CN110110398A - A kind of super surface automatic design method based on convolution self-encoding encoder - Google Patents
A kind of super surface automatic design method based on convolution self-encoding encoder Download PDFInfo
- Publication number
- CN110110398A CN110110398A CN201910315408.0A CN201910315408A CN110110398A CN 110110398 A CN110110398 A CN 110110398A CN 201910315408 A CN201910315408 A CN 201910315408A CN 110110398 A CN110110398 A CN 110110398A
- Authority
- CN
- China
- Prior art keywords
- super surface
- encoding encoder
- convolution self
- convolution
- method based
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
- G06F17/153—Multidimensional correlation or convolution
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Mathematical Physics (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Optimization (AREA)
- Pure & Applied Mathematics (AREA)
- Mathematical Analysis (AREA)
- Data Mining & Analysis (AREA)
- Computational Mathematics (AREA)
- Algebra (AREA)
- Software Systems (AREA)
- Databases & Information Systems (AREA)
- Computing Systems (AREA)
- Computer Hardware Design (AREA)
- Evolutionary Computation (AREA)
- Geometry (AREA)
- Radar Systems Or Details Thereof (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a kind of super surface automatic design method based on convolution self-encoding encoder, comprising the following steps: step 1, generate several super surface cell structures at random, the reflected phase and amplitude on each super surface are calculated separately using electromagnetic simulation software；Step 2, using the deep learning method based on convolution self-encoding encoder, by by the reflected phase being calculated in step 1 and amplitude while inputting convolution self-encoding encoder, corresponding super surface cell structure, Lai Xunlian deep learning model are exported；Step 3, it will be trained in the reflected phase of design object and amplitude input step 2 in the deep learning model finished, convolution self-encoding encoder used to complete the matching between feature extraction and feature and super surface matrix, that is, the super surface texture designed required for obtaining.The present invention can accomplish the reflected amplitudes and reflected phase that design super surface simultaneously, sweep ginseng process without complicated, can automatically generate surface texture, simplify design procedure, and design efficiency is high.
Description
Technical field
The invention belongs to super surface design fields, are related to a kind of super surface Automated Design based on convolution self-encoding encoder
Method.
Background technique
Super surface refers to that a kind of thickness is less than the artificial stratified material of wavelength.By the design of SPA sudden phase anomalies spatial distribution,
Super surface can realize the flexible Effective Regulation to propagation characteristics such as polarization of electromagnetic wave, amplitude, phase, polarization mode, communication modes.
Super surface can be considered the two-dimensional surface Asia wave structure array of Meta Materials.
The super surface of the phase gradient being widely used at present, the super surface of coding etc., the catadioptric width for needing to design simultaneously
Degree and phase.However realize that the method for amplitude and phase regulation mostly can only be in amplitude and phase currently based on super surface
Single parameter is regulated and controled, and is regulated and controled while being difficult to realize to amplitude and phase.Designer is difficult to design the knot on super surface simultaneously
Structure makes amplitude and phase reach the sizes of needs.Meanwhile in the prior art, it designs super surface and not only expends the time, but also expend
Computing resource, need to carry out complexity sweeps ginseng process, carries out modeling and parameter optimization, design procedure is many and diverse, and engineer is very
Hardly possible directly gives corresponding super surface by design object, and design efficiency is low.
Summary of the invention
The object of the present invention is to provide a kind of super surface automatic design method based on convolution self-encoding encoder, solves existing
It is difficult to realize present in technology to super surface amplitude and phase while being regulated and controled, and the problem that design efficiency is low.
The technical scheme adopted by the invention is that a kind of super surface automatic design method based on convolution self-encoding encoder, packet
Include following steps:
Step 1, several super surface cell structures are generated at random, calculate separately each super surface using electromagnetic simulation software
Reflected phase and amplitude；
Step 2, using the deep learning method based on convolution self-encoding encoder, pass through the reflection that will be calculated in step 1
Phase and amplitude inputs convolution self-encoding encoder simultaneously, exports corresponding super surface cell structure, Lai Xunlian deep learning model；
Step 3, it will be trained in the reflected phase of design object and amplitude input step 2 in the deep learning model finished,
The matching between feature extraction and feature and super surface matrix is completed using convolution self-encoding encoder, that is, is designed required for obtaining
Super surface texture.
