摘要
对内蒙古河套灌区土壤含盐量高光谱曲线进行分析,结果表明通过对原始光谱曲线进行连续统去除变换,可以突出波峰、波谷等重要波段位置,有利于建立逐步回归模型(R2=0.589,RMSE=0.327)对土壤含盐量进行定量反演。通过偏最小二乘法建模,不仅可以利用非正态分布的含盐量数据建模,且模型的精度(R2=0.695)和稳定性(R2=0.756)均优于逐步回归方法。其中,将光谱经过对数和一阶微分变换后,模型结果最佳(R2=0.989,RMSE=0.368)。
By conducting a hyperspectral study of soil salt content in Hetao irrigated area,Inner Mongolia,the results demonstrate that converting the original spectra to continuum reflectance can highlight the important band site i.e.peak and trough of wave,which is beneficial for establishing stepwise regression model （R2 =0.589,RMSE=0.327） for soil salt content＇s quantitative inversion.On the one hand,the partial least square regression method can establish models by means of non-normal data of soil salt content.On the other hand,the established model had higher accuracy （R2 =0.695） and more stability （R2 =0.756） than stepwise regression method.The optimum result （R2 =0.989,RMSE=0.368） was acquired after logarithm and first order differential transform of hyperspectra.
出处
《中国农村水利水电》
北大核心
2014年第4期1-3,6,共4页
China Rural Water and Hydropower
基金
国家"十二五"科技支撑计划课题"农业节水综合技术集成"(2011BAD25B07)
国家自然科学基金(51279142)
中央高校基本科研业务费专项资金(2012206020206)
关键词
高光谱
连续统去除
逐步回归
偏最小二乘
定量反演
hyperspectra
continuum reflectance
stepwise regression
partial least regression
quantitative inversion
