Reaching millikelvin resolution in Raman distributed temperature sensing using image processing

Image processing is proposed and experimentally demonstrated to improve the capabilities of Raman distributed optical fibre sensors. The here reported technique consists in stacking consecutive one-dimensional Raman Stokes and anti-Stokes traces in two-dimensional data arrays (one for each Raman component), which are then processed by an image denoising algorithm. Owing to the high level of correlation between consecutive measurements in conventional Raman sensing, it is experimentally demonstrated that this newly-proposed two-dimensional denoising approach provides a significant signal-to-noise ratio improvement, which in this case reaches 13.6 dB with no hardware modification to the conventional set-up. Experimental results demonstrate Raman distributed sensing with a remarkably enhanced temperature resolution of 4 mK at 9 km distance, which is obtained with 2 m spatial resolution and a short acquisition time of 35 s.