CN107958475A - Varied angle illumination based on deep learning generation network chromatographs method and device - Google Patents
Varied angle illumination based on deep learning generation network chromatographs method and device Download PDFInfo
- Publication number
- CN107958475A CN107958475A CN201711372608.7A CN201711372608A CN107958475A CN 107958475 A CN107958475 A CN 107958475A CN 201711372608 A CN201711372608 A CN 201711372608A CN 107958475 A CN107958475 A CN 107958475A
- Authority
- CN
- China
- Prior art keywords
- mrow
- sample
- network
- field
- layer
- 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
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T11/00—2D [Two Dimensional] image generation
- G06T11/003—Reconstruction from projections, e.g. tomography
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Biophysics (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention discloses a kind of varied angle illumination based on deep learning generation network to chromatograph method and device, wherein, method includes：Derive to obtain diffractional field and the distributed model of refraction field when light is successively propagated in non-homogeneous transparent medium according to Fourier's propagation model of the Helmholtz equation of ripple, light；Copy physical process to build and neutral net is generated in the deep learning that time domain and frequency domain are propagated with plural form；The output light complex field for not passing through sample collected is regard as input light plural number field data by angular spectrum propagation formula back-propagation, and using the complex field through sample to be reconstructed collected as output light plural number field data；According to resolution condition percentage regulation learning network parameter is rebuild, to be trained to network；The three-dimensional refractive index that the weight obtained by training solves to obtain sample is distributed, and realizes the tomographic reconstruction to sample.Low collection capacity, high-resolution tomographic reconstruction ability are the method achieve, effectively improves the resolution accuracy of sample tomographic reconstruction.
Description
Technical field
The present invention relates to calculating optical, computer vision and shooting technical field is calculated, it is more particularly to a kind of based on deep
The varied angle illumination chromatography method and device of degree study generation network.
Background technology
At present, it is current calculating light that high-resolution tomographic reconstruction is carried out to micro- sample, particularly living body biological sample
Study picture, computer vision, the hot research problem for calculating the ambits such as shooting.In relevant chromatographic technique, due to big
Most living body biological cells have the characteristics that weakly heterogeneous in intensity and High Defferential in phase, therefore widely use phase imaging technology
Studied.But existing phase chromatographic technique needs to gather substantial amounts of data mostly, includes the image of different angle irradiation
Or the image of different depth shooting is focused on, and the development and application of the rate limitation phase chromatography gathered.
Another common problem is in living body biological sample tomographic reconstruction, and existing reconstruction technique is in optical axis direction
More serious elongation phenomenon often occurs, causes larger error, so as to limit the chromatography resolution ratio of optical axis direction, shadow
Ring the effect of tomographic reconstruction, it is difficult to realize nano level chromatography.Meanwhile in the chromatography method proposed before, the light field that uses
Propagation model is the linear propagation model for ignoring multiple scattering mostly.So doing can make algorithm become more simple and convenient, still
It can influence chromatographic effect.
The content of the invention
It is contemplated that solve at least some of the technical problems in related technologies.
For this reason, an object of the present invention is to provide a kind of varied angle illumination chromatography based on deep learning generation network
Method, the method achieve low collection capacity, high-resolution tomographic reconstruction ability, effectively improve the resolution ratio of sample tomographic reconstruction
Precision.
It is another object of the present invention to propose that a kind of varied angle illumination based on deep learning generation network chromatographs to fill
Put.
To reach above-mentioned purpose, one aspect of the present invention embodiment proposes a kind of angle based on deep learning generation network
Illumination chromatography method is spent, is comprised the following steps：Derive to obtain according to Fourier's propagation model of the Helmholtz equation of ripple, light
Diffractional field and the distributed model of refraction field when light is successively propagated in non-homogeneous transparent medium；According to the distributed model of derivation
Copy physical process to build and neutral net is generated in the deep learning that time domain and frequency domain are propagated with plural form, wherein, wait to train
Weight is the body index distribution of sample to be chromatographed, and training sample is the field distribution of input light plural number and the output light of corresponding angle
Plural field distribution；Multigroup irradiation is carried out to sample in the range of predetermined angle, and camera is fixed on optical axis rear end and carries out data
Collection, regard the output light complex field for not passing through sample collected as input recovery by angular spectrum propagation formula back-propagation
Number field data, and using the complex field through sample to be reconstructed collected as output light plural number field data；Differentiated according to rebuilding
Rate condition percentage regulation learning network parameter, to be trained to network；The weight obtained by training solves to obtain sample
Three-dimensional refractive index is distributed, and realizes the tomographic reconstruction to sample.
The varied angle illumination chromatography method based on deep learning generation network of the embodiment of the present invention, can be by using base
In the delamination propagation model of beam propagation method, scattering and reflected that light propagates in multi-layer transparent sample have been considered
Journey, itself and deep learning this current effect optimization method the most prominent is combined, is coordinated with digital hologram collection side
Method, realizes low collection capacity, high-resolution tomographic reconstruction ability, effectively improves the resolution accuracy of sample tomographic reconstruction.
