变量消去
偏最小二乘回归
校准
近红外光谱
分光计
均方误差
稳健性(进化)
计算机科学
决定系数
数学
生物系统
人工智能
统计
化学
生物
光学
物理
生物化学
推论
基因
作者
Ziniu Zhao,Yihan Liu,Yang Shuo,Yurong Li,Yeshun Zhang,Hui Yan
标识
DOI:10.1016/j.infrared.2023.104818
摘要
Fast detection of the tenderness of mulberry leaves is an important process for the feeding of juvenile silkworms because the old leaves can damage their mouthparts, and it is a challenge for the sericulture industry. In this study, the near-infrared spectra of mulberry leaves were recorded by a portable near-infrared (NIR) spectrometer, and the spectral data were pretreated by the first derivative (1st Der), standard normal variate (SNV), and their combination. Competitive adaptive reweighted sampling (CARS), random frog (RF), and uninformative variable elimination (UVE) were used to select the sensitive variables to improve the model performance, and then partial least squares (PLS) regression was applied to establish the calibration model. The results showed that the combination of 1st Der and SNV was the optimal spectrum pretreatment method, and the calibration model with the variables selected by UVE has the highest calibration performance. The model was validated by the unknown samples, and it demonstrated robustness with the root mean square error of the prediction set (RMSEP) 2.59 and the coefficient of determination of the prediction set (R2P) 0.7105. Therefore, it is feasible that the portable NIR spectrometer can be used for the detection of the tenderness of mulberry leaves on-site and in-field, and it will be beneficial for the sericulture industry.
科研通智能强力驱动
Strongly Powered by AbleSci AI