CN115078305A - Zernike coefficient wavefront prediction algorithm based on graph neural network - Google Patents

Zernike coefficient wavefront prediction algorithm based on graph neural network Download PDF

Info

Publication number
CN115078305A
CN115078305A CN202210616475.8A CN202210616475A CN115078305A CN 115078305 A CN115078305 A CN 115078305A CN 202210616475 A CN202210616475 A CN 202210616475A CN 115078305 A CN115078305 A CN 115078305A
Authority
CN
China
Prior art keywords
zernike coefficient
neural network
atmospheric turbulence
zernike
network model
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Pending
Application number
CN202210616475.8A
Other languages
Chinese (zh)
Inventor
吴计
杨宗翰
邸江磊
赵建林
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Northwestern Polytechnical University
Original Assignee
Northwestern Polytechnical University
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Northwestern Polytechnical University filed Critical Northwestern Polytechnical University
Priority to CN202210616475.8A priority Critical patent/CN115078305A/en
Publication of CN115078305A publication Critical patent/CN115078305A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/41—Refractivity; Phase-affecting properties, e.g. optical path length
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/04—Architecture, e.g. interconnection topology
    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
    • G06N3/00—Computing arrangements based on biological models
    • G06N3/02—Neural networks
    • G06N3/08—Learning methods
    • G—PHYSICS
    • G01—MEASURING; TESTING
    • G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
    • G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
    • G01N21/17—Systems in which incident light is modified in accordance with the properties of the material investigated
    • G01N21/41—Refractivity; Phase-affecting properties, e.g. optical path length
    • G01N2021/4106—Atmospheric distortion; Turbulence
    • Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
    • Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
    • Y02A—TECHNOLOGIES FOR ADAPTATION TO CLIMATE CHANGE
    • Y02A90/00—Technologies having an indirect contribution to adaptation to climate change
    • Y02A90/10—Information and communication technologies [ICT] supporting adaptation to climate change, e.g. for weather forecasting or climate simulation

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Health & Medical Sciences (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • General Health & Medical Sciences (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Artificial Intelligence (AREA)
  • Computing Systems (AREA)
  • Biophysics (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • Molecular Biology (AREA)
  • Analytical Chemistry (AREA)
  • Biochemistry (AREA)
  • Immunology (AREA)
  • Pathology (AREA)
  • Chemical & Material Sciences (AREA)
  • Optical Communication System (AREA)

Abstract

The invention discloses a Zernike coefficient wavefront prediction algorithm based on a graph neural network, which is characterized in that a time sequence Zernike data set is established and the network is trained by constructing a graph neural network model, so that the Zernike coefficient at the correction moment of a self-adaptive optical system is accurately and efficiently predicted while the calculated amount is reduced. The method does not need to establish a turbulence time-varying mathematical physical model, is quick in calculation and free from iteration, reduces the influence of the environment on the device, and improves the turbulence correction capability of the self-adaptive optical system in extremely weak illumination and extremely far deep space.