In addition, the varied angle illumination chromatography method according to the above embodiment of the present invention based on deep learning generation network is also
There can be following additional technical characteristic：
Further, in one embodiment of the invention, the diffractional field and the distributed model of refraction field are expressed as：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Further, in one embodiment of the invention, the distributed model according to derivation copies physics mistake
Journey is built generates neutral net with plural form in the deep learning that time domain and frequency domain are propagated, and further comprises：
With the node that the envelope complex amplitude a (r) of the distribution field of light is the neutral net, for the node a of adjacent two layers
(x, y, z) and a (x, y, z+ δ z), operation relation are divided into two parts corresponding to the diffraction during light propagation and refraction：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator,For Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Further, in one embodiment of the invention, building the frame used during the neutral net is
TensorFlow, the layer used in network include input layer, diffracting layer, refracting layer and low-pass filtering layer, wherein, it is described to spread out
Penetrate layer and the refracting layer corresponds to above-mentioned diffraction process computing and refracting process computing respectively, the low-pass filtering layer corresponds to
The frequency domain characteristic of collection.
Further, in one embodiment of the invention, it is characterised in that in a network, the expression formula of loss function
It is as follows：
Loss=∑s | ypredict-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresent the data truly gathered, S represents sparse item constraint.
To reach above-mentioned purpose, another aspect of the present invention embodiment proposes a kind of change based on deep learning generation network
Angular light shines chromatographic apparatus, including：Derivation module, Fourier's propagation model for the Helmholtz equation according to ripple, light push away
Export obtains diffractional field and the distributed model of refraction field when light is successively propagated in non-homogeneous transparent medium；Module is built, is used for
Copy physical process to build according to the distributed model of derivation to give birth in the deep learning that time domain and frequency domain are propagated with plural form
Into neutral net, wherein, weight to be trained is the body index distribution of sample to be chromatographed, and training sample is input light complex field point
The output light plural number field distribution of cloth and corresponding angle；Acquisition module, it is multigroup for being carried out in the range of predetermined angle to sample
Irradiation, and camera is fixed on optical axis rear end and carries out data acquisition, the output light complex field for not passing through sample that will be collected
Input light plural number field data, and the plural number through sample to be reconstructed that will be collected are used as by angular spectrum propagation formula back-propagation
Field is used as output light plural number field data；Training module, for according to rebuild resolution condition percentage regulation learning network parameter, with
Network is trained；Module is rebuild, the three-dimensional refractive index that the weight for being obtained by training solves to obtain sample is distributed, real
Now to the tomographic reconstruction of sample.
The varied angle illumination chromatographic apparatus based on deep learning generation network of the embodiment of the present invention, can be by using base
In the delamination propagation model of beam propagation method, scattering and reflected that light propagates in multi-layer transparent sample have been considered
Journey, itself and deep learning this current effect optimization method the most prominent is combined, is coordinated with digital hologram collection side
Method, realizes low collection capacity, high-resolution tomographic reconstruction ability, effectively improves the resolution accuracy of sample tomographic reconstruction.
In addition, the varied angle illumination chromatographic apparatus according to the above embodiment of the present invention based on deep learning generation network is also
There can be following additional technical characteristic：
Further, in one embodiment of the invention, the diffractional field and the distributed model of refraction field are expressed as：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Further, in one embodiment of the invention, the module of building is additionally operable to the envelope of the distribution field of light
Complex amplitude a (r) is the node of the neutral net, and for the node a (x, y, z) and a (x, y, z+ δ z) of adjacent two layers, computing is closed
System is divided into two parts corresponding to the diffraction during light propagation and refraction：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Further, in one embodiment of the invention, building the frame used during the neutral net is
TensorFlow, the layer used in network include input layer, diffracting layer, refracting layer and low-pass filtering layer, wherein, it is described to spread out
Penetrate layer and the refracting layer corresponds to above-mentioned diffraction process computing and refracting process computing respectively, the low-pass filtering layer corresponds to
The frequency domain characteristic of collection.
Further, in one embodiment of the invention, in a network, the expression formula of loss function is as follows：
Loss=∑s | ypredict-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresent the data truly gathered, S represents sparse item constraint.
The additional aspect of the present invention and advantage will be set forth in part in the description, and will partly become from the following description
Obtain substantially, or recognized by the practice of the present invention.
Brief description of the drawings
Of the invention above-mentioned and/or additional aspect and advantage will become from the following description of the accompanying drawings of embodiments
Substantially and it is readily appreciated that, wherein：
Fig. 1 is the varied angle illumination chromatography method that network is generated based on deep learning according to one embodiment of the invention
Flow chart；
Fig. 2 shows for the emulating image based on beam propagation method delamination propagation model according to one embodiment of the invention
It is intended to；
Fig. 3 is the neural network framework structural representation that network is generated based on deep learning according to one embodiment of the invention
Figure；
Fig. 4 is based on varied angle illumination holographic acquisition system structure diagram according to one embodiment of the invention；
Fig. 5 is a kind of tomographic reconstruction result schematic diagram for emulating bead according to one embodiment of the invention；
Fig. 6 is the varied angle illumination chromatography side that network is generated based on deep learning according to one specific embodiment of the present invention
The flow chart of method；
Fig. 7 is the varied angle illumination chromatographic apparatus that network is generated based on deep learning according to one embodiment of the invention
Structure diagram.
Embodiment
The embodiment of the present invention is described below in detail, the example of the embodiment is shown in the drawings, wherein from beginning to end
Same or similar label represents same or similar element or has the function of same or like element.Below with reference to attached
The embodiment of figure description is exemplary, it is intended to for explaining the present invention, and is not considered as limiting the invention.
The varied angle light based on deep learning generation network proposed according to embodiments of the present invention is described with reference to the accompanying drawings
According to chromatography method and device, describe to propose according to embodiments of the present invention first with reference to the accompanying drawings generates network based on deep learning
Varied angle illumination chromatography method.
