Research On Recognition Method Of Strong Convective Thunderstorms Based On Satellite Cloud Images

Strong convective weather mainly refers to thunderstorms,strong winds,hail and short-term heavy rainfall and other weather conditions.It is one of the main disastrous weather.The occurrence of strong convective weather has a great impact on economic development and people's production and life.Based on Fengyun 4 meteorological satellite data,the identification of strong convective thunderstorms is studied in this paper.The main contents of this paper are as follows:1.The satellite data from July to September 2018 are preprocessed to obtain satellite images of each channel.The original satellite data is stored in the form of matrix,which is transformed into satellite cloud images through radiation calibration and projection conversion steps.In addition,the radar data are preprocessed to obtain the radar combined reflectance map,and the radar data is used as the verification set.2.Using threshold method to recognize convective cloud clusters in satellite images and extract their characteristics;studying the evolution process of cloud clusters,using area overlap method to match the target cloud clusters at different times to obtain the same sequence of target cloud clusters under time information.In the matching process,firstly,the coarse screening region is set up and the rectangular coordinates outside the target cloud clusters are proposed.The method of relationship greatly shortens the time of searching for matching target.3.The connection of the center of mass in the process of cloud cluster movement is the sequence of cloud cluster movement.The classification and prediction of cloud cluster sequence are based on the bag of words model algorithm.According to the intensity of cloud cluster development,cloud cluster sequence can be divided into two categories: severe thunderstorm sequence and non-severe thunderstorm sequence.According to the development time of cloud cluster sequence,the classification label of cloud cluster sequence is classified.In the process of feature extraction,time information is added to reconstruct the feature,and new 9-dimensional features are obtained,which are three mean,three variance and three slope.After constructing the word bag model,each cloud cluster sequence can be represented by a histogram,and classified and predicted by histogram similarity measure.The prediction information includes the category label,the current moment and the end time.In summary,based on satellite data,the strong convective thunderstorm clouds are studied in this paper.Firstly,the strong convective thunderstorm cloud clusters are identified by threshold method,and the evolution process of cloud clusters is studied.Then,the classification and prediction of cloud cluster sequence based on word bag model algorithm is proposed.It provides a new idea for the study of severe convective weather and theoretical support for weather early warning.