Detection of edible and nonedible mushrooms using electronic nose and artificial intelligence

Non-use of non-edible and toxic mushrooms by common people in edible mushrooms is a dangerous challenge that can lead to mushroom poisoning. In this research, it is possible to use the smell machine system based on metal semiconductor sensors as non-destructive tools for non-edible mushrooms. Food was evaluated. PCA principal component analysis, LDA linear discriminant analysis, QDA 2 degree linear discriminant analysis, SVM support vector machine and ANN artificial neural network were the methods used to achieve this goal. The results obtained from the analyzes showed that the best performance of the PCA method was in the separation of 5 samples of edible and non-edible mushrooms with 91% accuracy. The obtained results showed that the QDA analysis method separated the mushroom samples into three different categories with 100% accuracy and better performance than the LDA method. In the meantime, the C-SVM method could only fully and 100% detect A. bisporus among the samples of edible and non-edible mushrooms, and it did not have a good performance in differentiating other samples. The ANN method had a good performance in classifying the samples so that it could classify the samples of A. bisporus, S. comptus and R. delica 100%. According to the results obtained in this research, it can be said that the QDA method was higher than other methods in classifying types of fungi. Considering the proper performance of machine smell in separating edible mushrooms from non-edible mushrooms, it seems that using this technology will be a promising method in different types of mushroom.