The features of the present invention also characterized in that:
Electromagnetic simulation software in step 1 is CST STUDIO SUITE MWS.
Detailed process is as follows for step 1:
The matrix of the super surface cell structure generated at random using the output of MATLAB software first, wherein labeled as " 1 "
Region indicates that the area filling has metal, indicates the region blank labeled as the region of " 0 ", reuses electromagnetic simulation software CST
STUDIO SUITE MWS carries out reflected phase and amplitude that super surface cell is calculated.
Detailed process is as follows for training deep learning model in step 2:
Step 2.1, the reflection phase of super surface cell data acquisition generation and preprocessing process: is drawn using MATLAB language
The image of position and amplitude, gray scale and pixel characteristic to the image are normalized operation, set image as 32 × 32 × 1 list
Channel grayscale image limits the pixel of image between 0~1, and using image of this pixel between 0~1 as deep learning
The input of model；
Step 2.2, using convolution self-encoding encoder extract feature process: by the pixel obtained in step 2.1 0~1 it
Between image input convolution self-encoding encoder, pass through convolution several times, pondization operation and the accordingly up-sampling of number, deconvolution first
The reflected phase of input and amplitude curve compression of images are reconfigured in a representation space by operation, then further according to this expression
The output data matrix that space extracts to the end to data；Then under the coded portion of convolution self-encoding encoder being disassembled
Come, input image data is encoded, original input image is compressed into expression vector, is rolled up by I iteration adjustment of activation primitive
Product core obtains expression vector when loss function I minimizes until the minimum of loss function I；
Step 2.3, the super surface texture matrix matching process based on convolution self-encoding encoder: pass through activation primitive II and loss
Function II is trained using the artificial neural network that connects entirely, the expression that the coded portion of input convolution self-encoding encoder generates to
Moment matrix exports super surface texture matrix, until it is minimum to complete loss function I in step 2.2 when loss function II minimizes
The matching indicated between vector characteristics and super surface texture when change, deep learning model training finish.
Activation primitive I isWherein x represents the output number after each layer of linear weighted function of neural network
According to；Loss function I is cross entropy cost functionWherein x represents output data
Number, y are desired output, and y' is actual output.
Activation primitive II uses ReLU=max (0, x), after wherein x represents each layer line weighted sum of neural network
Output data；Loss function II is mean square errorWherein yi is desired output, y 'iFor reality output.
The beneficial effects of the present invention are: solve it is existing in the prior art be difficult to realize it is same to super surface amplitude and phase
Shi Jinhang regulation, and the problem that design efficiency is low, are tied super surface by the deep learning model based on convolution self-encoding encoder
Structure matrix is matched with its electromagnetic property, can be realized simultaneously the design to super surface reflection phase and reflected amplitudes；And
Unsupervised learning may be implemented using the method for convolution self-encoding encoder, this method copys the vision noticing mechanism building of biology, knot
The operation such as convolution sum pond of convolutional neural networks is closed, ginseng process is swept without complicated, without modeling and parameter optimization, to super table
The design in face may be implemented to automate, and simplify design procedure, and design efficiency is high.
Detailed description of the invention
Fig. 1 is the unit on super surface described in a kind of super surface automatic design method based on convolution self-encoding encoder of the present invention
Structural schematic diagram；
Fig. 2 is the deep learning model training flow chart in the present invention based on convolution self-encoding encoder；
Fig. 3 is the super surface structure design flow chart in the present invention based on convolution self-encoding encoder；
Fig. 4 is convolution self-encoding encoder described in a kind of super surface automatic design method based on convolution self-encoding encoder of the present invention
Structural schematic diagram；
Fig. 5 (a) is the reflected amplitudes proposed in the embodiment of the present invention and the design object of phase；Fig. 5 (b) is for Fig. 5
(a) value of super surface reflection amplitude and phase designed by the design object in.
In figure, 1. dielectric layers, 2. structure sheafs, 3. metal backings.
Specific embodiment
The following describes the present invention in detail with reference to the accompanying drawings and specific embodiments.
