AU2020102396A4

AU2020102396A4 - Machine learning-based method for analyzing characteristics of laser beam propagation through turbulent atmosphere

Info

Publication number: AU2020102396A4
Application number: AU2020102396A
Authority: AU
Inventors: Chunyi Chen; Xin Feng; Zhengang JIANG; Shoufeng TONG; Huamin YANG
Original assignee: Changchun University of Science and Technology
Current assignee: Changchun University of Science and Technology
Priority date: 2020-09-23
Filing date: 2020-09-23
Publication date: 2020-11-05
Anticipated expiration: 2028-09-23

Abstract

The disclosure discloses a machine learning-based method for analyzing characteristics of laser beam propagation through a turbulent atmosphere. The method includes: firstly, performing a measuring operation for the normalized variance of received optical signal intensity fluctuations under different external propagation conditions and creating a training sample set for a multi-layer feedforward neural network; secondly, training the multi-layer feedforward neural network by using the created training sample set to approximate a functional relation between the normalized variance of received optical signal intensity fluctuations and external propagation conditions; and finally, obtaining the normalized variance of received optical signal intensity fluctuations under particular external propagation conditions using the multi-layer feedforward neural network. Since dimensionless parameters are used as inputs to the multi-layer feedforward neural network in the method, the multi-layer feedforward neural network built according to the method can be used to predict the normalized variance of received optical signal intensity fluctuations under those external propagation conditions that are not applied for experimental measuring. DRAWINGS 101 102 End-------------- ---- ---------------- Ed 201 202 End A 17End B urbulent Atmosphere Channel 203 204 FIG.1I

Description

DRAWINGS

101 102

End-------------- ---- ---------------- Ed

201 202

End A 17End B

urbulent Atmosphere Channel 203 204 FIG.1I

Editorial Note 2020102396 There is only ten pages of the description

MACHINE LEARNING-BASED METHOD FOR ANALYZING CHARACTERISTICS OF LASER BEAM PROPAGATION THROUGH TURBULENT ATMOSPHERE TECHNICAL FIELD The disclosure belongs to the technical field of optical signal propagation over atmospheric channels, and relates to a machine leaming-based method for analyzing characteristics of laser beam propagation through a turbulent atmosphere. BACKGROUND A laser beam propagating through a turbulent atmosphere usually experience scintillations, which may lead to random fluctuations in received optical signal intensity at an optical receiver. The normalized variance of received optical signal intensity fluctuations is an important indicator for the measurement of the effect of atmospheric turbulence on laser signal propagation. The

= P2 / P normalized variance is defined as " , where P represents the received optical signal intensity, and the angle brackets represent an ensemble average thereof. Scholars have done a great number of analytical and theoretical studies on a normalized variance of received optical signal intensity fluctuations of a laser beam after propagation through a turbulent atmosphere. However, the experimental measurement results can reflect the statistical characteristics of received optical signal intensity fluctuations on a turbulent atmosphere channel, with each measurement result corresponding to a specific external propagation condition. It is a challenge to estimate, based on the existing experimental measurement results, a normalized variance of received optical signal intensity fluctuations under other external propagation conditions which are not applied for measuring. The disclosure will provide a method to solve the technical problem. According to Laser beam propagation through random media, 2nd edition, published by SPIE Publications in 2005, regarding Gaussian beam propagation, a Gaussian beam in an emitting plane is typically depicted by two dimensionless parameters, namely Fresnel Ratio Ao and Curvature Parameter 00, where Ao=2L/(kWo 2), and 80=1-L/Fo, with L representing a propagation length, k representing an optical wave number, Wo representing the radius of the Gaussian beam in the emitting plane, and F0 representing the wavefront radius of curvature of the Gaussian beam in the emitting plane. An actual optical receiver typically collects optical signals by means of a circular aperture of a particular size. Assuming that the radius of a receiving aperture is WG, a dimensionless receiving aperture-related Fresnel number DG=2L/(kWG)may be defined to depict a relative size of the receiving aperture. According to Modern Atmospheric Optics published by Science Press in 2012, the variance of angle-of-arrival fluctuations of a spherical wave received by a finite aperture after propagation through a turbulent atmosphere may be expressed as: 2 =5.675(2WG,) L fC2('L )jS3d