Description

Zernike coefficient wavefront prediction algorithm based on graph neural network
Technical Field
The invention belongs to the field of adaptive optics, and relates to a prediction adaptive optics technology.
Background
For the current adaptive optics system, the time delay of 2-3 sampling periods exists between the wave front sensing and the wave front correction. Predictive adaptive optics is an adaptive optics technique that incorporates wavefront prediction. In the whole prediction adaptive optical system, a wavefront sensor detects turbulence distortion information in a period of time, and a wavefront prediction algorithm predicts turbulence distortion wavefronts after 2-3 sampling periods in the future by using the past wavefront information. The predicted wavefront is loaded on a deformable mirror to correct for atmospheric turbulence. Existing wavefront prediction algorithms are mainly classified into 3 classes. The first type is a wave front prediction method based on a linear operator, the method does not consider the space-time coupling characteristic of turbulent evolution, and the prediction precision is limited. The second category is based on control algorithms, which do not exploit the physical characteristics of turbulent evolution and require repeated iterations. The third type is based on a deep learning algorithm, and the method does not take the characteristics of an object subjected to turbulent evolution into consideration, so that the calculation amount is large, and the real-time property of the self-adaptive optical system is influenced.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a Zernike coefficient wavefront prediction algorithm based on a graph neural network. The wavefront prediction algorithm predicts the Zernike coefficients at the wavefront correction time by using the Zernike coefficients in a period of time. Compared with the existing deep learning wavefront prediction method, the method has the advantages of small calculated amount and high prediction precision, provides a new technical route for predicting the self-adaptive optical system, and is favorable for ensuring the detection effectiveness of the self-adaptive optical system in an extremely long distance and an extremely atmospheric turbulence environment.
The technical scheme adopted by the invention for solving the technical problem comprises the following steps:
s1, acquiring Zernike coefficient data of atmospheric turbulence, wherein the data is marked as Zi, i is 1,2,3,4 … T, and T is the maximum frame number of the acquired time sequence Zernike coefficient; the time intervals between two adjacent Zernike coefficients are equal;
s2, processing the T-frame atmospheric turbulence Zernike coefficients into N parts of isometric atmospheric turbulence Zernike coefficient data, wherein each part of data comprises an atmospheric turbulence Zernike coefficient sequence ZN ═ { Zt-N +1, …, Zt } and an atmospheric turbulence Zernike coefficient Zt + r at the correction moment, N is the length constant of the input atmospheric turbulence Zernike coefficient sequence, T represents the moment of the atmospheric turbulence Zernike coefficients, and r is the time delay between turbulence detection and turbulence correction;
s3, establishing a neural network model, initializing network model parameters, taking an atmospheric turbulence Zernike coefficient sequence ZN as the input of the network, taking an atmospheric turbulence Zernike coefficient Zt + r as the gold standard of the network, calculating a loss function value of the output of the network and the gold standard, reversely propagating the gradient of the loss function value to update the parameters of the network model until the loss function value is smaller than a set condition, and stopping updating the network model parameters to obtain a completely Trained neural network Trained _ GPnet;
s4, actually measuring a section of atmospheric turbulence Zernike coefficient sequence Z 'N with the length of N as { Z'1, …, Z 'N } as input of the Trained _ GPnet, and estimating to obtain an atmospheric turbulence Zernike coefficient Z' N + r at the correction moment.
The neural network model adopts a graph neural network model; the graph neural network model is any graph neural network model with a graph learning layer and a graph convolution layer; the graph learning layer is used for learning a adjacency matrix of Zernike coefficients.
And the maximum value of the data N is rounded down after T-N + 1.
The length constant n of the input atmospheric turbulence Zernike coefficient sequence is more than 10.
The loss function is an MSE loss function, and the set condition that the updating of the corresponding network model parameters is stopped is that the MSE loss function value is less than 0.002; the loss function is MSE loss function, and the set condition that the updating of the corresponding network model parameter is stopped is that the MSE loss function value is less than 0.002.
The invention has the beneficial effects that: the Zernike coefficients at the correction time can be predicted only by a group of Zernike coefficients in a known time period, and the prediction is accurate and rapid. The invention does not need to use complex turbulence information, can greatly reduce the calculated amount of the prediction adaptive optical algorithm by utilizing the Zernike coefficient prediction, reduces the time cost of a hardware system, and meets the real-time requirement of the adaptive optical system. The neural network of the graph adopted by the invention has a simple structure, and can automatically learn the adjacency matrix of the Zernike coefficients.
Drawings
FIG. 1 is a flowchart of a method of example 1 of the present invention;
in the figure, the solid line part is a training stage, and the dashed line part is a testing stage;
fig. 2 is a structural diagram of a graph neural network (GPnet) used in embodiment 1;
FIG. 3 is an optical path diagram for data acquisition of a Zernike coefficient wavefront prediction algorithm based on a graph neural network in example 1;
