A novel convolutional autoencoder-based approach for reconstruction of finite energy Airy–Hermite–hollow Gaussian beams propagating in atmospheric turbulent link
Physica Scripta, cilt.101, sa.34, 2026 (SCI-Expanded, Scopus)
- Yayın Türü: Makale / Tam Makale
- Cilt numarası: 101 Sayı: 34
- Basım Tarihi: 2026
- Doi Numarası: 10.1088/1402-4896/ae8e0b
- Dergi Adı: Physica Scripta
- Derginin Tarandığı İndeksler: Science Citation Index Expanded (SCI-EXPANDED), Scopus, Chemical Abstracts Core, Compendex, INSPEC, zbMATH
- Anahtar Kelimeler: atmospheric turbulence, CAE, denoising, FAHHG beam, free-space optical communication, prediction
- Sivas Cumhuriyet Üniversitesi Adresli: Evet
Özet
In this study, a convolutional autoencoder (CAE)-based approach is proposed to estimate the undistorted versions of finite energy Airy–Hermite–hollow Gaussian beam (FAHHGB) profiles that have been degraded by atmospheric turbulence. The primary novelty of this study lies in the first experimental use and modeling of FAHHGB beam data in the literature, introducing both a previously unexplored data type and a deep learning-based reconstruction methodology. A total of 3240 data samples were recorded, of which 1620 were turbulent and 1620 were non-turbulent; initial experiments using a baseline CAE model gave average structure similarity index method (SSIM) and peak signal-to-noise ratio (PSNR) values of 0.8973 and 23.32 dB, respectively. Subsequently, an enhanced CAE architecture with batch normalization and skip connections achieved improved metrics of 0.9489 SSIM and 28.1 dB PSNR. The results demonstrate that CAE-based architectures can effectively reconstruct FAHHGB profiles from turbulence-induced distortions. By introducing FAHHGB beam data to deep learning-based turbulence mitigation for the first time, this study provides a unique and valuable dataset and establishes a new reference point for future research in this area.