Chinese Optics Letters, Volume. 19, Issue 11, 110601(2021)
Compensation of turbulence-induced wavefront aberration with convolutional neural networks for FSO systems
Figures & Tables(6)
Fig. 1. Block diagram of an AO system with deep learning for FSO communication. BS, beam splitter. Inset: AlexNet structure.
Fig. 2. (a) Normalized power as a function of mode count and (b) phases of the first ten Zernike modes. Test error and power penalty for different (c), (d) numbers of Zernike modes (K), (e), (f) quantization bits, and (g), (h) CNN structures. (c)–(f) D/r0 = 16. (g), (h) D/r0 = 0–16.
Fig. 3. Comparison in power penalty among SPGD, SA, and AlexNet-based CNN.
Fig. 4. Experimental setup for evaluation of a CNN-based AO system and the corresponding block diagram.
Fig. 5. (a) Loss performance versus epochs for training CNN. (b) Estimated Zernike coefficients and absolute errors. (c) Wavefront aberration and (d) corresponding intensity images (D/r0 = 16).
Fig. 6. Power penalty in the weak/strong turbulence case. Inset: power penalty versus RMS of estimated wavefront errors (D/r0 = 16).
Tools
Get Citation
Copy Citation Text
Min’an Chen, Xianqing Jin, Shangbin Li, Zhengyuan Xu. Compensation of turbulence-induced wavefront aberration with convolutional neural networks for FSO systems[J]. Chinese Optics Letters, 2021, 19(11): 110601
Paper Information
Category: Fiber Optics and Optical Communications
Received: Mar. 4, 2021
Accepted: Apr. 15, 2021
Posted: Apr. 16, 2021
Published Online: Aug. 13, 2021
The Author Email: Xianqing Jin (xqjin@ustc.edu.cn), Zhengyuan Xu (xuzy@ustc.edu.cn)
© Copyright 2018-2021 | Chinese Laser Press.
All Rights Reserved 沪ICP备15018463号-20