Precipitation And Thunderstorm Identification In Weather Radar Images For Civil Aviation Based On CNN

Safety is an eternal theme in civil aviation.Since each process of flying can be affected by meteorological conditions,there is a close relationship between flying safety and meteorological conditions.Civil aviation meteorology becomes an essential part of modern civil aviation,which devotes to reducing and minimizing the influence on safety and efficiency from weather.With a dramatically increasing in China's civil aviation transportation,the total number of flights each year is going higher,meteorological work becomes more and more important.Along with the development of the aviation in the last few decades,civil aviation meteorological works in China have made much progress.Meteorological equipments have been improved.A large number of advanced equipments are brought into work in recent years especially the Doppler weather radar.The practical application of Doppler weather radar improves the safety and efficiency of flying under thunderstorm weather.Till now data of the weather radar becomes an indispensable intelligence to ensure flying safety.Monitoring the operation of Doppler weather radar also becomes a momentous work for civil aviation meteorological engineers under thunderstorm weather.This article expounds the basic principles of weather radar and its traditional method of measuring quantitative precipitation.Collecting pinches of clouds scanning images and rainfall data from “ Yi-zuo-yun ” weather radar station(belongs to Yunnan Air Traffic Management Sub-bureau)in the last three years.After filtering and preprocessing,comparing the features of radar echo,actual measurement results with manual prediction results,trying to predict the area where there is going to generate new thunderstorms and make date sets.Training Faster R-CNN and SSD convolution neural network models(which are more general used than other models)with observed date.Trying to identify rainfall,thunderstorm area and rain intensity in one radar image by computer deep learning.Comparing the experiment results from these two models.Choosing radar images and observed date of Kunming on July 20 th 2019(with super strong precipitation)to verify the feasibility and applicability of these two convolutional neural network models.Discussing the applicability and feasibility of these two models.Prospecting the application of this method in future civil aviation meteorological work.