Application Of The Classification Of Dispersion Trend Of Lighting Whistle In The Calculation Of Electron Concentration

In 2018,China seismo-Electromagnetic Satellite(CSES)was launched from Jiuquan Launch Center.It is very important to extract the hidden lightning whistling waves and calculate the ionospheric electron concentration according to their dispersion trend,which is very important for our future research on ionospheric disturbances and related phenomena in space physics.The unique advantages of ZH-1satellite,such as large dynamic,wide perspective and all-weather,help China's space earth observation technology to stride forward on a new journey,but at the same time,the big data challenges it brings should not be overlooked.Traditional artificial extraction and classification of lighting whistle work often face the problems of poor classification effect and low efficiency.In recent years,with the continuous development of machine learning technology,its strong ability to learn to handle huge amounts of electromagnetic data.In order to further improve classification accuracy and efficiency,this paper uses the support vector machine classifier for lightning whistler dispersion trends for fine classification,The main research work is as follows:(1)In view of the large amount of noise in the time-frequency image of lightning whistle collected in this experiment,the time-frequency images are processed accordingly.One is the Gaussian noise that often appears in the background of time-frequency images,which presents irregular pixels or pixel blocks.Gaussian(2)denoising can be very good to eliminate it.Another kind of complex noise is the platform noise under the image.Because it is roughly the same color as the lightning whistle wave,it will cause computer misjudgment,so that the original non diffusion lightning whistle can be judged as diffusion lightning whistle.To eliminate this kind of noise needs to find a suitable standard,the standard line by cutting at the bottom of the image.Finally,we processed 2303 acoustic images of lightning whistle into a data set.(3)Because we need to classify the extremely special forms of lightning whistle images,the traditional image processing methods can not directly describe the diffusion and non diffusion states with a single feature.Therefore,this paper proposes features according to the waveform change of lightning whistle wave.Finally,only three eigenvalues are extracted from each image,but its accuracy can reach 94.35%,and a good classification effect is achieved.