Bidirectional reflectance distribution function modeling of one-dimensional rough surface in the microwave band

In this study, the bidirectional reflectance distribution function(BRDF) of a one-dimensional conducting rough surface and a dielectric rough surface are calculated with different frequencies and roughness values in the microwave band by using the method of moments, and the relationship between the bistatic scattering coefficient and the BRDF of a rough surface is expressed. From the theory of the parameters of the rough surface BRDF, the parameters of the BRDF are obtained using a genetic algorithm. The BRDF of a rough surface is calculated using the obtained parameter values. Further, the fitting values and theoretical calculations of the BRDF are compared, and the optimization results are in agreement with the theoretical calculation results. Finally, a reference for BRDF modeling of a Gaussian rough surface in the microwave band is provided by the proposed method.     

Bidirectional reflectance one-dimensional rough distribution function modeling of surface in the microwave band

In this study, the bidirectional reflectance distribution function （BRDF） of a one-dimensional conducting rough surface and a dielectric rough surface are calculated with different frequencies and roughness values in the microwave band by using the method of moments, and the relationship between the bistatic scattering coefficient and the BRDF of a rough surface is expressed. From the theory of the parameters of the rough surface BRDF, the parameters of the BRDF are obtained using a genetic algorithm. The BRDF of a rough surface is calculated using the obtained parameter values. Further, the fitting values and theoretical calculations of the BRDF are compared, and the optimization results are in agreement with the theoretical calculation results. Finally, a reference for BRDF modeling of a Gaussian rough surface in the microwave band is provided by the proposed method.     

支持向量机和神经网络在粗糙面参数反演中的比较

首先介绍支持向量机和神经网络方法及其在内部网络训练上的不同.分别利用支持向量机和神经网络对高斯粗糙面的均方根高度和相关长度进行反演.通过仿真结果和误差对比分析,发现在小样本情况下,支持向量机的反演结果比神经网络好,而在具有大量样本的情况下,神经网络的反演精度有显著提高,而且反演时间比支持向量机少很多.