高频地波雷达海洋回波谱的Bragg峰和二次谐波峰中，均蕴含着浪高信息，因此发展出了基于Bragg峰功率(Power of Bragg Peak，PB)、二次谐波峰与 Bragg 功率比(Power Ratio of Second Harmonic Peak to Bragg Peak，RSB)、双频 Bragg 峰功率比(Power Ratio of Dual Frequency Bragg Peak，RDB)等浪高反演算法。然而，三种算法均无法实现不同海况、不同距离下浪高的精确反演。本文研究发现，PB算法适用于低海况，RSB算法在近距离高海况下表现良好，而RDB算法适用于远距离测量，即三种算法高性能测量区间存在强互补性。在此基础上，本文提出了一种基于多特征(PB、RSB和RDB)融合的浪高反演算法，其中采用反向传播(Back Propagation，BP)神经网络作为特征融合器。实验表明:本文算法在测量精度、波高适应范围、距离适应范围上均优于现有波高反演算法。
Wave Height Inversion Algorithm Based on Multi-Feature Fusion of HF Radar Echo Spectrum
The Bragg peaks and second harmonic peaks(SHPs)in the ocean echo spectrum of high frequency surface wave radar contain wave height information.Therefore,the inversion algorithms based on the power of Bragg peak(PB),the power ratio of SHP to Bragg peak(RSB)and the power ratio of dual frequency Bragg peak(RDB)have been proposed.However,the three algorithms can't measure wave height accurately under different sea states and different distances.In this paper,it is found that the PB-based algorithm is suitable at low sea state,the RSB-based algorithm performs well at high sea state in near distance,and the RDB-based algorithm is suitable in long distance.Thus,their suitable ranges are complementarity.On this basis,this paper proposes a wave height inversion algorithm based on multiple features fusion and BP neural network.Experiment data shows that this algorithm is superior to the traditional algorithms in measuring accuracy,adaptive ranges of wave height and distance.
high frequency surface wave radarwave height inversionecho analysismulti-feature fusionBP neural network
NETL
NSTL
万方数据
