improvement SVA and CS as described in claim 1 force down method, which is characterized in that
The process of acquisition M × N-dimensional SAL the first matrixs of consequence of image described in step 2 includes:
2a) to M × N-dimensional data matrix X of initial SAL images into row distance to improvement SVA algorithm process, to each column data
Real and imaginary parts carry out identical improvement SVA algorithm process respectively, by taking real part as an example, the real part of each column data is denoted as M
Dimensional signal vector I, M are positive integer, and carrying out non-integral multiple Nyquist resamplings to signal vector I obtains output signal vector I0:
I0(n)=I (n)+α (n) * (I (n-1/Ns)+I(n+1/Ns))
Wherein, I (n) indicates n-th of sample point data of signal vector I, and n is the serial number of X line numbers, and 1≤n≤M, n are positive integer,
1/NsFor non-integral multiple Nyquist sample rates, 0 < Ns< 1, α (n) are the weighting function of data item sum at equal intervals before and after n, I0(n)
For output valve of n-th of the sample point data I (n) in signal vector I after non-integral multiple Nyquist resamplings;
2b) calculate weighting function:
The solution expression formula of weighting function α (n) is:
α (n)=- I (n)/(I (n-1/Ns)+I(n+1/Ns))
2c) construction improves SVA algorithm models:
According to the difference of weighting function α (n) value, any sample point data I (n) is obtained in signal vector I through non-integral multiple
Nyquist resamplings improve the different output valve I after SVA algorithm process0m(n):
Wherein, γminFor minus real number, γmaxFor the real number more than 1/2;
Same treatment is carried out to the imaginary part of the N column signal vectors of M × N-dimensional data matrix X of initial SAL images, and generates M × N
SAL the first matrixs of consequence of image are tieed up, M indicates that the line number of M × N-dimensional SAL the first matrixs of consequence of image, N indicate M × N-dimensional SAL images
The columns of first matrix of consequence.
3. a kind of SAL data secondary lobes based on improvement SVA and CS as described in claim 1 force down method, which is characterized in that
The process of sparse signal needed for the construction CS includes in step 3:
Whole M dimensional signals vectors of M × N-dimensional SAL the first matrixs of consequence of image 3a) are subjected to identical normalized, note is complete
One of portion's M dimensional signals M dimensional signal vectors are I', and the expression formula to signal vector I' normalizeds is:
I "=- orth (I')
Wherein, orth indicate to vector normalization operation, I " indicate to the M dimensional signals after signal vector I' normalizeds to
Amount, | I " |=1;
3b) signal indicates rarefaction, and M dimensional signal vectors I " is expressed as:
I "=ψ * s
Wherein, ψ indicates that the sparse basis array of M dimensional signal vectors I ", s indicate sparse coefficients of the I " on ψ, is the dimensional vectors of M × 1.
4. a kind of SAL data secondary lobes based on improvement SVA and CS as described in claim 1 force down method, which is characterized in that
In step 4 group moment is observed needed for the M dimensional signals vector I " every to M × N-dimensional SAL the first matrixs of consequence of image constructions CS
The processing procedure of battle array is identical, includes:
4a) determine the reasonable dimension of observation basic matrix:
Note observation basic matrix is the line number that A × B ties up matrix φ, A expression φ, and B indicates the columns of φ, and B should be with signal vector I's "
Dimension M is equal, and A should meet calculation expression:
A=fix (K*ln (M/K) * β)
Wherein, K indicates that the number of the nonzero element contained by M dimensional signal vectors I ", ln expressions seek natural logrithm, and fix is indicated
Truncation floor operation, β are observed differential.
4b) the observation basic matrix of design M dimensional signal vectors I ":
The design of the observation basic matrix of M dimensional signal vectors I ", should make observation basic matrix φ uncorrelated to sparse basis array ψ, and make
The matrix being made of arbitrary A column vector in observation basic matrix φ is nonsingular, and it is independent same point of A × B dimensions that observation basic matrix, which is arranged,
The observation basic matrix expression formula of the gaussian random matrix of cloth, M dimensional signal vectors I " is:
φ=randn (A, B)+i*randn (A, B)
Wherein, randn (A, B) expressions are sought in mean value being 0, and variance is that A × B of 1 normal distribution ties up random matrix, and i indicates empty
Number unit;
4c) compression processing M dimensional signals vector I ":
Compression processing is carried out to M dimensional signal vectors I " using observation basic matrix, compression processing expression formula is:
Y=φ * I "
Wherein, Y indicates compressed treated the M dimension compression result vectors of M dimensional signal vectors I ".
