Research On Tracking And Extrapolation Of Rain Clouds Based On Doppler Radar Images

Nowcasting is described as current weather conditions and 0 to 2 hours weather forecast,it is a local emergency prevention services of severe storms and other severe weather forecasting which gives benefits to socio-economic.Short-term nowcasts of disaster weather and automatic weather forecasting business in China have long been the urgent need for improving weather forecasting and disaster prevention.Although short-term forecasts have achieved good results,but there is still a big gap with foreign advanced technology.Firstly the paper introduces working principle,structure and data of Doppler radar,then discusses Z-R relationship between reflectivity and rainfall intensity,analysis centroid method,SCIT and TITAN on tracking storm cell,then do research on cross-correlation method to calculate the general weather precipitation echoes,at the situlation of echoes changing rapidly,conclude through experimentation that traditional cross-correlation method will produce disorder vectors but cross-correlation method based on difference image can get consistent motion vector,in order to improve the shortcoming of cross-correlation method when calculating echoes rapid evoluted,the paper introduce optical flow method in image processing,as for strenuous exercise between typhoon radar echo images lead L-K optical flow constraints can not be met,the paper improves L-K optical flow method,adding pyramid technology when caluculate large movement between radar echo images.The above methods are used to calculate the vectors of radar images as foundation for extrapolation forecast.The paper then study linear prediction of nowcasting,for the case of linear prediction can not handle the change of echo intensity,the paper analyzes the method of calculating the reflectivity variety of centroid position in linear extrapolation,experiment verificates that the method can improve echoes extrapolation accuracy at the case of generation or disappear.In the case of rotation vector in linear prediction,the paper uses two-dimensional time layer Semi-Lagrangian method to calculate,paper improves the implementation strategy when calculating the initial reflectivity,reducing number of interpolation to once and using spline interpolation method at the same time,it will effectively improve computational efficiency and reduce time for warning response when dealing with the case of slow intensity change between typhoon radar echoes.This paper also achieve tracking rain clouds.Pretreatment is to remove the regions in radar images which can not form rainstorm in a short time.The method for segmentation uses a region-based active contour model,level set function will firstly be binarized and then uses Gaussian filter method to smooth,which avoids complicated initialization calculating of signed distance function of C-V model,combined with the GAC and the C-V model to achieve local and global segmentation at the same time,which can obtain smoothly and effectively edges.Tracking is based on the character that there is just a little movement between rain clouds,adding area item to the area overlap method to achieve,it can achieve cloudy groups tracking.At the situations of clouds split or merge,generation or disappearance etc,it can track effectively either.Finally,the paper also discusses the use of rain clouds tracking when analyzing future weather,and analysis the prospect of automatic weather forecasts at the same time.