Research On Optical Fiber Sensing Technology Based On Compressed Sensing

Among all kinds of optical fiber sensors,the distributed optical fiber sensors are applied for sensing environment requiring continuous long-distance sensing capability,while Bragg grating based optical fiber sensors and Fabry-Perot optical fiber sensors are considered as point sensors,and widely used in various applications.Compressed sensing theory asserts that compressible signals can be recovered from much smaller number of measured data points than traditional method uses.After the discrete data points are collected from the signal with a sampling rate far less than that required by the Nyquist theorem,the original signal can be reconstructed without distortion from the collected data.It has been widely used in the fields of medical imaging,signal processing,and electronic engineering.We focus on spectral reconstruction of distributed fiber sensors based on stimulated Brillouin scattering,and Fabry-Perot optical fiber sensors,Bragg grating based optical fiber sensors which are widely used in various applications.This thesis includes the following works:Firstly,we introduce the principle,applications and research status of optical fiber sensing technology,and describe the principle of compressed sensing.Then we introduce discrete cosine transform and fast Fourier transform,which are the two kinds of sparse transforms used in the experiments.We also demonstrate the feasibility of reconstructing spectral data by simulations.Then,we use compressed sensing technique to reconstruct the spectra of distributed fiber sensors based on stimulated Brillouin scattering.We achieve the sparse representation of spectral data with discrete cosine transform and fast Fourier transform,and use the greedy iterative reconstruction algorithms such as Orthogonal Matching Pursuit algorithm,Matching Pursuit algorithm and Compressive Sampling Matching Pursuit algorithm,and the convex optimization algorithms such as Basis Pursuit algorithm and the smooth l₀ algorithm to reconstruct the optical spectra.After repeated experiments,we compare the different sparse transforms and algorithms,and find out the best suitable algorithm for stimulated Brillouin scattering based distributed optical fiber sensors.We have concluded that using Orthogonal Matching Pursuit algorithm together with fast Fourier transform,one can reconstruct the original Brillouin spectra with only 25%of the measured data;therefore,the acquisition time is very much reduced in practical applications.Next,we apply the compressed sensing technique from distributed optical fiber sensors to point sensors,we reconstruct the spectral data of Fabry-Perot optical fiber sensors and Bragg grating based optical fiber sensors with compressed sensing technique and compare the different algorithms and sparse transforms to find out best suitable algorithm toward point sensors.We have concluded that using Orthogonal Matching Pursuit algorithm and smooth l₀algorithm with fast Fourier transform,one can reconstruct the spectra of Fabry-Perot optical fiber sensors with only 25%of the measured data;while for Bragg grating based optical fiber sensors,50%of the measured data is needed to reconstruct the original spectra.In optical fiber sensing technology,a large amount of optical fiber sensors are based on spectral measurement to achieve sensing purpose,and generally,it provides better accuracy compared to the sensors based on intensity modulation.Since,the spectral data of the optical fiber sensor are compressible,the compressed sensing technique can be applied to the optical fiber sensors based on spectral measurement.After collecting the spectral data with compressed sampling to reduce the amount of sampled data points,one can use algorithms to reconstruct the original spectrum,so that it is possible to reduce the hardware cost of the optical fiber sensing system,improve information processing efficiency,and reduce propagation error from the fitting process.