Establishment of Detection Model of Soybean Quality Traits by Near Infrared Spectroscopy

Abstract

Background: Rapid prediction with near infrared (NIR) spectroscopy on quality traits is pretty popular recently, for the convenience and simple operation. But to make good use of this technology, precise and suitable calibration equations are very important to get dependable result. In this study, we mostly refer to the building of the equation and how the pretreatment effect them. Methods: In this paper, near infrared (NIR) spectroscopy was used to simultaneously predict the quality traits of soybean, including oil content, protein content, oleic acid content, linoleic acid content, stearic acid content. Near infrared spectral data of a total of 112 samples is collected from given materials in Chongqing. Samples were scanned from 1000 nm to 2500 nm using a monochromator instrument (SuperNIR-2700). Calibration equations were developed from NIR data using partial least squares (PLS) regression with internal cross validation. In addition, in this study, we also cover the affection of different pre-treatments to the different calibration equations predicting different quality traits. And measure the effect with three indicators including R, SECV and RPD. Result: Eventually we find the most suitable combination of pre-treatments for each calibration equation predicting a certain trait soybean. The present study would lay the foundations of rapid detection of quality traits in soybean.