5. a kind of SAL data secondary lobes based on improvement SVA and CS as described in claim 1 force down method, which is characterized in that
The process of the solution compressed sensing underdetermined equation includes in steps of 5:
5a) compressed sensing underdetermined equation mathematic(al) representation is:
Y=φ * x
Wherein, x indicates that sparse signal, φ indicate that observation basic matrix, y indicate compression result vector;
The theoretically reconstruct of original signal, by l0Signal reconstruction optimization problem is solved under norm, reconstruct expression formula is:
Wherein, α indicates that sparse coefficients of the x on ψ, x=ψ * α, the meaning for reconstructing expression formula are, item is constrained in y=φ * ψ * α
Under part, seek making the optimum estimation value α ' when nonzero element number minimum in vectorial α;
5b) reconstruct SAL image original signals:
X=I ", α=s, y=Y are enabled, sparse coefficient s, M of M dimensional signal vectors I ", I " on its sparse basis array ψ is tieed up into compression
Result vector Y substitutes into compressed sensing underdetermined equation and reconstruct expression formula;
Convert the optimization problem of theoretically signal reconstruction to linear programming for solution problem, expression formula is:
Thus to obtain the estimation s' of sparse coefficient s;
The compressed sensing estimated value expression formula of any M dimensional signals vector I " of M × N-dimensional SAL the first matrixs of consequence of image is:
I " '=ψ * s'
Wherein, ψ indicates sparse basis array, and I " ' indicates the compressed sensing estimated value of original signal I ", I " '=s', I " ' be M tie up to
Amount.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711214945.3A CN108318891B (en) | 2017-11-28 | 2017-11-28 | SAL data side lobe depression method based on improved SVA and CS |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201711214945.3A CN108318891B (en) | 2017-11-28 | 2017-11-28 | SAL data side lobe depression method based on improved SVA and CS |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108318891A true CN108318891A (en) | 2018-07-24 |
CN108318891B CN108318891B (en) | 2021-09-10 |
Family
ID=62892468
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201711214945.3A Active CN108318891B (en) | 2017-11-28 | 2017-11-28 | SAL data side lobe depression method based on improved SVA and CS |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108318891B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109347482A (en) * | 2018-08-03 | 2019-02-15 | 西安电子科技大学 | Frequency Hopping Signal compressed sensing reconstructing method based on parameter Estimation |
CN110068833A (en) * | 2019-05-05 | 2019-07-30 | 中国科学院电子学研究所 | A kind of synthetic aperture laser radar imaging method, instrument and system |
CN110333489A (en) * | 2019-07-24 | 2019-10-15 | 北京航空航天大学 | The processing method to SAR echo data Sidelobe Suppression is combined with RSVA using CNN |
CN110632616A (en) * | 2019-11-08 | 2019-12-31 | 北京环境特性研究所 | Micro-motion imaging method of airborne inverse synthetic aperture laser radar under sparse sampling |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110254728A1 (en) * | 2010-04-20 | 2011-10-20 | Peterson Timothy T | Three quarter spatially variant apodization |
CN102721965A (en) * | 2012-06-08 | 2012-10-10 | 北京航空航天大学 | Omni-directional side lobe suppression method for synthetic aperture radar imaging processing |
CN106680785A (en) * | 2017-03-06 | 2017-05-17 | 浙江工业大学 | Method for suppressing SAR image sidelobe based on wavelet space apodization |
CN106772276A (en) * | 2016-12-21 | 2017-05-31 | 南京信息工程大学 | A kind of geostationary orbit circular track SAR horizontal planes two dimension side lobe suppression method |
-
2017
- 2017-11-28 CN CN201711214945.3A patent/CN108318891B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20110254728A1 (en) * | 2010-04-20 | 2011-10-20 | Peterson Timothy T | Three quarter spatially variant apodization |
CN102721965A (en) * | 2012-06-08 | 2012-10-10 | 北京航空航天大学 | Omni-directional side lobe suppression method for synthetic aperture radar imaging processing |
CN106772276A (en) * | 2016-12-21 | 2017-05-31 | 南京信息工程大学 | A kind of geostationary orbit circular track SAR horizontal planes two dimension side lobe suppression method |