A kind of super surface automatic design method based on convolution self-encoding encoder of the present invention, as shown in Figure 1,2 and 3, including with
Lower step:
Step 1, several super surface cell structures are generated at random, calculate separately each super surface using electromagnetic simulation software
Reflected phase and amplitude；
Step 2, using the deep learning method based on convolution self-encoding encoder, pass through the reflection that will be calculated in step 1
Phase and amplitude inputs convolution self-encoding encoder simultaneously, exports corresponding super surface cell structure, Lai Xunlian deep learning model；
Step 3, it will be trained in the reflected phase of design object and amplitude input step 2 in the deep learning model finished,
The matching between feature extraction and feature and super surface matrix is completed using convolution self-encoding encoder, that is, is designed required for obtaining
Super surface texture.
Electromagnetic simulation software in step 1 is CST STUDIO SUITE MWS.
Detailed process is as follows for step 1:
The matrix of the super surface cell structure generated at random using the output of MATLAB software first, wherein labeled as " 1 "
Region indicates that the area filling has metal, indicates the region blank labeled as the region of " 0 ", reuses electromagnetic simulation software CST
STUDIO SUITE MWS carries out reflected phase and amplitude that super surface cell is calculated.
Detailed process is as follows for training deep learning model in step 2:
Step 2.1, the reflection phase of super surface cell data acquisition generation and preprocessing process: is drawn using MATLAB language
The image of position and amplitude, and sets the gray scale and pixel of the image, sets image as 32 × 32 × 1 single channel grayscale image, by
In grayscale image pixel coverage between 0~255, to keep feature distribution more uniform, operation is normalized to feature, will be schemed
The pixel of picture is limited between 0~1, and using image of this pixel between 0~1 as the input of deep learning model；
Step 2.2, using convolution self-encoding encoder extract feature process: by the pixel obtained in step 2.1 0~1 it
Between image input convolution self-encoding encoder, pass through convolution several times, pondization operation and the accordingly up-sampling of number, deconvolution first
The reflected phase of input and amplitude curve compression of images are reconfigured in a representation space by operation, then further according to this expression
The output data matrix that space extracts to the end to data；Then under the coded portion of convolution self-encoding encoder being disassembled
Come, input image data is encoded, original input image is compressed into expression vector, is rolled up by I iteration adjustment of activation primitive
Product core obtains expression vector when loss function I minimizes until the minimum of loss function I；
Step 2.3, the super surface texture matrix matching process based on convolution self-encoding encoder: pass through activation primitive II and loss
Function II is trained using the artificial neural network that connects entirely, the expression that the coded portion of input convolution self-encoding encoder generates to
Moment matrix exports super surface texture matrix, until it is minimum to complete loss function I in step 2.2 when loss function II minimizes
The matching indicated between vector characteristics and super surface texture when change, deep learning model training finish.
Activation primitive I isWherein x represents the output number after each layer of linear weighted function of neural network
According to；Loss function I is cross entropy cost functionWherein x represents output data
Number, y are desired output, and y' is actual output.
Activation primitive II uses ReLU=max (0, x), after wherein x represents each layer line weighted sum of neural network
Output data；Loss function II is mean square errorWherein yi is desired output, y 'iFor reality output.
Embodiment
As shown in Figure 1, the super surface cell structure in step 1 is by dielectric layer 1, structure sheaf 2 and 3 groups of metal backing layer
At the lower surface of dielectric layer 1 is metal backing layer 3, and the upper surface of dielectric layer 1 is structure sheaf 2, super surface cell in the present embodiment
The material of the dielectric layer 1 of structure is glass-epoxy copper-clad plate (Fr4), dielectric layer 1 with a thickness of 1.5mm, metal back
Plate layer 3 with a thickness of 0.018mm, be copper in the material of microwave frequency band flowering structure layer 2 and metal backing layer 3；Structure sheaf 2 divides
For 64 with " 0 " or the region of one token, the region for being labeled as " 1 " indicates that the area filling has metal, is labeled as the area of " 0 "
Domain representation region blank, therefore structure sheaf 2 can be indicated with matrix, 1 material of dielectric layer and with a thickness of fixed value in the case where
Super surface cell structure can also be indicated with matrix.
As shown in Fig. 2,3 and 4, the 2000 groups of super surface cell structures generated at random are exported using MATLAB software first
8*8 matrix reuses electromagnetic simulation software CST STUDIO SUITE MWS and carries out that 2000 groups of super surface cells are calculated
Reflected phase and amplitude.