( where C(L ) represents an atmospheric refractive-index structure constant at a location with z=4L. According to Numerical Simulation of Optical Wave Propagation with Examples in MATLAB published by SPIE publications in 2010, the atmospheric coherence length of a spherical wave may be expressed as:

rol.= O.423k2L C ( L) 51d L JL (2)

The formula (1) and the formula (2) may be combined to give: -3/1 0. 09 3 k 2 2WC13

= (O.O939k 2 3) (3)

In the field of laser beam propagation through the turbulent atmosphere, the atmospheric coherence length of a spherical wave is often used as a parameter to depict the strength of atmospheric turbulence along an entire propagation path. Similar to the receiving aperture radius,

a dimensionless coherence length-related Fresnel number Dro=2L/(kr ) can also be defined to depict the strength of the atmospheric turbulence. The normalized variance of received optical signal intensity fluctuations of a laser beam after propagation through the turbulent atmosphere is dependent on a profile of atmospheric refractive-index structure constants along a path. In practice, it is difficult to accurately measure a profile of atmospheric refractive-index structure constants along a path. Chinese invention patent application No. 201811382166.9 uses three discrete random phase screens, which are approximately equivalent to the effects of continuous atmospheric turbulence along a propagation path on optical propagation; and each random phase 2 12

screen corresponds to an effective refractive-index structure constant, expressed as ,

62 and , respectively. A relative magnitude relation of the effective refractive-index structure

constants for the three random phase screens can be depicted by q and

,= ,1 defined based on . Actually, ' and '7 depict the general variations of the atmospheric refractive-index structure constants along the propagation path. Regarding Gaussian beam propagation through a turbulent atmosphere, external propagation

conditions can be depicted by Ao, 0, G, QrO, '7 and 7' . The normalized variance of received

optical signal intensity fluctuations is supposed to be a function of Ao, 0, G, Oro, '7 and ' , namely having a value within a six-dimensional space defined by Ao, 00, DG, ,o, '7 and q7 After numerous measurement results of the normalized variance of received optical signal intensity fluctuations under different conditions (corresponding to different positions in the six-dimensional space) are obtained, such measurement results and their respective external propagation conditions can be used as a training sample set to train a multi-layer feedforward neural network, allowing the multi-layer feedforward neural network to approximate the functional relation between the normalized variance of received optical signal intensity fluctuations and AO, 0 0, DG, 0o, '7 and q' . Multi-layer feedforward neural networks have been described in detail in Principles of Artificial Neural Networks and Simulation Examples published by China Machine Press in 2003. In the forgoing training sample set, each sample corresponds to an input vector [Ao, 00, DG, Qro,'7, ' ]and the desired output normalized variance of received optical signal intensity fluctuations. The desired output normalized variance of received optical signal intensity fluctuations corresponding to a particular sample may be obtained by measuring, and the values of the components of its corresponding input vector [Ao,

, DG, ro, '7 , ' ]correspond to the external propagation conditions where the measuring operation is carried out. When the measuring operation is carried out, the values of AO, 00 and DG

can be artificially controlled, and the values of 0o, '7 and '7 depend on actual conditions of

atmospheric turbulence. Generally, the values of 0o, '7 and q' corresponding to the same

propagation path may differ in different durations. During the analysis of the characteristics of laser beam propagation through the turbulent atmosphere, it is desirable to give a propagation length L, an optical wave number k, the initial radius WO of a Gaussian laser beam in an emitting plane, the initial wavefront radius FO of curvature of the Gaussian laser beam in the emitting plane, the radius WG of a reeiving aperture

and a profilecn(z) of atmospheric refractive-index structure constants along a propagation

path first. In the absence of the accurate form of a profilecn(z) of atmospheric refractive-index

structure constants along a propagation path, c can be usually assumed as a constant with regard to a horizontal propagation path, or a profile form can be assumed based on experience. C 2 "2

and can be calculated based on , specifically by solving the following set of equations:

1K =1 )5, (1 )/ 5, 1 Cj(JL)1 C d C+ C,+ + C

C 2L j )5/ 1 5 1 (L (1 d=- C - + - C C3 |

)5/6 )5/6 )5/6 5/6 )56 1F( 5 2

+ 3 k36) 4) 36)

