基于傅里叶近红外光谱和电子鼻技术的苹果霉心病无损检测

Nondestructive detection of apple moldy core based on FT-NIR and electronic nose technology

为研究傅里叶近红外光谱技术(Fourier transform near infrared spectroscopy,FT-NIRS)和电子鼻技术分别结合化学计量学方法对苹果霉心病的判别效果,以“红富士”霉心病苹果和健康苹果为试材,利用近红外光谱技术,基于主成分分析建立Fisher判别和多层感知器(multi-layer perceptron,MLP)神经网络模型;同时利用电子鼻技术分别结合Fisher判别、MLP神经网络和径向基函数神经网络3种化学计量学的方法建立判别模型。根据建模集和验证集的预测准确率综合考虑,基于主成分分析建立的MLP神经网络模型和电子鼻结合MLP神经网络模型对苹果霉心病的判别效果最好,验证集中的正确判别率分别达到87.7%和86.2%。说明电子鼻和近红外光谱技术均可以较好地判别苹果霉心病。

Fourier near-infrared spectroscopy(FT-NIRS)and electronic nose technique were used to identify the efficacy of apple moldy core in combination with chemometrics.With“Red Fuji”moldy core and healthy apples as raw materials,the near-infrared spectroscopy model of multi-layer perceptron(MLP)neural network and Fisher discriminant was established based on the principal component analysis.Meanwhile,a discriminant model was established by the electronic nose combined with Fisher discrimination,MLP neural network and radial basis function(RBF)neural network,respectively.According to the comprehensive consideration of the prediction accuracy of the modeling set and the verification set,the MLP neural network model based on principal component analysis and the electronic nose combined with the MLP neural network model had the best discriminating effect on the apple moldy core disease and the correct discriminating rate of verification set reached 87.7%and 86.2%respectively.It shows that the electronic nose and near-infrared spectroscopy used for distinguishing apple mold core are feasible.