Rapid noise removal based dual adversarial network for the Brillouin optical time domain analyzer

Abstract

We propose a dual adversarial network (DANet) to improve the signal-to-noise ratio (SNR) of the Brillouin optical time domain analyzer. Rather than inferring the conditional posteriori distribution in the conventional maximum a posteriori (MAP) framework, DANet constructs a joint distribution from two different factorizations corresponding to the noise removal and generation tasks. This method utilizes all the information between the clean–noisy image pairs to preserve data completely without requiring traditional image priors and noise distribution assumptions. Additionally, the clean–noisy image pairs produced by the generator can expand the original dataset to retrain and enhance the denoising effect. The performance of DANet is verified using the simulated and experimental data. Without spatial resolution deterioration, an SNR improvement of 35.51 dB is observed in the simulation, and the Brillouin frequency shift (BFS) uncertainty along the fiber is reduced by 3.56 MHz. Experiments yield a maximum SNR improvement of 19.08 dB, with the BFS uncertainty along the fiber reduced by 0.93 MHz. Significantly, DANet has a processing time of 1.26 s, which is considerably faster than conventional methods, demonstrating its potential for rapid noise removal tasks.

Improving BFS measurement accuracy of BOTDR based on Cauchy proximal splitting

“…Nevertheless, these image-denoising methods suffer from challenges related to degradation of frequency accuracy and spatial resolution, as details of the data are weakened and small data sampling point number leads to insufficient information for denoising. Another approach to denoising is through neural network algorithms, which include convolutional neural networks [16], dual adversarial neural networks [17], and denoising and extracting convolutional neural networks [18]. Neural networks possess the ability to learn and extract relevant features from extensive datasets without necessitating analytical definitions or prior knowledge of the signal.…”

Signal-to-noise ratio enhancement using block-matching and 3D filtering for both Brillouin gain and phase spectra in sweep-free Brillouin optical time-domain analyzer

“…One-dimensional digital denoising methods, such as the wavelet (packet) denoising (WD) 21 and empirical mode decomposition, 22 , 23 have been proposed, but the denoising performance of these methods is greatly affected by the selection of wavelet basis functions or modal aliasing. Machine learning technology is also proposed for denoising 24 . Two-dimensional image denoising methods are also good choice to improve the SNR, which have been proved to achieve significant SNR improvement without modifying the system structure 21 , 25 , 26 .…”

“…Machine learning technology is also proposed for denoising. 24 Two-dimensional image denoising methods are also good choice to improve the SNR, which have been proved to achieve significant SNR improvement without modifying the system structure. 21,25,26 Two common image denoising techniques, nonlocal means (NLM) and WD, have been proposed to improve the SNR and enable the performance enhancement by two orders.…”