Research On Denoising Algorithm Of BOTDA System Based On Noise Level Estimation

This dissertation summarizes the domestic and international research situations of Brillouin Optical Time Domain Analysis(BOTDA)technology and signal-to-noise improving methods.Due to the limitation of low SNR,the traditional BOTDA system cannot meet the requirements of high accuracy and long sensing distance in large-scale engineering.Therefore,the SNR play a key role in the performance of BOTDA system.It has become one of the urgent problems of BOTDA system that how to further reduce the influence of system noise on the detection signal,ensure the measurement accuracy and reduce the data processing time.The cumulative average and the Non-Local Means(NLM)image processing approaches are used to removing the noise from data acquired by temperature sensing experiment based on RBOTDA,and the denoising behavior of NLM affected by SNR of raw sensing image is analyzed.The results show that the noise reduction effect of NLM algorithm is better than the cumulative average,and data processing time is short.Only when the SNR of BOTDA system is above 0d B,the NLM algorithm can avoid the distortion of Brillouin Gain Spectrum(BGS)and effectively recover the data.By numerical simulation study,we analyze the effectiveness and adaptability of the three noise level estimation algorithms and propose to apply the noise level estimation algorithm based on flat blocks and local statistics to the BOTDA system,thereby setting the key parameter noise level of the NLM algorithm.Particle Swarm Optimization(PSO)is used to adaptively select the parameters of NLM algorithm based on noise level estimation,and the effectiveness of the proposed algorithm is verified by temperature sensing experiment based on R BOTDA.It indicates that the NLM algorithm based on noise level estimation of flat patches and local statistics achieves a signal-to-noise ratio from 1.3485 d B to 19.2576 d B and the SNR is improved obviously.Integral image is used to accelerate the proposed algorithm and the data processing time is reduced to 30%.Finally,the Lorentz Curve Fitting(LCF)is performed on the denoised data and the fitting results are basically consistent with the experimental condition.It is proved that the proposed algorithm can effectively improve system SNR,ensure well BFS accuracy and reduce the average number of BOTDA.