KEYWORDS: Signal to noise ratio, Filtering (signal processing), Electronic filtering, Digital signal processing, Signal processing, Reflectometry, Optical filters, Interference (communication), Wavelets, Denoising
Phase-sensitive optical time-domain reflectometry (ϕ-OTDR) has been widely used to interrogate multipoint vibration events in the health monitoring of large-scale infrastructures. Since high signal-to-noise ratio (SNR) is the core parameter for evaluating the performance of ϕ-OTDR, many researchers have presented digital signal processing (DSP) methods for SNR enhancement. However, the DSP methods using fixed parameters cannot achieve the optimum SNR for different vibration events. In addition, although transform domain analysis methods such as wavelet transform could provide multiscale observation for target events, a large amount of computation would be required simultaneously. Matched filtering is a commonly used technique to extract noisy signals in traditional wireless radar systems. Mathematical theory has indicated that matched filtering is also suitable for signal extraction in ϕ-OTDR. The optimization of system performance could be realized when the filter scale matches the length of vibration events. For different lengths of disturbance region, a multiscale matching filtering method has been proposed, which makes it possible to choose appropriate filter scale and finally obtain the optimal SNR. Experimental results have shown that the proposed multiscale matching filtering method could improve the SNR by over 6 dB even under strong noise influence and reach the lowest locating uncertainty of 0.49 m with low time consumption, compared to conventional methods.