Fig. 1 is the flow of the varied angle illumination chromatography method based on deep learning generation network of one embodiment of the invention
Figure.
As shown in Figure 1, it should be comprised the following steps based on the varied angle illumination chromatography method of deep learning generation network：
In step S101, derive to obtain light non-according to Fourier's propagation model of the Helmholtz equation of ripple, light
Diffractional field and the distributed model of refraction field when successively being propagated in homogeneous transparent medium.
It is understood that the embodiment of the present invention can be according to the Helmholtz equation of ripple, Fourier's propagation model of light
And a series of rational approximating assumptions derive diffractional field when light extraction is successively propagated in non-homogeneous transparent medium and reflect field
Distributed model.
For example, as shown in Fig. 2, the BPM (Beam Propagation Method, the light that are derived in the embodiment of the present invention
Beam transmission method) delamination propagation model, theoretical model can be emulated using computer.Input light complex field is set for height
Image space, is focused on the central core of sample, output image by the amplitude and phase distribution of this light beam.
Alternatively, in one embodiment of the invention, diffractional field and the distributed model of refraction field are expressed as：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Specifically, the basic principle of the embodiment of the present invention is Helmholtz equation in inhomogeneous medium and its derives
Paraxial ripple propagation field distribution formula, be respectively equation 1 and equation 2,
Wherein, r=(x, y, z) representation space position distribution, u are the distribution fields of r positions light,
For Laplace operator, I is feature operator,It is the wave number of the light of r positions.
Wherein,n0It is the refractive index of background media, a (r) represents the complex amplitude envelope of u (r)
Two kinds of approximations are introduced to upper two formula to simplify.The change that the first envelope for being approximately considered plane wave answers amplitude is slow
, i.e.,Second is approximately under the premise of index distribution disturbance δ n (r) are less, ignores (δ n (r)
)2And other higher order terms.It is possible thereby to push away：
Equation 3 is the Helmholtz equation of paraxial ripple.
Peer-to-peer 3, which carries out Fourier transformation, finally to be released,
Equation 4 can be divided into diffraction and refraction two parts represent：
In step s 102, physical process is copied to build with plural form in time domain and frequency domain according to the distributed model of derivation
The deep learning generation neutral net of propagation, wherein, weight to be trained is the body index distribution of sample to be chromatographed, training sample
For the field distribution of input light plural number and the output light plural number field distribution of corresponding angle.
That is, the embodiment of the present invention can copy the physical model to build according to the optical propagation model derived
Neutral net is generated in the deep learning that time domain and frequency domain are propagated with plural form, weight to be trained is divided for the body refractive index of sample
Cloth, training sample are the field distribution of input light plural number and the output light plural number field distribution of corresponding angle.Wherein, deep learning generates
The frame structure of neutral net is as shown in Figure 3.
For example, the concept of application of embodiment of the present invention infinitesimal, by sample infinitesimal, gridding, it is believed that each infinitesimal
In refractive index be identical.And the size of this infinitesimal, also embody the size for chromatographing resolution ratio.Using being derived by before
Formula (4), network can be built.
Further, in one embodiment of the invention, according to the distributed model of derivation copy physical process to build with
Plural form generates neutral net in the deep learning that time domain and frequency domain are propagated, and further comprises：With the envelope of the distribution field of light
Complex amplitude a (r) is the node of neutral net, for the node a (x, y, z) and a (x, y, z+ δ z) of adjacent two layers, operation relation point
For two parts corresponding to the diffraction during light propagation and refraction：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Specifically, the knot that the embodiment of the present invention can be with the envelope complex amplitude a (r) of the distribution field of light for neutral net
Point, for the node a (x, y, z) and a (x, y, z+ δ z) of adjacent two layers, its operation relation can be divided into two parts：
Correspond respectively to the diffraction during light propagation and refraction.
Further, in one embodiment of the invention, building the frame used during neutral net is
TensorFlow, the layer used in network include input layer, diffracting layer, refracting layer and low-pass filtering layer, wherein, diffracting layer
Correspond to above-mentioned diffraction process computing and refracting process computing respectively with refracting layer, the frequency domain that low-pass filtering layer corresponds to collection is special
Property.
It is understood that the frame that the embodiment of the present invention can use during deep learning neutral net is built is
TensorFlow, the layer used in network include input layer (Input), diffracting layer (DiffractionLayer), refracting layer
(RefractionLayer) and low-pass filtering layer (LowPassLayer), wherein diffracting layer and refracting layer correspond to above-mentioned respectively
Diffraction process computing and refracting process computing, low-pass filtering layer correspond to acquisition system in harvester frequency domain characteristic.It is right
For given sample, diffracting layer is identical, therefore can be shared；Parameter δ n (r) to be asked are included in refracting layer, can not be total to
Enjoy, for layer to be trained, δ n (r) are its weight to be trained.
Alternatively, in one embodiment of the invention, in a network, the expression formula of loss function is as follows：
Loss=∑s | ypredic-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresent the data truly gathered, S represents sparse item constraint.
Specifically, in a network, loss function (loss function) expression formula of definition is as follows：
Loss=∑s | ypredict-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresenting the data truly gathered, S represents sparse item constraint,
Its expression formula is as follows：
Wherein, γ1Represent data item sparsity constraints, γ2Represent differential term sparsity constraints, w represents weight, i.e. δ n
(r)。
In step s 103, multigroup irradiation is carried out to sample in the range of predetermined angle, and after camera is fixed on optical axis
End carries out data acquisition, and the output light complex field for not passing through sample collected is made by angular spectrum propagation formula back-propagation
For input light plural number field data, and using the complex field through sample to be reconstructed collected as output light plural number field data.
