基于近红外技术和偏最小二乘判别法对卫青萝卜内部品质的快速判别

Identification of Green Radish Internal Quality Based on Near-infrared Spectroscopy and Partial Least Square-discriminant Analysis

为实现天津卫青萝卜内部品质的快速无损检测,以天津卫青萝卜为试验对象,对其进行近红外光谱采集,并对其可溶性固形物、水分、糠心3个品质指标进行检测,根据光谱数据、品质指标构建偏最小二乘判别分析(PLS-DA)模型进行聚类判别。结果表明,经MA平滑处理后的样本光谱,在波长700~750 nm,900~950 nm范围内,样本的吸光度值范围较宽,因此选择在波长700~950 nm范围的光谱数据进行建模。卫青萝卜水分含量、可溶性固形物含量的平均值±样本标准差(x±s)分别为92.02%±0.01%,7.61±0.69°Brix。偏最小二乘法建立的模型具有良好的预测性,卫青萝卜可溶性固形物预测值和真实值的决定系数R2为0.8236,水分含量R2为0.8740,糠心模型准确率高达94.44%。由此可见,近红外光谱技术对天津卫青萝卜内部品质的快速无损检测具有可行性。

In order to distinguish internal quality of Weiqing green radishes quickly and without damage,the samples of Weiqing green radishes in Tianjin were collected by near-infrared spectroscopy and detected three quality indexes of soluble solid content(SSC),moisture content and chaff hollowness.According to the spectral data and quality indexes,PLS-DA model was constructed for cluster analysis.The results showed that the sample spectrum has a wide range of absorbance values in the wavelength range of 700~750 nm and 900~950 nm,which processed by MA(Moving Average)smoothing processing.Therefore,the spectral data were selected for modeling in the wavelength range of 700~950 nm.The average±standard deviation(x±s)of the moisture content and soluble solid content were 92.02%±0.01%,7.61±0.69°Brix.The model established by partial least squares method has good predictability.The decision coefficient(R2)of SSC between the predicted value and the measurement value was 0.8236,and R2 of moisture was 0.8740.The accuracy of the hollowness model was 94.44%.It can be seen that nondestructive measurement of internal qualities of green radishes using NIR spectroscopy technique was feasible.