神经网络在大米蛋白质含量预测模型中的应用  

Determination of protein content of rice by near infrared spectroscopy based on neural networks

作　　者：刘建学[1] 吴守一[2] 方如明[2] 

机构地区：[1]河南科技大学食品与生物工程学院,河南洛阳471003 [2]江苏大学生物与环境工程学院,江苏镇江212013

出　　处：《江苏大学学报（自然科学版）》2004年第3期196-198,共3页Journal of Jiangsu University：Natural Science Edition

基　　金：河南省自然科学基金资助项目(1347001)

摘　　要：大米蛋白质含量是影响大米营养及食味特性的一个重要指标 借助主成分分析,确立了用于近红外光谱分析的BP神经网络的输入输出模式对;并用BP神经网络方法建立了不同类型、不同粒度的大米样品蛋白质含量预测模型;考察了模型的预测能力,其预测值与用标准方法取得的化学分析值间具有良好线性关系,相关系数达0 90以上.Protein content (PC) is one of the important parameters in affecting the nutrition and taste characteristics. The chemical measurement of rice PC is expensive and time consuming. The input pattern of BP neural networks based on near infrared spectroscopy and principal component regression is presented, and the BP neural networks (BPNN) model of PC for the rice samples of different particles and varieties is built by optimizing the studying vector. The correlation coefficient between the evaluated value of rice PC by BPNN algorithm and chemical value is 0.94. It shows that the BP neural networks can reduce the predicting difference produced by the rice samples of different particles.

关 键 词：大米 蛋白质含量 近红外光谱 BP神经网络 预测模型 

分 类 号：O433.3[机械工程—光学工程] TP39[理学—光学]