SNR enhancement for Φ-OTDR system based on SG filtering and double outlier detection combined with linear interpolation

Abstract

In phase-sensitive optical time domain reflectometry (Φ-OTDR) sensing systems, noise can reduce the vibration positioning signal-to-noise ratio (SNR) and distort the recovered phase signal. This poses the difficulty of vibration detection. To solve this problem, a Φ-OTDR signal processing method based on Savitzky–Golay (SG) filtering, double outlier detection (DOD), and linear interpolation (LI), termed SG-DOD-LI is proposed in this work. This method first preprocesses the demodulated amplitude signal through the SG filtering. Subsequently, it adopts the density-based adaptive local outlier factor algorithm to identify vibration regions on the optical fiber. Finally, it suppresses the noise peaks in the non-vibration regions by the secondary outlier detection and LI. Furthermore, the phase signals of the vibration regions are optimized. A Φ-OTDR sensing system was constructed, and 0.5 V sinusoidal vibration at frequencies of 500 Hz, 100 Hz, and 10 Hz were applied to the optical fiber. The results show that the proposed method achieved mean positioning SNR of 34.10 dB, 35.83 dB, and 29.23 dB at the three frequencies. Compared with the conventional methods, the average positioning SNR is improved by more than 20 dB. For phase signal optimization, the proposed method achieves approximately 2–6 dB greater noise reduction than the NLM filtering, empirical mode decomposition and Pearson correlation coefficient, and variational mode decomposition and Pearson correlation coefficient in some experiments. Meanwhile, the reductions in the mean SNR by the proposed method at 500 Hz, 100 Hz, and 10 Hz reach 14.54 dB, 15.78 dB, and 13.20 dB, respectively. This method provides an effective solution for vibration detection in Φ-OTDR systems and shows a good prospect of engineering applications.