SUMMARY The disclosure aims to provide a machine learning-based method for analyzing characteristics of laser beam propagation through a turbulent atmosphere, so that the normalized variance of received optical signal intensity fluctuations under other external propagation conditions which are not applied for measuring can be obtained according to existing experimental measurement results of the normalized variance of received optical signal intensity fluctuations. The technical solution of the method is implemented as follows. A machine learning-based method for analyzing characteristics of laser beam propagation through a turbulent atmosphere includes the following process. Firstly, a laser beam propagation path through the turbulent atmosphere is selected, and parameters including the radius of a Gaussian laser beam in an emitting plane, the wavefront radius of curvature of the Gaussian laser beam in the emitting plane and the radius of a receiving aperture are specified. Experimental measuring operation for the normalized variance of received optical signal intensity fluctuations is performed in different

durations. The values of a vector [Ao, 00, G, ,o,'7 ] corresponding to external propagation

conditions where the experimental measuring operation is performed each time and the

corresponding measured normalized variance a of received optical signal intensity fluctuations are recorded and added as a sample to a training sample set, with Ao representing a Fresnel ratio, 8o representing a curvature parameter, DG representing a reeiving aperture-related 77 Fresnel number , Qo representing a coherence length-related Fresnel number,

j/ , and , 12 and ,3 representing effective refractive-index structure constants for three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation, respectively. When the experimental measuring operation for the normalized variance of received optical signal intensity fluctuations is performed, the variations of the parameters including the radius of the Gaussian laser beam in the emitting plane, the wavefront radius of curvature of the Gaussian laser beam in the emitting plane and the radius of the receiving aperture are needed in addition to the change in duration, so that the measurement results of the normalized variance of received optical signal intensity fluctuations under different external propagation conditions are obtained. Each external propagation condition and the measurement result of the normalized variance of received optical signal intensity fluctuations under the condition are then added as a sample to the training sampling set. Secondly, a multi-layer feedforward neural network is trained with all the samples in the training sample set to approximate a functional relation between the normalized variance of received optical signal intensity fluctuations and AO, 0 0 , DG, 0ro, 7 and 7' . Thirdly, during the analysis of the characteristics of laser beam propagation through the turbulent atmosphere, the value of each component of the vector [Ao, 00, DG, r, ,' ] is calculated based on the parameters (i.e., the radius of the Gaussian laser beam in the emitting plane, the wavefront radius of curvature of the Gaussian laser beam in the emitting plane and the radius of the receiving aperture) of a system for laser beam propagation through a turbulent atmosphere to be analyzed and a considered profile of atmospheric refractive-index structure constants along the path, and the vector [Ao, 00, DG, r, ,' ] is input to the multi-layer feedforward neural network which then produces an output, i.e., the resulting normalized variance of received optical signal intensity fluctuations from the analysis. As shown in FIG. 1, a system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere is arranged in parallel to a system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere. The system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere includes an emitting-receiving device A (101) and an emitting-receiving device B (102). The system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere includes a laser emitting device (201), an optical receiving subsystem (202), a photoelectric detector (203) and a computer (204). A laser beam emitted by the laser emitting device (201) propagates on an turbulent atmosphere channel to a receiving aperture of the optical receiving subsystem (202); an optical signal received by the receiving subsystem (202) is incident on the photoelectric detector (203), and a signal output by the photoelectric detector (203) is transmitted to a signal acquisition card of the computer (204). The effective refractive-index structure constants for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation can be calculated based on the measurement data obtained by the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere, and then the values of the external propagation condition parameters ' and

' can be obtained. The system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere is configured to measure the normalized variance of received optical signal intensity fluctuations. The first part of the method involves performing the measuring operation for the normalized variance of received optical signal intensity fluctuations and creating the training sample set for the multi-layer feedforward neural network, which specifically includes the following steps: Step 101: selecting a propagation path PATH1 having a length of L through the turbulent atmosphere with the laser emitting device (201) located at end A of the path PATHI and the optical receiving subsystem (202) located at end B of the path PATH1; emitting a Gaussian laser beam having an optical wave number k by the laser emitting device (201); translating the path PATH1 a distance Ds in a direction perpendicular to the path PATH1 to obtain a path PATH2 with the emitting-receiving device A (101) located at end A of the path PATH2 and the emitting-receiving device B (102) located at end B of the path PATH2; and emptying the training sample set; Step 102: let ni=0, n2=0, and n3=0; and performing the following operation: Step 102-1: let x=-bi+ni 1 , x 2=b2+n26 2, and x3=-b3+n 33, with bi being a positive real number, b 2 being a real number less than 1, b 3 being a positive real number, 1 =2bi/N1

