本研究运用近红外光谱无损检测技术，开发了一种适用于面粉品质检测的在线测量系统。本系统在硬件平台基础上，采用C++Builder 6.0对NIR 1.7/S微型光谱仪进行二次开发，编写了具有光谱采集、面粉品质预测、模型更新和数据存储等功能的软件。对市售170种面粉进行试验，以面粉水分含量为代表性指标。通过对比不同光谱预处理方法建模结果，发现不进行任何预处理时的面粉水分偏最小二乘回归（PLS）得到的模型精度最高。建模集和验证集决定系数（R2）分别为0.947，0.841；均方根误差（RMSE）分别为0.146%，0.198%；RPD值为2.53。模型导入软件后对30份新样品进行外部验证，预测值与测量值决定系数（R2）为0.883，均方根误差为0.206%。结果表明，该系统能够初步实现面粉水分的实时预测，为近红外在线检测技术应用提供了一定的技术参考。

In this work, an on-line detection system for flour quality testing was developed by near infrared spectroscopy. C++Builder 6.0 was applied for secondary development of NIR 1.7/S micro spectrometer based on hardware platform. A custom software with functions such as spectral acquisition, flour quality prediction, model update and data storage was also wrote. A total of 170 wheat flour samples collected from the market were subjected to NIR analysis and water content was chosen as the representative index for wheat flour quality. Different spectral pretreatment methods were tried and the spectra without any treatment combined with partial least square (PLS) regression obtained the best performance. The determination coefficient (R2) of calibration and prediction obtained were 0.947 and 0.841, respectively. The root mean square error (RMSE) of calibration and prediction as well as RPD value were 0.146%, 0.198% and 2.53, respectively. The established model was then imported into the software, and 30 samples were used for external validation. Results indicated that determination coefficient (R2) between the predicted value and the reference value was 0.883, and the RMSE was 0.206%. In conclusion, the prediction of the moisture of wheat flour by this on-line system is feasible, which can provide a technical reference for the application of on-line NIR spectroscopy technique.

国家重点研发计划（2017YFD0400401）；国家自然科学基金（31772061）；江苏省研究生科研与实践创新计划（KYCX17_1213）；江苏高校优势学科建设工程资助项目（2014-124）