Noise reduction in a Brillouin optical time-domain sensor by a frequency-domain feature filter

A novel frequency-domain image processing method is proposed, to the best of our knowledge, to filter the noise from data collected by distributed optical fiber sensors based on Brillouin optical time-domain sensing (BOTDS). In the proposed method, we first divide a data image into subimages, and then we filter the noisy subimages by retaining the useful frequency information corresponding to the Lorentz-shape frequency spectrum and Brillouin frequency shift (BFS) transitions. The denoising performance improvements are verified by simulation and experiment. The performances in terms of temperature/strain measurement uncertainty, spatial resolution, and processing time achieved by the proposed filter are then compared with those by using a Gaussian filter and a nonlocal means (NLM) filter. In a proof-of-concept experiment with a 5.2 km length G657 sensing fiber, we achieve a temperature measurement uncertainty improvement of 27% compared with the results obtained by using the Gaussian filtering method. Furthermore, the processing speed of the proposed method is 22 times faster than that of the NLM filter under the same temperature measurement uncertainty.