Method for identifying vortex beam mode number

By combining GS algorithm and ViT neural network, using Fourier domain and spatial domain information, the recognition difficulty and inaccuracy caused by beam distortion in vortex beam pattern recognition is solved, and high-precision pattern recognition is achieved.

CN119004257AInactive Publication Date: 2024-11-22HAINAN UNIV
View PDF 1 Cites 0 Cited by

Patent Information

Application Number
CN202411052206.9
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-01
Publication Date
2024-11-22
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

The prior art, when identifying the vortex beam mode, changes in the light intensity distribution caused by beam distortion, lead to difficulty and inaccuracy of recognition, which makes it difficult to meet the requirements of high accuracy.

Method used

Combining the Gerchberg-Saxton (GS) algorithm and Vision Transformer (ViT) neural network, through phase recovery and feature extraction, the Fourier domain and spatial domain information of the vortex beam are obtained, and the self-attention mechanism is used to identify the pattern number of light beams.

Benefits of technology

The accuracy and accuracy of vortex beam pattern recognition is significantly improved, and through multi-level feature extraction and representation learning, the phase characteristics of the beam are captured, providing more comprehensive and detailed optical characteristic data.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119004257A_ABST
    Figure CN119004257A_ABST
Patent Text Reader

Abstract

The invention discloses a method for identifying the mode number of a vortex light beam, which is used for efficiently and accurately identifying an unknown light beam by combining a GS algorithm and a ViT neural network. The method comprises the following steps: firstly, performing phase recovery on a vortex light beam distorted in a transmission process through a calibration system, gradually optimizing phase distribution of the light beam through multiple iterations of a GS algorithm, and extracting the phase of the distorted vortex light beam to obtain a preliminarily recovered phase image; and then the ViT neural network is introduced into an identification process. A powerful attention transfer mechanism of ViT is utilized, and through multi-level feature extraction and representation learning, the phase features of the vortex beam are accurately identified. The method shows high identification precision and stability in transmission channels with different distances, and can maintain high identification performance even in a complex underwater environment.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention belongs to the technical field of optical communications, and in particular relates to a method for identifying the number of vortex light beam modes. Background Art

[0002] In the field of optics, vortex beams are widely used in optical communications, optical sensing, laser processing and other fields. Vortex beams have a special axial structure, and their phase changes in a spiral shape, which can carry more degrees of freedom and information. However, in practical applications, due to the non-ideality of optical systems and optical elements and the influence of external factors such as turbulence, vortex beams often produce serious distortion, and the loss of the required phase structure and spatial characteristics has a great impact on the recognition of the beam.

[0003] Vortex beams of different modes have different light intensity distribution characteristics, and the beam pattern can be identified by a neural network method. It is known that at a given channel length, when the light beam is transmitted through the channel, it will be distorted due to the influence of the transmission medium. This distortion will cause a change in the light intensity distribution, making it difficult and inaccurate to identify the pattern of the vortex beam by light intensity. Specifically, the vortex beam will be affected by factors such as turbulence, scattering and diffraction during transmission, resulting in changes in the light intensity distribution. For example, vortex beams of different modes ideally have a specific annular spot structure and intensity distribution, but these features may be blurred or distorted during transmission. Therefore, relying solely on light intensity to identify the pattern of the vortex beam will introduce significant errors due to beam distortion. To make it more complicated, even at the same transmission distance, the degree of distortion of the light intensity distribution will vary due to the instability and randomness of the medium conditions in the transmission path. This further increases the difficulty and uncertainty of identifying the vortex beam pattern by light intensity. Taking the Laguerre-Gaussian beam as an example, the light intensity distribution of the beam presents a typical "donut" structure. The radius of the "donut" of a Laguerre-Gaussian beam with different mode numbers is different, and as the mode number increases, the radius of the "donut" also increases. This feature allows the Laguerre-Gaussian beam to distinguish different mode numbers through the light intensity distribution during pattern recognition. However, when the Laguerre-Gaussian beam is transmitted in a medium, it will be affected by the transmission medium, causing the beam to expand. This expansion phenomenon will make the light intensity distribution of the beam blurred, and the originally clear "donut"-shaped structure will be stretched, deformed, and even experience intensity attenuation. As the propagation distance increases, the expansion phenomenon becomes more and more obvious, thereby affecting the recognition accuracy of the beam. Therefore, the method of relying solely on light intensity recognition has great limitations in practical applications and is difficult to meet the requirements of high-precision recognition. Summary of the invention

[0004] In order to overcome the above-mentioned shortcomings of the prior art, the purpose of the present invention is to provide a method for identifying the mode number of a vortex beam. By combining the GS algorithm with the ViT neural network, the phase information of the vortex beam can be more accurately extracted and identified, thereby greatly improving the recognition accuracy.

