论文部分内容阅读
为建立近红外光谱无损检测鸡蛋脂肪含量的方法,在近红外光谱全波段内采集鸡蛋样品的漫反射光谱图,用酸水解法测定鸡蛋样品中的脂肪含量。对采集的光谱进行最小-最大归一化(Min-max Normalization,MMN)、矢量归一化(Vector Normalization,SNV)、平滑、一阶导数(First Derivative,FD)及多元散射校正(Multiplicative Scatter Correction,MSC)处理,用偏最小二乘法(Partial Least Squares,PLS)对鸡蛋脂肪含量建模验证。结果表明,经多元散射校正(MSC)法预处理,偏最小二乘法(PLS)建模以及杠杆校正(Leverage Correction)检验,鸡蛋的脂肪含量与其近红外光谱信号之间存在线性关系,校正集和验证集相关系数R2分别0.947 5,0.906 3,校正均方差RMSEE为0.173 2,预测均方差RMSEP为0.231 4,模型效果最好,可用于鸡蛋中脂肪含量的无损检测。
In order to establish a method for non-destructive testing of fat content in eggs by near infrared spectroscopy, the diffuse reflectance spectra of egg samples were collected in the whole wave band of near infrared spectroscopy and the fat content in egg samples was determined by acid hydrolysis. The collected spectra were subjected to Min-max Normalization (MMN), Vector Normalization (SNV), Smoothing, First Derivative (FD) and Multiplicative Scatter Correction , MSC), and the egg fat content was modeled and verified by Partial Least Squares (PLS). The results showed that there was a linear relationship between the fat content of the egg and the near infrared spectroscopy signal through the pretreatment of MSC method, PLS modeling and Leverage Correction. The calibration set and The correlation coefficients R2 of the validation set were 0.947 5 and 0.906 3 respectively. The mean square error of calibration (RMSEE) was 0.173 2. The mean square error of prediction (RMSEP) was 0.231 4. The model was the best and could be used for non-destructive testing of fat content in eggs.