, 62=2(1-b 2)/N2 , and 63=2b /N 3 3 ; calculating A0= X2 , and DG=103 ; calculating

Wo=[2L/(kAo)]1/2, Fo=L/(1-8O), and WG=[2L/(kG) 1/2; and adjusting the radius of the Gaussian laser beam emitted by the laser emitting device (201) to Wo, the wavefront radius of curvature of the Gaussian laser beam emitted by the laser emitting device (201) to Fo and the radius of the circular receiving aperture of the optical receiving subsystem (202) to WG; Step 102-2: performing the operation from Step 102-2-1 to Step 102-2-3 at intervals of H hours over a duration DUR: Step 102-2-1: enabling the system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere and the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere to work normally; Step 102-2-2: sampling, by the computer (204) and using a signal acquisition card, a signal AOO1 output by the photoelectric detector (203) at a frequency off Hz, thereby obtaining NUM

sampled values A002 of the signal A001, and calculating the normalized variance " of the

NUM sampled values A002 of the signal A001; meanwhile, measuring the effective

refractive-index structure constants , and for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation by using the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere in combination with a method for approximately measuring the profile of atmospheric refractive-index structure constants along the path, recording the measured

variance 'a7B of angle-of-arrival fluctuations of an optical wave propagating from the emitting-receiving device A (101) to the emitting-receiving device B (102), and calculating

n,3 n4l ,22,and Qro=2L/(kro"sw);and Step 102-2-3: creating a new sample A003, supposing an input vector corresponding to the

new sample A003 to be [Ao, 0 0, ® G, ro, 77 , ] and the desired output normalized variance of

received optical signal intensity fluctuations corresponding to the new sample A003 to

normalized variance " , and adding the new sample A003 to the training sample set;

Step 102-3: if ni is less than Ni, let ni=ni+1, and skipping to Step 102-1, or let ni=0, and skipping to Step 102-4; Step 102-4: if n2 is less than N2, let n2=n2+1, and skipping to Step 102-1, or let n2=0, and skipping to Step 102-5; and Step 102-5: if n3 is less than N 3 , let n3=n3+1, and skipping to Step 102-1, or skipping to Step 103; and Step 103: completing the creation of the training sample set. The second part of the method involves using the training sample set created in the first part of the method to train a multi-layer feedforward neural network, allowing the multi-layer feedforward neural network to approximate the functional relation between the normalized variance of received optical signal intensity fluctuations and Ao, 00, G, Oro, '7 and ' , which specifically includes the following steps: Step 201: building a multi-layer feedforward neural network having six input nodes at an input layer thereof and one output node at an output layer thereof; and Step 202: realizing the training of the multi-layer feedforward neural network with the values

of six components of the input vector [Ao, 00, DG, Qro, 7 , "' ]corresponding to each sample in

the training sample set as respective inputs to the six input nodes of the multi-layer feedforward neural network and the desired output normalized variance of received optical signal intensity fluctuations corresponding to each sample as the desired output of the multi-layer feedforward neural network, allowing the multi-layer feedforward neural network to approximate the functional relation between the normalized variance of received optical signal intensity fluctuations and A0 o, 00, G, ro, '7 and'I'

