基于高光谱数据的橡胶树叶片氮素含量估算模型研究

分别采用光谱指数法和偏最小二乘法（PLS）构建了橡胶树叶片氮素含量高光谱估算模型。结果表明,光谱指数模型中,基于NDSI（R1025,R736）建立的二次多项式模型最佳,调整R2达到0.792。基于不同Savitzky-Golay（SG）平滑模式预处理结果建立的PLS模型,估算精度差异明显,最优的SG平滑模式为零阶导数、53点平滑点数、二次多项式。利用测试集对光谱指数模型和PLS模型进行验证,结果显示,均方根误差RMSE分别为1.772和1.432 g/kg,相对分析误差RPD分别为2.598和3.215,平均相对误差ARE分别为5.189%和4.057%,PLS模型具有较高的估算精度。...

Full description

Saved in:
Bibliographic Details
Published in西南农业学报 Vol. 28; no. 2; pp. 569 - 574
Main Author 黎小清 刘忠妹 丁华平 许木果 李春丽
Format Journal Article
LanguageChinese
Published 云南省热带作物科学研究所,云南景洪,666100 2015
Subjects
估算模型
偏最小二乘法
橡胶树叶片
氮素
高光谱
Rubber tree leaves
Estimation models
偏最小二乘法
Hyper-spectral
Nitrogen
估算模型
Partial least squares
高光谱
橡胶树叶片
氮素
Online AccessGet full text
ISSN1001-4829
DOI10.16213/j.cnki.scjas.2015.02.022

Cover

More Information
Summary:分别采用光谱指数法和偏最小二乘法（PLS）构建了橡胶树叶片氮素含量高光谱估算模型。结果表明,光谱指数模型中,基于NDSI（R1025,R736）建立的二次多项式模型最佳,调整R2达到0.792。基于不同Savitzky-Golay（SG）平滑模式预处理结果建立的PLS模型,估算精度差异明显,最优的SG平滑模式为零阶导数、53点平滑点数、二次多项式。利用测试集对光谱指数模型和PLS模型进行验证,结果显示,均方根误差RMSE分别为1.772和1.432 g/kg,相对分析误差RPD分别为2.598和3.215,平均相对误差ARE分别为5.189%和4.057%,PLS模型具有较高的估算精度。
Bibliography:The hyper-spectral estimation model of rubber tree leaf nitrogen was established using spectral index and partial least squares（ PLS）,respectively. The results showed that the quadratic polynomial regression model based on NDSI（ R1025,R736） was the best in all spectral index models,and the adjusted R2 was 0. 792. Significant differences were observed in the estimation accuracy of the PLS models based on results of preprocessing in various Savitzky-Golay（ SG） smoothing models. Optimal SG smoothing model was 0 order derivative,53 smooth points number and quadratic polynomial. The Spectral index model and PLS model were validated by test set,and the results showed that the RMSE were 1. 772 and 1. 432 g / kg,the RPD were 2. 598 and 3. 215,and the average relative error were 5. 189 % and 4. 057 %,respectively. The PLS model has the highest estimation accuracy.
51-1213/S
LI Xiao-qing, LIU Zhong-mei, DING Hua-ping, XU Mu-guo, LI Chun-li （Yunnan Institute of Tropical Crops,Yunnan Jinghong 666100,China）
Hyper-spectral Ru
ISSN:1001-4829
DOI:10.16213/j.cnki.scjas.2015.02.022