in the figure, 1-laser, 2-collimating lens, 3-atmospheric turbulence pool, 4-imaging lens, 5-Shack-Hartmann wavefront sensor;
Detailed Description
The present invention will be further described with reference to the following drawings and examples, which include, but are not limited to, the following examples.
The embodiment 1 of the invention provides an atmospheric turbulence phase space-time estimation algorithm based on deep learning, which comprises two stages of training and estimation as shown in fig. 1, and comprises the following steps:
a. the training phase comprises the following steps:
s1, acquiring Zernike coefficient data of atmospheric turbulence, wherein the data is marked as Zi, i is 1,2,3,4 … T, and T is the maximum frame number of the acquired time sequence Zernike coefficient; the time intervals between two adjacent Zernike coefficients are equal;
s2, processing the T-frame atmospheric turbulence Zernike coefficients into N parts of isometric atmospheric turbulence Zernike coefficient data, wherein each part of data comprises an atmospheric turbulence Zernike coefficient sequence ZN ═ { Zt-N +1, …, Zt } and an atmospheric turbulence Zernike coefficient Zt + r at the correction moment, N is the length constant of the input atmospheric turbulence Zernike coefficient sequence, T represents the moment of the atmospheric turbulence Zernike coefficients, and r is the time delay between turbulence detection and turbulence correction;
s3, establishing a neural network model, initializing network model parameters, taking an atmospheric turbulence Zernike coefficient sequence ZN as the input of the network, taking an atmospheric turbulence Zernike coefficient Zt + r as the gold standard of the network, calculating a loss function value of the output of the network and the gold standard, reversely propagating the gradient of the loss function value to update the parameters of the network model until the loss function value is smaller than a set condition, and stopping updating the network model parameters to obtain a completely Trained neural network Trained _ GPnet;
b. the estimation stage comprises the following steps:
s4, actually measuring a section of atmospheric turbulence Zernike coefficient sequence Z 'N with the length of N as { Z'1, …, Z 'N } as input of the Trained _ GPnet, and estimating to obtain an atmospheric turbulence Zernike coefficient Z' N + r at the correction moment.
The neural network model adopts a graph neural network model; the graph neural network model is any graph neural network model with a graph learning layer and a graph convolution layer; the graph learning layer is used for learning a adjacency matrix of Zernike coefficients.
And the maximum value of the data N is rounded down after T-N + 1.
The length constant n of the input atmospheric turbulence Zernike coefficient sequence is more than 10.
The loss function is an MSE loss function, and the set condition that the updating of the corresponding network model parameters is stopped is that the MSE loss function value is less than 0.002; the loss function is MSE loss function, and the set condition that the updating of the corresponding network model parameter is stopped is that the MSE loss function value is less than 0.002.
Embodiment 1 of the present invention employs a phase measurement optical path as shown in figure 3, comprising a laser 1, a collimating lens 2, an atmospheric turbulence cell 3, an imaging lens 4 and a Shack-Hartmann wavefront sensor 5.
The training phase of example 1 of the present invention, using the optical path shown in figure 3, begins with the acquisition of a wavefront deviation lattice Bi over a period of time from an atmospheric turbulence pool 4 using a Shack-Hartmann wavefront sensor 5, where i is 1,2,3,4 … T. And calculating the Zernike coefficient Zi of the corresponding atmospheric turbulence according to the relation between the wave front slope and the Zernik coefficient, wherein i is 1,2,3,4 … T. Processing T frame atmospheric turbulence Zernike coefficients into N equal-length time sequence atmospheric turbulence Zernike coefficient data, wherein each data comprises an atmospheric turbulence Zernike coefficient sequence ZN ═ { Zt-N +1, …, Zt } of a known time period and an atmospheric turbulence Zernike coefficient sequence Zt + r of a correction time period, wherein N is an input Zernike coefficient sequence length constant, m is an output Zernike coefficient sequence length constant, r is time delay between turbulence detection and turbulence correction, T ═ 1,2,3,4 … N, and in the N data, 70% is a training set and 30% is a testing set. Establishing a neural network model as shown in fig. 2, initializing network model parameters, taking n atmospheric turbulence Zernike coefficient sequences Zn which are continuously changed as the input of the network, taking the atmospheric turbulence Zernike coefficients Zt + r as the gold standard of the network, calculating the MSE loss function value of the output of the network and the gold standard, and reversely propagating the gradient of the loss function value to update the parameters of the network model until the loss function value is less than 0.002, stopping updating the network model parameters, and obtaining a completely Trained neural network Trained _ GPnet, wherein the network training parameters are as follows: the learning rate is 0.0003, the Batch size is 32, and the Epoch is 120. The training stage is only needed to be executed once, and the trained network model can be used for deployment and used for atmospheric turbulence phase space-time estimation.
In the test stage of embodiment 1 of the present invention, the atmospheric turbulence phase map sequence Z 'N ═ { Z'1, …, Z 'N } obtained by using the optical path and dot matrix demodulation algorithm shown in fig. 3 is used as an input of the Trained _ GPnet, and the atmospheric turbulence phase map Z' N + r at the predicted correction time is obtained.

