基于电子鼻表征霉心病苹果特征气味及无损检测模型建立

Characterization of characteristic odor and establishment of nondestructive detection model of core rot apples based on electronic nose

为探究电子鼻检测技术对霉心病苹果的判别效果,以富士健康苹果和霉心病苹果为试材,基于SIMCA软件对采集的不同病变程度霉心病苹果的电子鼻信息进行表征,基于SPSS 23.0软件建立霉心病苹果Fisher函数、多层感知器神经网络(muhilayer perceptron neural network, MLPNN)和径向基函数神经网络(radial basis function neural network, RBFNN)判别模型。结果表明,健康果的特征传感器有W5C、W1C和W3C,重度果的特征传感器有W1S、W2S、W1W和W5S;MLPNN模型的判别效果最好,其对训练集和验证集的总体预测判别率分别为88.61%和88.46%;RBFNN模型的判别效果次之,其对训练集和测试集的总体预测判别率分别为93.50%和80.95%;Fisher判别函数判别效果最差,其对训练集和验证集的总体预测判别率分别为91.50%和79.27%。另外,3种判别模型对健康果和重度果都有很好的判别效果,对健康果和轻度果的判别效果不理想,需要在今后的研究中进一步优化。

To explore the discrimination effect of the electronic nose on core rot apples, “Fuji” moldy core and healthy apples were used as test materials and the electronic nose information of core rot apples was characterized based on SIMCA software. In addition, the fisher discriminant function, multilayer perceptron neural network(MLPNN) and radial basis function neural network(RBFNN) discriminant models of core rot apple was established. The results showed that W5C, W1C and W3C were the characteristic sensors of healthy fruit, while W1S, W2S, W1W and W5S were the characteristic sensors of severely diseased fruit. The model of MLPNN had the best discrimination effect, and the overall prediction accuracy for the modeling and validation was 88.61% and 88.46%, respectively. The model of RBFNN was secondary with the overall prediction accuracy for the modeling and validation was 93.50% and 80.95%, respectively. While, the model of fisher discriminant function had the worst discrimination effect, and the overall prediction accuracy for the modeling and validation was 91.50% and 79.27%, respectively. Even though the three discriminant models had good discriminant effect on healthy and severe fruit, and the discriminant effect on healthy and mild fruit was not ideal, which needs to be further optimized in future.