Enhancing BOTDR SNR Using Learned-Dictionary Sparse Reconstruction of Brillouin Gain Spectrum Matrix

A method of enhancing signal-to-noise ratio (SNR) in Brillouin optical time domain reflectometer (BOTDR) is proposed using sparse reconstruction of the Brillouin gain spectrum (BGS) matrix. In the sparse reconstruction process, the measured raw BGS matrix is treated as a noisy image, which can be represented as a linear combination of sparse basis and reconstructed by iteratively solving the optimal sparse coefficients. During the iterative operation, the random noise is suppressed as residuals and hence enhance the SNR. The basic principle, flow chart and solving algorithm are presented and further verified by numerical simulation. In the experiments, the temperature sensing is implemented by BOTDR to evaluate the denoising performance of BGS sparse reconstruction over 24.14 km fiber with 3m spatial resolution. Under 65°C, the fluctuation range and the root mean square error (RMSE) in the heating region are reduced from 13.9 MHz to 5 MHz and from 5.47 MHz to 1.65 MHz, respectively, with only 200 average counts. According to the calculation of distance-window RMSE, the SNR enhancement of 6.19 dB at the fiber end is achieved. Both simulation and experimental results show good performance in noise suppression.