基于神经网络的双频雷达海面风速反演方法研究

High frequency ground wave radar is an effective technique to observe the dynamic parameters of ocean surface in real-time. The real-time monitoring of ocean current based on this device has realized commercial operation. However, there are still some problems in the monitoring of the waves and the ocean surface wind field, such as how to improve the distance of the detection and how to improve the accuracy of the inversion results, etc. At present, the application of radar data is mainly focused on the extraction of related ocean information features of a Bragg frequency point. In order to improve the efficiency and accuracy of the acquisition of sea waves and wind speed, a comprehensive analysis of Bragg scattering information based on more frequency points is needed. According to the needs of the above research, a method of extracting sea surface wind velocity using first order scattering echo is proposed in this paper. This project is based on the dual frequency marine radar system developed recently, and combines the 40 years small scale hindcast model data in Europe. Based on the auxiliary correction of the hindcast mode data, combined with the methods of neural networks, the nonlinear mapping relationship between the Bragg wave spatial diffusion characteristics and the wind speed will be inversed through the trained network, in this way, the coverage in space and accuracy of wind speed would be further improved from the HF radar backscatter signals.

高频地波雷达是一种能够实时监测海洋表面动力学参数的先进手段，目前海流监测应用方面已经商业化运行。但实现更精确的海洋信息提取和分析海仍然存在较多尚待解决的问题，比如海洋表面风速反演主要依赖于二阶谱，存在着探测距离近，低风速反演能力不足等问题。基于上述问题中风速获取方法的研究需要，本项目提出了一种利用多频点Bragg散射信息分析的一阶散射回波海面风速提取方法。本研究以双频海洋雷达风速反演应用为应用背景，结合欧洲HIPOCAS的40年小尺度后报模式数据，该数据包含了完整的海面风速、风向以及有向波浪谱数据，可以提取分析特定频点上的波浪谱数据。通过该后报模式数据分析研究，并结合复杂时延神经网络的工具，进而研究双频点的Bragg波浪空间扩散特性与风速之间的映射关系。利用该映射，在基于雷达实测数据的反演中，进一步提升风速测量的空间范围和精度。

高频地波雷达是一种能够实时监测海洋表面动力学参数的先进手段，目前海流监测应用方面已经商业化运行。但实现更精确的海洋信息提取和分析海仍然存在较多尚待解决的问题，比如海洋表面风速反演主要依赖于二阶谱，存在着探测距离近，低风速反演能力不足等问题。基于上述问题中风速获取方法的研究需要，本项目提出了一种利用多频点Bragg散射信息分析的一阶散射回波海面风速提取方法。本研究以双频海洋雷达风速反演应用为应用背景，结合欧洲HIPOCAS的40年小尺度后报模式数据，该数据包含了完整的海面风速、风向以及有向波浪谱数据，可以提取分析特定频点上的波浪谱数据。通过该后报模式数据分析研究，并结合复杂时延神经网络的工具，进而研究双频点的Bragg波浪空间扩散特性与风速之间的映射关系。利用该映射，在基于雷达实测数据的反演中，进一步提升风速测量的空间范围和精度。