Identification of saffron origin and determination of crocin content in saffron based on machine vision and visible-near infrared spectroscopy

Abstract

Saffron is a kind of medicine food homologous products with ornamental, edible, medicinal and economic value, and the identification of the origin of saffron and the determination of the content of its effective components play an important role in the evaluation of its economic value and medicinal efficacy. In this study, the content of crocin in saffron samples from different habitats was detected by UPLC technology, the color and shape features of saffron were extracted by machine vision technology, and the spectral feature variables of saffron were screened by visible-near infrared technology combined with variables; k-nearest neighbor algorithm (KNN) model was established to identify the origin of saffron. The result shows that machine vision combined with visible-near infrared spectral variables could identify the origin of saffron, and the recognition rate of KNN model is 90.32%. Back propagation neural network (BP-ANN) was established to determine the content of crocin in saffron. The result shows that visible-near infrared spectroscopy could better predict the content of crocin in saffron. The correlation coefficient of crocin I in prediction set was 0.9883, the correlation coefficient of crocin Ⅱ in prediction set was 0.9982. The above results showed that the technology could be used to identify the origin of saffron and determine the content of crocin rapidly.