基于植物表型复合参数体系的烤烟等级初分检测方法

Primary Grade Detection Method for Flue-cured Tobacco Based on Plant Phenotypic Compound Parameter System

为实现快速无损智能判定烤烟等级,基于植物表型复合参数体系设计了一款便携式烤烟等级快速检测仪,通过试验确定了37个与烤烟等级值具有显著或极显著相关的表型参数组成了烤烟等级特征复合参数体系,选取K-Means聚类法、系统聚类法、 CHAID决策树算法、BPNN算法作为烟叶判定模型的建模对比算法,从500株云烟87烟株中选取了6个等级的595片烟叶作为样本源,每个等级的50片烟叶y样本作为建模数据,其余的烟叶作为验证数据,结果表明:基于贝叶斯分类器算法的等级判定准确率为75.59%,高于企业所要求的烟农初分准确度标准(准确度≥60%)。基于表型复合参数体系的烟叶判定模型可用于构建便携式轻量化的烤烟等级检测仪,为后续开展不同生态区烤烟等级快速检验研究奠定了基础。

A portable grade detector was designed based on plant phenotypic compound parameter system to determine the grade of flue-cured tobacco quickly and intelligently.Through the experimental study,37 phenotypic parameters that were significantly or extremely significantly correlated with the grade value of flue-cured tobacco were determined to form the compound parameter system of flue-cured tobacco grade characteristics.K-means cluster,hierarchical cluster,CHAID(Chi-squared automatic interaction detection)decision tree algorithm and BPNN(back propagation neural network)algorithm were selected as comparison algorithms of tobacco grade determining modeling.595 tobacco leaves of six grades were selected from 500‘yunyan 87’tobacco plants as sample sources,and 50 tobacco leaves of each grade were used as modeling data and the remaining tobacco leaves as verification data.The results show that the determination accuracy based on Bayesian classifier algorithm is 75.59%,higher than the accuracy standard required by the enterprise(≥60%).In this paper,the tobacco leaf determination model based on phenotypic compound parameter system can be used to construct a portable lightweight flue-cured tobacco grade detector,which can lay a foundation for the subsequent rapid inspection of flue-cured tobacco grade in different ecological areas.