Determination of average times for Brillouin optical time domain analysis sensor denoising by non-local means filtering

Integration of signal average and image restoration can improve the performance of distributed optical fiber sensors. Here we illustrate the denoising behavior of non-local means filtering (NLM) affected by number of average. On the one hand, the signal average can effectively remove the high-frequency noise but be time-consuming; on the other hand, the NLM can accelerate decent in noise but is dependent on level of noise contained in raw sensing image. To best of our knowledge, this is the first time to apply NLM to conventional Brillouin optical time domain analysis sensor (BOTDA) with theoretically spatial resolution of 10 ns. Moreover, we propose an optimal averaging number determination method for further denoising by NLM to achieve trade-off between accuracy and average times. The results indicated that the optimal-times for BOTDA configured 10 ns and 10km fiber-loop is 13 with Brillouin frequency shift uncertainty of 0.85MHz. The proposed method has potential to apply in engineering applications and commercialized software for BOTDA denoising by image restoration.