基于近红外光谱技术的小麦粉灰分含量检测方法研究

采用近红外光谱技术建立小麦粉灰分含量的快速检测方法。使用两种不同的近红外光谱仪器采集小麦粉的近红外光谱数据,以常规分析法的测定值作为建模数据,采用偏最小二乘(PLS)回归分析法建立小麦粉灰分的定量分析模型,比较两种不同的近红外光谱仪器扫描的小麦粉近红外光谱图对模型的影响。结果表明,MicroNIR-1700近红外光谱仪扫描的谱图所建校正集模型的相关系数R~2为90.69,均方根误差(RMSECV)为0.031 2,预测集模型的均方根误差(RMSEP)为0.021 7;VERTEX70傅里叶变换近红外光谱仪扫描的谱图所建校正集模型的相关系数R~2为89.40,均方根误差(RMSECV)为0.035 0,预测集模型的均方根误差(RMSEP)为0.036 6。两种仪器都能用于小麦粉光谱采集,并进行灰分含量快速检测,MicroNIR-1700在小麦粉灰分检测方面有更好的应用。

Detection of flour ash based on near infrared spectroscopy

Rapid detection of ash content in flour was built based on the near infrared spectroscopy technology. The near infrared spectral data of flour was collected by two different near-infrared spectrometers; conventionally measured value as model data, a quantitative analysis model of flour ash content was established by partial least square regression analysis; the impact of near infrared spectrograms of flour scanned by two different near-infrared spectrometers on the model were compared. The results showed that the correlation coefficient of calibration set model of MicroNIR-1700 scanning near-infrared spectrometer was 90.69, with the root mean square error RMSECV 0.0312; the root mean square error of prediction set model RMSEP was 0.0217; the correlation coefficient of calibration set model of VERTEX70 Fourier transform near infrared spectrometer was 89.40, with the root mean square error RMSECV 0.0350; the root mean square error of prediction set model RMSEP was 0.0366. Both instruments can be used for collecting flour spectrograms, and rapid detection of ash content, while MicroNIR-1700 had good application in flour ash detection.