. The third part of the method involves obtaining the normalized variance of received optical signal intensity fluctuations under particular external propagation conditions using the resulting multi-layer feedforward neural network from the second part of the method, which specifically includes the following steps: Step 301: specifying a propagation length L, an optical wave number k, the radius Wo of a Gaussian laser beam in an emitting plane, the wavefront radius FO of curvature of the Gaussian laser beam in the emitting plane, the radius WG of a reeiving aperture, andcalculating 22 C2 (Z Ao=2L/(kWo ),2 Qo=1-L/Fo, and G=2L/(kWG2); specifying a profile C(z) of atmospheric refractive-index structure constants along the propagation path, with z representing a distance away from the laser emitting device in the propagation path, and calculating the atmospheric coherence length of a spherical wave:

ro. = 0O.423k2 LfI C 2( L 53d |3/ r2 L ;calculating Qr0=2L/(k o"); calculating

F(17 (17 (53

C =M 3J C2 (L)(1- ) dM ) ) )

C2 1 3 fC2 (Lj )(1 - ) d 3 3 L t3L L3J (L)( -S<d,j where 6 K 36 and

calculating 12 1 and , ;and

Step 302: combining Ao, 00, DG, ro, '7 and ' obtained in Step 301 into the input vector

[Ao, ®0,£G, QrO, '7, ' ]to the resulting multi-layer feedforward neural network from the

second part of the method, and then producing, by the multi-layer feedforward neural network, an output, i.e., the corresponding normalized variance of received optical signal intensity fluctuations. The method has the advantage that the normalized variance of received optical signal intensity fluctuations under other external propagation conditions which are not applied for the measuring operation can be obtained readily according to existing measurement results of the normalized variance of received optical signal intensity fluctuations. Since dimensionless parameters are used as inputs to the multi-layer feedforward neural network in the method, the multi-layer feedforward neural network built according to the method can be used to predict the normalized variance of received optical signal intensity fluctuations under those external propagation conditions that are not applied for the measuring operation. For example, the multi-layer feedforward neural network built according to the method can be used to predict the normalized variance of received optical signal intensity fluctuations under propagation length conditions which are not applied for the measuring operation. BRIEF DESCRIPTION OF DRAWINGS FIG. 1 is a schematic diagram illustrating parallel arrangement of a system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere and a system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere. DETAILED DESCRIPTION In this example, effective refractive-index structure constants ' and C, 3 for three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along a propagation path on optical propagation are measured by the method disclosed in Chinese invention patent application No. 201811382166.9. The Chinese invention patent application No. 201811382166.9 provides a system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere, which can measure the variance of angle-of-arrival fluctuations of an optical wave. A Gaussian laser beam emitted by a laser emitting device (201) has a waveform of 800 nm, and a corresponding optical wave has a wave number k--2/. Photoelectric detector (203) is a PIN photoelectric detector; a distance Ds is 3 m; bi=3, b 2 =0, b 3=3, N 1=7, N2=5, N3=7, L=2 km, a duration DUR=5 days, H=1, f=100, and NUM=10000; and a multi-layer feedforward neural network includes two hidden layers. Path PATH1 and path PATH2 are parallel and not far from each other, and therefore, the effective refractive-index structure constants ', and C,3 for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path PATH1 on optical propagation and the '2 '

1 effective refractive-index structure constantsC , 1 2. and C 3 for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path PATH2 on optical propagation can be regarded as approximately the same. The effective refractive-index structure constants for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation can be calculated based on the measurement data obtained by the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere, and then the values of the external propagation condition parameters ' and q' can be obtained. The system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere is configured to measure the normalized variance of received optical signal intensity fluctuations.

Editorial Note 2020102396 There is only five pages of the claim

Claims

What is claimed is: 1. A machine leaming-based method for analyzing characteristics of laser beam propagation through a turbulent atmosphere, wherein firstly, a laser beam propagation path through the turbulent atmosphere is selected, and parameters comprising the radius of a Gaussian laser beam in an emitting plane, the wavefront radius of curvature of the Gaussian laser beam in the emitting plane and the radius of a receiving aperture are specified; experimental measuring operation for the normalized variance of received optical signal intensity fluctuations is performed in different

durations; the values of a vector [Ao, 00, G, 0ro, '7 , ] corresponding to external propagation

conditions where the experimental measuring operation is performed each time and the

corresponding measured normalized variance " of received optical signal intensity fluctuations are recorded and added as a sample to a training sample set, with Ao representing a Fresnel ratio, 80 representing a curvature parameter, DG representing a receiving aperture-related

