基于近红外光谱的山核桃蛋白质、脂肪含量的测定

Determination of protein and fat content in pecan based on near infrared spectroscopy

为实现山核桃品质快速鉴定,采集了119份山核桃样品的近红外光谱,并分别测定其蛋白质和脂肪含量。在剔除异常样本后,将110份样本按照3∶1的比例划分为校正集和验证集。使用不同的光谱预处理和特征波段选取方法,结合偏最小二乘法(partial least squares)建立了山核桃蛋白质和脂肪含量的近红外模型。结果表明:使用一阶导数(first derivative)预处理的模型预测性能最佳;经过竞争性自适应权重取样法(competitive adapative reweighted sampling,CARS)处理后,光谱变量数目大大减少,模型的精度有部分提升,可实现对山核桃蛋白质和脂肪含量的快速检测。

In order to quickly identify the quality of pecan,the near infrared spectra of 119 pecan samples were collected,and the protein and fat contents were determined,respectively.After eliminating abnormal samples,110 samples were divided into correction set and verification set according to the ratio of 3∶1.Using different spectral preprocessing and feature band selection methods,combined with partial least squares,the near-infrared model of protein and fat content of pecan was established.The results showed that the model with first derivative pretreatment had the best prediction performance.After the competitive adapative reweighted sampling(CARS)processing,the number of spectral variables was greatly reduced,and the accuracy of the model was partially improved,which could realize the rapid detection of protein and fat in pecan.