Fig 6 - available via license: Creative Commons Attribution 4.0 International
Content may be subject to copyright.
(a) Accuracy of LG light getting through atmospheric turbulence with various C 2 n when the distance travelled is 1000m. (b) Accuracy of MFFCNN recognition for various number of images in training set for one single mode with C 2 n = 1 × 10 −14 . (c) Accuracy of recognition for various number of images in training set for one single mode with C 2 n = 1 × 10 −14 using previous methods. (d) Behavior of MFFCNN and one dimensional CNNs with different number of convolutional layers at z = 1000m and C 2 n = 1 × 10 −14 . x coordinates represent the structure of convolutional layers. For example, '16-32' means that there are two convolutional layers with 16 and 32 channels respectively. (e) Behavior of MFFCNN and one dimensional CNNs with different number of channels at z = 1000m and C 2 n = 1 × 10 −14 .
Source publication
Due to countless orthogonal eigenstates, light beams with orbital angular momentum(OAM) have a large potential information capacity. Recently, deep learning has been extensively applied in recognition of OAM mode. However, previous deep learning methods require a constant distance between laser and receiver. The accuracy will drop quickly if the di...
Contexts in source publication
Context 1
... is shown in Fig.6. (a) that the accuracy of intensity one dimensional recognition starts to fall when the structure constant of refractive index C 2 n reaches 1×10 −15 . For angular spectrum one dimensional recognition, the accuracy of recognition begins to slide when C 2 n reaches 4×10 −15 , which shows that angular spectrum alone is also a more accurate ...
Context 2
... might as well refer the size for each class satisfying requirement above as converge point (CP) in this letter. It can be seen in Fig.6. (b) (c) that the longer LG light propagated that is to say the severer LG light is affected, the larger training set is needed. ...
Context 3
... propagates 1000m and 1200m, the CP is 300 for MFFCNN. Meanwhile, the CPs are 600 and 750 for z = 1400 and z = 1600 respectively. This result is reasonable, because we usually need larger training set to find best map for classification when the input gets more noise. As for previous methods, CP points are 600 for various distances as shown in Fig. 6. (c). In a nut shell, the size of training set required by the MFFCNN is only half that for previous method which can make better use of the potential advantage of large OAM state ...
Context 4
... Fig.6.(d) (e), we can see that MFFCNN does not need a complex CNN structure to reach a very high accuracy. ...
Context 5
... we can see that MFFCNN does not need a complex CNN structure to reach a very high accuracy. In Fig.6.(d), we can also see that MFFCNN can get 95% accuracy even with only one convolutional layer, meanwhile accuracy for one dimensional CNN is only 65%. ...
Context 6
... features that cannot be extracted from angular spectrum dimension do not need complex CNN to extract from intensity dimension to extract. As shown in Fig.6.(e), intensity one dimensional recognition needs larger number of channels to reach a stable point, which also means that intensity one dimensional recognition is computationally expensive. ...
Similar publications
The InSight mission, featuring continuous high‐frequency high‐sensitivity pressure measurements, is in ideal position to study the active atmospheric turbulence of Mars. Data acquired during 1.25 Martian year allows us to study the seasonal evolution of turbulence and its diurnal cycle. We investigate vortices (abrupt pressure drops), local turbule...
With satellite laser communication, advantages can be achieved in high communication rate, precise beam pointing, high confidentiality, and light‐small terminal. Satellite–ground (SG) laser links also play a crucial role as trunk links. However, the SG link is influenced by atmospheric turbulence and platform vibration, these two factors will cause...
Citations
... Third, it has good scalability. Like the Graph Pad Prism tool [29], many image features and computational methods of laser process parameters can be researched in this proposed model; and other machine learning tools, such as the deep learning network [30] can also be designed in our next research step. Clearly, the proposed method also has some drawbacks. ...
