基于图像识别的人工影响天气业务的研究  

Research on Weather Modification Based on Image Recognition

作　　者：王山海 刘谦[1] 马鑫鑫[1] WANG Shan-hai;LIU Qian;MA Xin-xin(Weather Modification Center of Henan Province,Zhengzhou 450000,China)

机构地区：[1]河南省人工影响天气中心,河南郑州450000

出　　处：《计算机技术与发展》2019年第5期172-177,共6页Computer Technology and Development

基　　金：中国气象局气象关键技术集成与应用项目(CMAGJ2014M33);河南省气象科研项目(Z201508)

摘　　要：在人工影响天气业务中,对人影作业潜力区的划定,目前还没有准确的定量化指标。传统方法依靠人工读图,通过分析卫星云图、雷达回波图、模式资料等来确定适于开展人影作业的潜力区。这些方法对经验依赖性较大,随意性较强,可靠性不高,人影业务的效果受到很大影响。当前基于深度学习的图像识别技术发展迅猛,在学术界和工业界都有了大量应用。深度学习能有效地提取图像中的丰富信息,为解决人影作业潜力区的划定难题提供了新思路和新方向。文中阐述了云雾降水基本机制,分析了卫星云图、雷达回波图在人影领域的应用。介绍了基于深度学习的图像识别原理,分析了三种不同的深度学习结构模型,提出利用卷积神经网络对卫星云图、雷达回波图等海量图片数据进行处理,充分挖掘图片中的重要信息,准确地划定人影作业潜力区,提高人影作业的针对性和时效性,使人影业务发挥更大的作用。In the weather modification business there is no accurate quantitative index for delineation of the potential area of weather modification.Traditional methods rely on manual map reading and determine potential areas suitable for weather modification business by analyzing satellite cloud image,radar echo image and pattern data.They depend on experience with great randomness and low reliability.So the effect of weather modification business has been greatly affected.Currently,image recognition technology based on deep learning has developed rapidly and widely applied in academia and industry.Deep learning can effectively mine the rich information in the image,and provide a new way and new direction to solve the problem of delineating the potential area of weather modification.The basic mechanism of cloud precipitation is expounded,and the application of satellite cloud and radar echoes in the field of weather modification is analyzed.We introduce the principle of image recognition based on deep learning,analyze three different models of deep learning structure,and put forward the use of convolution neural network to deal with mass picture data such as satellite cloud map and radar echo map to determine the potential area of weather modification operation,which can improve the pertinence and timeliness of weather modification and make the weather modification play a greater role.

关 键 词：深度学习 卷积神经网络 图像识别 人工影响天气 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]