[0005] To achieve the above object, the present invention provides a method for identifying the number of vortex beam modes, which is performed by the following steps:

[0006] 1. A method for identifying the number of vortex beam modes, comprising the following steps:

[0007] S1, when the vortex beam passes through the turbulent channel and reaches the receiving end, the phase of the vortex beam distorted during the transmission is recovered through the calibration system. The phase distribution of the beam is gradually optimized through multiple iterations of the GS algorithm. The phase of the distorted vortex beam is extracted to obtain a preliminary recovered phase image.

[0008] S2 uses the ViT neural network to divide the phase image of the distorted light beam restored by S1 into multiple small blocks, converts these small blocks into corresponding feature vectors, and performs self-attention operations between these feature vectors to capture the relationship between different areas in the image. Through multi-level feature extraction and representation learning, the phase characteristics of the vortex beam are identified and the mode number of the vortex beam is obtained.

[0009] Furthermore, the step of S1 phase recovery is:

[0010] S11, when the distorted vortex light beam passes through the beam splitter BS1 of the calibration system, it is split into two beams, one of which is used for identification processing and the other is used for information collection;

[0011] S12, the beam of light used for information collection in S11 is split into two beams again by the beam splitter BS2 of the calibration system, one of which passes through a Fourier lens, and the camera CCD2 in the calibration system measures the intensity information in the Fourier domain, which is used to obtain the characteristics of the vortex light beam in the frequency domain; the other beam of light does not pass through the Fourier lens, and the camera CCD1 in the calibration system directly measures the intensity information in the spatial domain, which records the transmission characteristics and light intensity distribution of the vortex light beam in the spatial domain;

[0012] S13, bringing the measured intensity information in the Fourier domain and the spatial domain into the data processor, and performing phase recovery through the GS algorithm.

[0013] Furthermore, the specific steps of the GS algorithm of S13 are as follows:

[0014] S131, the spatial domain intensity information measured by the CCD1 camera, i.e., the initial amplitude distribution f0(x, y), is combined with the randomly generated undistorted vortex beam phase Combined to get the initial light field As input data;

[0015] S132, perform Fourier operation on u0 in S131 to obtain the Fourier domain phase distribution U0=F(k x ,k y )ψ(k x ,k y ), determine whether the normalized root mean square error Δ≤ξ, where ξ is a given threshold. If it is satisfied, the iteration is terminated; otherwise, the amplitude F(k x ,k y ) is replaced by the amplitude F of the distorted beam in the Fourier domain measured by the CCD2 camera m (k x ,k y ), generate a new U0, and perform inverse Fourier processing on it to obtain a new The normalized RMS Δ is:

[0016]

[0017] Where N is the maximum number of iterations;

[0018] S133, replace the amplitude f(x, y) in u0 obtained in S132 with the amplitude f0(x, y) measured by the CCD1 camera, combine the new amplitude and phase, and u0 is updated again;

[0019] S134, perform Fourier transform on u0 obtained in S133 again, and then execute step S132 until the iteration termination condition is met or the maximum number of iterations is reached, and output Obtain the distorted phase screen.

[0020] Furthermore, the ViT neural network in S2 will construct a training data set through the multiple phase information recovered by S1. The data set covers the phase information of vortex beams with different mode numbers, different transmission distances and different turbulence intensities. The sample set is divided into training set, test set and verification set in proportion.

[0021] Furthermore, the training set, test set and verification set are divided into 8:1:1 ratios, the number of training rounds is set to 1000, and the Adam loss function is used.

[0022] Furthermore, the multiple small blocks of S2 are two-dimensional matrices, the size of which can be customized, and each small block represents a local feature of the image.

