摘要
为研究用高光谱数据反演悬浮泥沙质量浓度的方法,以近海悬浮泥沙为实验材料,配制了不同质量浓度悬浮泥沙样品,利用2种方法对样品进行建模:一种是将可见光和近红外光谱进行小波变换并将小波低频系数以偏最小二乘回归建模,另一种为波段组合法.利用交叉检验方法,分析了交叉检验结果,并将2种方法建立的模型进行了比较.研究表明:波段组合法中,TM4与TM1波段反射率比值的指数模型效果最好;在预测能力上,小波偏最小二乘模型总体上比波段组合模型精度高并且稳定性好,适用于悬浮泥沙光谱定量分析.
In order to develop the models to retrieve the suspended sediment concentration information by use of the spectral data, the experiments were carried out on suspended sediment samples （made of materials from coastal suspend sediment） with different concentrations. Two types of models were established to predict the suspend sediment concentration by spectral data. One was the partial least-squares regression （PLSR） model through the wavelet transformation of visible light and near infrared light spectra by use of the wavelet low frequency coefficient. The other was the band composite model. The cross-validation results of the two models were analyzed and compared. The analysis shows that as for the band composite model, the TM4/TM1 reflectance exponential model is the best, while with regard do the prediction capacity, the wavelet-PLS model is of better precision and stability than the band composite model in general, and it is suitable for the quantitative analysis of spectra for the suspend sediment samples.
出处
《河海大学学报（自然科学版）》
CAS
CSCD
北大核心
2008年第6期738-742,共5页
Journal of Hohai University(Natural Sciences)
基金
国家自然科学基金(40606044)
关键词
悬浮泥沙
小波系数
偏最小二乘回归
光谱反射率
suspended sediment
wavelet coefficient
partial least-squares regression
spectral reflectance