CN106680785A (en) * | 2017-03-06 | 2017-05-17 | 浙江工业大学 | Method for suppressing SAR image sidelobe based on wavelet space apodization |
Non-Patent Citations (2)
Title |
---|
KEMKEMIAN, S ET AL.: "On the Challenges of Incomplet Sampling in SAR Imagery", 《2014 15TH INTERNATIONAL RADAR SYMPOSIUM ( IRS) INTERNATIONAL RADAR SYMPOSIUM PROCEEDINGS》 * |
倪崇 等: "一种基于改进SVA的SAR旁瓣抑制算法", 《中国科学:技术科学》 * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109347482A (en) * | 2018-08-03 | 2019-02-15 | 西安电子科技大学 | Frequency Hopping Signal compressed sensing reconstructing method based on parameter Estimation |
CN109347482B (en) * | 2018-08-03 | 2021-04-06 | 西安电子科技大学 | Frequency hopping signal compressed sensing reconstruction method based on parameter estimation |
CN110068833A (en) * | 2019-05-05 | 2019-07-30 | 中国科学院电子学研究所 | A kind of synthetic aperture laser radar imaging method, instrument and system |
CN110068833B (en) * | 2019-05-05 | 2021-10-29 | 中国科学院电子学研究所 | Synthetic aperture laser radar imaging method, instrument and system |
CN110333489A (en) * | 2019-07-24 | 2019-10-15 | 北京航空航天大学 | The processing method to SAR echo data Sidelobe Suppression is combined with RSVA using CNN |
CN110632616A (en) * | 2019-11-08 | 2019-12-31 | 北京环境特性研究所 | Micro-motion imaging method of airborne inverse synthetic aperture laser radar under sparse sampling |
CN110632616B (en) * | 2019-11-08 | 2021-07-27 | 北京环境特性研究所 | Micro-motion imaging method of airborne inverse synthetic aperture laser radar under sparse sampling |
Also Published As
Publication number | Publication date |
---|---|
CN108318891B (en) | 2021-09-10 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108318891A (en) | It is a kind of that method is forced down based on the SAL data secondary lobes for improving SVA and CS | |
US7279895B2 (en) | Fast generalized autocalibrating partially parallel acquisition image reconstruction algorithm for magnetic resonance imaging | |
Lorintiu et al. | Compressed sensing reconstruction of 3D ultrasound data using dictionary learning and line-wise subsampling | |
Dong et al. | A novel compressive sensing algorithm for SAR imaging | |
US8587307B2 (en) | Systems and methods for accelerating the acquisition and reconstruction of magnetic resonance images with randomly undersampled and uniformly undersampled data | |
US11030780B2 (en) | Ultrasound speckle reduction and image reconstruction using deep learning techniques | |
US20190266761A1 (en) | System and method for sparse image reconstruction | |
CN109615675B (en) | Image reconstruction method for multi-channel magnetic resonance imaging | |
CN106802418A (en) | A kind of method for designing of the high-effect sparse dictionary in synthetic aperture compressed sensing ultrasonic imaging | |
US8026720B1 (en) | Rapid auto-calibrated parallel reconstruction using synthetic target coil | |
CN111127320B (en) | Deep learning-based photoacoustic image super-resolution reconstruction method and device | |
WO2016155239A1 (en) | Method and device for ultrasonic imaging by synthetic focusing | |
RU2626184C2 (en) | Method, device and system for reconstructing magnetic resonance image | |
CN104076360B (en) | The sparse target imaging method of two-dimensional SAR based on compressed sensing | |
JP2022505511A (en) | Adaptive beamforming methods and systems for ultrasonic signals | |
Schretter et al. | Ultrasound imaging from sparse RF samples using system point spread functions | |
CN110880196A (en) | Tumor photoacoustic image rapid reconstruction method and device based on deep learning | |
DE102017205566A1 (en) | Diffraction Source Compensation in Medical-Diagnostic Viscoelastic Ultrasound Imaging | |
Ellis et al. | Super-resolution image reconstruction using diffuse source models | |
CN112617798B (en) | Parallel magnetic resonance imaging reconstruction method based on Lp norm combined total variation | |
CN109766646B (en) | Ultrasonic imaging method and device based on sparse channel echo data reconstruction | |
Wang et al. | Plane-wave ultrasound imaging based on compressive sensing with low memory occupation | |
CN108931770B (en) | ISAR imaging method based on multi-dimensional beta process linear regression | |
CN108618799B (en) | Ultrasonic CT imaging method based on spatial coherence | |
CN114219843B (en) | Method and system for constructing terahertz spectrum image reconstruction model and application |
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 |