[0023] Compared with the prior art, the present invention has the following advantages and beneficial effects:

[0024] Compared with the prior art, the present invention can obtain comprehensive optical property data by simultaneously obtaining the Fourier domain and spatial domain information of the vortex light beam. The intensity information in the Fourier domain provides the spectral distribution and phase structure of the vortex light beam, while the intensity information in the spatial domain provides the transmission characteristics and distortion of the light beam. The advantage of this method is that through the setting of the beam splitter and different optical paths, the information of the vortex light beam in the Fourier domain and the spatial domain is obtained at the same time. In this way, the characteristics of the light beam can be more fully understood, and more accurate and detailed data can be provided for subsequent recognition processing and analysis. Secondly, the powerful feature extraction capability and attention mechanism of the VIT neural network are utilized to accurately identify the recovered phase, thereby greatly improving the recognition accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The drawings described herein are used to provide a further understanding of the present invention and constitute a part of this application. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0026] Figure 1 A schematic diagram of a calibration system for a method for identifying a vortex beam mode number provided in an embodiment of the present application;

[0027] Figure 2 A schematic diagram of the GS algorithm for identifying the number of vortex beam modes provided in an embodiment of the present application;

[0028] Figure 3 A ViT neural network flow chart of a method for identifying the number of vortex beam modes provided in an embodiment of the present application;

[0029] Figure 4 A method for identifying the mode number of a vortex beam provided in an embodiment of the present application is provided, which is a phase recovery diagram of an LG beam at different distances and different mode numbers;

[0030] Figure 5 A mean absolute error diagram at different distances of a method for identifying a vortex beam mode number provided in an embodiment of the present application;

[0031] Figure 6 A graph of average relative errors at different distances for a method for identifying the number of vortex beam modes provided in an embodiment of the present application;

[0032] Figure 7 A root mean square error diagram at different distances of a method for identifying the number of vortex beam modes provided in an embodiment of the present application;

[0033] Figure 8 A correlation coefficient diagram at different distances of a method for identifying vortex beam mode number provided in an embodiment of the present application.

[0034] The realization of the purpose, functional features and advantages of the present invention will be further explained in conjunction with embodiments and with reference to the accompanying drawings. DETAILED DESCRIPTION

[0035] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0036] The embodiment of the present invention provides a method for identifying the mode number of a vortex beam, and combines the Gerchberg-Saxton (GS) algorithm with the Vision Transformer (ViT) neural network to efficiently and accurately identify an unknown beam.

[0037] To achieve the above purpose, the following steps are mainly performed:

[0038] S1, when the vortex beam passes through the turbulent channel and reaches the receiving end, the phase of the vortex beam distorted during the transmission is recovered through the calibration system. The phase distribution of the beam is gradually optimized through multiple iterations of the GS algorithm. The phase of the distorted vortex beam is extracted to obtain a preliminary recovered phase image.

[0039] S2 uses the ViT neural network to divide the phase image of the distorted light beam restored by S1 into multiple small blocks, converts these small blocks into corresponding feature vectors, and performs self-attention operations between these feature vectors to capture the relationship between different areas in the image. Through multi-level feature extraction and representation learning, the phase characteristics of the vortex beam are identified and the mode number of the vortex beam is obtained.

[0040] In this joint method, the traditional GS algorithm is first used for preliminary phase recovery. Then, the recovered phase is deeply identified and analyzed by the ViT neural network.

[0041] Specifically, at the transmitting end of the vortex optical system, Figure 1As shown, a distorted vortex beam is obtained after the light beam is disturbed by turbulence. The calibration system is mainly composed of two charge coupled devices (CCD cameras), a Fourier lens and two beam splitters (BS). When the distorted light beam is divided into two beams by BS1, one beam is used to identify information, and the other beam is used for information collection, and is split again by BS2. The intensity of the light beam in the time domain is obtained by the CCD1 camera, and the intensity of the light beam in the Fourier domain is obtained by the CCD2 camera. Both are input into the data processor to obtain a phase estimate of the distorted light beam. It should be noted that the CCD camera measures the distribution of light intensity, while the embodiment of the present invention uses the amplitude distribution, and the measured data needs to be squared before use.

