论文部分内容阅读
采用近红外光谱测定蜂蜜中果糖、葡萄糖、蔗糖和麦芽糖含量。测定结果表明:K-S(Kennard-Stone)法是一种有效的划分校正集和验证集的算法。采用蒙特卡罗交互验证法,剔除校正集中的奇异值。果糖、葡萄糖、蔗糖和麦芽糖在奇异值剔除前模型预测值和实际值之间的决定系数分别为0.945,0.927,0.901和0.610,奇异值剔除后其决定系数分别提高到0.971,0.940,0.949和0.750。蒙特卡罗交叉验证扣除奇异值的方法能够改善数据分布,提高模型准确性;小波变换算法处理后由光谱建立的蜂蜜PLS校正模型与变换前精确度无明显变化,但是采用小波变换算法处理后的光谱建模所用时间由3.65 s减少到2.76 s。通过上述几种化学计量学方法,得到蜂蜜中4种主要糖分的近红外光谱模型,为蜂蜜快速检测和实时在线质量控制提供依据。
Determination of fructose, glucose, sucrose and maltose in honey by near infrared spectroscopy. The results show that the K-S (Kennard-Stone) method is an effective algorithm for dividing the calibration set and validation set. Using Monte Carlo cross-validation, excluding the singular value of the calibration set. The coefficients of determination of the predicted and actual values of fructose, glucose, sucrose and maltose before the singular value elimination are 0.945, 0.927, 0.901 and 0.610, respectively. After the singular value is removed, the coefficient of determination is increased to 0.971, 0.940, 0.949 and 0.750 . Monte Carlo cross-validation method can deduce the singular value of the data to improve the distribution and improve the accuracy of the model; Wavelet transform algorithm to deal with the spectrum of the honey established by the PLS calibration model and accuracy before transformation no significant change, but using wavelet transform algorithm The time required for spectral modeling decreased from 3.65 s to 2.76 s. Through the above several chemometrics methods, the near-infrared spectrum model of the four main sugars in honey was obtained, which provided the basis for rapid honey detection and real-time on-line quality control.