In long-range Brillouin optical time-domain reflectometer (BOTDR) sensors, owing to the excessive attenuation of optical signals and low signal-to-noise ratio (SNR), the BFS fluctuation range measured at a certain position of the fiber can be significantly larger than that at other positions. Solving this problem by only increasing the average number of traces is challenging. Increasing the average trace time significantly increases the measurement time and degrades the real-time performance of the sensor. Therefore, an effective denoising method must be devised to prevent the BFS fluctuation range from increasing, decrease the trace average time, and improve the measurement accuracy of long-range BOTDRs.

In this study, an backward Raman amplification scheme was used to improve the energy of a Brillouin backscattered signal, avoid nonlinear effects, and increase the sensing distance. Simultaneously, block matching and 3D filtering algorithm (BM3D) were introduced to solve the problem of expanding BFS fluctuation range in long-range BOTDRs. The search window used to identify similar blocks in the BM3D algorithm was adjusted to a rectangle, which increased the search range in the distance direction and rendered it more adaptive to the BGS-distance matrix.

At a sensing distance of 100 km, with a spatial resolution of 2 m and trace average number of 100, the BFS fluctuation range reduced from ±15.761 to ±2.276 MHz before and after denoising (Fig. 6), and the BFS fluctuation range is consistent throughout the fiber. The BFS fluctuation range is ±6.598 MHz under trace average number of 10000, and it continues to increase at the position featuring the lowest SNR. The maximum RMSE value is reduced from 6.477 to 1.028 MHz before and after denoising, and the maximum root mean square error (RMSE) value is ±3.246 MHz under an average of 10000 trace times (Fig. 8). In addition, the linearity and spatial resolution remains almost unchanged before and after denoising (Figs. 9 and 10). Finally, the performance of the BM3D algorithm is compared with the trace average and wavelet threshold denoising algorithm, and the R_{SNR En} of each method is calculated (Fig. 11, Table 1). The results show that a 6.301-dB SNR_En is achieved by the BM3D algorithm and that the BFS fluctuation range processed by the BM3D algorithm is the smallest.

In this study, the measurement accuracy of long-distance BOTDRs is improved and the trace average times required to measure the BFS is reduced. Experimental results show that the RMSE of the system after denoising is 1.028 MHz and that the BFS fluctuation range is ±2.276 MHz under a spatial resolution of 20 m, a 100-km-long sensing fiber, and trace average number of 100. The linearity and spatial resolution are not significantly affected. The proposed system improves the engineering application potential of BOTDRs in the monitoring of large-scale infrastructures, such as long-distance high-voltage transmission cables and submarine fiber-optic communication cables.