It is understood that the embodiment of the present invention can carry out multigroup irradiation in the range of certain angle to sample, by phase
Machine is fixed on optical axis rear end and carries out data acquisition, and the output light complex field for not passing through sample collected is propagated public affairs by angular spectrum
Formula back-propagation is as input light plural number field data, using the complex field through sample to be reconstructed collected as output light plural number
Field data.
For example, it is fixed as shown in figure 3, the embodiment of the present invention can change the angle of incident light by rotating galvanometer
Collection terminal camera position is constant；By the interference effect with reference light, hologram is photographed, so as to obtain the plural number of output light
.
In step S104, according to resolution condition percentage regulation learning network parameter is rebuild, to be trained to network.
It is understood that the embodiment of the present invention can be according to the requirement percentage regulation study generation network for rebuilding resolution ratio
Parameter, including reconstruction sample parameter --- rebuild three-dimensional network size, resolution elements size etc. and deep learning surpasses ginseng
Number --- initial learning rate, batch processing size and sparsity constraints item parameter etc. are trained network.
For example, the deep learning neutral net that content is built before the embodiment of the present invention can utilize, according to reality
The amplification factor of the scale of sample, resolution requirements and collection terminal, sets the parameter of neutral net, and the number of plies such as network is (corresponding
In axial resolution), every layer of size (corresponding to horizontal plane resolution ratio), batch processing size (depend on the group of gathered data
Number), initial learning rate and iterations etc., start training network afterwards.
In step S105, the weight obtained by training solves to obtain the three-dimensional refractive index distribution of sample, realizes to sample
This tomographic reconstruction.
It is understood that the embodiment of the present invention can be trained by using adaptive moments estimation (Adam) optimization method,
Solve to obtain the body index distribution of sample by obtained weight, realize the three-dimensional reconstruction to sample.
Specifically, after network training, trained weight parameter, i.e. δ n (r) are taken out, it is by its stratification drawing, i.e., real
The chromatography to sample is showed.Wherein, Fig. 5 is the tomographic reconstruction result schematic diagram for emulating bead.
In one particular embodiment of the present invention, as shown in fig. 6, the method for the embodiment of the present invention comprises the following steps：
Step S1, light in media as well diffraction occasion refraction field distributed model；
Step S2, copies physical process to build the deep learning generation network of plural form；
Step S3, different angle coherent light carry out multigroup irradiation；
Step S4, output terminal gather the complex field of light；
Step S5, adjusts network parameter, is trained；
Step S6, according to training weight calculation body index distribution, realizes chromatography.
To sum up, the embodiment of the present invention coordinates varied angle plane wave illumination acquisition system using deep learning neutral net, real
Existing high speed acquisition, the function of precisely chromatographing.
The varied angle illumination chromatography method based on deep learning generation network proposed according to embodiments of the present invention, Ke Yitong
Cross using the delamination propagation model based on beam propagation method, considered scattering that light propagates in multi-layer transparent sample and
Refracting process, itself and deep learning this current effect optimization method the most prominent is combined, is coordinated with digital hologram
Acquisition method, realizes low collection capacity, high-resolution tomographic reconstruction ability, effectively improves the resolution ratio essence of sample tomographic reconstruction
Degree.
The varied angle light based on deep learning generation network proposed according to embodiments of the present invention referring next to attached drawing description
According to chromatographic apparatus.
Fig. 7 is the structure of the varied angle illumination chromatographic apparatus based on deep learning generation network of one embodiment of the invention
Schematic diagram.
As shown in fig. 7, it should be included based on the varied angle illumination chromatographic apparatus 10 of deep learning generation network：Derivation module
100th, module 200, acquisition module 300, training module 400 are built and rebuilds module 500.
Wherein, derivation module 100 is used to derive to obtain according to the Helmholtz equation of ripple, Fourier's propagation model of light
Diffractional field and the distributed model of refraction field when light is successively propagated in non-homogeneous transparent medium.Module 200 is built to be used for according to pushing away
The distributed model led copies physical process to build and generates neutral net in the deep learning that time domain and frequency domain are propagated with plural form,
Wherein, weight to be trained is the body index distribution of sample to be chromatographed, and training sample is the field distribution of input light plural number and correspondence
The output light plural number field distribution of angle.Acquisition module 300 is used in the range of predetermined angle carry out sample multigroup irradiation, and will
Camera is fixed on optical axis rear end and carries out data acquisition, and the output light complex field for not passing through sample collected is passed by angular spectrum
Formula back-propagation is broadcast as input light plural number field data, and using the complex field through sample to be reconstructed collected as output
Recover number field data.Training module 400 be used for according to rebuild resolution condition percentage regulation learning network parameter, with to network into
Row training.Rebuild module 500 to be used to solve to obtain the three-dimensional refractive index distribution of sample by the weight that training obtains, realize to sample
This tomographic reconstruction.The device 10 of the embodiment of the present invention realizes low collection capacity, high-resolution tomographic reconstruction ability, effectively carries
The resolution accuracy of high sample tomographic reconstruction.
