Hail Recognition Index Design Based On Feature Extraction Of Radar Images

Hail is a very severe meteorological disaster caused by strong convective weather system.Its formation should have enough water vapor supply and strong air convectio n.It has a strong randomness,sudden and obvious regional characteristics,although mo st of the range of hail and short duration,but with fierce intensity,and often accomp anied by other disastrous weather,bring huge losses to Our country's economic develo pment.Our country is a country with frequent hail disasters,although meteorological o bservatory can predict hail disaster in advance according to satellite detection and rada r monitoring,but the accuracy is still not high enough.Weather radar,satellite,compu ter and other advanced equipment in a large number of meteorological use,so that th e accuracy of hail prediction greatly improved.The main work of this paper:(1)Collect and hail hail cloud radar images according to their order of mobile cloud storage.The edge of the image contains most of the main information of the im ageIn this paper,the edge extraction of radar reflectivity nephogram is obtained.The e dge extraction graph is obtained,and the advantages and disadvantages of traditional e dge detection operator and CNN edge detection operator are compared,and the advant age of CNN edge detection template is obtained.(2)Edge extraction based on the image texture feature extraction,shape feature extraction and extraction of the original image color features.Texture feature extraction using gray level co-occurrence matrix,obtained in 0°?45°?90°?135°in the four dire ctions of energy,entropy,moment of inertia,correlation,inverse difference moment;sh ape feature extraction using shape feature extraction based on invariant moments,seve n invariant moments to calculate the target image;image extraction the color feature,R,G,B extracted component digital image features,and the image of the one or two and three order color moment.(3)Hail Cloud Radar Digital features obtained by second steps to establish a da tabase,and analyze the characteristics of the hail hail and display graphics data,the d esign of hail identification index.(4)Using Bayes analysis,AdaBoost classifier were tested to hail the selected indicators,analysis results show that the selected indicators in the two methods under the high accurate.This paper first introduces the preprocessing radar hail cloud image,which shows the advantages of the five order cellular neural network edge extraction operator;sec ondly extract digital features of radar image,and establishment of the database;finally using data extraction feature library to analyze the image features,to identify whether the hail clouds.