Fresnel number , Qo representing a coherence length-related Fresnel number,

/ 62 c' ,'= 2 ,/ , and 621 ", 1 and "3 representing effective refractive-index structure constants for three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation, respectively; when the experimental measuring operation for the normalized variance of received optical signal intensity fluctuations is performed, the variations of the parameters comprising the radius of the Gaussian laser beam in the emitting plane, the wavefront radius of curvature of the Gaussian laser beam in the emitting plane and the radius of the receiving aperture are needed in addition to the change in duration, so that the measurement results of the normalized variance of received optical signal intensity fluctuations under different external propagation conditions are obtained; each external propagation condition and the measurement result of the normalized variance of received optical signal intensity fluctuations under the condition are then added as a sample to the training sampling set; secondly, a multi-layer feedforward neural network is trained with all the samples in the training sample set to approximate a functional relation between the normalized variance of

received optical signal intensity fluctuations and Ao, 00, G, ro, '7 and ' ; thirdly, during the

analysis of the characteristics of laser beam propagation through the turbulent atmosphere, the

value of each component of the vector [AO, 0 0 , Go, ,0,' , '' ] is calculated based on the

parameters (i.e., the radius of the Gaussian laser beam in the emitting plane, the wavefront radius of curvature of the Gaussian laser beam in the emitting plane and the radius of the receiving aperture) of a system for laser beam propagation through a turbulent atmosphere to be analyzed and a considered profile of atmospheric refractive-index structure constants along the path, and

the vector [Ao, ®0, G, ,ro, '7 , ' ] is input to the multi-layer feedforward neural network which then produces an output, i.e., the resulting normalized variance of received optical signal intensity fluctuations from the analysis; a system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere is arranged in parallel to a system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere; the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere includes an emitting-receiving device A and an emitting-receiving device B; the system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere includes a laser emitting device, an optical receiving subsystem, a photoelectric detector and a computer; a laser beam emitted by the laser emitting device propagates on an turbulent atmosphere channel to a receiving aperture of the optical receiving subsystem; an optical signal received by the receiving subsystem is incident on the photoelectric detector, and a signal output by the photoelectric detector is transmitted to a signal acquisition card of the computer; the effective refractive-index structure constants for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation can be calculated based on the measurement data obtained by the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere, and then the values of the external propagation condition parameters ' and '' can be obtained; the system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere is configured to measure the normalized variance of received optical signal intensity fluctuations; the first part of the method involves performing the measuring operation for the normalized variance of received optical signal intensity fluctuations and creating the training sample set for the multi-layer feedforward neural network, which specifically comprises the following steps: Step 101: selecting a propagation path PATH1 having a length of L through the turbulent atmosphere with the laser emitting device located at end A of the path PATH1 and the optical receiving subsystem located at end B of the path PATH1; emitting a Gaussian laser beam having an optical wave number k by the laser emitting device; translating the path PATH1 a distance Ds in a direction perpendicular to the path PATH1 to obtain a path PATH2 with the emitting-receiving device A located at end A of the path PATH2 and the emitting-receiving device B located at end B of the path PATH2; and emptying the training sample set;

Step 102: let ni=0, n2=0, and n3=0; and performing the following operation: Step 102-1: let x1 =-bi+nio1, x 2=b2+n26 2, and x3=-b3+n 3 3, with bi being a positive real number, b 2 being a real number less than 1, b 3 being a positive real number, 1 =2bi/N1

, 62=2(1-b 2)/N 2 , and o3 =2b 3/N3; calculating A = 10 , 0o = x, and 9G=10' ; calculating

Wo=[2L/(kAo)]1/2, Fo=L/(1-80), and WG=[2L/(kG)] 1; and adjusting the radius of the Gaussian laser beam emitted by the laser emitting device (201) to Wo, the wavefront radius of curvature of the Gaussian laser beam emitted by the laser emitting device (201) to Fo and the radius of the circular receiving aperture of the optical receiving subsystem (202) to WG Step 102-2: performing the operation from Step 102-2-1 to Step 102-2-3 at intervals of H hours over a duration DUR: Step 102-2-1: enabling the system for measuring the normalized variance of received optical signal intensity fluctuations of a laser beam propagating through a turbulent atmosphere and the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere to work normally; Step 102-2-2: sampling, by the computer and using a signal acquisition card, a signal AOO1 output by the photoelectric detector at a frequency of f Hz, thereby obtaining NUM sampled