Further, in one embodiment of the invention, diffractional field and the distributed model of refraction field are expressed as：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Further, in one embodiment of the invention, module 200 is built to be additionally operable to answer with the envelope of the distribution field of light
Amplitude a (r) is the node of neutral net, and for the node a (x, y, z) and a (x, y, z+ δ z) of adjacent two layers, operation relation is divided into
Corresponding to the diffraction during light propagation and two parts of refraction：
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
Further, in one embodiment of the invention, building the frame used during neutral net is
TensorFlow, the layer used in network include input layer, diffracting layer, refracting layer and low-pass filtering layer, wherein, diffracting layer
Correspond to above-mentioned diffraction process computing and refracting process computing respectively with refracting layer, the frequency domain that low-pass filtering layer corresponds to collection is special
Property.
Further, in one embodiment of the invention, in a network, the expression formula of loss function is as follows：
Loss=∑s | ypredict-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresent the data truly gathered, S represents sparse item constraint.
It should be noted that the foregoing solution to the varied angle illumination chromatography embodiment of the method based on deep learning generation network
The varied angle illumination chromatographic apparatus based on deep learning generation network that explanation is also applied for the embodiment is released, it is no longer superfluous herein
State.
The varied angle illumination chromatographic apparatus based on deep learning generation network proposed according to embodiments of the present invention, Ke Yitong
Cross using the delamination propagation model based on beam propagation method, considered scattering that light propagates in multi-layer transparent sample and
Refracting process, itself and deep learning this current effect optimization method the most prominent is combined, is coordinated with digital hologram
Acquisition method, realizes low collection capacity, high-resolution tomographic reconstruction ability, effectively improves the resolution ratio essence of sample tomographic reconstruction
Degree.
In the description of the present invention, it is to be understood that term " " center ", " longitudinal direction ", " transverse direction ", " length ", " width ",
" thickness ", " on ", " under ", "front", "rear", "left", "right", " vertical ", " level ", " top ", " bottom " " interior ", " outer ", " up time
The orientation or position relationship of the instruction such as pin ", " counterclockwise ", " axial direction ", " radial direction ", " circumferential direction " be based on orientation shown in the drawings or
Position relationship, is for only for ease of and describes the present invention and simplify description, rather than indicates or imply that signified device or element must
There must be specific orientation, with specific azimuth configuration and operation, therefore be not considered as limiting the invention.
In addition, term " first ", " second " are only used for description purpose, and it is not intended that instruction or hint relative importance
Or the implicit quantity for indicating indicated technical characteristic.Thus, define " first ", the feature of " second " can be expressed or
Implicitly include at least one this feature.In the description of the present invention, " multiple " are meant that at least two, such as two, three
It is a etc., unless otherwise specifically defined.
In the present invention, unless otherwise clearly defined and limited, term " installation ", " connected ", " connection ", " fixation " etc.
Term should be interpreted broadly, for example, it may be fixedly connected or be detachably connected, or integrally；Can be that machinery connects
Connect or be electrically connected；It can be directly connected, can also be indirectly connected by intermediary, can be in two elements
The connection in portion or the interaction relationship of two elements, unless otherwise restricted clearly.For those of ordinary skill in the art
For, the concrete meaning of above-mentioned term in the present invention can be understood as the case may be.
In the present invention, unless otherwise clearly defined and limited, fisrt feature can be with "above" or "below" second feature
It is that the first and second features directly contact, or the first and second features pass through intermediary mediate contact.Moreover, fisrt feature exists
Second feature " on ", " top " and " above " but fisrt feature are directly over second feature or oblique upper, or be merely representative of
Fisrt feature level height is higher than second feature.Fisrt feature second feature " under ", " lower section " and " below " can be
One feature is immediately below second feature or obliquely downward, or is merely representative of fisrt feature level height and is less than second feature.
In the description of this specification, reference term " one embodiment ", " some embodiments ", " example ", " specifically show
The description of example " or " some examples " etc. means specific features, structure, material or the spy for combining the embodiment or example description
Point is contained at least one embodiment of the present invention or example.In the present specification, schematic expression of the above terms is not
It must be directed to identical embodiment or example.Moreover, particular features, structures, materials, or characteristics described can be in office
Combined in an appropriate manner in one or more embodiments or example.In addition, without conflicting with each other, the skill of this area
Art personnel can be tied the different embodiments or example described in this specification and different embodiments or exemplary feature
Close and combine.
Although the embodiment of the present invention has been shown and described above, it is to be understood that above-described embodiment is example
Property, it is impossible to limitation of the present invention is interpreted as, those of ordinary skill in the art within the scope of the invention can be to above-mentioned
Embodiment is changed, changes, replacing and modification.
Claims (10)
1. a kind of varied angle illumination chromatography method based on deep learning generation network, it is characterised in that comprise the following steps：
Derive to obtain light in non-homogeneous transparent medium successively according to Fourier's propagation model of the Helmholtz equation of ripple, light
Diffractional field and the distributed model of refraction field during propagation；
Physical process is copied to build with plural form in time domain and the depth of frequency domain propagation according to the distributed model of derivation
Generation neutral net is practised, wherein, weight to be trained is the body index distribution of sample to be chromatographed, and training sample is input light plural number
Field distribution and the output light plural number field distribution of corresponding angle；
Multigroup irradiation is carried out to sample in the range of predetermined angle, and camera is fixed on optical axis rear end and carries out data acquisition, with
The output light complex field for not passing through sample collected is regard as input light plural number number of fields by angular spectrum propagation formula back-propagation
According to, and using the complex field through sample to be reconstructed collected as output light plural number field data；
According to resolution condition percentage regulation learning network parameter is rebuild, to be trained to network；And
The three-dimensional refractive index that the weight obtained by training solves to obtain sample is distributed, and realizes the tomographic reconstruction to sample.
