Researchon Recognition Technology Of Convective Clouds Based On Himawari-8 Meteorological Satellite Data

Convective weather is a common weather in summer.In the process of its occurrence and development,it is often accompanied by strong air flow,wind shear,ice accumulation,lightning,heavy precipitation,downburst and other phenomena,which has a great impact on the normal operation of civil aviation.Compared with radar data,satellite data has great advantages in timeliness.At the same time,satellites can monitor it in relatively remote areas.This paper mainly identifies convective clouds according to their characteristics in each stage of birth,development and maturity.Firstly,using the data of Himawari-8Meteorological geostationary satellite,the noise of early cloud image is processed by guided filtering,and the image fusion of infrared cloud image and visible cloud image is realized by Laplacian algorithm.In the initial stage of convection,the birth convection is identified mainly according to mb06 algorithm,and the causes of wrong prediction are analyzed.In the rapid development stage,two indexes are added on the basis of rdca algorithm.According to the data set establishment requirements proposed by rdca algorithm,based on the ground lightning monitoring data of the Central Meteorological Observatory,the data set of rapid development convection is established by using the sliding window algorithm,and the random forest algorithm is used to rank the importance of rdca algorithm indexes,so as to improve the rdca algorithm,According to the logistic regression,compared with the original algorithm,the accuracy is improved from 0.703 to 0.752,and the false positive rate is reduced from 0.531 to 0.523.Finally,the cumulonimbus clouds in the mature stage are identified,and the impact on the airport is analyzed in combination with the flight production of the airport.In terms of convection tracking,Canny operator is used to extract cloud information,Hu moment invariant method,area overlap method,cross-correlation algorithm and discriminant cross-correlation filtering algorithm are used to track 30 mesoscale convection.Finally,the cross correlation method is selected.According to the mass center displacement index,the next motion trajectory of cloud cluster is predicted by using linear extrapolation and vector extrapolation methods.