Research On Intelligent Retrieving Algorithms And Experiment Of Mie Scattering Lidar

Mie scattering lidar is a powerful tool for detecting the optical properties of atmospheric aerosols.However,there are two unknown parameters in the Mie scattering lidar equation,extinction coefficient and backscattering coefficient.In the mean time,the lidar echo signal is non-stationary,it is very difficult to eliminate the noise in the lidar echo signal.The two problems listed above are critical for Mie scattering lidar data retrieving.According to the statistical properties of the noise in the lidar echo signal,an adaptive filter is designed to achieve optimal filtering.Without making assumptions on the relationship between extinction coefficient and backscattering coefficient,the BP network optimized by genetic algorithm is proposed to solve the Mie scattering lidar equation and retrieve aerosol extinction coefficient from Mie scattering lidar signal finely.The research is of great significance for the fine retriving of the Mie scattering lidar signal.The statistical properties of the noise in the Mie scattering lidar echo signal is analyzed by statistical hypotheses testing method.Based on this,an adaptive f ilter is proposed to eliminate the noise.The least mean square error algorithm is used to achieve optimal filtering,in which the MSE is minimized by adjusting the filter's weight matrix.The validity of the adaptive filter is verified by numerical simulation and experimental data retrieving.In numerical simulation,the output SNR of the adaptive filter is larger than that of the wavelet transform filter,and the MSE of the adaptive filter is less than that of the wavelet transform filter.In experimental data retrieving,the filtered Mie echo signals of adaptive filter and wavelet transform filter are used to retrieve extinction coefficient respectively under different weather conditions.The amplitude of the ripples in the extinction coefficient profile of adaptive filter is less than that of wavelet transform fi lter.The details of extinction coefficient are displayed more clearly in the profile of adaptive filter.The feasibility of BP network for solving the Mie scattering lidar equation is studied in depth.The structure and ain p arameters of BP network are designed.The BP network is trained with the Mie scattering lidar echo signal and the extinction coefficient retrieved by Raman method.Then the m athematical relationship between the Mie scattering lidar echo signal and the extinction coefficient is stored in the BP network.The experimental results show that when the original echo signal is selected as input signal,the network structure is selected as 26-28-26,the activation function is selected as tansig-tansig,the expected error is selected as 0.01,and the learning rate is selected as 0.1,the BP network retrived extinction coefficient is of high accurance.When the number of training samples increases,the retrieve accuracy and convergence speed of the BP network increase firstly and then tend to be stable.In order to banlance the contradiction between the inversion accuracy and computing load,the number of training samples is set to 900.Finally,the number of hidden layer units is determined by the impro ved empirical formula method.In order to improve the convergence rate and prevent falling into local minimum,the initial weights and thresholds of BP network are optimized b y genetic algorithm.The experimental results show that,the GA-BP network has higher convergence speed and higher retrieve accuracy than the BP network unoptimized.Finally,the aerosol extinction coefficient is retrieved by GA-BP network,Raman method and Fernald method respectively under different time and weather conditions.And the applicability of GA-BP network for retrieving extinction coefficient is analyzed.The experimental results show that,except cloudy weather,the GA-BP network can retrieve extinction coefficient finely.