[0042] Among them, the camera CCD2 in the calibration system measures the intensity information in the Fourier domain, which is used to obtain the characteristics of the vortex light beam in the frequency domain; the camera CCD1 in the calibration system measures the intensity information in the spatial domain, and this beam of light records the transmission characteristics and light intensity distribution of the vortex light beam in the spatial domain;

[0043] The measured intensity information in the Fourier domain and spatial domain is brought into the data processor, and phase recovery is performed using the GS algorithm.

[0044] Specifically, the specific steps of the GS algorithm are as follows: Figure 2 As shown, the details are as follows:

[0045] S131, the spatial domain intensity information measured by the CCD1 camera, i.e., the initial amplitude distribution f0(x, y), is combined with the randomly generated undistorted vortex beam phase Combined to get the initial light field As input data;

[0046] S132, perform Fourier operation on u0 in S131 to obtain the Fourier domain phase distribution U0=F(k x ,k y )ψ(k x ,k y ), determine whether the normalized root mean square error Δ≤ξ, where ξ is a given threshold. If it is satisfied, the iteration is terminated; otherwise, the amplitude F(k x ,k y ) is replaced by the amplitude F of the distorted beam in the Fourier domain measured by the CCD2 camera m (k x ,k y ), generate a new U0, and perform inverse Fourier processing on it to obtain a new The normalized RMS Δ is:

[0047]

[0048] Where N is the maximum number of iterations;

[0049] S133, replace the amplitude f(x, y) in u0 obtained in S132 with the amplitude f0(x, y) measured by the CCD1 camera, combine the new amplitude and phase, and u0 is updated again;

[0050] S134, perform Fourier transform on u0 obtained in S133 again, and then execute step S132 until the iteration termination condition is met or the maximum number of iterations is reached, and output Obtain the distorted phase screen.

[0051] ViT neural model Figure 3 shown.

[0052] (1) First, divide the image into multiple small patches (patches). Each patch is a two-dimensional matrix whose size can be customized. Each patch represents a local feature of the image.

[0053] (2) Each patch is sequentially input into the linear projection of the flattened patches, i.e., Linear Projection of Flattened Patches. Each patch gets a corresponding vector, which can be called a token. The function of the Linear Projection of Flattened Patches module is to convert each image patch into the input form required by the Transformer model. It usually includes flattening the image patches and adding position encoding to indicate their position information in the image.

[0054] (3) The embedding vector of each image patch adds a positional encoding to indicate its position in the original image, which helps the model capture the spatial structure of the image.

[0055] (4) A special token, called a class token, is added in front of the representation of all image patches. This token is used to capture the global information of the entire image and is ultimately used for image classification tasks.

[0056] (5) Next, each new image block representation is input into the encoder (Transformer Encoder). The Transformer Encoder consists of multiple Transformer blocks, the number of which L can be set.

[0057] (6) Finally, the class token is input into the Transformer Encoder and the corresponding output is input into the Multilayer Perceptron (MLP Head) to obtain the final classification result. When using your own data, the MLP Head can be regarded as a fully connected layer (Linear).

[0058] Transformer Encoder extracts features from the received sequence through a multi-layer self-attention mechanism and feedforward neural network, integrates the position encoding information, dynamically adjusts the weights of different positions in the sequence, and converts image blocks into feature representations.

[0059] The following is an example of restoring the phase of a vortex beam based on the GS algorithm, and the specific implementation of the present invention is described in conjunction with the accompanying drawings: When the laser and spatial light modulator at the transmitting end generate a vortex beam and pass through a turbulent channel, the distorted vortex beam will be subjected to the following steps when it reaches the receiving end:

[0060] (1) A Laguerre-Gaussian (LG) beam with a topological charge value of l=3 is passed through a beam splitter (BS1).

[0061] When the distorted vortex beam passes through BS1 to split a beam into two beams, one beam is used for identification processing, while the other is used for information collection. The further process is as follows:

[0062] The beam of light responsible for information collection passes through BS2 again, which further divides it into two beams. One of the beams passes through the Fourier lens, and then the intensity information in the Fourier domain is measured by the CCD2 camera. This step is to obtain the characteristics of the vortex beam in the frequency domain in order to more comprehensively analyze its amplitude information. The CCD2 camera records and transforms the intensity distribution image after the Fourier lens. The other beam of light does not pass through the Fourier lens, and the intensity information in the spatial domain is directly measured by the CCD1 camera. This beam of light records the transmission characteristics and light intensity distribution of the vortex beam in the spatial domain. The CCD camera records and transforms the intensity distribution image in the spatial domain.

