Noise Reduction For Lidar Return Signal And The Inversion Of Size Distribution Of Dust Particles Based On Wavelet Analysis

In this paper,we mainly study the Mie scattering lidar signal denoising and the aerosol particle spectrum distribution.Because of the shortcomings of the lidar system,the measured signal-to-noise ratio of the return signal is not high,which seriously affects the study of extinction coefficient,water vapor mixing ratio and polarization.In addition,the aerosol particle spectrum is morbid and not unique,which is caused by the aerosol extinction coefficient equation that belongs to the first kind of Fredholm integral equation.In order to solve these problems,this paper mainly uses the wavelet method to denoise the llidar signal and inversion the aerosol particle spectrum distribution.The research of return signal denoising algorithm is mainly based on self-adaptive wavelet neural network.The orthogonal wavelet basis functions are used as the node functions of hidden layer,searching algorithm is used to select the optimal parameters?thresholds and nodes number of hidden layer in the network.And the fastest decline Levenberg-Marquardt algorithm is selected in the paper.Adjust the whole wavelet neural network by comparing the mean square error of the fitting output and the given mean square error,until the parameter is optimal.What different with other wavelet neural network is that the number of hidden layer nodes is not given but obtained by search algorithm.Constructed the The adaptive BP wavelet neural network by this method is adjusted adaptively with the change of external environment.In the study of the inversion of aerosol particle spectral,firstly,assuming the particle is spherical and calculate the extinction coefficient and scattering coefficient by Mie theory;secondly,measure the optical thickness with CE-318;and finally,combining wavelet Galerkin and Tikhonov regularization algorithm to solve size distribution of aerosol particle.Comparing with the traditional iterative algorithm,this method is convenient ?simple and calculate easily,which uses wavelet Galerkin method to discretize the relation between the optical thickness and the gas particle spectrum into linear equations,then useTikhonov regularization method to calculate the regular solution.Which can overcome the instability and non uniqueness of the solution of the equation.In order to verify the feasibility of the two methods,the former compare with the wavelet threshold denoising algorithm,the latter select typical optical thickness to inversion aerosol particle spectrum in yinchuan area from March.2015 to December.2016,and comparing with actual weather conditions.The results show that the self-adaptive wavelet neural network denoising algorithm is superior to the wavelet threshold denoising;and the weather conditions is analyzed by using wavelet Galerkin method to inversion the aerosol particle spectrum in Yinchuan,which is matched with historical data of Meteorological Bureau.