期刊论文

Rapid Discrimination of Maturity of Tobacco Leaf Based on Near-infrared Spectroscopy and Random Forest

中文

西南农业学报

1001-4829

2017

30

4

931-936

10.16213/j.cnki.scjas.2017.4.035

湖南省研究生科研创新项目“烟草烘烤过程近红外光谱在线无损监测及变化规律的研究”（CX2015B237）；

本校为第一机构

【目的】成熟度是烟叶品质的中心因素,田间成熟是获得优质烟叶的前提和基础,如何准确判断烟叶成熟程度是一个难题。【方法】本文利用近红外光谱技术结合化学计量学方法对不同成熟度的新鲜烟叶进行了探讨。【结果】随机森林(RF)分类模型参数优化简单、泛化能力强、预测结果较好,能有效识别不同成熟程度的烟叶样本,实现烟叶成熟度的快速判别,上、中和下部烟叶不同成熟度模型预测集的分类正确率分别为0.9231、0.90和0.9091,预测正确率优于主成分分析(PCA)、K最近邻(KNN)和支撑向量机(SVM)等方法。【结论】因此,近红外光谱技术结合随机森林方法简单、快速、准确,可为客观辨别烟叶成熟度的等级、优劣...

[ Objective ] Maturity is the center of the tobacco leaf quality factors, and the maturity in field is the promise and foundation to obtain high quality tobacco leaf, but how to accurately determine tobacco maturity is a difficult problem. [ Method ] A new method by using near infrared spectroscopy combined with chemometrics for classifying different maturity of fresh tobacco leaves was proposed in this paper. [ Resuit] Compared with principal component analysis（ PCA）, K nearest neighbor （KNN） and support vector machine （SVM） qualitative ...