摘要
提出了一种采用可见-近红外光谱分析技术快速鉴别可乐品牌的新方法。采用美国ASD公司的便携式光谱仪对三种不同品牌的可乐进行光谱分析,各获取55个样本数据。将样本随机分成150个建模样本和15个预测样本,采用平均平滑法和标准归一化方法对样本数据进行预处理,再用主成分分析法对光谱数据进行聚类分析并获得各主成分数据。将建模样本的主成分数据作为BP网络的输入变量,可乐品牌作为输出变量,建立三层人工神经网络鉴别模型,并用模型对15个预测样本进行预测。结果表明,预测准确率为100%,实现了可乐品牌快速、准确的鉴别。
A new method was developed to fast discriminate the brands of cola by means of visual-near infrared spectroscopy （NIRS）. Three different brands of cola （Coca-cola, Pepsi-cola and Future-cola） were analyzed using a handheld near infrared spectrometer produced by ASD Company. Fifty five samples were used for each brand of cola, and they were divided randomly into a group of 150 samples as calibrated samples and one of 15 samples as prediction samples. The samples data were pretreated using average smoothing and standard normal variable method, and then the pretreated spectra data were analyzed using principal component analysis （PCA）. The principal component data of calibrated samples were used as the inputs of back-propagation artificial neural network （ANN-BP）, while the values of cola brands used as the outputs of ANN-BP, and then the three layers ANN-BP discrimination model was built. The 15 unknown prediction samples were analyzed by the ANN-BP model. The result showed that the distinguishing rate was 100~ it was realized to discriminate different brands of Cola rapidly and exactly.
出处
《光谱学与光谱分析》
SCIE
EI
CAS
CSCD
北大核心
2007年第8期1543-1546,共4页
Spectroscopy and Spectral Analysis
基金
国家科技支撑项目(2006BAD10A04
2006BAD10A09)
国家自然科学基金资助项目(30600371)
高等学校优秀青年教师教学科研奖励计划(02411)资助
关键词
可见-近红外光谱
可乐
主成分分析
人工神经网络
鉴别
Visual-near infrared spectral
Cola
Principal component analysis （PCA）~ Artificial neural network~ D/scrimination