Claims (5)

1. A Zernike coefficient wavefront prediction algorithm based on a graph neural network is characterized by comprising the following steps of:
s1, acquiring Zernike coefficient data of atmospheric turbulence, wherein the data is marked as Zi, i is 1,2,3,4 … T, and T is the maximum frame number of the acquired time sequence Zernike coefficient; the time intervals between two adjacent Zernike coefficients are equal;
s2, processing the T-frame atmospheric turbulence Zernike coefficients into N parts of isometric atmospheric turbulence Zernike coefficient data, wherein each part of data comprises an atmospheric turbulence Zernike coefficient sequence ZN ═ { Zt-N +1, …, Zt } and an atmospheric turbulence Zernike coefficient Zt + r at the correction moment, N is the length constant of the input atmospheric turbulence Zernike coefficient sequence, T represents the moment of the atmospheric turbulence Zernike coefficients, and r is the time delay between turbulence detection and turbulence correction;
s3, establishing a neural network model, initializing network model parameters, taking an atmospheric turbulence Zernike coefficient sequence ZN as the input of the network, taking an atmospheric turbulence Zernike coefficient Zt + r as the gold standard of the network, calculating a loss function value of the output of the network and the gold standard, reversely propagating the gradient of the loss function value to update the parameters of the network model until the loss function value is smaller than a set condition, and stopping updating the network model parameters to obtain a completely Trained neural network Trained _ GPnet;
s4, actually measuring a section of atmospheric turbulence Zernike coefficient sequence Z 'N with the length of N as { Z'1, …, Z 'N } as input of the Trained _ GPnet, and estimating to obtain an atmospheric turbulence Zernike coefficient Z' N + r at the correction moment.
2. A Zernike coefficient wavefront prediction algorithm based on a graph neural network as claimed in claim 1, characterized in that the neural network model adopts a graph neural network model; the graph neural network model is any graph neural network model with a graph learning layer and a graph convolution layer; the graph learning layer is used for learning a adjacency matrix of Zernike coefficients.
3. A Zernike coefficient wavefront prediction algorithm based on a graph neural network as claimed in claim 1, characterized in that the maximum value of the data N is rounded down after T-N + 1.
4. A figure neural network-based Zernike coefficient wavefront prediction algorithm as claimed in claim 1, characterized in that the input atmospheric turbulence Zernike coefficient sequence length constant n is larger than 10.
5. A Zernike coefficient wavefront prediction algorithm based on graph neural network as claimed in claim 1, characterized in that the loss function is MSE loss function, and the setting condition for stopping updating of the corresponding network model parameter is that the MSE loss function value is less than 0.002; the loss function is MSE loss function, and the set condition that the updating of the corresponding network model parameter is stopped is that the MSE loss function value is less than 0.002.
CN202210616475.8A 2022-06-01 2022-06-01 Zernike coefficient wavefront prediction algorithm based on graph neural network Pending CN115078305A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202210616475.8A CN115078305A (en) 2022-06-01 2022-06-01 Zernike coefficient wavefront prediction algorithm based on graph neural network

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202210616475.8A CN115078305A (en) 2022-06-01 2022-06-01 Zernike coefficient wavefront prediction algorithm based on graph neural network

Publications (1)

Publication Number Publication Date
CN115078305A true CN115078305A (en) 2022-09-20

Family

ID=83248401

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202210616475.8A Pending CN115078305A (en) 2022-06-01 2022-06-01 Zernike coefficient wavefront prediction algorithm based on graph neural network

Country Status (1)

Country Link
CN (1) CN115078305A (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115374712A (en) * 2022-10-24 2022-11-22 中国航天三江集团有限公司 Method and device for calibrating optical transmission simulation parameters under influence of laser internal channel thermal effect

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN115374712A (en) * 2022-10-24 2022-11-22 中国航天三江集团有限公司 Method and device for calibrating optical transmission simulation parameters under influence of laser internal channel thermal effect

Similar Documents

Publication Publication Date Title
AU2020103905A4 (en) Unsupervised cross-domain self-adaptive medical image segmentation method based on deep adversarial learning
CN107765347B (en) Short-term wind speed prediction method based on Gaussian process regression and particle filtering
CN111130839A (en) Flow demand matrix prediction method and system
US6801810B1 (en) Method and device for state estimation
CN108764568B (en) Data prediction model tuning method and device based on LSTM network
KR101898629B1 (en) Method and device for post-correction of forecasted meteorological parameters by using a H-Infinity filter
CN111931983B (en) Precipitation prediction method and system
CN112464567B (en) Intelligent data assimilation method based on variational and assimilative framework
CN115078305A (en) Zernike coefficient wavefront prediction algorithm based on graph neural network
CN116432543B (en) Method for predicting remaining life of power semiconductor module, terminal device and storage medium
CN114330647A (en) Model training method and device and silicon rod weight prediction method
EP3206094B1 (en) Predicted value modification system, control system, predicted value modification method, control method, and predicted value modification program
CN112990585A (en) Hen laying rate prediction method based on LSTM-Kalman model
CN114626307B (en) Distributed consistent target state estimation method based on variational Bayes
CN115308558A (en) Method and device for predicting service life of CMOS (complementary Metal oxide semiconductor) device, electronic equipment and medium
CN113988415B (en) Medium-and-long-term power load prediction method
CN114964313A (en) RVM-based fiber optic gyroscope temperature compensation method
Van Gorp et al. Learning neural networks with noisy inputs using the errors-in-variables approach
CN114239796A (en) Power system state estimation method based on extended Kalman filtering
CN110986946B (en) Dynamic pose estimation method and device
CN116910534A (en) Space-time intelligent prediction method and device for ocean environmental elements in different sea areas
CN115936236A (en) Method, system, equipment and medium for predicting energy consumption of cigarette factory
CN116048956A (en) Software defect occurrence prediction method, device, equipment and storage medium
CN113239614A (en) Atmospheric turbulence phase space-time prediction algorithm
CN112818455A (en) Bridge structure response monitoring method and system

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