基于电子鼻和BP神经网络对‘黑珍珠’鲜食玉米产地的区分和识别

Distinction and Recognition of the 'Black Pearl' Fresh Corn Origin Based on Electronic Nose and BP Neural Network

以‘黑珍珠’鲜食玉米为研究对象,采用电子鼻技术分别测定了黑龙江、陕西两大产区共计200个鲜食玉米样品的气味传感器响应值原始数据,通过主成分分析(Principal component analysis,PCA)、判别因子分析(Discriminant function analysis,DFA)对不同产地鲜食玉米的挥发性风味进行了区分,采用软独立建模分析(Soft independent modeling class analogy,SIMCA)建立了黑龙江‘黑珍珠’鲜食玉米的判定模型,并通过Pytorch软件建立了反向传播神经网络(Back propagation neural network,BP神经网络)模型,对不同产地的‘黑珍珠’鲜食玉米进行鉴别区分。结果表明,不同产地的‘黑珍珠’鲜食玉米的挥发性风味虽有相似之处但具有明显的产地特征,SIMCA模型可实现对未知样品是否来自黑龙江产区的有效识别(正确率为97%),BP神经网络模型则可对未知产地的‘黑珍珠’鲜食玉米样品进行产地预测及鉴别,平均正确率达99.44%。采用电子鼻技术结合BP神经网络模型可以准确的区分和识别‘黑珍珠’鲜食玉米产地。

'Black pearl'fresh corns from different regions were analyzed using an electronic nose to capture the aroma profile.Principal component analysis(PCA)and discriminant function analysis(DFA)were used for multivariate statistical analysis of 200 data from two regions.Based on this,the judgment model of samples from Heilongjiang production area was built using a soft independent modeling class analysis(SIMCA)algorithm,and a back propagation neural network model was established by Pytorch software to identify and differentiate samples from different regions.The results illustrated that,although the volatile flavor of'black pearl'fresh corns from different origins were similar,it also showed obvious origin characteristics.SIMCA model could effectively distinguish whether unknown samples come from Heilongjiang (the accuracy rate was 97%), while BP neural network model could predict and identify the origin of 'black pearl' fresh corns from unknown production areas, and the average accuracy rate was 99.44%. The combination of electronic nose technology and BP neural network model could accurately distinguish and identify the origin of 'black pearl' fresh corns.