Since the two dimensional image of input is a series of two-dimensional arrays for computer angle, intuitively it is difficult therefrom to extract
Suitable feature is matched with output matrix.Therefore, feature extraction operation is carried out to input picture first.It is self-editing using convolution
The unsupervised learning to image may be implemented in the method for code device, and this method copys the vision noticing mechanism building of biology, in conjunction with volume
The convolution sum pondization operation of product neural network, to realize feature extraction.
As shown in figure 3, the single channel input picture for being 32 × 32 × 1 for size, carries out convolution operation and pond to it
Operation, the size and depth difference of every layer of convolution kernel, but stride is 1, and zero padding mode is equivalent filling (same
Padding), pondization operation uses maximum pond function (MaxPooling), the basic parameter at model based coding end in Fig. 3 are as follows:
Coded portion:
First layer convolution: 7 × 7 convolution kernel 32, step-length 1, filling mode are equivalent filling (same padding)；
First layer pond: pond filter size is 4 × 4, maximum pond function；
Second layer convolution: 5 × 5 convolution kernel 64, remaining parameter is consistent with first layer convolution；
Second layer pond: pond filter size is 4 × 4, maximum pond function；
Third layer convolution: 3 × 3 convolution kernel 128, remaining parameter is consistent with first layer convolution；
Third layer pond: pond filter size is 2 × 3, maximum pond function.
After the completion of the above operation, i.e., matrix compression that size is 32 × 32 × 1 will be originally inputted into 1 × 1 × 128 square
Battle array, then data are reconstructed by decoded portion, the basic parameter of Fig. 3 model decoding end are as follows:
First layer up-sampling: sampling core size is 2 × 2；
First layer deconvolution: 3 × 3 convolution kernel 64, step-length 1, filling mode are equivalent filling (same
padding)；
Second layer up-sampling: sampling core size is 4 × 4；
Second layer deconvolution: 5 × 5 convolution kernel 32, step-length 1, filling mode are equivalent filling (same
padding)；
Third layer up-sampling: sampling core size is 4 × 4；
Second layer deconvolution: 7 × 7 convolution kernel 1, step-length 1, filling mode are equivalent filling (same
padding)。
After the completion of the above decoding operate, i.e., by coding side compression 1 × 1 × 128 matrix reconstructions at 32 × 32 × 1 square
Battle array.
The implementation procedure of the convolution self-encoding encoder as seeks the process of cross entropy cost function minimum value.According to given
Input picture, by the continuous iteration adjustment convolution kernel of activation primitive I, so that loss function I minimizes, the image and original of reconstruct are defeated
It is most like to enter image.After the completion of the training of convolution self-encoding encoder, its coded portion is dismantled, takes out the 1 of coded portion compression
× 1 × 128 matrixes as it is after feature extraction as a result, i.e. characteristic dimension be compressed to 1 by 32 × 32 × 1=1024 dimension of script ×
1 × 128=128 dimension, characteristic dimension become original 12.5%, and little to the influential effect of image.
Super surface texture matrix matching process based on convolution self-encoding encoder be for the expression vector characteristics that newly extract with
Matching between super surface texture matrix, is trained using the artificial neural network connected entirely, input convolution self-encoding encoder
The expression vector matrix that coding side generates exports super surface texture matrix.
The matching between 128 dimensional features and super surface texture matrix that newly extract, the net are realized in the embodiment of the present invention
The input of network is 128 newly-generated dimensional feature matrixes, and non-linear factor is added in neural network using activation primitive II, is lost
The super surface texture matrix that output size is 8 × 8 when function II minimizes.The model may be defined as:
Y=f (X β)
Wherein, X is input vector, and β is parameter vector, and Y is output vector, and f (*) is activation primitive.
The learning process of the neural network is to seek the process of mean square error minimum value, is asked according to the setting of initial β parameter
Obtain an output y 'i, calculate y 'iWith desired output yiBetween mean square error value, mean square error is determined using gradient descent algorithm
The minimum value of difference, and pass through the setting that Back Propagation Algorithm (BP algorithm) constantly iteration updates each β parameter in neural network,
To obtain the value of neural network β parameter in the case of loss function II minimizes.
After model training, as shown in Fig. 5 (a), the design object in the present embodiment is at 15.9GHz and 18GHz
Super surface cell with microwave absorbing property, from 8GHz to 14.5GHz, phase has the change of 180 degree.
