基于近红外高光谱图像技术的栗果品质无损检测

Non-destructive detection of Chinese chestnut(Castanea mollissima) nut qualities based on near-infrared hyperspectral imaging techniques

提出一种基于近红外高光谱图像技术的板栗果实品质快速无损检测方法。分别选取3个不同品种栗果、1个品种的霉变栗果和1个品种的虫害栗果各30个样品,采集供试样品的近红外高光谱数据;采用偏最小二乘法(PLS)建立栗果中总糖和淀粉含量预测模型,预测值与实际值的相关系数为0.9313～0.9587,均方根误差为0.062 4～0.225 0;结合主成分分析法(PCA),建立不同品种栗果鉴别以及识别霉变、虫害、正常栗果的判别分析(DA)模型,模型的识别率分别为96.7%和98.6%。结果表明,近红外高光谱图像技术可用于栗果总糖和淀粉的定量预测,以及不同品种栗果和霉变、虫害果的快速定性识别。

In this paper,a rapid and non-destructive detection method was proposed for quality inspection of Castanea mollissima nuts using near-infrared hyperspectral imaging techniques.Thirty samples of three different types,i.e.normal nuts,mouldy nuts and insect infested nuts were selected,and their near-infrared hyperspectral data were collected.The quantitative analysis model for the contents of sugar and starch was established using partial least squares(PLS)method.The coefficients between the predictive values and the actual values were between 0.9313 and 0.9587,and the root mean squared errors were between 0.0624 and 0.2250.Combining principal component analysis(PCA),the discriminant analysis(DA)model for the detection of different nut varieties and for the identification of the mouldy,the insect infested and the normal nuts was established,and their recognition accuracies of the established model were 96.7%and 98.6%,respectively.The results showed that near-infrared hyperspectral imaging techniques can not only be used for quantitative analysis of total sugar and starch in Castanea mollissima nuts,but also be used for the rapid detection of different nut varieties and the mouldy or the insect infested nuts.