基于BP神经网络的SAR海表风场反演——以长江入海口及临近海域为例  

Retrieval of sea surface wind field of SAR based on BP neural network- Taking the Yangtze River estuary with its adjacent seas as an example

作　　者：胡敏[1,2] 况润元[1,2] 王利花[1,2] 周耀鑫 HU Min;KUANG RunYuan;WANG LiHua;ZHOU YaoXin(Jiang xi University of Science and Technology,Jiangxi Ganzhou 341000;Chengdu University of Information Technology,Sichuan Chengdu 610225)

机构地区：[1]江西理工大学,江西赣州341000 [2]成都信息工程大学,四川成都610225

出　　处：《海洋湖沼通报》2020年第2期35-42,共8页Transactions of Oceanology and Limnology

基　　金：江西省教育厅科学技术研究项目(GJJ160617)资助。

摘　　要：研究基于合成孔径雷达SAR影像数据提出了一种快速反演我国近海表风场的BP神经网络算法。采用2DFFT算法与BP神经网络算法进行海表风向反演结果对比,验证了BP神经网络算法反演海表风向的可行性和直接性。选取2012年2月4日及2012年2月23日两景ENVISAT ASAR影像对我国近海海表风场进行反演,并加以ERA-Interim、NCEP、HY-2微波散射计三种风场产品,进行结果验证。实验表明:BP神经网络算法反演的海表风场与ERA-Interim、NCEP、HY-2微波散射计三种风场产品矢量(风向°,风速m/s)的、CCmax、RESEmax、BIASmax、NRMSEmax分别为(0.8,0.7)、(13.6,1.6)、(10.5,0.46)、(11.4,0.99)、(0.06,0.13)。本文的研究亦可为"21世纪海上丝绸之路"的风能开发、极值风速研究等提供技术支持。Based on the SAR image data, a BP neural network algorithm for quickly retrieving the offshore wind field in China is proposed. The 2 DFFT algorithm and BP neural network algorithm are used to compare the wind direction retrieval results. The feasibility and directness of BP neural network algorithm inversion of sea surface wind direction are verified. The ENVISAT ASAR images on February 4, 2012 and February 23, 2012 were selected to retrieve the offshore wind field in China. Three wind products from the ERA-Interim, NCEP and HY-2 microwave scattering were used for verification of results. Experiments show that the CCmax、RESEmax 、BIASmax 、NRMSEmax between the BP neural network algorithm retrievals and the ERA-Interim, NCEP, HY-2 microwave scatterometer are(0.8,0.7),(13.6,1.6),(10.5,0.46),(11.4,0.99),(0.06,0.13) in term of wind direction(°) and wind speed(m/s). The research can provide technical support for the wind energy development and extreme wind speed research in the project of the "21 st Century Maritime Silk Road".

关 键 词：海表风场 SAR ASAR 2DFFT BP神经网络 

分 类 号：P732[天文地球—海洋科学]