Fast and high-accuracy temperature extraction of BOTDR sensor based on wavelet convolutional neural network

We propose an instantaneous temperature measurement method based on wavelet convolutional neural network (wavelet-CNN) to extract Brillouin frequency shift (BFS) in Brillouin optical time domain reflectometer (BOTDR) sensor and map the BFS to the temperature. Compared to Lorentzian curve fitting (LCF), both the simulation and experimental results show the wavelet-CNN has better accuracy and shorter processing time.