FDNet:A Deep Learning Approach with Two Parallel Cross Encoding Pathways for Precipitation Nowcasting

作     者：闫碧莹 杨超 陈峰 Kohei Takeda Changjun Wang Bi-Ying Yan;Chao Yang;Feng Chen;Kohei Takeda;Changjun Wang

作者机构：University of Chinese Academy of SciencesBeijing 100049China Institute of SoftwareChinese Academy of SciencesBeijing 100190China School of Mathematical SciencesPeking UniversityBeijing 100871China Peng Cheng LaboratoryShenzhen 518052China Guiyang Academy of Information TechnologyGuiyang 550081China NTT DATA CorporationTokyo 163-8001Japan NTT DATA Institute of Management Consulting Inc.Tokyo 163-8001Japan 

出 版 物：《Journal of Computer Science & Technology》 (计算机科学技术学报（英文版）)

年 卷 期：2023年第38卷第5期

页      面：1002-1020页

核心收录：

学科分类：12[管理学] 1201[管理学-管理科学与工程(可授管理学、工学学位)] 081104[工学-模式识别与智能系统] 08[工学] 0835[工学-软件工程] 0811[工学-控制科学与工程] 0812[工学-计算机科学与技术（可授工学、理学学位）] 

基　　金：supported in part by the National Key Research and Development Program of China under Grant No.2018YFC0831500 the Beijing Natural Science Foundation under Grant No.JQ18001,and the Beijing Academy of Artificial Intelligence 

主　　题：spatio-temporal predictive learning precipitation nowcasting neural network 

摘      要：With the goal of predicting the future rainfall intensity in a local region over a relatively short period time,precipitation nowcasting has been a long-time scientific challenge with great social and economic *** radar echo extrapolation approaches for precipitation nowcasting take radar echo images as input,aiming to generate future radar echo images by learning from the historical *** effectively handle complex and high non-stationary evolution of radar echoes,we propose to decompose the movement into optical flow field motion and morphologic *** this idea,we introduce Flow-Deformation Network(FDNet),a neural network that models flow and deformation in two parallel cross *** flow encoder captures the optical flow field motion between consecutive images and the deformation encoder distinguishes the change of shape from the translational motion of radar *** evaluate the proposed network architecture on two real-world radar echo *** model achieves state-of-the-art prediction results compared with recent *** the best of our knowledge,this is the first network architecture with flow and deformation separation to model the evolution of radar echoes for precipitation *** believe that the general idea of this work could not only inspire much more effective approaches but also be applied to other similar spatio-temporal prediction tasks.