摘要
烟叶的红外光谱特征不仅与其化学成分密切相关,而且与其内部结构有一定的关系。提取90份烤烟烟叶样本的近红外光谱进行二阶微分去噪,并进行定量分析。加入外观特征并采用非线性主成分分析法进行数据压缩和特征提取,利用因子分析的方法进行烟叶内在质量和外光特征间的相关性分析,并利用BP神经网络建模对相关性进行预测。分析结果显示,烟碱、蛋白质、宽长比、叶片长度、色调标准差等5个变量有相似的变化趋势,总糖、还原糖、叶片厚度、平均亮度等4个变量间有相似的变化趋势。依据所得结论,可以辅助烟叶评级标准,使烟叶评级标准更加的客观。
Infrared spectrum of tobacco leaves have close connections with not only their internal structures but also their chemical compositions.Using second order differential methods eliminate the noises of near-infrared spectrum of ninety tobacco samples and do the quantitative analysis.Compressed data and extract figures based on the internal qualified factors together with the appearance characteristic of the tobaccos used the nonlinear principal component analysis method.Then using factor analysis method analyzes the correlation between internal qualified factors and appearance characteristics and BP neural network predicts the effect of correlation analyses.Analysis results show that variables of nicotine,protein,ratio of width to length,length of tobacco leaf and hue standard deviation have similar trend;variables of total sugar,reducing sugar,thickness of tobacco leaf and average brightness have similar trend.Based on the analysis,it can help tobacco rating standard to improve objectivity.
出处
《科学技术与工程》
2010年第10期2543-2546,2555,共5页
Science Technology and Engineering
关键词
近红外光谱
非线性主成分分析
因子分析
烟叶评级
near infrared reflectance spectroscopy(NIRS) nonlinear principal component analysis factor analysis tobacco leaves grading