基于全极化SAR数据反演鄱阳湖湿地植被生物量  

Retrieval of Wetland Vegetation Biomass in Poyang Lake Based on Quad-polarization Image

作　　者：刘菊[1,2] 廖静娟[1] 沈国状[1] 

机构地区：[1]中国科学院对地观测与数字地球科学中心,北京100094 [2]中国科学院研究生院,北京100049

出　　处：《国土资源遥感》2012年第3期38-43,共6页Remote Sensing for Land & Resources

基　　金：中国科学院对地观测与数字地球科学中心主任科学基金项目(编号:Y1ZZ05101B)资助

摘　　要：鄱阳湖是中国最大的淡水湖,也是国际重要湿地,对其生物量进行长期、定量研究有助于加深对区域乃至全球碳平衡的认识和理解。探讨了利用全极化Radarsat-2 C波段数据反演鄱阳湖湿地生物量的方法,改进了基于辐射传输理论的植被冠层散射模型,模拟了C波段湿地植被的后向散射特性;应用极化分解技术,增加了神经网络训练数据,并用后向反馈神经网络(BP)算法,反演了鄱阳湖湿地植被生物量。与野外实测生物量比较的结果表明:将改进的植被冠层散射模型和全极化分解得到的后向散射系数引入BP神经网络算法,能够有效降低生物量反演误差;全极化SAR数据在生物量反演中具有广阔的应用前景。The Poyang Lake is the largest freshwater lake in China as well as an internationally important wetland. Long -term quantitative study of vegetation biomass in this area helps deepen our understanding of regional and global carbon balance. The authors investigated the approach and method of Radarsat - 2C - Band quad - polarization imagery for biomass retrieval in wetland vegetation. The vegetation canopy scattering model was modified and used to simulate the backscattering characteristics. Polarization decomposition was adopted to prepare the testing data with the model output for BP neural network. After obtaining the retrieval values of vegetation biomass, the values were compared with the filed - measured values. The results show that the introduction of the output data of vegetation canopy scattering model and polarimetric decomposition technique to the BP neural network algorithm could reduce the retrieval error effectively, and that the Quad - polarization imagery has broad application prospect in the field of biomass retrieval.

关 键 词：生物量 植被冠层散射模型 全极化分解 BP神经网络 RADARSAT-2 

分 类 号：TP79[自动化与计算机技术—检测技术与自动化装置]