High-accuracy strain sensing of frequency-scanned phase-sensitive optical time domain reflectometry based on block matching.

A strain demodulation method of frequency-scanned -OTDR based on block matching is proposed, which improves the demodulation performance under short pulse width and large strain. The method of frequency shift estimation based on image block matching improves the similarity between Rayleigh backscattered spectra and the accuracy of demodulation results by combining local frequency domain data and local spatial domain data. The image of the matching degree is denoised by a block-matching three-dimensional filter (BM3D), and the performance is further improved by three-dimensional transformation and collaborative filtering of similar image blocks. The results demonstrate that the frequency shift standard deviations with 10 ns pulse width demodulated by the proposed method are always below 0.0164 GHz (i.e. 0.11), the overall accuracy is increased by more than 4.25 times compared with Pearson correlation coefficient, Euclidean distance, LMS, Cosine distance, the traditional correlation coefficient. And the accuracy of the large strain is increased by dozens of times. In addition, the root mean square errors of frequency shift with 10 ns pulse width demodulated by the proposed method are always below 0.0685 GHz (i.e., 0.46), which is 16.76% lower than the aforementioned methods. And the root mean square errors of the results with large strain demodulated by the proposed method are reduced by 77.4% and 90.4% for 20 ns pulse width and 10 ns pulse width, respectively. The proposed method can effectively improve the accuracy of frequency-scanned -OTDR strain sensing with no hardware modification.