摘要
本研究尝试利用近红外光谱技术测量冬枣的Vc含量,用连续投影算法(SPA)在校正模型中选择有效的近红外光谱波长变量,然后用筛选出的变量建立偏最小二乘(PLS)模型。该模型的预测标准偏差(RMSEP)为0.249 3,预测相关系数(RP)为0.919 7,并将SPA筛选的变量建立的PLS模型与全光谱建立PLS模型结果进行比较。结果表明,SPA优选出全光谱1 557个变量中的24个变量,建立的PLS模型预测效果要好于全光谱建立的PLS模型,SPA能够有效地选取待测成分的特征波长,在冬枣Vc无损检测方面提供理论基础。
The study attempts to detect fresh jujube Vc content by NIRS, Choose effective NIR wavelength variable in adjusted model by continuous projection algorithm （SPA）, then establish partial least squares（PLS） model based on selected variables. The standard deviation of the predicted value（RMSEP） was 0. 249 3, the correlation coefficient （RP） of the predicted value was 0. 919 7, comparative the model based on the variables SPA screening with the model basted on the full spectrum. The results show, 24 variables were preferred from 1 557 full spectrum variables on SPA screening, the predicted value of PLS model established with the 24 variables is better than the predicted value of PLS model established with full - spectrum, the characteristic wavelength can be elected effectively by SPA, this provide a theoretical basis for non - destructive testing in winter jujube Vc.
作者
石鲁珍
张景川
蒋霞
陈杰
白铁成
Shi Luzhen Zhang Jingchuan Jiang Xia Chen Jie Bei Tiecheng(College of Information Engineering, Tarim University, Alar, Xinjiang 843300 Xinjiang Production＆ Construction Corps Key Laboratory of Protection and Utilization of Biological Resources in Tarim Basin, Alar, Xinjiang 843300 College of Mechanic and Electrical Engineering, Tarim University, Alar, Xinjiang 843300)
出处
《塔里木大学学报》
2015年第4期93-98,共6页
Journal of Tarim University
基金
塔里木大学校长基金硕士项目(TDZKSS201413)
关键词
近红外光谱
PLS
SPA
VC
NIRS
least squares（PLS）
continuous projection algorithm （SPA）
Vc