Using phase and the design object of reflection coefficient as input, deep learning model can export the matrix of 8*8 automatically:
In order to verify the accuracy of design result, the structure which is indicated in CST STUDIO SUITE MWS into
Row calculates, and as shown in Fig. 5 (b), the electromagnetic property and design object of model are very close, at the 15.9GHz and 18GHz there are two
Strong absworption peak presents strong microwave absorbing property；From 8GHz to 14.5GHz, super surface realizes the phase change of 180 degree, this proof
The validity of design result.
A kind of super surface automatic design method based on convolution self-encoding encoder of the present invention, the advantage is that and be: by super table
The reflected amplitudes and phase of face structure are connected by the super surface design method based on convolution self-encoding encoder, are being designed
Cheng Zhong can accomplish the reflected amplitudes and reflected phase that design super surface simultaneously；It can be real using the method for convolution self-encoding encoder
Existing unsupervised learning, without modeling and parameter optimization, this can automatically generate super surface texture, simplify design procedure, mention
Design efficiency is risen.
Claims (6)
1. a kind of super surface automatic design method based on convolution self-encoding encoder, which comprises the following steps:
Step 1, several super surface cell structures are generated at random, calculate separately the anti-of each super surface using electromagnetic simulation software
Penetrate phase and amplitude；
Step 2, using the deep learning method based on convolution self-encoding encoder, pass through the reflected phase that will be calculated in step 1
Convolution self-encoding encoder is inputted simultaneously with amplitude, exports corresponding super surface cell structure, Lai Xunlian deep learning model；
Step 3, it will train in the deep learning model finished, use in the reflected phase of design object and amplitude input step 2
The convolution self-encoding encoder completes the matching between feature extraction and feature and super surface matrix, that is, designs required for obtaining
Super surface texture.
2. a kind of super surface automatic design method based on convolution self-encoding encoder according to claim 1, which is characterized in that
The electromagnetic simulation software in the step 1 is CST STUDIO SUITE MWS.
3. a kind of super surface automatic design method based on convolution self-encoding encoder according to claim 2, which is characterized in that
Detailed process is as follows for the step 1:
The matrix of the super surface cell structure generated at random using the output of MATLAB software first, wherein being labeled as the region of " 1 "
It indicates that the area filling has metal, indicates the region blank labeled as the region of " 0 ", reuse the electromagnetic simulation software CST
STUDIO SUITE MWS carries out reflected phase and amplitude that the super surface cell is calculated.
4. a kind of super surface automatic design method based on convolution self-encoding encoder according to claim 1, which is characterized in that
Detailed process is as follows for training deep learning model in the step 2:
Step 2.1, data acquisition generate and preprocessing process: using MATLAB language draw super surface cell reflected phase and
Operation is normalized in the image of amplitude, gray scale and pixel characteristic to the image, sets image as 32 × 32 × 1 single channel
Grayscale image limits the pixel of image between 0~1, and using image of this pixel between 0~1 as deep learning model
Input；
Step 2.2, using convolution self-encoding encoder extract feature process: by the pixel obtained in the step 2.1 0~1 it
Between image input convolution self-encoding encoder, pass through convolution several times, pondization operation and the accordingly up-sampling of number, deconvolution first
The reflected phase of input and amplitude curve compression of images are reconfigured in a representation space by operation, then further according to this expression
The output data matrix that space extracts to the end to data；Then under the coded portion of convolution self-encoding encoder being disassembled
Come, input image data is encoded, original input image is compressed into expression vector, is rolled up by I iteration adjustment of activation primitive
Product core obtains expression vector when loss function I minimizes until the minimum of loss function I；
Step 2.3, the super surface texture matrix matching process based on convolution self-encoding encoder: pass through activation primitive II and loss function
II is trained using the artificial neural network connected entirely, the expression moment of a vector that the coded portion of input convolution self-encoding encoder generates
Battle array exports super surface texture matrix, until when loss function I minimizes in completion step 2.2 when loss function II minimizes
The matching indicated between vector characteristics and super surface texture, deep learning model training finishes.
5. a kind of super surface automatic design method based on convolution self-encoding encoder according to claim 4, which is characterized in that
The activation primitive I isWherein x represents the output data after each layer of linear weighted function of neural network；
The loss function I is cross entropy cost functionWherein x represents output data
Number, y be desired output, y' be actual output.
