Identification of Greengrocery Seeds Based on NIR and Different Pretreatment Methods

An identifiable model based on near-infrared spectra (NIR) was proposed to distinguish the classification of greengrocery seeds. The performance of five pretreatment methods: Original, Smoothing, MSC (Multiplication scatter correction), SNV (Standard Normalized Variable) and FD (First Derivative) were utilized to reduce the noise in the original spectrum. The effective wavelengths were selected to remove the redundancy existing in the spectra by simulating stepwise regression. The performances of the model were optimized by the combination of pretreatments and effective wavelengths selection in this paper. Compared with the five pretreatment methods, SNV was superior to other methods with an accuracy of 100%. It is concluded that SNV coupled with simulating stepwise regression could be used to identify greengrocery seeds effectively.