2. the varied angle illumination chromatography method according to claim 1 based on deep learning generation network, it is characterised in that
The diffractional field and the distributed model of refraction field are expressed as：
<mrow>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>+</mo>
<mi>&delta;</mi>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mi>e</mi>
<mrow>
<msub>
<mi>jk</mi>
<mn>0</mn>
</msub>
<mrow>
<mo>(</mo>
<mi>&delta;</mi>
<mi>n</mi>
<mo>(</mo>
<mi>r</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mi>&delta;</mi>
<mi>z</mi>
</mrow>
</msup>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are between optical axis direction adjacent layer in hierarchical mode
Away from,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωy, it is Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
3. the varied angle illumination chromatography method according to claim 1 based on deep learning generation network, it is characterised in that
The distributed model according to derivation copies physical process to build with plural form in time domain and the depth of frequency domain propagation
Generation neutral net is practised, is further comprised：
With the envelope complex amplitude a (r) of the distribution field of light be the neutral net node, for adjacent two layers node a (x, y,
Z) and a (x, y, z+ δ z), operation relation are divided into two parts corresponding to the diffraction during light propagation and refraction：
<mrow>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>+</mo>
<mi>&delta;</mi>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mi>e</mi>
<mrow>
<msub>
<mi>jk</mi>
<mn>0</mn>
</msub>
<mrow>
<mo>(</mo>
<mi>&delta;</mi>
<mi>n</mi>
<mo>(</mo>
<mi>r</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mi>&delta;</mi>
<mi>z</mi>
</mrow>
</msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Wherein, wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are optical axis direction adjacent layer in hierarchical mode
Spacing,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωySat for Fourier
Mark,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
4. the varied angle illumination chromatography method according to claim 3 based on deep learning generation network, it is characterised in that
It is TensorFlow to build the frame used during the neutral net, and the layer used in network includes input layer, diffraction
Layer, refracting layer and low-pass filtering layer, wherein, the diffracting layer and the refracting layer correspond to above-mentioned diffraction process computing respectively
With refracting process computing, the low-pass filtering layer corresponds to the frequency domain characteristic of collection.
5. the varied angle illumination chromatography method of network is generated based on deep learning according to claim 1-4 any one of them, its
It is characterized in that, in a network, the expression formula of loss function is as follows：
Loss=∑s | ypredct-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresent the data truly gathered, S represents sparse item constraint.
6. a kind of varied angle illumination chromatographic apparatus based on deep learning generation network, including：
Derivation module, Fourier's propagation model for the Helmholtz equation according to ripple, light derive to obtain light non-homogeneous
Diffractional field and the distributed model of refraction field when successively being propagated in transparent medium；
Module is built, copies physical process to build with plural form in time domain and frequency domain for the distributed model according to derivation
The deep learning generation neutral net of propagation, wherein, weight to be trained is the body index distribution of sample to be chromatographed, training sample
For the field distribution of input light plural number and the output light plural number field distribution of corresponding angle；
Acquisition module, for carrying out multigroup irradiation to sample in the range of predetermined angle, and by camera be fixed on optical axis rear end into
Row data acquisition, the output light complex field for not passing through sample collected is used as by angular spectrum propagation formula back-propagation defeated
Enter to recover number field data, and using the complex field through sample to be reconstructed collected as output light plural number field data；
Training module, for according to rebuild resolution condition percentage regulation learning network parameter, to be trained to network；And
Module is rebuild, the three-dimensional refractive index that the weight for being obtained by training solves to obtain sample is distributed, and is realized to sample
Tomographic reconstruction.
7. the varied angle illumination chromatographic apparatus according to claim 6 based on deep learning generation network, it is characterised in that
The diffractional field and the distributed model of refraction field are expressed as：
<mrow>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>+</mo>
<mi>&delta;</mi>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mi>e</mi>
<mrow>
<msub>
<mi>jk</mi>
<mn>0</mn>
</msub>
<mrow>
<mo>(</mo>
<mi>&delta;</mi>
<mi>n</mi>
<mo>(</mo>
<mi>r</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mi>&delta;</mi>
<mi>z</mi>
</mrow>
</msup>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are between optical axis direction adjacent layer in hierarchical mode
Away from,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
8. the varied angle illumination chromatographic apparatus according to claim 6 based on deep learning generation network, it is characterised in that
The module of building is additionally operable to the node that the envelope complex amplitude a (r) of the distribution field of light is the neutral net, for adjacent two
The node a (x, y, z) and a (x, y, z+ δ z) of layer, operation relation are divided into two corresponding to the diffraction during light propagation and refraction
Part：
<mrow>
<mi>a</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>+</mo>
<mi>&delta;</mi>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>=</mo>
<msup>
<mi>e</mi>
<mrow>
<msub>
<mi>jk</mi>
<mn>0</mn>
</msub>
<mrow>
<mo>(</mo>
<mi>&delta;</mi>
<mi>n</mi>
<mo>(</mo>
<mi>r</mi>
<mo>)</mo>
<mo>)</mo>
</mrow>
<mi>&delta;</mi>
<mi>z</mi>
</mrow>
</msup>
<mi>f</mi>
<mrow>
<mo>(</mo>
<mi>x</mi>
<mo>,</mo>
<mi>y</mi>
<mo>,</mo>
<mi>z</mi>
<mo>)</mo>
</mrow>
<mo>,</mo>
</mrow>
Wherein, x, y, z is the three-dimensional coordinate of sample body index distribution, and δ z are between optical axis direction adjacent layer in hierarchical mode
Away from,Respectively Fourier transformation operation symbol and inverse Fourier transform operator, ωx、ωyFor Fourier coordinate,For wave number, n0For background media refractive index, j is imaginary unit, and δ n (r) are weight to be trained.