6. a kind of super surface automatic design method based on convolution self-encoding encoder according to claim 4, which is characterized in that
The activation primitive II uses ReLU=max (0, x), and wherein x represents the output after each layer line weighted sum of neural network
Data；The loss function II is mean square errorWherein yiFor desired output, y 'iFor reality output.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910315408.0A CN110110398B (en) | 2019-04-19 | 2019-04-19 | Super-surface automatic design method based on convolution self-encoder |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910315408.0A CN110110398B (en) | 2019-04-19 | 2019-04-19 | Super-surface automatic design method based on convolution self-encoder |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN110110398A true CN110110398A (en) | 2019-08-09 |
| CN110110398B CN110110398B (en) | 2023-05-12 |
Family
ID=67485763
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910315408.0A Active CN110110398B (en) | 2019-04-19 | 2019-04-19 | Super-surface automatic design method based on convolution self-encoder |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN110110398B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110647024A (en) * | 2019-10-30 | 2020-01-03 | 武汉大学 | Method for realizing circuitous phase coding multiplexing based on super-surface array structure |
| CN111241700A (en) * | 2020-01-19 | 2020-06-05 | 中国科学院光电技术研究所 | Intelligent design method of microwave broadband super-surface absorber |
| CN111430903A (en) * | 2020-04-01 | 2020-07-17 | 中国人民解放军空军工程大学 | Radiation scattering integrated low-RCS antenna housing and design method thereof |
| CN111767678A (en) * | 2020-07-06 | 2020-10-13 | 桂林电子科技大学 | Structure on-demand design method of metamaterial electromagnetic induction transparent device based on deep learning neural network |
| CN113205179A (en) * | 2021-05-08 | 2021-08-03 | 湖南太观科技有限公司 | Deep learning framework for medium laser acceleration |
| CN113822424A (en) * | 2021-07-27 | 2021-12-21 | 湖南大学 | All-optical diffraction neural network system based on super-structured surface |
| CN115167088A (en) * | 2022-07-12 | 2022-10-11 | 中国人民解放军空军工程大学 | Customizable holographic super-surface design method based on deep learning |
| CN115964955A (en) * | 2023-01-11 | 2023-04-14 | 哈尔滨工业大学 | Discrete coding super surface unit electromagnetic response prediction method and device |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014191781A (en) * | 2013-03-28 | 2014-10-06 | Asahi Kasei Corp | Meta material design method, meta material structure, meta material design system, and program |
| CN107967386A (en) * | 2017-11-27 | 2018-04-27 | 上海交通大学 | The design method and its device of electromagnetism metamaterial unit |
| CN108539427A (en) * | 2018-06-16 | 2018-09-14 | 中国人民解放军空军工程大学 | The super surface Ai Li Beam generators and design method regulated and controled simultaneously based on amplitude and phase |
| CN108733924A (en) * | 2018-05-21 | 2018-11-02 | 东南大学 | A kind of intellectualized design method of digital coding metamaterial unit |
| CN109489485A (en) * | 2018-11-28 | 2019-03-19 | 浙江大学 | A kind of adaptive super surface electromagnetism stealth clothing system and its working method |
-
2019
- 2019-04-19 CN CN201910315408.0A patent/CN110110398B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2014191781A (en) * | 2013-03-28 | 2014-10-06 | Asahi Kasei Corp | Meta material design method, meta material structure, meta material design system, and program |
| CN107967386A (en) * | 2017-11-27 | 2018-04-27 | 上海交通大学 | The design method and its device of electromagnetism metamaterial unit |
| CN108733924A (en) * | 2018-05-21 | 2018-11-02 | 东南大学 | A kind of intellectualized design method of digital coding metamaterial unit |
| CN108539427A (en) * | 2018-06-16 | 2018-09-14 | 中国人民解放军空军工程大学 | The super surface Ai Li Beam generators and design method regulated and controled simultaneously based on amplitude and phase |
| CN109489485A (en) * | 2018-11-28 | 2019-03-19 | 浙江大学 | A kind of adaptive super surface electromagnetism stealth clothing system and its working method |
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110647024A (en) * | 2019-10-30 | 2020-01-03 | 武汉大学 | Method for realizing circuitous phase coding multiplexing based on super-surface array structure |
| CN111241700B (en) * | 2020-01-19 | 2022-12-30 | 中国科学院光电技术研究所 | Intelligent design method of microwave broadband super-surface absorber |
