Origin Identification of Jujubes Based on Hyperspectral Imaging Technology

Abstract

To achieve non-destructive, rapid, batch detection and accurate identification of jujube origins, a recognition model for jujube origin was constructed in this study based on hyperspectral technology combined with machine learning methods. Jujubes from six different provinces—Liaoning, Xinjiang, Yunnan, Hebei, Hubei, and Guangxi—were selected as research subjects, and their hyperspectral data were collected. The original spectral data were preprocessed using multiplicative scatter correction (MSC), standard normal variate (SNV), Savitzky-Golay smoothing (SG), first derivative (FD), and second derivative (SD) methods. Subsequently, support vector machine (SVM), random forest (RF), and partial least squares discriminant analysis (PLS-DA) were employed to establish jujube origin recognition models, which were evaluated using confusion matrix visualization and F1 score. The results showed that the SNV-SVM combination model achieved the highest F1 score of 99.44% and a classification accuracy of 99.43%, which were 0.15% and 0.31% higher in F1 score, and 0.16% and 0.32% higher in accuracy, respectively, compared to the SD-PLS-DA and SD-RF combination models. In summary, the combination of hyperspectral technology and the SNV-SVM method enables non-destructive, rapid, and batch identification of jujubes from different origins, providing strong technical support for jujube research and production.