values A002 of the signal A001, and calculating the normalized variance " of the NUM sampled values A002 of the signal A001; meanwhile, measuring the effective refractive-index "2 ~ c structure constants , and for the three random phase screens that are equivalent to the effects of continuous atmospheric turbulence along the propagation path on optical propagation by using the system for measuring the variance of intensity fluctuations and the variance of angle-of-arrival fluctuations of an optical wave bidirectionally propagating through a turbulent atmosphere in combination with a method for approximately measuring the profile of atmospheric refractive-index structure constants along the path, recording the measured variance

oAof angle-of-arrival fluctuations of an optical wave propagating from the emitting-receiving

' 2 ' 2

device A to the emitting-receiving device B, and calculating C 2 2 3 )-/ 2

2 '009 2~B , and Qro=2L/(kr");and Step 102-2-3: creating a new sample A003, supposing an input vector corresponding to the

new sample A003 to be [AO, 00, DG, Qro, 7 , '7 ]and the desired output normalized variance of received optical signal intensity fluctuations corresponding to the new sample A003 to

normalized variance " , and adding the new sample A003 to the training sample set;

Step 102-3: if ni is less than Ni, let ni=ni+1, and skipping to Step 102-1, or let ni=0, and skipping to Step 102-4; Step 102-4: if n2 is less than N2, let n2=n2+1, and skipping to Step 102-1, or let n2=0, and skipping to Step 102-5; and Step 102-5: if n3 is less than N3 , let n3=n3+1, and skipping to Step 102-1, or skipping to Step 103; and Step 103: completing the creation of the training sample set; the second part of the method involves using the training sample set created in the first part of the method to train a multi-layer feedforward neural network, allowing the multi-layer feedforward neural network to approximate the functional relation between the normalized

variance of received optical signal intensity fluctuations and A 0 , DG, ro, '7 and 7' , which

specifically comprises the following steps: Step 201: building a multi-layer feedforward neural network having six input nodes at an input layer thereof and one output node at an output layer thereof; and Step 202: realizing the training of the multi-layer feedforward neural network with the values

of six components of the input vector [Ao,00, DG, ,o, '7 ,'7 ] corresponding to each sample in

the training sample set as respective inputs to the six input nodes of the multi-layer feedforward neural network and the desired output normalized variance of received optical signal intensity fluctuations corresponding to each sample as the desired output of the multi-layer feedforward neural network, allowing the multi-layer feedforward neural network to approximate the functional relation between the normalized variance of received optical signal intensity

fluctuations and A 0 , ,G, o, '7 and ' ;

the third part of the method involves obtaining the normalized variance of received optical signal intensity fluctuations under particular external propagation conditions using the resulting multi-layer feedforward neural network from the second part of the method, which specifically comprises the following steps: Step 301: specifying a propagation length L, an optical wave number k, the radius Wo of a Gaussian laser beam in an emitting plane, the wavefront radius FO of curvature of the Gaussian laser beam in the emitting plane, the radius WG of a reeiving aperture, and calculating A-2 T" \(2 2II; C Ao=2L/(kWo2), 0o=1-L/Fo, and DG=2L/(kWG); specifying a profile (z) of atmospheric refractive-index structure constants along the propagation path, with z representing a distance away from the laser emitting device in the propagation path, and calculating the atmospheric coherence length of a spherical wave: ro = F0.42 3k 2'L IC L 53d | r-3 2 L calculating QrO=2L/(k "ro); calculating

3f C2 (Lj ) j53d 5 / 1/ 1/ I-2 2 3 j )5/ 6 26 C M - 3 C (Lj)()- )d

L~3 C [3f 3 C (L )( -j )51 d 363 , wherein 6 1 36 and

calculating 11 , 2 and 3 and

Step 302: combining A o, 00, QG, Oo, '7 and q' obtained in Step 301 into the input vector

[Ao, 00, QG, rO, '7, ' ]to the resulting multi-layer feedforward neural network from the

second part of the method, and then producing, by the multi-layer feedforward neural network, an output, i.e., the corresponding normalized variance of received optical signal intensity fluctuations.