| CN111241700A (en) * | 2020-01-19 | 2020-06-05 | 中国科学院光电技术研究所 | Intelligent design method of microwave broadband super-surface absorber |
| CN111430903A (en) * | 2020-04-01 | 2020-07-17 | 中国人民解放军空军工程大学 | Radiation scattering integrated low-RCS antenna housing and design method thereof |
| CN111430903B (en) * | 2020-04-01 | 2021-08-10 | 中国人民解放军空军工程大学 | Radiation scattering integrated low-RCS antenna housing and design method thereof |
| CN111767678A (en) * | 2020-07-06 | 2020-10-13 | 桂林电子科技大学 | Structure on-demand design method of metamaterial electromagnetic induction transparent device based on deep learning neural network |
| CN113205179A (en) * | 2021-05-08 | 2021-08-03 | 湖南太观科技有限公司 | Deep learning framework for medium laser acceleration |
| CN113822424A (en) * | 2021-07-27 | 2021-12-21 | 湖南大学 | All-optical diffraction neural network system based on super-structured surface |
| CN113822424B (en) * | 2021-07-27 | 2023-10-20 | 湖南大学 | All-optical diffraction neural network system based on super-structured surface |
| CN115167088A (en) * | 2022-07-12 | 2022-10-11 | 中国人民解放军空军工程大学 | Customizable holographic super-surface design method based on deep learning |
| CN115167088B (en) * | 2022-07-12 | 2024-04-16 | 中国人民解放军空军工程大学 | Customizable holographic super-surface design method based on deep learning |
| CN115964955A (en) * | 2023-01-11 | 2023-04-14 | 哈尔滨工业大学 | Discrete coding super surface unit electromagnetic response prediction method and device |
| CN115964955B (en) * | 2023-01-11 | 2023-12-15 | 哈尔滨工业大学 | Discrete coding super-surface unit electromagnetic response prediction method and device |
Also Published As
| Publication number | Publication date |
|---|---|
| CN110110398B (en) | 2023-05-12 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110110398A (en) | A kind of super surface automatic design method based on convolution self-encoding encoder | |
| CN111814875B (en) | Ship sample expansion method in infrared image based on pattern generation countermeasure network | |
| CN110119780B (en) | Hyper-spectral image super-resolution reconstruction method based on generation countermeasure network | |
| CN103646410B (en) | Fast magnetic resonance parametric formation method and system | |
| Kelly et al. | FrankenGAN: guided detail synthesis for building mass-models using style-synchonized GANs | |
| Mahdizadehaghdam et al. | Sparse generative adversarial network | |
| CN112686817B (en) | Image completion method based on uncertainty estimation | |
| CN114821261B (en) | Image fusion method | |
| CN101950365A (en) | Multi-task super-resolution image reconstruction method based on KSVD dictionary learning | |
| CN110276148A (en) | The feature extraction of micro-structure dimensionality reduction and reconstruct implementation method based on self-encoding encoder | |
| CN108983228A (en) | A kind of RCS Near-far fields transfer method based on deep neural network | |
| CN108230223A (en) | Light field angle super-resolution rate method and device based on convolutional neural networks | |
| CN109598676A (en) | A kind of single image super-resolution method based on Hadamard transform | |
| CN103337085A (en) | Efficient portrait face distortion method | |
| CN107220936A (en) | A kind of image super-resolution reconstructing method and system | |
| CN114463183A (en) | Image super-resolution method based on frequency domain and spatial domain | |
| CN108520513A (en) | A kind of threedimensional model local deformation component extraction method and system | |
| CN108495132A (en) | The big multiplying power compression method of remote sensing image based on lightweight depth convolutional network | |
| Jang et al. | Dual path denoising network for real photographic noise | |
| CN113096015B (en) | Image super-resolution reconstruction method based on progressive perception and ultra-lightweight network | |
| CN107105245A (en) | High speed JPEG method for compressing image based on TMS320C6678 chips | |
| CN114494003A (en) | Ancient character generation method combining shape transformation and texture transformation | |
| CN114022356A (en) | River course flow water level remote sensing image super-resolution method and system based on wavelet domain | |
| CN111241700B (en) | Intelligent design method of microwave broadband super-surface absorber | |
| CN112862946B (en) | Gray rock core image three-dimensional reconstruction method for generating countermeasure network based on cascade condition |
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 |