[0063] By simultaneously acquiring the Fourier domain and spatial domain information of the vortex beam, comprehensive optical property data can be obtained. The intensity information in the Fourier domain provides the spectral distribution and phase structure of the vortex beam, while the intensity information in the spatial domain provides the transmission characteristics and distortion of the beam.

[0064] The advantage of this method is that by setting up a beam splitter and different optical paths, the information of the vortex beam in the Fourier domain and the spatial domain can be obtained at the same time. This can provide a more comprehensive understanding of the characteristics of the beam and provide more accurate and detailed data for subsequent identification processing and analysis.

[0065] (2) The measured intensity information in the Fourier domain and the spatial domain is brought into the data processor.

[0066] The data processor process is also the GS algorithm process, such as Figure 2 As shown, the input is the amplitude distribution f0(x,y) of the distorted beam in the spatial domain and the amplitude F m (k x ,k y ) and randomly generated phases

[0067] The specific data processor process is as follows:

[0068] 1) The initial amplitude distribution f0(x,y) and phase Combined to generate the initial light field As the input data of GS algorithm.

[0069] 2) Perform Fourier operation on u0 to obtain U0 = F(k x ,k y )ψ(k x ,k y ). Determine whether the normalized root mean square (RMS) error △≤ξ (ξ is a given threshold, ξ=0.000001). If it is satisfied, the iteration is terminated. Otherwise, the amplitude F(k x ,k y ) is replaced by the intensity F measured by the CCD2 camera m (k x ,k y ), generate a new U0, and perform inverse Fourier processing on it to obtain a new u0. The normalized root mean square △ is:

[0070]

[0071] N is the maximum number of iterations, N=200.

[0072] 3) In u0, replace the amplitude f(x,y) with the amplitude f0(x,y) measured by the CCD1 camera. Combine the new amplitude and phase, and u0 is updated.

[0073] 4) Perform Fourier transform on the obtained u0 again, and then execute step (2) until the iteration termination condition is met or the maximum number of iterations is reached, and output Obtain the distorted phase screen.

[0074] This process extracts the phase of the distorted vortex beam by combining the measured amplitude information with the phase information of the known undistorted vortex beam, and obtains the phase diagram, such as Figure 4 shown.

[0075] In order to explore the reliability and robustness of the phase recognition vortex beam, the embodiment of the present invention comprehensively evaluates the performance of the method by setting transmission channels of different distances. Specifically, the experiment evaluates the effect of the phase recognition method by measuring and analyzing the following four key indicators:

[0076] (1) Mean Absolute Error (MAE)

[0077] The accuracy of the recognition method can be evaluated by calculating the mean absolute error between the recognized beam phase and the actual beam phase. The smaller the MAE, the closer the recognition result is to the true value and the higher the accuracy of the method.

[0078]

[0079] (2) Mean Relative Error (MRE)

[0080] The stability and reliability of the identification method under different conditions are evaluated by calculating the mean relative error between the identification results and the actual values. The smaller the MRE, the higher the accuracy of the method at different distances and turbulence intensities.

[0081]

[0082] (3) Root Mean Square Error (RMSE)

[0083] By calculating the root mean square error between the recognition result and the actual value, the overall error level of the recognition method can be evaluated. RMSE combines the square average of the error and can more sensitively reflect larger errors, which helps to fully understand the performance of the method.

[0084]

[0085] (4) Correlation Coefficient (R)

[0086] By calculating the correlation coefficient between the recognition result and the actual value, the accuracy and consistency of the recognition method can be evaluated. The closer the correlation coefficient is to 1, the stronger the linear relationship between the recognition result and the actual value, and the higher the accuracy of the method.

[0087]

[0088] Where N is the total number of test samples, is an estimated value, is the true value, for yi The average value of .

