基于近红外光谱技术快速检测稻谷水分含量

目的:建立一种无损、快速高效的稻谷水分含量检测方法。方法:研究收集了不同年份的稻谷样品161份,运用近红外光谱结合化学计量学方法,通过剔除异常光谱和光谱预处理,采用偏最小二乘法建立稻谷水分含量预测模型。结果:采用主成分分析结合马氏距离的方法剔除异常光谱样品15个,最佳的光谱预处理方式为消除常数偏移量。训练集建立的预测模型(R_CAL²)为0.9943,模型标准偏差(RMSEC)为0.21%,模型交叉验证决定系数(R_CV²)为0.9936,模型交叉验证标准偏差(RMSECV)为0.32%,表明预测模型交叉验证预测样品水分含量准确度高。用验证集样品检验预测模型,模型验证集验证决定系数R 2 VA L为0.9801,模型验证集验证标准偏差(RMSEP)值为0.36%,相对分析误差(RPD)值为7.14,表明预测模型对未知样品的预测准确度高。验证集样品实测值与预测值均值方程T检验结果P值(双侧)为0.879,验证集样品实测值与预测值之间差异不显著,表明预测模型的预测结果可信度高,验证集样品预测值与实测值的误差在±1%,且90%以上的验证集样品其预测值与实测值的误差都在±0.5%以内。结论:建立的稻谷水分预测模型可以实现收储稻谷的无损、快速、准确检测。

Research on rapid prediction model of rice moisture content based on near infrared spectroscopy

Objective:In order to established a non-destructive, rapid and efficient method for detecting the moisture content of rice. Methods:This study, 161 rice samples were collected from 5 different regions were studied by near infrared spectroscopy combined with stoichiometry. By eliminating abnormal spectra and preprocessing the spectra, the prediction model of rice moisture content was established by partial least squares regression. Results:15 abnormal spectrum samples were eliminated using the method of principal component analysis combined with mahalanobis distance. The best spectral pretreatment was to eliminate the constant offset. The prediction model R2CAL established in the training set was 0.994 3, root mean square error of calibration (RMSEC) was 0.21%, R2CV was 0.993 6, and root mean square error of cross validation (RMSECV) was 0.32%, which indicated that the cross-validation of the prediction model had high accuracy in predicting sample moisture content. The prediction model was tested with the validation set samples. The validation determination coefficient R2VAL of the model validation set was 0.980 1, the root mean square error of prediction (RMSEP) was 0.36%, and the relative percent deviation (RPD) was 7.14, which indicated that the prediction model had high prediction accuracy for the unknown samples. The P-value (two-sided) of the mean equation T-test of the measured and predicted values of the samples in the validation set was 0.879, and the difference between the measured and predicted values of the samples in the validation set was not significant, indicating that the prediction results of the prediction model were highly reliable. The error between the predicted value and the measured value of the verification set samples was within ±1%, and more than 90% were within ±0.5%. Conclusion:The established rice moisture prediction model can be applied to actual production, and it provides a non-destructive, rapid and high-accuracy detection method for the inspection of rice harvesting and storage.