9. the varied angle illumination chromatographic apparatus according to claim 8 based on deep learning generation network, it is characterised in that
It is TensorFlow to build the frame used during the neutral net, and the layer used in network includes input layer, diffraction
Layer, refracting layer and low-pass filtering layer, wherein, the diffracting layer and the refracting layer correspond to above-mentioned diffraction process computing respectively
With refracting process computing, the low-pass filtering layer corresponds to the frequency domain characteristic of collection.
10. the varied angle illumination chromatographic apparatus of network is generated based on deep learning according to claim 6-9 any one of them, its
It is characterized in that, in a network, the expression formula of loss function is as follows：
Loss=∑s | ypredct-ytrue|+S,
Wherein, ypredictRepresent the data of network generation, ytrueRepresent the data truly gathered, S represents sparse item constraint.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711372608.7A CN107958475B (en) | 2017-12-19 | 2017-12-19 | Variable-angle illumination chromatography method and device based on deep learning generation network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201711372608.7A CN107958475B (en) | 2017-12-19 | 2017-12-19 | Variable-angle illumination chromatography method and device based on deep learning generation network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107958475A true CN107958475A (en) | 2018-04-24 |
| CN107958475B CN107958475B (en) | 2020-06-02 |
Family
ID=61959395
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201711372608.7A Active CN107958475B (en) | 2017-12-19 | 2017-12-19 | Variable-angle illumination chromatography method and device based on deep learning generation network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107958475B (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109932818A (en) * | 2019-05-07 | 2019-06-25 | 东北大学 | The optical fiber of a kind of pair of unknown structure parameter carries out nondegenerate mode power decomposition method |
| CN110243827A (en) * | 2019-07-18 | 2019-09-17 | 华中科技大学 | A kind of rapid three dimensional imaging process suitable for light transparent sample |
| CN110309916A (en) * | 2019-06-05 | 2019-10-08 | 清华大学 | The full optical depth learning system of multistage null tone domain Modulation and Nonlinear and method |
| CN110378473A (en) * | 2019-07-26 | 2019-10-25 | 清华大学 | Method and device is chromatographed based on deep learning and the phase of random pattern |
| CN111199574A (en) * | 2018-11-16 | 2020-05-26 | 青岛海信激光显示股份有限公司 | Holographic image generation method and equipment |
| CN111582468A (en) * | 2020-04-02 | 2020-08-25 | 清华大学 | Photoelectric hybrid intelligent data generation and calculation system and method |
| CN112639817A (en) * | 2018-06-20 | 2021-04-09 | 特斯拉公司 | Data pipeline and deep learning system for autonomous driving |
| CN112926159A (en) * | 2021-03-16 | 2021-06-08 | 中国科学院声学研究所 | Second-order Helmholtz resonator design method based on deep learning |
| WO2022195754A1 (en) * | 2021-03-17 | 2022-09-22 | 株式会社エビデント | Data processing method, data processing device, three-dimensional observation device, learning method, learning device, and recording medium |
| CN116310719A (en) * | 2023-02-10 | 2023-06-23 | 中国人民解放军军事科学院国防科技创新研究院 | Time-frequency domain-based optical diffraction complex model training method and image processing method |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0943539A (en) * | 1995-07-28 | 1997-02-14 | Konica Corp | Three-dimensional image method and its device |
| CN105070293A (en) * | 2015-08-31 | 2015-11-18 | 武汉大学 | Audio bandwidth extension coding and decoding method and device based on deep neutral network |
| CN105976408A (en) * | 2016-04-28 | 2016-09-28 | 北京大学 | Digital holographic compression transmission method of quantum backward propagation nerve network |
| CN106124449A (en) * | 2016-06-07 | 2016-11-16 | 中国科学院合肥物质科学研究院 | A kind of soil near-infrared spectrum analysis Forecasting Methodology based on degree of depth learning art |
| CN106770287A (en) * | 2016-12-07 | 2017-05-31 | 广东工业大学 | A kind of one camera balanced type optical coherence tomography scanning means and method |
-
2017
- 2017-12-19 CN CN201711372608.7A patent/CN107958475B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH0943539A (en) * | 1995-07-28 | 1997-02-14 | Konica Corp | Three-dimensional image method and its device |
| CN105070293A (en) * | 2015-08-31 | 2015-11-18 | 武汉大学 | Audio bandwidth extension coding and decoding method and device based on deep neutral network |
| CN105976408A (en) * | 2016-04-28 | 2016-09-28 | 北京大学 | Digital holographic compression transmission method of quantum backward propagation nerve network |
| CN106124449A (en) * | 2016-06-07 | 2016-11-16 | 中国科学院合肥物质科学研究院 | A kind of soil near-infrared spectrum analysis Forecasting Methodology based on degree of depth learning art |
| CN106770287A (en) * | 2016-12-07 | 2017-05-31 | 广东工业大学 | A kind of one camera balanced type optical coherence tomography scanning means and method |
Non-Patent Citations (2)
| Title |
|---|
| 潘哲朗: ""微小生物体的数字全息显微层析成像技术研究"", 《中国博士学位论文全文数据库（电子期刊）基础科学辑》 * |
| 马利红等: ""高精度数字全息显微衍射层析成像"", 《光子学报》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112639817A (en) * | 2018-06-20 | 2021-04-09 | 特斯拉公司 | Data pipeline and deep learning system for autonomous driving |