[0089] The experimental results show that the method exhibits high recognition accuracy and stability in transmission channels of different distances. The mean absolute error and mean relative error are both low, the root mean square error is controlled within a reasonable range, and the correlation coefficient is close to 1, indicating that the recognition method can accurately capture the phase information of the vortex beam and maintain high recognition performance even in complex underwater environments.

[0090] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0091] The professionals should also be further aware that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented with electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in terms of function in the above description. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0092] The above is only an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application should be included in the scope of the claims of the present application.

Claims

1. A method for identifying a vortex beam mode number, characterized in that: The following steps are involved: S1, when the vortex beam passes through the turbulent channel and reaches the receiving end, the phase of the vortex beam distorted during the transmission is recovered through the calibration system. The phase distribution of the beam is gradually optimized through multiple iterations of the GS algorithm. The phase of the distorted vortex beam is extracted to obtain a preliminary recovered phase image. S2 uses the ViT neural network to divide the phase image of the distorted light beam restored by S1 into multiple small blocks, converts these small blocks into corresponding feature vectors, and performs self-attention operations between these feature vectors to capture the relationship between different areas in the image. Through multi-level feature extraction and representation learning, the phase characteristics of the vortex beam are identified and the mode number of the vortex beam is obtained.

2. A method for identifying the number of vortex beam modes according to claim 1, characterized in that: The steps of S1 phase recovery are: S11, when the distorted vortex light beam passes through the beam splitter BS1 of the calibration system, it is split into two beams, one of which is used for identification processing and the other is used for information collection; S12, the beam of light used for information collection in S11 is split into two beams again by the beam splitter BS2 of the calibration system, one of which passes through a Fourier lens, and the camera CCD2 in the calibration system measures the intensity information in the Fourier domain, which is used to obtain the characteristics of the vortex light beam in the frequency domain; the other beam of light does not pass through the Fourier lens, and the camera CCD1 in the calibration system directly measures the intensity information in the spatial domain, which records the transmission characteristics and light intensity distribution of the vortex light beam in the spatial domain; S13, bringing the measured intensity information in the Fourier domain and the spatial domain into the data processor, and performing phase recovery through the GS algorithm.

3. A method for identifying the number of vortex beam modes according to claim 2, characterized in that: The specific steps of the GS algorithm of S13 are as follows: S131, the spatial domain intensity information measured by the CCD1 camera, i.e., the initial amplitude distribution f0(x, y), is combined with the randomly generated undistorted vortex beam phase Combined to get the initial light field As input data; S132, perform Fourier operation on u0 in S131 to obtain the Fourier domain phase distribution U0=F(k x ,k y )ψ(k x ,k y ), determine whether the normalized root mean square error Δ≤ξ is satisfied, where ξ is a given threshold. If satisfied, the iteration terminates; otherwise, the amplitude F(k x ,k y ) is replaced by the amplitude F of the distorted beam in the Fourier domain measured by the CCD2 camera m (k x ,k y ), generate a new U0, and perform inverse Fourier processing on it to obtain a new The normalized RMS Δ is: Where N is the maximum number of iterations; S133, replace the amplitude f(x, y) in u0 obtained in S132 with the amplitude f0(x, y) measured by the CCD1 camera, combine the new amplitude and phase, and u0 is updated again; S134, perform Fourier transform on u0 obtained in S133 again, and then execute step S132 until the iteration termination condition is met or the maximum number of iterations is reached, and output Obtain the distorted phase screen.

4. The method for identifying the vortex beam mode number according to claim 1, characterized in that: The ViT neural network in S2 will construct a training data set through multiple phase information recovered by S1. The data set covers phase information of vortex beams with different mode numbers, different transmission distances and different turbulence intensities. The sample set is divided into training set, test set and verification set in proportion.

5. A method for identifying the number of vortex beam modes according to claim 4, characterized in that: The training set, test set and verification set are divided into 8:1:1 ratio, the number of training rounds is set to 1000 rounds, and the Adam loss function is used.

6. The method for identifying the vortex beam mode number according to claim 1, characterized in that: The multiple small blocks of S2 are two-dimensional matrices, the size of which can be customized, and each small block represents a local feature of the image.

Citation Information

Patent Citations

  • Distortion vortex light correction system and method based on GS algorithm and deep learning

    CN117908248A