Research And Application Based On Marking Large-scale Time Series Samples

The main work of this paper is about the cell tracking algorithm and the 3D visualization of weather radar data.Based on the above algorithm proposed,the main work and result are as follows:(1)To solve the large amount of labeled samples,an automatic method is proposed that bases on storm segmentation algorithm and cell tracking algorithm.(2)CNN can be used to distinguish common cell and wind cell.In this paper,MARC,liner mesoscale convective system and other features are used to correct the result of CNN Experiments show that the hit rate reaches 82.7% and CSI reaches 64.4% when false positive is 27.7%.(3)We designed and implemented assessment system for hail suppression work.The system function are as follows: a)Based on tracking algorithm,it can provide line graph which shows changes of features before and after work.b)Based on improved marching cubes algorithm,it can provide 3D visualization of hail cell and shell aviation track.(4)Based on weather radar data and convective weather recognition system,we designed and implemented convective weather analysis system that contains following function: a)management of convection sample.b)Expression of cell evolution in convection process.c)Feature distribution of different cell.d)Build and debug of decision tree that can be used to classify sever convection cell and common cell.In summary,the innovations are as follows:(1)Cell tracking algorithm based on optical flow and associated conditions.(2)Improved marching cubes that can be applied to non-Euclidean coordinates.(3)Auto marking for wind cell based on tracking algorithm,which solves the large amount of labeled samples.After analysis,we propose two effective fusion mechanisms for identifying convective winds.(4)Two business platforms: a)convective weather analysis system;b)assessment system for hail suppression work.