首页 | 官方网站   微博 | 高级检索  
     

改进变分光流法并行算法实现
引用本文:王志斌,肖艳姣,吴涛,崔方.改进变分光流法并行算法实现[J].计算机应用与软件,2019,36(1):105-110.
作者姓名:王志斌  肖艳姣  吴涛  崔方
作者单位:中国气象局武汉暴雨研究所 湖北武汉430205;武汉中心气象台 湖北武汉430074;中国电力科学研究院新能源研究中心 江苏南京210009
基金项目:国家重点研发计划;湖北省雷电专项;湖北省气象局重点课题;国家电网有限公司总部科技项目
摘    要:通过天气雷达回波并使用变分光流方法的改进算法对雷达数据进行风场反演,计算回波的光流场变化而得到运动的矢量场。把Horn&Schunck方法和Lucas&Kanade方法结合在一起,加入高阶平滑项,再运用九点平滑算法得到矢量场。外推预报采用Semi-Lagrange方法进行。实验表明,该方法的效果优于Horn&Schunck方法和Lucas&Kanade方法。通过Join-Fork模式运用共享内存处理并行算法实现了大范围三维矢量场的快速计算,其加速比可达5. 05,减少实时业务所需时间。

关 键 词:天气雷达  临近预报  变分方法  光流方法  并行计算

IMPLEMENTATION OF IMPROVED VARIATIONAL OPTICAL FLOW PARALLEL ALGORITHM
Wang Zhibin,Xiao Yanjiao,Wu Tao,Cui Fang.IMPLEMENTATION OF IMPROVED VARIATIONAL OPTICAL FLOW PARALLEL ALGORITHM[J].Computer Applications and Software,2019,36(1):105-110.
Authors:Wang Zhibin  Xiao Yanjiao  Wu Tao  Cui Fang
Affiliation:(Institute of Wuhan Heavy Rain,China Meteorological Administration,Wuhan 430205,Hubei,China;Wuhan Central Meteorological Observatory,Wuhan 430074,Hubei,China;Renewable Energy Center,China Electric Power Research Institute,Nanjing 210009,Jiangsu,China)
Abstract:A wind field inversion was conducted by means of weather radar echo and an improved variational optical flow method,and the motion vector field was obtained by calculating the optical flow changes of echo.We combined the Horn&Schunck method with Lucas&Kanade method,in which the higher order smoothing term was added.The vector field was obtained by using the nine dot smoothing algorithm.The Semi-Lagrange method was adopted for extrapolation prediction.The experimental results show that the effect of the improved method is better than Horn&Schunck and Lucas&Kanade method.Shared memory process parallel algorithm is used in Join-Fork mode to realize the fast calculation of a large range of three-dimensional vector field.Experiments show that the speedup of the parallel algorithm can reach 5.05,which reduces the time required for real-time services.
Keywords:Weather radar  Nowcasting  Variational method  Optical flow method  Parallel computing
本文献已被 维普 万方数据 等数据库收录!
设为首页 | 免责声明 | 关于勤云 | 加入收藏

Copyright©北京勤云科技发展有限公司    京ICP备09084417号-23

京公网安备 11010802026262号