西北寒旱灌区裸露地表粗糙度SAR反演建模方法研究

发布时间：2018-05-02 12:10

本文选题：地表粗糙度 + 神经网络　；参考：《灌溉排水学报》2017年06期

【摘要】：快速获取大范围土壤地表粗糙度的空间分布是一个急需解决的科学难题。以快速获取内蒙古河套灌区解放闸灌域土壤地表粗糙度为目的,研究了RADARSAT-2雷达影像数据的地表粗糙度信息提取技术,通过剖面板法实测地表粗糙度数据。利用BP(Back Propagation)人工神经网络和LMBP(Levenberg-Marquardt Back Propagation)人工神经网络模型2种方法建立了土壤地表粗糙度的定量反演模型,并对模型进行验证。结果表明,LMBP模型的反演效果优于BP模型,其决定系数R2分别为0.888 3、0.689 2。建立的雷达后向散射系数反演土壤地表粗糙度的人工智能模型,能够在一定程度上满足快速获取土壤地表粗糙度的需要,为微波遥感监测土壤墒情及土壤盐渍化提供重要基础参数。
[Abstract]:It is an urgent scientific problem to quickly obtain the spatial distribution of soil surface roughness. In order to obtain the surface roughness of the irrigated area of Jiefang Gate in Hetao Irrigation area of Inner Mongolia, the surface roughness information extraction technology of RADARSAT-2 radar image data was studied. The surface roughness data were measured by the method of cross-section plate. Two methods, BP(Back propagation artificial neural network and LMBP(Levenberg-Marquardt Back propagation artificial neural network model, are used to establish the quantitative inversion model of soil surface roughness, and the model is verified. The results show that the inversion effect of LMBP model is better than that of BP model, and the determination coefficient R ~ 2 is 0.888 ~ 0.689 ~ 2, respectively. The artificial intelligence model for the retrieval of soil surface roughness by radar backscattering coefficient can meet the need of obtaining soil surface roughness quickly. It provides important basic parameters for monitoring soil moisture and soil salinization by microwave remote sensing.
【作者单位】：内蒙古农业大学水利与土木建筑工程学院测绘工程系;
【基金】：国家自然科学基金项目(51569018,51249007) 内蒙古自然科学基金项目(2013MS0609) “十三五”国家重点研发计划项目(2016YFC0501301)
【分类号】：S127;S156.41

【相似文献】