| CN111199574A (en) * | 2018-11-16 | 2020-05-26 | 青岛海信激光显示股份有限公司 | Holographic image generation method and equipment |
| CN109932818B (en) * | 2019-05-07 | 2020-03-20 | 东北大学 | Method for carrying out nondegenerate mode power decomposition on optical fiber with unknown structural parameters |
| CN109932818A (en) * | 2019-05-07 | 2019-06-25 | 东北大学 | The optical fiber of a kind of pair of unknown structure parameter carries out nondegenerate mode power decomposition method |
| US11600060B2 (en) | 2019-06-05 | 2023-03-07 | Tsinghua University | Nonlinear all-optical deep-learning system and method with multistage space-frequency domain modulation |
| CN110309916A (en) * | 2019-06-05 | 2019-10-08 | 清华大学 | The full optical depth learning system of multistage null tone domain Modulation and Nonlinear and method |
| CN110309916B (en) * | 2019-06-05 | 2021-09-17 | 清华大学 | Multi-stage space-frequency domain modulation nonlinear all-optical deep learning system and method |
| CN110243827A (en) * | 2019-07-18 | 2019-09-17 | 华中科技大学 | A kind of rapid three dimensional imaging process suitable for light transparent sample |
| CN110378473A (en) * | 2019-07-26 | 2019-10-25 | 清华大学 | Method and device is chromatographed based on deep learning and the phase of random pattern |
| CN110378473B (en) * | 2019-07-26 | 2021-12-21 | 清华大学 | Phase chromatography method and device based on deep learning and random pattern |
| CN111582468A (en) * | 2020-04-02 | 2020-08-25 | 清华大学 | Photoelectric hybrid intelligent data generation and calculation system and method |
| CN111582468B (en) * | 2020-04-02 | 2022-08-09 | 清华大学 | Photoelectric hybrid intelligent data generation and calculation system and method |
| CN112926159A (en) * | 2021-03-16 | 2021-06-08 | 中国科学院声学研究所 | Second-order Helmholtz resonator design method based on deep learning |
| WO2022195754A1 (en) * | 2021-03-17 | 2022-09-22 | 株式会社エビデント | Data processing method, data processing device, three-dimensional observation device, learning method, learning device, and recording medium |
| CN116310719A (en) * | 2023-02-10 | 2023-06-23 | 中国人民解放军军事科学院国防科技创新研究院 | Time-frequency domain-based optical diffraction complex model training method and image processing method |
| CN116310719B (en) * | 2023-02-10 | 2024-04-19 | 中国人民解放军军事科学院国防科技创新研究院 | Time-frequency domain-based optical diffraction complex model training method and image processing method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107958475B (en) | 2020-06-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107958475A (en) | Varied angle illumination based on deep learning generation network chromatographs method and device | |
| CN106373167B (en) | A kind of compression sensing magnetic resonance imaging method employing based on deep neural network | |
| CN109300166A (en) | The method and apparatus and storage medium of CT image reconstruction | |
| CN106796716B (en) | For providing the device and method of super-resolution for low-resolution image | |
| CN104834015B (en) | The four-dimensional measuring method of pulsating flame chromatographed based on light deflection | |
| CN109962688A (en) | The quick predict and reverse geometry design method of all dielectric Meta Materials filter transfer characteristic based on deep learning neural network | |
| CN108389226A (en) | A kind of unsupervised depth prediction approach based on convolutional neural networks and binocular parallax | |
| CN106204447A (en) | The super resolution ratio reconstruction method with convolutional neural networks is divided based on total variance | |
| CN104331579A (en) | Simulation method of low-permeability reservoir crude oil boundary layer | |
| CN106067161A (en) | A kind of method that image is carried out super-resolution | |
| CN106097441A (en) | Compound regularization Bioluminescence tomography reconstruction method based on L1 norm Yu TV norm | |
| CN105717145B (en) | Multi-joint dress 3 D cone-beam computer tomography method and device | |
| CN109117944A (en) | A kind of super resolution ratio reconstruction method and system of steamer target remote sensing image | |
| CN110415307A (en) | A kind of multipotency CT imaging method based on tensor completion, device and its storage equipment | |
| CN1916923A (en) | Computer simulation method for visualized information of substance metabolism functions inside human body | |
| CN102393969B (en) | Optical three-dimensional imaging method based on biological tissue specificity | |
| CN104851080A (en) | TV-based 3D positron emission tomography (PET) image reconstruction method | |
| CN109655890B (en) | Depth domain shallow-medium-deep layer combined chromatography inversion speed modeling method and system | |
| Sun et al. | A reconstruction method for cone-beam computed laminography based on projection transformation | |
| CN107991854A (en) | Solid artificial holography object reproducting method based on chromatography | |
| CN110084809A (en) | Diabetic retinopathy data processing method, device and electronic equipment | |
| CN104504735B (en) | A kind of material distribution detection method based on phase recovery | |
| Hu et al. | Nonlinear optimization for mixed attenuation polyenergetic image reconstruction | |
| CN109358481A (en) | A kind of multiplexing body holographic lens reproducting method | |
| CN106503302A (en) | A kind of method and device for setting up blood electromagnetic simulation model |
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 |