Identification of Baha'sib mung beans based on Fourier transform near infrared spectroscopy and partial least squares

Abstract

In the study, Fourier transform near infrared spectroscopy technology (FT-NIR) was adopted to protect a geographically iconic product, Baha Siber mung bean. Based on partial least squares analysis method, the origin-variety dual tracing model was firstly established. Then, the data of Baha Siber mung beans from four counties (Durbert Mongolian Autonomous County (Dumeng County for short) and Baicheng, Tailai and Chifeng) were processed with different preprocessing methods to establish the origin tracing model for the analysis and comparison. Among different preprocessing methods, the vector normalization preprocessing method yielded the higher precision of corresponding model (R2 =98.02). A variety identification model was established for five varieties: mung bean, Xiaoyinggelu, Dayinggelu, Lufeng 2 and Chilu 3 from Dumeng. The preprocessing method based on multivariate scattering correction yielded the higher precision (R2 = 96.83). According to the partial least squares method-discriminant analysis results (PLS-DA), the correct recognition rate of Dumeng mung beans obtained with the origin tracing model was 92.31 %; the correct recognition rate of Xiaoming mung beans obtained with the variety identification model was 90.00 %; the correct recognition rate of Baha'sib mung obtained with the origin-variety dual tracing model was 96.67 %. Therefore, the origin-variety dual tracing model based on FT-NIR and PLS improved the correct recognition rate of Bahaxibo mung beans. This method provides a new brand protection way for geographical indication products of mung bean.It also provides a new strategy for the identification of other high value-added agricultural products.