利用近红外分析技术测定大豆水分含量方法的研究

Measurement of moisture content in soybean by near-infrared spectroscopy technique

建立了基于FOSS XDS近红外光谱分析仪快速测定大豆水分含量模型,对光学处理和数学处理手段等因素对模型的影响进行了探讨,对模型进行了内部和外部验证。实验结果表明最佳的建模参数为:光学处理选用标准正常化处理(SNV Only),数学处理选用1.4.4.1方法,大豆水分定标方程的交互定标决定系数(1-VR)为0.990 8,定标决定系数(R2)为0.993 9,定标标准误差(SEC)为0.096 7,交互定标标准误差(SECV)为0.127 3,现有数据预测标准偏差(SEP)为0.136。利用该模型对大豆水分含量进行检测,达到了代替常规标准方法的要求,可以应用于快速检测。

The model of determining soybean moisture was built using near- infrared spectroscopy （NIRS） and the FOSS XDS system as the analyzer. The influences on the model of factors, such as the mathematics methods and optics treatment methods were studied. The calibrations of the moisture content in soybean were also performed. The results showed that the best factors were SNV only for optics treatment method and 1.4.4. 1 for mathematics method. The correlation coefficient （ 1 - VR） was 0. 990 8 ,the average determina- tion coefficient of validation （R2） was 0.993 9, the square error of cross（SEC） was 0.096 7, the square error of cross validation （SECV） was 0. 127 3 ,and the standard error of prediction （SEP） was 0. 136. This model could substitute for traditional chemical method to determine the moisture content in soybean.