An image analysis-based Reflection Coefficient (RC) estimation method of femtosecond laser surface processing for the blackening of X-ray imaging sensor shell is proposed. The Support Vector Regression (SVR) is used for RC computation and both an offline and an online steps are considered in this method. Regarding offline step, the typical laser process parameters are set to perform surface processing and Scanning Electron Microscope (SEM) images are recorded. Then a series of image features are computed and both the computed image features and typical laser parameters are used to train SVR: the training dataset includes the laser line space, laser beam diameter, natural logarithm of laser power divided by laser frequency, and image features of Gray-Level Co-occurrence Matrix (GLCM); the supervising data are laser ablation diameters. As for online step, when SEM image data are recorded after laser processing, the trained SVR is used to predict laser ablation diameter and then the RC can be computed by laser ablation model. Many experiment results have verified the effectiveness of our proposed method, and the RC estimation accuracy can be better than 90.0%.
... Apart from recognizing OAM modes in fractions other than integers, distinguishing between positive and negative topological charges is also of interest to the researchers. OAM modes that only differ in the signs of topological charges cannot be distinguished by a direct intensity measurement because they have identical intensity distributions [31]; therefore, an astigmatic transformation was utilized in [32] to generate new images that are distinguishable for those OAM modes. Every new transformed image was combined with its original direct image to form a so-called "two-channel" image that would be fed to a residual network. ...
Accurate recognition of orbital angular momentum (OAM) modes is a major challenge for OAM-based optical communications over atmospheric turbulence channels. The turbulence-induced distortions cause difficulties for the receiver to distinguish between adjacent OAM modes. Deep learning, such as convolutional neural networks (CNN), has been a promising technique in solving this problem. To improve the recognition performance, we propose a vortex modulation method that can magnify the subtle differences between closely adjacent OAM states. This allows the CNN to capture the image features more effectively and to recognize the topological charges more accurately. Numerical results show high recognition accuracy for both integer topological charges and fractional ones even under strong turbulence intensity and long propagation distance, which demonstrate the utility of the proposed method.
... All these methods require a well-defined wavefront, which is only applicable to transmission in free space. Recently, learningbased methods have been introduced to identify OAM states in long-distance optical communication, in which the effect of atmospheric turbulence [15][16][17][18][19][20] needs to be considered. ...
The inhomogeneity of turbid medium disrupts the coherent vortex structure of vortex beam and causes the formation of speckle pattern. Here, we propose a new, flexible approach to measure the orbital angular momentum (OAM) of vortex beams through turbid medium by convolutional neural network (CNN). The proposed technique directly recognizes the speckle image and obtains the corresponding OAM mode information. The accuracies exceed 99% and 97% in simulations and experiments respectively. In addition, the relationship between the recognition accuracy and environmental noise level, is present. The results show great potential in fiber communication, biomedical imaging and astronomical application, etc.
Angular momentum of light can be divided into spin angular momentum (SAM) and orbital angular momentum (OAM). Due to the theoretically unlimited orthogonal states, the physical dimension of OAM provides a potential solution to boost the information capacity. The OAM multiplexing and modulation techniques have been implemented to meet the continuous growth of bandwidth requirements, resulting in the concept of OAM optical communication. However, the performances of the traditional optical OAM detection techniques degrade seriously in the practical application of OAM optical communications. Thanks to the powerful data analysis advantages, the cutting-edge machine learning (ML) algorithms have been widely used in the field of image processing, laying the technical foundation for OAM recognition. This paper reviews the recent advances on OAM optical communications that are enhanced by ML methods. More than the traditional OAM detection methods, the OAM demodulation methods based on multiple network architectures, including the support vector machine (SVM), self-organizing map (SOM), feed-forward neural network (FNN), convolutional neural network (CNN), and diffractive deep optical neural network (D ² NN), have been summarized. We also discuss the development of the spiking neural network and on-chip D ² NN, opening a possible way to facilitate the future ultra-low power and ultra-fast OAM demodulation technology.
