Study on Discriminant Model of Flue-Curing Stage of Tobacco Leaves in Curing Process

Automatic discrimination of curing stage in the process of intensive curing is very important to promote intelligent curing of tobacco leaves. In order to realize the fast and accurate identification of the curing stage. The image features of tobacco leaves in the curing process are extracted, and the feature subset is obtained by variable clustering analysis and correlation analysis. The traditional machine learning GA-SVM model is established by taking the feature subset as input variables, and the original image as input variables. The depth learning algorithm AlexNet model, ResNet18 model and DenseNet201 model are established. The above four models are used to distinguish and compare the curing stage of tobacco leaves. The discrimination accuracy of GA-SVM model is 88.20%. The discrimination accuracy of AlexNet model is 83.67%. The discrimination accuracy of Resnet 18 model is 95.67%. The discrimination accuracy of Denet 201 model is 93.89%. The results show that the discrimination effect of ResNet18 model is better than that of DenseNet201 model, that of DenseNet201 model is better than that of GA-SVM model, and that of GA-SVM model is better than that of AlexNet model.