期刊论文

Assessment of Influence of Year and the Oil Content Variability on Robustness of Near Infrared Models for Oil Content of Vernicia fordii Seeds

中文

中国粮油学报

1003-0174

2019

34

6

92-96

10.3969/j.issn.1003-0174.2019.06.016

湖南省教育厅重点项目(14A155).

本校为第一机构

收集了2014和2015年的油桐籽样本,用偏最小二乘法(PLS)分别建立了单一年份、混合年份及单一年份不同含油率范围的油桐籽含油率近红外光谱(NIR)检测模型,并验证.结果 显示单一年份模型对本年份样本有较好预测,而对另一单一年份样本的预测精度明显下降,而混合年份模型对各年份样本都有较好预测;同年份不同含油率范围样本所建模型,含油率范围大,则模型预测精度下降,但稳定性更好.利用竞争性自适应重加权(CARS)算法筛选出30个变量,并结合PLS对混合年份样本建模,既简化了模型,又提高了模型预测性能,验证集相关系数为0.929,均方根误差为1.765,相对标准偏差为3.31％.因此,建立油桐籽含油率NIR检测模型时...

The samples of Vernicia fordii Seeds from 2014 to 2015 were collected. Partial least squares (PLS) was used to establish NIR calibration models for oil content of Vernicia fordii seeds by the samples from single year, hybrid years and the samples of the different oil content range from the same year, respectively, and then the models were used to predict the different validation samples. The results showed that the models developed by the samples from particular year could be effectively applied to the samples from same year but not another yea...