摘要
根据烟叶样本的近红外光谱定量分析得出3种烟叶成分:烟碱、还原糖和蛋白质的含量,加入8种烟叶外观特征,然后使用因子分析方法对概率神经网络输入进行压缩和特征提取,在简化样本的同时对概率神经网络进行优化.应用概率神经网络根据外观质量因素对降维后的烟叶样本建立内在质量的数学预测模型并获得较理想的预测效果.
The content of nicotine,reducing sugar and protein of the tobacco was determined based on the near-infrared spectrum of tobacco samples.Principal component analysis method was used to compress data and extract figures when combining with eight appearance characteristics of tobacco.The dimension of sample was simplified and the input of probabilistic neural network was optimized.Then the correlation between inner qualities and appearance characteristics of tobacco was analyzed by the probabilistic neural network.The results indicated that the inner quality could be predicted by the probabilistic neural networks.
出处
《北京工商大学学报（自然科学版）》
CAS
2010年第6期66-70,共5页
Journal of Beijing Technology and Business University:Natural Science Edition
关键词
烤烟烟叶
近红外光谱
因子分析
概率神经网络
flue-cured tobacco leaf
near-infrared spectrum
principal component analysis
probabilistic neural network
