Objective Brillouin optical time-domain reflectometer (BOTDR) is employed in numerous fields with practical applications owing to its various advantages such as measuring multiple physical parameters, long sensing distance, and single-end measurement. However, Brillouin backscattered signals become weak along the fibers. Moreover, noises can be introduced by optical or electrical components in the BOTDR system, resulting in poor signal-to-noise ratio (SNR) as well as difficulty in accurately measuring the Brillouin frequency shift for long fiber, which limits the transmission distance. Reducing noise to enhance SNR is an essential for solving the problem. Traditional denoising methods include cumulative average denoising, traditional wavelet threshold denoising, and lifting wavelet threshold denoising. The cumulative average denoising method has limitations such as the time of accumulation and the storage capacity of the hardware. Meanwhile, whether traditional wavelet threshold denoising or lifting wavelet threshold denoising, they depend heavily on the characteristics of the signal and choice of parameters, such as the wavelet basis, wavelet decomposition layer, threshold rule, and threshold function. Therefore, for low SNR signals, these two methods are combined with other methods to achieve the expected denoising results. Considering the good denoising performance of the cumulative average method and the short denoising time of lifting wavelet transform, a novel denoising scheme combining both lifting wavelet threshold and the cumulative average is proposed to improve the measurement distance.
Methods A denoising method for the BOTDR system that combines the lifting wavelet threshold with cumulative average is presented. The characteristics of spontaneous Brillouin scattering signals are analyzed. Then, we introduce the principle of lifting wavelet transform and discuss its denoising flowchart (Fig. 2). Furthermore, the parameters of the wavelet threshold denoising suitable for the BOTDR signal are analyzed and optimized through simulations. This paper compares the effectiveness of five denoising schemes (Fig. 5), which are cumulative average denoising, traditional wavelet threshold denoising, lifting wavelet threshold denoising, combined scheme of both traditional wavelet threshold and cumulative average denoising, and a combined scheme of both lifting wavelet threshold and cumulative average denoising. Finally, an experimental platform is built (Fig. 6) and the effectiveness of the proposed method is verified. In addition, the measurement results for the proposed method are compared with other denoising methods.
Results and Discussions The combination of lifting wavelet threshold and cumulative average denoising exerts higher SNR, shorter running time, and longer sensing distance than other denoising methods. Simulations and experiments are performed on two methods to verify the effectiveness of the proposed scheme. Brillouin signal power and SNR are replaced with transmission distance and are obtained from simulations (Fig. 1). The wavelet threshold denoising parameters are explored and discussed for different SNR scenarios (Fig. 3 & Fig. 4 & Table 1), which show the universality of the selected parameters. The optimal parameters for wavelet threshold denoising are obtained, including wavelet threshold denoising of db4, decomposition levels of 7, the threshold rule of Sqtwolog, and soft threshold function. The effectiveness of the five denoising schemes is compared, as shown in Table 2. The schemes combining two denoising algorithms show a better noise reduction effect than the others with one algorithm. Also, the SNR was improved greatly while the processing time reduced. Moreover, denoising using the lifting wavelet is better than that using the traditional wavelet. After combining the method of cumulative average noise reduction, the denoising and processing speed of the lifting wavelet remain better. The experimental platform is built to verify the reliability of the proposed scheme (Fig. 6). The results show that, compared with the simple cumulative average method, the proposed method improves the sensing distance from 16.22 km to 39.45 km, while processing time only increases 2.22 s. Also, when compared with the method of combining wavelet threshold with cumulative average denoising, the proposed method has a 0.63 km improvement in sensing distance and 0.69 s better in processing time reduction (Fig.7).
Conclusions A novel denoising scheme that combines lifting wavelet threshold with cumulative average is proposed to further improve the measurement distance. Combining lifting wavelet threshold with cumulative average denoising exerts higher SNR, shorter run time than the other four schemes of cumulative average denoising, traditional wavelet threshold denoising, lifting wavelet threshold denoising, combination of traditional wavelet threshold and cumulative average denoising. The reliability of the proposed scheme is verified by the experimental platform. Therefore, this study serves as an important reference for exploring noise reduction of Brillouin scattering signal.
