当前位置: 首页 > 文章 > 概率神经网络在烟叶自动分级中的应用 农机化研究 2011,33 (12) 32-35
Position: Home > Articles > Application of Probabilistic Neural Network in Tobacco Automatical Grading Journal of Agricultural Mechanization Research 2011,33 (12) 32-35

概率神经网络在烟叶自动分级中的应用

作  者:
张乐明;申金媛;刘剑君;刘润杰
单  位:
郑州大学信息工程学院河南省激光与光电信息技术重点实验室;河南省烟草公司郑州分公司
关键词:
烟叶分级;概率神经网络;红外光谱
摘  要:
烟叶自动分级一直都是国内外烟草研究领域的重难点。为此,以红外光谱作为烟叶的特征,采用概率神经网络对11个等级的烟叶进行分组和分级。对光谱信号做消除基线漂移预处理,然后将其作为神经网络的输入样本,选择50%左右的样本作为学习训练样本,其余为测试样本;训练好的模型不论是分组还是分级,对于训练样本的正确吻合率为100%,测试样本的平均正确吻合率在90%以上。结果表明,概率神经网络可以进行烟叶自动分级,为烟叶的自动分级开辟了一条新途径。
译  名:
Application of Probabilistic Neural Network in Tobacco Automatical Grading
作  者:
Zhang Leming1,Shen Jinyuan1,Liu Jianjun2,Liu Runjie1(1.School of Information Engineering,Zhengzhou University,Henan Key Laboratory of Laser and Opto-electric Information Technology,Zhengzhou 450001,China;2.Zhengzhou Filiale,Tobacco Company of Henan Province,Zhengzhou 450001,China)
关键词:
tobacco leaves grading;probabilistic neural network;infrared spectrum
摘  要:
The tobacco automatical grading has always been being the key and difficulty in the field of tobacco study.In this paper,probabilistic neural network(PNN) is proposed to grade the tobacco leaves.The infrared spectra used as the features of tobacco leaves are employed as the input patterns of PNN.The infrared spectra of some tobacco leaves including 11 levels are collected and the operation of subtracting the minimum value is carried on to remove the baseline drifting.In order to train the network,about a half of leaves are chosen as the trainning samples.The others are uesed as test samples to verify the performance of the trained network.The results of part classfication and grading are that the mathc rate are 100% and ≥90% respectly for trainning and test samples.This means that the PNN can be used to automatically grade the tobacco leaves.It gives a new way for the automatic grading of tobacco.

相似文章

计量
文章访问数: 7
HTML全文浏览量: 0
PDF下载量: 0

所属期刊

推荐期刊