Geographical Origin Identification of Glycyrrhiza Uralensis Fisch Seeds by Hyperspectral Imaging Technology

Glycyrrhiza Uralensis Fisch is an important medical herb in China. Qualities of Glycyrrhiza Uralensis Fisch vary greatly according to the geographical origin. In this paper, a non-destructive and fast geographical origin identification method for Glycyrrhiza Uralensis Fisch seeds was proposed based on the near infrared hyperspectral imaging technology. To avoid subjectivity of operators and improve identification efficiency, a hyperspectral image threshold segmentation method was used to extract regions of interest (ROIs), instead of ENVI software. In this method, reflectance calibration was applied to the seed hyperspectral image to obtain the percent reflectance value. The principal component analysis (PCA) was used to reduce dimension, and the support vector machine (SVM) was employing to build classification model. The average correct recognition rate reached 91%, which was superior to the model by the manual selection method. The experimental results demonstrated that the threshold segmentation method was more effective than the manual method for extracting ROIs. The proposed method can identify the geographical origins of Glycyrrhiza Uralensis Fisch seeds more correctly.