Research On Key Steps Of Interferogram Preprocessing For Doppler Asymmetric Spatial Heterodyne Interferometer

With the increasing development of science and technology and the deepening of human exploration of the universe,the study of wind field in the middle and upper atmosphere has important application value and significance in many fields.Doppler asymmetric spatial heterodyne spectrometer(DASH)is developed from spatial heterodyne spectroscopy(SHS)technology,which has great application value in atmospheric wind field detection.The DASH for atmospheric wind field detection is based on the principle of optical Doppler,which can deduce the velocity of particles by detecting the airglow spectrum's Doppler frequency shift,and then the atmospheric wind speed can be obtained.The degree accuracy of wind speed inversion can be directly affected by the quality of the interferogram.The Doppler frequency shift of airglow spectrum caused by atmospheric wind speed is very small(10^-8?10^-6wavelength).Therefore,it is necessary to eliminate the noise,drift and error caused by instrument and environment in the retrieval process of interferogram.In this thesis,the key steps of interferogram preprocessing of DASH are studied.This paper briefly introduces the necessity of wind field detection in the middle and upper atmosphere and the advantages of DASH in atmospheric wind field detection.The domestic and foreign development status of DASH and its interferogram preprocessing are reviewed.The basic working principle of DASH is analyzed.First of all,the system structure of DASH and its basic principle for detecting atmospheric wind field are introduced.This paper analyzes the phase retrieval algorithm of atmospheric wind speed of DASH,and the retrieval process is divided into five stages:interferogram preprocessing,interferogram Fourier transform,target spectral line extraction,inverse Fourier transform and phase calculation.The causes and characteristics of various noises,drifts and errors of interferogram are studied,and the methods of interferogram preprocessing are given.The main steps of preprocessing include low frequency baseline removal,dark current noise removal,spike removal,pixel transmission effect removal,flat field,phase shift correction.Combined with the characteristics of interferogram and the cause and expression of spike,based on the phase retrieval model of DASH,the influence of spike on the accuracy of phase retrieval is theoretically deduced,and the theoretical model of retrieval phase error and various characteristic parameters of spike is given.The simulation study is carried out with the experimental data of DASH.The results show that the retrieval phase error changes periodically with the position of the spike,and is positively correlated with the intensity and peak width of the spike.The interference data with spike measured by the laboratory DASH is compared with the theoretical model,and the experimental results are fond to the theoretical model,which verifies the theoretical model of the impact of spike on the inversion retrieval.At the same time,it is proved that the impact of spike on the retrieval phase is serious and the necessity of spike correction is also proved.Based on the characteristics of interference data and spike of DASH,the principles and advantages and disadvantages of several commonly used spike de-noising methods,such as mean filtering,median filtering and global threshold method,are analyzed.A method combining the nearest neighbor comparison method and the switching median filtering method is proposed to locate and correct the spike of the DASH data.Using the interference data collected by the laboratory DASH under different signal-to-noise ratio,the spike of interferogram is corrected by the above methods.Compared with other methods,this method can accurately locate and correct the spike without impacting on the original interferogram,and can correct the phase error caused by the spike.When the method is used in different period and signal-to-noise ratio interference data,the correction of retrieval phase error is more than 90%,which is universal for different interference data.