Intensifying the SNR of BOTDA using adaptive constrained least squares filtering

We propose and experimentally demonstrate an adaptive image restoration method based on constrained least squares filtering (CLS) and noise level estimation by principal component analysis of low-rank patches (PCA) that enhances the signal-to-noise ratio (SNR) of Brillouin optical time domain analysis sensor (BOTDA). The proposed method employs PCA to estimate the initial noise level of the sensing image and subsequently the input parameter for CLS can be adjusted by a constraint factor. Furthermore, to effectively remove the noise in the raw data, we introduce a calculation procedure for which the noise level of the denoising image serves as a criterion. This procedure is adaptive and can restore the optimal Brillouin gain spectrum. In the experiment, the proposed method shows an increase of 12.2 dB in SNR and that the temperature accuracy can be increased to 1.3 °C. The adaptive CLS is also applied to sensing data with the different noise level, which validates that this method is a valid denoising technique. This report describes that image restoration, including non-local means, wavelet filter and CLS, has been applied to conventional BOTDA with the spatial resolution of 1m. Due to the influence of linewidth broadening for a 10 ns pulse, these methods all show sub-meter distortion in resolution, whereas the proposed method has the least impact on the spatial resolution with 1.41m. The proposed method is adaptive and can provide faster processing speed; thus it can be considered to be an alternative technique to enhance the performance of distributed optical fiber sensors, especially for long-distance and high-resolution sensing.